commit 700745cdbb34e964e1abda86183809fd8dd95796 Author: Matteo Frigo Date: Thu May 24 08:00:45 2018 -0400 Bump FFTW_MINOR_VERSION for fftw-3.3.8 commit 902d0982522cdf6f0acd60f01f59203824e8e6f3 Author: Matteo Frigo Date: Thu May 24 07:43:02 2018 -0400 update NEWS commit 41b0d9eff394891ba3327b9062811d48677bb411 Author: Matteo Frigo Date: Thu May 24 07:35:36 2018 -0400 CFLAGS: don't use -ffast-math -ffast-math is a relic from 1999 when it was kind of necessary for full use of FMA on powerpc. Nowadays it is just a liability. For example, 'gcc-8 -ffast-math' ignores the disctintion between +0 and -0, thus breaking the avx and avx2 implementations in fftw-3.7. commit 19eeeca592f63413698f23dd02b9961f22581803 Author: Matteo Frigo Date: Thu May 24 07:29:00 2018 -0400 Fixes for gcc-8 It looks like 'gcc-8 -ffast-math' does honor the distinction between +0.0 and -0.0 in floating-point constants. I suppose that technically -ffast-math has the right to do so. For good measure, this patch encodes such constants as their explicit binary representation. A separate patch will disable -ffast-math. commit bf478afbf2367df0f38c77f31d1f912aeeb82585 Author: Miklos Espak Date: Thu Apr 26 18:31:57 2018 +0100 Define include directory for installed targets (#141) commit ab888adf510338c03ea8ac49b4aab91fb57f1479 Author: Steven G. Johnson Date: Sat Apr 14 11:40:39 2018 -0400 don't need both identifier and name fields commit 2b999c600c58c78b8acb78c3352b02d9df6f6e60 Author: Steven G. Johnson Date: Fri Apr 13 08:43:35 2018 -0400 JSON doesn't like trailing commas commit 92eee8bbc4252c871aa870d2dce88eb98d0c7d18 Author: Steven G. Johnson Date: Fri Apr 13 08:38:50 2018 -0400 list both C and OCaml (as explained in codemeta/codemeta#181) commit 35e5609f17e212bf1c40da9b2ebe66784ad37052 Author: Steven G. Johnson Date: Thu Apr 12 12:01:15 2018 -0400 add codemeta file commit eba07c46b5d2f7824d293ab59aa5c29a25034963 Author: Matteo Frigo Date: Mon Feb 19 09:30:29 2018 -0500 Call _mm256_zeroupper() when leaving avx512 code Carsten Steger says: simd-avx512.h defines VLEAVE as nothing in FFTW 3.3.7. However, the current Intel® 64 and IA-32 Architectures Optimization Reference Manual, chapter 15.18, recommends the following: - When you have to mix group B instructions with Intel SSE instructions, or you suspect that such a mixture might occur, use the VZEROUPPER instruction whenever a transition is expected. - Add VZEROUPPER after group B instructions were executed and before any function call that might lead to Intel SSE instruction execution. - Add VZEROUPPER at the end of any function that uses group B instructions. - Add VZEROUPPER before thread creation if not already in a clean state so that the thread does not inherit Dirty Upper State. (Group B are instruction types that modify bits 128-511 of vector registers 0-15.) Therefore, I believe it would be prudent to define VLEAVE as _mm256_zeroupper in simd-avx512.h (see the attached patch). At https://software.intel.com/en-us/forums/intel-isa-extensions/topic/704023 Mark Charney says: To be clear, we very much still recommend using VZEROUPPER on Skylake. Even though it does not have the same penalties as earlier designs in that family for mixing AVX and SSE code, we definitely recommend using VZEROUPPER on Skylake. Yes it would obviously be better if there were one solution. For code that has to run on both families, the "common code" solution is to use the Xeon guidelines. If Mark Charney recommends VZEROUPPER, that's good enough for me. commit b267008613d082975b108252ed596ba0916ffa31 Author: Matteo Frigo Date: Wed Nov 22 12:54:18 2017 -0500 fftw3-mpi.f03 should be regenerated when Makefile changes commit 708b202fd593cf1002cf97dce0863e2a438e3720 Merge: 2e0cfdda 8ba34c40 Author: Matteo Frigo Date: Mon Nov 20 09:37:17 2017 -0500 Merge pull request #113 from xantares/mingw CMake enhancements commit 2e0cfddacacccc8a1e6e679c5e3fa81fb0219bda Author: Matteo Frigo Date: Mon Nov 20 07:07:30 2017 -0500 Attempt to strengthen language in README.md commit 8ba34c40fef38f661c9c413781990a7c021ba22b Author: Michel Zou Date: Thu Nov 9 22:33:51 2017 +0100 Preliminary Fortran support commit bd753a7679ecca2799640e7c8ced6f1f784f1b51 Author: Michel Zou Date: Mon Nov 6 23:00:29 2017 +0100 CMake MinGW fixes Mostly fixes the SSE2 macro in config.h, otherwise minor detection fixes commit da5372a175bcb09578359960869c76da74c9fda3 Author: Matteo Frigo Date: Tue Oct 31 20:21:17 2017 -0400 EXTRA_DIST += README-perfcnt.md commit 1b64d9269254e9d0a0f0b088e5eceb0db92d531f Merge: b5ccc557 2be183c3 Author: Matteo Frigo Date: Tue Oct 31 20:19:13 2017 -0400 Merge pull request #112 from alexeicolin/PR--armv7-pmccntr-counter-and-docs Pr armv7 pmccntr counter and docs commit 2be183c3a44d58aaa11909ba8882310fb44d598c Author: Alexei Colin Date: Tue Oct 31 23:34:38 2017 +0000 perf counters: name ARMv8 PMCCNTR_EL0 explicitly For consistency with the rest. commit 504ece7f8ffc60c2a03b28d977e9825230052d48 Author: Alexei Colin Date: Tue Oct 31 23:28:48 2017 +0000 perf counters: add PMCCNTR for ARMv7 and add docs The existing armv7 counter (CNTVCT) does need enabling from kernel mode (so updated the configure help), and the enable bit is different from the PMU enable bit (described in the new docs). Tested on XU4: printed the returned counter values and they look reasonable. commit b5ccc557fd2e57bfc955f0db9b5182e92f9cb55c Author: Matteo Frigo Date: Sun Oct 29 08:13:04 2017 -0400 fftw-mpi.h should include , not "fftw3.h" commit 9e3f8da20e65f1e34e677768e550086b06d77f16 Author: Matteo Frigo Date: Sun Oct 29 08:09:35 2017 -0400 NEWS: warn that cmake support is experimental and not well tested commit 9616fb9ff1c2694f5cfa2c4a59efa96094ae6812 Author: Matteo Frigo Date: Sun Oct 29 07:48:43 2017 -0400 Update NEWS for upcoming fftw-3.3.7 commit 62edb203fc09c8c8ac2c2d5ac3299ea8d4dc7838 Author: Matteo Frigo Date: Tue Oct 10 18:58:37 2017 -0400 Ditch --enable-debug-malloc and --enable-debug-alignment We wrote DEBUG_MALLOC in 1997 to debug memory leaks. Nowadays DEBUG_MALLOC is just confusing. Better tools are available, and DEBUG_MALLOC is not thread-safe and it does not respect SIMD alignment. It confused at least one user. In the gcc-2.SOMETHING days, gcc would allocate doubles on the stack at 4-byte boundary (vs. 8) reducing performance by a factor of 3. That's when we introduced --enable-debug-alignment, which is totally obsolete by now. commit 6ed4297e85e5ef24a18ce428b18e020d8e48413a Author: Matteo Frigo Date: Fri Sep 29 19:27:43 2017 -0400 Use armv7a cycle counter unconditionally if HAVE_ARMV7A_CNTVCT It looks like __ARM_ARCH_7A__ is not always defined. If the user says HAVE_ARMV7A_CNTVCT, trust the user. commit 2dd77382319ceb99c32b38418716783eec8adad4 Merge: 04590cb1 e09ab8ca Author: Matteo Frigo Date: Thu Sep 21 22:42:38 2017 -0400 Merge pull request #110 from junghans/cmake Minor cmake fixes commit e09ab8cac98c0f206968bbd962a6f76cf26e7437 Merge: 890dac59 76427f30 Author: Christoph Junghans Date: Thu Sep 21 16:13:43 2017 -0600 Merge commit 'refs/pull/109/head' of github.com:FFTW/fftw3 into cmake commit 04590cb11baa11bbfdebe101fa90186bbf48423c Author: Matteo Frigo Date: Thu Sep 21 18:00:58 2017 -0400 simd-vsx.h: don't use vpermxor It seems like gcc-6 generates incorrect code when using vpermxor (tested with qemu emulator, so there is a chance that gcc is right and qemu is wrong). Disable the use of vpermxor and do the simple thing (one multiplication + one permutation). commit 76427f30080e2cab3ca5047193ce8ffe6110f047 Author: Michel Zou Date: Thu Sep 21 23:44:15 2017 +0200 No need to list includes commit e47e9a81c41454e5e128cd68505b38152ad60500 Author: Matteo Frigo Date: Thu Sep 21 17:13:14 2017 -0400 Remove AC_FUNC_{MALLOC,REALLOC,MMAP} They don't do what I thought. E.g., AC_FUNC_MALLOC checks that malloc(0) returns NULL, and defines malloc to be rpl_malloc otherwise. We don't support rpl_malloc() and we don't care about malloc(0). commit 5aebc02ff30af12d2dc3be6c762e821a38f56595 Author: Matteo Frigo Date: Thu Sep 21 10:09:02 2017 -0400 Dead-Code Police commit d97394a17250d71d6a722ae64dcc3123130cf08f Author: Matteo Frigo Date: Thu Sep 21 09:54:36 2017 -0400 Fixup fftw3-mpi.h fftw3-mpi.h must include "fftw3.h", not "api/fftw3.h", because both fftw3-mpi.h and fftw3.h will ultimately be installed in /usr/include. Thus, as a special exception, mpi/Makefile.am must specify the include path -I $(top_srcdir)/api. commit 890dac59aca4c153e7e22add0a8de00766227670 Merge: 4ebda892 106582aa Author: Christoph Junghans Date: Wed Sep 20 14:44:04 2017 -0600 Merge commit 'refs/pull/109/head' of github.com:FFTW/fftw3 into cmake commit 4ebda89297b6b38632c3d91bd5a673a1bee4ffff Author: Christoph Junghans Date: Wed Sep 20 14:05:13 2017 -0600 autotools: fix install of FFTW3ConfigVersion.cmake commit e9a66d5f748037f9cb9c0f5b8d824d73c0425042 Author: Christoph Junghans Date: Wed Sep 20 13:29:29 2017 -0600 cmake: use GNUInstallDirs commit 4fbb72ad294e2070d64a83b24f89a601d4f624c6 Author: Matteo Frigo Date: Wed Sep 20 13:11:55 2017 -0400 Generate codlist.c only when MAINTAINER_MODE The user is not supposed to regenerate .c files. In addition, the generation rule is subtly nonportable (it depends on whether or not '#' can be escaped in Makefiles, an issue that does not appear settled.) commit f243f8ce48be61952527d43da222096296fdd2f9 Author: Matteo Frigo Date: Wed Sep 20 11:54:13 2017 -0400 Generate {dft,rdft}/simd/{sse,sse2,avx,...}/*.c only when MAINTAINER_MODE Users are not supposed to generate them. Apart from that, the generation rule uses '$*' in an explicit make rule, which is technically a GNU extension. (Works with {open,free}bsd, but breaks Solaris.) commit 106582aa8f97257f53730cbac81f98e8659b084c Author: Michel Zou Date: Wed Sep 20 15:46:51 2017 +0200 Fix includes, export target commit 1a24e67165ba56447f814bcdc12b9d6e083f1670 Author: Matteo Frigo Date: Wed Sep 20 07:24:58 2017 -0400 Restore the ability to build out of tree. Before 1f3704b9, we had "-I $(top_srcdir)/foo -I $(top_srcdir)/bar". After 1f3704b9, we had no -I specification at all, but automake wants an explicit -I $(top_srcdir) in order to build out of tree. commit 919b795940d1e86a948a4430193dbd0853f47272 Merge: 6076339a f7a64365 Author: Matteo Frigo Date: Wed Sep 20 06:41:50 2017 -0400 Merge pull request #107 from xantares/config-mode Config mode commit f7a6436509d324297783eb77df54010320b062f8 Author: Michel Zou Date: Wed Sep 20 11:46:05 2017 +0200 Build bench according to BUILD_TESTS commit 82cec28b7e14280ad11878978e23a3680bb0e983 Author: Michel Zou Date: Wed Sep 20 11:41:20 2017 +0200 Use cmake config mode Installs FFTW3Config.cmake instead of a FindFFTW3.cmake Also configures the pkgconfig file from cmake commit 6076339a342b12b0d0cfd9f6d967bfa9fbf6b1b2 Author: Matteo Frigo Date: Tue Sep 19 23:38:27 2017 -0400 Fix performance regression with gcc-3.3 commit f4c37657cb32b2552c5e86f0540c0308d4f451ef Author: Matteo Frigo Date: Tue Sep 19 23:24:08 2017 -0400 get rid of the sse2-nonportable.c hack It was necessary to support some broken compiler 15 years ago. Remove it and see if anybody complains. commit 362ae5c7b8a9df76b5ec0de4433131db33bae0ae Author: Matteo Frigo Date: Tue Sep 19 21:44:13 2017 -0400 configure.ac Police Remove some obsolete AC_CHECK_HEADERS, add new checks suggested by autoscan. commit a56b5b4b149e56fce43778172a56f77d30352833 Author: Matteo Frigo Date: Tue Sep 19 21:43:45 2017 -0400 Include Police fftw-wisdom.c was including instead of "api/fftw3.h" commit 1f3704b9eff4b7e80ef7d775fb13f5bb8de0a5f1 Author: Matteo Frigo Date: Tue Sep 19 21:12:22 2017 -0400 Do not set include path ("-I") in Makefile.am .[ch] files should specify their own paths explicitly. Setting paths in the Makefile was always a bad idea, but it is totally untenable if we are supporting cmake. commit 6e0ae04bad14a7dd9b4928f22d7a01e887dfdc03 Author: Matteo Frigo Date: Tue Sep 19 19:31:55 2017 -0400 Fix OpenBSD build Using $< in a non-suffix rule context is a GNUmake idiom and OpenBSD doesn't like it. commit 31a53789197f90d6bf349dd230ab86023e5fb83c Author: Matteo Frigo Date: Tue Sep 19 19:24:34 2017 -0400 EXTRA_DIST += FindFFTW3.cmake.in commit ae1a764ce88166e8e1f05a25888f105ec8f1939d Merge: 5fdca1d9 97b273d8 Author: Matteo Frigo Date: Tue Sep 19 17:13:58 2017 -0400 Merge pull request #69 from junghans/cmake Build und install cmake module commit 5fdca1d9b0a0b2e6491c98f63873dcf600355e09 Merge: b521e530 66506470 Author: Matteo Frigo Date: Tue Sep 19 15:57:59 2017 -0400 Merge pull request #92 from tklauser/armv7a-cycle-counter Fix ARMV7-A cycle counter detection commit b521e5305a7317c1c0f1d454beb6580eaf4de1db Author: Matteo Frigo Date: Tue Sep 19 15:51:03 2017 -0400 cmake: don't check for dlfcn.h We don't use it commit fc852fcdfa80fab30eac2284249686853efa2e4b Author: Matteo Frigo Date: Tue Sep 19 15:43:02 2017 -0400 Remove ancient paranoia In the '90s we used to run autoconf three times, just in case (because it really didn't work the first time). "Three" was modeled after the "sync; sync; sync; reboot" incantation of the '80s. Hopefully we are past this by now. commit 34738e7f669882c6abc12c2744c8acc347c91719 Author: Matteo Frigo Date: Tue Sep 19 15:32:39 2017 -0400 Flip boolean in a way that makes more sense to me commit a2bfd859d9ad08490d02252d8a80c5994dd82747 Author: Matteo Frigo Date: Tue Sep 19 15:28:56 2017 -0400 Various CMakeLists.txt fixes * AVX2 codelets require -mfma * --enable-avx2 automatically enables the 128-bit avx2 codelets in *dft/simd/avx2-128 * bump FFTW_VERSION to 3.3.7, SOVERSION to 3.5.7 * build bench always, irrespective of Threads_FOUND commit 93ac6e1075e73c0275a9e0006fe9161c3b6fae38 Merge: a71f3dd3 d3a8d13f Author: Matteo Frigo Date: Tue Sep 19 14:31:03 2017 -0400 Merge pull request #103 from xantares/cmake Add user cmake support Still needs work, but let's move forward and move this contribution into the official repository commit d3a8d13f74361a7ffc4c48c229181a86b35e9a7d Author: Michel Zou Date: Tue Jul 18 12:16:43 2017 +0200 Add user cmake infrastructure commit a71f3dd355f802dc362a52674a977ff81daadf9d Author: Matteo Frigo Date: Wed Jul 5 06:33:40 2017 -0400 Disable ISA_EXTENSION_PREFERS_FMA for now I still don't understand whether or not avx2 should use FMA codelets. Ryzen is faster with the non-FMA version. Haswell prefers the FMA version. However, I suspect that Haswell prefers FMA because of a quirk of the micro-architecture. Haswell has two floating-point "ports". You can issue an addition only through one "port", but you can issue two FMA in parallel on both ports, so FMA appears to be faster. Skylake apparently restores balance (but I haven't tried yet). Suspend judgment for now until I gather more data. commit f82b8c94596868897987b71a648eaa664590602a Author: Matteo Frigo Date: Tue Jul 4 20:06:57 2017 -0400 Rationalize HAVE_FMA Distinguish ARCH_PREFERS_FMA, for architectures that "naturally" prefer FMA (e.g., powerpc), from ISA_EXTENSION_PREFERS_FMA, for instruction-set extensions that favor FMA where the base architecture does not (e.g., avx2 on x86). Previously, --enable-avx2 would use FMA code for scalar and avx codelets, which is wrong. This change improves performance by a few percent on Ryzen (where FMA doesn't really do anything), and is a wash on Haswell. commit 0869f4e51b8b0aeb7da1b21b2683c30cd4e10a5e Author: Steven G. Johnson Date: Tue May 9 09:14:37 2017 -0400 document that howmany ≥ 0 (closes #95) commit 665064700b26c01c0836e4c12a5ee0eab3923858 Author: Tobias Klauser Date: Wed Mar 29 16:15:45 2017 +0200 Fix ARMV7-A cycle counter detection Check for the correct pre-processor define HAVE_ARMV7A_CNTVCT from config.h (instead of ARMV7A_HAS_CNTVCT) to fix the detection of the cycle counter for ARMv7-A in the configure script (and actually use it in the built library). Without this fix, even the following ./configure call: ./configure --enable-neon --enable-single --enable-armv7a-cntvct \ --host=arm-linux-gnueabihf --disable-fortran \ CC="arm-linux-gnueabihf-gcc -march=armv7-a" will emit the warning: checking whether a cycle counter is available... no *************************************************************** WARNING: No cycle counter found. FFTW will use ESTIMATE mode for all plans. See the manual for more information. *************************************************************** With this fix applied, ./configure will correctly detect the cycle counter register: ... checking whether a cycle counter is available... yes ... commit cc5fc8ce7ffd77f467740554f649aab4d3f71344 Merge: 102f2fd0 950b1539 Author: Matteo Frigo Date: Tue Mar 14 07:21:45 2017 -0400 Merge pull request #91 from fornwall/android-clock-gettime Avoid trying to use CLOCK_SGI_CYCLE on Android commit 950b153910f7f0dde9cc20cddeee5dc9048d25b7 Author: Fredrik Fornwall Date: Mon Mar 13 23:41:35 2017 +0100 Avoid trying to use CLOCK_SGI_CYCLE on Android The Android headers defines CLOCK_SGI_CYCLE but the call fails at runtime as it's not implemented. Combined with getticks() not checking the return value of clock_gettime() this causes bogus values to be returned from getticks(). commit 102f2fd0249dca301d195b4df1b94e7b339b8c60 Author: Matteo Frigo Date: Wed Feb 22 14:59:30 2017 -0500 Compute mflops() in 64 bit precision Old code was overflowing for N>2^32 commit 2b63fc2eaae645a5c2ef4a97c384beb2adefd58d Author: Matteo Frigo Date: Fri Jan 27 16:06:27 2017 -0500 Update NEWS for 3.3.6-pl2 commit d2ca54234956ad8be82ba050305ccf979fd631a7 Author: Matteo Frigo Date: Fri Jan 27 16:01:42 2017 -0500 Get ready for fftw-3.3.6-pl2 commit 83092f8efbf872aefe7cfc6ee8fa43412f8e167a Author: Matteo Frigo Date: Fri Jan 27 15:52:18 2017 -0500 Fix scrips that generate the MPI F03 interface It turns out that the scripts were using fftw3.h from /usr/include, not ../api, and were failing silently if fftw3.h was not installed. This bug led to a fftw-3.3.6pl1 release with incomplete mpi/f03 header files. commit ab402b00f9a003daa10863b9bcdbe0810b26f541 Author: Steven G. Johnson Date: Wed Jan 25 13:03:15 2017 -0500 mention mkdist.sh and summarize the build process in README.md (closes #85) commit fa9f00b3831177f0a9582092f21efb14e3d4601f Author: Matteo Frigo Date: Sun Jan 22 14:51:44 2017 -0500 add __cdecl decorators to fftw3.h functions on Windows This patch re-does 1f19d597 in a more disciplined way. Also, Whitespace Police. commit 42c0036e839b78a7af651d5504add62ed57f9961 Author: Matteo Frigo Date: Sun Jan 22 14:32:32 2017 -0500 Revert "add __cdecl decorators to fftw3.h functions on Windows, in case someone compiles with a non-default calling convention, as discussed in #80" This reverts commit 1f19d59793eb629dd8228e8a41f4f8618c20a246. The chosen syntax FFTW_EXTRN(T) X(name) is improper because __cdecl appertains to the declarator and not to the return type. (As is clear, e.g., in void (__cdecl *foo)(void)). This forces monstrosities such as FFTW_EXTRN(R *) X(name) that contradict the C declaration syntax. I'll redo the patch in a way that looks like C: FFTW_EXTERN R *FFTW_CDECL X(name) commit 1f19d59793eb629dd8228e8a41f4f8618c20a246 Author: Steven G. Johnson Date: Thu Jan 19 23:09:23 2017 -0500 add __cdecl decorators to fftw3.h functions on Windows, in case someone compiles with a non-default calling convention, as discussed in #80 commit 596b924b86340456771fb75559016ec2cc1b44c4 Author: Matteo Frigo Date: Mon Jan 16 10:25:37 2017 -0500 Assert that CURRENT-AGE=3 This is an attempt to prevent the 3.3.6 version screwup from occurring again. In any reasonable universe, libraries would have a version H and they would specify a L such that the library is compatible with all versions in [L..H]. Any sensible programmer would never change L, as this breaks backward compatibility and screws users. A new version would increase H and be done. Instead, libtool wants CURRENT=H and AGE=H-L (a new version change two variables). Furthermore, the name of the library in the file system is a combination of L and H-L. The two changes of basis arent't even orthogonal. Pure madness. This change attempts to impose sanity by asserting that that the implied L is 3, since we never intend to break backward compatibility with fftw-3.3, which was version L=3. commit 6fb9cd7b6359f29ce488a5802793139971d59c6c Author: Matteo Frigo Date: Mon Jan 16 09:06:06 2017 -0500 Release 3.3.6-pl1 commit 18b7e53c54727303703db29373e61a35fb8d5db8 Author: Matteo Frigo Date: Mon Jan 16 08:56:53 2017 -0500 Fix #82: FFTW3 3.3.6 shared version rollback commit 64a5a288e56c6ff4462b69531f4f34d740fdc12c Author: Matteo Frigo Date: Mon Jan 16 08:42:01 2017 -0500 Improve documentation of fftw_make_planner_thread_safe Specifically, tell people not to use it unless they must. commit 811a672bdaedec4363272d9f7ed5fae56086aeb1 Author: Matteo Frigo Date: Sun Jan 15 17:40:37 2017 -0500 rm obsolete simd/ directory We switched to simd-support/ many years ago, not sure why it is still in git. This was not a problem when the repository was private, but the directory probably confuses people on github. commit 5c9bead1ea35b3a21fb33f17011d6802722ba44b Author: Matteo Frigo Date: Sun Jan 15 07:25:40 2017 -0500 Warnings Police * suppress dead code in genfft/simd.ml * fix on size_t/int confusion * fix one float*/double* confusion (should have been void* because we only check the alignment of the pointer, not its type). commit 41b191ee128fefe28a228ab706dfdfb65d32c2e1 Author: Matteo Frigo Date: Sun Jan 15 07:02:40 2017 -0500 Update configure.ac, NEWS for 3.3.6 commit fc3ada6e6bd790341fb5d91c6775b8afd686bad7 Author: Matteo Frigo Date: Sun Jan 15 06:40:23 2017 -0500 Ansi C Police fftw is supposed to compile with c89/c90. Restore this property so that I can test with gcc -ansi. This change may seem needlessly reactionary, but in the last release I accidentally inserted an assertion before a declaration and I broke the Visual Studio build, so we must be careful not to use C99 constructs. There are a few non-ANSI function calls in tests, e.g. isnan(), drand48(), snprintf(). Since nobody has complained about those in years, I am leaving them alone. commit 50dacdaba79694c873965ab23d11c8ca3b94d436 Author: Matteo Frigo Date: Sat Jan 7 09:01:47 2017 -0500 Revert simd-avx.h changes from b606e3191 They didn't improve performance at all as far as I can tell, and they ended up breaking the PGI compiler. It is always tempting to use the fancy addsub instructions in FFTW to do complex multiplications, but the reality is that FFTW is designed to avoid complex multiplications in most cases (we started in the SSE days), and thus they don't make any difference. We are better off using the minimal possible set of AVX instructions to minimize the chance of triggering compiler bugs. The same statement holds for _mm256_shuffle_pd() versus _mm256_permute_pd(): in theory the latter is better, in practice either one is rarely used. However, SHUFFLE is older (since the SSE days) and has a higher chance of working. commit 5fa55dc130e18cc4b3f4d88b8a159307eecf51d0 Merge: 1637e8aa aa00ba84 Author: Matteo Frigo Date: Sun Nov 13 05:49:09 2016 -0500 Merge pull request #77 from rolandschulz/master Fix AVX512 load+store commit aa00ba84079a272637666c9ae941821087f712b8 Author: Roland Schulz Date: Sat Nov 12 20:52:49 2016 -0800 Fix AVX512 load+store FFTW alignment is only 16 bytes. AVX512 requires 64 bytes. Thus unaligned load/store is required. AVX256 does the same. commit 1637e8aace6e91d67837901b5a4cbbc87c42aca9 Merge: 3e7ee221 a538bf2c Author: Matteo Frigo Date: Thu Nov 3 11:24:44 2016 -0400 Merge pull request #76 from forandom/patch-2 Update simd-vsx.h to support building with IBM XLC commit a538bf2c4a17ec509f2cec37bffe48874702c671 Author: forandom Date: Thu Nov 3 23:06:17 2016 +0800 Update simd-vsx.h to support building with IBM XLC defined(__POWER8_VECTOR__) && defined(__GNUC__) && defined(__LITTLE_ENDIAN__) is true for IBM XLC compiler for which we should use the intrinsic __vpermxor instead of __builtin_crypto_vpermxor. commit 3e7ee2211ae1bd5e76901bbe1bcca67b31f84ccb Author: Matteo Frigo Date: Sat Sep 24 06:39:01 2016 -0400 Do not run programs at configure time, ever. configure was running a program to detect the ARM cycle counter, thus preventing cross-compiling. Sorry about that. commit fee0f966b2d3fae18019dd03a9bae338b4108d42 Merge: 3a3173b0 cca0c6e5 Author: Matteo Frigo Date: Fri Sep 9 06:49:23 2016 -0400 Merge pull request #72 from tkelman/patch-1 #include in threads.c for windows build commit cca0c6e5a8c717df10f380411709f3360ceea6e9 Author: Tony Kelman Date: Fri Sep 9 03:24:30 2016 -0700 #include in threads.c for windows build otherwise an i686-w64-mingw32 cross compile is giving ``` libtool: link: i686-w64-mingw32-gcc -march=pentium4 -m32 -std=gnu99 -shared -Wl,--whole-archive kernel/.libs/libkernel.a dft/.libs/libdft.a dft/scalar/.libs/libdft_scalar.a dft/scalar/codelets/.libs/libdft_scalar_codelets.a rdft/.libs/librdft.a rdft/scalar/.libs/librdft_scalar.a rdft/scalar/r2cf/.libs/librdft_scalar_r2cf.a rdft/scalar/r2cb/.libs/librdft_scalar_r2cb.a rdft/scalar/r2r/.libs/librdft_scalar_r2r.a reodft/.libs/libreodft.a api/.libs/libapi.a simd-support/.libs/libsimd_support.a simd-support/.libs/libsimd_sse2_nonportable.a dft/simd/avx/.libs/libdft_avx_codelets.a rdft/simd/avx/.libs/librdft_avx_codelets.a threads/.libs/libfftw3f_threads.a -Wl,--no-whole-archive -march=pentium4 -m32 -O3 -mtune=native -malign-double -Wl,--stack -Wl,8388608 -o .libs/libfftw3f-3.dll -Wl,--enable-auto-image-base -Xlinker --out-implib -Xlinker .libs/libfftw3f.dll.a libtool: link: i686-w64-mingw32-gcc -march=pentium4 -m32 -std=gnu99 -shared -Wl,--whole-archive kernel/.libs/libkernel.a dft/.libs/libdft.a dft/scalar/.libs/libdft_scalar.a dft/scalar/codelets/.libs/libdft_scalar_codelets.a rdft/.libs/librdft.a rdft/scalar/.libs/librdft_scalar.a rdft/scalar/r2cf/.libs/librdft_scalar_r2cf.a rdft/scalar/r2cb/.libs/librdft_scalar_r2cb.a rdft/scalar/r2r/.libs/librdft_scalar_r2r.a reodft/.libs/libreodft.a api/.libs/libapi.a simd-support/.libs/libsimd_support.a simd-support/.libs/libsimd_sse2_nonportable.a dft/simd/avx/.libs/libdft_avx_codelets.a rdft/simd/avx/.libs/librdft_avx_codelets.a threads/.libs/libfftw3_threads.a -Wl,--no-whole-archive -march=pentium4 -m32 -O3 -mtune=native -malign-double -Wl,--stack -Wl,8388608 -o .libs/libfftw3-3.dll -Wl,--enable-auto-image-base -Xlinker --out-implib -Xlinker .libs/libfftw3.dll.a threads/.libs/libfftw3_threads.a(libfftw3_threads_la-threads.o):threads.c:(.text+0x121): undefined reference to `_mm_pause' threads/.libs/libfftw3_threads.a(libfftw3_threads_la-threads.o):threads.c:(.text+0x581): undefined reference to `_mm_pause' collect2: error: ld returned 1 exit status threads/.libs/libfftw3f_threads.a(libfftw3f_threads_la-threads.o):threads.c:(.text+0x121): undefined reference to `_mm_pause' threads/.libs/libfftw3f_threads.a(libfftw3f_threads_la-threads.o):threads.c:(.text+0x581): undefined reference to `_mm_pause' collect2: error: ld returned 1 exit status make[4]: *** [Makefile:627: libfftw3f.la] Error 1 make[4]: *** [Makefile:627: libfftw3.la] Error 1 make[3]: *** [Makefile:672: all-recursive] Error 1 make[2]: *** [Makefile:536: all] Error 2 make[3]: *** [Makefile:672: all-recursive] Error 1 make[1]: *** [/home/Tony/julia32/deps/fftw.mk:46: scratch/fftw-3.3.5-single/build-compiled] Error 2 make[1]: *** Waiting for unfinished jobs.... make[2]: *** [Makefile:536: all] Error 2 make[1]: *** [/home/Tony/julia32/deps/fftw.mk:46: scratch/fftw-3.3.5-double/build-compiled] Error 2 make: *** [Makefile:81: julia-deps] Error 2 ``` commit 97b273d87dcc797e688709e207f119dd4dfca015 Author: Christoph Junghans Date: Wed Aug 31 14:24:05 2016 -0600 Build und install cmake module commit 3a3173b018f30d03df5f3166d459888f2669fe25 Author: Matteo Frigo Date: Wed Aug 31 06:14:51 2016 -0400 C++ compatibility Although FFTW is a C program, we try to make it compilable by a C++ compiler as well. Implicit cast void * ==> double * is not allowed in C++. commit 5fd9609eaed60360ce84d98add5d9548093e0bdc Author: Matteo Frigo Date: Fri Aug 12 04:24:52 2016 -0400 Updated NEWS commit 402d2508fe970770d9316d9c83f21d6fc268ba12 Author: Matteo Frigo Date: Fri Aug 12 04:21:33 2016 -0400 Fix race condition when destroying a plan. More generally, this patch calls the planner hooks when destroying a plan. The intended usage is that the hooks do in fact acquire a lock. commit 432835f2cd37d2cb8b9528ac8ef983b3b38738f2 Author: Matteo Frigo Date: Tue Aug 9 05:29:39 2016 -0400 MSVC fixes by Carsten Steger * don't mix declarations and statements, stick to ANSI C * suppress some warnings with Intel cc * undefined variable in x86-cpuid.h when (_MSC_VER > 1500) || (_MSC_VER == 1500 & _MSC_FULL_VER >= 150030729) commit c018cbe430fd6b2af31d594c27a0aaf711292567 Author: Matteo Frigo Date: Thu Aug 4 06:36:29 2016 -0400 Fix SIMD autodetection on amd64 when (_MSC_VER > 1500) commit d5055c9ae2e60f191f6cc2e8b5200fd06dbdb6be Author: Matteo Frigo Date: Sun Jul 31 13:42:00 2016 -0400 revise README.md language commit 0af8d8b9eea0750add8be0e6dec18841ee61424e Author: Matteo Frigo Date: Sun Jul 31 13:39:49 2016 -0400 revise README.md language commit 0d026e09f9b514cb86bbc7977ad0a03b664b95de Author: Matteo Frigo Date: Sun Jul 31 13:37:09 2016 -0400 Attempt to tell users to download official tarballs from fftw.org instead of github commit b405994456f9a87f2170ba19536d4c4d8278682f Author: Matteo Frigo Date: Sat Jul 30 16:33:22 2016 -0400 update AUTHORS commit 4d0c1894fb37c61b0f0a42b50afd435d226f6b9e Author: Matteo Frigo Date: Sat Jul 30 15:18:06 2016 -0400 Fixes for Windows cross-compilation These days mingw by default produces binaries that depend on libgcc-sjlj-1.dll, which defeats the whole historical point of mingw (produce vanilla win32 binaries with no GNU stuff). Add a hack to link with -static-libgcc, which avoids the problem. commit a17d44eeb3100780ba106a22f497d47a43be7642 Author: Matteo Frigo Date: Sat Jul 30 11:39:09 2016 -0400 Misc fixes. * sed s/avx[_- ]128[-_ ]fma/avx-128-fma * avoid some signed/unsigned casts commit f3688be112ed0099b4c57970db74c08373f3604d Author: Matteo Frigo Date: Sat Jul 30 10:52:53 2016 -0400 Fix SIMD autodetection * AVX was not testing for OSXSAVE support * AVX2 was broken (issuing XGETBV without checking for its presence---failing on atom) * AVX512 was broken in the same way as AVX2, I have guessed a fix but I have no way to test it. commit 7fce2ae37f8338bd7e021b1a406c75b213c31c77 Author: Matteo Frigo Date: Fri Jul 29 07:48:10 2016 -0400 document fftw_make_planner_thread_safe() commit 6167b92e3362f2d116274daa561c0d788fb670d4 Author: Matteo Frigo Date: Fri Jul 29 07:28:03 2016 -0400 rm README-bench It appears in tests/README commit cc9640cbbaa70e6645a0ea46be0508268905c2ba Author: Matteo Frigo Date: Fri Jul 29 07:27:25 2016 -0400 Add README-bench commit d82fe4f3e06bdbf92b09324e36f4d477bc5fe376 Author: Matteo Frigo Date: Fri Jul 29 07:25:00 2016 -0400 Do not enable avx128-fma unless the user asks for it. Adding SIMD instruction sets automatically is user-hostile behavior. Also, update the manual to reflect the new SIMD support commit dc32329871d304de8d95ad290973844dfbc6101f Author: Matteo Frigo Date: Fri Jul 29 07:00:55 2016 -0400 Update NEWS for 3.3.5 commit 2ed010c62b1bc8ca6b23bfda2e09b8c28e1e8bcc Author: Matteo Frigo Date: Sun Jun 5 07:07:15 2016 -0400 Clean up some int<->size_t confusion commit ea86c49ac7470a646d1e6a4fa007ecbda6ab56c4 Author: Matteo Frigo Date: Sat Jun 4 20:33:15 2016 -0400 Unused Variable Police commit d9a3f48343bda0a88c8a87cab329d95426ddfcb9 Author: Matteo Frigo Date: Sat Jun 4 20:30:12 2016 -0400 Integral Type Police clear some int/unsigned/size_t confusions commit 29cee6cc95d434321292d013d6a7be4c55379a49 Author: Matteo Frigo Date: Sat Jun 4 19:50:10 2016 -0400 Cast Police Eliminate some useless (but harmless) int<->size_t conversions. commit d7c566eb98523c7c0bafae734c7894a5a3595771 Author: Matteo Frigo Date: Sun Mar 13 17:50:45 2016 -0400 Clarify ambiguous/wrong documentation of halfcomplex output format. commit 6543818e3091ea788a1aac41d06ca343e672f103 Author: Matteo Frigo Date: Wed Jan 20 18:18:14 2016 -0500 Cleanup Rewrite Unique.make in more idiomatic caml style, strongly typed. commit 4965e33c6c98484b66787f1891cfe4f689becee5 Merge: 119aa4c4 f8a73593 Author: Matteo Frigo Date: Wed Jan 20 15:14:32 2016 -0800 Merge pull request #53 from artemkin/master Fixed unique token generation in genfft commit f8a73593a499efc751103460ff2f07d8b1e2ff0c Author: Stanislav Artemkin Date: Thu Jan 21 01:17:30 2016 +0400 Fixed unique token generation in genfft Unique token generation was based on the assumption that OCaml compiler won't inline a given piece of code. Starting from 4.02.0 it does more aggressive inlining and breaks this functionality. commit 119aa4c4a893f32dfd837a84fac9453b6dae6680 Merge: 8c7a7af1 e41df2c3 Author: Matteo Frigo Date: Wed Sep 30 15:38:00 2015 -0400 Merge pull request #48 from rleonid/master Replace depracted usage of Pervasives or. commit e41df2c3cac7c3e69586c07f80f1bb0a24dccd5a Author: Leonid Rozenberg Date: Wed Sep 30 15:22:16 2015 -0400 Replace depracted usage of Pervasives or. commit 8c7a7af184a63064325fa542a8d1d7f4e3b4b8aa Author: Matteo Frigo Date: Tue Sep 8 10:35:18 2015 -0400 Clarify how to bootstrap fftw from the git repository commit a0cbff67eae9ab66f6f2b4cf2ea79de6c95d7d61 Author: Matteo Frigo Date: Tue Sep 8 10:28:49 2015 -0400 Clarify that ocamlbuild is necessary for --enable-maintainer-mode commit f6339eadef8a62432ea2f2017ce0b4a1954ea738 Author: Steven G. Johnson Date: Wed Aug 5 11:11:26 2015 -0400 fix LaTeX typo, thanks to Gael Lorieul commit 8cd9bfa347289143a00fa0d5eea30f4766192d46 Author: Erik Lindahl Date: Wed May 27 00:15:57 2015 +0200 Update VSX SIMD to avoid inline assembly Thanks to some help from Michael Gschwind of IBM, this removes the remaining inline assembly calls and replace the with vector functions. This avoid interfering with the optimizer both on GCC and XLC, and gets us another 3-10% of performance when using VSX SIMD. Tested with GCC-4.9, XLC-13.1 in single and double on little-endian power 8. commit 579cec9a64cc177e673f006eb112d488be21b230 Author: Erik Lindahl Date: Tue May 26 19:27:58 2015 +0200 Enable SSE2 automatically with AVX,AVX2, or AVX512. 256-bit AVX can be significantly slower than 128-bit SIMD. Despite recommendations many distributions appear to only enable AVX, but not SSE. This fixes the problem by also enabling SSE when we use the wider SIMD instructions. commit dd80210ec433938876575e2435e12d7e630872e7 Author: Erik Lindahl Date: Tue May 26 19:09:40 2015 +0200 Turn AVX-128 into AMD-specific AVX-128-FMA The only platform where AVX-128 really matters is AMD (since the compute units can execute a single 256-bit or two 128-bit SIMD instructions), so now we only use it there which means we can also enable FMA instructions. commit b3105ed9529846ca8dd9267e46d7bcd2ebb12ff6 Author: Matteo Frigo Date: Mon May 25 17:33:15 2015 -0400 Fix broken avx/32-bit compilation commit d3442a8395e1fc6e77490c3f34c868b1998e4e96 Author: Matteo Frigo Date: Mon May 25 17:27:31 2015 -0400 rm hooks api's, add fftw_make_planner_thread_safe() api fftw_make_planner_thread_safe() installs a lock around the planner. It is guaranteed to be atomic and idempotent. I wrote an emulation of pthread mutex initializers on Windows, but I haven't even compiled the Windows code yet. commit 842596fe6bfb277effc8f8f8db7e1c4008bb59e1 Author: Matteo Frigo Date: Mon May 25 11:11:58 2015 -0400 Add TODO's commit eff7dfcd526e90539f169cfff2374ceb2fd2dd0e Author: Matteo Frigo Date: Mon May 25 10:50:21 2015 -0400 add TODOs commit 0e53e3e9d2640a895a69c5aff4a676c156271141 Author: Matteo Frigo Date: Mon May 25 10:48:03 2015 -0400 Add argument to planner hooks commit 94ef591d61c8e78e87c65f9779eb003fc7f1ba6b Author: Matteo Frigo Date: Mon May 25 10:37:24 2015 -0400 Update shared-version-info We added an API (planner hooks), so the shared version info needs to be bumped. commit 9ef9ec85588c026e6a80b1475df24ba20d098e8e Author: Matteo Frigo Date: Mon May 25 10:27:25 2015 -0400 Bump version to 3.3.5 commit 593d55932959366918e209fbbd2f4719d39d448c Author: Romain Dolbeau Date: Wed May 13 16:42:23 2015 +0200 Typo ; Fixes #41 commit cd2b27d1600d80ba719f1b70094886e39cf145a3 Author: Erik Lindahl Date: Thu May 7 17:45:43 2015 +0200 Separate routines to query 128-bit AVX support This also disables 256-bit AVX for current AMD processors that work better with 128-bit AVX. Note that this is not detected by the timing routines since the effect is only apparent when using multiple cores. commit a1cf4158dd829853bd9f6b8c4c4951d7495c9e64 Merge: b6135085 0331b39c Author: Romain Dolbeau Date: Mon Apr 20 22:01:15 2015 +0200 Merge branch 'master' of github.com:FFTW/fftw3 commit 0331b39cd3641a8ac89be27dbde3e41204fd1888 Merge: 38b93ccf d2ea399c Author: Erik Lindahl Date: Mon Apr 20 21:09:44 2015 +0200 Merge branch 'experimental-simd' Merged in new SIMD architectures from separate branch. commit d2ea399c46174db45838ca6a3b917cf880970921 Author: Romain Dolbeau Date: Thu Apr 16 08:54:58 2015 +0200 Fix stack alignment (alloca) for generic256. commit b6135085bd1ee2e2c6c82b06e78d492e4f242cca Author: Romain Dolbeau Date: Tue Apr 14 10:17:38 2015 +0200 Add a configure option to disable building the documentation in doc/. This is useful if some documentation tools are missing on the host (i.e. fig2dev in maintainer mode). commit 38b93ccfc3786d1c23726dc939de558f4dd2a2d3 Author: Romain Dolbeau Date: Mon Apr 13 14:16:28 2015 +0200 In maintainer mode, detect whether 'indent' is available and is GNU indent. BSD 'indent' in e.g. MacOSX doesn't support -kr, which is the default style in FFTW3. Fallback to 'indent' with no option for non-GNU 'indent', or 'cat' if 'indent' is not available. This should fix GitHub issue #13. commit 96eb0ad31c8d0c226a6aeb95e68bda90dd7e6f6a Author: Romain Dolbeau Date: Sun Apr 12 13:49:39 2015 +0200 missing AC_ARG_ENABLE for --enable-fma commit 24ff943f4e99458d41db543305a2c945ba2ba429 Author: Romain Dolbeau Date: Sun Apr 12 13:47:24 2015 +0200 Revert "reinstate --enable-fma ; not enabled by default for AVX2 & AVX-512 (will ad a warning later)" This reverts commit 40691a49eea40d305405fe527e174e8067606dae. A different fix is needed. commit 37a0dbc10f0199fd431f0b5d8b42143cc2a2a88d Author: Romain Dolbeau Date: Sun Apr 12 13:39:28 2015 +0200 Add sanity check & warning in AVX2 & AVX-512 commit 40691a49eea40d305405fe527e174e8067606dae Author: Romain Dolbeau Date: Sun Apr 12 13:38:29 2015 +0200 reinstate --enable-fma ; not enabled by default for AVX2 & AVX-512 (will ad a warning later) commit 7960d08a3fe74b38d8dfdd20917efb52d141d53a Author: Erik Lindahl Date: Wed Apr 8 22:55:28 2015 +0200 Improved compiler flags for OS X Separate detection for AVX/AVX2 on gcc and clang. Clang works for AVX, but AVX2 leads to a compiler crash. Issue 20471870 has been filed with Apple. When using gcc, we now request to use the external system assembler, or the AVX/AVX2 instructions will cause errors. commit 91928338b767b84742e8ec86da6b4864381ed889 Author: Erik Lindahl Date: Wed Apr 8 22:54:41 2015 +0200 Fix alignments for generic simd. commit eaaec9b6ea9dc0f0656d953639c325855cb3bbee Author: Erik Lindahl Date: Wed Apr 8 21:16:13 2015 +0200 Made api versions more verbose for 128-bit AVX. commit 4b3dbf7009b020bffe7c9c96a5b24c87496fd058 Author: Erik Lindahl Date: Wed Apr 8 21:09:50 2015 +0200 Make 128/256 bit generic simd separate options These will only be used on esoteric and/or new architectures, which likely also miss cycle counters. In this case the widest simd would be picked automatically based on flops estimates, so to give the user more control it is better to provide separate options to enable/disable these two choices. commit cbe2a4a64064d12b9b817235906d61a996c00be1 Author: Romain Dolbeau Date: Mon Mar 30 13:35:47 2015 +0200 AVX-512: minor fix(sp)&improvement(dp) to VDUPL/VDUPH commit 5379243044ea4113b9cbde25fd097195817b3653 Author: Romain Dolbeau Date: Sun Mar 29 14:17:13 2015 +0200 AVX-512: fix typo; fix shuffle parameter in SP; _mm512_set1 exists now. commit 90f9610ee6708efc11c848b5e078dd92997ffa25 Author: Romain Dolbeau Date: Sun Mar 29 14:14:02 2015 +0200 Fix typo in KCvi commit 2b44c9213a11816506b1bd3d6b7316ed1ed65a15 Merge: 5c5bed23 1f28d2d3 Author: Romain Dolbeau Date: Sun Mar 29 10:49:42 2015 +0200 Merge branch 'experimental-simd' of github.com:FFTW/fftw3 into experimental-simd commit 5c5bed2365693b5f57503f6aad35264a3b3d86b9 Author: Romain Dolbeau Date: Sun Mar 29 10:48:13 2015 +0200 typo for AVX-512 commit 1f28d2d3693bd919c674e08dc76726cf56d3648f Author: Romain Dolbeau Date: Sun Mar 29 10:48:13 2015 +0200 typo commit a8845007ecb07fd0cc91994c574b1008c64708e0 Author: Romain Dolbeau Date: Thu Mar 19 19:22:44 2015 +0100 Generic SIMD support for gcc While not as optimized as the specific SIMD ports, this enables the usage of gcc's generic vector representation, which usually gets implemented with SIMD instructions on most hardware. Double precision implementations for 256 and 128 bits by Romain Dolbeau, merged into a single generic SIMD implementation and single precision added by Erik Lindahl. The option --enable-generic-simd will turn on both 128 and 256 bit versions, and the timers will choose the fastest codelets. commit 56bbdbab2f03ebae92fd2c52b9509fb3b8ffe226 Author: Erik Lindahl Date: Tue Mar 24 19:35:31 2015 +0100 Added Power8 VSX SIMD support Power8 is a descendent of Power7, but the switch to little endian means the old altivec SIMD will not work due to shifts on load and store, and the new VSX instructions are much improved. This adds support for both single and double precision VSX SIMD, using either gcc (tested with version 4.9) or IBM xlC (tested with version 13.1.2, slower than gcc). Clang from llvm-3.7 is still too buggy to compile VSX code correctly, but flags and detection has been added so it will work with a correct clang. commit 8aa91763af07767f3ebb71a9836a69e3b3385cab Author: Romain Dolbeau Date: Tue Feb 24 09:27:07 2015 +0100 Double precision Neon SIMD for aarch64 --enable-neon now works in double precision for 64-bit Arm. Support added for the generic timer virtual counter in armv7a (optional, available in A15 & A7) & armv8. They are privileged, but should be made user-readable in recent linux (> 3.19.1 for v7a, most for v8). Architecture Reference Manual ARMv7-A and ARMv7-R edition: E.7.16 CNTVCT, Virtual Count register, system level Architecture Reference Manual ARMv8, for ARMv8-A architecture profile Beta: D7.5.17 CNTVCT_EL0, Counter-timer Virtual Count register commit aa26395250c9c4d6831e8e5017650ea70af56a28 Author: Romain Dolbeau Date: Thu Sep 5 10:53:42 2013 +0200 AVX-512 SIMD support New configure flag --enable-avx512 commit de81bfdb66b9bc867e389bbaf67b56490ca2e2cd Author: Erik Lindahl Date: Wed Mar 25 15:49:33 2015 +0100 128-bit AVX2 SIMD support Add 128 bit support for AVX2. Similar to AVX-128, this improves slightly on SSE2 due to more efficient instructions, and the shorter SIMD width is beneficial in some cases. Both 128- and 256-bit flavors will be built automatically with --enable-avx2, and the timing routines will chose the best one automatically. commit da988fa4c53fb63fafe2eeff3da4abad93e7d014 Author: Erik Lindahl Date: Sat Mar 28 12:52:52 2015 +0100 AVX2 kernels and CPUID support Initial AVX2 code from Romain Dolbeau. Modifications, cpuid and more compiler flags from Erik Lindahl. New --enable-avx2 configure flag supported. commit d7d9b9d2b71bd93bc4d4fa82d46a9c013291b7fe Author: Romain Dolbeau Date: Tue Sep 3 10:02:53 2013 +0200 KCvi [Knight Corner Vector Instructions] SIMD support This adds SIMD support for the first generation of Xeon Phi. commit b606e3191e5b65e2e13f67ef7dad5b1e7c40206c Author: Erik Lindahl Date: Wed Mar 25 01:44:17 2015 +0100 Improved AVX SIMD Previously, some kernels were actually faster with the old SSE2 SIMD, which made it necessary to compile with both sse2 and avx for good performance. This adds 128-bit AVX kernels which are enabled together with the standard AVX kernels. Apart from being encoded with AVX rather than SSE instructions (depending on compiler flags), it also uses a couple of new instructions only available with AVX that use fewer micro-ops. These instructions have also been added to the 256-bit AVX SIMD implementation. No new configure flags needed, it is just faster. commit 131027afcd3ed5d7c0185611036431c1035a734a Merge: 0ea3051f 56af330f Author: Matteo Frigo Date: Sun Mar 22 16:24:29 2015 -0400 Merge pull request #37 from maxlevesque/patch-1 add indent to requirements commit 56af330fc1600a856241968482ecd443bc2c26aa Author: Maximilien Levesque Date: Sat Mar 21 22:41:26 2015 +0100 add indent to requirements Without indent, `make` reports errors in somewhere in a directory called codelets. It also reports "/bin/bash: indent: command not found" not far away. For my Ubuntu 14.04 flavor, a simple `sudo apt-get install indent` made `make` work as expected. commit 0ea3051f99f2931a46a66aef8862517cd9c5f3c7 Merge: 506c1634 69a82a6c Author: Matteo Frigo Date: Mon Feb 16 06:36:46 2015 -0500 Merge pull request #32 from psteinb/2d_mem_layout_complying_to_text changed ny to n1 and nx to n0 so that the labels match the text commit 69a82a6c5ff4d2169a7f8a0afda9f3c68f24eb2a Author: Peter Steinbach Date: Mon Feb 16 09:10:23 2015 +0100 changed ny to n1 and nx to n0 so that the labels match the text, it's quite confusing otherwise commit 506c16346f9fc57444b179e542e88225e3c3e923 Author: Steven G. Johnson Date: Fri Jan 30 15:02:17 2015 -0500 fix #29 commit d94666815b9a0073e4ac8cc48f88e18ac931bd45 Merge: ded00512 9831bbd1 Author: Steven G. Johnson Date: Sun Dec 7 16:39:15 2014 -0500 Merge pull request #27 from mpip/master avoid segfaults due to double free commit 9831bbd14ca5b963ad1dba260c86151c94e000ee Author: Michael Pippig Date: Sat Dec 6 15:03:33 2014 +0100 avoid segfaults due to double free If fftw_mpi_mkplans_posttranspose() fails, the plans cld3, cld2rest, and cld2 are destroyed at nada and must be set to NULL. Otherwise, a second destroy at nada in mkplan() will cause a segfault. commit ded0051238f129fb65846e822191706c9b1f5221 Author: Matteo Frigo Date: Sun Nov 16 09:45:16 2014 -0500 Conciseness Police commit 113e1086966fdff4c172672753cc880e6bc74d3d Author: Matteo Frigo Date: Sun Nov 16 09:41:05 2014 -0500 add {before,after}_planner_hooks FFTW now calls fftw_before_planner_hook() before creating a plan, and fftw_after_planner_hook() afterwards. This allows users, e.g., to grab a lock. TBD: add arguments. commit 28635e1d5f0a0079af3e7d00cd0678c4745e2c2b Author: Matteo Frigo Date: Sun Nov 16 09:22:47 2014 -0500 avoid multiple declarations of fftw_alignment_of() fftw_alignment_of() was declared both in the API header file fftw3.h and in the internal header file ifftw.h. While there is nothing wrong with this, it breaks the property that all exported symbols are defined in the API directory. E.g., I am not sure what happens on windows without the proper DLLEXPORT nonsense. To avoid any issues, rename the internal routine to fftw_ialignment_of(), and define an API wrapper. commit 36597576e4c3c5dc3efd7d8b57a1bbad505715aa Author: Steven G. Johnson Date: Tue Aug 12 12:29:50 2014 -0400 fix #21 (don't use float128 on Portland compilers, which pretend to be gcc) commit cde4559ba9b822166cb88a84a0994fdb83a2061c Author: Matteo Frigo Date: Sat Jul 26 20:09:38 2014 -0400 Avoid transforming uninitalized data. In r2c/c2r transforms when using 4-way SIMD, sometimes FFTW uses the following hack: to transform an odd number of inputs, it copies the input into a buffer that holds space for one extra input; it transforms the buffer (now comprising an even number of inputs, as required by SIMD); it copies back the odd number of transformed inputs, ignoring the padding element. The extra input was uninitialized until now. This is ok because we ignore the transform of the uninitialized input. Transforming uninitialized data may cause floating-point exceptions, an effect that is observable. This patch initializes the additional elements to zero, thus avoiding the problem. This patch also includes a test, but the test is disabled by default because it is nonportable. To observe the FP exception, one must use feenableexcept(), which appears to be a GNU-ism. commit 2493129c332197c5195ecb6796cfeb5e8d92e09a Author: Steven G. Johnson Date: Thu Jul 10 10:41:38 2014 -0400 fix #19: missing Fortran interface for fftwq_alloc_real commit 07ef78dc1b273a40fb4f7db1797d12d3423b1f40 Author: Steven G. Johnson Date: Tue Jul 8 11:14:15 2014 -0400 fix #18 (disable float128 for CUDACC) commit 2fd372f31ab7c6417de0634199bcd5b7765df926 Author: Matteo Frigo Date: Sat Jun 28 17:17:19 2014 -0400 git rm *~ commit 203e0d610ec1e413bb426a7d60fd5e2a206a2830 Author: Matteo Frigo Date: Thu Apr 3 15:46:19 2014 -0400 Fix wrong boolean precedence in hppa cycle counter. Thanks Jens Keiner for the bug report. For some reason nobody noticed this in years. commit f8048af3e30cb3f65befd0aa2f3d16de3eeb5583 Author: Steven G. Johnson Date: Tue Mar 4 15:23:11 2014 -0500 more .gitignore additions commit 5a51b3fe98509cc7e7ba5d3e17a3381777ad4731 Author: Steven G. Johnson Date: Tue Mar 4 15:21:31 2014 -0500 some fixes for make distcheck commit 836af27f5d780970c87e436da882c9928e09c0f0 Author: Steven G. Johnson Date: Tue Mar 4 13:42:02 2014 -0500 copyright year update commit 853f9f7cad1a8d3e92e6767562e4cd2d336164d2 Author: Steven G. Johnson Date: Tue Mar 4 13:39:18 2014 -0500 updates for 3.3.4 commit bf30f5aeea1c2927b302b13dc7579acf9f2adc12 Author: Steven G. Johnson Date: Tue Mar 4 13:30:00 2014 -0500 another file in .gitignore commit c82b4fd61796715b1043982b1d4af49047f90238 Author: Steven G. Johnson Date: Tue Mar 4 10:39:09 2014 -0500 add .gitignore commit 146fa8d61fca4a06a85c70d7167ac925575df02b Author: Steven G. Johnson Date: Thu Jan 16 10:47:40 2014 -0500 added fftw_sprint_plan to output plan info to a string (so that the caller can be more flexible about how it is displayed) commit 53e1fdbc07133b53ffbbd51c56b57a89880c0b21 Author: Steven G. Johnson Date: Thu Jan 16 10:26:48 2014 -0500 document fftw_alignment_of (since I found it useful in Julia, other people may too) commit 7dbc7067e99477312acae30a9001c0dffa9bb428 Author: Matteo Frigo Date: Tue Nov 19 19:08:44 2013 -0500 group together AC_CHECK_DECLS for functions in stdlib.h commit fb70e413bddca578b2b72e0cac281d5c9a3c4101 Author: Matteo Frigo Date: Tue Nov 19 07:33:59 2013 -0500 Be more careful in detecting sinl(), cos(), memalign(), posix_memalign() commit 40f59a1d0f9bf384826595c499b0e7fe99aa1df5 Author: Matteo Frigo Date: Tue Nov 19 06:20:43 2013 -0500 Fix autodetection of cosl(), sinl() Autoconf must have changed since we last looked. Thanks Åke Sandgren for the fix. commit 7e66dc5a495edc855dc9e156767172eaeabee335 Author: Matteo Frigo Date: Sat Oct 26 17:31:00 2013 -0400 Fix wrong example in fftw-wisdom-to-conf.1 Thanks Julian Taylor for the bug report. commit dd3283ac7d2e916bdeccb7229a669fcc2ef7ff83 Author: Matteo Frigo Date: Tue Oct 1 07:03:51 2013 -0400 Use "bench$(EXEEXT)" instead of "bench" Seems to be necessary on Windows. commit c6acf03f53d217c9041eae3d381ad41b942dc9d9 Author: Steven G. Johnson Date: Fri Sep 20 09:26:08 2013 -0400 fix typo commit f230f8cf903f1e2bb1261ed2f8657a99ac12a9ca Author: Matteo Frigo Date: Thu Jul 11 19:28:50 2013 -0400 Fix fftw-wisdom-to-conf Apparently we broke fftw-wisdom-to-conf many years ago and nobody noticed. Thanks Florian Oppermann for the bug report. commit c74775bff0c164611377b29d95b3f6a6e8192005 Author: Matteo Frigo Date: Tue Jun 4 10:31:33 2013 -0400 Somehow the NEWS entries for fftw-3.1.[23] were missing from the trunk. commit c87bdc8bfd7cbf753bbe0635ea82613bc271220d Author: Matteo Frigo Date: Mon Jun 3 06:59:35 2013 -0400 note fixes for texinfo-5 commit ed390e3a385832e1faa452032f170510be6ed280 Author: Matteo Frigo Date: Sun Jun 2 09:10:01 2013 -0400 version.texi should not be in git commit aff23d05642705f738f788648c060085bdc476d6 Author: Matteo Frigo Date: Sun Jun 2 09:05:50 2013 -0400 Fix the manual to work with both texinfo-4 and texinfo-5. Texinfo has been stable for the first 15 years of FFTW's history. Then some genius, with too much time in his hands and on a mission to deliver the world from the evil of the C language, decided to rewrite makeinfo in Perl, the old C version of makeinfo being, as I said, evil. The official excuse for the rewrite was that now I can have my manual in XML format, as if XML were a feature. The result of this stroke of genius is that texinfo-5 has different rules for macro expansion than texinfo-4 does, specifically regarding whether or not spaces after a macro are ignored. Texinfo-4 had weird rules, but at least they were constant and internally more or less consistent. Texinfo-5 has different rules, and even worse the rules in texinfo-5 are inconsistent between the TeX and HTML output processors. This situation makes it almost impossible for us to produce a manual that works with both texinfo 4 and 5 in all modes (TeX, info, and html). The @noindent/@refill hack is my best shot at patching this situation. commit b0308275bb63a9cb3edb5847fa130f901deaf47e Author: Matteo Frigo Date: Fri May 17 11:39:05 2013 -0400 fftw_wisdom.1.in: document the --threads option. commit 7eb9af0354f7663fa89daa56163d5bc5865bcec1 Author: Matteo Frigo Date: Wed Mar 27 13:12:15 2013 -0400 Define the version number as M4 macros, so that it is defined only once. We used to have two version strings: the package number FFTW-X.Y.Z and the libtool number CURRENT:REVISION:AGE with the invariant that REVISION==Z. Unfortunately in the fftw-3.3.3 release we forgot to change REVISION, with the result that fftw-3.3.3 and fftw-3.3.2 have the same REVISION number. Hopefully this patch will prevent similar screwups in the future. commit e67d16ed0b81b0e01124c7cdee2320bfea2fd6e4 Author: Matteo Frigo Date: Sun Mar 17 19:47:57 2013 -0400 note that these scripts are not meant for normal users commit b892c705c7051bcd27c0939070e515fe85f68bb2 Author: Matteo Frigo Date: Sun Mar 17 19:45:52 2013 -0400 add README.md for github's convenience commit 029db460f692ea9fefc7d0efdf8e11b5d51215b0 Author: Matteo Frigo Date: Sun Mar 17 19:44:00 2013 -0400 Update README for people who download the git repository. commit 128e9ddbde76a63c7f255b258ff02bd6766cbaa4 Author: Matteo Frigo Date: Sun Mar 17 19:37:40 2013 -0400 Remove reference to obsolete mailing list. commit 4b1d3b63d64a01216c023389565e40ca03d6be2f Author: Matteo Frigo Date: Sat Mar 16 09:47:59 2013 -0400 mkdist.sh: use git instead of darcs Use git instead of darcs to generate the ChangeLog. Also, refuse to create a distribution if git HEAD does not have a tag, to prevent a common mistake. commit fe84f5d9eeef9f773b0785ba91032ab5b7677461 Author: Matteo Frigo Date: Sat Mar 16 09:26:03 2013 -0400 remove some junk '*~' files created by the darcs->git conversion commit 2f9d0a41eec86def6fe9160dac1be172bb241eba Author: stevenj Date: Thu Feb 7 14:22:12 2013 -0500 use Win32 threads, not pthreads, if both are present (it's not clear why Windows users would ever want the latter); see also https://github.com/JuliaLang/julia/issues/2015 commit b8623189967ed7eb1ca50e80f2b5ee2d6f3ca0ad Author: stevenj Date: Thu Feb 7 14:13:40 2013 -0500 I found it useful in the Julia interface to call fftw_alignment_of (in order to check plan applicability), in which case we need to IFFTW_EXTERN it for Windows; might be worth considering documenting this function commit 23b1bf3e19198d2c4575bf6f305cd41ebc17124c Author: athena Date: Sat Jan 12 15:35:46 2013 -0500 Increase timing interval to 5000 cycles on x86_64 commit 6e615417da7e1c7cba49b1a73c9edcc15d938cf9 Author: athena Date: Tue Dec 4 15:58:22 2012 -0500 make -lm a private library in fftw.pc.in Julian Taylor says: Make -lm a private library, libfftw is linked against it so clients don't need it. You can use pkg-configs --static flag for static linking. this works on all systems where indirect linking works, probably on all others pkg-config is not supported anyway (wild guess). commit 553849f32bcc82a17de11c76cfa9b2f672bfb89e Author: athena Date: Tue Dec 4 15:54:58 2012 -0500 Escape minus signs in man pages http://lintian.debian.org/tags/hyphen-used-as-minus-sign.html says: By default, "-" chars are interpreted as hyphens (U+2010) by groff, not as minus signs (U+002D). Since options to programs use minus signs (U+002D), this means for example in UTF-8 locales that you cannot cut and paste options, nor search for them easily. Thanks Julian Taylor for the patch. commit 715c7ea8347a5dad7d97a0c5d81a87801e826ada Author: athena Date: Tue Dec 4 15:53:28 2012 -0500 Change texinfo category to Development commit 01810ba2a427ee086a4a5323e991dd19e2d715be Author: stevenj Date: Sat Nov 24 22:37:54 2012 -0500 fixed deadlock bug caused by bogosity flag getting out of synch between processes; thanks to Michael Pippig for the bug report commit 69aa82642e26a8eb5292a8a7b83250e8df619065 Author: athena Date: Wed Nov 21 18:34:29 2012 -0500 Updated NEWS commit e98f888b9457ce5855491279c6c0ef72e23a374c Author: athena Date: Wed Nov 21 18:33:15 2012 -0500 use 2x2 AVX transposition instead of individual stores. This seems to improve single-precision AVX on Sandy Bridge machines. commit 466f579cb8856a0709da1e6c6b5ca03360bc61a0 Author: stevenj Date: Tue Nov 20 12:18:00 2012 -0500 revert part of Taylor patch to acx_mpi.m4: do not link -lmpi if mpicc works without libraries, as -lmpi may be some completely different MPI implementation commit 610460226f6d5d2d7c4c53896b5aff9b1f108e4b Author: stevenj Date: Tue Nov 20 11:44:57 2012 -0500 fix deadlock bug (thanks to Michael Pippig for the bug report and patch, and to Graham Dennis for the bug report) in which some processes called MPI_Alltoall and some called MPI_Alltoallv commit 512d8d783d6af373fca8376f79255b794df5bd31 Author: athena Date: Mon Oct 29 15:20:01 2012 -0400 fix texinfo quirk commit ff329890540002506c47717ebbc3959de30e5066 Author: athena Date: Mon Oct 29 09:16:43 2012 -0400 clarify that padding only applies to in-place transforms commit 905ded711f93fa3c94faa7623a5093525338fdeb Author: athena Date: Sun Oct 28 18:42:48 2012 -0400 make the index-computation logic less paranoid The problem is that for each K and for each expression of the form P[I + STRIDE * K] in a loop, most compilers will try to lift an induction variable PK := &P[I + STRIDE * K]. In large codelets we have many such values of K. For example, a codelet of size 32 with 4 input pointers will generate O(128) induction variables, which will likely overflow the register set, which is likely worse than doing the index computation in the first place. In the past we (wisely and correctly) assumed that compilers will do the wrong thing, and consequently we disabled the induction-variable "optimization" altogether by setting STRIDE ^= ZERO, where ZERO is a value guaranteed to be 0. Since the compiler does not know that ZERO=0, it cannot perform its "optimization" and it is forced to behave sensibly. With this patch, FFTW is a little bit less paranoid. FFTW now disables the induction-variable optimization" only when we estimate that the codelet uses more than ESTIMATED_AVAILABLE_INDEX_REGISTERS induction variables. Currently we set ESTIMATED_AVAILABLE_INDEX_REGISTERS=16. 16 registers ought to be enough for anybody (or so the amd64 and ARM ISA's seem to imply). commit 1dacef5bde5cb6599f9d98e42495f7897f109787 Author: athena Date: Sun Oct 28 18:33:24 2012 -0400 silence warnings commit fb08724b27a05ca890c1da062c8d0385c22c02eb Author: athena Date: Sat Oct 27 09:58:49 2012 -0400 bump version to 3.3.3 commit c4d6abbc1c80eb612b2abccce728a06189780a69 Author: athena Date: Sat Oct 27 09:55:15 2012 -0400 evaluate plans for >1ms when using gettimeofday() The previous limit 10ms was too paranoid, and it made life difficult on machines without an "official" cycle counter, such as ARM. commit 172dd3def821c0898822a5ca72c3f5391553536c Author: athena Date: Sat Oct 27 09:46:04 2012 -0400 use 4-way NEON SIMD instead of 2-way Kai-Uwe Bloem tried to warn me a year ago that 128-bit NEON was better than 64-bit NEON even on machines with a 64-bit pipe, but I foolishly did not listen. Now that 128-bit NEON pipes are starting to appear on the market it is definitely time to switch. commit 1c9c469f8727ab1780533226283746e7e9098694 Author: athena Date: Wed Sep 26 14:21:12 2012 -0400 Note that fftw-3.3 includes MPI support commit 2de12d67e5f4e34d39119f2e730e9d70e4df0c4e Author: athena Date: Wed Jul 18 11:25:40 2012 -0400 remove obsolete unused function commit 6bc94ae7db56490e35c256bd4840608eea6ec150 Author: stevenj Date: Fri Jun 29 15:57:14 2012 -0400 whoops, call omp_get_max_threads; thanks to Hanno Rein for the bug report commit 747ece1503281aad7beb32448af9a7cad05eba52 Author: athena Date: Sat Apr 28 10:55:09 2012 -0400 Fix libfftw3/libfftw3_threads chicken-egg problem On most systems we want to build libfftw3 first, so that libfftw3_threads can depend upon libfftw3. When producing a single combined-thread library (e.g. on Windows) we want the opposite, so that libfftw3 can include libfftw3_threads. commit 4bcfb67da7211171d5f4b80a90845770bbd2e147 Author: athena Date: Sat Apr 28 10:11:28 2012 -0400 updated NEWS for 3.3.2 commit cb553a8315ae9a700558956a190aac4658064b83 Author: athena Date: Thu Apr 26 19:36:11 2012 -0400 change revision to 3.3.2 commit 98229b0d7673cfa15a8c339d305b09367b97d670 Author: athena Date: Thu Apr 26 19:31:02 2012 -0400 Remove old aligned_main() hack. On i386, in our benchmark program we used to manually aligned the stack to 16-byte boundary via asm trickery. This was a good idea in 1999 (and it was actually necessary to make things work) but the hack is now obsolete and it seems to break gcc-4.7. So the hack is now gone. commit 4e4c680e7497ee8bb87bb31451d10d71b8c205e2 Author: athena Date: Thu Mar 29 16:26:16 2012 -0400 Bugfix: a couple of uninitialized values in the benchmark program commit 229d864bde7f95f45ee23608b756926a3a20ddd0 Author: athena Date: Tue Mar 20 19:03:47 2012 -0400 make libfftw{threads,mpi} depend upon libfftw for libtool purposes Thanks Julian Taylor for the patch commit f9a05701f2027906b68ff913713166310e18c8cc Author: stevenj Date: Tue Mar 6 04:44:00 2012 -0500 formatting tweak commit cec6c01b2647796f5909cfe2d90ce040380da5c5 Author: athena Date: Mon Mar 5 21:05:27 2012 -0500 destroying => overwriting commit ad79a0ae3baf83548bc5c9597b0aae94e5226073 Author: stevenj Date: Fri Mar 2 10:31:20 2012 -0500 note that WISDOM_ONLY is a documented flag commit adf4cdab68e50778689170eb91270bb189aea1ac Author: stevenj Date: Fri Mar 2 10:27:08 2012 -0500 check for icc pretending to be gcc before including quad-precision decls; thanks to Michael Anselmi for the bug report commit 99aeb386978a56c1a33f6fe7c9b7942421cf550e Author: stevenj Date: Fri Mar 2 10:23:19 2012 -0500 foo_CFLAGS needs to manually include AM_CFLAGS; thanks to Henry Gomersall for the Windows bug report commit 6060dbccef23e01e603f44d42ab602ab89a9442e Author: athena Date: Sat Feb 25 15:21:39 2012 -0500 update for latest mingw commit 06fff6523418bd7ee2478c081cec418eb73adee5 Author: stevenj Date: Mon Feb 20 23:06:13 2012 -0500 added Fortran NEWS commit 51e33866a2c0cc189954b8197e5053b3275fc564 Author: stevenj Date: Mon Feb 20 23:00:13 2012 -0500 move non-portable extended/quad precision F03 interfaces into separate .f03 files (while keeping double/single in fftw3.f03 for minimal ABI breakage) commit 545c90a1db20c5cd50f30d0f31d2334ec10cf8ac Author: athena Date: Mon Feb 20 11:21:57 2012 -0500 rm mpi/fftw3-mpi.f03 at make clean time; thanks Tyler Luchko for the bug report. commit b7c0fcdee76b9f2efe194ec13d3b1ca97fadf376 Author: athena Date: Mon Feb 20 11:18:24 2012 -0500 Disable a Visual Studion warning that was obnoxious enough for Sebastian Schuberth to send us a patch. commit 4a0de08f4d1f026454fba2e053bd0573afb984d2 Author: athena Date: Mon Feb 20 11:18:06 2012 -0500 Change version to 3.3.1 commit 49783f83e58d9bc7dc3ea7f1822fefe961361935 Author: athena Date: Mon Feb 20 11:03:15 2012 -0500 Integrated Visual Studio AVX patches by Carsten Steger commit bc9dc18402445ce9b357dea5ff2b20720773ddc5 Author: stevenj Date: Wed Nov 9 10:13:32 2011 -0500 typo commit 659e85e6065429fc6a6e4a005bc5f5068f707fb3 Author: stevenj Date: Tue Nov 8 22:45:09 2011 -0500 add missing F77 set_timelimit function; thanks to Martin Diehl for the bug repory commit cf1a5563a691fd0c25c67910926adb2ef936cbb7 Author: athena Date: Sun Sep 25 10:54:56 2011 -0400 note requirement of /machine:x64 in windows x64 README commit cb216e1fd71ab751f2de5e083bc0237cc98535c9 Author: athena Date: Sun Sep 18 09:28:20 2011 -0400 AVX detection for MSVC commit 8acf60e9e0c6417a91d9a21dddb1467feb4fd23e Author: athena Date: Tue Sep 13 14:58:29 2011 -0400 compile with C89 commit f3edf46ebdda4a632ccb0ece6c5fa0559014d24e Author: athena Date: Sat Sep 3 16:25:50 2011 -0400 use the same search pruning heuristics for threaded plans as for nonthreaded plans commit f004d764307d0e1815c2f936a9a398825d367f2c Author: athena Date: Sat Sep 3 16:12:11 2011 -0400 shorten ESTIMATE planning time for certain weird sizes FFTW includes a collection of "solvers" that apply to a subset of "problems". Assume for simplicity that a "problem" is a single 1D complex transform of size N, even though real "problems" are much more general than that. FFTW includes three "prime" solvers called "generic", "bluestein", and "rader", which implement different algorithms for prime sizes. Now, for a "problem" of size 13 (say) FFTW also includes special code that handles that size at high speed. It would be a waste of time to measure the execution time of the prime solvers, since we know that the special code is way faster. However, FFTW is modular and one may or may not include the special code for size 13, in which case we must resort to one of the "prime" solvers. To address this issue, the "prime" solvers (and others) are proclaimed to be SLOW". When planning, FFTW first tries to produce a plan ignoring all the SLOW solvers, and if this fails FFTW tries again allowing SLOW solvers. This heuristic works ok unless the sizes are too large. For example for 1044000=2*2*2*2*2*3*3*5*5*5*29 FFTW explores a huge search tree of all zillion factorizations of 1044000/29, failing every time because 29 is SLOW; then it finally allows SLOW solvers and finds a solution immediately. This patch proclaims solvers to be SLOW only for small values of N. For example, the "generic" solver implements an O(n^2) DFT algorithm; we say that it is SLOW only for N<=16. The side effects of this choice are as follows. If one modifies FFTW to include a fast solver of size 17, then planning for N=17*K will be slower than today, because FFTW till try both the fast solver and the generic solver (which is SLOW today and therefore not tried, but is no longer SLOW after the patch). If one removes a fast solver, of size say 13, then he may still fall into the current exponential-search behavior for "problems" of size 13*HIGHLY_FACTORIZABLE_N. If somebody had compleined about transforms of size 1044000 ten years ago, "don't do that" would have been an acceptable answer. I guess the bar is higher today, so I am going to include this patch in our 3.3.1 release despite their side-effects for people who want to modify FFTW. commit 610f7976d8d31f385e2bce4fd8da0a5c770fb877 Author: athena Date: Sat Aug 27 13:55:24 2011 -0400 Fix typo fftw_execute_dft_r2r => fftw_execute_r2r Thanks KIU Shueng Chuan for the bug report. commit 76f7c5558bb5eedb80ff3e67db798b4e493a5872 Author: athena Date: Fri Aug 26 06:13:55 2011 -0400 In Rader's algorithm, compute the generator lazily. The planner was spending a lot of time computing generators for plans that were immediately discarded. Now we compute generators only when absolutely needed. commit e1b527d72aad02ddea04f266f6831fb13768fbc3 Author: athena Date: Sun Aug 21 16:16:38 2011 -0400 Release notes for 3.3.1-beta1 commit 7079b5216c27e2320215f1eb10f6c6554a6c1ac1 Author: stevenj Date: Fri Aug 19 19:59:17 2011 -0400 make fftw_mpi_block routine 10x faster, since it is being called zillions of times (thanks to Tom Vacek for the profiling) commit e5c7931a01f350aa1f756bfa76307b317e2208e1 Author: athena Date: Thu Aug 18 14:19:36 2011 -0400 Implement autodetection of NEON extensions commit 16600d97d52a81152e4ef9ac140c336a5a1ca126 Author: athena Date: Sun Aug 14 14:12:29 2011 -0400 Update the FSF address. The FSF moved downtown. commit 8609b388c7872e1b39baa6d72349dbbb476ade97 Author: stevenj Date: Thu Aug 11 14:54:38 2011 -0400 allow specifying TRANSPOSE_{IN/OUT} transpose plans, since libbench does not canonicalize rnk=1 n=1 plans as rnk=0 commit 341d3e142ea250096b5c76778f77fba4bf6de622 Author: stevenj Date: Thu Aug 11 14:17:24 2011 -0400 check.pl should occasionally check DESTROY_INPUT problems too (especially since those enable slightly different algorithms in MPI) commit 2845a3c0be7059c93b704b624cc482f9e4778188 Author: stevenj Date: Thu Aug 11 12:37:51 2011 -0400 unify post-MPI transpose handling in pairwise and alltoall solvers; should make the former faster in the destroy-input out-of-place case, and the latter more widely applicable commit f02c57b267809e16e0e29ae8f7c1301323eda1cb Author: athena Date: Mon Aug 8 10:06:14 2011 -0400 Add support for ARM NEON commit 8ec2b52144bcf72a028cb84c273719b0ecc7730e Author: stevenj Date: Fri Aug 5 17:25:32 2011 -0400 more C++ paranoia commit 6047a7079ad3a55d982e68e03cbe158c5e476ee0 Author: stevenj Date: Fri Aug 5 17:02:00 2011 -0400 tentative version bump for 3.3.1 commit 6fbb0639cf17563d4603c6d545e9335d19e3ac43 Author: stevenj Date: Fri Aug 5 16:52:28 2011 -0400 fixes so that MPI code compiles when MPICC is a C++ compiler, even if the serial code is compiled with a C compiler; thanks to Kyle Spyksma for the bug report commit d7feb2daba8f23e5bca6c4e4e3f4177ad9a49e87 Author: stevenj Date: Fri Aug 5 16:04:06 2011 -0400 use correct precision in f03-wrap.c, avoiding a (harmless) implicit pointer cast that prevented compilation under C++; thanks to Kyle Spyksma for the bug report commit 50d12441bd0b3410c799d11784717e76147b5474 Author: stevenj Date: Fri Aug 5 14:04:32 2011 -0400 manual typo commit 9e45ff08aca4e28ec61c947284188a01aed45fe6 Author: athena Date: Tue Jul 26 20:55:45 2011 -0400 Honor WITH_OUR_MALLOC in libbench2 commit 2cfcd40d46731a41a400ed4a4d2eeeb954422568 Author: athena Date: Tue Jul 26 20:27:28 2011 -0400 fixed typo: incorrect name of combined threads library on Windows commit a81ea0083c5cc32c9acecfd5f57f38d0e4f07bb5 Author: stevenj Date: Mon Jul 25 14:38:20 2011 -0400 3.3 version bump & NEWS commit df2116046636b6ed6b989fa8b706f3e9db53e17e Author: stevenj Date: Mon Jul 25 14:37:48 2011 -0400 use int(..., C_SIZE_T) rather than declaring another variable in the Fortran examples commit 19ebb38d4180b0a70f4579d37dedfd6e678edecd Author: stevenj Date: Wed Jul 13 05:02:32 2011 -0400 typo, thanks to Rhys Ulerich for the comment commit a221f0e99a9206e1edaa8016e04f891261cc5196 Author: athena Date: Mon Jul 11 14:39:52 2011 -0400 Fix bug in bubblesort Bubblesort was not sorting. This was a bug in the benchmark library (not in FFTW per se), and it impacted the benchmark program with --report-time and --report-mflops causing it to output an incorrect value for the median. (The minimum, maximum, and average value were correct.) Thanks Dima Baksheev of Intel for reporting this bug. commit d33f4f7b648b658d0d232f8561c85fd9b007c105 Author: stevenj Date: Fri Jul 8 13:35:59 2011 -0400 small manual typos commit 047c6636b6cbbdde2d8f4e5a62e26013336fd3e7 Author: athena Date: Wed Jul 6 10:49:40 2011 -0400 Detection of altivec.h requires $ALTIVEC_CFLAGS commit 8cb56c732d86e26edca3ea53e63440756a434031 Author: athena Date: Tue Jul 5 19:58:47 2011 -0400 Introduce fake dependency so that my-getopt.c is recompiled my-getopt.c does not depend on anything, and so it is not rebuilt when reconfiguring for a different ISA (e.g., CC="gcc -m32" vs CC="gcc -m64"). Add a fake dependency on so that the file is recompiled. commit e50fbe175fe6b482d4eabf554a9d923fe1cb727a Author: stevenj Date: Tue Jul 5 18:53:36 2011 -0400 support compiling/installing --enable-threads --enable-openmp at the same time, although in this case the test program only uses the threads variety. Update documentation accordingly, and in general expand the documentation of the OpenMP support commit 1b13a7673c31c9f98151186ab5ad96952f0c8cc2 Author: stevenj Date: Tue Jul 5 16:04:03 2011 -0400 call omp_set_num_threads in fftw-bench so that the number of OpenMP threads corresponds with the number of FFTW threads commit fcd3d63bce6f23ca8274e739ca83a0fcb8b63a99 Author: stevenj Date: Tue Jul 5 16:03:06 2011 -0400 when --enable-openmp, install as fftw3_omp rather than fftw3_threads, so that both the POSIX threads and OpenMP variants of FFTW can be installed at once commit 4c8bae967265bf7b4c9705d6efe87cf7e9151fce Author: stevenj Date: Sat Jul 2 02:21:22 2011 -0400 don't even declare an fftw_execute interface in Fortran, since it is unsafe and we recommend against it anyway; thanks to Arjen Markus for the suggestion commit b1741fcc12ccd46d1ce538398c78ca8da98b2448 Author: athena Date: Fri Jul 1 14:35:44 2011 -0400 consistently use the order single, double, long double commit 02d76b0f908814ec69eb9f4edf423e6794d63720 Author: athena Date: Wed Jun 29 17:27:06 2011 -0400 MSVC AVX 64-bit detection does not work, punt for now. commit 5a057b2b67c6eac3cb59c5dd555e1e4093d0ce8e Author: stevenj Date: Wed Jun 29 15:52:27 2011 -0400 fixed typo, added note on transposed flags for r2c/c2r; thanks to Rhys Ulerich for the suggestions [empty commit message] commit 21db43d01a6a55f3bce9bbb3bd01fc968a8fb4ac Author: athena Date: Wed Jun 29 09:41:39 2011 -0400 fixes for compiling with MSVC (untested) commit dcbc5ebfe7dd814f3ef8ee85fb5b2ccb4a3671f0 Author: athena Date: Tue Jun 28 16:48:36 2011 -0400 comment commit f71799bf38f03deaeea50a8b4178757e826854c0 Author: stevenj Date: Mon Jun 27 21:01:56 2011 -0400 rm extraneous line break in HTML output ... I hate texinfo [empty commit message] commit 9ae9c2b534eb9064d8153f235e6d013a4c8b50c1 Author: stevenj Date: Mon Jun 27 00:47:33 2011 -0400 maintainer-clean should delete html directory (otherwwise we keep obsolete HTML files in the dist tarball, sigh) [empty commit message] commit f66d29622c87134e4a790fdab1e25413fac8d33d Author: stevenj Date: Sun Jun 26 23:36:32 2011 -0400 update copyright year in manual [empty commit message] commit 3799446cb5d30354dc69a36f07e8bdf87ed5cb34 Author: stevenj Date: Sun Jun 26 22:52:54 2011 -0400 whoops, don't dist .f03 headers, since those are built by the user's Makefile [empty commit message] commit 8c336f8396e94752233e91433a0e64a72e137599 Author: stevenj Date: Sun Jun 26 22:43:49 2011 -0400 fix embarrassing deadlock/crashing bug in my previous nowisdom_hook fix -- I forgot to handle the case where one process has wisdom and another one doesn't, requiring a nowisdom_hook in the latter case; this should only affect MPI transforms since otherwise these hook functions are NULL [empty commit message] commit e32aa9704f9a6e0811638809bbf764dc748116fb Author: stevenj Date: Sun Jun 26 21:02:15 2011 -0400 subsubheadings, MPI transpose reference [empty commit message] commit e2759a2102797af24072573371a9d94d4943f1ff Author: stevenj Date: Sun Jun 26 20:48:53 2011 -0400 add MPI plan reference [empty commit message] commit 5fd0d86ab1801a09997624a79a56029f76e4c718 Author: stevenj Date: Sun Jun 26 17:07:21 2011 -0400 portions of MPI reference docs; tweaks to NEWS [empty commit message] commit b280b47c049bc941297a69b5d668a7fdf4a81977 Author: stevenj Date: Sun Jun 26 12:40:43 2011 -0400 use $(CHECK_PL_OPTS) more consistently [empty commit message] commit f78b49c77e0442702bb0bb3b1e52b82795c36358 Author: athena Date: Sun Jun 26 10:04:54 2011 -0400 accept \r\n as well as \n. Grrr... commit 4449361639ee599d5221557c1b7021c85954c2d9 Author: athena Date: Sun Jun 26 09:52:11 2011 -0400 new configure option --with-incoming-stack-boundary=N This option selects CFLAGS to align the stack at all externally-callable functions. This currently comprises api/* and threads/* commit 44191f4b3b5109c9e1befb9a3eefb1f34a1fd63e Author: athena Date: Sun Jun 26 09:51:37 2011 -0400 add -fomit-frame-pointer back Somehow -O3 does not imply -fomit-frame-pointer on ia32 commit 94f1e0517794a91b91b81bc46695d0bcf5d23ca9 Author: athena Date: Sun Jun 26 07:20:27 2011 -0400 Note that removal of mips-ps is temporary. commit 6ec5e833bf16b843f2893e894f786a67721cf647 Author: stevenj Date: Sat Jun 25 23:15:03 2011 -0400 update copyright year [empty commit message] commit 27117ddc70e191d20cc88be0a2285f454a1409cd Author: stevenj Date: Sat Jun 25 21:33:13 2011 -0400 updated NEWS [empty commit message] commit 2e1f81718cf9d9073a65e907c6aecebc1333a4a1 Author: stevenj Date: Sat Jun 25 20:29:55 2011 -0400 fixes to Fortran interface and docs [empty commit message] commit 32e0027e573cbfcae6c39e535a1a3549f602b97a Author: stevenj Date: Sat Jun 25 17:43:31 2011 -0400 initial stab at MPI Fortran docs [empty commit message] commit f1b33feb4669f8b92467a448172c0c0734c5ac48 Author: stevenj Date: Sat Jun 25 16:43:31 2011 -0400 correct description of what MPI standard says about I/O (I can't believe this crap) [empty commit message] commit 0329701daca6ace6a4167366fda97a86d7d3b6f4 Author: stevenj Date: Sat Jun 25 15:14:07 2011 -0400 more MPI documentation; mention `fftw_alloc' functions earlier in the manual [empty commit message] commit c4a68ffe9d9332bc0fc9e9db4f41a7c0728ec663 Author: stevenj Date: Sat Jun 25 13:40:19 2011 -0400 clarification about --enable-sse2 [empty commit message] commit cbf6b823ec85dfecab28d071db39d5f92cdcb561 Author: athena Date: Sat Jun 25 13:31:25 2011 -0400 Update mingw build scripts for fftw-3.3 commit 1c97317e9689ac1376ba51c408adde1514475140 Author: athena Date: Sat Jun 25 08:52:13 2011 -0400 Fix typo: EXTRADIST => EXTRA_DIST commit 0ab873d4daf2a047ec04e273f0a8046f8919961c Author: stevenj Date: Fri Jun 24 23:52:19 2011 -0400 finished draft "modern fortran" chapter [empty commit message] commit 35f278113d2e2b4532514f40b44fe468b5c7d729 Author: stevenj Date: Fri Jun 24 20:47:49 2011 -0400 include FFTW_EXTERN prototypes for wrappers, so that they are properly exported to DLLs on Windows (sigh) [empty commit message] commit d2a1f24513b42464d64fa4fea5b97326f4251646 Author: athena Date: Fri Jun 24 16:52:30 2011 -0400 use malloc() instead of alloca() for large buffers The proximate cause for this patch is that OpenBSD/i386 reserves 256KB stack size per thread. We were allocating a buffer of size 128*130*sizeof(fftw_complex) that exceeds the stack. While 128*130*sizeof(fftw_complex) = 260KiB is the worst case for normal configurations, it is a good idea to limit stack allocation just in case. Also, the generic solver might in principle generate unbounded buffers, even though it is normally disabled for n > 137. So, as an added precaution, we now never stack-allocate buffers larger than 64KiB, which ought to be enough for anybody. commit 3b1c71b8e61a7fbfa88589ddf418d494a672ed78 Author: stevenj Date: Fri Jun 24 16:32:30 2011 -0400 don't imply that AVX is available on Pentium III; note that MIPS Paired Single is currently only in FFTW 3.2.x [empty commit message] commit 02153f462b97b2733b47d298a3fc5cc57b45ba86 Author: stevenj Date: Fri Jun 24 16:05:27 2011 -0400 silence annoying gfortran warnings [empty commit message] commit dff007f6cc14bdd1ea710466aa3f6fb3a5408c0a Author: stevenj Date: Fri Jun 24 14:59:30 2011 -0400 a couple MPI Fortran 2003 fixes; changed MPI flags to not use 1<<31 since Fortran (not having unsigned integers) does not allow us to declare that constant in a portable way [empty commit message] commit ff330ebfa25ed56b29ea61ea99bf293cec079dde Author: athena Date: Fri Jun 24 15:05:05 2011 -0400 Fix libtool shared version info. FFTW-3.3.x should be a direct drop-in replacement for all FFTW-3.x.y versions. commit 6c3c5cd3040d318a184fc0e6ee6ee4e3429ef8ce Author: stevenj Date: Fri Jun 24 14:38:47 2011 -0400 add MPI Fortran API and wrappers [empty commit message] commit 02d3e72585a1254f2685014f5f8de5c7730b8ec5 Author: athena Date: Fri Jun 24 14:51:12 2011 -0400 Do not require fig2dev on the user's machine Distribute the manual's figures in PDF/PS/PNG form instead. commit 5169fc22863b9b6ea7bfbaafc0ca523e82114ee1 Author: athena Date: Fri Jun 24 11:52:44 2011 -0400 Remove --enable-portable-binary, --with-gcc-arch from documentation. commit f6d1274e5c9cc173d07c58df0a3535c9f4767e48 Author: athena Date: Fri Jun 24 11:48:48 2011 -0400 Forget about specifying nonportable CFLAGS. Let the user do it if he wants. commit 74872e79034a2379c537c23a7c29d0b6f43d2437 Author: athena Date: Fri Jun 24 11:48:25 2011 -0400 Add "-avx" to version string when appropriate. commit 2d6800ac3b8070da86d6d825d89fea05fbd44b78 Author: athena Date: Fri Jun 24 10:26:38 2011 -0400 change 3.3-alpha => 3.3-beta1 commit 97a4d17f30831d71b4075eec8b7f4b7c12deedb7 Author: athena Date: Fri Jun 24 09:25:49 2011 -0400 Extend OUR_MALLOC16 to larger alignments Make it work for 32-byte alignment and beyond, as needed by AVX. Rename --with-our-malloc16 to --with-our-malloc. Keep old --with-our-malloc16 flag for compatibility. commit fd31e415cd0731c428daafe791386d79ff34b8ca Author: athena Date: Fri Jun 24 09:19:38 2011 -0400 Fix typo commit 4f8a370b687860b92a93c49fed128218e6fb9f9b Author: athena Date: Fri Jun 24 09:10:26 2011 -0400 One pass over the manual. commit ba838fa07395a4f365eb16aa8ba1bb108f533dd5 Author: athena Date: Fri Jun 24 08:19:03 2011 -0400 eliminate the WITH_ALIGNED_STACK hack This is 2011 and I have no system with incorrect stack alignment. commit 7e32fb649dcd2a78a3b2d216140fc218cb69c334 Author: athena Date: Fri Jun 24 07:49:47 2011 -0400 enable both threaded and unthreaded wisdom in tools/fftw-wisdom commit 7543b3029bce4fc595c5efc7e0d31a71b4ec5cce Author: stevenj Date: Fri Jun 24 02:40:04 2011 -0400 clarification [empty commit message] commit 81589ce427090a8aea9f4362a9b7f9ba7e76e111 Author: stevenj Date: Fri Jun 24 02:24:01 2011 -0400 check for error code in example [empty commit message] commit 3c57716a6525c37f0f485e925c7df8f9819c66fd Author: stevenj Date: Fri Jun 24 02:22:18 2011 -0400 cleanup - since NATIVE_MALLOC is always malloc, delete this #define [empty commit message] commit 644b3ee0d54eb80e78a35710d2b8027f3104fbe6 Author: stevenj Date: Fri Jun 24 02:19:44 2011 -0400 document wisdom string import/export in Fortran [empty commit message] commit bee8d24fa16cd40bba5612938edc915399526c03 Author: stevenj Date: Fri Jun 24 02:11:40 2011 -0400 bug fix - NATIVE_MALLOC should always be plain malloc, even in debug_malloc mode, because it is used in the API to return things that should be deallocated with free(); correspondingly, be sure to use free() ansd not X(free) with this [empty commit message] commit 119eb3d276de6b3178c5436fd0cee98a35f0aabf Author: stevenj Date: Fri Jun 24 01:35:27 2011 -0400 declaration style [empty commit message] commit 8916c213d4b3121e65af9ee4989ed8ae0fc21db9 Author: stevenj Date: Fri Jun 24 01:25:36 2011 -0400 document wisdom file export/import from Fortran; add export/import_to/from_filename functions for convenience [empty commit message] commit 9fb007e826ee94927e9ff4a9de14c6b80ec06e69 Author: stevenj Date: Thu Jun 23 19:19:43 2011 -0400 more fortran docs [empty commit message] commit 3f0d26b4fd7e8e501fbd2f6ff24337ff0ad3c97c Author: stevenj Date: Thu Jun 23 17:50:30 2011 -0400 enforce 132-character line-length limit that is the default in Fortran [empty commit message] commit 6e69de25872aed3a4eb13523a43bcfdf9a4f731c Author: stevenj Date: Wed Jun 22 23:27:31 2011 -0400 the F03 standard is ambiguous about whether types can be assigned to wider types as formal parameters with VALUE attributes, and e.g. gfortran interprets it to disallow this code [empty commit message] commit a91a52952fb6d32423f351afdda8de2d04b71e38 Author: athena Date: Thu Jun 23 18:12:10 2011 -0400 Add md5 hash of fftw's configuration to wisdom file People were already confused by threaded vs unthreaded wisdom, and now things will be even worse because we enable/disable AVX codelets at runtime. Accept incoming wisdom only if it was produced by the same configuration (modulo MD5). commit 0de6ca5c7f7720457124d7b7a3ca35153d06f761 Author: athena Date: Thu Jun 23 09:01:27 2011 -0400 distribute fftw3.f03.in commit 6422ed65a6cfb1704c1b428f37a6034baeba212c Author: stevenj Date: Wed Jun 22 22:02:18 2011 -0400 more Fortran documentation [empty commit message] commit 9c1e1d48f0f97b49b62e5ee42e92673964e7d7cf Author: stevenj Date: Wed Jun 22 20:10:39 2011 -0400 correct comment [empty commit message] commit ce8bb23e571efe5595d6d9578a45769e48bad7fe Author: athena Date: Wed Jun 22 20:26:18 2011 -0400 Use "sh FOO.sh" instead of "./FOO.sh" to avoid chmod +x. commit 660905636ebbefe3338d405d4a6fb30b89e07912 Author: stevenj Date: Wed Jun 22 19:19:05 2011 -0400 document fftw_alloc_real/complex ... should we switch to using these in the tutorial examples? [empty commit message] commit ea8c9f7fedc0ca183449b70bee978a4251f552e1 Author: stevenj Date: Wed Jun 22 19:07:49 2011 -0400 whoops, added missing file [empty commit message] commit 6caf08feba78084741e749e022c8c2e523c7ab44 Author: athena Date: Wed Jun 22 18:46:01 2011 -0400 Note addition of AVX. commit a17b6a6a493d19c93fd3a6b8eaacef8174033cd7 Author: athena Date: Wed Jun 22 18:43:43 2011 -0400 In SSE2, AVX: use FMA macros when applicable. Makes it easier to play with fma4 and fma3 when it comes out. commit d4dfffc05a4f9c3f7aa7c2c2ef9fa416f9257a5f Author: stevenj Date: Wed Jun 22 18:16:45 2011 -0400 all modern Fortran compilers can call FFTW's C interfface directly -- support this, and in particular generate a Fortran 2003 interface file from fftw3.h so that Fortran code calling FFTW can be typechecked ((addressing the source of a lot of Fortran-user problems) [empty commit message] commit 10a9a86ddf3d94750c57ca8b20c0fca39cea6541 Author: athena Date: Wed Jun 22 13:10:02 2011 -0400 some cleanup of SSE2 macros commit f301a0adb53b56a2e4fd74ef61cf29d66b745286 Author: athena Date: Wed Jun 22 07:38:18 2011 -0400 don't use -xHost on ICC -xHost with ICC is problematic. On icc-12.0.0, "-mavx -xHost" overrides -mavx with -xHost, generating SSE2 code instead of AVX code. ICC does not seem to support -mtune=host or equivalent non-ABI changing flag. commit 9e7758ff431947863cec44354413c27067f0fda8 Author: athena Date: Tue Jun 21 20:35:36 2011 -0400 Complete AVX implementation for split codelets commit a547e3d5dba38863f6e2c9acb45ffb94351fe3a5 Author: stevenj Date: Tue Jun 21 19:37:14 2011 -0400 whoops, missing altivec conf patches [empty commit message] commit 9dc6263714fa685f384fb61ed08d398b38b26329 Author: stevenj Date: Tue Jun 21 19:12:45 2011 -0400 some BSD ar versions (e.g. on MacOS X) give an error if there are no object files, so we cannot build empty libraries [empty commit message] commit 32dcba1b5d68cf9e33ec3f81e0405f2f909389dd Author: stevenj Date: Tue Jun 21 19:12:12 2011 -0400 re-insertion of Altivec code [empty commit message] commit 931617ee1ba9f9257117ad8e1df38dfe055f9cee Author: athena Date: Tue Jun 21 16:26:09 2011 -0400 Implement faster AVX loads/stores. commit f76d6c1d278b34b2ccac7cff57522cab7ec90864 Author: athena Date: Tue Jun 21 16:03:24 2011 -0400 Initial AVX256/single implementation This should be correct but slow. I need to figure out how to implement noncontiguous loads/stores efficiently. commit 164cc4c8fa9bfdf2b02d9cb9364c8f3f36e420e7 Author: athena Date: Tue Jun 21 14:13:57 2011 -0400 fix AVX alignment commit c277833627164aeb649d187ba0409a3fdca9166d Author: athena Date: Tue Jun 21 14:07:28 2011 -0400 rename avx256d -> avx AVX will work in both double and single precision, like SSE2. commit f97162a135eee43630825fccaa29b735ee284fff Author: athena Date: Tue Jun 21 13:52:20 2011 -0400 remove CODELET_OPTIM In the old 32-bit gcc-3.x days we used to play games with gcc to force it to produce decent code. Now gcc has gotten smarter and it produces indecent code no matter what we do, so it is safe to remove these hacks. commit 57baa1a6761a441c6aa47b0b63503bfd5270d9a9 Author: athena Date: Tue Jun 21 09:57:31 2011 -0400 work around gcc/icc quirks commit deeea8205c4cf4511b41a8a7cd85195b1e4eb7f6 Author: athena Date: Tue Jun 21 09:56:07 2011 -0400 Add remarks in places where we work around gcc quirks commit b357b3a90ba2060915fa03d888f0b2306be1f17a Author: stevenj Date: Mon Jun 20 21:17:59 2011 -0400 remove the libbench directory (which we have kept lingering in the repository for years due to CVS's inability to remove directories) [empty commit message] commit 1ff7bbcc3b0dcfa3bd8f16ec1c672e5671cc0f99 Author: stevenj Date: Mon Jun 20 21:17:14 2011 -0400 update URLs [empty commit message] commit ddcac323fe11f273f8e983836edaec002237a3ca Author: stevenj Date: Mon Jun 20 20:53:31 2011 -0400 whoops, forgot to check in alignment change [empty commit message] commit 5eac5571d61bf5515946dab5750f5db82632cd54 Author: athena Date: Mon Jun 20 20:22:23 2011 -0400 "test X = Y" requires spaces around "=" commit 60d4535475d937207b8fd0c6ebef966d95601154 Author: stevenj Date: Mon Jun 20 19:18:52 2011 -0400 indenting [empty commit message] commit 22cdbb3e3f35d6c289f4c250fbeba696090c6cc9 Author: stevenj Date: Mon Jun 20 18:57:10 2011 -0400 merge back in SSE support, now combined with SSE2; --enable-sse2 now works in both single and double precision, and simd-sse2.h contains both the double- and single-precision code (which overlap a lot); in single precision it is still compiled for SSE-only (SSE2 is only required for double) [empty commit message] commit 2d767316e1ba0cf9fd4f5eb3134c6341b2d87a29 Author: athena Date: Mon Jun 20 16:02:07 2011 -0400 Implement AVX autodetection (gcc-only so far) commit 1ed535ea5c0ae847edb64b1696c7c40ea6022fbd Author: athena Date: Mon Jun 20 14:25:54 2011 -0400 Add VZEROUPPER at the end of AVX codelets If the Intel Optimization Manual is to be believed, we need to wave a dead chicken before transitioning from AVX code to SSE code. I am supposed to believe that there is a transition penalty for doing so, unless one uses a magic VZEROUPPER instruction that apparently has zero cost. Whatever. commit 1b26ff69ef0065d12689cd77ae65a7a049a37150 Author: athena Date: Mon Jun 20 10:21:25 2011 -0400 Move RDFT to new simd scheme commit 02b63c9ba5acf94a24d0b948436026df702681a9 Author: athena Date: Mon Jun 20 09:23:38 2011 -0400 New SIMD build system We now support multiple SIMD extensions in the same binary, e.g. --enable-sse2 --enable-avx. This patch adds the necessary infrastructure for SSE2/AVX and complex DFT. Later patches will add RDFT and SSE/ALTIVEC/etc. commit 3409ea120286bc180d314be65f949ecb62f954cb Author: stevenj Date: Sun Jun 19 12:29:27 2011 -0400 italicize Latin quote [empty commit message] commit 760f9aec6ca8c45219a357605b8216fef71ff04f Author: stevenj Date: Sun Jun 19 12:26:34 2011 -0400 work around incredibly annoying makeinfo bug -- for HTML output, in any paragraph ending with an @index command, two blank lines are needed to create a paragraph break ... our HTML output has apparently been screwed up for years [empty commit message] commit f7a34a1e53ec9e1b7c263d6c8a4cf8020c1de89e Author: stevenj Date: Sun Jun 19 12:01:39 2011 -0400 conjugate-pair algorithm turns out not to be due to djb, but it was pointed out to us by djb [empty commit message] commit 34e740878c4a449ef31a6df2f538a67387504b57 Author: stevenj Date: Sun Jun 19 11:59:21 2011 -0400 new-array execute functions are *not* the same for MPI, since the problems are different [empty commit message] commit a0b07a03fd52e6192dd0023054cc6359dd651554 Author: stevenj Date: Sun Jun 19 11:57:52 2011 -0400 tweaks to MPI manual [empty commit message] commit 9fa05d4bcd100d5ed0ade53a1c28f00a01cd89fc Author: stevenj Date: Sun Jun 19 00:47:23 2011 -0400 only call MPI_Init_thread for MPI version >= 2 [empty commit message] commit e2f3cf7c8965653eb94a03c25ab7ce8f4f09ed8e Author: stevenj Date: Sun Jun 19 00:35:44 2011 -0400 document quad precision in Fortran [empty commit message] commit f5dd231509aa0a32aa1d0fa1024adf8ebba30aa9 Author: stevenj Date: Sun Jun 19 00:16:10 2011 -0400 use -lquadmath for quad-precision library in pkgconfig file [empty commit message] commit d0775d11f358dc46f0f2925be4830b3127564ab4 Author: stevenj Date: Sun Jun 19 00:10:33 2011 -0400 document quad precision [empty commit message] commit a6142f7e745906e6b392369e5ffc5bf83f57d573 Author: stevenj Date: Sat Jun 18 21:26:24 2011 -0400 there is currently no standard quad-precision type in MPI, so don't allow MPI support to be compiled with --enable-quad-precision [empty commit message] commit 34067999298aa71f65b2ac33386693f03bebd725 Author: stevenj Date: Sat Jun 18 21:25:51 2011 -0400 make quad-precision library libfftwq [empty commit message] commit 28db5c98edf2ce5508cc8a91118b41ede6476183 Author: stevenj Date: Sat Jun 18 21:19:50 2011 -0400 use --estimate in check script with --enable-random-estimator [empty commit message] commit e16e119568222dd95a717242b191a4adf72ec2de Author: stevenj Date: Sat Jun 18 20:36:51 2011 -0400 bug fix - correct crashing interaction between threads and debug-malloc [empty commit message] commit 25a1b5252eb203919634a13d0c5deb8f8ddad940 Author: stevenj Date: Sat Jun 18 18:41:48 2011 -0400 whoops, typo [empty commit message] commit 2d8b12a9624e30de0054d035d35474c57d115f49 Author: stevenj Date: Sat Jun 18 18:35:14 2011 -0400 clarification of when fftw_cost may return 0 [empty commit message] commit 9d7a9545df20eb5b0c4ea6535161108225abce21 Author: stevenj Date: Sat Jun 18 18:31:26 2011 -0400 corrected manual and test program for proper interaction of MPI and threads [empty commit message] commit a950b94a168ed22d850db5394fd633eee3309ebf Author: stevenj Date: Sat Jun 18 17:13:52 2011 -0400 need --mpi restriction in mpi+threads check [empty commit message] commit 7a8106a510a8147ce2f8f986a0ea6f1fc935b419 Author: stevenj Date: Sat Jun 18 12:09:04 2011 -0400 split fftw3.texi into multiple files for ease of editing [empty commit message] commit cb26916dace1d5d264f7561b760a89d5ee972d3b Author: stevenj Date: Fri Jun 17 17:26:50 2011 -0400 merge recent Cell deletion with MPI branch [empty commit message] commit 1595e9847b428d6b439d7f908d3d6f06b6746d1f Author: stevenj Date: Fri Jun 17 01:52:59 2011 -0400 whoops, incorrect assertion [empty commit message] commit 158a22057fd700263ff39e20dafbf506982bad25 Author: stevenj Date: Fri Jun 17 01:52:51 2011 -0400 comment fix [empty commit message] commit e0d118305f05ad4f429fda2879547b9285c362ea Author: stevenj Date: Thu Jun 16 23:30:27 2011 -0400 check if pln creation failed (e.g. for split input) bbefore calling setup_gather_scatter, to prevent crashes [empty commit message] commit b2254795d8d3f65d1205053f39d5e837bec6e4d0 Author: stevenj Date: Thu Jun 16 23:26:48 2011 -0400 bug fix -- transpose-recurse is only applicable if subtransposes fit in the same space (unless I change the allocation routine, but this would seem to require looking at all possible recursive invocations of transpose-recurse) [empty commit message] commit 25faa34d32d8b7577ad1107769f3e87e7c87cef0 Author: stevenj Date: Mon Apr 11 17:58:03 2011 -0400 yikes, any_true check on subplan creation should be in comm, not comm2, so that all processes know if failure occurred [empty commit message] commit c0b90d9331fbcc167f07c04a3ce1298cc8d88593 Author: stevenj Date: Mon Apr 11 17:00:46 2011 -0400 add wisdom_ok_hook to enforce wisdom synchronization on MPI problems, apparently fixing a longstanding deadlock/crash bug [empty commit message] commit 651a2f59b8fdeecf79246c3f65b776be567b2ddb Author: stevenj Date: Mon Apr 11 15:10:22 2011 -0400 add a check (in DEBUG mode only) that all processes produce the same hash of MPI problems; don't include alignment in MPI problem hash because it may differ between processes for unaligned malloc [empty commit message] commit 23bb1cb665dc571a773eb5125371b1079e0a4243 Author: stevenj Date: Fri Apr 8 18:46:54 2011 -0400 use cost_hook in random_estimate [empty commit message] commit 21229fc46c88d7ca15c9ba06c1f731d01eecd77a Author: stevenj Date: Sun Mar 6 23:33:53 2011 -0500 added mpi new-array execute functions; thanks to Guo Luo for the bug report [empty commit message] commit 7335ef11cd5e2e9962dbc41c42e5c14e156e6f82 Author: stevenj Date: Wed Feb 9 21:29:17 2011 -0500 MPI may not support tags > 2^15-1 (e.g. Cray MPI requires tags < 2^24); thanks to Jonathan Bentz for the bug report. [empty commit message] commit 155887d9e173f2a72cb63cf5b7b36ff49bf52356 Author: stevenj Date: Wed Feb 2 12:21:30 2011 -0500 fix merge conflicts [empty commit message] commit 56c3bf01ecfcc23f2d76efed77dac234a59b9234 Author: stevenj Date: Sat Nov 15 20:33:33 2008 -0500 version bump for 3.3alpha1 [empty commit message] commit 136cf63d2cdcb5889a38163c4ccf9b1198e47e04 Author: stevenj Date: Sun Oct 26 22:47:07 2008 -0400 re-added mpi/Makefile [empty commit message] commit 3bfab1285c5390285e734dc910b728a328d9f7fd Author: stevenj Date: Sat Oct 25 17:14:42 2008 -0400 re-add MPI to dist [empty commit message] commit b2470178928b190c6b50d3bef353925061db3d08 Author: athena Date: Sat Jun 18 08:50:13 2011 -0400 remove obsolete Cell code commit 38bfc62ffa5051da294faa46a8ab29fc7225a97b Author: stevenj Date: Fri Jun 17 23:31:33 2011 -0400 bug fix in accuracy test, which prevented us from consiistently determining accuracy in > double precision [empty commit message] commit 7356645134bcb66286da0c00ad3d993e90e425af Author: athena Date: Fri Jun 17 20:05:13 2011 -0400 do not check for gcc version before checking for gcc commit d94f234b00d3ced13ccfc3551e0a20cd66645922 Author: stevenj Date: Fri Jun 17 18:56:37 2011 -0400 require gcc 4.6.0 or later for --enable-quad-precision, to match fftw3.h header file; no need to mark this as EXPERIMENTAL (make check passes, and support in gcc 4.6 seems reasonably complete) [empty commit message] commit 28ebf4da08a1fc607b0ea41ed24ea3707a256548 Author: stevenj Date: Fri Jun 17 18:51:41 2011 -0400 need ugly __attribute__ to use __float128 with _Complex, ugh [empty commit message] commit b00cf20a5c45b953d2e6d954570543727b72cf9a Author: stevenj Date: Fri Jun 17 18:23:05 2011 -0400 --verify tolerance in quad precision changed to 1e-29 [empty commit message] commit 015e365952344e8395a0516c9fe0e3c736550b9e Author: stevenj Date: Fri Jun 17 18:22:38 2011 -0400 quad-precision F77 api should use "qfftw" prefix [empty commit message] commit af4a1c37d8eb5e14678449741fd185cb65c4c10b Author: stevenj Date: Fri Jun 17 18:22:27 2011 -0400 rm extraneous space from fftw3.h [empty commit message] commit 654e73b0cfe2079498eadaf15b19361f71ef18ab Author: stevenj Date: Fri Jun 17 18:05:10 2011 -0400 use cosq etcetera with libquadmath in libbench2, so that --verify correctly gives ~33 decimal places in shift test [empty commit message] commit 00bac1ae1e651e5e85c507294c3e215dde1bb85b Author: stevenj Date: Fri Jun 17 17:52:51 2011 -0400 libquadmath ships with gcc 4.6.0, so we should require this library for sinq/cosq with --enable-quad-precision; also, include the __float128 FFTW functions in the header file for gcc >= 4.6 on i86/x86_64/ia64 [empty commit message] commit a17a0720767ca177a799d685a5ac7b63331dd559 Author: stevenj Date: Fri Jun 17 16:54:01 2011 -0400 typo in manual for fftw_cost [empty commit message] commit 4aab5d51dbc8b798c25f857cfa07ce7f25f9360d Author: stevenj Date: Fri Jun 17 16:48:24 2011 -0400 fix fftw_cost function: pcost needs to be saved in mkapiplan, since the plan is re-created from wisdom [empty commit message] commit c031d561c14a97b9a04197ba07c19cbf769648f9 Author: athena Date: Fri Jun 17 16:42:25 2011 -0400 removed support for the Cell Broadband Engine commit 34f5ad2eaae86075973d26ee5adcc7ceb04f4924 Author: athena Date: Tue May 24 06:51:07 2011 -0400 Undo previous change; the typo was not a typo after all. commit b9b72d1a42b162f74dfe9dc073ce5f5fc873113b Author: athena Date: Mon May 23 05:08:05 2011 -0400 Fix typo in manual commit f41fb798be412eff5c26c539e07385dbbc6ac56a Author: athena Date: Sat May 21 17:37:50 2011 -0400 clarify intent about canonicalization of tensor in tensor_compress_contiguous() commit 438c3b46f428023ce986422a697c9d33687ae9c1 Author: athena Date: Sat May 21 17:30:31 2011 -0400 avoid useless canonicalization in tensor_compress_contiguous() commit 8fbf77ca71756bd32bb5c71e709ec717e388451e Author: athena Date: Sat May 21 17:24:57 2011 -0400 Fix tensor_compress_contiguous tensor_compress_contiguous() was supposed to sort dimensions by descending istride, and then compress adjacent dimensions. This property was lost once we changed the canonical order of strides to be sorted by descending min{istride,ostride}. Change tensor_compress_contiguous() to sort by descending istride again, which is necessary for its correctness, and then canonicalize at the end. commit 5ad70bf5c1bdeadefcaac86cc3e9a76b31f75724 Author: athena Date: Sun May 8 18:47:26 2011 -0400 Don't distribute obsolete .depend commit c48d28bc5b7fb54ea8e037579cb0af3fae339543 Author: athena Date: Sun May 8 18:05:36 2011 -0400 Use ocamlbuild for building genfft Remove the old Makefile cruft to support ocaml, and use ocamlbuild instead. commit d183b35663d030b1ad789795fa441941961472c0 Author: athena Date: Sun May 8 18:03:07 2011 -0400 Do not use __float128 unless BENCHFFT_QUAD is defined Otherwise, compilation fails on compilers that do not support __float128. commit 833ec738fe3c3234382a3fc160c8fe54060dc860 Author: stevenj Date: Fri Apr 8 13:15:54 2011 -0400 fix configure --help string for --disable-alloca (since default is enabled) [empty commit message] commit 68538e1ffa84d206cae95558c636d9fa490888bd Author: stevenj Date: Fri Apr 8 13:09:56 2011 -0400 add "random estimator" for debugging purposes; note that this is best used with ESTIMATE_PATIENT mode [empty commit message] commit 56d274a97c1868b43a2294bab6a4d6d245849608 Author: stevenj Date: Tue Apr 5 14:47:56 2011 -0400 add AC_CHECK_DECLS for srand48; thanks to Ralf Wildenhues for the bug report [empty commit message] commit 50465ef2118b72f9f868e9ec53ee7e53bb050259 Author: stevenj Date: Sat Feb 5 17:00:40 2011 -0500 experimental support for gcc's __float128 quad-precision type [empty commit message] commit 2471f34097daef93ae593743403cf40820a0673a Author: athena Date: Sun Oct 24 14:33:59 2010 -0400 guarantee that "timelimit < 0" means "no timeout" "timelimit < 0" was always meant to be equivalent to "timelimit = HUGENUM", but this was not true in all cases, causing some obscure wisdom behavior. Thanks William Andrew Burnson for the bug report. commit c16bc87b770bb0757dead223c13dc1966e3c2e3e Author: athena Date: Sun Oct 24 14:32:20 2010 -0400 compile with --enable-fma and SSE, SSE2 Allow compilation with --enable-fma and --enable-sse, --enable-sse2. This is a bad idea performance-wise, but people will try anyway. commit eb34fdf7b5233e8f8a5b44c7f275e0c950ead3e7 Author: athena Date: Sun Jul 11 13:34:06 2010 -0400 Make threads.c compiler with c++ commit a80ce9ee8210615480efcaf03989278540ad900e Author: athena Date: Sun Jul 11 10:05:05 2010 -0400 Attempt at clarifying the advanced interface doc. commit 537372cf3446b014e53ad2b2bfd636748abfe44f Author: athena Date: Sun Jul 11 07:37:27 2010 -0400 rename rfftwnd html picture It turns out that texinfo with pdf output reads .png files in preference to .pdf files (when did this change?). I renamed the .png figure to avoid producing an ugly pdf file. commit 7be5dbd77b2c719b804b53731ffc9e27100b48f6 Author: stevenj Date: Tue Mar 30 19:43:22 2010 -0400 added fftw_cost function; this is the second time people have asked for this, and there is a reasonable use for it in comparing e.g. oout-of-place vs. in-place plans [empty commit message] commit 101fc17a6dbc1feb1e5cc7625a251068fac7c202 Author: stevenj Date: Tue Mar 2 18:55:49 2010 -0500 documented that --enable-debug-malloc causes fftw_execute to be thread-unsafe (thanks to Alexis Rohou for the problem report) [empty commit message] commit e9b894f0a2b7d65d559d8fded2d7298bdfc90694 Author: athena Date: Fri Jan 22 19:42:08 2010 -0500 Added FAQ about how to transpose matrices using FFTW. [empty commit message] commit 18462a4c21b99482fcb9b2dc7338b83f8bbd790b Author: stevenj Date: Thu Jan 7 20:16:57 2010 -0500 catch FMS (instead of generating FMA(_,_,NEG(_)) with h -generic-arith option [empty commit message] commit 90015026798cd3bad02d8c4144f527dfdc5f1ca7 Author: athena Date: Fri Dec 11 07:01:26 2009 -0500 note future wisdom enhancements. [empty commit message] commit 03747c3de44cda09224d08e8d580483cc23b6687 Author: athena Date: Mon Oct 19 20:21:05 2009 -0400 Use SIMD flags when checking for xmmintrin.h This prevents an obnoxious warning from configure. commit 9211b77226bd67a674d5be4b26843b466a24d377 Author: athena Date: Sat Aug 29 20:47:56 2009 -0400 new bug [empty commit message] commit 4d7817c4b3f6476892515b47aca61d2830ba2e5c Author: stevenj Date: Sun Jul 26 00:40:11 2009 -0400 typo (s/man1/many) [empty commit message] commit 4de43b59f0ef8a445810e2b96e746d95a63b39b1 Author: stevenj Date: Tue Jul 14 14:19:08 2009 -0400 BUILD-MINGW32 script, updated Windows README [empty commit message] commit fe7ce32f22a1cfe7049d095ecbea5024915e93ac Author: fftw Date: Mon Jul 13 09:40:38 2009 -0400 cleanup BUILD-MINGW64.sh commit 18ddd3f4d23697d8f6dcbf9c122ca281ed17bdde Author: athena Date: Sun Jul 12 06:34:46 2009 -0400 Update NEWS, version number for 3.2.2 release. commit 54f4ad016522892f54c2955ecaa5dc06dbff260a Author: athena Date: Sat Jul 11 22:28:38 2009 -0400 Reintroduce the pruning heuristic in ESTIMATE mode for r2r problems. Somehow, we lost this feature between fftw-3.1.3 and fftw-3.2. commit 18e2d0a785627f6360e756d1cd93c78a8f7b6ba3 Author: athena Date: Thu Jun 25 07:39:04 2009 -0400 don't use pshared=1 in sem_init pshared is really not necessary, and it is not supported on GNU/kFreeBSD. Thanks Petr Salinger for the bug report. commit e58f95716b84021e2175673ccbbd4fe1bc180ad3 Author: fftw Date: Thu Jun 11 19:35:40 2009 -0400 Add mingw64 build file so that we can track it. commit 03c7f0e0ade24de0e4d08bc6899f78db917ffaca Author: fftw Date: Wed Jun 10 12:10:58 2009 -0400 note 3.2.2 NEWS commit ddd2281898dca1a30c48cff89c42c2d9c631faa3 Author: fftw Date: Wed Jun 10 12:04:54 2009 -0400 add --disable-alloca to configure It looks like alloca() is broken on mingw64, and thus we need to disable it explicitly. commit 902651afe12f9de0efb131bffa42db8189516595 Author: athena Date: Sun Apr 26 16:33:10 2009 -0400 Note in FAQ that --enable-k7 has been discontinued. commit 0d50e05674406773adea157318c85b8f9c94d9c0 Author: athena Date: Tue Mar 24 09:16:18 2009 -0400 clarified small confusion in fftw_cleanup documentation commit bea8d6909782b15db9d6a591c8344e8179444746 Author: stevenj Date: Thu Mar 19 13:18:06 2009 -0400 fix documentation of dfftw_init_threads to indicate thaat it takes an argument (since the C version returns a value); thanks t Hans Johnnston for the bug report [empty commit message] commit d6eaf145d2cc51db18fd69b49ef24cf628313c01 Author: fftw Date: Thu Mar 12 13:12:13 2009 -0400 if possible, use a 128-bit type for copy commit b1d732a8e58c303b9f3be4feb082f5b5ac27628c Author: fftw Date: Tue Mar 10 12:49:51 2009 -0400 add size-128 simd codelets It's about time commit d1f4ac081fe3fa267db692b6da4da3b2023e2e8a Author: athena Date: Mon Mar 9 20:29:16 2009 -0400 copy two floats as a double when possible Resurrect the old hack of copying two floats as a double, which makes some difference in these days of 64 bit boxes. commit aa6865bac3cfb4050d6f662ddf744c86c8324393 Author: athena Date: Sun Mar 8 18:08:04 2009 -0400 fixed (harmless) confusion of strides RS and VS were swapped in dftw-direct.c. This is a bug, but it is harmless unless one uses fixed-stride codelets, which we do not. commit f34f1f3fee5d0bd112c8e55c9292d47bd643552f Author: athena Date: Sun Mar 8 10:29:49 2009 -0400 oops, I checked in debug code accidentally. commit d5a07965857012694d310ac28800d47741abccfd Author: athena Date: Sat Feb 14 19:01:00 2009 -0500 Change TLO email address since Magdalen is no longer there. [empty commit message] commit 8418ed8d856c8c50e6463828e015c9d80be1285d Author: stevenj Date: Sat Feb 14 18:18:45 2009 -0500 quote arguments to bench in test script on the off-chance that '*' would be expanded by the shell into a valid filename, and also to avoid shell confusion on Cygwin that "//" begins the name of a Windows network mountpoint [empty commit message] commit 55f747d248139ddffad00ede4d649fc6eb612839 Author: athena Date: Sun Mar 8 10:02:59 2009 -0400 stricter conditions for Cooley-Tukey being ugly It turns out that m=2 in the leaf of Cooley-Tukey may be advantageous in certain cases, eg. i512v512 on AMD Shanghai: (dft-buffered-512-x128/512-6 (dft-ct-dit/4 (dftw-direct-4/24-x128 "t2fv_4") (dft-vrank>=1-x4/1 (dft-ct-dit/64 (dftw-direct-64/504-x128 "t2fv_64") (dft-vrank>=1-x64/1 (dft-direct-2-x128 "n2fv_2"))))) (dft-r2hc-1 (rdft-rank0-tiled/2-x128-x512)) (dft-nop)) Presumably this works around the 2 way associativity of the L1 cache. commit b2acc4c668acebf2ded61cc3a939606bbc73a3e0 Author: stevenj Date: Mon Feb 9 19:46:00 2009 -0500 disable Windows QueryPerformanceCounter code, since it requires us to pull in windows.h in ifftw.h and causes namespace conflicts; gettimeofday seems to work well enough and has had few complaints [empty commit message] commit 1b3884da38a34cbc1c8f33f78394eeca588f0786 Author: stevenj Date: Wed Feb 4 22:55:54 2009 -0500 version bump for 3.2.1, updated NEWS [empty commit message] commit e12e5cb04667aa089cf606fb163ce788247d1c2b Author: stevenj Date: Wed Feb 4 22:27:28 2009 -0500 recommend that users avoid fftw_execute in Fortran, instead using dfftw_execute_dft and friends so that the compiler knows that the input/output arrays are used [empty commit message] commit db43aa9ae291dd66fc542c13afc2dd577750ef75 Author: stevenj Date: Wed Jan 21 16:02:08 2009 -0500 prefer windows queryperformancecounter to gettimeofday on Windows, thanks to David Price for the suggestion [empty commit message] commit 338b5272f6924179a0b345e70f44fd1e5edafc23 Author: athena Date: Sun Feb 1 14:34:49 2009 -0500 compilation fixes in case snprintf() is defined as a macro. [empty commit message] commit 321141568010f66f31c36912a230005ab10d54d3 Author: athena Date: Wed Jan 28 20:19:04 2009 -0500 Automake does not like continuation lines beginning with a comment. [empty commit message] commit 618225923a36a2ef96214e2f2a0c6c12b0fb89be Author: athena Date: Wed Jan 28 18:24:39 2009 -0500 Add r2cb_2.c r2cb_2.c is needed for problem rb2, which is not equivalent to rf2 (unlike kb2, which is equivalent to kf2). This change would not matter much except that rb2 is generated when reducing backward rdft2 to dft, and the absence of the codelet was preventing radix 2 from being employed at all in this case. commit dc4c6cb9e1ae4df2be85e77c3fea172f24e1523b Author: athena Date: Sat Jan 10 06:47:22 2009 -0500 handle the case vecsz->rnk == 0 correctly. [empty commit message] commit 3ca4f694d0b27bb0b1e84ea028e3dadcfdf5b236 Author: stevenj Date: Fri Dec 19 15:20:36 2008 -0500 Macs are no longer ppc-based; thanks to Charles Collicutt for the FAQ update [empty commit message] commit bbfa5e2c5102a5f923eed3b31d37ec7b75616edd Author: stevenj Date: Mon Dec 8 18:08:33 2008 -0500 use new multiple-nbuf code in rdft/buffered, like for dft/buffered [empty commit message] commit 679ab9ffd5738b9df115adfb64f72fd015fe7d6b Author: stevenj Date: Sat Dec 6 16:34:36 2008 -0500 make x86_cpuid macro work on x86_64 [empty commit message] commit 9771718d2e4e57450b03c89bb0c06663c59242c8 Author: athena Date: Sat Dec 6 09:20:37 2008 -0500 Allow automatic choice of buffer size in dft/buffered.c Try a couple of different buffer sizes in buffered transforms, since this seems to make a difference on some Core2 models. commit c4888a4f4fb2961e0f067c55489989da8f3223f5 Author: athena Date: Tue Dec 2 19:18:30 2008 -0500 libbench2: do not assume that split-complex arrays are stride-1 [empty commit message] commit e05f9234129ed82f1f5094465788742c271d1f4b Author: stevenj Date: Tue Dec 2 18:39:43 2008 -0500 updated NEWS [empty commit message] commit 3cfe589fdc4393ff549c1bacbeb2b23c27562339 Author: stevenj Date: Tue Dec 2 18:30:00 2008 -0500 date fix [empty commit message] commit 81326cf16c11ddc12f3d3c1fda7861853abba308 Author: stevenj Date: Tue Dec 2 18:29:06 2008 -0500 updated icc flags -- now prefer -xHost (-xN etc. seem t be obsolete), check for new spelling -ansi-alias, and use -malign-double like we do for gcc [empty commit message] commit 2e94f29d37f25690825b11ed436d726d5077dba6 Author: stevenj Date: Tue Dec 2 18:28:03 2008 -0500 use $ax_cv_c_compiler_vendor rather than $GCC, as the former is more reliable (icc incorrectly self-identifies as gcc on MacOS where we don't use -no-gcc) [empty commit message] commit a5edcdb0c705b3b42f8aec48e41457fd3771bed7 Author: stevenj Date: Tue Dec 2 17:55:36 2008 -0500 don't use -no-gcc for icc on MacOS [empty commit message] commit 2ce94a7fa1858ca3bfdf38f47f7f2d62ceae5262 Author: stevenj Date: Tue Dec 2 17:34:04 2008 -0500 document some more bench options [empty commit message] commit 315a8ae3389d84d6c062a1afc5c2c4eddee4bb77 Author: stevenj Date: Wed Nov 19 16:55:13 2008 -0500 make it clearer that --enable-openmp and --enable-threads are mutually exclusive; thanks to Long To for his comments [empty commit message] commit 4bca25954f1e56299cf45e61bec3877007f8cbc4 Author: stevenj Date: Mon Nov 17 20:16:28 2008 -0500 version bump to 3.2.1, use explicit Makefile.am for m4 subdirectory so that tarball does not include random files in there when you do 'make dist' [empty commit message] commit 7728d69ca6e4f30747b182a3f0d30ec7c25bf26d Author: stevenj Date: Sat Nov 15 21:12:58 2008 -0500 document behavior of FFTW guru arrays, and in particular the odd behavior of the plan_guru_r2r routine in Fortran (thanks to Alexander Pozdneev for the bug report) [empty commit message] commit ba5c08b8b8a3c0e69ba228e071d71664f72c76ba Author: stevenj Date: Mon Nov 10 20:21:32 2008 -0500 version bump to 3.2, updated copyright year [empty commit message] commit a25226c3cd7b9451e6690a813cc3266b15acd7c2 Author: athena Date: Wed Nov 5 16:40:31 2008 -0500 Store GPLv2 in darcs because automake installs GPLv3 these days. [empty commit message] commit f80a90668573e199b3509a7737ba2d071f4e3974 Author: athena Date: Thu Oct 30 15:03:41 2008 -0400 stylistic changes, comments [empty commit message] commit b8f22edf8aadff2aea6d4e69b7651506951810d0 Author: athena Date: Thu Oct 30 14:40:14 2008 -0400 oops [empty commit message] commit bfaec2f123eb8d8297ce405590e07d682cf80293 Author: athena Date: Thu Oct 30 14:30:08 2008 -0400 simplification of the threading machinery [empty commit message] commit c471651b5bc46868c1e8231a89ec6d459c868854 Author: athena Date: Thu Oct 30 14:22:40 2008 -0400 typo [empty commit message] commit e1f64989da3c427c36c9df3495ef9d24ab01993d Author: athena Date: Thu Oct 30 13:42:07 2008 -0400 [SECOND ATTEMPT] do not assume that a semaphore can be freed just because nobody is using it Let S be a semaphore, initially 0. Let thread A execute UP(S); let thread B execute DOWN(S); free(&S); It is unclear whether this code is correct with posix semaphores. The problem is whether UP() uses S after allowing DOWN() to continue; this seems to be the case in the glibc-2.7 implementation, and thus the pattern above seems to be incorrect. Avoid using such a pattern, and introduce a global semaphore for the unavoidable case when nothing else can be depended upon. commit cddb0450696d51a99ca1d9663d4e4a606c45725e Author: stevenj Date: Wed Oct 29 20:09:39 2008 -0400 updated cpu codes from x86-1.21 [empty commit message] commit 2b8ab85fb21f1fd637cfe7a04ad0acf1610b5713 Author: athena Date: Wed Oct 29 16:24:16 2008 -0400 Previous change was bogus, need to find another way. [empty commit message] commit 0123295bb8dd2064d858a59a48242611219b020d Author: athena Date: Wed Oct 29 12:22:20 2008 -0400 do not assume that a semaphore can be freed just because nobody is using it Let S be a semaphore, initially 0. Let thread A execute UP(S); let thread B execute DOWN(S); free(&S); It is unclear whether this code is correct with posix semaphores. The problem is whether UP() uses S after allowing DOWN() to continue; this seems to be the case in the glibc-2.7 implementation, and thus the pattern above seems to be incorrect. Avoid using such a pattern, and introduce a global semaphore for the unavoidable case when nothing else can be depended upon. commit 1b1dd4a34bb653d4bd63775d2760b435cb4f0d1d Author: stevenj Date: Mon Oct 27 23:38:02 2008 -0400 don't need PROG_AS any more [empty commit message] commit f7f67160af9b208e74304378ef54b4b5608fb96a Author: stevenj Date: Sun Oct 26 23:41:11 2008 -0400 use AC_CONFIG_MACRO_DIR macro [empty commit message] commit d6334fa841da6e5f06a7323ce1d31346fb79b4f4 Author: athena Date: Sun Oct 26 10:08:44 2008 -0400 Remove mpi/Makefile from configure.ac Otherwise, the tarball breaks because mpi/ is not in the distribution. commit dda1fd64e3ab5fbb80578cfe41c69191cdfcc7f5 Author: stevenj Date: Sat Oct 25 17:13:50 2008 -0400 remove MPI from dist until FFTW 3.3 [empty commit message] commit a3a16288f18899e7fc8351da1c500024709174cd Author: stevenj Date: Sat Oct 25 17:12:35 2008 -0400 use MPIRUN even for -np 1 [empty commit message] commit 1b2a86c0ab92772ce43bf6b0c5e0bbae2dfb7fee Author: stevenj Date: Fri Jul 18 17:17:08 2008 -0400 use new gcc arch=native flag as fallback [empty commit message] commit 73944e9472d933cdafcff9c3e5b85efc9661ea5b Author: athena Date: Sat Oct 25 13:36:40 2008 -0400 Use sem_t to implement mutexes Use sem_t instead of pthread_mutex_t to implement mutexes. It seems like pthread mutexes hang on linux-2.6.22 after several days of tests; the hang does not occur on linux >= 2.6.24 or when we use sem_t instead of pthread_mutex_t. The situation is still quite mysterious but this code seems to work. commit 82a0159088bdacf0cb5d5ecb5547f51ae34f77ab Author: athena Date: Thu Oct 23 13:32:45 2008 -0400 print informative message when pstring is NULL. [empty commit message] commit ddc681c6a5de3aee386cf990428dcfba95d32369 Author: athena Date: Sun Oct 19 16:00:07 2008 -0400 Fix incorrect alignment in dftw-generic. Multithreaded dftw-generic is supposed to process only a slice of the array, but we were planning with the alignment of the original array rather than the slice. This led to unaligned accesses in certain obscure situations. commit d35ad1d719daf7a8b1c7658b88bf962e86dd050a Author: Matteo Frigo Date: Mon Aug 18 17:27:26 2008 -0400 Paranoia: do not create OS threads while holding locks. Glibc at least plays silly games such as keeping a global variable that records whether there is more than one thread in the process, and it does not perform atomic operations if the variable says that there is only one thread. Who knows how this interacts with creating threads while holding a lock. Some day some genius will come up with some ``optimization'' that breaks everything. commit 9ae439e37bf0cb024de699e15f98b5f5074d116c Author: athena Date: Wed Aug 6 07:41:46 2008 -0400 Welcome to the quadcore era [empty commit message] commit be6af68ab86ada70645a79ef9ac5da11ea787eba Author: stevenj Date: Mon Jun 16 16:46:39 2008 -0400 backslash is technically not allowed in "echo" arguments; thanks to Debian Bug#486046 for pointing out problem and solution (and Raphael Geissert and Vincent Zweije, in particular) [empty commit message] commit fd1ef499dca6f079bd9980cbecd1499e08fe99d0 Author: stevenj Date: Sun May 4 12:15:24 2008 -0400 note problem with test program in gcc 4.1.2-4.2; thanks to Raymond Rogers for reporting it [empty commit message] commit 771f298272494232c994bdca79978f00cbd0a0ac Author: stevenj Date: Fri May 2 19:21:30 2008 -0400 output count of constants along with other statistics [empty commit message] commit 8a8a1bb47539bb8be624af291c28c77cc541ba4e Author: athena Date: Sat Apr 19 14:15:03 2008 -0400 Lower priority of unaligned SIMD codelets. List t1[fb]uv_* codelets before the corresponding aligned codelets, since the estimator picks the latter ones in case of a tie and aligned codelets are preferable. In other words, this is a hack. commit d4d0ed3f7cee7e5c06409b3162fbcf4bfd42fea9 Author: Matteo Frigo Date: Sat Apr 19 08:55:46 2008 -0400 There is no point in using higher radices for unaligned codelets. [empty commit message] commit c09bbbbc4d2c7236b861b7b2f67b77d4821bb8fc Author: stevenj Date: Fri Apr 18 19:01:27 2008 -0400 support generating loopless, strideless r2r codelets [empty commit message] commit 5d2811a46f654c7dbbade5d2d65921e056c6b3a2 Author: stevenj Date: Fri Apr 18 19:00:25 2008 -0400 added Magic.threemult to use 3+3 complex-multiply variant when possible [empty commit message] commit b21cf57a0c4d5711ea4cdb085b068b366f93c916 Author: stevenj Date: Thu Apr 10 19:53:31 2008 -0400 fix documentation bug - export_wisdom_to_string returns a string that should be deallocated with free, not fftw_free (thanks to Stein Vidar Hagfors Haugan for the bug report) [empty commit message] commit 2a8ac0ba37e9087af75bd0edc9563279424c909d Author: stevenj Date: Mon Jan 21 01:11:44 2008 -0500 bsd calls x86_64 "amd64"; thanks to Fernando Herrero Carron for the bug report [empty commit message] commit 9d150bc32eea5f0404562a6e26c8f5af7571174c Author: stevenj Date: Tue Jan 1 12:29:56 2008 -0500 fix typo in manual, thanks to Yinon Ehrlich [empty commit message] commit a46838157abb2d1d08cf36d882b6e6979c90a059 Author: stevenj Date: Mon Dec 3 13:57:13 2007 -0500 note problem with gcc 3.4.4 on x86_64, thanks to Uwe Hollerbach for the report [empty commit message] commit 14def93d5b1ba54d64d86d0b9635dfbb41795197 Author: stevenj Date: Tue Nov 13 16:19:22 2007 -0500 bump shared-lib revision [empty commit message] commit ac56042c777020dd5edd04a142c522d0ea3d55d9 Author: stevenj Date: Tue Nov 13 16:16:49 2007 -0500 update NEWS for alpha3 [empty commit message] commit d90bca8d53eaa2ecde6c54123b290ea2cccfeda1 Author: stevenj Date: Tue Oct 2 13:53:04 2007 -0400 fixed URL [empty commit message] commit 2e992067b2172b9ef10c068d5b3b1f5f0e336790 Author: stevenj Date: Mon Sep 17 19:38:29 2007 -0400 added missing prototype [empty commit message] commit a9a3696f1daa50cfcea8e9264912fd7010f59edf Author: stevenj Date: Tue Aug 14 22:35:06 2007 -0400 terminology tweak [empty commit message] commit 7e0c1a370fdec22df30379eb952943a8b9f16080 Author: stevenj Date: Wed Aug 1 18:44:21 2007 -0400 check for pathscale compilers (thanks to Julian Cummings) [empty commit message] commit dabff4a5303cc1bde1a1ac88508f3301caeb6e9f Author: athena Date: Sat Sep 15 18:02:32 2007 -0400 Avoid possible conflict with Windows include files. [empty commit message] commit 1090ecb91cd0da452cca31e8ef926494895a83bc Author: athena Date: Tue Aug 7 21:26:05 2007 -0400 Distribute codlist.c for SIMD codelets in the commercial tarball. [empty commit message] commit 9c132ba2c88c1d5f9e35ad89c8d9e7b012f3741f Author: stevenj Date: Wed Aug 1 10:33:41 2007 -0400 some documentation clarifications, and documented FFTW_WISDOM_ONLY, at the suggestion of Mario Emmenlauer and Phil Dumont [empty commit message] commit 79a73f23fc62044a1edd421d031c35d9d17345dc Author: stevenj Date: Tue Jul 31 16:52:56 2007 -0400 bug fix in test program for vrank-3 transpose plans with vl=1 [empty commit message] commit 13dcde33151a281f5dd4084f3a65277223d444eb Author: stevenj Date: Sun Jul 29 17:02:46 2007 -0400 only run mpi checks for --enable-mpi [empty commit message] commit da920b9d9649c89291980b342a38cd31e689d04c Author: stevenj Date: Sun Jul 29 16:45:30 2007 -0400 check for NULL return from spe_context_create in case SPE_MAP_PS not supported [empty commit message] commit 8b5208ab2d26b33f10864d23ae032a575877cdb3 Author: stevenj Date: Sun Jul 29 15:56:57 2007 -0400 use problem-state pointer to write SPE mailbox with lower latency (makes a significant performance difference for N < 32k), thanks to Jan Wagner for suggestion [empty commit message] commit 35435685af71440fc9601b845163491e61845b4b Author: stevenj Date: Sun Jul 29 14:22:08 2007 -0400 port cell code to SDK2.1 (libspe2), since libspe1 API is deprecated and can't be used in code that also uses libspe2 API [empty commit message] commit 21dc1b9f90f96120a92469077cabfd80dd7fcb70 Author: stevenj Date: Sun Jul 29 11:46:24 2007 -0400 bug fix: ego->W allocated with cell_aligned_malloc, so deallocate with free, not X(ifree0) [empty commit message] commit 7c9f576ed6672631a9d36698f5d9824d173e06ea Author: stevenj Date: Mon Jul 2 15:57:12 2007 -0400 removed obsolete reference to CVS id [empty commit message] commit 2efeadcf8d2b6562d8c18707a0b7eb2e3e5f73d3 Author: athena Date: Mon May 21 14:25:39 2007 -0400 cycle counter for sun compiler [empty commit message] commit 0b59ce4b61ae5c052d9c673807e7b3386d7bdaa2 Author: stevenj Date: Wed May 9 19:49:11 2007 -0400 use __inline instead of inline for AIX routines (__inline is supported by gcc and xlc, whereas apparently "inline" is only supported by xlc if you specify -qlanglvl=stdc99 or similar); thanks to Jeff Haferman for the bug report [empty commit message] commit a3dc6f8631568cca3039a6932aa4a66e46456a79 Author: stevenj Date: Mon Apr 30 15:37:56 2007 -0400 fixed incorrect type prefix (fftw_ vs. X(...)) in mpi/wisdom-api.c; thanks to Eric A. Borisch for the bug report [empty commit message] commit 122d2b4a77a11b949e61e503681975f2da3d7d4f Author: stevenj Date: Wed Apr 25 21:21:39 2007 -0400 some cleanups in MPI make check [empty commit message] commit 4d26d141fb5e018b49133b1e080acbff744d97c0 Author: stevenj Date: Wed Apr 25 21:19:27 2007 -0400 re-enable heuristic in the common case where we are not compiling for Cell [empty commit message] commit 28a27bde79e55d51dd1fb81d82ca418587106a62 Author: athena Date: Tue Apr 24 17:42:43 2007 -0400 Removed duplicate codelet names, was breaking linker. [empty commit message] commit 1a4a3dcbbe5a046f07654fa3734a5b3568d51b32 Author: stevenj Date: Tue Apr 24 11:38:16 2007 -0400 added more codelets of sizes 5/10/20/25 to improve speed for round decimal sizes (speed improvements of 10-20%, at cost of 10-30% in library size) [empty commit message] commit 35443ed785a0ce0a05e16d1f6419769f7641f415 Author: stevenj Date: Sat Mar 24 18:40:47 2007 -0400 for 1d prime sizes, punt and return serial plan [empty commit message] commit 3cf27d0073f9f911a2b15283b0d1acebeaf7b599 Author: stevenj Date: Sat Mar 24 18:24:55 2007 -0400 output reminders of the problem during bench --verify [empty commit message] commit e68227acb0ebf9ad3cb0b022382c6df9fcf0d8b1 Author: stevenj Date: Sat Mar 24 18:10:24 2007 -0400 bug fix - missing solver->destroy initializer in rdft2-rdft [empty commit message] commit 59a3e77ab1cba10afa53a627849abef6dd93152d Author: stevenj Date: Fri Mar 23 11:12:19 2007 -0400 -static, in --enable-debug, doesn't work on MacOS X (according to Daniel Oberhoff) [empty commit message] commit 1bcacddfa919627af62568a28b0713b368549612 Author: stevenj Date: Wed Mar 21 22:23:06 2007 -0400 fix MPI r2c/c2r to work with howmany > 1 [empty commit message] commit 86e99768530258b6184733a382920feae222ae55 Author: stevenj Date: Wed Mar 21 18:44:41 2007 -0400 rm MPI version from TODO [empty commit message] commit b55ed34cb35d64e452aac41b5661536d75c492d4 Author: stevenj Date: Wed Mar 21 18:34:40 2007 -0400 added 'make bigcheck' for MPI (no paranoid-check, unfortunately), and properly get MPIRUN from configure [empty commit message] commit e11b28e739bf5b888cfdf0ec97337166fbb6c425 Author: stevenj Date: Wed Mar 21 18:23:18 2007 -0400 bug fix - incorrect local_size returned for 1d bigvec case [empty commit message] commit 3c4171a56630a623798d71e1a6218c1400ea3e46 Author: stevenj Date: Wed Mar 21 03:13:54 2007 -0400 hack to specify MPI_TRANSPOSED_IN/OUT via "[" and "]" in libbench2 problem [empty commit message] commit b6643c4d6de6ac41e771a65accc67af6d515009f Author: stevenj Date: Wed Mar 21 02:58:11 2007 -0400 added MPI 'make check', still needs a bit of work [empty commit message] commit 5c4ca6bf40ab9683f717ef89a5bbb2c7da031680 Author: stevenj Date: Wed Mar 21 02:47:10 2007 -0400 bug fix in r2r transposed-input case [empty commit message] commit 518bfe4ddbe9a727866374bb0b0fb49a2d0a9f2b Author: stevenj Date: Wed Mar 21 02:46:25 2007 -0400 don't output more than 300 erroneous outputs (unless verbose > 2) [empty commit message] commit 5ae1f03689a0f37d5db6becf54c8e54395541407 Author: stevenj Date: Wed Mar 21 01:48:54 2007 -0400 fixed bug in transposed-in c2r MPI transforms ... seems to be working, finally [empty commit message] commit 6ff00891c3f7fcfe5399e652b9aeb1538bf9c8d6 Author: stevenj Date: Wed Mar 21 00:41:32 2007 -0400 some fixes to MPI r2c/c2r transforms with transposed output/input [empty commit message] commit fc68b9bdd4975fec0d3f9b9ef6a2d9e052f8e7ce Author: stevenj Date: Wed Mar 21 00:40:25 2007 -0400 typos [empty commit message] commit 75dce53511a209c32f881a782af96bf68fdf41af Author: stevenj Date: Tue Mar 20 19:53:02 2007 -0400 bug fix for mpi-bench with r2c/c2r: allocate a little bit extra to make sure that padding is allocated [empty commit message] commit 1720fcb4dc2220e66e50a2baa9201f6c58913bf4 Author: stevenj Date: Tue Mar 20 19:19:13 2007 -0400 fix typo, thanks to Ernest Turro for the bug report [empty commit message] commit 3a9a95e347c10d98ad7d7ef0c3dca97217e2137a Author: stevenj Date: Tue Mar 20 01:39:06 2007 -0400 spacing tweaks [empty commit message] commit 27cc0f277fb7839e10996eb97c07e4ea9e5bb94c Author: stevenj Date: Tue Mar 20 00:53:11 2007 -0400 Ralf Wildenhues is the one who pointed out that the self-communication could fill in the stalls in the pairwise schedule [empty commit message] commit 3879a995d7146eb6be6ab3df826ebdf2660451f5 Author: stevenj Date: Tue Mar 20 00:22:25 2007 -0400 add TRANSPOSED_OUT/IN support for r2c/c2r, respectively [empty commit message] commit 5a4f8df7a48f55926d1a2017e234903e75fbf35b Author: stevenj Date: Mon Mar 19 21:45:34 2007 -0400 yikes! fixed likely deadlock bug in MPI [empty commit message] commit e5514b08ccfb3c99dfa7034276872af1e3a15b84 Author: stevenj Date: Mon Mar 19 21:38:52 2007 -0400 comment [empty commit message] commit 2ccef4a6915eeebe969120c75c5790791905fd37 Author: stevenj Date: Mon Mar 19 21:30:44 2007 -0400 s/alpha1/alpha2/ [empty commit message] commit 4c069ca435d517243da0ea52594b8101723303b1 Author: stevenj Date: Mon Mar 19 00:39:47 2007 -0400 include README in dist tarball [empty commit message] commit ea9cd7ed69c82e7b129bf88b99dc58238d856c0c Author: stevenj Date: Mon Mar 19 00:35:43 2007 -0400 added MPI r2c/c2r transforms, some more documentation [empty commit message] commit 539fd2ff41bbf5d9955ff83880d983d422e4f4bd Author: stevenj Date: Sun Mar 18 23:14:29 2007 -0400 set version to 3.2alpha2 [empty commit message] commit 5b7625b7daebd8aba2e0a876083a944e2c7b520e Author: stevenj Date: Sun Mar 18 19:12:18 2007 -0400 changed --enable-mips_ps to --enable-mips-ps; added Cell section to manual (from README.Cell); many minor updates to manual [empty commit message] commit c8cd95869bb81b8d85d87a91e0d65402f9de2288 Author: stevenj Date: Sun Mar 18 15:27:06 2007 -0400 whoops, need to sync costs in problem_mpi_rdft [empty commit message] commit dfc055b714f7d4b63b6615bb4b00c86d1600b7de Author: stevenj Date: Sun Mar 18 12:44:49 2007 -0400 documented guru64 interface [empty commit message] commit e6a8b5ed239bd9e150f62c8f773e0dcdc97df31a Author: stevenj Date: Sun Mar 18 02:57:46 2007 -0400 typo [empty commit message] commit 4625ba2558f8f51201b06cc14102507dd3f2731d Author: stevenj Date: Sun Mar 18 02:45:09 2007 -0400 bumped copyright year to 2007 [empty commit message] commit 66392e6b7c997772c49e9c38d275fe79cc25ed33 Author: stevenj Date: Sun Mar 18 01:41:40 2007 -0400 noted CodeSourcery in AUTHORS [empty commit message] commit 2c18cc0507bb6ea17abd5d54bacf15bd7ccbca13 Author: stevenj Date: Sun Mar 18 01:25:00 2007 -0400 more MPI documentation [empty commit message] commit 498f7ef52ac60aa1467d76bbfbd0d2224b9ccb10 Author: stevenj Date: Sat Mar 17 23:15:04 2007 -0400 added MPI multi-dimensional rdft solvers & tests [empty commit message] commit 871ff1554eded2f68e184ecff1d3befd19aa2679 Author: stevenj Date: Sat Mar 17 22:52:00 2007 -0400 whoops [empty commit message] commit 5a46acce32021f8bcdf12188ca3d764ce7f1cc85 Author: stevenj Date: Sat Mar 17 22:43:54 2007 -0400 clarification - fftw_mpi_init should be called before importing wisdom [empty commit message] commit 3e2d1704698a4609579e332b904502f5b30370fa Author: stevenj Date: Sat Mar 17 19:49:37 2007 -0400 kindx/y/z -> kind0/1/2 for consistency [empty commit message] commit 73c018e5b29c759aaf1012ee39853b025024334c Author: stevenj Date: Sat Mar 17 19:34:02 2007 -0400 typo [empty commit message] commit 1d9eeb0231c02f554470a9b6150b07df35e85a4a Author: stevenj Date: Sat Mar 17 19:14:16 2007 -0400 some refactoring in preparation for mpi-rdft [empty commit message] commit 3a5f38381e0f65c6da82ab93eefe1be2789c9749 Author: stevenj Date: Sat Mar 17 18:12:45 2007 -0400 documented more stuff for MPI [empty commit message] commit f833ef7087c898b684d1e0945fb28164e7d5fc02 Author: stevenj Date: Sat Mar 17 15:41:23 2007 -0400 added NEWS for 3.2alpha [empty commit message] commit 75d1f8189e12f1104a11d92da913592e69b37227 Author: stevenj Date: Sat Mar 17 14:50:22 2007 -0400 documented MPI transpose routines [empty commit message] commit 4fdc9e45f4c6f587f12edb5c0bbe0c60a499d0a9 Author: athena Date: Sat Mar 17 08:57:30 2007 -0400 Removed unused variables [empty commit message] commit a1bd09375e2342d3dec8dbaba75321c278b50861 Author: athena Date: Fri Mar 16 14:47:10 2007 -0400 Preparing for interim release of Cell code. [empty commit message] commit bd1f6de1d0c69ee9a7b7be715797ae2e5cb28ed0 Author: athena Date: Thu Feb 8 12:23:43 2007 -0500 Added README.Cell [empty commit message] commit 4125ae1b42049d9828b51cb9f45398601fa25e5c Author: athena Date: Sat Mar 10 19:17:40 2007 -0500 Synchronized with main branch [empty commit message] commit fa8f1748c92e9255b456b995a9c2d439110fb1b5 Author: athena Date: Mon Jan 22 17:43:56 2007 -0500 Adapted vrecur heuristic to Cell. [empty commit message] commit 31c0788d89e5d2db56d1949df2e61171360ad282 Author: athena Date: Thu Jan 18 20:29:22 2007 -0500 Increased MAX_N to 32K/sizeof(R). [empty commit message] commit e0e08abd04fec6f16a5aa4b8dbec2f614b845bde Author: Matteo Frigo Date: Thu Jan 18 13:43:51 2007 -0500 Added pointer to solver->destroy which is used in the Cell branch. [empty commit message] commit dbb33d0b51897749feff0ef26e63af7769cd4fa5 Author: athena Date: Thu Jan 18 12:09:26 2007 -0500 Updated copyright notices [empty commit message] commit bdb23b08725ce86827dc72f39ace915e594e0ddd Author: athena Date: Fri Jan 12 12:54:43 2007 -0500 Use mfc_read_tag_status_all() instead of spu_mfcstat(2), since the former seems to be standardized. [empty commit message] commit d958d4081d9d131a8c331795b51c38392e038f5f Author: athena Date: Thu Jan 11 14:55:08 2007 -0500 Silence some int/INT warnings. [empty commit message] commit a465f3a820c88855cec17a8b62093f151cf4a75b Author: athena Date: Wed Jan 10 18:19:53 2007 -0500 Note incompatibility of --enable-cell with --enable-threads [empty commit message] commit f8d67adca8de472032a5bb176caef2df069ac301 Author: athena Date: Wed Jan 10 17:57:10 2007 -0500 forgot to add file [empty commit message] commit 2b494f41afa950fa4174d4588070c49879c89acb Author: athena Date: Wed Jan 10 17:45:16 2007 -0500 64-bit cleanup [empty commit message] commit d4d591f6547fba8ab96d982a76747e0248e94031 Author: athena Date: Wed Jan 10 13:47:20 2007 -0500 Use -mcpu=cell where appropriate. [empty commit message] commit bbff6b92337e5462c8e01c3c6f200f0841422229 Author: athena Date: Tue Dec 26 21:35:59 2006 -0500 synchronized with main [empty commit message] commit 011961c240152fa9ff6c791c430f0104ca2dada7 Author: athena Date: Sun Dec 24 20:58:25 2006 -0500 synchronized with main branch [empty commit message] commit 563b10fee948f8cf7089fc91f97ccd07b92169ae Author: athena Date: Sun Dec 24 13:47:37 2006 -0500 synchronized with main branch, updated to new sdk. [empty commit message] commit 8a7076d3c731fa116d61cc072416ab46beddfc03 Author: athena Date: Thu Dec 21 17:17:41 2006 -0500 removed obsolete file [empty commit message] commit b4139c9ef725c1910b97628667ab1fa9ab98e635 Author: athena Date: Tue Dec 19 15:17:20 2006 -0500 synchronized with main branch [empty commit message] commit 7df6b5623e1db3fe7e9dff26fd4d23f484495649 Author: athena Date: Tue Dec 19 11:27:38 2006 -0500 Synchronized with main branch [empty commit message] commit 6ea2fd75cd1aae82fe1d1a510f104646ab7aa7c9 Author: athena Date: Fri Dec 15 16:04:31 2006 -0500 resolved conflict with main branch [empty commit message] commit 65515a62ffff8fb5649200627cb0c93ef813a9c0 Author: athena Date: Fri Dec 8 14:43:50 2006 -0500 Fixes for compilation in subdirectories [empty commit message] commit d39013569f9cb742b50238ca622f3bd52cade5b5 Author: athena Date: Fri Dec 8 12:46:00 2006 -0500 Silence warning [empty commit message] commit 565f828f3b33be86ac3263046fae490fda62d059 Author: athena Date: Fri Dec 8 12:24:19 2006 -0500 silence warning [empty commit message] commit 5ae3fbed814693f0172abe16b6875b17df64616d Author: athena Date: Thu Dec 7 15:18:17 2006 -0500 Commented a particularly obscure piece of code. [empty commit message] commit b4f78f41fb5960b0ec300d23d6653e80f498eee2 Author: athena Date: Thu Dec 7 11:53:29 2006 -0500 Reorganized, clarified conditions for applicability of the DFT solver. [empty commit message] commit 4d31f1609b233b3951f5f660bc7d2f8f98d7bd38 Author: athena Date: Mon Dec 4 21:33:49 2006 -0500 Minor changes [empty commit message] commit b4cd386c5a2ebd85f8c53be407642199914c26ed Author: athena Date: Mon Dec 4 17:43:28 2006 -0500 Clarified comment [empty commit message] commit aafb3252fb01b21a852ed938b9bc6b2e8b852517 Author: athena Date: Mon Dec 4 16:49:06 2006 -0500 Less incorrect conditions for fitting into local store. [empty commit message] commit 985d9f04682d20ed877a04bfc1d5c0cb73af903c Author: athena Date: Mon Dec 4 16:08:24 2006 -0500 Implemented DECDIF+TRANSPOSE on Cell [empty commit message] commit 00969378eff1f94e8858105bd22015e622da58d0 Author: athena Date: Fri Dec 1 17:42:55 2006 -0500 relaxed conditions of applicability of SPE [empty commit message] commit 4df1c8eb6351f9f6dd2e869a33044d8b36f8dd54 Author: athena Date: Fri Dec 1 16:28:10 2006 -0500 tweaks [empty commit message] commit 92a232ae0edf2fe3f92ca5485861d866aa4c96b0 Author: athena Date: Fri Dec 1 14:35:17 2006 -0500 Implemented Cell opcounts [empty commit message] commit b2e38a76dac60881d4e14e3d9a6ad2b236086a68 Author: athena Date: Fri Dec 1 13:38:44 2006 -0500 minor cleanup [empty commit message] commit 0ab21aa8b736fd4002a92db39449d9e140c39606 Author: athena Date: Fri Dec 1 11:16:52 2006 -0500 use [c0 s0 c1 s1] format for Cell twiddle factors, rather than [c0 c1 s0 s1]. This makes life easier and there is no speed penalty on Cell (unlike Altivec). [empty commit message] commit d080990161fe731d5e1af92f9534c4bd86e06d1f Author: athena Date: Wed Nov 29 18:02:54 2006 -0500 Implemented SPE-accelerated copies [empty commit message] commit 23f9c35ec5d4c123a664c907e5aaca2d9704888c Author: athena Date: Wed Nov 29 12:11:08 2006 -0500 allow SPEs to compute vrank-0 problems. [empty commit message] commit 2f71518adcd364d1cc45272d4f254028c779ff83 Author: athena Date: Tue Nov 28 18:03:07 2006 -0500 eliminated DMA lists [empty commit message] commit 8620a7ab00d1bd4d288513998c82d14cce47b98f Author: athena Date: Tue Nov 28 14:22:05 2006 -0500 Conservatively force all dimensions to be 0 (mod VL) in cell, since otherwise it is too hard to get all cases right. [empty commit message] commit 8343bf57ddaff7d35756635dd9a4a6aa8d31e964 Author: athena Date: Tue Nov 28 12:39:01 2006 -0500 Check alignment of strides when transposing on Cell. [empty commit message] commit 7e28410dc1fbaa1bbf53007869937b84d44c37f3 Author: athena Date: Tue Nov 28 12:19:09 2006 -0500 consistent usage of FFT_SIGN [empty commit message] commit 66491d40310d02c982e54ebee2f9f6c7cb0ae7db Author: athena Date: Tue Nov 28 11:35:38 2006 -0500 clever transposition algorithm without buffering [empty commit message] commit 33e68acd59c892e969a722ded4292aef57fab0a9 Author: athena Date: Mon Nov 27 14:08:28 2006 -0500 Fixed tracking of dependencies [empty commit message] commit 1eab08a3242ae935de04debda8abee056fb3d978 Author: athena Date: Mon Nov 27 14:03:53 2006 -0500 implemented 1D transforms, various tweaks [empty commit message] commit 14891530341581ba7a2422754d83d0c621c71daa Author: athena Date: Wed Nov 22 15:43:36 2006 -0500 no need to poll mailbox on spu side [empty commit message] commit bc0822718ba87089828bd9362b112c8a9bef878c Author: athena Date: Wed Nov 22 14:08:24 2006 -0500 increased maximum size handled by spe [empty commit message] commit 0fb0144374b505502ed768b7f13c191d775bf870 Author: athena Date: Tue Nov 21 16:23:17 2006 -0500 allow vrank<=2 problems in SPEs to avoid the vecloop overhead (grrr...) [empty commit message] commit 26017102cbb70e6e0292087249267b6560414f1c Author: athena Date: Mon Nov 20 14:41:45 2006 -0500 added emacs mode [empty commit message] commit ae2ed8341f9860cd659dc62def4f0c3712e856ab Author: athena Date: Mon Nov 20 09:34:12 2006 -0500 revised transpose, cleanup [empty commit message] commit 7de0b7799021747ff1eef31aca59f7b229750e93 Author: athena Date: Sun Nov 19 20:20:23 2006 -0500 added file [empty commit message] commit 7383afd328c1e98b5ec25c32094a28b0312fb7b0 Author: athena Date: Sun Nov 19 20:18:35 2006 -0500 removed file [empty commit message] commit fa6b1a88cae92cf2e4e5ab247d7a9d93e722c405 Author: athena Date: Sun Nov 19 20:15:38 2006 -0500 better automake integration [empty commit message] commit d1af1e31717e8df8126e7f44197d10e652adbaab Author: athena Date: Sat Nov 18 20:14:29 2006 -0500 changed algorithm for computing chunk size [empty commit message] commit 631ad019b179b1e260a5197ef0012e38e5adfb9c Author: athena Date: Sat Nov 18 19:18:11 2006 -0500 implemented transpose, various fixes. [empty commit message] commit 4d0d4332d36eb952e188eb44f039249a78dc2545 Author: athena Date: Thu Nov 16 16:33:50 2006 -0500 Added explicit destructor to all solvers to help with the cell port. [empty commit message] commit c668de4bc5e8677b2c78830b34214aa832631281 Author: athena Date: Thu Nov 16 15:22:15 2006 -0500 consistent use of #if vs. #ifdef [empty commit message] commit 34192d4eaa06426a1168b5ac743332adb1cd6039 Author: athena Date: Thu Nov 16 15:15:34 2006 -0500 Additional Cell double codelets, better automake integration [empty commit message] commit 4b19e0b192a7f797d21adc5b8b39b126bf809e53 Author: athena Date: Thu Nov 16 12:43:34 2006 -0500 Use dma lists. [empty commit message] commit 398bcee9091aa5c56d753877957f367e7041e6a7 Author: athena Date: Thu Nov 16 11:03:46 2006 -0500 converted to automake [empty commit message] commit caf4303b1448b64d7f82cf9ba36eee8071674421 Author: athena Date: Wed Nov 15 18:00:12 2006 -0500 Initial port to Cell Broadband Engine. [empty commit message] commit 45eebf6ef925ca51e2749ea8658cfb39216fd5b7 Author: athena Date: Wed Mar 14 10:19:53 2007 -0400 Remove Codesourcery contributions from commercial tarball. [empty commit message] commit 71e740a2b810c009c637addb3f87bba3338fa0d2 Author: athena Date: Wed Mar 14 08:59:18 2007 -0400 Added FFTW_WISDOM_ONLY, at the request of Phil Dumont. [empty commit message] commit 8c4485fd3fffb1cfd1aacddfecb58250b5b69607 Author: stevenj Date: Tue Mar 13 00:32:05 2007 -0400 fixed potential MPI deadlock if timer misbehaves [empty commit message] commit 193dbead568fc6582fce99e2b1824f7aac2c66b1 Author: stevenj Date: Mon Mar 12 23:31:52 2007 -0400 more work on MPI documentation [empty commit message] commit 4374a330a301a85267faf67eb71833daeeeefa72 Author: stevenj Date: Tue Feb 27 13:48:43 2007 -0500 index [empty commit message] commit 8dd26fb6008dec917db3ff3f34bbd437b21ba12f Author: stevenj Date: Tue Feb 27 13:46:45 2007 -0500 rename "new-data execute" to "new-array execute", since of course you do not need a new array to have new data [empty commit message] commit 89ebde1693423d225eb9a50b56dc0a5703d30384 Author: stevenj Date: Tue Feb 27 13:43:55 2007 -0500 consistency with manual (guru execute -> new-data execute) [empty commit message] commit 8cce0d9c67bf193b5cf177483ad0500e71a18a84 Author: stevenj Date: Tue Feb 27 13:42:24 2007 -0500 texinfo fixes; renamed "guru execute" section to "new-data execute", since previously it seemed to lead to endless confusion with the guru planner API [empty commit message] commit 7188f6250c91692fb25976542298900e557d092a Author: stevenj Date: Mon Feb 26 18:57:11 2007 -0500 consistently use n0/n1/.. everywhere instead of nx/ny/... (for consistency with d-dimensional case n[0], n[1], ...) ... first start at MPI documentation [empty commit message] commit 385b92bc1fa159e0423f02059cd15c93b7444c92 Author: athena Date: Sat Mar 10 18:48:05 2007 -0500 Changed C++-style comment into K&R [empty commit message] commit 32f8fc24e66030c4e5fdc42b9ec503c50d163435 Author: athena Date: Sat Mar 10 18:47:12 2007 -0500 Forgot to add file [empty commit message] commit acf05fd6f7275e013b16abcfafbc2db7437145f1 Author: athena Date: Sat Mar 10 18:44:39 2007 -0500 Note removal of K7 support. [empty commit message] commit e768b9aeeddd3937eacf72bf4bcd1fe6b67681b8 Author: athena Date: Sat Mar 10 18:41:52 2007 -0500 Updated manual for MIPS PS [empty commit message] commit 56c7d29b2740a24da19b5c022569e60a9bf1abaa Author: athena Date: Sat Mar 10 18:37:07 2007 -0500 Adopted MIPS_PS patches from Codesourcery. [empty commit message] commit 117c18b54c2398c74c00d3f375e60e490cfd0a55 Author: athena Date: Sun Feb 25 11:34:51 2007 -0500 Incorrect initialization of win32 semaphores [empty commit message] commit 835fb99c05fc32b63c000aaa65fa8f098d66d1a1 Author: stevenj Date: Tue Jan 30 11:43:09 2007 -0500 win32 fixes (I think, still untested) [empty commit message] commit 22544bfee3ccdd6810c7f7b9552eb8ec67f58562 Author: stevenj Date: Fri Jan 19 17:31:47 2007 -0500 message-size heuristic in tranpose-recurse [empty commit message] commit c3f9a60853f7d021b8e2e84aca81167fe0742499 Author: athena Date: Tue Jan 30 08:53:55 2007 -0500 Threading layer for Win32, completely untested. [empty commit message] commit e9103c2fa36af2816f233d39aa4b4e6aad4bccd2 Author: athena Date: Mon Jan 29 14:26:30 2007 -0500 Check for EINTR after sem_wait(), as suggested by Chip Salzenberg. [empty commit message] commit 719f223ad7ec385208d4d10171374f44d6dcbfa9 Author: athena Date: Mon Jan 22 13:58:23 2007 -0500 Force vector recursion by means of a separate function pointer. I need this for Cell. [empty commit message] commit 44d62282fe44de7c794ce22ec8a5a3120e71d57f Author: athena Date: Mon Jan 22 09:28:35 2007 -0500 Merge multiplications by twiddle with multiplications by i for faster r2c transforms. [empty commit message] commit 37defea1b213e2cb3e5f73fc481e34551ca72e59 Author: athena Date: Sun Jan 21 19:02:44 2007 -0500 Disabled vector recursion, too messy. [empty commit message] commit f6f7ab5fd044a6ed0b9803c8ea10e176c37137dd Author: athena Date: Sun Jan 21 14:23:35 2007 -0500 Changed heuristics for vector recursion. As in fftw-3.1, NO_VRECURSE disables vector recursion. As an exception, however, vector recursion is allowed when the predicate VRECURSE_ANYWAYP is true. We need some form of vector recursion to obtain decent plans on Cell, and this solution captures the common cases without increasing planning time too much. commit 3612cb7be2fce875627ec720c48a70f9204b42f6 Author: athena Date: Sun Dec 17 22:31:17 2006 -0500 fixed hc2c for vector-recursion branch [empty commit message] commit af9505fcc532b01fb7d7d4e4df0793f1d58bbedd Author: athena Date: Tue Dec 5 12:52:36 2006 -0500 switch to default vector recursion [empty commit message] commit d49ea8d1b8d123219d25c7279a06f0146ff0020b Author: athena Date: Sat Jan 20 23:37:33 2007 -0500 Smarter algorithm for selection of nbuf. [empty commit message] commit 983a3c8a18351c2aa89b096d17419c8ecc8ee4eb Author: athena Date: Sat Jan 20 22:15:33 2007 -0500 Increased buffer sizes according to Moore's law. [empty commit message] commit 3063d37e369e9b607af9a1870c3e9c67966969d3 Author: stevenj Date: Fri Jan 19 16:02:00 2007 -0500 fix another MPI synchronization bug -- several more places where cost_hook must be called to synchronized process timings (sigh) [empty commit message] commit ed26acb97814e71ca8961385f95d136fb532d3e5 Author: athena Date: Fri Jan 19 12:08:07 2007 -0500 Set havewisdom=0 when calling forget_wisdom() in the test program. [empty commit message] commit 9c8fc20de720a1f8588230add1f732504a489797 Author: stevenj Date: Fri Jan 19 10:29:56 2007 -0500 remove redundant check [empty commit message] commit ea709110aaac1eac97acdc9e6d6dccb1a319f491 Author: stevenj Date: Thu Jan 18 22:37:59 2007 -0500 fixed potential (unlikely) bug in wisdom import (triggered when importing impatient wisdom after creating more patient plans, but apparently only for nonstandard configure.c configurations) [empty commit message] commit c30ae9a7d180707e86d8a42ce607c6e7717b49e6 Author: stevenj Date: Thu Jan 18 21:50:14 2007 -0500 added functions to gather/broadcast wisdom for MPI [empty commit message] commit a87ad4116c7bf1ac3e28709b2dc7a3f942beba34 Author: stevenj Date: Thu Jan 11 18:33:17 2007 -0500 whoops, another int/INT bug [empty commit message] commit 51ddf455e30f2f8448b94dc40b8a19a8f296067b Author: stevenj Date: Thu Jan 11 17:42:24 2007 -0500 whoops, fixed bug in transpose-recurse for r != m [empty commit message] commit d2550926efaedd18154f03ae20b464f57ebbc71f Author: stevenj Date: Thu Jan 11 17:25:36 2007 -0500 canonicalize mpi-transposed flags by setting TRANSPOSED_IN/OUT where possible [empty commit message] commit 0129b3159f5ffd78b1d5e8c99a80e5aac5ae1743 Author: stevenj Date: Thu Jan 11 17:16:24 2007 -0500 replace transpose-radix2 with much more general transpose-recurse solver [empty commit message] commit b5399f6884419e5aac9bd45b2f99a55c722dbae6 Author: stevenj Date: Wed Jan 10 20:23:48 2007 -0500 rename transpose-inplace to transpose-pairwise, as the algorithm is not restricted to inplace operation [empty commit message] commit 1db83491ac2308011e874a0e14867ab82285ca87 Author: stevenj Date: Wed Jan 10 14:39:08 2007 -0500 whoops, some int/INT bugs [empty commit message] commit 7c54c7285fedadb55997fa5032a86721a5d73c00 Author: stevenj Date: Tue Jan 9 18:50:07 2007 -0500 fix FAQ Makefile for vpath builds [empty commit message] commit 1f9ce0c767890a637491a26dc6d671cb48d899e1 Author: athena Date: Tue Jan 9 20:22:11 2007 -0500 Missing ``static'' keyword. [empty commit message] commit 5719264a71b3d5a725179d6c6f38fc9844f614c7 Author: athena Date: Tue Jan 9 20:13:18 2007 -0500 Minor cleanup. [empty commit message] commit 31f23769946e8cbd668eae280cf6fa5e0d731cc6 Author: stevenj Date: Tue Jan 9 00:04:03 2007 -0500 interleave twiddle mults with DFTs (should we use dftw?) [empty commit message] commit 5ee274d2bfff3ccfa48faf75d4c5ba4254b6403a Author: stevenj Date: Mon Jan 8 18:35:41 2007 -0500 simplified (and somewhat sped up) dft-rank1 by exploiting dft-rank1-bigvec [empty commit message] commit 6603c476a81bd7d9a84eeec0106ce87ea7af55eb Author: stevenj Date: Sun Jan 7 00:31:31 2007 -0500 rearranged TRANSPOSED format, numerous speedups Split the TRANSPOSED and non-TRANSPOSED rank-geq2 solvers, and changed the DFT TRANSPOSED format to be more like fftw2 (both globally and locally transposed). In general, more emphasis on arranging the data contiguously for the DFTs, and more flexibility in intermediate transposed formats. Also disable NO_SLOW when planning transposes, since otherwise non-square in-place transposes gratuitously put the planner in SLOW mode. Currently, dft-rank1-bigvec has 5 variants (or 10, if DESTROY_INPUT). It looks like only 2 of these are commonly used, so I should probably add some UGLY tags once I do more benchmarking. commit 8efa4e83812fc0d52b20291b0ae6b6d863d873b4 Author: stevenj Date: Thu Jan 4 19:13:17 2007 -0500 add bench_cost_postprocess to prevent deadlocks in mpi-bench [empty commit message] commit ad8fbe7775bfe2a214cefd3759493f11d3330532 Author: stevenj Date: Thu Jan 4 16:46:29 2007 -0500 whoops [empty commit message] commit f1d13c4b532737e65ce9f8cdb058875fed16aac7 Author: stevenj Date: Wed Jan 3 14:23:42 2007 -0500 pass proper pointer types as arguments, so that ACX_PTHREAD still works with C++ and -Werror (thanks to Ewald Arnold for the suggestion) [empty commit message] commit 85662c73ef1053f67e55830adb20a7660c7f546e Author: athena Date: Mon Jan 1 19:30:43 2007 -0500 Renamed [io]vs => [io]vs_by_nbuf, which is more appropriate and would have saved me 30mins debugging. [empty commit message] commit 011d6fa311a3126c66527f22f76a55acababb0f3 Author: stevenj Date: Mon Jan 1 18:52:38 2007 -0500 add --with-g77-wrappers option & always include g77 wrappers on GNU systems and/or with gfortran Upcoming GNU/Linux distros will most likely switch to configuring FFTW with gfortran by default, since g77 isn't even included with recent gcc versions. However, we still want to include g77-compatible wrappers in this case (two underscores) in addition to gfortran wrappers (one underscore) lest we silently break binary compatibility and provoke lots of annoying emails. commit fbb0f99f47d4c09c87cd81573b3532809b44fee1 Author: stevenj Date: Mon Jan 1 16:48:36 2007 -0500 use AC_HELP_STRING for --disable-fortran [empty commit message] commit 4c6880164b7e43be156bd10825038bc5fe83b9b3 Author: stevenj Date: Mon Jan 1 15:56:12 2007 -0500 terminology [empty commit message] commit ae75dc0d2a5fb8286ebadc5fa70a1ff2e17ed7ba Author: athena Date: Sat Dec 30 16:18:35 2006 -0500 Free buffers before calling cldrest. [empty commit message] commit 053f45629c9af2bc608086640e6684ef1e76bf0b Author: athena Date: Fri Dec 29 10:52:15 2006 -0500 Removed obsolete code. [empty commit message] commit e6ffd09841ff145db9112e4fca774cc3454e1170 Author: athena Date: Thu Dec 28 21:37:48 2006 -0500 Attempt to work-around old gcc bugs in a more efficient fashion that does not lose performance on newer gcc's. [empty commit message] commit b1d16645f02bf03f9934c6acfe86d41705734cf6 Author: athena Date: Thu Dec 28 16:10:33 2006 -0500 Make sure that the speed() input is zero even in paranoid mode. [empty commit message] commit bd281a3248526dd660f3cc5db5662a38af6aca70 Author: athena Date: Thu Dec 28 11:41:46 2006 -0500 cld0 and cldm problems must be tainted because they are used in a v-loop. [empty commit message] commit 762203e35e50a636ebbe34f1bb4a9a72dbdfceae Author: athena Date: Wed Dec 27 17:17:45 2006 -0500 Run paranoid-check in patient mode. [empty commit message] commit 13fcf5a8a0073c72a967a6b5c21009dc09b0e63e Author: athena Date: Wed Dec 27 10:51:42 2006 -0500 Fixed incorrect initialization to zero. [empty commit message] commit 957a6a68a80da76a90adfd5c50e6570a0102a174 Author: athena Date: Wed Dec 27 09:33:02 2006 -0500 Fixed wrong TAINT() [empty commit message] commit 131e00d9d05b13400d93ba18bddd02ee53db276a Author: athena Date: Tue Dec 26 22:50:28 2006 -0500 Grrrrr... [empty commit message] commit e30b1ccf6d58829fdae52ee072c605802f728761 Author: athena Date: Tue Dec 26 22:48:44 2006 -0500 Give up trying to verify rdft2 when vrank=-infinity. [empty commit message] commit 50b70cffda5f011216fd43162aa1b6b69f7ef912 Author: athena Date: Tue Dec 26 22:31:38 2006 -0500 typo [empty commit message] commit 4cdf2b4a16da17795eadebf1a1215e1060f1c573 Author: athena Date: Tue Dec 26 21:54:53 2006 -0500 Correctly verify rdft2 when vrank = -infinity. [empty commit message] commit a93befe83478f18c088fc474973185660c237154 Author: athena Date: Tue Dec 26 21:25:02 2006 -0500 rdft/buffered2.c now generates rdft2 subproblems, not rdft. The old rdft2->rdft reduction is now in rdft/rdft2-rdft.c and still does way too much. commit 994d04b97146dcfb849bc7d83136402cb1a0a070 Author: athena Date: Tue Dec 26 14:03:27 2006 -0500 Buffer the input in hc2r problems, as opposed to the output. [empty commit message] commit 47f2f5a1335a6cc49ed95c73655fa08a19958606 Author: athena Date: Tue Dec 26 10:02:59 2006 -0500 streamlined buffered solvers [empty commit message] commit add79ce24b3c20348a098ca15c7431ce95835a54 Author: athena Date: Mon Dec 25 16:08:22 2006 -0500 c++ compatibility [empty commit message] commit 7caedf2e6196972d2a5c4745ff5badc6856c1b29 Author: athena Date: Sun Dec 24 20:27:23 2006 -0500 Gratuitous renaming of directories and files since the old naming was becoming too inconsistent for my taste. [empty commit message] commit ab5397b31582ea9372345a7868a165a340a8aefb Author: athena Date: Sun Dec 24 20:11:50 2006 -0500 Fixed another dftw bug (sigh) [empty commit message] commit aa1c7cfcbfb0b6bedd716d02d2b84ea86432f03b Author: athena Date: Sun Dec 24 11:48:01 2006 -0500 Removed debugging leftovers. [empty commit message] commit 401d4cd84bec4348694725e268ccb78c5cbbb71c Author: athena Date: Sun Dec 24 11:34:17 2006 -0500 Moved dftw-generic* to new dftw protocol. [empty commit message] commit 91d2ba6242b230cf1195cf9c8e7d61f2ee226a28 Author: athena Date: Sun Dec 24 09:37:19 2006 -0500 Oops [empty commit message] commit ef28c24515ebda3edb9fcb0e67682b3e608f7e67 Author: athena Date: Sun Dec 24 09:31:46 2006 -0500 Fixed wrong verification of rank-1 rdft2 [empty commit message] commit a0f01cb9aaaa491d23686acf3ccedc38c47cd8bb Author: athena Date: Sat Dec 23 20:11:29 2006 -0500 minor tweaks [empty commit message] commit de3d507c8f454e02a382e0cbf7d9453f04621021 Author: athena Date: Sat Dec 23 19:50:03 2006 -0500 Removed obsolete items. [empty commit message] commit 67ca9cb9258b4fb320d26445040b6fae0e450594 Author: athena Date: Sat Dec 23 17:56:37 2006 -0500 Modified the problem_dftw invocation protocol. apply() now requires pointers to the beginning of the full array. Each thread processes a slice mb <= m < me. This protocol is consistent with the one used in hc2hc, where there is no other choice. commit 378686a490f47d469eedfb5383cb46f6500cf835 Author: athena Date: Sat Dec 23 16:18:25 2006 -0500 typo [empty commit message] commit a15a3176f43a12770d66407d77b3fc138f278a53 Author: athena Date: Sat Dec 23 16:06:56 2006 -0500 changed hc2hc twiddle storage to be the same as hc2c [empty commit message] commit a4d048b7080396fae83e41bd64c4740ba0ab9f7a Author: athena Date: Sat Dec 23 15:16:36 2006 -0500 Allowed extra_iter in dftw-direct. Rationalized twiddle factors in hc2c. [empty commit message] commit 4c0bf02653b7e58b218d47e0cf01d719edc1d015 Author: athena Date: Sat Dec 23 10:37:11 2006 -0500 Implemented unmentionable hack to use 4-way SIMD with an odd number of iterations. commit 992d3ce4a54640d5af4d942ef17eb880f56ec36e Author: athena Date: Fri Dec 22 22:13:30 2006 -0500 altivec support for new codelets [empty commit message] commit e0908cfd6fe22ae0544576667bec649b71dcb922 Author: athena Date: Fri Dec 22 19:09:15 2006 -0500 fixed incorrect computation of W [empty commit message] commit df8bd57748278b92e0975a66062bd71bb5ac2e8d Author: athena Date: Fri Dec 22 18:51:22 2006 -0500 Implemented 4-way simd hc2cdftv Also eliminated the twiddle_shift hack. A zillion changes dictated by this choice, which was in turn necessary for the hc2cdftv thing to work. commit 7bac8d1f715f737bfed8742521fe60d5dec6b963 Author: athena Date: Fri Dec 22 08:45:46 2006 -0500 Fixed verification of rdft2 problems with new format. [empty commit message] commit c18f29a56027a5e08aa164530d10ff55c1950170 Author: athena Date: Fri Dec 22 00:05:59 2006 -0500 Added file [empty commit message] commit aafef0ef88f37d8b0e63a31afec168dad67a29f4 Author: athena Date: Fri Dec 22 00:02:50 2006 -0500 Hmm, previous commit did not work [empty commit message] commit 21c810018e7f4993ebadf9a05682f3bd0a6d2c8f Author: athena Date: Thu Dec 21 23:58:33 2006 -0500 Added SIMD r2cdft codelets. [empty commit message] commit fe2f5075f1443d522b445b31027cacb32e8add18 Author: athena Date: Thu Dec 21 21:19:21 2006 -0500 Bug in buffering, grrr... [empty commit message] commit 32f34cf494321ef860f20924b84df527d63e0ce0 Author: athena Date: Thu Dec 21 20:58:14 2006 -0500 Oops, memory leak. [empty commit message] commit 179cfb2f9ddf45916458e2dfdd0e0adbcf762044 Author: athena Date: Thu Dec 21 17:12:31 2006 -0500 minor changes, cleanup. [empty commit message] commit 2505062c2e9ded3822b01c123e82033d37968917 Author: athena Date: Wed Dec 20 22:09:28 2006 -0500 Unified hc2hc-direct, hc2hc-directbuf. Cleanup. [empty commit message] commit 5f80c36202acc746148c007e394a4260457f1f60 Author: athena Date: Wed Dec 20 17:55:56 2006 -0500 removed obsolete rdft2-radix2 [empty commit message] commit 18d4fddebb71049478d41152af043e33ed90d014 Author: athena Date: Wed Dec 20 17:51:20 2006 -0500 implemented reduction rdft2->dft [empty commit message] commit f84c7e67e6c77dc3b8ef4c0703277aa884852ab0 Author: athena Date: Wed Dec 20 09:29:39 2006 -0500 Implemented buffered direct-r2c, direct-hc2c. Also, removed some old cruft: * okp() functions were never used and a pain to maintain---now they are gone. * ``m'' in hc2hc and hc2c codelets is now the number of iterations, not the ``logical'' m. commit de904f19b230a114ead0b9580646689ab8519a29 Author: stevenj Date: Tue Dec 19 17:07:04 2006 -0500 added memcpy-loop rank0 solver (it makes a 5-20% difference for transposes of large tuples) [empty commit message] commit d91736d4442e92910eabaa0e923d0cda833213cc Author: stevenj Date: Tue Dec 19 16:15:54 2006 -0500 new variable to disable libbench2's problem allocation during speed benchmarking (to benchmark MPI transforms where the array does not fit into the memory of a single process) [empty commit message] commit 8c1f9aabc4cd60f5509b287de2850c2767d07fd8 Author: stevenj Date: Tue Dec 19 14:55:08 2006 -0500 allow transpose-inplace to use input as scratch for DESTROY_INPUT plans (to avoid non-square in-place transpositions) ... on supersgj, the planner often prefers transpose-inplace to transpose-alltoall in this case (apparently MPI_Alltoall in LAM MPI isn't that great) [empty commit message] commit 762d58ccde3a29468b9b522c8426ba6f48f0e74b Author: athena Date: Tue Dec 19 17:07:14 2006 -0500 For some reason HB2 codelets were not generated. [empty commit message] commit 928be47a01cfc332b729fd60775949d699d60795 Author: athena Date: Tue Dec 19 15:12:39 2006 -0500 split rdft/direct.c into direct-r2r and direct-r2c, since the file was getting out of control. [empty commit message] commit cf38c33836313129b7e98c192434dae261777810 Author: stevenj Date: Tue Dec 19 02:59:35 2006 -0500 added dft-rank1 solver - MPI now supports 1d complex DFTs! [empty commit message] commit 061b341f302122d768db24c7aab043ade2e2dbb8 Author: stevenj Date: Tue Dec 19 01:27:20 2006 -0500 fftw_flops must call cost_hook directly; iestimate_cost always uses COST_MAX [empty commit message] commit ebf61714b949775e7004b86b828112ae82b69726 Author: stevenj Date: Tue Dec 19 01:16:54 2006 -0500 fftw_flops and fftw_estimate_cost must now be called from every process, to prevent deadlocks in the MPI code (since they sum/max the cost over all processes) [empty commit message] commit bea9d02f1cc9edd02ad6d30a11795bc11ff64d1a Author: stevenj Date: Tue Dec 19 00:55:34 2006 -0500 whoops, typo in assert [empty commit message] commit aee20bd2d616611547ef7826e5d30bf033994736 Author: stevenj Date: Tue Dec 19 00:51:07 2006 -0500 remove multiplication by FFT_SIGN [empty commit message] commit 8583a34cfe3ae51cd835c45d9035d80a0f944b52 Author: stevenj Date: Mon Dec 18 22:51:45 2006 -0500 need to synchronize ESTIMATE costs in MPI planner, and sum ESTIMATE costs for flop reporting: generalize measure_hook to cost_hook(..., {COST_SUM, COST_MAX}) [empty commit message] commit 97ebd8fa0c58fd24345e3386b28f1c7abba8cb8d Author: stevenj Date: Mon Dec 18 15:36:15 2006 -0500 previous patch slowed down transpose-alltoall when TRANSPOSED_IN and DESTROY_INPUT; now allow planner to choose old behavior in this case [empty commit message] commit 045a9c7e5b7ac5d91eb7567f34a2b4e307adeeeb Author: stevenj Date: Mon Dec 18 15:15:39 2006 -0500 transpose-alltoall doesn't require input to be destroyed if TRANSPOSED_IN is set [empty commit message] commit 43ab77408bc3d76457a0e89ad02aec84f4949bf3 Author: athena Date: Mon Dec 18 17:41:25 2006 -0500 Added t2-style hc2c codelets, fixed typos. [empty commit message] commit 7e431baa8e5da11432db111e201a4da9b19d6f49 Author: athena Date: Mon Dec 18 16:02:42 2006 -0500 Renamed certain variables to avoid calling an output stride `is'. [empty commit message] commit 141dcad10c248a45577b80c26d1e396530597d3b Author: athena Date: Mon Dec 18 15:54:43 2006 -0500 Oops---wrong stride [empty commit message] commit 23386506c8294fc1d61bc7cfcedb0bebc4e8fe60 Author: athena Date: Mon Dec 18 14:59:16 2006 -0500 Oops, forgot to add file [empty commit message] commit 8e69f0617fe6f90d09c23d4ce8f125f1822eb363 Author: athena Date: Mon Dec 18 14:17:02 2006 -0500 Renamed r2hc/hc2r codelets to r2c After the recent changes, r2hc/hc2r codelets became rdft2 problems, so I renamed them accordingly to r2cf/r2cb. Codelet parameters are now a real array and a complex array, instead of an input array and an output array, and forward and backward codelets have the same type, which removes some clutter from the rdft code. commit 91b8d21aa599744cf6f9bb2141bcd4193fcdb957 Author: athena Date: Mon Dec 18 10:48:07 2006 -0500 Implemented backward radix-2k rdft2. [empty commit message] commit c91a0bf1ed85466bcb46d2f55128399855c90f06 Author: stevenj Date: Mon Dec 18 13:56:09 2006 -0500 move extract_reim into kernel, since it is used by internal MPI stuff and not just in the API code any more [empty commit message] commit ab0e79e7a1319598924b4d434f1a0ce57bc58a43 Author: athena Date: Mon Dec 18 08:40:14 2006 -0500 Do not check r1==cr unless rnk>0 [empty commit message] commit 4a203a4d39cb5b02d1f6d83f2f525a6a6c0cf065 Author: athena Date: Sun Dec 17 21:03:50 2006 -0500 Implemented radix-2k RDFT2, forward only for now [empty commit message] commit 6d86c9dd27b6fbfb45faf91980565df40ec8d825 Author: stevenj Date: Mon Dec 18 01:23:45 2006 -0500 separate TRANSPOSED/SCRAMBLED flags internally (this is required so that dft-rank1-bigvec and the future dft-rank1 won't have incompatible SCRAMBLED formats) [empty commit message] commit 7920d86807a6fe9829cca1cb4e633ab3156c8b38 Author: stevenj Date: Mon Dec 18 01:02:27 2006 -0500 ops_add -> ops_add2 where possible, to shrink code [empty commit message] commit 81d5eddab81d4c0265863e1da6302b63fb1a9a80 Author: stevenj Date: Mon Dec 18 00:43:02 2006 -0500 added dft-rank1-bigvec solver (easy case for 1d parallel transforms) [empty commit message] commit bf7a77840dcbe0b1d5a8a1d7568877f093299e69 Author: stevenj Date: Sun Dec 17 20:42:21 2006 -0500 rewrote MPI stuff to use dtensor data structure A dtensor is an ordered tuple of triplets (n, ib, ob) giving the size of a dimension (n) and its input and output block sizes of a distributed row-major multi-dimensional array. An MPI DFT (etc.) is now specified in terms of dtensors, which provide a much more flexible data layout. For example, we can now describe multidimensional block distributions, which are important if the number of processors is greater than the size of any given dimension. Currently, we only have solvers for 1d slab distributions, and this is all that is supported in the basic and advanced APIs. The guru API allows one to specify more general distributions, however, which will be useful when/if we have solvers for this case. We now also don't need a TRANSPOSED flag, at least internally, since TRANSPOSED multi-dimensional DFT plans just correspond to dtensors where the input and output block distributions are different. Other changes include the use of the XM(foo) macro for X(mpi_foo). commit c9e4b997dd8cd38b753a8c40fb0004ab07124ca7 Author: athena Date: Fri Dec 15 16:01:23 2006 -0500 Distinguished mutexes from semaphores. The distinction is useful because the linux implementation of sem_post() in unnecessarily slow when semaphores are used for mutual exclusion. This change made spinlocks messier to implement, so I excised them. commit 352252ac74f50d01ae2c996f0072533f84c9a043 Author: athena Date: Fri Dec 15 12:46:11 2006 -0500 Use posix semaphores where available. Paranoid declaration of all shared variables as ``volatile''. Paranoid initialization of all shared variables within locks. commit 3364aa89b74e5e6a3514c653dc4ef4ebeebddae4 Author: Matteo Frigo Date: Wed Dec 13 20:12:13 2006 -0500 paranoia [empty commit message] commit 0be1cd7e754cf3b9550bcf6f60c2b31643d4512a Author: stevenj Date: Tue Dec 12 17:28:13 2006 -0500 punt on detecting unsolvable rdft2 problems; make r==iio rdft2 problems unsolvable, since it doesn't look like we've consistently checked for this case and it's not clear why we would want to support it (it was also not documented in the manual) [empty commit message] commit 8f3194d212eeb8c2382a60a4db59ff1cf935faef Author: athena Date: Sat Dec 9 12:14:13 2006 -0500 Obey stupid const rules [empty commit message] commit 1f7938759fd6c3a52293ffeffea1961692a22e72 Author: stevenj Date: Fri Dec 8 18:21:50 2006 -0500 added unsolvable check for rdft2 problem An in-place rdft2 problem is ill-formed if the real data, including the extra "padding" elements, do not coincide with the complex data. CHANGE: the new code considers all in-place split r2c and c2r problems to be ill-formed. Previously, these could be done, but only if the entire multi-dimensional array fit into the buffer, which is kind of stupid. I'm not sure it's worth it to even try to support the split in-place r2c case. commit e5241fedc1b2a9be448809abfc8b812e07598801 Author: stevenj Date: Fri Dec 8 13:47:53 2006 -0500 check in-placeness after joining taints [empty commit message] commit b7a9db50bbf343d80e6b5d0cbaae7cd4d8653f1e Author: athena Date: Fri Dec 8 13:43:44 2006 -0500 Grrr... paranoid-check was not testing in exhaustive mode [empty commit message] commit d34bae7a5b89e8298450b98759be43f22e7c99d3 Author: Matteo Frigo Date: Fri Dec 8 10:00:30 2006 -0500 Implemented PROBLEM_UNSOLVABLE. In-place DFT and RDFT problems with inconsistent I/O strides are now unsolvable, and we don't check for them any longer in solvers. While I was at it, declared all problem pointers to be ``const'' for extra safety. commit db6e8b81d60e8cbe9f49ac3035a5151759afc88b Author: Matteo Frigo Date: Thu Dec 7 20:13:46 2006 -0500 Avoid qsort'ing one element. [empty commit message] commit a6d29bce88258799bb4bc6ee27c36aa5ccf4cbcd Author: Matteo Frigo Date: Thu Dec 7 18:25:47 2006 -0500 In-place vrank>=1 is now applicable only if the problem is really in-place. [empty commit message] commit f0e0bda2dacfef167a5387c26d0c5631b59aaa0c Author: athena Date: Tue Dec 5 12:21:38 2006 -0500 unused variable [empty commit message] commit 07dc6817f4991f8690c6b6952bc3879c4712a624 Author: Matteo Frigo Date: Sun Dec 3 19:16:33 2006 -0500 Removed CVS $Id$ everywhere, since darcs does not update them. [empty commit message] commit 2cf2355d9987a09710a6b6b345cf232abf8c1a8b Author: Matteo Frigo Date: Sun Dec 3 16:11:17 2006 -0500 generalized dftw to encompass q codelets. As a side effect, q codelets are now threaded. [empty commit message] commit 942c7b34d673282b52aacbb51237b38f3ffef3f5 Author: stevenj Date: Sat Nov 25 16:34:38 2006 -0500 add missing __declspec attribute to threads API functions when compiling for Windows (thanks to Robert O. Morris for the bug report) [empty commit message] commit 8c4b9a9a79e7fdbbd7eab18f4aac4ac94c8ce2cb Author: stevenj Date: Mon Nov 20 17:39:20 2006 -0500 add AC_SUBST to AX_OPENMP, thanks to Sebastien Maret for the suggestion [empty commit message] commit 02d141e6157d606dce0bf146248958313ea19466 Author: stevenj Date: Mon Oct 16 23:02:29 2006 -0400 not gcc bug for MIPS (thanks to Jonathan Day) [empty commit message] commit 87fc8a66ef9cf8161a4cf23c0580f91ec9e86e25 Author: stevenj Date: Sat Sep 23 17:52:36 2006 -0400 in maintainer/debug mode, don't modify CFLAGS if they were explicitly set (-pedantic seems to cause problems with LAM's mpicc, so I need a way to override) [empty commit message] commit d1ebd06376db08bd5afecbe45d6059f60f7cd09d Author: athena Date: Wed Nov 22 18:14:47 2006 -0500 Removed obsolete comment. [empty commit message] commit b7bdd19e3b444a9c0ce68991739644a0fd4d9166 Author: athena Date: Sun Nov 19 11:21:44 2006 -0500 Use p->v when comparing TW_FULL fields. [empty commit message] commit fd6481c30c72bf15c3316dd7db6664e5b801160e Author: athena Date: Thu Nov 16 14:49:05 2006 -0500 removed useless definition [empty commit message] commit d161dc974c19cf43addd4b6cb516ae6b770827e7 Author: athena Date: Mon Nov 13 09:18:32 2006 -0500 paranoid avoidance of integer overflows [empty commit message] commit e5a1cce0ead5ae9d73c2c38c48e66c3bf059a874 Author: athena Date: Mon Nov 13 09:00:11 2006 -0500 avoid potential overflows in cycle counters At the suggestion of Alex Cichowski, convert all ticks to double before operating on them, to avoid potential signed/unsigned confusion and integer overflow. commit 98e4c9b9751d26d0adecc56634cc77e5689357bc Author: Matteo Frigo Date: Sun Nov 5 09:00:52 2006 -0500 Removed unused struct field [empty commit message] commit 303349e158e3fdf0231790fe32a6831aa671f895 Author: Matteo Frigo Date: Sat Nov 4 09:43:13 2006 -0500 use pthread condition variables instead of semaphores Condition variables are more likely to be portable everywhere, and somehow they appear to introduce less overhead at least on my linux box. commit eb7fb1efca70242568f0e74266ea88e8c9a45eff Author: Matteo Frigo Date: Tue Oct 31 20:45:24 2006 -0500 Imprecise help message. [empty commit message] commit 08bdb758c515972281738ca7567e38d3aeb05cf4 Author: athena Date: Mon Oct 30 20:13:35 2006 -0500 Experimental implementation of spinlocks. This patch implements spinlocks via a semi-portable hack, and adds the -ospinlocks option to the bench program so that we can play with them. commit fe239f5afbec9a3868f4330849cfe3d6dccb54b0 Author: athena Date: Thu Oct 26 22:29:18 2006 -0400 Updated manual for new openmp configure options. [empty commit message] commit 9809db57ce3f7e945e7bf04105a3fce74cdf7f47 Author: athena Date: Thu Oct 26 21:52:39 2006 -0400 Added back openmp. Rationalized threads naming conventions: * threads explicitly managed by us are enabled by --enable-threads, predicated on HAVE_THREADS, etc. * openmp is enabled by --enable-openmp, predicated on HAVE_OPENMP, etc. * SMP denotes either THREADS or OPENMP. commit 22cd21b038b6dd972444d5b00a6ebd00a932aa7f Author: athena Date: Mon Oct 23 20:14:31 2006 -0400 different thread protocols [empty commit message] commit f61da0e9c9db3c22dfd61c108f5f65bc4afe78bd Author: athena Date: Sun Oct 22 14:49:32 2006 -0400 fix memory leak Added pthread_attr_destroy to avoid memory leak. commit 6f6532928b29c6ac7599424c9dc834c41fc7fcf0 Author: athena Date: Sun Oct 22 14:23:30 2006 -0400 Experimental new pthread implementation that recycles threads. [empty commit message] commit e8c76bbab164fd6ad784bcfd9ec0fe5f002bfb43 Author: athena Date: Tue Oct 24 23:28:10 2006 -0400 switched buddies Switched order of buddies in rdft2 rank-geq2 for consistency with analogous dft and rdft solvers. Furthermore, this change reduces the MEASURE planning time for rank == 3. commit 3e13c85d1651dfc698143631f47a45b4c3947d12 Author: Matteo Frigo Date: Sat Oct 21 10:10:00 2006 -0400 Typo. [empty commit message] commit 7151038f7642b5ca050afb037fd91719a6f733b8 Author: Matteo Frigo Date: Sat Oct 21 09:56:50 2006 -0400 Out of place is the default. Thanks to Kirk Kern for pointing this out. [empty commit message] commit 281e20106cf076681392cb66050e11d2ac758dd2 Author: stevenj Date: Fri Sep 29 01:36:11 2006 -0400 rename "test" to "tst", since a user (Igor Levicki) reports that "test" is a reserved words in some x86 assemblers [empty commit message] commit 859e712b9c8051f6a1fb5a6500472042f194712f Author: Matteo Frigo Date: Tue Sep 26 09:01:08 2006 -0400 Stylistic change. [empty commit message] commit d024e575f98dc4a3452b9e0d5115a6650a7bf951 Author: athena Date: Tue Sep 26 08:45:37 2006 -0400 Do not set tmin=1e10, since a large FFT may take longer than that. [empty commit message] commit f31fffd32e9497675200784973ee8420ef3d60db Author: athena Date: Sat Sep 23 22:07:10 2006 -0400 Disable certain gcc optimizations. When PRECOMPUTE_ARRAY_INDICES is #define'd, array indices have the form array[stride[k]] for compile-time constant k. Apparently new gcc's copy stride[k] onto the stack before the codelet loop, which is an idiotic optimization if ever there was one. This patch confuses gcc enough to prevent this optimization. commit 0cc1f20ac1bbadb2e82d6465941755a9754d69d7 Author: stevenj Date: Sat Sep 23 13:02:58 2006 -0400 re-enable TOMS algorithm - it is the best for large vector lengths, since for such sizes the cache line is not an issue and the bookkeeping overhead is negligible [empty commit message] commit 458c7ee2e058d2cbe6fc90d23780b59bb29fbb52 Author: stevenj Date: Thu Sep 21 15:40:15 2006 -0400 add measure_hook so that MPI can synchronize timing measurements (otherwise different processors might end up with different MPI plans, yikes!) [empty commit message] commit 9fa6c37e2e09ea6bd226bde9a62f95d7c66f33bb Author: stevenj Date: Tue Sep 19 21:26:19 2006 -0400 added O(p log p) transpose algorithm (radix 2) [empty commit message] commit 1316689f96089a3b53799a3733d15add7c2f267c Author: stevenj Date: Tue Sep 19 21:05:09 2006 -0400 comments [empty commit message] commit c4b69d82b2c7de2fa2a963e27e3e498544c80262 Author: stevenj Date: Tue Sep 19 19:54:58 2006 -0400 whoops [empty commit message] commit 19265283619d376581c5815adab98a2b51fb0cb2 Author: stevenj Date: Tue Sep 19 19:49:01 2006 -0400 synchronize planning so that if one process fails to create a plan then all of them do. [empty commit message] commit e1b970b70562528b08d665ad2f7a17ee31f23e47 Author: stevenj Date: Tue Sep 19 18:17:38 2006 -0400 call MPI_Alltoall instead of MPI_Alltoallv for equal-blocks case, in case MPI implementation has special optimizations for the common case of equal sizes [empty commit message] commit 3adcd54e6a54928afc2c58b7aab4bc75664be1d1 Author: stevenj Date: Tue Sep 19 12:07:35 2006 -0400 whoops [empty commit message] commit 31b25a3bccaa0a8290dab5010199da8a5fec619d Author: stevenj Date: Tue Sep 19 02:20:06 2006 -0400 typo in comment [empty commit message] commit 519395bdb6ed85a2f18ba95bc82a68b2d42ddaae Author: stevenj Date: Tue Sep 19 01:58:55 2006 -0400 more filename simplifications [empty commit message] commit 2870207d79654158ecf7ae1d68d37382be5e39da Author: stevenj Date: Tue Sep 19 01:49:52 2006 -0400 canonicalize file names (hyphens, not underscores) [empty commit message] commit 41c88ac7e2cc2df2ff894c7223d55a41800b4c98 Author: stevenj Date: Tue Sep 19 01:34:07 2006 -0400 add mpi-dft-serial [empty commit message] commit cbf1beef8656c2b0bc205c4787e6986cf32b405a Author: stevenj Date: Tue Sep 19 01:21:47 2006 -0400 silence warnings [empty commit message] commit ff7df52317b96acc1966b2d5920e46a3e368bded Author: stevenj Date: Tue Sep 19 00:31:59 2006 -0400 make "t" problem semantics match FFTW_MPI_TRANSPOSED [empty commit message] commit 694244ed8ede7153eca565e43ff553a26db39b8f Author: stevenj Date: Mon Sep 18 23:50:43 2006 -0400 whoops, fixed backwards mpi_dft [empty commit message] commit a36a49ee7709c1f3010f32039814f655e502850a Author: stevenj Date: Mon Sep 18 22:26:31 2006 -0400 initial stab at rank-geq2 mpi-dft; seems to be mostly working [empty commit message] commit c8e0a65f63c0d8eb6a148cee255c5aca2ff4c68b Author: stevenj Date: Sun Sep 17 13:41:32 2006 -0400 support SCRAMBLED_OUT in alltoall transpose [empty commit message] commit 72887de15eec06aeb7426d6a7cc527fc171821b0 Author: stevenj Date: Sun Sep 17 12:34:30 2006 -0400 skeleton of future support for block-cyclic [empty commit message] commit 62b562d1df02409e10395385348f56318e46a2e7 Author: stevenj Date: Sun Sep 17 12:11:19 2006 -0400 test program now checks scrambled in/out via -obflag=28/29 [empty commit message] commit b3ef0c11ca26e02875c29eb154f1a5b9bf386ad5 Author: stevenj Date: Sun Sep 17 11:58:36 2006 -0400 added -obflag to make it easier to set high-order bits [empty commit message] commit a4cbe985fe508661ccf587b5331e62b0a6526289 Author: stevenj Date: Sun Sep 17 01:30:51 2006 -0400 use proper child plans for 2nd transpose in transpose_alltoall; implement opcount in transpose_inplace [empty commit message] commit ab2dd6cfd9903487e3ac3cf9401c9a1f35de4862 Author: stevenj Date: Sun Sep 17 01:08:01 2006 -0400 fix in test program for transposes of vectors -- transpose routines seem to completely work now (except for scrambled in/out, which is untested) [empty commit message] commit aa5eecbf266020f2e2788bba862c13f6575d1ce8 Author: stevenj Date: Sun Sep 17 01:01:16 2006 -0400 fixed bug in transpose_alltoall for unequal blocks [empty commit message] commit 202f232a54abc2cb04e0a3d0d32bdb727c306c4e Author: stevenj Date: Sat Sep 16 15:29:46 2006 -0400 correctly handle cld2rest [empty commit message] commit 36668c0dd144f2bc2f6b2bf1f10eb1677593b9c6 Author: stevenj Date: Sat Sep 16 15:29:31 2006 -0400 some debugging code and other fixes [empty commit message] commit 66dcf1f5c673fd16b2f0f88988c4aaf388eeaf27 Author: stevenj Date: Sat Sep 16 14:54:30 2006 -0400 whoops, forgot to check in mpi_bench.c file [empty commit message] commit 3d96f316225934ef4485bcc2432314b89292914b Author: stevenj Date: Sat Sep 16 14:54:02 2006 -0400 added bench_exit routine so that it can be overridden (by MPI_Abort) if needed [empty commit message] commit b14337a969f6fee88bda25464c7ef7c0e56b5c00 Author: stevenj Date: Sat Sep 16 14:52:56 2006 -0400 bug fix in mpi_transpose_inplace for case where some processors are idle [empty commit message] commit 43fd42786f54710bffe85528beae2fff76e4a58e Author: stevenj Date: Fri Sep 15 18:47:13 2006 -0400 allow vecloop for sz->rnk==0 in exceptional (SLOW) cases, e.g. it is necessary for loops of non-square transposes (otherwise e.g. ik1v5:200:200x10:20:1x20:1:10 planning fails) [empty commit message] commit 121eaa69908a7b465f21f3529f74e983a63801ad Author: stevenj Date: Thu Sep 14 23:36:48 2006 -0400 first pass at working mpi_bench test program; transpose seems to work iff dimensions are divisible by #processors [empty commit message] commit 51101a902b4fdaef585e1d9e975238100951601c Author: stevenj Date: Wed Sep 13 17:28:07 2006 -0400 whoops [empty commit message] commit 7986cd7f00327db5f156e8d4d1458456f309e37a Author: stevenj Date: Tue Sep 12 22:27:03 2006 -0400 initial stub for mpi_bench [empty commit message] commit 279ca0155c7cb9dcd9bb9c75149a24bb1f44ba50 Author: stevenj Date: Tue Sep 12 21:54:31 2006 -0400 do no output at all if verbose < 0 (for use with MPI, where we only want output from process 0) [empty commit message] commit 481f3838af04ae3db7aee15094ecf748f71d03da Author: stevenj Date: Tue Sep 12 21:39:15 2006 -0400 whoops [empty commit message] commit 792aaa1acca61e89b5605cbed49e9dd86bfbc2b1 Author: stevenj Date: Tue Sep 12 21:31:40 2006 -0400 split bench.c into bench.c and fftw_bench_common.c so that we can re-use some of the code in the MPI test program [empty commit message] commit 3b3b95ad0b1ab373687a9df59cdf2ec4bcdd502b Author: stevenj Date: Tue Sep 12 21:00:36 2006 -0400 MPI stuff at least compiles now [empty commit message] commit f5092f54bc2b8ee7289a2fb5148fc5315cbb2ee8 Author: stevenj Date: Mon Sep 11 22:26:36 2006 -0400 initial (nonfunctional) start at MPI support (similar to FFTW 2.x in spirit, but mostly rewritten) [empty commit message] commit 64d68fafe0c7f4433aec4a0925ce5972c33c78b4 Author: stevenj Date: Mon Sep 11 22:25:38 2006 -0400 make X(plan_awake) work for NULL argument to reduce code size [empty commit message] commit ed05c503c219544b0fe91af61db02d9cbb4027b5 Author: stevenj Date: Fri Sep 15 23:47:08 2006 -0400 -mt should go before -mthreads to avoid spurious warnings on HPUX (thanks to Peter O'Gorman for the bug report) [empty commit message] commit d737c7b3eca4ff8d7f372273f114dfd4e765b70c Author: stevenj Date: Mon Sep 11 13:53:44 2006 -0400 Fortran init_threads wrapper didn't return result; thanks to Markus Wetzstein for the bug report [empty commit message] commit ba5664a7958d533904b9251a4bfaa56b0f338a8a Author: stevenj Date: Thu Sep 7 18:43:55 2006 -0400 make sure wrappers are included even if Fortran compiler was not detected (unless --disable-fortran was specified explicitly) ... this was supposed to be done before, but the definition was in the wrong place, grr [empty commit message] commit 188c9dde71d0bc56ba30a052b82d02b6676f20ed Author: stevenj Date: Thu Aug 31 19:33:29 2006 -0400 I'm sick of answering this question about non-deterministic results [empty commit message] commit 336fb6116c43aa5559392ea2d0759606efd6f275 Author: Matteo Frigo Date: Tue Aug 22 21:27:29 2006 -0400 Add --tag=CC flag to libtool. This change is consistent with the libtool invocation in the latest automake, and is required to compile with (some version of) xlc. commit d98d86f9115ca1a836e92d8df8e061f98f329032 Author: athena Date: Mon Aug 21 21:40:36 2006 -0400 avoid ``fma'' because it is defined in c99. [empty commit message] commit 841eb8db14a22936ba8ef81f439f42cb2411073d Author: Matteo Frigo Date: Sun Aug 20 11:40:53 2006 -0400 Obey -standalone flag. [empty commit message] commit f270abac4732fe5f77708bef5f0d0cdc599bdb61 Author: Matteo Frigo Date: Sat Aug 19 13:34:27 2006 -0400 obey -standalone when generating simd codelets [empty commit message] commit f573bbe2aaafabfbb21daf7da62972b8b071167f Author: Matteo Frigo Date: Sat Aug 19 13:33:43 2006 -0400 removed obsolete athfft [empty commit message] commit 84e5b7792da92198e101b168d10710f0b81df5e8 Author: stevenj Date: Thu Aug 17 21:50:50 2006 -0400 updated citation to Proc. IEEE paper [empty commit message] commit 5ca3a79e05b95a688c21e7cb37a1ef7fa42a7f04 Author: stevenj Date: Thu Aug 17 21:47:05 2006 -0400 use darcs changes --summary to make nice changelog; emacs fill-region hack is obsolete [empty commit message] commit b80ff1b7affc5ba9c62bde0b06a548c3baf7c615 Author: athena Date: Mon Aug 14 17:53:19 2006 -0400 removed timer calibration Timer calibration seems not to work any longer on recent processors--- too much noise. I have remove it completely. commit 8986b3fa943f3f424a2f75541f8627a86af31a0a Author: Matteo Frigo Date: Mon Aug 14 10:47:15 2006 -0400 removed k7 Removed obsolete k7 support. commit 820835bfa680e9a0193435bfbcaf21923df9e7fc Author: athena Date: Sun Aug 13 11:02:11 2006 -0400 Use darcs instead of cvs. [empty commit message] commit 818c52da26a5d0781db8d9b45d4026403fb7e922 Author: Matteo Frigo Date: Wed Jul 19 08:52:15 2006 -0400 Treat a the string "-" as a nonoption. commit 4e8a814e90696ee38898bfb5f079ac9bb6b614c2 Author: Steven G. Johnson Date: Tue Jul 4 17:10:47 2006 -0400 comment out pkginclude dir for now commit 1bc4dd79b8cc59be7b18676f338c78013da54dab Author: Steven G. Johnson Date: Mon Jul 3 20:51:08 2006 -0400 make sure CCAS = CC to avoid libtool confusion commit ebddd6bce119dec0b9a970a6d6194131321bdc5e Author: Steven G. Johnson Date: Fri Jun 23 04:07:31 2006 -0400 install x77.h guru.h guru64.h in pkgincludedir commit e272fe53d7d822aa7d5ce03277f40c87aa843eef Author: Steven G. Johnson Date: Fri Jun 23 04:03:42 2006 -0400 whitespace commit d6d23fdac18d0d01e363ff60bdba1285be017d0c Author: Steven G. Johnson Date: Fri Jun 23 02:33:45 2006 -0400 support cycle counter with xlc on Linux/ppc commit 677dd906902cf9dd2215c576a8f9d9e6755cc7cd Author: Matteo Frigo Date: Tue Jun 20 08:16:08 2006 -0400 Stylistic change. commit bb6bed2564fdec63eb8439031bc45caf8436b378 Author: Steven G. Johnson Date: Tue Jun 20 02:20:34 2006 -0400 bump date commit 6944a35c403fdcbf6b7b46f1aa9df9288991efca Author: Steven G. Johnson Date: Tue Jun 20 02:20:06 2006 -0400 correct bug reported by Andrew Salamon ... --enable-portable-binary was ignored (or rather, treated unpredictably) due to typo, grrr commit ad98ebc35798f8713ac299ebe9ce74ca9fefe2f1 Author: Steven G. Johnson Date: Thu Jun 1 20:30:06 2006 -0400 install 'internal' header files into includedir/fftw3/, includedir/fftw3f/, etcetera....this will make it easier to write external libraries that plug into FFTW internals, e.g. to add new solvers commit 4ce51f61d823524e8bebc4bc92ad2b17b6e7b53a Author: Steven G. Johnson Date: Mon May 29 23:59:19 2006 -0400 bug fix, thanks to James Donald for the bug report (only affects experimental semaphore stuff) commit ca9e38be107c761af7cd66a3ce9f0cfe93e9c069 Author: Steven G. Johnson Date: Mon May 29 23:58:16 2006 -0400 comment commit 7e4b4be5e1bcdd9706a3ded5e2f59010ff751401 Author: Steven G. Johnson Date: Mon May 29 21:02:50 2006 -0400 whoops commit 5af69a3ec3b932c0d7e3e2dfdbcbff2aa067c5bf Author: Steven G. Johnson Date: Sat May 27 19:36:15 2006 -0400 version bump commit 27dd43e42fa0b4ccea275b2143a9056f42f8c7f9 Author: Steven G. Johnson Date: Sat May 27 14:54:47 2006 -0400 only check for xlc_r/cc_r if we are not using gcc commit c222c025be6649da84164ba5d2334fdcf0b3ac0b Author: Steven G. Johnson Date: Fri May 26 15:00:38 2006 -0400 use ptrdiff_t (it's C89 and standard C++, hooray) commit c3450d7f654ac2adf06bbbe9687f99cf1c6641b5 Author: Steven G. Johnson Date: Fri May 26 12:59:33 2006 -0400 version bump commit 0be4f57c071dc97314660a66f4d46eee4ac143e9 Author: Steven G. Johnson Date: Fri May 26 12:57:32 2006 -0400 noted 64-bit guru API commit 90455678a81def7a9aa3bc14f17047deb714271b Author: Steven G. Johnson Date: Fri May 26 12:53:09 2006 -0400 note that newer versions of VC++ support long long commit efddf05184fe6977af120842d10faf89399f14e0 Author: Steven G. Johnson Date: Fri May 26 12:46:09 2006 -0400 try harder to get a portable 64-bit type commit 1d34caa16af08ff47fd75006c7576242e4643d17 Author: Steven G. Johnson Date: Thu May 25 22:04:18 2006 -0400 added draft guru64 API commit f987e828891ddd69efa3c664c68d231c1d3fc460 Author: Steven G. Johnson Date: Mon May 22 16:41:44 2006 -0400 added FIXME note commit 245cd07427cd24c953e4f1eea383790c7f557701 Author: Steven G. Johnson Date: Mon May 22 16:40:30 2006 -0400 check for xlc_r in addition to cc_r; thanks to Guy Moebs for the bug report commit 8a76c773855a145883608d47ca0aaa369e3ec408 Author: Steven G. Johnson Date: Fri Apr 21 12:35:25 2006 -0400 added note about gcc 4.0.1 on MacOS/Intel commit 9bb0ec78947a8597e0642379e7348e6b1c03af0b Author: Steven G. Johnson Date: Thu Apr 20 23:08:42 2006 -0400 added code for Core Duo; thanks to Eric Branlund commit d7a2e4a3ad51c01ec1bffbbadad602bb643da270 Author: Steven G. Johnson Date: Thu Apr 20 20:21:03 2006 -0400 fixed failure for -fPIC or for gcc-4 on Apple Intel machines; thanks to Eric Branlund for the bug report commit 3cb3cea549b4e8e0f9a16a1952eae4b4d8be1189 Author: Matteo Frigo Date: Tue Apr 11 20:00:31 2006 -0400 Use -maltivec when checking for altivec.h. commit e2fb474c726118343e25059e2e1e8d2da6a21f62 Author: Steven G. Johnson Date: Mon Apr 3 15:52:44 2006 -0400 note planner overwriting input in planner-flags reference commit 89a78d79d22078ee258d43d581cee6aaa3ba1d80 Author: Matteo Frigo Date: Tue Mar 28 09:05:26 2006 -0500 FAQ entry about --enable-k7 in 64-bit mode. commit bfc115831ce70cd5cbf96fc005710862cf10bef6 Author: Steven G. Johnson Date: Mon Mar 27 23:41:05 2006 -0500 sprintf -> snprintf, to avoid (harmless) complaints by users/compilers commit a2e4f6bfa281ed8b11c3a42e4cf32570e6a5c4d9 Author: Steven G. Johnson Date: Mon Mar 27 23:30:22 2006 -0500 silence compiler warning commit 2c39d368d18c97bb079456491d60bb9a0c4c4342 Author: Matteo Frigo Date: Fri Mar 17 09:20:10 2006 -0500 Remove dft/codelets/inplace, add simd/nonportable to list of directories to be compiled on non-unix systems. commit 01fa8ec4e8e6bd7a560437afe4ce4e37c13e0806 Author: Steven G. Johnson Date: Sat Mar 4 16:17:56 2006 -0500 whoops commit 6687db156af27c4ba2a4ddab66b6aa0a951b1a35 Author: Steven G. Johnson Date: Sat Mar 4 16:13:08 2006 -0500 note that we align the stack ourselves if necessary, with gcc and icc commit eee84dd2a9317a44c05e2f4dc2c05ff42709a973 Author: Steven G. Johnson Date: Sat Mar 4 16:08:16 2006 -0500 clearer distinction between static and automatic storage in C commit bc1aba15a1ddb5cd37b8088ea70f81ea6093e8d2 Author: Steven G. Johnson Date: Sat Feb 25 20:27:01 2006 -0500 rm unused var commit d93efe4d5783cc4d5791894d58524c93644d5cb2 Author: Matteo Frigo Date: Sat Feb 25 17:30:28 2006 -0500 Improved usage of goto (Dijkstra miserere nostri) commit 36a203c3ada1b6257109162fee8dc563da9c4bc2 Author: Steven G. Johnson Date: Sat Feb 25 14:19:15 2006 -0500 boilerplate commit 4041499e9299726d5840a0d119af094517810bda Author: Steven G. Johnson Date: Sat Feb 25 14:14:40 2006 -0500 update for upcoming 3.1.1 commit 579c413f3bb5e24ac92d433aa17a063b1f11f8a1 Author: Steven G. Johnson Date: Sat Feb 25 13:57:34 2006 -0500 replace obsolete IMPATIENT with MEASURE commit e0e594ba308e101ba93aacdceabdf0a35b4b0221 Author: Steven G. Johnson Date: Sat Feb 25 13:52:25 2006 -0500 corrected comment commit 5c1e2c07d1d8e21c219853b35212ba7373b35b45 Author: Matteo Frigo Date: Sat Feb 25 10:19:26 2006 -0500 -v does not take an argument. commit dca8aaed07eadc0d1db6fe19b4a86d00ff7a328b Author: Matteo Frigo Date: Sat Feb 25 10:17:18 2006 -0500 Obey the unix convention that -ab = -a -b commit 95450e7e4d3ed287b4ff36d6ccd1250023cc06a2 Author: Steven G. Johnson Date: Fri Feb 24 23:13:49 2006 -0500 minor fixes (return error on unrecognized option) commit af67fa909fc8ad31f5163b26da1693b1f9a61649 Author: Steven G. Johnson Date: Fri Feb 24 22:46:12 2006 -0500 ugh commit fec17358e3fbfb5e049933495db198312f9e10f9 Author: Matteo Frigo Date: Fri Feb 24 21:42:56 2006 -0500 require exact match for long options. commit 5538e310cb61df6e3b5bd880ae604d86ec8f6121 Author: Matteo Frigo Date: Fri Feb 24 21:38:02 2006 -0500 better fix commit 6f17dbe2db5e098604a35c03a7a1514040ea47d6 Author: Matteo Frigo Date: Fri Feb 24 21:37:06 2006 -0500 Fix commit f10cae7e4cf944a6ef5928afbbaead482a4692d3 Author: Matteo Frigo Date: Fri Feb 24 21:25:48 2006 -0500 nothing commit 01a4d4b5c204ef2c6d0afc7402a72481ad4a6c3e Author: Steven G. Johnson Date: Mon Feb 20 17:37:21 2006 -0500 rm transpose-indirect-inplace solver, which was buggy commit baa641d48341281eb3a9d4d9792f4482042836b4 Author: Matteo Frigo Date: Wed Feb 15 08:43:05 2006 -0500 Comment fix. commit 207d1eae51bef5d4d14c7c670d6d16dadd8c8edf Author: Matteo Frigo Date: Wed Feb 15 08:18:41 2006 -0500 Cycle counter for Visual C++ x86-64, courtesy of Dirk Michaelis commit 0aefc1a066f619ed6f5b54791b00ab0acffe2901 Author: Steven G. Johnson Date: Tue Feb 14 19:17:30 2006 -0500 rfftwnd.png is in builddir commit 1799b5fa46fef72b8f116de92605ea4b8118a3db Author: Steven G. Johnson Date: Tue Feb 14 19:03:27 2006 -0500 fixed typo: --enable-portable-binary, not --with commit f75d618a590c61c3bdba28ad0155f327670e231f Author: Matteo Frigo Date: Mon Feb 13 07:59:06 2006 -0500 estimator tweaks. commit 1c0cc8d7bb3570ce31320d87bbe35eb5b03ef38d Author: Matteo Frigo Date: Sun Feb 12 20:43:39 2006 -0500 sse/sse2 support for t3?v codelets commit b1116627def6398d97dc443ba0f9bba1e2989f86 Author: Matteo Frigo Date: Sun Feb 12 20:39:22 2006 -0500 Use CEXP instead of SIN/COS. commit dd361f593b5e85eeea97f2a2c048d3a8dce9d7e5 Author: Matteo Frigo Date: Sun Feb 12 20:12:10 2006 -0500 bug in randomized cse eliminator. commit 9d329a9d010b44b728449d0f566eaa9356682a82 Author: Matteo Frigo Date: Sun Feb 12 18:34:12 2006 -0500 Added support for t2-style simd codelets. This is altivec only for now; sse/sse2 don't even compile yet. commit b187b797d17ccda6efb853eb93b27aa42474a02a Author: Matteo Frigo Date: Sun Feb 12 15:30:27 2006 -0500 Added support for t2-style simd split-complex codelets. commit 78281302f4fd1a0d42f0a0baf64cb364076aedb3 Author: Steven G. Johnson Date: Fri Feb 10 18:21:28 2006 -0500 [empty commit message] commit 948abfe0c2c7db954b7de51ff8da674dab3258ee Author: Steven G. Johnson Date: Fri Feb 10 18:19:46 2006 -0500 punctuation commit e24cb9776a48a5cb0673e4ee8d75d142fcf2b117 Author: Steven G. Johnson Date: Fri Feb 10 18:00:35 2006 -0500 windows DLL stuff for Fortran interface commit fd7272f2e39eddd4491666e0bfe1e5c69c1eb04c Author: Matteo Frigo Date: Fri Feb 10 09:48:52 2006 -0500 Bumped version to 3.1.1 commit 81a965e9d51e8e647f2eeaa12a973f7ef5188314 Author: Matteo Frigo Date: Fri Feb 10 09:18:39 2006 -0500 Precompute array indices on x86-64. Speeds up Pentium IV and makes no appreciable difference on AMD. commit 5dbfa49ad81db0d3dca7e419507654fc0adc63fe Author: Matteo Frigo Date: Tue Feb 7 22:01:36 2006 -0500 Check whether the processor supports CPUID before issuing the instruction. (Grrr...) Code contributed by Eric J. Korpela. commit ce017677182a7662b7b1db85f32c6a8f34773703 Author: Matteo Frigo Date: Tue Feb 7 21:36:47 2006 -0500 icc supports x86_64 these days. commit a7f132f06de9d343ee68b436c089bd37e6b7fc17 Author: Matteo Frigo Date: Sun Feb 5 18:19:55 2006 -0500 Paranoia. commit 8645d5236a621db86ff7094b0e1a3e2946abc9fc Author: Steven G. Johnson Date: Mon Jan 30 15:27:53 2006 -0500 whoops, fixed assert (y <= x) commit 13864d94f5c655cee3914be9d8751e184f86b8c9 Author: Steven G. Johnson Date: Mon Jan 30 15:26:22 2006 -0500 note that safe_mulmod requires {x,y} < p (or at least < 2p), and added assert commit 849af348d142662e71fc4f3efe2866907e3bc745 Author: Matteo Frigo Date: Mon Jan 30 11:09:32 2006 -0500 fixed aix/xlc lossage commit 106ee57674a134e1f876b6b6b77accd3a3b7a5f8 Author: Matteo Frigo Date: Sun Jan 29 20:42:51 2006 -0500 In the impuse test, normalize the impulse so that the impulse and the random vectors have roughly the same L2 norm. This change reduces the number of bits that we lose because of floating-point cancellation, so that we can focus on the bits that we lose because of bugs. commit 45098b30a5e91b87bb97474de6ff2c16fd7373e3 Author: Matteo Frigo Date: Sun Jan 29 20:37:47 2006 -0500 Compute omega in trigreal precision, as opposed to R. commit 2c5480453a0fd877bdec040eb421b975eb2c63f4 Author: Steven G. Johnson Date: Fri Jan 27 19:16:22 2006 -0500 add --with-combined-threads option as workaround to Windows inability to build shared libs with dependencies commit 68fde0a7351209d643634dfc19367da685c7e455 Author: Steven G. Johnson Date: Fri Jan 27 17:20:45 2006 -0500 libfftw3_threads should *not* used -no-undefined because, in fact, it is not true -- this library depends on -lfftw3, and is not self-contained commit ba85fd54aba2401c937c7acbff52a7c557956f68 Author: Steven G. Johnson Date: Thu Jan 26 22:04:34 2006 -0500 updated commit 42feb604758692ce9936076f37e10c0f4098d46d Author: Matteo Frigo Date: Thu Jan 26 21:10:50 2006 -0500 Added paranoid stack alignment when awaking plans. While I was at it, removed obsolete, redundant AWAKE macro. commit 6b9831ddefcd83bf50aeafd90a6aa1effb44183e Author: Matteo Frigo Date: Thu Jan 26 20:54:39 2006 -0500 Updated for 3.1. commit 4f2fadf55d8ba8d714bc96fb2236dfa981e3d244 Author: Matteo Frigo Date: Thu Jan 26 19:15:12 2006 -0500 ditched one alignment check and noted that we should eliminate the rest as well commit 7c89983f07d925a997e5c293f8cdd5fbe577e3fb Author: Matteo Frigo Date: Wed Jan 25 23:02:19 2006 -0500 alignment hack commit 34f414ddf79840e5a7a9122c98e97bb2a09ecbd7 Author: Matteo Frigo Date: Wed Jan 25 22:05:11 2006 -0500 detect pentium M commit 4f5853890a7ad01e763186bd03e44a5b20e5ef4a Author: Steven G. Johnson Date: Wed Jan 25 18:42:58 2006 -0500 don't trust host_cpu if it claims we are on i386/i486, and call cpuid anyway (if it fails we use no arch flag). This is needed on FreeBSD commit 16caea410e6cc85276555146cf41c370534074b6 Author: Steven G. Johnson Date: Wed Jan 25 18:00:04 2006 -0500 suggest --with-our-malloc16 in error message commit 7574f2ae7e872c4f05f34b73057069b57fb2df01 Author: Steven G. Johnson Date: Tue Jan 24 19:53:34 2006 -0500 ditto for -no-gcc commit 04b5cc720e50ebe4cd2360425e79d6767356288a Author: Steven G. Johnson Date: Tue Jan 24 19:51:08 2006 -0500 flags required for successfull compilation should be added even if the user overrides CFLAGS commit 623ce195f6fe58d67f3bc8928ecc173f753e55db Author: Steven G. Johnson Date: Tue Jan 24 18:43:59 2006 -0500 upcoming gcc OpenMP support uses -fopenmp commit abec9a1443aa235af2e2cfbd86f636599bcfea5e Author: Steven G. Johnson Date: Tue Jan 24 18:26:59 2006 -0500 note that PGI uses -mp as well commit 883f0f18de0f8528fe77129192f521b1a77bfece Author: Matteo Frigo Date: Mon Jan 23 15:31:24 2006 -0500 my best guess at how to fix the microsoft crap du jour commit b457e9e371cf5e13bb818868495b38dce7ccdcd6 Author: Steven G. Johnson Date: Mon Jan 23 14:05:14 2006 -0500 use -Masmkeyword for PGI cycle counter, grr commit 91a65bac96c2fc134d11c8d551eb410de6d18bc2 Author: Matteo Frigo Date: Sun Jan 22 18:09:06 2006 -0500 Bumped version number to 3.1. commit 1745639a638440b17bfcea324c068679f655df24 Author: Matteo Frigo Date: Sat Jan 21 10:03:59 2006 -0500 Report that --enable-k7 is incompatible with --enable-shared. commit 840da056365df79ea63fc3d5a21b1ab5a13707e9 Author: Matteo Frigo Date: Sat Jan 21 09:17:54 2006 -0500 Do not use empty libraries in LIBADD, since otherwise the linker fails on Solaris. commit 4228f20154f81216ab4ddae092d7661bb8af1652 Author: Steven G. Johnson Date: Wed Jan 18 10:47:59 2006 -0500 warn end-users away from this file commit b6e0f0a6eac2561efe417f2cfc0eb8686196a385 Author: Matteo Frigo Date: Tue Jan 17 16:16:42 2006 -0500 Gcc sucks. commit 8560506aa44b2740ea378c83c403373dfce2a662 Author: Matteo Frigo Date: Tue Jan 17 11:48:55 2006 -0500 Disabled checks that may turn out to be too paranoid. commit 782888694f5690298d87cc67cf9963f97aabc412 Author: Matteo Frigo Date: Tue Jan 17 10:35:03 2006 -0500 Some paranoid checks. commit c93e48fa31d081994b9e3b11cca9f1ab25bdf6a1 Author: Matteo Frigo Date: Tue Jan 17 09:31:08 2006 -0500 Flush stdout after printing. commit 8a84f237ca9d96babf1f4edeecb181c47cd74dbe Author: Matteo Frigo Date: Tue Jan 17 08:28:18 2006 -0500 Run the leak detector in all cases, not just when verbose > 2. commit 884a08a129046af3f84ce0fc138f385976f5a5a9 Author: Matteo Frigo Date: Tue Jan 17 08:11:41 2006 -0500 Eliminate calls to pow(), rint(). commit 370ddffe8e4854e4826b1ff4ea14c617d1eea504 Author: Steven G. Johnson Date: Tue Jan 17 00:45:06 2006 -0500 put # in first column, for stylistic consistency commit ede00270785b328279288ada254a11f7314bcd6c Author: Matteo Frigo Date: Tue Jan 17 00:17:27 2006 -0500 Made timeout part of impatience flags, in order to improve the usability of wisdom. Also, fixed bogus error recovery logic in planner.c:imprt(). commit 693f01973548254de258d7efa4217cabea005e79 Author: Steven G. Johnson Date: Mon Jan 16 23:03:34 2006 -0500 make timelimit < 0 .eq. FFTW_NO_TIMELIMIT commit 5af63c169becdefc68db3b4f2df8e788b9867c98 Author: Matteo Frigo Date: Mon Jan 16 21:52:01 2006 -0500 Eliminated the FFTW_TIMELIMIT flag in favor of this simpler logic: fftw_set_timelimit(0) disables time limit. fftw_set_timelimit(X), X>0 sets the time limit to X. commit 8a9d6dd6b442050ad202a6f7154926d145e359b1 Author: Matteo Frigo Date: Mon Jan 16 08:38:04 2006 -0500 Force the use of the estimator when wisdom fails because of md5 collisions, otherwise the planner takes forever. commit 7c6a1a3f7e16df4dca8f78ee994d0488278977fb Author: Matteo Frigo Date: Sun Jan 15 21:30:31 2006 -0500 Ranted about how broken gcc-4 is. commit 383c1374f7af522dfcfe363c508d7fb630e83746 Author: Steven G. Johnson Date: Sun Jan 15 19:59:38 2006 -0500 change fftw_timelimit global var to fftw_set_timelimit(double) function, for simpler usage with shared libraries and for consistency with e.g. set_numthreads commit 584641592c2c273e233b919c8e68e1dbb840d72f Author: Matteo Frigo Date: Sun Jan 15 19:32:27 2006 -0500 Minor tweaks. commit f8fd8093b44aca863601612ae0b4818e91cca853 Author: Matteo Frigo Date: Sun Jan 15 16:32:54 2006 -0500 tweaks to make sure that time_n() is always called from the same stack position. commit 1a5445769d483d86df5d2de6e41f4c5e9515a4f1 Author: Matteo Frigo Date: Sun Jan 15 16:09:53 2006 -0500 Major simplification of the timer calibration logic. Also, use an FFT as a unit of work instead of the old pointer chasing, because God knows how pointer chasing interacts with the idiotic cache-hit speculation on the Pentium IV. commit 1838fc3c1290495355ce10791c9a9f376dd7522d Author: Matteo Frigo Date: Sun Jan 15 15:12:08 2006 -0500 Fixed broken aligment checks when sizeof(R)==12. commit 17b67db5fa051c8eef9c962abfb698b51d11f303 Author: Matteo Frigo Date: Sun Jan 15 10:36:40 2006 -0500 Manual unrolling of loop. commit 7465e7b2ab1c48d06a50189a8545af2b0d98fdda Author: Matteo Frigo Date: Sun Jan 15 10:12:55 2006 -0500 Various improvements to timer calibration routines. commit 049684aa3e3f0411c535a71b0f4adc0ecff7327d Author: Matteo Frigo Date: Sat Jan 14 22:16:09 2006 -0500 cygwin defines __CYGWIN__, not __WIN32__ etc. commit c418027ffb0304f446af729a93415df506d093f3 Author: Matteo Frigo Date: Sat Jan 14 20:40:12 2006 -0500 fixed confusion between libbench and user timers commit 64c27bbd6553e2d8e2e988456890e3e31266b89e Author: Steven G. Johnson Date: Sat Jan 14 12:32:44 2006 -0500 update commit 1b00d512abfe62699264bc556e84a50e89c9a377 Author: Matteo Frigo Date: Sat Jan 14 10:24:11 2006 -0500 Comment. commit 7fa69534cca1e07e3c1260151ed8dee5e5a645cc Author: Matteo Frigo Date: Sat Jan 14 10:19:28 2006 -0500 Workaround gcc bug. commit 686f1af1d0f1c63d99f2891a47c5de2b459b92d2 Author: Matteo Frigo Date: Fri Jan 13 19:13:18 2006 -0500 Switched to -beta2. commit 9f370230780a4f3c03643f6c35f4114fada8c1e7 Author: Matteo Frigo Date: Thu Jan 12 22:21:57 2006 -0500 Fixed technically correct but highly obfuscated use of the enum tag R2HC as a null pointer. commit 38965981e1187e5d0574e129690f3e02b4bc1cae Author: Steven G. Johnson Date: Thu Jan 12 19:25:20 2006 -0500 --enable-unsafe-mulmod is obsolete commit 3f29e7d2e1993de8b7a9759bc879955cb8ae569d Author: Matteo Frigo Date: Thu Jan 12 19:23:18 2006 -0500 More thoughts. commit 84082b78ddfaf5133e49453cc3a62c3d3dde9c04 Author: Matteo Frigo Date: Thu Jan 12 19:17:57 2006 -0500 Removed loop unrolling because it slows things down on at least one powerpc and it generates clumsy x86 code. commit 51caa62b55dcdb8e1aeb9da2d10a40874cef875a Author: Steven G. Johnson Date: Thu Jan 12 19:17:35 2006 -0500 tweaks commit e29d0b0a8ceacbe19adba501d8e5799c7647bb87 Author: Steven G. Johnson Date: Thu Jan 12 15:55:52 2006 -0500 MacOSX x86 ABI specifies that the stack is kept 16-byte aligned commit 487e03a1ee35bc63877b6ec2c2e410da5f3dd4d5 Author: Matteo Frigo Date: Thu Jan 12 12:46:49 2006 -0500 ``ret'' is a reserved word in the evil empire. commit ef109b1d0703dbf67144c7ef5afe9ee4dd0ef489 Author: Matteo Frigo Date: Thu Jan 12 08:31:43 2006 -0500 Changed ret => result because ret ``is a reserved word'' in the evil empire. commit 7dfbcb39afd28daaa10eba6e7909e0e8e3dd56bc Author: Matteo Frigo Date: Wed Jan 11 19:30:42 2006 -0500 Workaround Visual c++ lossage. commit 93876be963fec88768744d04a2c027a4c14f49f0 Author: Matteo Frigo Date: Wed Jan 11 19:26:16 2006 -0500 Workaround visual c++ lossage. commit b2e9544d09112da7db08f07f268e2ed3ad707634 Author: Matteo Frigo Date: Wed Jan 11 19:10:52 2006 -0500 isprint() is guaranteed to work for unsigned char + EOF only. commit 5b926765db935776483660d88b2ce02dca54081e Author: Steven G. Johnson Date: Wed Jan 11 13:47:49 2006 -0500 rm obsolete fixme commit 9237b1a5063d1190e4a8a79d924599a240706756 Author: Steven G. Johnson Date: Wed Jan 11 13:38:46 2006 -0500 [empty commit message] commit 72c1af743cd4da543e142aa9d51b600f47811378 Author: Steven G. Johnson Date: Wed Jan 11 13:32:26 2006 -0500 fix comment commit 47b608a52f08027e1429325bd1639ee4f176aea2 Author: Matteo Frigo Date: Wed Jan 11 12:27:05 2006 -0500 Paranoid use of K(x) for all constants x, to avoid runtime double->float conversions on sufficiently stupid compilers. commit fd9ac529906a8db6d171aa999e4a848b495a8fde Author: Matteo Frigo Date: Tue Jan 10 20:10:38 2006 -0500 Workaround to gcc nonsense. commit 90aaf565346f372e580fb899f9212558ff87d0d2 Author: Steven G. Johnson Date: Tue Jan 10 18:44:28 2006 -0500 bug fix: infinite loop in transpose-cut planning commit 5cd8a9482a90f25c76df01e4f8ea4b2a3386c449 Author: Steven G. Johnson Date: Tue Jan 10 18:12:14 2006 -0500 clarified comment commit 4c5e2af8af80c06734bac116adefdc9e346caa47 Author: Steven G. Johnson Date: Tue Jan 10 18:10:32 2006 -0500 more Windows decorations commit c0bb01fdec0e5c0d9636187641f43690cdb771e2 Author: Steven G. Johnson Date: Tue Jan 10 17:57:45 2006 -0500 added FIXME comment commit 82eb98885049d8d0b69490915a39614c17d5263c Author: Steven G. Johnson Date: Tue Jan 10 17:52:07 2006 -0500 'make clean' should not delete codlist.c since it is included in the dist tarball commit 7fece302306db2d854caf017c680e29eadf79cb7 Author: Matteo Frigo Date: Tue Jan 10 17:50:12 2006 -0500 Change threshold for ``large'' Cooley-Tukey to 256K from 64K, since it seems to benefit the Pentium IV with sse and the planning cost is not too horrible. commit 46c94f013ca0ab45344996479ff3059a6b835241 Author: Steven G. Johnson Date: Tue Jan 10 17:45:11 2006 -0500 more missing Windows DLL decorations commit 67d487e555c8d365aa9530173dba788656f2d91b Author: Steven G. Johnson Date: Tue Jan 10 17:41:28 2006 -0500 remove unused var commit 3356ac92a38eea7582b9712a6cef2067dd9ccf28 Author: Steven G. Johnson Date: Tue Jan 10 14:00:50 2006 -0500 allow compiler threads, if enabled, to take precedence over explicit threads commit d2c3905718e0dbe3bb6e67befc3a2f4c63badbeb Author: Steven G. Johnson Date: Tue Jan 10 12:30:09 2006 -0500 [empty commit message] commit 755b3ecdb747b9b6f1dbc52036cf1d08f724596d Author: Steven G. Johnson Date: Tue Jan 10 12:21:56 2006 -0500 [empty commit message] commit 22db6a0e68da61729239444ff941e84f8de1b336 Author: Matteo Frigo Date: Tue Jan 10 09:13:20 2006 -0500 Fixed comment typo. commit 219609390fc443e6defd5f4940aa36e059b0e6c6 Author: Matteo Frigo Date: Tue Jan 10 08:59:22 2006 -0500 Rearranged timeout checks so as to eliminate one of them. commit 5d22885da57a28d4ce96128650dba99a3ea76481 Author: Matteo Frigo Date: Tue Jan 10 08:56:55 2006 -0500 Converted residual CK() -> A(). commit d82a20e3e3a4d47aebbd2ce4350da4976ba32652 Author: Matteo Frigo Date: Tue Jan 10 08:36:13 2006 -0500 Maintain the invariant TIMED_OUT ==> NEED_TIMEOUT_CHECK. commit 08f674254d16c7770944dc9e7c0eaa3579f333bb Author: Matteo Frigo Date: Tue Jan 10 08:24:41 2006 -0500 silence some 64-bit warnings commit 37aaadd4f3be4c0a5f03fffae1df96e82e8064c9 Author: Matteo Frigo Date: Tue Jan 10 07:58:48 2006 -0500 Assertions. commit 3ee7cd888752144ff48442480446982dcbf3bba3 Author: Steven G. Johnson Date: Tue Jan 10 00:14:00 2006 -0500 some condensing commit 667419d3ac72c2cc43df2d10f704111a40320338 Author: Steven G. Johnson Date: Tue Jan 10 00:03:32 2006 -0500 eliminate X(seconds) in favor of X(elapsed_since), in paranoia of clock wrap commit f696d1fe43a496c64fcf0daaa83060ac70c789fb Author: Steven G. Johnson Date: Mon Jan 9 23:21:21 2006 -0500 [empty commit message] commit 95280e070545a42bbd407c52877e6f8b48d778d6 Author: Steven G. Johnson Date: Mon Jan 9 23:21:06 2006 -0500 hmm, a bit more pessimistic about clock wrapping commit 47d7479fd14d1baf3102a699e72eb7158fede8bd Author: Matteo Frigo Date: Mon Jan 9 23:20:26 2006 -0500 Revert to md5uint = unsigned int whenever possible, so as to avoid wasting space for unsigned long on 64-bit machines. commit 887d8a089ff5b925f88a198bf1b1cc7ddf61392a Author: Steven G. Johnson Date: Mon Jan 9 23:12:27 2006 -0500 note why clock() wrap should not be a concern commit 2991a94ba406fa1f245b62216e463a778f646bbd Author: Steven G. Johnson Date: Mon Jan 9 22:57:16 2006 -0500 bugfix in recent timeout changes - check for case where last solver times out commit 3a0c958aa20c64dab514ec5afe74531e933ac77a Author: Steven G. Johnson Date: Mon Jan 9 22:40:26 2006 -0500 started changes list from beta commit 63922f26968d87550c4fcfd47f41225d056e4977 Author: Matteo Frigo Date: Mon Jan 9 22:34:13 2006 -0500 Paranoia. commit 5bf3d3dd9c6ab2be2e450cd03cabb775bb3db4ca Author: Matteo Frigo Date: Mon Jan 9 22:27:37 2006 -0500 Paranoid assertions. commit 967f0848ad28ffe357a4758d477c5826075a4bac Author: Matteo Frigo Date: Mon Jan 9 22:13:32 2006 -0500 Added FIXME comment stating the 64-bit uncleaniness of fftw_tensor_to_bench_tensor(). commit 0a2228df28268ba4855063849942199ed2c86d31 Author: Matteo Frigo Date: Mon Jan 9 22:06:05 2006 -0500 Another 64-bit bug. commit dc20e0d303713ae5664b91eb4762b4a0f5cf0623 Author: Steven G. Johnson Date: Mon Jan 9 21:54:07 2006 -0500 more Windows DLL nonsense commit d884e3edc5ca24864e92470966ed04aeaccab8f9 Author: Steven G. Johnson Date: Mon Jan 9 21:18:25 2006 -0500 some additional dllexport tags required to build the test program, due to internal stuff called by hook.c commit c0fc6ffb403456e03f5e8dc425182e6607c1cd2a Author: Steven G. Johnson Date: Mon Jan 9 20:31:15 2006 -0500 [empty commit message] commit a85549d03edbe4bee47b0248fac7d76d9cdfeb06 Author: Steven G. Johnson Date: Mon Jan 9 20:30:19 2006 -0500 comment commit 1f72b4d52a499bd63cd52ec259ae2585c6df2b66 Author: Steven G. Johnson Date: Mon Jan 9 20:20:28 2006 -0500 [empty commit message] commit fb0eb0e86fcfffb4c1b5ba17cc520e27914c4ff4 Author: Steven G. Johnson Date: Mon Jan 9 20:16:50 2006 -0500 clarification commit 29de1846aefcd05cb14e0dd286a8374a4a9b18e8 Author: Steven G. Johnson Date: Mon Jan 9 20:12:23 2006 -0500 define FFTW_DLL if DLL_EXPORT (defined by libtool) is supplied commit b99426a4b113d6c8017a6feeb22d89b8a32211f7 Author: Steven G. Johnson Date: Mon Jan 9 20:05:11 2006 -0500 whoops commit 5bb395fcc8798c697d5158b9242d4d914922d9db Author: Steven G. Johnson Date: Mon Jan 9 20:00:47 2006 -0500 another stab at Windows DLL mess commit 9453c5ed5d0c160deb3aef127870b7d65c26d8d1 Author: Matteo Frigo Date: Mon Jan 9 19:23:42 2006 -0500 64-bit clean SIMD header file. I missed those because sparse does not know vector types. Grrr... commit a27e044b39f52abb6066e070b1a3492b6be2e155 Author: Steven G. Johnson Date: Mon Jan 9 19:08:36 2006 -0500 this option is called AC_DISABLE_SHARED in the documentation commit 25cd95982a2acda3a3d6220728768164d6d9c890 Author: Steven G. Johnson Date: Mon Jan 9 17:34:13 2006 -0500 fixed --with-gcc-arch to work when cross-compiling commit c0b9d3122ba267c448b98b0ede12bcf27b9b4e02 Author: Matteo Frigo Date: Mon Jan 9 12:04:04 2006 -0500 Moved the timeout check back into the search loop, sicut erat in principio. This gives us a precise control over the timeout. To avoid the overhead of X(seconds)(), only call X(seconds)() if some time measurement was taken since the last call to X(seconds)(). commit 8c4448e2b69fb02b70e85405bf58a77ec4c13de9 Author: Steven G. Johnson Date: Mon Jan 9 00:07:40 2006 -0500 comments commit 3cd770cab6fac7657b7cd55d6d98f3f516a20fb3 Author: Steven G. Johnson Date: Sun Jan 8 23:58:23 2006 -0500 generalized transpose-cut routine to be able to call transpose-gcd recursivly; TOMS follow-the-cycles algorithm now seems to be completely superseded commit 7ce8a67fabd9ed925a7aee905fa50c658ed2fd20 Author: Steven G. Johnson Date: Sun Jan 8 20:53:18 2006 -0500 [empty commit message] commit ee3cbdc7ad5dd5bfcb2f1f3df6b5ace55f121f32 Author: Steven G. Johnson Date: Sun Jan 8 20:52:16 2006 -0500 ignore errors from setscope -- POSIX standard does not require PTHREAD_SCOPE_SYSTEM to be supported, and PTHREAD_SCOPE_PROCESS is usually okay in that case commit 0ee88684468fc0dae5fd08cc684b8c174d885dd9 Author: Steven G. Johnson Date: Sun Jan 8 15:58:40 2006 -0500 added TODO comment commit 839a6d3192f804e3bc018419b90d18aa82d00292 Author: Steven G. Johnson Date: Sun Jan 8 15:39:28 2006 -0500 whoops commit d3fdf3fcd234dbb12aea0ab1029db2f121356f1e Author: Matteo Frigo Date: Sun Jan 8 14:44:23 2006 -0500 Boasted ``much faster altivec performance''. commit 6d85298a5a4c987cf192ae2df25673a8250d265d Author: Matteo Frigo Date: Sun Jan 8 11:44:52 2006 -0500 Added a new pass to the generator to schedule for the pipeline latency. (This schedule modifies the ``optimal'' cache-oblivious schedule and hence it uses more registers.) This pass is currently: * disabled for non-fma code, under the assumption that this will run on a register-starved fma. * enabled for non-simd fma code, under the assumption that this will run on a processor with 32 or more FP registers. The latency of 4 is conservative and does not introduce too much register pressure. * enabled for simd fma code, under the assumption that this will run on altivec. The latency of 8 seems to produce the best results. commit 1e7e0cd308f67033c681c0ae52836283f874fe51 Author: Steven G. Johnson Date: Sun Jan 8 03:13:53 2006 -0500 fixed estimator for vrank3-transpose commit 13dd2e84c1d331d07eaaef76bb78110dbe941446 Author: Steven G. Johnson Date: Sun Jan 8 02:02:11 2006 -0500 more detail on VC++ workaround commit 699008e51d100801bb19d99b2dbe595b1e33c445 Author: Steven G. Johnson Date: Sun Jan 8 00:19:19 2006 -0500 typo commit 2f842c52fbd7cac9b7564045378e1d649af6dbbf Author: Steven G. Johnson Date: Sun Jan 8 00:16:20 2006 -0500 screw it, just use planner for all sub-transposes in vrank3-transpose (still just use memcpy for contiguous copies, though) commit e6908d5d37b50dff661acfecd3687d1a9fd3300c Author: Steven G. Johnson Date: Sat Jan 7 23:13:45 2006 -0500 add an assert commit 000c5f8a4f6f83143f85268a03709d12ba1e896e Author: Steven G. Johnson Date: Sat Jan 7 21:57:34 2006 -0500 vrank3-transpose now uses planner to decide whether to use cpy2d, cpy2d_tiled, etc. commit f9db072d1270330e0fde90db33e71576d4a2e141 Author: Steven G. Johnson Date: Sat Jan 7 20:57:16 2006 -0500 too annoying to have isqrt unexpectedly fail for n==0 commit 9c8847c394cdd9bdd3d02a127a2497e09bab2d28 Author: Steven G. Johnson Date: Sat Jan 7 17:49:37 2006 -0500 clarifications commit 9fdeaf83ff81ca6931d74b65f8477f95fcfae323 Author: Steven G. Johnson Date: Sat Jan 7 16:39:20 2006 -0500 comment fix commit ad7b11b215b77bad24047e811e8bdaa2ee320edb Author: Steven G. Johnson Date: Sat Jan 7 15:16:22 2006 -0500 more faq updates commit 8bc87da1486f5f1a451cc418a345bb12b95479fc Author: Steven G. Johnson Date: Sat Jan 7 15:12:16 2006 -0500 enable fma on hppa, update FAQ entry commit 479aa905ff4136d48a86ef8ea28e46c06c07ee79 Author: Matteo Frigo Date: Sat Jan 7 14:06:31 2006 -0500 Accomodate different semantics of 'const' in C and C++ commit cf0d153fd10cf6e894520f58c2ce1e6259b683c9 Author: Matteo Frigo Date: Fri Jan 6 23:40:53 2006 -0500 Altivec is called VMX in IBM land. commit a46734a158edbc1e170c0e043d64fb3a320c8d80 Author: Matteo Frigo Date: Fri Jan 6 23:40:16 2006 -0500 Noted faster altivec support. commit 4e7329c580102980a2862964df1474c403d59f9d Author: Steven G. Johnson Date: Fri Jan 6 21:49:10 2006 -0500 updated icc flag detection commit 2de66ca6567360268fa4f1653c787903471a2ab7 Author: Matteo Frigo Date: Fri Jan 6 10:01:50 2006 -0500 Note ``memoize triggen''. commit c19609ea4726f8e842db68cbf15f2ee94abdf33d Author: Matteo Frigo Date: Fri Jan 6 09:36:51 2006 -0500 Use --enable-threads to generate dependencies in the threads/ directory. commit 7538d17a7e277e5f3099b285f85944ee81df6a7c Author: Matteo Frigo Date: Fri Jan 6 09:26:29 2006 -0500 Workaround to icc #defining __GNUC__. commit 3623ea4c4e5649470d360af6c89410b22da9b9ef Author: Matteo Frigo Date: Fri Jan 6 09:21:19 2006 -0500 Switched name to 3.1-beta1. commit 5022d2e2f5e385f82c9b298f958a6935de39233c Author: Matteo Frigo Date: Thu Jan 5 23:08:44 2006 -0500 More thoughts. commit d6262891e97139b27fdb2ca73addf122be568d17 Author: Matteo Frigo Date: Thu Jan 5 22:30:51 2006 -0500 Note wish that (block_size % 4) == 0. commit d6779fe4008a3ff1b5341cc82946e24a6e0cf418 Author: Matteo Frigo Date: Thu Jan 5 22:19:09 2006 -0500 Check alignment of mstart, mcount in SIMD codelets. commit 3d4fc920479d90ecc75a2256c6306c148d2a7bd8 Author: Matteo Frigo Date: Thu Jan 5 21:56:19 2006 -0500 Enable threads at bootstrap time, so I get the compiler warnings that I would otherwise ignore. commit 90f3ef0fb9b081f29eae1e1923e94ea3bb29d7ba Author: Matteo Frigo Date: Thu Jan 5 18:23:15 2006 -0500 made compilable by c++ commit b68d5ed7c28299cf92764bff3ab8b8f06ec1cf00 Author: Matteo Frigo Date: Thu Jan 5 17:39:02 2006 -0500 FIXED: incorrect twiddle_shift() commit b56739cdd0ea335b6ca48c8dd34103316cc43785 Author: Matteo Frigo Date: Thu Jan 5 16:01:51 2006 -0500 Replaced remnants of awake flag with the new enum wakefulness type. commit 8871d572d270aa76dea86073fc11362c6d516c9a Author: Matteo Frigo Date: Thu Jan 5 11:20:59 2006 -0500 Oops---there is no need to find a free slot. commit fedf131be6c553e13212c16f7a8f474a0e61fed6 Author: Matteo Frigo Date: Thu Jan 5 09:41:58 2006 -0500 Assertions. commit ff66bb4a211ea2640f833ae48bedb1b34a0b47f2 Author: Matteo Frigo Date: Thu Jan 5 09:29:55 2006 -0500 Commented the hash table lookup algorithm. commit 4bafb30ddfc85ff74bb758a23532ce60bb621d19 Author: Matteo Frigo Date: Thu Jan 5 09:12:00 2006 -0500 Fixed infinite loop in hashtable lookup/insert. Grrr... commit 02a5374038e878b9e0cfe88ee88b0389bf20a255 Author: Steven G. Johnson Date: Wed Jan 4 22:04:28 2006 -0500 updated copyright years to 2006 commit b2d48f50aa87d2b9e5f57c6c04959b7ce0984732 Author: Steven G. Johnson Date: Wed Jan 4 21:57:23 2006 -0500 whoops commit 24baeff279c41dbe00c5fd1b13844175e8f70cfe Author: Steven G. Johnson Date: Wed Jan 4 21:52:18 2006 -0500 whoops commit 490a044a9e2b2f599506ef415c3f87c2b64ba83a Author: Steven G. Johnson Date: Wed Jan 4 21:51:40 2006 -0500 more updates for recent pentia/amd commit 21fc6cf5d45450edd194c6d83d328dd7c27c8142 Author: Matteo Frigo Date: Wed Jan 4 20:57:47 2006 -0500 Pruned TODO. commit 96c862a6929365a5a78a2196cd72c5037082c5d8 Author: Matteo Frigo Date: Wed Jan 4 20:43:41 2006 -0500 Prototype of problem_destroy() commit 700b7dcd5331fe4317b214d64086771a404814ef Author: Steven G. Johnson Date: Wed Jan 4 20:43:13 2006 -0500 rm obsoleted TODOs commit f722e923cd823d4501bc8c3a730fbc09d2c26e06 Author: Matteo Frigo Date: Wed Jan 4 20:37:24 2006 -0500 Fallback to 970 if neither -mcpu=power5 nor -mcpu=power4 are supported. commit b5823feffb1b189d536e5c562959969c247a61c3 Author: Steven G. Johnson Date: Wed Jan 4 20:29:07 2006 -0500 NEWS updates, clarifications, and reorganization commit fffa543ce9d6cb43d2c09bf401c029b5f6830356 Author: Steven G. Johnson Date: Wed Jan 4 19:54:41 2006 -0500 remove some compiler warnings, add an assert check, make estimator work properly for nop plans commit 3c4889a04995ac9f01ffdb3c4dd0ddc4ef42dc53 Author: Matteo Frigo Date: Tue Jan 3 19:34:04 2006 -0500 Two big changes: 1) revised the twiddle generation machinery, to avoid generating twiddles when measuring, and to use a faster O(sqrt(N)) table when this entails no loss of precision. 2) implemented new ALLOW_PRUNING estimator hack. commit 30e3e40e0439f7109a75c063ebb0544bbe68a0c7 Author: Matteo Frigo Date: Sat Dec 24 22:08:29 2005 -0500 Estimator tweaks, mostly to favor generic over rader for small n. commit 2e0e06d43cef1259a6fdda21744c8fa71960ea69 Author: Matteo Frigo Date: Sat Dec 24 17:55:47 2005 -0500 Grrr... missing break statement in switch. commit 12348cb25f94416b730862ea4d0a5e85eb2c98b2 Author: Matteo Frigo Date: Sat Dec 24 16:08:50 2005 -0500 Swapped fields TW and OPS in struct ct_desc_s, to make k7 asm code insensitive to -malign-double. For consistency, changed struct hc2hc_desc_s in the same way. commit 33a820de9270d537b4079f08fe258a969c410632 Author: Matteo Frigo Date: Sat Dec 24 16:00:42 2005 -0500 Wrong check for infeasible slvndx in imprt(). commit 4b5008a48fbfaf95504f2816b980f971d6678326 Author: Matteo Frigo Date: Sat Dec 24 15:56:59 2005 -0500 Removed obsolete function invoke_solver_if_correct_kind(). commit e1959cade352dd407f5c1c87cf37580ef60f6eb3 Author: Matteo Frigo Date: Sat Dec 24 14:22:12 2005 -0500 Faster implementation of safe_mulmod(), avoiding divisions altogether. Works for 0 <= p <= INT_MAX. commit f827b89e687419b19b7133b64651c3a2f10de064 Author: Matteo Frigo Date: Sat Dec 24 12:05:54 2005 -0500 FFTW_ALLOW_LARGE_GENERIC must belong to flags->l, it cannot be overridden by fftw. commit 5dbe4dcaa75797cb76e09e4349b526993fb435b2 Author: Steven G. Johnson Date: Fri Dec 23 20:46:24 2005 -0500 no more need for limits.h, add some explanatory comments commit 1dba2396d5d50261e6c82014e279b4ac035120f2 Author: Matteo Frigo Date: Fri Dec 23 17:50:25 2005 -0500 Paranoia. commit a09014d7cc40be154096f5b14b0b136985ac39fb Author: Matteo Frigo Date: Fri Dec 23 17:40:41 2005 -0500 Fixed subtle bug involving overflow of the slvndx field in flags_t. commit 1a5304605e6f104eb147f96a5bc76dad55ad9dbf Author: Matteo Frigo Date: Fri Dec 23 16:33:56 2005 -0500 Note 64-bit clean. commit 7d6e177477acee44216776a7afff2306b58eb963 Author: Matteo Frigo Date: Fri Dec 23 15:34:32 2005 -0500 Threads are now 64-bit clean commit 208ba330fb9eaaa58a138350dc9f9e965b95bd2c Author: Matteo Frigo Date: Fri Dec 23 13:00:31 2005 -0500 Restored the old numbering TW_NEXT=3 etc, because the k7 code depends on it. commit 5a7e2e7cbedf9021d8b278afdd9762f3fe0cc697 Author: Matteo Frigo Date: Fri Dec 23 11:58:00 2005 -0500 Portable implementation of MULMOD() and safe_mulmod(). Removed all unnecessary AC_CHECK_SIZEOF() from configure.ac. commit e515294ed8f991b8efb4dc7a0891c16562783679 Author: Matteo Frigo Date: Thu Dec 22 11:12:29 2005 -0500 Inline the loop body in r2r codelets like we do everywhere else. commit 94210bafc8387499f631cdd6187ab293943261a2 Author: Matteo Frigo Date: Thu Dec 22 10:48:53 2005 -0500 Oops. commit 2dcf5d5b1908062b236d6aa2fba93b28937e9488 Author: Matteo Frigo Date: Thu Dec 22 10:25:15 2005 -0500 Renamed X(sin_and_cos)() to X(cexp)(). commit de2f6ff5df500a8d15c1cb36f620d277994ec098 Author: Matteo Frigo Date: Wed Dec 21 22:49:58 2005 -0500 Somewhat faster generation of twiddle factors. commit 2bda3ba8833c53949694b05f2518b57b2cda80a3 Author: Matteo Frigo Date: Tue Dec 20 23:50:01 2005 -0500 tweaks commit 86c8779bcf89bca6fad1812b716a0171b7ab0f91 Author: Matteo Frigo Date: Tue Dec 20 22:29:19 2005 -0500 Sped up planner, esp. in estimate mode. The planner now classifies all solvers into DFT, RDFT, and RDFT2, and it only invokes solvers appropriate for the problem being planned. Because we have several hundred solvers, the overhead of calling irrelevant solvers is significant, and this modification mitigates the issue somewhat. commit 98ea24afbd44d88617f25cd467def39b934cbed5 Author: Matteo Frigo Date: Mon Dec 19 22:04:00 2005 -0500 Eliminated all calls to sprintf() in favor of own routines, so as not to force users to link stdio and the associated locale/pthreads crap. commit 112a5e19c813a918315e26a80ed9e1f427aa59c3 Author: Matteo Frigo Date: Mon Dec 19 21:27:25 2005 -0500 Implemented routine to print INT, removing the need for c99's %td format. commit 5c20f07423e4661b32498afa8071e1f6dacd47c7 Author: Matteo Frigo Date: Mon Dec 19 12:06:33 2005 -0500 info->n is size_t commit 6ae75f3b9b700352da7e3ad728d49d988f80e864 Author: Matteo Frigo Date: Sun Dec 18 18:15:04 2005 -0500 Explicit casts in front of pointer difference in printf() context, just in case INT != ptrdiff_t. commit 25abe60b6b82d9cab328fbfc8dc17f33ffd6803a Author: Matteo Frigo Date: Sun Dec 18 16:52:38 2005 -0500 Forgot to add %D to print.c commit 7e07750df2164e8f8c88185b8857c527f145b444 Author: Matteo Frigo Date: Sun Dec 18 16:43:26 2005 -0500 Use %D as format character for type INT. commit 1bf67aff56a4e6b2f0fc41cb8b66e9b09d4b2ea0 Author: Matteo Frigo Date: Sun Dec 18 15:14:03 2005 -0500 Changed type of an_int_guaranteed_to_be_zero. Changed name as well. commit 602b07fee7f1fbb86b429e682fbce4a4f886e0d1 Author: Matteo Frigo Date: Sun Dec 18 14:41:31 2005 -0500 converted %o -> INT commit e99c67870f4d09190598610fc7c1bd5df8e4515e Author: Matteo Frigo Date: Sat Dec 17 20:28:50 2005 -0500 Major 64-bit cleanup. commit 3cd29a6839b31e093a5c715d6deb2867eafb1b15 Author: Steven G. Johnson Date: Wed Dec 7 22:39:01 2005 -0500 PGI x86-64 cycle counter, courtesy Cristiano Calonaci commit 7b830d38cb785513bde604f14a3253e171a75e0c Author: Matteo Frigo Date: Mon Dec 5 21:25:57 2005 -0500 Must insert into hash table when wisdom_state == WISDOM_ONLY, otherwise wisdom does not work. commit 9cfa064f6635afd41f01788e5a16a7a56babfca0 Author: Steven G. Johnson Date: Sat Oct 8 18:08:44 2005 -0400 comment commit 7fd8f4a4ff768b59317a318d3d83ac0726609868 Author: Matteo Frigo Date: Sun Oct 2 11:49:13 2005 -0400 Paranoia: made planner robust against MD5 collisions. commit 55004ef918346e933b7d46aa529fc76258c0b673 Author: Matteo Frigo Date: Tue Sep 27 22:33:18 2005 -0400 Note that --enable-3dnow is unsupported. commit 317d36cb4265710fe5ccbf3518f15f7f24c076cb Author: Matteo Frigo Date: Tue Sep 27 22:31:04 2005 -0400 * Removed --enable-3dnow support. * SIMD support for split complex arrays. commit 2f87ee31a3c1a416b983aee2ad2441b0624f6839 Author: Matteo Frigo Date: Tue Sep 27 22:28:41 2005 -0400 Removed --enabled-3dnow, since it is becoming useless as the world moves to x86-64, and it is a pain to maintain. (We should probably remove the k7 stuff as well.) commit e5a5da39405e5960f93478937fea04c98feabf49 Author: Matteo Frigo Date: Tue Sep 27 21:59:16 2005 -0400 Missing BEGIN_SIMD(), END_SIMD() statements. commit 7898dae11c979e9b069616b3d922b09b23b8750f Author: Matteo Frigo Date: Tue Sep 27 12:16:08 2005 -0400 Tweaks commit 3bc850803f4f000f1c979a3576bdd066c37eaafe Author: Matteo Frigo Date: Tue Sep 27 10:04:32 2005 -0400 Fixed wrong opcount for simd codelets. commit 2c35b6d0d3217976f3597d04403cfac7a4f7da57 Author: Matteo Frigo Date: Tue Sep 27 09:25:50 2005 -0400 Fixed wrong opcount for simd codelets. commit 27aa07803ba692bbdbc563607e6531222bb56488 Author: Matteo Frigo Date: Mon Sep 26 22:58:19 2005 -0400 fixed flop counts commit 97b8e6bc0d2daddf10da0eb41c94e8e8c4e92bf1 Author: Matteo Frigo Date: Mon Sep 26 22:34:40 2005 -0400 Silence warnings commit 804b1a4d34edaba87c4aa0f6f7fe3f173bb926f8 Author: Matteo Frigo Date: Mon Sep 26 20:52:36 2005 -0400 Implemented split-complex SIMD codelets commit 4c34b9513f4003ec04ebc836dd009d15d4f913f1 Author: Matteo Frigo Date: Sun Sep 25 22:25:35 2005 -0400 Generalized the ``store pairs'' trick (now called ``store multiple''). commit c8eb4f532fe1b280cd93313eab57b1e51cd6d4cf Author: Matteo Frigo Date: Sun Sep 25 18:58:20 2005 -0400 Silence some warnings. commit 7ecbbeacf952a07cbc1a338fa9bdc9612d99b7bb Author: Matteo Frigo Date: Sat Sep 24 12:37:16 2005 -0400 Removed obsolete cruft commit 9a8a94ca3fb2d0ee33268ae8527f65260631d958 Author: Matteo Frigo Date: Mon Sep 19 22:55:19 2005 -0400 Re-enabled check for because OSX requires it. commit 2525a542b0277af07f89f45a3e68c2ac022d4189 Author: Matteo Frigo Date: Sun Sep 11 11:03:03 2005 -0400 Check for sizeof(unsigned int) unconditionally, because the result is used by ifftw.h. commit 5750c658cabc6d64ab0f9817312b2399d75f4041 Author: Matteo Frigo Date: Sun Sep 11 10:59:40 2005 -0400 Higher size limit for t2 codelets. commit c5134ff6de3bfe5306428398c14cb7dcc9a09afe Author: Matteo Frigo Date: Sun Sep 11 10:50:37 2005 -0400 Heuristic: do not use t2 simd codelets for N>1024. commit 8c4b74a02763d61fd64c98f01fd2658bf80fbc68 Author: Matteo Frigo Date: Mon Sep 5 22:22:50 2005 -0400 Larger tolerance in timer calibration routine. commit ed07b941c3be22c7f19c569bd29230c683783b47 Author: Matteo Frigo Date: Mon Sep 5 16:03:33 2005 -0400 #include unconditionally. (There is no point in checking.) commit f03e0aced4c470b2b24d8d5abb94be526833a2b1 Author: Matteo Frigo Date: Mon Sep 5 15:23:27 2005 -0400 Removed SSE and SSE2 asm because it was bitrotting. Use the Intel API instead, which seems to be supported by gcc >= 3.3. Moved files that require -msse, -msse2 to new directory. commit a12a85c774d25cb85391f200a8e6d62da2572cce Author: Matteo Frigo Date: Mon Sep 5 12:56:28 2005 -0400 Parse cputypes of the form 7447A,altivecsupported commit 1d5a7d722689e83fdcccae9edae36ec276b68241 Author: Matteo Frigo Date: Mon Sep 5 12:52:30 2005 -0400 Distinguish powerpc 7400 from the 7450, which has a different pipeline. commit b363c2bb7fe126fe80afcd974a463349e63a48a6 Author: Matteo Frigo Date: Mon Sep 5 12:46:00 2005 -0400 Paranoia: define RIGHT_CPU unconditionally. commit 558789684b3fa4435a4fab4d86769f2a5ee53b57 Author: Matteo Frigo Date: Thu Aug 11 20:56:41 2005 -0400 Removed obsolete name fftw-wisdom2c. commit d73fb7f9d84bc1acccdf9c8f7f2b71e10b3d7854 Author: Matteo Frigo Date: Thu Aug 11 20:55:59 2005 -0400 Avoid creation of temporary files---use cpp magic instead. This fix solves a security bug and avoids nonportable tempfile creation hacks. commit a74941c286a12d9a008c3b89ba558cfab82587af Author: Matteo Frigo Date: Fri Aug 5 10:03:02 2005 -0400 Workaround for with gcc-3.3 altivec bug. commit 259f7d688fec2615a29b1aeb22321568cdcc4bc4 Author: Steven G. Johnson Date: Wed Jun 15 21:36:46 2005 -0400 solaris fix: check -pthreads first since gcc does not like -pthread but chokes due to stubbed libc (grr) commit 261b7c0fcfaa8c8e6a34d06b051c4355bcac60b1 Author: Steven G. Johnson Date: Fri Jun 3 17:19:56 2005 -0400 note that VC++ bug was fixed in 2005 commit 14832d8b25d4091667d3f0e5c8fd8fa1c14f8ce1 Author: Steven G. Johnson Date: Mon May 30 16:30:45 2005 -0400 generalized ax_cc_vendor to ax_compiler_vendor commit b13949fd1df86e14fcd73495557bea7532b49b8c Author: Steven G. Johnson Date: Mon May 30 15:55:07 2005 -0400 updated message commit ead701adfc138233d26e86258f0daa8041a41d37 Author: Steven G. Johnson Date: Mon May 30 15:45:14 2005 -0400 update for new AC archive format commit 56c34ca4db1ff26982040ff00e1cb549653ab720 Author: Steven G. Johnson Date: Mon May 23 23:12:22 2005 -0400 [empty commit message] commit c04871b2f43fe56cd9e921b4864a26ad354cf3f5 Author: Steven G. Johnson Date: Mon May 23 18:17:38 2005 -0400 [empty commit message] commit c4afbfd4ef5235b1b88715bac592b8f091d76d13 Author: Steven G. Johnson Date: Mon May 23 18:13:08 2005 -0400 more notes commit 1cf10c2f758f89da2c0f8bd68f0a8c974e93f33c Author: Steven G. Johnson Date: Sun May 22 23:37:08 2005 -0400 whoops commit 568dac7da89c3fe5dbab61ff28e2aa6dc52ca71f Author: Steven G. Johnson Date: Sun May 22 22:37:50 2005 -0400 note icc 8.x annoyance commit 1b1f5c242db3f55c2dfadb248a9fb292981c5e6b Author: Steven G. Johnson Date: Sun May 22 22:36:04 2005 -0400 [empty commit message] commit f66bc7b513029ac91ec983bb3279f3c0dec3468c Author: Steven G. Johnson Date: Sun May 22 22:35:34 2005 -0400 note gcc 3.4.[0123] bug, which is fixed in gcc 3.4.4 commit 0f2a7eb61a2bcf44583bd41245ad55c7e78eb70f Author: Steven G. Johnson Date: Sun May 22 22:21:26 2005 -0400 added automatic detection of icc architecture flag commit 7b90a23bc9ceeeb03131b4774aa0ff5d04e91c63 Author: Steven G. Johnson Date: Sun May 22 21:47:19 2005 -0400 add -no-gcc to icc flags...even if it is Intel's fault, I'm sick of dealing with bug reports about this commit ff0439a0bc1dc149d302630cb96062fc7fb053f1 Author: Steven G. Johnson Date: Sun May 22 21:40:59 2005 -0400 added @cindex portability commit e18637fa933a8a75ef831024c4c966d6a2dff76b Author: Steven G. Johnson Date: Sun May 22 21:34:10 2005 -0400 note --without-gcc-arch commit 7131ee53a750ff084f05b97c67e34a39e1a7011c Author: Steven G. Johnson Date: Sun May 22 20:54:54 2005 -0400 bsd ppc detection; some odd 603 types commit 7f439b2ab6289af0e08134c659480f9589b13387 Author: Steven G. Johnson Date: Sun May 22 11:53:20 2005 -0400 [empty commit message] commit 32419ec5a48e285cbcbee2f0a4c49e628fcf6ccb Author: Steven G. Johnson Date: Sat May 21 20:34:52 2005 -0400 ensure no spaces in cputype commit 7a6288d8a7617720cb8c46fc9152a31c7dab793a Author: Steven G. Johnson Date: Sat May 21 20:31:41 2005 -0400 nevermind commit b9bac647b7039e381615e0faac27fc3a8de06eb4 Author: Steven G. Johnson Date: Sat May 21 20:30:08 2005 -0400 more bsd stuff commit f1c985e46f8c17122e47ece0e9696258638be1f1 Author: Steven G. Johnson Date: Sat May 21 20:28:40 2005 -0400 added BSD cpu detection for SPARC and better super/hypersparc detection commit e35c028649be9cc1568401e9e39eb2e19d1cda3b Author: Steven G. Johnson Date: Sat May 21 20:22:11 2005 -0400 comment commit a0582b1056c2562cd639c18f2827fc124dd79fa6 Author: Steven G. Johnson Date: Fri May 20 19:40:09 2005 -0400 "alternate" == "alternative" is US-centric commit 333d9eb5086ed1afa77719e9f24142a8bd5dada9 Author: Steven G. Johnson Date: Fri May 20 19:36:26 2005 -0400 typo commit e2d0b93f5de6abb830a0d28324399d4689850b09 Author: Steven G. Johnson Date: Fri May 20 01:28:34 2005 -0400 clarification commit f8a4a4af8c47ae8e572e5f169c0eeb0720eb7473 Author: Steven G. Johnson Date: Tue May 17 18:56:46 2005 -0400 print out estimate-planner time from can_do in verbose>2 mode commit e1bbc2ce6ff2b094ad3549a5140d6acd0218b7d8 Author: Steven G. Johnson Date: Mon May 9 00:47:19 2005 -0400 comment commit 2e2b68117557549932c89d24586be1852a189462 Author: Steven G. Johnson Date: Thu May 5 23:47:55 2005 -0400 fixes for building Windows DLLs with Cygwin; thanks in part to Stephane Fillod commit bb8fc9fb4dda639b9f0b1f13ef448e39d71a4b39 Author: Steven G. Johnson Date: Fri Apr 22 19:47:43 2005 -0400 -ffast-math seems to produce code that is either about the same speed or slightly faster (gcc 3.3 and 4.0, x86) commit 2f7b1f2707810c171bb85b330c99a94196a257d0 Author: Steven G. Johnson Date: Fri Apr 22 19:18:23 2005 -0400 power5 fallback to power4 sched for older gcc's commit 169cba437dfb6f553bb1a8e2a404ca2bf74a5b56 Author: Steven G. Johnson Date: Fri Apr 22 19:14:53 2005 -0400 check for power5 commit 1978d7cd087b7e6e93133c7b4aa2c612f664203d Author: Matteo Frigo Date: Tue Apr 19 21:55:13 2005 -0400 Removed clause #3 commit 3c385073178a321cc4108d4b88f121276b5d0020 Author: Steven G. Johnson Date: Tue Apr 19 21:44:57 2005 -0400 license clarification commit ab865d9025afbb6c923e94956c3e7ebdd64ef75d Author: Matteo Frigo Date: Tue Apr 19 21:42:51 2005 -0400 Changed license of fftw3.h to X11. commit d851f36c4ff5e1febbc2ed47cb08eba3f8dbaf19 Author: Steven G. Johnson Date: Mon Apr 11 13:15:12 2005 -0400 delete fixed-input code commit cc673385bfc98894c37272241fcb6135756d2c14 Author: Matteo Frigo Date: Sun Apr 10 16:33:24 2005 -0400 joned L-U-planner branch commit d4b2b38d4a6b40919a6229bb574ecd49884ad58f Author: Steven G. Johnson Date: Thu Apr 7 23:15:02 2005 -0400 ref commit 8895af84fb9e4970420b21451977fde49072c2b9 Author: Steven G. Johnson Date: Thu Apr 7 00:11:13 2005 -0400 whoops commit 6dbfe38e27a7f4a5090917f8b53a03e334a40881 Author: Steven G. Johnson Date: Wed Apr 6 22:06:21 2005 -0400 added (optional) new split-radix algorithm, enabled with -newsplit; also new -standalone option to omit desc; also -unitary, -normalization, and -normsqr options to generate r2r codelets with various normalization (to match lit. in DCT-II, use: -unitary -normsqr 2) commit 5e1deadac7dbe4d60d493b86f66b37474388b11e Author: Matteo Frigo Date: Fri Mar 25 08:59:43 2005 -0500 Moved timeout check outside the search loop, because X(seconds) is expensive. commit 094cbe955f1ad43c143f7781eb524ede71d164bc Author: Matteo Frigo Date: Sun Mar 20 18:35:53 2005 -0500 Enable vector recursion for in-place problems, otherwise dftw-genericbuf works only in PATIENT mode. commit 14a9b596a784705637abb9cd5a47595ed2a4bcbd Author: Matteo Frigo Date: Sun Mar 20 17:53:58 2005 -0500 oops commit 7ea889cca28101323df5287b988ee6bd96c531a0 Author: Matteo Frigo Date: Sun Mar 20 17:49:13 2005 -0500 make solver UGLY for small N commit a4abcfa708787e3e18b32fc37506992215578c4b Author: Matteo Frigo Date: Sun Mar 20 17:16:37 2005 -0500 new dftw-genericbuf solver commit 70997fbe34952f59b14245e68e5fd4614d13c3ac Author: Matteo Frigo Date: Sun Mar 20 16:12:44 2005 -0500 new dftw-genericbuf solver commit 3d40d10cca6f0fb8ed0e327ae23d569829a43768 Author: Matteo Frigo Date: Thu Mar 17 21:48:19 2005 -0500 Hmm... what was I thinking? commit b27eff441bd1e24148569ed9ee02c05c08b46ea4 Author: Matteo Frigo Date: Thu Mar 17 19:20:54 2005 -0500 Workaround for a MSVC bug. commit 433960d78aef7dc12c5611baa3213b4db99f99cc Author: Matteo Frigo Date: Thu Mar 17 08:18:39 2005 -0500 Workaround for a MSVC bug that was reported by Eddie Yee. commit 0c4f3dfe86c936003eed705208a100c11a5bcce6 Author: Matteo Frigo Date: Tue Mar 15 13:25:53 2005 -0500 try both contiguous input and contiguous output when in doubt commit 155f07c46c6589d374f886a8ed86f985a64642e3 Author: Matteo Frigo Date: Tue Mar 15 08:44:41 2005 -0500 Added genfft flag -precompute-twiddles which moves the computation of the twiddle factors before the main schedule. This flag produces smaller code everywhere, and slightly faster code on powerpc. I observe no speed difference on x86. commit 5cc6165f9756f2faeab137eed5f8c25ebac08773 Author: Steven G. Johnson Date: Mon Mar 14 21:43:53 2005 -0500 sp commit 255c6db9915f31c3b323cee61a7900999c7b4cfe Author: Steven G. Johnson Date: Mon Mar 14 21:43:05 2005 -0500 whoops, spelling error (thanks to Steve Eddins for bug report) commit 556965536b7671795bc6e4ef86edfffe75b2ffd9 Author: Matteo Frigo Date: Sat Mar 12 15:03:47 2005 -0500 Do not approximate pcost = vl * child->pcost unless child is guaranteed not to be a simple codelet. commit a5282a50ce6211585a443fa099e2fa6e47450ceb Author: Matteo Frigo Date: Wed Mar 9 20:00:02 2005 -0500 Relaxed applicability conditions. commit 2496640b61c0ac594325d4fa68e3729873c004bd Author: Matteo Frigo Date: Wed Mar 9 00:05:47 2005 -0500 Minor optimization commit 81c49148f9fb58b0c541b2636b37bd8422a458b1 Author: Matteo Frigo Date: Tue Mar 8 22:14:02 2005 -0500 Interpret K to mean *1024. Similarly for M. commit b94f2eb04282f6c3c511944e3767a8895a19ef77 Author: Matteo Frigo Date: Tue Mar 8 20:44:25 2005 -0500 Hmm... somehow some previous commit got lost. commit 55b8abdbbc3a2bbb26f005735bd9d121634c4055 Author: Matteo Frigo Date: Tue Mar 8 20:30:42 2005 -0500 Paranoia commit 752db4c71fd1a447d9ed1699ed0382e042d4f89c Author: Steven G. Johnson Date: Mon Mar 7 14:30:01 2005 -0500 whoops commit 6c18ecea25e2a9f685131b49c7365fc35b8c4c7c Author: Steven G. Johnson Date: Mon Mar 7 14:29:43 2005 -0500 move fftw-specific HP/UX tweak into configure.ac commit 3916e3b25257834172ce4eb126a2d745b8943123 Author: Steven G. Johnson Date: Mon Mar 7 14:19:24 2005 -0500 ax_cc_family -> ax_cc_vendor (vendor names are easier to remember), add checks for many new compilers, use in ax_cc_maxopt commit a0ad3ef6add8118e82611c08b4c252ec8346efea Author: Matteo Frigo Date: Sun Mar 6 21:36:05 2005 -0500 Count FMA as one flop in estimator when HAVE_FMA commit 10a57b3a5a428bab777ec22f4eb83203498a743a Author: Matteo Frigo Date: Sun Mar 6 19:16:06 2005 -0500 Do not try radix-2 generic. commit e38ef2e30e6f41fb2301acf208ff7f9b775de0ac Author: Matteo Frigo Date: Sun Mar 6 13:04:23 2005 -0500 Use -O3 for xlc now that we use -O for CODELET_OPTIM commit be3c47c96bcc1ef146a296202b53db7a457b3230 Author: Matteo Frigo Date: Sun Mar 6 13:02:41 2005 -0500 New AX_CC_FAMILY macro, that detects the compiler based on symbols that it defines (as opposed to the name of the compiler). We need to start use this strategy everywhere else. commit 562882d5c889b0bce256013a056ce07f55c27dfb Author: Matteo Frigo Date: Sun Mar 6 11:33:15 2005 -0500 Runtime checks to guarantee small strides. commit 1fcf24126783752b3ab8f35f480a0e5d0fa90aab Author: Matteo Frigo Date: Sat Mar 5 20:09:25 2005 -0500 Reduced the search space for rank-0 transforms commit 77cbffe7c30bbac4d294cd2c7321163054732418 Author: Steven G. Johnson Date: Fri Mar 4 17:50:29 2005 -0500 little assert commit 495b9d7617c0167346817c4d5620fe80ee2d1194 Author: Matteo Frigo Date: Tue Mar 1 09:19:16 2005 -0500 Implemented directbuf, enabled for now. commit 1869b027f29cef23f101026dee512744fba87eaa Author: Matteo Frigo Date: Mon Feb 28 22:21:14 2005 -0500 Unified dftw-direct, dftw-directbuf in an attempt to tame code growth commit 82fce69cd912d2a58b86a5699c04d2eea3b9a536 Author: Steven G. Johnson Date: Sun Feb 27 13:51:24 2005 -0500 fixed copyright commit 7d1a5530230d76d105f3ed4aeebdf4f708ed0e8a Author: Matteo Frigo Date: Sat Feb 26 22:21:03 2005 -0500 silence warnings commit 753ab3b636f099eedb841e643898aed3e8c5c817 Author: Matteo Frigo Date: Sat Feb 26 22:19:16 2005 -0500 oops commit a64fecb2ccd2670c6b37d40d70558d553e4cb17d Author: Matteo Frigo Date: Sat Feb 26 21:28:39 2005 -0500 Tweaking while thinking about a higher-rank transposer (bitreverser) commit 9c7a7d3c45be7ca132fdece876ebea7eb053fad7 Author: Matteo Frigo Date: Sat Feb 26 20:06:49 2005 -0500 Transposed the buffer, and skewed it. This allows for contiguous copy operations, and the codelet should not incur associativity conflicts if the buffer is large. commit 521fa92ebcf99b32b35cb4c26b304f42a2812e22 Author: Steven G. Johnson Date: Sat Feb 26 18:14:11 2005 -0500 make tensor_max_index more reasonable (take maximum of input and output max indices, computed separately) commit c6c2bcbb2b8c8f3b1da7d5465e4bee93905c8d32 Author: Matteo Frigo Date: Sat Feb 26 10:04:30 2005 -0500 Use cpy2d instead of cpy2d_tiled, because vl may be too large. commit 269e71f3db6c3d1bcf8dc77e25983dcc9989d5f7 Author: Matteo Frigo Date: Sat Feb 26 00:31:52 2005 -0500 Fixed old bug that was introduced with yesterday's changes. commit e769a1735dd71165677025498471db8a41271198 Author: Matteo Frigo Date: Fri Feb 25 21:54:23 2005 -0500 ``Interesting'' switch statement. commit 7e729390b41355c7abf6c2a3901dec6cb40c4c23 Author: Matteo Frigo Date: Fri Feb 25 12:29:54 2005 -0500 Disabled -reorder-loads -reorder-stores, since they seem to do nothing. commit 4350026ea3252e1dbc25b1539941ee79b3cb6124 Author: Steven G. Johnson Date: Fri Feb 25 12:19:10 2005 -0500 Because of the recent changes to kernel/pickdim.c, splitrnk=0 is no longer equivalent to splitrnk=1 for rnk < 4, where the latter is the FFTW2 behavior. For small rnk, however, I observe the planner to pretty consistently choose the FFTW2 behavior (splitrnk=1), despite its not being asymptotically optimal in the cache oblivious sense. So, make splitrnk=1 instead of splitrnk=0 the default in FFTW_MEASURE and FFTW_ESTIMATE modes (rnk > 3 is pretty rare in practice anyway). commit 3bfeb642d11098a707ca70b7332077b6472917d6 Author: Steven G. Johnson Date: Fri Feb 25 00:33:27 2005 -0500 tweak commit 24560b26faac0a352c23e15c892c38a762bbb453 Author: Steven G. Johnson Date: Fri Feb 25 00:29:09 2005 -0500 slight relaxation commit cadf7b9d5561d14d8042ad3b051f7f95a010cb1f Author: Steven G. Johnson Date: Fri Feb 25 00:21:00 2005 -0500 cruft commit 42d46a1c8af18b951c978ee2cf1cc57ca106929f Author: Steven G. Johnson Date: Fri Feb 25 00:03:14 2005 -0500 added experimental indirect-transpose solver: when transforming the columns of the matrix, allow us to do a transpose to make the DFTs contiguous commit eec7f69ff78e1b95f1bdd09a2f96b3be5cf1b407 Author: Steven G. Johnson Date: Thu Feb 24 23:04:58 2005 -0500 check for abort() commit e1d0f900a4e4444b4ef0fa230de11da87a48a192 Author: Steven G. Johnson Date: Thu Feb 24 23:04:43 2005 -0500 call abort() on failed assertion commit 4d8aee345fa2da4b2383722a482d245d38288dad Author: Matteo Frigo Date: Thu Feb 24 21:17:23 2005 -0500 Forgot to change X(isqrt) -> isqrt_maybe commit 47e79fca2a795dcd96ecf59852cdc53bc883f9d1 Author: Steven G. Johnson Date: Thu Feb 24 20:18:59 2005 -0500 require finite_rnk commit 7e29047649fc202d7061c007ce3ba8a3962ed38c Author: Steven G. Johnson Date: Thu Feb 24 20:07:38 2005 -0500 #ifdef HAVE_STRING_H must come after rdft.h so that we get config.h commit d0b93533d99e69f85e2aaf759989f652311206ac Author: Matteo Frigo Date: Thu Feb 24 18:59:40 2005 -0500 Implemented reordering of loads and stores so that the real and imaginary part are loaded/stored together. This should improve out-of-cache performance in the presence of associativity conflicts, and maybe worsen in-cache performance because of worse scheduling. Enabled for now, for experimental purposes. commit 827ad1c139031037135765c5600dcf05b58030e4 Author: Steven G. Johnson Date: Thu Feb 24 18:10:49 2005 -0500 fix comment commit 35e5d61fd3b5f769ea631e357ac6f55002f74f96 Author: Steven G. Johnson Date: Thu Feb 24 18:10:23 2005 -0500 better message commit d2c6d9c9d37a6ea058c48c7445fbaca7089a6489 Author: Steven G. Johnson Date: Thu Feb 24 18:08:36 2005 -0500 use gcc version > 3.0 as fallback in check for alignment bug commit 9efbf189a95137e78b39f48e223e66df384eb89c Author: Steven G. Johnson Date: Thu Feb 24 18:02:31 2005 -0500 don't use -malign-double unconditionally (it is only available on x86) commit 858b560880b60856698a28728dd44964d456b7cf Author: Matteo Frigo Date: Thu Feb 24 12:03:30 2005 -0500 Subtler selection of tilesz. commit c44a6cff160e0ecd38f2a4f56bff4e34ddda2b59 Author: Matteo Frigo Date: Thu Feb 24 11:52:25 2005 -0500 Call cpy2d_tiledbuf, not cpy2d_tiled. commit 826a2387489dd9efde0ed09afc92e91e50a6d578 Author: Matteo Frigo Date: Thu Feb 24 11:29:28 2005 -0500 buffer sizes were wrong :-( commit fdabdfc4ef5010ed7965168b1ab583c296db3637 Author: Matteo Frigo Date: Thu Feb 24 11:19:01 2005 -0500 Single function for computing tile size. Eliminate spurious assertions. commit add19c2d3c32f843ff951cc227dc4ce1221fafb6 Author: Matteo Frigo Date: Thu Feb 24 10:00:02 2005 -0500 Do tiling recursively. commit 203fc5647fea6fe99f2d23cc43a24eeea47aee49 Author: Matteo Frigo Date: Thu Feb 24 09:40:30 2005 -0500 Reworked tiled transposes; provide tiling with and without buffering. I can't believe that one has to waste his life with this @#$%. commit c92a1fc69c9315d97f71a3070003d37923ac02b8 Author: Matteo Frigo Date: Wed Feb 23 22:21:19 2005 -0500 Clarified logic. I am not sure why the code was so confusing to begin with. The computation of *dp in the which_dim == 0 case was also wrong, returning e.g. *dp == -1 if sz->rnk == 1. commit 44692fa46d7313f08a624ec68bd421e282fa139f Author: Matteo Frigo Date: Wed Feb 23 22:00:15 2005 -0500 Enable aggressive inlining in codelets only, to avoid code bloat. commit e94240f1731b33ff9ad18ffe4c14a08a7d66d65a Author: Matteo Frigo Date: Wed Feb 23 21:51:50 2005 -0500 Removed cache-oblivious copy/transpose algorithms in favor of explicitly blocked algorithms. The cache-oblivious algorithms fail if there are associativity conflicts, in which case buffering is necessary, as per Carter and Gatlin. Once you set the buffer size, there is no point whatsoever to do the algorithm recursively, and you may as well use blocking. commit 77aeedee308c8b7bce0ff4c36986f715ced6748c Author: Steven G. Johnson Date: Wed Feb 23 18:46:12 2005 -0500 --disable-fortran now differs from --enable-fortran that fails commit 3cb3e167e76d53336c1307cecb6b1eb975bdda61 Author: Steven G. Johnson Date: Wed Feb 23 18:42:21 2005 -0500 comment tweak commit e0f881c48bd199f098eaa764fb17982cf1435475 Author: Steven G. Johnson Date: Wed Feb 23 18:41:14 2005 -0500 If a Fortran compiler was not detected, just make our best guess at what wrappers to use...I'm sick of dealing with user complaints from cases where wrapper detection fails for whatever reason. commit aa2c11cd3b47c6352d13b8f869f858082bb7a52a Author: Steven G. Johnson Date: Wed Feb 23 18:10:40 2005 -0500 fflush(stdout) after print_plan, in case F77 doesn't commit 76bdaf349e332587c7b5b4ae1fe55f4d3c0cc92d Author: Matteo Frigo Date: Tue Feb 22 22:54:42 2005 -0500 --enable-sse is necessary after all, to generate all dependencies correctly. commit 5844ac653fc5e937e4f2939d8a73dcc282657fd2 Author: Matteo Frigo Date: Tue Feb 22 22:32:06 2005 -0500 Put cpy2d_pair into its own file, so that I can experiment with buffering of nontwiddle codelets. commit e7d485c4f71be2a762c91d4d7e96a321afdfe858 Author: Matteo Frigo Date: Tue Feb 22 20:07:11 2005 -0500 Copy rfftwnd.png from ${srcdir}, not $PWD commit 0c56019ec6dc8f3c778b628a8a0b6094cd8a31d0 Author: Matteo Frigo Date: Tue Feb 22 17:08:48 2005 -0500 Do not bother memcpy-ing complex numbers. commit 6accb53a30744a5793b451670a70afb371cceeff Author: Matteo Frigo Date: Tue Feb 22 16:20:46 2005 -0500 Tighther layout of buffers. I am not sure it matters, but just in case... commit 0f5938fa6bcc89ad947656aa949a89feb73b7c77 Author: Matteo Frigo Date: Tue Feb 22 10:13:02 2005 -0500 Usec cpy1d for rank-0 copies commit 24a0b716253a1914882d738969bc8b101b70380f Author: Matteo Frigo Date: Tue Feb 22 10:06:13 2005 -0500 Implemented in-place transposes with buffering. Moved copy/transposition routines into own files, so that we can reuse them from multiple places. TODO: merge vrank3-transpose.c with rank0.c, or rename vrank3-transpose.c to rank0-fancy.c or something like that; decide whether square in-place transposes should be in rank0.c or vrank3-transpose.c; apply FIXME's in vrank3-transpose.c. commit 52f669f4280a8ad0834f201919290dc382898a4c Author: Matteo Frigo Date: Mon Feb 21 23:29:52 2005 -0500 Indentation should be printed after newline, not at the beginning of print() commit decdf03722050f50fba24b8152927c2327109e16 Author: Matteo Frigo Date: Mon Feb 21 10:07:24 2005 -0500 generalized in anticipation of more complicated solvers. commit 2a7b91a46dd814576f0dbfa54f17d38380bd35f0 Author: Matteo Frigo Date: Sun Feb 20 22:18:59 2005 -0500 Implemented buffered recursive transpose commit 4ce9d94def9d52633bb76b107aba65caa8c4fcf4 Author: Matteo Frigo Date: Sun Feb 20 18:27:29 2005 -0500 Fixed comment commit ac7a99027ee51e48f6be6dadcf00eb593d6017d9 Author: Matteo Frigo Date: Sun Feb 20 18:22:15 2005 -0500 grand unification of rank0 solvers commit 20af4f6724d7080f17a83aae996a6fd00e08ae7b Author: Matteo Frigo Date: Sun Feb 20 15:35:24 2005 -0500 manual tail-recursion optimization commit e834b974175d946c82b66c99c7bf18593f85cd8c Author: Matteo Frigo Date: Sat Feb 19 17:57:44 2005 -0500 implemented check for transpositions commit 6f6c5d224ae74b757b7013102ab25c018d7f9a30 Author: Matteo Frigo Date: Sat Feb 19 17:28:43 2005 -0500 Previous fix was wrong for rdft2 problems. commit 6bd660a504ef0345ea0f55db133690f9de7218ec Author: Matteo Frigo Date: Sat Feb 19 17:23:36 2005 -0500 vecsz->rnk must be finite for this solver to apply. commit 05d2a86385b2655cca135d882688ff493eccaa22 Author: Matteo Frigo Date: Sat Feb 19 17:15:19 2005 -0500 unified the various simple'' transposers commit e67ffc01608a1ebeedd99bb1390ff0ad58e33c0c Author: Matteo Frigo Date: Sat Feb 19 16:55:29 2005 -0500 Fixed stupid bug in rec_transpose_swap. Fixed stupid verifier that did not catch the bug. commit 49f3542f8f1ee7aa2bc7ddb12ded96d4b330b452 Author: Matteo Frigo Date: Sat Feb 19 15:24:03 2005 -0500 Minor cleanup of transposition routines. commit 770952578791d8ac1394ba8e19890fce2779ad67 Author: Matteo Frigo Date: Sat Feb 19 09:31:14 2005 -0500 Make the batch size B=Theta(r) instead of B=Theta(1) in buffered twiddle solvers. Theory: for cache line size L, we want B = Omega(L) to utilize the cache line fully. We also want B*r =O(Z), where Z is the size of the cache. It is safe to assume that Z = Theta(L^2): cache designers will tend to make L as large as they can get away with, because they don't have to program the machines that they build, and Z < Theta(L^2) will screw up the little matrix transposition benchmarks that they use to design the cache. Hence, B=Theta(r) is the right number. commit 0fc1650f8f411bc3fd1b6019b33d8e67d54b43a3 Author: Steven G. Johnson Date: Fri Feb 18 23:47:22 2005 -0500 for --enable-portable-binary, only try -mcpu=$arch and -m$arch on x86, since these generate non-portable code on every other target (and some other targets, like Alpha, don't support -mtune=$arch). commit 77be37a9825edf45432db688f9b6e307fc779320 Author: Matteo Frigo Date: Thu Feb 17 21:15:42 2005 -0500 gcc/aix defines _POWER, not __powerpc__ like the rest of the world does. commit da4852a84de13f2ed74462052a1081a8517fac9c Author: Matteo Frigo Date: Wed Feb 16 22:30:27 2005 -0500 enable fma for ia64, since it seems to help with the hpux compiler. commit e9b2b83177aabb8ff8d42f4b239e9eda1fbd10bf Author: Matteo Frigo Date: Wed Feb 16 21:47:48 2005 -0500 [empty commit message] commit 9f01f364832d025554f5912bd4f71c3c0b972d5c Author: Matteo Frigo Date: Wed Feb 16 15:27:18 2005 -0500 Fixes for darwin commit ff3f2d0d66afc832a1ec7f70d14e6d1520e40858 Author: Matteo Frigo Date: Wed Feb 16 14:27:42 2005 -0500 Made the correctness of the code more obvious. commit 0eaea796c7d8dfc833c38cc2485c68004bcb9d4c Author: Steven G. Johnson Date: Wed Feb 16 12:30:29 2005 -0500 s/with-portable-binary/enable-portable-binary/ to be GNUlly correct; I'm sticking with --with-gcc-arch=arch, however, as --enable-gcc-arch=arch has the wrong connotations for me commit 1f54539fae28f217a239c3dbc5c66a31784dbcd9 Author: Steven G. Johnson Date: Wed Feb 16 11:44:48 2005 -0500 whoops commit 743d6f8aa35cf29485b805e657e72afb83e401cf Author: Steven G. Johnson Date: Wed Feb 16 11:23:38 2005 -0500 bless wisdom with patience used to create it commit 741a55c0cb7529ae5ce8b1b3a01375a3f176a5e0 Author: Steven G. Johnson Date: Wed Feb 16 11:18:56 2005 -0500 whoops commit ab2c1f6788b6309abe08b585fa21ac7254e02f07 Author: Steven G. Johnson Date: Wed Feb 16 10:50:28 2005 -0500 whoops commit aa37add40de415143b25c5c3fa09d3d212af9ec2 Author: Steven G. Johnson Date: Tue Feb 15 23:53:53 2005 -0500 added 'timed' planner option commit 79f70936e6e19cb09dafb45f8ead8d9fff715111 Author: Matteo Frigo Date: Tue Feb 15 23:08:29 2005 -0500 Do not use SIMD_CFLAGS. The theory is that if taint.c is unsafe with SIMD_CFLAGS, then all files in this directory are as well. Conversely, if these files require SIMD_CFLAGS because they include "simd.h", then taint.c requires SIMD_CFLAGS as well, and thus we need some other hack. commit f9e6da507bcacf5aa503ce42e7cd73c0c501cbe2 Author: Matteo Frigo Date: Tue Feb 15 22:49:05 2005 -0500 Do not override CFLAGS in Makefile.am. commit 932e8f656a8a592700a3ca153c416e3e1504d278 Author: Matteo Frigo Date: Tue Feb 15 10:30:12 2005 -0500 Allow users to build long double version even if sizeof(long double) == sizeof(double) commit b35d88cdc10b06342c2c39a8d2012a71875aecf6 Author: Matteo Frigo Date: Mon Feb 14 19:55:38 2005 -0500 Updated for 3.1 commit 3c20661d7ca87a19ec855d94791bd24a3202e30d Author: Matteo Frigo Date: Mon Feb 14 19:07:14 2005 -0500 Oops, version.h is no longer used commit 485e6dbbea69f8e6438ec11fdb265cbe3b786464 Author: Matteo Frigo Date: Mon Feb 14 18:51:05 2005 -0500 unified fma and non-fma versions commit 800ea93e6f610aa9a7c15f1e9e7ed779dedefcfa Author: Matteo Frigo Date: Mon Feb 14 14:12:09 2005 -0500 forgot to remove inplace/Makefile from configure.ac commit 48bfe71f273d592eb0010911c4df16e12df1b9b4 Author: Matteo Frigo Date: Mon Feb 14 12:08:52 2005 -0500 Merged dft/codelets/inplace with the main dft/codelets/standard directory. This step makes dft codelets consistent with the rest of the naming conventions, and will simplify the eventual merge of fma and non-fma codelets. commit 1f70ee8f508d17b3cb0b694d838c71d4b411d740 Author: Matteo Frigo Date: Mon Feb 14 11:16:15 2005 -0500 inline altivec constants, since gcc seems to generate better code this way. commit 454930e2baceefbda8523cfbc103db0061604799 Author: Matteo Frigo Date: Sun Feb 13 18:17:32 2005 -0500 group altivec constants into a single array, for faster access commit 6cfc3df81b5b843ac0641d7aff61b76d29f82a63 Author: Matteo Frigo Date: Sun Feb 13 18:15:37 2005 -0500 code cleanup commit e8d683e0260b327eeedec8e25249bfd8c81cdda9 Author: Matteo Frigo Date: Sun Feb 13 10:29:32 2005 -0500 removed some unused stuff commit d495f6e14d219a63d1ed2a3e77e2c526e185a82c Author: Matteo Frigo Date: Sat Feb 12 22:04:40 2005 -0500 New twiddle scheme for altivec, 3dnow commit 510cdba23c47b1838f8a027da5680ad9ff21dcf3 Author: Matteo Frigo Date: Sat Feb 12 20:17:35 2005 -0500 Implemented new twiddle scheme for sse2 commit fd74e1eb06f6460dc3f0d8b6c5504fc005f98806 Author: Matteo Frigo Date: Sat Feb 12 19:57:46 2005 -0500 Implemented experimental t2* codelets, which store twiddle factors in a more convenient format, at the expense of twice the storage. Currently only SSE works; I have to port SSE2, altivec, etc. to the new scheme. After this, we will decide whether these codelets are worth the price. commit 9ba2ad18ff0a5c9a683120d7737cc6d343b83246 Author: Matteo Frigo Date: Fri Feb 11 08:07:12 2005 -0500 Forgot to define SIMD_STRIDE_OKPAIR commit 24aa1c39dc04c158a5275310b779bec639962a38 Author: Matteo Frigo Date: Thu Feb 10 22:20:00 2005 -0500 fixed sse2, 3dnow, and altivec, as promised commit fa8ee16c80d02c0a0a19391f9aa5897b37ac004b Author: Matteo Frigo Date: Thu Feb 10 21:47:40 2005 -0500 Generate n2?v_* codelets in such a way that we may or may not pair stores, depending on which mode happens to work best on a particular SIMD implementation. sse2, 3dnow, and altivec are currently broken---will fix soon. commit 8a141e0f8570683466ef4cf2aa4e8027d7ea698e Author: Matteo Frigo Date: Thu Feb 10 08:53:22 2005 -0500 instantiate altivec constants only once commit b23eef5ad62b650caafba583fae089d173718eac Author: Matteo Frigo Date: Thu Feb 10 06:37:56 2005 -0500 Fixed alignment checks for new SIMD scheme commit bf8b613b6a4299e8fcc3b36c1c0ec6c61ae944d6 Author: Matteo Frigo Date: Wed Feb 9 21:35:01 2005 -0500 Change n2?v_* codelets to store pairs of vectors, with implicit 2x2 transposition. Works for 2-way SIMD as well. Tested with sse and sse2. I haven't tried altivec yet, but I observed a huge speedup when I transformed one codelet by hand. commit b45f5e7af8fe63c291238eded48cff440ad1f4b9 Author: Matteo Frigo Date: Tue Feb 8 21:28:38 2005 -0500 Resurrected old DIF codelets for experimental purposes. They are disabled for now, but I am keeping the setup around for future reference. commit 2b2271e7df0c994e8ed02a49304a2ef279c084d2 Author: Steven G. Johnson Date: Tue Feb 8 20:10:19 2005 -0500 [empty commit message] commit c06695785e699d90aab66ce15e718ccab31f42bc Author: Steven G. Johnson Date: Tue Feb 8 19:37:09 2005 -0500 clarifications, document --with-portable-binary and --with-gcc-arch commit 4658829ef2505ec43aab6986fdc4778314c3e0bf Author: Steven G. Johnson Date: Tue Feb 8 19:23:41 2005 -0500 [empty commit message] commit 44be70997db3875b83dfe5dee436014717bdf235 Author: Steven G. Johnson Date: Tue Feb 8 01:36:22 2005 -0500 more change comments commit b7802bbb738b279d8d061756f90f03caecd0767a Author: Steven G. Johnson Date: Tue Feb 8 00:41:38 2005 -0500 fma is definitely beneficial on Itanium with the HP/UX compiler commit 95f76ca2081a043388616e815c0364bc6ffde166 Author: Matteo Frigo Date: Mon Feb 7 22:58:47 2005 -0500 Silence warnings. commit fe63ebfa96d081c7d45183e96a8d904d3dcfd226 Author: Steven G. Johnson Date: Mon Feb 7 22:55:49 2005 -0500 when we compile our own getopt, change symbol names to avoid conflicts (e.g. avoid build failure on MacOS X with --enable-shared) commit 151717343ac9ebd9197dfa0065de4176fa9d0894 Author: Steven G. Johnson Date: Mon Feb 7 22:36:42 2005 -0500 grr, more bugfixes for in-place case commit ca853db7099972e3b3840be7d1d3ee1abff00d04 Author: Matteo Frigo Date: Mon Feb 7 22:29:35 2005 -0500 removed relics of FRANZ mode commit b5015c430276d969565a9b6fe816a55556f8d6f7 Author: Matteo Frigo Date: Mon Feb 7 18:48:36 2005 -0500 Somehow xlc does not like ``vector int dummy;'' commit e8ba7b5c1885c85755dd33973ec8d2c5305f41e9 Author: Matteo Frigo Date: Mon Feb 7 13:59:47 2005 -0500 There is no need to enable sse to make the distribution. This might have been true in the past but not anymore. commit fea3ce788e0bd8cfd350e05d05c418e90b27ec63 Author: Matteo Frigo Date: Mon Feb 7 13:55:17 2005 -0500 Oops---included fortran file in C sources commit 2f4c935bb52c2e34940f4ad58ea6fd26ba30740f Author: Matteo Frigo Date: Mon Feb 7 13:42:45 2005 -0500 Set version string at ``make dist'' time, not at ``configure'' time, so we know whether a user is using the fma version or not. commit fcd17cfa8271300c8a41d87c9abd4968502ebaca Author: Matteo Frigo Date: Sun Feb 6 17:00:33 2005 -0500 Removed useless files commit 2707963bd735e791f7f5b8200c8c9d4f155bc4f8 Author: Matteo Frigo Date: Sun Feb 6 16:59:39 2005 -0500 Different (simpler?) way to prevent the compiler from optimizing loop inductive variables. We now explicitly corrupt stride variables by xor-ing them with another variable that happens to be zero (but the compiler does not know it). In this way, the compiler does not attempt to extract a zillion loop indices from codelets, which would overflow the register set. Set the -fno-loop-optimize flag to further help the process. Consequences: removed m* codelets. Smaller library size. Slightly faster code with gcc/powerpc (including altivec). Much faster code with xlc/powerpc. No changes for gcc/pentium. Maybe slightly faster with icc/pentium. commit 1e222893c8c84f35b16a63384ad1239e471ce684 Author: Steven G. Johnson Date: Sat Feb 5 18:51:08 2005 -0500 paranoia about in-place rodft00 plans commit 1d442744933c7161e86dd825d65aeb3d0c640e53 Author: Steven G. Johnson Date: Sat Feb 5 18:39:55 2005 -0500 don't believe pcost when using the estimator...there is no point, and it screws up estimator hacks to prefer in-codelet loops to vecloops commit 9ad39d1cad4ef56e0c29fc64a12a76e2e6195c52 Author: Matteo Frigo Date: Sat Feb 5 18:34:25 2005 -0500 Reduced optimization level from -O3 to -O for xlc, since -O generates faster code. commit 91fa9ff722538be49b29c22a3174cef3fdce9c25 Author: Steven G. Johnson Date: Sat Feb 5 16:26:58 2005 -0500 whoops, only applicable to redft00/rodft00 plans commit 6591b1e69eec3c3d11199ec3f84c341aa8e754db Author: Steven G. Johnson Date: Sat Feb 5 16:22:39 2005 -0500 fixed in-place operation, and don't create size-0 sub-plans commit f01834e572803db476083af9b0a0906b951ac9d0 Author: Matteo Frigo Date: Fri Feb 4 11:30:30 2005 -0500 Autodetect altivec on linux. This code works with gcc-3.4 and -maltivec, with or without -mabi=altivec. The code *should* work with gcc-3.3 without -mabi=altivec. However, disabling -mabi=altivec on gcc-3.4 produces much worse code (I don't know why). commit 2ac42677bbf31c868ad589a378f93887163910c1 Author: Steven G. Johnson Date: Fri Jan 28 00:04:58 2005 -0500 update reference commit 2f3db335dab469a165ed2d9a4f19435371ef9590 Author: Steven G. Johnson Date: Thu Jan 27 15:48:28 2005 -0500 note that DCT-II/III are often called the'' DCT/DCT commit a2480b0a7742cb4792f0a17ef54fcfa47bf9299f Author: Steven G. Johnson Date: Fri Jan 21 14:42:04 2005 -0500 added MSVC++ for ia64 (based on information at http://www.intel.com/cd/ids/developer/asmo-na/eng/19949.htm?prn=Y) commit fa86c1be03f8a3ac77ad8f17a4a0db76b8a08d04 Author: Steven G. Johnson Date: Fri Jan 21 14:22:50 2005 -0500 vc++ defines _M_AMD64 on x86-64, apparently commit 905e261576a2333fdc356609f6f6533740716663 Author: Steven G. Johnson Date: Tue Jan 18 22:30:27 2005 -0500 avoid gratuitous breakage with -Werror, requested by Simon Perreault commit 6fb09d4fad8df9be7c5cadda330234fbcf6bdecd Author: Steven G. Johnson Date: Mon Jan 17 18:54:55 2005 -0500 comment typo commit 2f9aac9cff6654101febb130659eab9345b58783 Author: Steven G. Johnson Date: Sat Jan 15 16:56:23 2005 -0500 bumped shared-lib revision# commit c793a51d3eafa054b132ebbc6095810261ac56b6 Author: Steven G. Johnson Date: Sat Jan 15 16:35:42 2005 -0500 add X(estimate_cost) to get estimator cost, and print from bench, to aid in tweaking estimator commit ef81def3aef05a8e513d2c28f9eba162af22020b Author: Steven G. Johnson Date: Sat Jan 15 14:57:56 2005 -0500 [empty commit message] commit 1b90ee6f155399994c4234601dfdce43c854555e Author: Steven G. Johnson Date: Sat Jan 15 12:57:07 2005 -0500 formatting fix commit 2abab58ebcf1286120285091b31ff706fa81cf81 Author: Steven G. Johnson Date: Sat Jan 15 12:31:28 2005 -0500 tweaks commit 044466122b66a254d87c396cbf0b17039543fd13 Author: Steven G. Johnson Date: Sat Jan 15 12:03:24 2005 -0500 use less buffer space commit 3e78c0361397476b699825b883be3d32331e8439 Author: Steven G. Johnson Date: Sat Jan 15 01:41:58 2005 -0500 added split-radix-based dct/dst I for odd n commit d994d2ded5077bfb54d19ee5c062e607b73ce73a Author: Steven G. Johnson Date: Fri Jan 14 21:50:08 2005 -0500 [empty commit message] commit cf8ef77af5eddfdda0d6c952ae0ae1955890bca4 Author: Steven G. Johnson Date: Fri Jan 14 21:49:55 2005 -0500 warn silly users who confuse CVS id with FFTW version commit e7ab0f25025fb3be5f73408419e51a2fcf54f031 Author: Steven G. Johnson Date: Fri Jan 14 16:57:36 2005 -0500 get sparc cpu type on solaris as well as with linux commit e82ef68d349c8df79cb772c944164b79b7f2c77a Author: Steven G. Johnson Date: Thu Jan 13 19:21:58 2005 -0500 detect prescott mobile (f37) commit 3622c28434b7292df2153c577f8262a2974fd6ce Author: Steven G. Johnson Date: Thu Jan 13 18:09:52 2005 -0500 use cpuid for x86_64 as well as i[56]86 commit ba6d8352bbd435da164d15a693e824711bcd86ce Author: Steven G. Johnson Date: Thu Jan 13 17:59:55 2005 -0500 update with x86info 1.7 and other sources (identify k8, nocona, etc), handle nonzero leading bytes in eax commit 92d9e4b244a2689bc7fb64105d20c874d09f9cca Author: Steven G. Johnson Date: Thu Jan 13 16:30:33 2005 -0500 compactified check for JOINABLE; use AC_DEFINE_UNQUOTED instead of AC_DEFINE for PTHREAD_CREATE_JOINABLE (thanks to Oliver Niekrenz for the bug report) commit 5440f786f094cdfb2b624e1e9050ba74a06ad780 Author: Matteo Frigo Date: Wed Jan 12 12:22:13 2005 -0500 The scheduler hack was incorrect because it swapped instructions of the form A = *B and *B = C. Fixed. commit 124a19a9d293ffa06f8b50519fc1e53ced2ca1ab Author: Matteo Frigo Date: Tue Jan 11 22:13:24 2005 -0500 Quote expressions such as ``if test $FOO = yes'' when $FOO may be empty. Also, $GCC is set to either ``yes'' or empty, never to ``no''. commit d52e4f122a2b71ab9272261bfec25931b8d9cd5b Author: Matteo Frigo Date: Tue Jan 11 19:30:47 2005 -0500 Hmm---somehow the previous commit did not work. commit ca5f6331f2b2432591707b129dc343705209e482 Author: Matteo Frigo Date: Tue Jan 11 16:54:45 2005 -0500 Fixed various gcc-related problems on powerpc: - gcc-3.4 becomes totally confused by expressions like vec_add(a, vec_add(b, vec_add(c, ...))) The compiler uses gigabytes of memory and then crashes, presumably because of the exponential-time search problem involved in typing the above expression (since vec_add can take either ints or floats). I changed VADD and similar macros to be inline functions, thus constraining the type system. - New flags --param inline-unit-growth=1000 --param large-function-growth=1000 to work around limitations of the gcc-3.4 inliner. commit 43a34b10c3383ccf1277216826b8201c3a0f3276 Author: Matteo Frigo Date: Mon Jan 10 21:27:24 2005 -0500 Check for HAVE_ALTIVEC_H commit 558d64554efbc303c104513b4f6243d2178335a8 Author: Matteo Frigo Date: Mon Jan 10 21:09:30 2005 -0500 Remove support for altivec using gcc builtins, since these keep changing across gcc versions. These changes work on gcc-3.4/linux; I haven't tried MacOS X yet. (The altivec ``spec'' differs between Motorola/Apple and gcc, grrr...) commit d9289c88276c6c878bd61c454049052d420013fa Author: Matteo Frigo Date: Mon Jan 10 18:57:30 2005 -0500 Stylistic changes commit 34b131fad38155ebb215614d02fba749c35c10ed Author: Matteo Frigo Date: Mon Jan 10 17:34:41 2005 -0500 Changed incorrect ugliness condition. commit 79acbd2e7f6820db37155925cdb1411a808b5bb4 Author: Steven G. Johnson Date: Mon Jan 10 16:09:43 2005 -0500 note x86info version number that was used, to make it easier to update the cpuid for changes in later versions commit 114d644618ae24c93784d908c47981dfbd32719e Author: Matteo Frigo Date: Mon Jan 10 15:00:51 2005 -0500 Make dft-r2hc non-UGLY for rank-0 problems commit 1a81406a03ef105ec8c188ee2f77dd605c25d422 Author: Matteo Frigo Date: Mon Jan 10 14:50:23 2005 -0500 Do not use -mcpu=970 on power4 processors, because power4 does not have altivec. commit 393ce48d0f58f8f7788198d59fa203e19f36db69 Author: Matteo Frigo Date: Mon Jan 10 14:48:47 2005 -0500 Note gcc-3.4 problem with inlining. commit c25eb53aa1676a746b9243a7463a62e3d753fb0f Author: Matteo Frigo Date: Mon Jan 10 13:51:08 2005 -0500 Oops, forgot to remove ``static'' from the declaration of noninlinable functions. commit 1447d501267177b4d3f4b5160a7e3b4fc16e7aba Author: Matteo Frigo Date: Mon Jan 10 12:31:26 2005 -0500 Recognize power4. Use ``head -n COUNT'' instead of obsolete ``head -COUNT'' (which fails on gentoo). commit 3de5bb754f1b9eb2514402c3b542a3735009f223 Author: Matteo Frigo Date: Sun Jan 9 22:12:16 2005 -0500 Remind to add FAQ entry concerning gcc-3.4.[1-3] crashes. commit 669ca8a3c4968477bf695ebc2961279779e0ec37 Author: Steven G. Johnson Date: Sun Jan 9 21:53:08 2005 -0500 whoops commit dcaa702e5c8d172b42b79a0c8ae14a1c8525f0a3 Author: Steven G. Johnson Date: Sun Jan 9 21:48:02 2005 -0500 support checking for major.minor.patchlevel commit 584fa85e1d7ca47d71b72c14f7dab1ac448048ec Author: Matteo Frigo Date: Sun Jan 9 21:40:18 2005 -0500 Revert CODELET_OPTIM to -O on IA32, which is faster than -O2. commit ec5ec6cbc0d0325a26eda54206f7f17253b39bae Author: Matteo Frigo Date: Sun Jan 9 20:30:12 2005 -0500 /bin/sh allows no spaces in assignments. commit 2b5a7ef73ab8bd55c32f63badf3120d1c4a62a28 Author: Matteo Frigo Date: Sun Jan 9 20:05:55 2005 -0500 Make non-inlinable functions external, so that gcc becomes confused and does not try to inline them. commit 321304bb50c85a0d5353f7bed5116d33b865dc4b Author: Matteo Frigo Date: Sun Jan 9 13:44:25 2005 -0500 Add -fno-web to CFLAGS, because -fweb destroys FMAs. commit 2f4f3044ed140d5b0edf1cf7415e0c0035392b40 Author: Matteo Frigo Date: Sun Jan 9 10:31:47 2005 -0500 Allow -mcpu=970 besides -mcpu=G5 commit e00f75f258a1b31526633b408804ed3c231cef68 Author: Matteo Frigo Date: Sun Jan 9 10:26:20 2005 -0500 configure was not using -fno-schedule-insns :-( commit e86fb1669da7d88ee98278e686d078ed205237c3 Author: Matteo Frigo Date: Sun Jan 9 08:52:40 2005 -0500 In mkplan() and elsewhere, use solver index instead of solver *pointer*, which looks marginally clearer. commit 446a3894d345237cabc59f659d5a2186c1f26554 Author: Matteo Frigo Date: Sun Jan 9 08:15:36 2005 -0500 Split planner hash table into two tables, for blessed and unblessed solutions respectively. Now an unblessed solution never overwrites a blessed solution, thus avoiding wisdom leakage by construction. Further, forget() is now a O(1) operation, which speeds up the estimator when the wisdom table is large. commit ee5380a2af1b55803ff5d64557ff5b9a2005b54b Author: Matteo Frigo Date: Sat Jan 8 21:19:45 2005 -0500 New TODO idea. commit 8bf4164bfd1d6aef62dac0e09eb5c5ef712ed8f4 Author: Matteo Frigo Date: Thu Jan 6 11:02:29 2005 -0500 Split search() into two routines to make the UGLY/NO_UGLY logic obvious. commit 1f170904d9848a43935bbd9a7c95d0249fa39138 Author: Steven G. Johnson Date: Fri Dec 17 16:08:54 2004 -0500 push/pop 64-bit registers on ia64; thanks to Orion Poplawski for the fix commit c53a0b8fa44ee6e63d41cdf2e4eb12589981f43d Author: Steven G. Johnson Date: Thu Dec 9 21:41:09 2004 -0500 patch from FreeBSD ports - FreeBSD does not have memalign, but its malloc is 16-byte aligned commit 31b763b9455632deddfb6425b630c4ce458b444e Author: Steven G. Johnson Date: Tue Nov 23 17:06:47 2004 -0500 don't compile taint.c with SIMD_CFLAGS (fixed Debian bug #259612) commit 5f505f2c11b292e769afc7de1e1fbb9bb75d1495 Author: Steven G. Johnson Date: Thu Nov 18 11:37:32 2004 -0500 revert incorrect change -- codlist.c should be rebuilt, but it is built in the build directory and not in the source directory commit 247e871cccf86dee2fa5543473c76373e5c46b34 Author: Steven G. Johnson Date: Wed Nov 17 22:53:53 2004 -0500 $(CODLIST) should be rebuilt only if Makefile.am changes, or alternatively only in maintainer mode, to prevent stomping in the source directory during user builds. (Thanks to Grant Cook for the bug report.) commit 7b6e452ba1709033b19a1056184ef5e7865773c3 Author: Steven G. Johnson Date: Sat Nov 13 13:43:01 2004 -0500 corrected #ifdef for icc/ia64, thanks to Matt Boman commit 80176573959dd2f034b41ab5d38c541281a5987a Author: Steven G. Johnson Date: Sat Nov 13 13:34:55 2004 -0500 spelling correction (Larsen, not Larson) commit 4e72b0ba4a2ee4245a1c996aabcea979753ded6e Author: Steven G. Johnson Date: Mon Nov 8 22:12:39 2004 -0500 use standard withval commit 38a050f2474601bd6fc7f1e9faca33e8656f0a63 Author: Steven G. Johnson Date: Mon Nov 8 22:09:16 2004 -0500 match doc commit 8d34c77d933aba00013d63875fb0a8cfdb5c5058 Author: Steven G. Johnson Date: Mon Nov 8 22:00:34 2004 -0500 formatting commit f354a059a0559c7816da1f1bfcbf30fef2965584 Author: Steven G. Johnson Date: Mon Nov 8 21:59:33 2004 -0500 make sure OPENMP_CFLAGS environment variable is used correctly commit caffdb38e0d057c260d21dcd45fee9d04ba48520 Author: Steven G. Johnson Date: Mon Nov 8 21:46:50 2004 -0500 replace ax_check_cc_flags with more generic ax_check_compiler_flags commit bc44b190250c3a55ddc841fdb85623efef8a1d04 Author: Steven G. Johnson Date: Mon Nov 8 17:49:42 2004 -0500 separate macro for OpenMP test commit 7bdd20309c710d7f29cb11cd2a130a2a453252ca Author: Steven G. Johnson Date: Fri Nov 5 16:24:22 2004 -0500 typo commit ba62ab6d94914626b1bb5c4fa59d239a92f5789a Author: Steven G. Johnson Date: Fri Oct 29 00:48:13 2004 -0400 [empty commit message] commit bbe80b4b34e5e86fb09b40b44a0f686b07bbd17b Author: Steven G. Johnson Date: Thu Oct 28 00:09:38 2004 -0400 better guessing of sparc type on Linux commit 93d85f0ab3a391bf35f1eb8c51e0d693736fa416 Author: Steven G. Johnson Date: Wed Oct 27 13:44:08 2004 -0400 note default commit 78065724b3f4e1170788d4d75cc1c1e318663b06 Author: Steven G. Johnson Date: Wed Oct 27 13:41:57 2004 -0400 tweak commit e43858fa862ad22519805870bef8be66593db88c Author: Steven G. Johnson Date: Wed Oct 27 13:34:25 2004 -0400 comment commit af53c27b20c589cc956cc567f7a85d05e5f9996d Author: Steven G. Johnson Date: Wed Oct 27 13:31:10 2004 -0400 whoops, m4 is EXTRA_DIST, not SUBDIR, since it doesn't have a Makefile commit fc7444822d899746b1c4e68cb06847ce95ff12b7 Author: Steven G. Johnson Date: Wed Oct 27 13:16:57 2004 -0400 silence warnings commit be281108e1c825de4313ece30b12fd918273b1a9 Author: Steven G. Johnson Date: Wed Oct 27 13:14:22 2004 -0400 clean up m4 macros; try to detect correct gcc -march flag on x86; new --with-portable-binary, --with-gcc-arch= flags; use -O2 for codelets with gcc 3.4 to work around bug commit 9403174ddea85728f959287755950e43901c2d39 Author: Steven G. Johnson Date: Tue Oct 26 16:46:14 2004 -0400 rename cexp -> mcexp to avoid conflict with C99 builtin commit d581a67939f4d7c95a0b07b3a4952d35e44bb17c Author: Steven G. Johnson Date: Mon Oct 25 16:58:23 2004 -0400 use basename , w/o args, for compiler-name comparisons; also detect Compaq ccc on alpha-linus commit a1d9fccd5bdda57ae410ba0ce15367e987d64f73 Author: Steven G. Johnson Date: Sun Oct 24 22:05:10 2004 -0400 note recent icc problems commit dfddc484065adab609af43ba17821394c23dc5cd Author: Steven G. Johnson Date: Sun Oct 24 02:10:12 2004 -0400 whoops, disable semaphores again (for now) commit a2dad5feeb7e13cc8d93adb55bb59ed0431341be Author: Steven G. Johnson Date: Sun Oct 24 02:04:58 2004 -0400 POSIX semaphores are *not* the same as SYSV semaphores commit 64a5d0fd73897b1f811382dc5238209dfe9672be Author: Steven G. Johnson Date: Sun Oct 24 01:18:14 2004 -0400 re-implement threaded stuff; dftw now takes parameters to indicate a portion of m loop commit 99fecf91b80dfe5aabdd4b3d69cc71639de2c483 Author: Steven G. Johnson Date: Thu Oct 21 20:44:51 2004 -0400 more C++ notes commit 77e885e9ca5d60b2b34f126b21cce95382cafc59 Author: Steven G. Johnson Date: Thu Oct 14 09:50:38 2004 -0400 note bug report for VC++ 6.0 from Dale Dickerhoof commit e5523dbd23cd0d4beff0d5b53ca76a275b7e5b1f Author: Steven G. Johnson Date: Fri Oct 1 16:06:59 2004 -0400 fmt commit 8e9f882720c1fc5f2c7c3b168a8f48608af95057 Author: Steven G. Johnson Date: Fri Oct 1 15:59:17 2004 -0400 comment typo commit 689ac491bc35a2728b1ae0ccc6e1698f84a04f4f Author: Steven G. Johnson Date: Fri Oct 1 15:48:09 2004 -0400 bug fix -- ishift/oshift only apply to execution of child plan commit 6438e86b96980ae10958e4483acf04e80573c1dd Author: Matteo Frigo Date: Thu Sep 30 21:12:47 2004 -0400 New planner that tries never to lose wisdom. commit 28f9e28b2b6d2c7d5969c93cdf3c460f6fd895a2 Author: Matteo Frigo Date: Thu Sep 30 13:36:43 2004 -0400 Nested comment was triggering a warning. commit 7f1f6a5fe5723ce3079588306a98c43289f6df32 Author: Steven G. Johnson Date: Fri Sep 10 15:20:07 2004 -0400 system "root" under dgjpp is /dev/env/DJDIR, not /dev/env/DJGPP, according to djgpp's libc.info; patch confirmed with J. M. Guerrero commit 354611ae36fd7494d3f90789fa33d6b26febeec2 Author: Steven G. Johnson Date: Wed Sep 8 18:50:03 2004 -0400 some minor portability fixes for djgpp; thanks to Juan Manuel Guerrero for the patch commit 133be56f2adeb3f4ab3c394a03da4254f758eacf Author: Steven G. Johnson Date: Thu Aug 19 12:41:23 2004 -0400 pointer to tutorial for quick start commit 6a23ed45415cae1a9825953e80dc99ceee5d185a Author: Steven G. Johnson Date: Thu Aug 19 12:39:50 2004 -0400 point users to manual commit b759a1ca992dee63a97a67ba2beddde782dba6c7 Author: Steven G. Johnson Date: Sat Aug 7 13:42:22 2004 -0400 minor typo commit 693ed3bc9f47a262b3502ad06b42be41f68ee47f Author: Steven G. Johnson Date: Sun Jul 18 18:54:18 2004 -0400 use __DECCXX for Compaq cxx, not Linux-specific symbol commit df4ddeeaad67144bd7d6f855f690cf06907f1d56 Author: Steven G. Johnson Date: Fri Jul 16 13:55:25 2004 -0400 patch by John Bowman to make cycle counter work with DEC cxx under Linux commit fd9cd11e5b8806245d6b5522fdef29b1626eda0f Author: Steven G. Johnson Date: Wed Jun 30 00:45:10 2004 -0400 updated pruned FFT discussion, with link to further details on www.fftw.org/pruned.html commit 243e4dafca54e62e83d796c176d4af2ce00690b1 Author: Steven G. Johnson Date: Mon Jun 14 20:08:27 2004 -0400 darwin is based on freebsd commit fa86af755d34199fa6ddf2a1e40dbedb9898f5bb Author: Steven G. Johnson Date: Thu Jun 3 14:23:41 2004 -0400 in --with-windows-f77-mangling, add lowercase + single underscore for Intel compilers, etc. (thanks to David Gomez for the bug report) commit 3f13a0eb176fe03d5937ef282b9defa42c258876 Author: Steven G. Johnson Date: Wed Apr 7 00:46:07 2004 -0400 whoops, extra alignment check commit f0e8345ba78e99831a1589192d6fc3b2e1e41e38 Author: Steven G. Johnson Date: Wed Apr 7 00:16:49 2004 -0400 disable most 2-float-as-double copying, add alignment check in one remaining place commit a12d8b846381396d54acbc1748e53cbc0c09baac Author: Steven G. Johnson Date: Tue Apr 6 13:49:13 2004 -0400 make sure it is clear that real-even/odd refers to symmetry, not size commit 920197fd649070eadef659b39572b155a8b0c36c Author: Steven G. Johnson Date: Mon Apr 5 20:18:29 2004 -0400 optimization commit c0d199f22910faaf1f4850900185c161a585f96b Author: Steven G. Johnson Date: Fri Apr 2 21:31:00 2004 -0500 separate cutoff for ugliness...these cutoffs are still not ideal commit e1920963de856b058811b84764d848947cab454f Author: Steven G. Johnson Date: Fri Apr 2 21:30:17 2004 -0500 transpose.c is gone commit a115ba2703fa6d7cfb8e1453904bf94cd9c25b7d Author: Steven G. Johnson Date: Fri Apr 2 21:18:27 2004 -0500 move all rank0 transforms to rdft commit 444b8f48586e952b107d4ee2ad58c56e357e5fbd Author: Steven G. Johnson Date: Fri Apr 2 20:35:35 2004 -0500 enable fp-moves/us comparison of rank-0 transforms commit 2cc4d9f4818fb41d6aa0c1be4224eb25a94b3ac7 Author: Steven G. Johnson Date: Thu Apr 1 16:13:22 2004 -0500 whoops commit b0ee7083fd7d7ff73366c8011fbf43d675380d8c Author: Steven G. Johnson Date: Thu Apr 1 15:25:30 2004 -0500 whoops commit 50854b83979e79e4a0a2f6e90404ca553d0a3d33 Author: Steven G. Johnson Date: Wed Mar 31 18:11:02 2004 -0500 sort tensor dims by stride absolute values, not strides commit 39cd8178427b5a70d5fa503a14c663c6a4f96edf Author: Steven G. Johnson Date: Tue Mar 30 20:22:50 2004 -0500 [empty commit message] commit 9b5e15aa7f7e515b9faa505be708a14d568ceb1b Author: Steven G. Johnson Date: Tue Mar 30 19:44:54 2004 -0500 added improved transpose algorithm for N x M where |N-M| is small commit 2db4ea7db1f587486546e244ea42930e51275806 Author: Steven G. Johnson Date: Tue Mar 30 19:41:14 2004 -0500 check to make sure SIMD matches precision, and make sure user doesn't select both SSE and SSE2 commit 8995d09da3bc5d0fd7daf6f6ad295fccd9e94893 Author: Matteo Frigo Date: Sun Mar 28 09:26:38 2004 -0500 Implemented hc2hc-generic hc2r. commit 581a83475a46b89a73b8a7fb3f2dccb140f72629 Author: Matteo Frigo Date: Thu Mar 25 11:19:25 2004 -0500 Inverted loop for stride-1 access. commit b0d68fa533d26d6bebf433c43c4dfee7b99a3701 Author: Matteo Frigo Date: Thu Mar 25 11:18:49 2004 -0500 Swapped j <-> k for consistency commit dc715359aa2bb496a60ae650612b42f0cdf998dc Author: Matteo Frigo Date: Tue Mar 23 12:08:07 2004 -0500 Require that R be odd commit 36e2199cf602f511e50a4bbc56e472d79c935e8f Author: Matteo Frigo Date: Tue Mar 23 11:49:01 2004 -0500 Implemented hc2hc-generic (DIT only for now). commit bc377e92e0d11be803dc1a3deb60f05a82799f85 Author: Matteo Frigo Date: Mon Mar 22 14:43:16 2004 -0500 Relax equality of twiddle description, since the `i' field is not used by TW_FULL or TW_HALF. commit ede9d975b188649b84cca9bf24c5f7feab3653c4 Author: Matteo Frigo Date: Mon Mar 22 13:22:44 2004 -0500 Do not allocate tw_instr's on the stack. Thus, the ``consistency check'' in twiddle.c becomes wrong. commit 19b8fbca72260c622266cd93466267c9dfb57cc3 Author: Matteo Frigo Date: Mon Mar 22 13:21:28 2004 -0500 Fixed incorrect malloc()/free() logic. commit 050be8cad10f411ab6ca025f59e5cffc3f7bf42d Author: Matteo Frigo Date: Mon Mar 22 09:04:37 2004 -0500 Silence warnings commit ae20d94938c08cb65f257fec653a9e3b1961a77b Author: Matteo Frigo Date: Mon Mar 22 09:02:55 2004 -0500 Separate file for hc2hc common routines commit e35b856a11108d5bdf61855976cddd7e8e7a84e9 Author: Matteo Frigo Date: Mon Mar 22 08:23:56 2004 -0500 (re)Implemented buffered hc2hc. Slight simplification of twiddle-factors management. commit de8ff3b06710f0dda76007150592239d4aa7565c Author: Matteo Frigo Date: Sun Mar 21 19:53:05 2004 -0500 Incremented libtool revision number before we forget. commit 5004b2e13de1b9b8635441bba800e8f6b850900a Author: Matteo Frigo Date: Sun Mar 21 19:25:56 2004 -0500 Fixed opcnt commit af360d8473ebdda79f57a3fa6bd3bbb2b7b041a2 Author: Matteo Frigo Date: Sun Mar 21 17:56:15 2004 -0500 Renamed files. These solvers are not really cooley-tukey. commit c6c735fb857127becb133e21c37544052b985806 Author: Matteo Frigo Date: Sun Mar 21 12:38:45 2004 -0500 Started moving rdft/ to the new cooley-tukey ontology commit 5df5843950df1fb50697f28d983ff0a9b8d5c5b9 Author: Matteo Frigo Date: Sun Mar 21 10:59:42 2004 -0500 Plans in ct-*.c are subtypes of plan_dftw, not plan_dft commit 5a4eb1dc842c864c311f175e9f97dde3f42dba2f Author: Matteo Frigo Date: Sun Mar 21 10:38:18 2004 -0500 Slight simplification commit 23b338208a4b752b307a0c6ff8d03f4e3f3c077f Author: Matteo Frigo Date: Sun Mar 21 10:20:06 2004 -0500 Minor simplification commit 71c684955ce3dfb91065a561e806edc213b2a1ae Author: Matteo Frigo Date: Sat Mar 20 08:43:57 2004 -0500 Workarounds for icc-8.0 nonsense. commit 446cbae42c628ae2ba7e6f63f4771355a10b5e0f Author: Matteo Frigo Date: Sun Mar 7 07:56:08 2004 -0500 FFTW_FORWARD is not technically an ``option''. commit 150af2bf6e6d380dc31ebffdcb79961e64d47f97 Author: Steven G. Johnson Date: Tue Feb 24 12:17:06 2004 -0500 Alejandro requested that his name be removed from @author commit 6948af91f0140722c52246a2b09faaeb7e664d99 Author: Steven G. Johnson Date: Mon Feb 23 17:42:56 2004 -0500 GNU Pth emulation library check commit b28089821d98c117e9688fdb7c65b4bfc0645345 Author: Steven G. Johnson Date: Sat Feb 21 17:51:13 2004 -0500 calling can-do calls the estimating-planner, which creates wisdom that we don't want ...we should be able to do all of the documented problems, anyway commit 262bd966d1d8394d17b2dc1ae7b76446b9300323 Author: Steven G. Johnson Date: Sat Feb 21 17:46:06 2004 -0500 don't forget_wisdom because of side effects commit b5c61a6821de885155d1b960a1d8b50a5464bc3d Author: Steven G. Johnson Date: Sat Feb 21 17:42:47 2004 -0500 forget wisdom from can_do commit 51442d30f17617100834e2fb27cbe7df79b3d61b Author: Steven G. Johnson Date: Thu Feb 19 14:11:14 2004 -0500 parenthesization commit 63bf06148e526cb5c90550fbd7b53a40fc73f2d4 Author: Matteo Frigo Date: Fri Feb 13 07:20:31 2004 -0500 Split malloc into kernel_malloc and API malloc commit 26fb1d12ecc37fda0f9760386b1f59a87e193e01 Author: Steven G. Johnson Date: Thu Feb 12 15:42:20 2004 -0500 X(malloc) must be extern "C" commit bb95c42e188e35ab4f22703978e5da7ba796eaff Author: Steven G. Johnson Date: Thu Feb 12 15:41:44 2004 -0500 satsify C++ compiler commit 5560fa42b5df5500dfd63303262c412f308ceb76 Author: Steven G. Johnson Date: Thu Feb 5 20:39:14 2004 -0500 with the new flags, fma is definitely beneficial on PA-RISC with HP/UX cc commit 795e5b6919fec5bbdf1fb9cff3be1db5f63ddeee Author: Steven G. Johnson Date: Thu Feb 5 19:52:17 2004 -0500 grr, Ofaster etcetera are not supported under older versions of the compiler. Note that +Ofltacc *disables* fp-reordering optimizations (which are enabled by +Oall). +Optrs_ansi is the older version of the aliasing stuff commit db287e0973e7d6ef19261b5a96979dff3b339b9a Author: Steven G. Johnson Date: Thu Feb 5 19:26:01 2004 -0500 +Otype_safety=ansi on hpux commit c98916ab1aa18fd95e8e43584e6eb618015de573 Author: Steven G. Johnson Date: Thu Feb 5 19:22:34 2004 -0500 just use +Ofaster on hpux (+O3 +Onolimit +Olibcalls +Ofltacc=relaxed -Wl,+mergeseg) commit 4e3bf163dca0615df17146ee2e18481d0a20a9e6 Author: Steven G. Johnson Date: Fri Jan 30 14:17:15 2004 -0500 check for win32 threads for mingw32; thanks to Alessio Massaro commit 86652c99050dcc4f52d17974597bdbf56a5998de Author: Steven G. Johnson Date: Thu Jan 29 15:23:33 2004 -0500 added missing 'static', thanks to Alessio Massaro commit 96566e4ddd0a9d00b23ad7c8ad04240cce7bfac1 Author: Steven G. Johnson Date: Fri Jan 9 16:36:48 2004 -0500 print more like bluestein commit 9ce57c8bf150fdf7d8177b0252abd7721d82d28b Author: Steven G. Johnson Date: Fri Jan 9 15:45:22 2004 -0500 fixed op count for R2HC_ONLY_CONV commit 1d28fc43969f2824c21efd2d1f2ce5a365dc0a07 Author: Steven G. Johnson Date: Fri Jan 9 15:41:50 2004 -0500 include DESTROY_INPUT in buffered flags for in-place...otherwise in-place hc2r uses rdft-dhtcvs diff commit bf6f542cb4702cf9d2c9346254cf09f0bce0e032 Author: Steven G. Johnson Date: Fri Jan 9 15:41:09 2004 -0500 resurrected R2HC_ONLY_CONV option to share plans and save on planning time commit 04d01b659718c6e66f9b14ad925fc5c630c7c1e8 Author: Steven G. Johnson Date: Fri Jan 9 14:47:00 2004 -0500 precompute folding for cyclic convolution commit 41947ea5b1fd5203da848d5afe3ecfed87f90d91 Author: Steven G. Johnson Date: Wed Jan 7 16:48:39 2004 -0500 minor commit 766e29f31c88457b71f94bc7607e81a39b445fdd Author: Steven G. Johnson Date: Wed Jan 7 16:48:25 2004 -0500 note reports of successful compilation on Windows commit a3b6ef73f675682810957a7770b13c5ede3c75a3 Author: Steven G. Johnson Date: Wed Jan 7 14:16:16 2004 -0500 citation year commit c6ff6592cd7196994610014cb3da0caafa4354df Author: Steven G. Johnson Date: Tue Jan 6 01:07:36 2004 -0500 comment commit 326cb17c5ee1b2d61f66a81bf90c012128148add Author: Steven G. Johnson Date: Tue Jan 6 01:07:08 2004 -0500 comment fix commit 16c7ff8cc5de0d74ad68a628e80e9ac8ede5e918 Author: Steven G. Johnson Date: Tue Jan 6 01:06:57 2004 -0500 fixed naming cruft commit 3c33d645c75a685c1b95f12be64b272dd01fb621 Author: Steven G. Johnson Date: Tue Jan 6 00:56:16 2004 -0500 space commit daf2625a4becbd9aac13358827eef8bbf2ef115e Author: Steven G. Johnson Date: Tue Jan 6 00:55:53 2004 -0500 comment commit 8bc4eaa51d81346cd07706771fc58bb8767bd428 Author: Steven G. Johnson Date: Tue Jan 6 00:54:07 2004 -0500 moved assert commit 679d41c77182afe28023a32c667fb6771f7df10d Author: Steven G. Johnson Date: Tue Jan 6 00:49:16 2004 -0500 comment commit 0ead6d3532ee317a8e8e99391655c0ef3b13bfd3 Author: Steven G. Johnson Date: Tue Jan 6 00:41:06 2004 -0500 delete old R2HC_ONLY_CONV hack, now defunct commit 4f45958b21e10e4b0ad128e4a36b3b1c456e7a22 Author: Steven G. Johnson Date: Tue Jan 6 00:32:58 2004 -0500 added padded real rader commit 2743f45c6e17d1cc72cce590faa7475c03e8ea74 Author: Steven G. Johnson Date: Mon Jan 5 22:56:58 2004 -0500 removed unused var commit 57cf035f36b08fae46144c4e03a52b3fb7a65f51 Author: Steven G. Johnson Date: Mon Jan 5 21:20:29 2004 -0500 handle both FFT_SIGN values commit 0f0e531863a2392a747c6b540a966b48b77debaa Author: Matteo Frigo Date: Fri Jan 2 06:07:51 2004 -0500 Oops: d->ros ==> d->ios commit 6ce67e298050d47829243bf692910eeef126601f Author: Matteo Frigo Date: Fri Jan 2 06:05:10 2004 -0500 Oops: d->ris should have been d->iis commit 3ddc923840694e141348bb8f48eb6bc9272b607b Author: Matteo Frigo Date: Thu Jan 1 16:00:07 2004 -0500 Removed rdft rader cooley-tukey, to be superseded by a generic reduction of rdft twiddle problems to dft + pre/post processing commit 3f82980635418e49dad204ff327021a8adf4bcfb Author: Matteo Frigo Date: Thu Jan 1 15:44:09 2004 -0500 In anticipation of the upcoming revision of rdft, removed rdft generic dit/dif cooley-tukey, in favor of generic rh2c and hc2r solvers. Cleaned up stuff that became unused after this change, such as TW_GENERIC. commit f7546dadff20223e87c698a9c3e8bcdb8496547e Author: Matteo Frigo Date: Thu Jan 1 12:59:30 2004 -0500 Removed useless file commit 822bd0498b7fe0a45a001af73cd4b317f33b0230 Author: Steven G. Johnson Date: Fri Dec 26 13:54:00 2003 -0500 whoops, don't call AC_F77_DUMMY_MAIN if no Fortran compiler is found; thanks to Charles Radley for the bug report. commit 6d8fa3754568aeb1979cbd6d1f6b91c90f524989 Author: Steven G. Johnson Date: Fri Dec 19 13:58:05 2003 -0500 guess good flags for Solaris/intel, suggested by J. Gregory Wright commit e393cf5533fdff4834a269b4d163641553f9532f Author: Steven G. Johnson Date: Fri Dec 5 19:55:13 2003 -0500 blah commit bcb1ecc806de458e9744b90452468221ab65d36d Author: Matteo Frigo Date: Sun Nov 30 06:59:41 2003 -0500 DIF generic solver was destroying the input. commit 1580db9c3c75c57928058fc81faa981295c7b6ca Author: Matteo Frigo Date: Sat Nov 29 19:28:39 2003 -0500 Fixed bug that caused HC2R transforms to destroy the input in certain cases, even if the user specified FFTW_PRESERVE_INPUT. commit 24f8af52f2a239ba51cd03e37e4c1c74befdc2f2 Author: Matteo Frigo Date: Sat Nov 29 16:49:01 2003 -0500 Implemented swap_io hack for r2r verifier. commit e15bf89a0e15d64e294ea23deb4f3de422e467b6 Author: Steven G. Johnson Date: Thu Nov 20 22:00:53 2003 -0500 citation commit f8afd813d4a9af4549ccca07ea7604c12c081761 Author: Matteo Frigo Date: Fri Nov 14 20:57:55 2003 -0500 Trying to get ``make paranoid-check'' to work. (Still broken.) commit 74399102001e6b633c21ca16f1b4369f63facab1 Author: Steven G. Johnson Date: Fri Nov 14 20:05:54 2003 -0500 fixes for input-preservation tests commit 29fc95192096b664dc7f23bfc6d3530f5880c66f Author: Matteo Frigo Date: Fri Nov 14 19:19:31 2003 -0500 Assume FFTW_PRESERVE_INPUT unless either the `d' flag is given in the problem, or the problem is multidimensional c2r (which fftw3 cannot without destroying the input). With this change, we can at least test that FFTW_PRESERVE_INPUT works in the c2r 1d case. commit 5a2907cf7121cc9e824150f654d83c9ff984aa92 Author: Steven G. Johnson Date: Fri Nov 14 19:14:40 2003 -0500 apply should copy back input for input-preservation check commit e5b287efdbd1a909467ac69e49f09c708462ea6d Author: Matteo Frigo Date: Fri Nov 14 19:01:36 2003 -0500 Undone previous bogus changes commit 55075f65e5f23bb045b5a076bd559d508e15fab8 Author: Matteo Frigo Date: Fri Nov 14 18:27:12 2003 -0500 Check dr[fb] in addition to r[fb] commit 74d5a2653d744d48adc748e8ebfaafdad0198cb7 Author: Matteo Frigo Date: Fri Nov 14 17:33:44 2003 -0500 Fixed conditions under which the rank-geq2-rdft2 solver is applicable. The old solver was not applicable for out-of-place problems unless DESTROY_INPUT. This is bogus. As long as the subsolvers honor !DESTROY_INPUT, the solver is always applicable. Changed semantics of test program, so that PRESERVE_INPUT is always true unless the problem specifies destroy_input explicitly. Without this change, there is no way to test the new solver. commit edcc72abc5fafc3147bfb8b802cea42d249c711b Author: Steven G. Johnson Date: Thu Oct 30 15:10:42 2003 -0500 added AIX OpenMP (-qsmp=omp) support; thanks to Greg Bauer commit 8dffe5112def767dc95ffe8c722d009bd2d3a5aa Author: Matteo Frigo Date: Thu Oct 30 10:11:39 2003 -0500 G5 CFLAGS commit 3cdf00d461370ae110601bf4612b31601a9b7100 Author: Steven G. Johnson Date: Fri Oct 24 04:17:39 2003 -0400 western FAQ commit 8ae00fe74ec955caecad22123ab716b908fb595a Author: Matteo Frigo Date: Thu Oct 23 11:34:11 2003 -0400 Oops. commit f7b3b4aae8910a377ab2c391da7855bad4c6875a Author: Matteo Frigo Date: Thu Oct 23 11:28:28 2003 -0400 Autodetect altivec commit 7458c1cb849f028a73209b18da00a2b63da0d861 Author: Steven G. Johnson Date: Wed Oct 22 01:14:10 2003 -0400 MinGW gets confused by a single / commit 3b8090b60a50893cb21b3e7442c5d8eec86c756e Author: Matteo Frigo Date: Fri Oct 17 10:46:41 2003 -0400 Paranoid portability fix commit 82175d1a4b25bd246759a5a9499a50037b51bc94 Author: Matteo Frigo Date: Thu Oct 16 11:07:46 2003 -0400 size -> length, which should make clear that we are not talking about arbitrary precision. commit b5f6b9f86f41b0429dd03c64101bc2cbd4a0b261 Author: Steven G. Johnson Date: Wed Oct 15 15:01:40 2003 -0400 pruned transforms are a FAQ commit 736d76c64078a558714dc87e31bc158117a274d7 Author: Steven G. Johnson Date: Wed Oct 8 23:54:17 2003 -0400 NO_SEARCH has already been mapped to FFTW_WISDOM_ONLY commit 02a8d7f029570cfb78c03215109a391d74f22ebd Author: Steven G. Johnson Date: Wed Oct 8 23:53:19 2003 -0400 newline commit 4177857895703cbd233c7b10e80cb95b8e8c25c5 Author: Steven G. Johnson Date: Sat Sep 27 20:27:32 2003 -0400 fix commit 694836051975049f0d2981df1372a23be9d9bc1b Author: Steven G. Johnson Date: Sat Sep 27 20:24:39 2003 -0400 clarification commit 0b20096c7a3afe66a13dc3efd84ba9e535748248 Author: Steven G. Johnson Date: Sat Sep 27 17:43:57 2003 -0400 minor fix commit 39ef965f44a5ba2245f12c7e1b182032746d6dda Author: Steven G. Johnson Date: Sat Sep 27 17:42:30 2003 -0400 grammar commit f7d34b13ca2adef07ab32520d97f05c403d34bd4 Author: Steven G. Johnson Date: Sat Sep 27 17:29:04 2003 -0400 html output fix commit 4b0c92ff25547e2f4fef16bf2089b6525407fc9e Author: Steven G. Johnson Date: Sat Sep 27 17:22:48 2003 -0400 mentioned sqrt(2) factors for DCT/DST commit 512c0e8650df0efa89bc8fea862a005f6f58f2cc Author: Steven G. Johnson Date: Sat Sep 27 17:07:18 2003 -0400 FFTW_WISDOM_ONLY flag (undocumented for now), suggested by Phil Dumont commit 54b4afc9f006bc10ce0423f09625b91af30d9dc3 Author: Steven G. Johnson Date: Tue Sep 23 23:36:19 2003 -0400 removed UpTime code commit 26c7e51b101004fe0cc9c2a5f90c732fd7bcf2b2 Author: Steven G. Johnson Date: Tue Sep 23 23:27:29 2003 -0400 updated documentation for mach_absolute_time commit 12a80b367661367374d32deeb29e01fd75e311e3 Author: Steven G. Johnson Date: Tue Sep 23 23:25:52 2003 -0400 use mach_absolute_time on MacOS/Darwin, as a fallback; don't bother checking for UpTime since it requires extra libs commit fa1787b57c9fb3539af76bd43c35ce224da7dace Author: Steven G. Johnson Date: Tue Sep 23 22:59:29 2003 -0400 support Apple UpTime function for asm-less xlc, grrr... commit 13e7c9ab7273a625f9b21015a75eff8ef163d468 Author: Steven G. Johnson Date: Tue Sep 23 15:42:29 2003 -0400 additional paranoia for xlc etc. commit 2da4a3de8376303c716cd7ee5b3b47b8f759983b Author: Steven G. Johnson Date: Mon Sep 22 15:28:56 2003 -0400 work around _Complex_I weirdness in xlc, reported by Greg Allen commit 3533775b3de13e397feb15794631f1d2fedd1e98 Author: Steven G. Johnson Date: Fri Sep 5 18:03:11 2003 -0400 typo commit ef0a55daee823f7ae55367f4304e32abf20c0d40 Author: Matteo Frigo Date: Fri Sep 5 13:11:40 2003 -0400 New script that produces commercial version. commit 4c4873b3aface20d5444216fcb5ea79221e5a289 Author: Matteo Frigo Date: Fri Sep 5 07:27:06 2003 -0400 Noted that VC++ is buggy. Noted that we know nothing about Windows. Noted that the sky is blue as well. commit a3d172eb32ea6f17849604998b55db355d60cc26 Author: Matteo Frigo Date: Tue Sep 2 09:04:19 2003 -0400 Noted that certain arrays are no longer used after the planner has completed. commit c68f62d97627bb869061a9433c4f4605f4dc8fd4 Author: Matteo Frigo Date: Tue Aug 26 08:22:38 2003 -0400 Typo commit b95fbd832374dacc4e8c9fb21b8c4085b33b3460 Author: Matteo Frigo Date: Mon Aug 25 21:27:43 2003 -0400 New item commit b2a02ef7112f03f73e25ccb67227265306b69b6c Author: Steven G. Johnson Date: Thu Aug 21 17:36:08 2003 -0400 try creating output file before planning (thanks to Phil Dumont for the suggestion) commit 766c3757fc99565ef8a14a9f3d5729740b1e7182 Author: Matteo Frigo Date: Tue Aug 19 10:08:07 2003 -0400 Clarified fftw_cleanup() commit 8eecb544f6ac10687ec258d3162a5c7508c6156d Author: Steven G. Johnson Date: Sat Aug 16 03:13:41 2003 -0400 typo commit 443c1d796f44813179c18d5b34c2836bd441a1e2 Author: Steven G. Johnson Date: Mon Jul 28 18:01:13 2003 -0400 use time() instead of clock() (FIXME: what to do for non-POSIX systems?) ...thanks to JP Sugarbroad and James A. Treacy for the bug report commit 9de40445c11cba14fa6eb7ae8e06d2792ebb2262 Author: Matteo Frigo Date: Thu Jul 24 18:58:10 2003 -0400 Need __volatile__ in sparc cycle counter. This is why the debian port hangs. commit 3da4cd5a3334e2b8415224657c5f5d9ce17eef12 Author: Steven G. Johnson Date: Sun Jul 20 16:02:43 2003 -0400 merged 3.0.1 notes commit cb6949db0271a00b15369efbd59d5c74af6d8c0c Author: Steven G. Johnson Date: Sun Jul 13 20:57:34 2003 -0400 whoops commit 906832308c9b7ba4ce3e55a3efa502ab55e6a676 Author: Matteo Frigo Date: Thu Jul 10 11:48:50 2003 -0400 Dealing with constants in a way that seems to confuse gcc less. commit e823de6c6b1433905ed2851aadcd1e11bf9b81b9 Author: Matteo Frigo Date: Wed Jul 9 17:39:23 2003 -0400 Enabled scheduler hack for FMA, where it seems to help. commit cff00fb9b000446f13c060876536184a03873ca5 Author: Matteo Frigo Date: Wed Jul 9 12:57:38 2003 -0400 Hmm---the new scheduler seems make things worse for gcc/x86, better for gcc/ppc, and about the same for icc/x86. Disabled for now. commit ca5556ba3956352649cd6d7342d16ffa660db23f Author: Matteo Frigo Date: Wed Jul 9 08:09:53 2003 -0400 New scheduling pass that keeps ``x = a + b'' and ``y = a - b'' close together. This property was no longer automatic for the dags generated in SIMD mode. I cannot measure any speed difference due to this change. However, the change is justified by a minimal-screwup argument. Moreover, the sse2 fftw library is now 1% smaller than it was before. commit 82ab8c034224f1364c148e9f725c90e6e16f721d Author: Matteo Frigo Date: Tue Jul 8 20:42:22 2003 -0400 -(FNMS()) => FMS() commit 471cc543a0805c19f676664861e5dc9d1fb5f1d7 Author: Steven G. Johnson Date: Sun Jul 6 13:53:23 2003 -0400 added more convenient target name commit 18303fef32c580b6c7dc03c4e8bfe30ca2a92724 Author: Steven G. Johnson Date: Sat Jul 5 13:30:10 2003 -0400 typo commit c30db69dc73fc965683cd147c948c3dba1f6aad7 Author: Matteo Frigo Date: Sat Jul 5 13:19:36 2003 -0400 Consistent naming commit da3b10c5645b3bb482b26cb42d0821dda213b994 Author: Matteo Frigo Date: Sat Jul 5 13:05:51 2003 -0400 Got rid of problemw. commit 21c3f87f755cadd9ef25945fe33448eeea7cb511 Author: Matteo Frigo Date: Fri Jul 4 06:56:26 2003 -0400 Increase TIME_MIN on intel only commit c749315d331fac4826ab1754a84f2a66e00197df Author: Matteo Frigo Date: Fri Jul 4 06:36:02 2003 -0400 A little hack to get more consistent scheduling. commit 1af463bd5b8ae6bd696ecb6a86d8b1952aaf6fd0 Author: Matteo Frigo Date: Thu Jul 3 16:47:42 2003 -0400 New experimental scheduler (currently disabled). The old scheduler is ``optimal'' in the sense that it minimizes register pressure. The only way to reduce register pressure is to schedule dependent instructions as closely as possible, so as to minimize the life time of registers. This strategy maximizes the number of pipeline stalls, however. With enough registers and short enough pipelines, this tradeoff is fine. This is no longer the case for the devilish pipeline of the Pentium IV or (probably) the PowerPC 970. The new scheduler switches to a ``list scheduler'' for dags smaller than a specified size. The list scheduler executes a butterfly left to right one column at the time. This amounts to the best possible pipeline utilization, and the worst possible register pressure. The ``specified size'' defaults to 0, i.e., no change from fftw2 and fftw-3.0. It seems like a value of 7--10 produces the best results for Pentium IV (probably screwing the G3/G4 powerpcs and sparc, but I haven't tried.) As time goes by, we may want to increase this number to favor newer processors over older processors. commit b4dc4ef39f37d036b9120f6e273549fd7bbaaa2a Author: Steven G. Johnson Date: Wed Jun 25 17:43:59 2003 -0400 remove non-portable use of tempfile; thanks to Nicolas Decoster for the patch commit 9630b97551ac80d0b1a8ecc26722ab98149a2be6 Author: Steven G. Johnson Date: Wed Jun 25 17:14:03 2003 -0400 increase stupid HP preprocessor limits commit a31916f0f41eb0a1398f38341da28991919123f0 Author: Matteo Frigo Date: Thu Jun 19 15:21:52 2003 -0400 Distribute gen_mdct.ml commit 367373d615cb66b335d9586544ead2f9d424effc Author: Matteo Frigo Date: Wed Jun 11 06:55:21 2003 -0400 Cleared int/ptrdiff_t confusions commit f1e0319906231b0e8d5675cc4ab24db00e560dc0 Author: Matteo Frigo Date: Tue Jun 10 22:15:42 2003 -0400 Cleared int/ptrdiff_t confusion commit 48d9ab9a8d3ddcb8fc21d4c8e90adfcbf7e4e200 Author: Matteo Frigo Date: Sun Jun 8 09:52:57 2003 -0400 Increased TIME_MIN. This seems to produce more reliable plans on Pentium IV. commit 3ba082c0f617d0c03cde783d3d7eba4392d13397 Author: Matteo Frigo Date: Sat Jun 7 21:43:00 2003 -0400 Removed relic -trivial-stores, which dates back to Franz's early experiments. Speed improved on SSE2, both with gcc and icc. commit 25a3b0e594d1b7fbe6b87c322ae34470bfdcccba Author: Steven G. Johnson Date: Thu Jun 5 22:29:52 2003 -0400 fix direntry commit 4ceff26934f86ada8712eabca96511462e7e8eaf Author: Steven G. Johnson Date: Thu Jun 5 13:41:34 2003 -0400 added imdct commit 1f23163e190c844ee3b2caf4564245cfbddd1c5b Author: Matteo Frigo Date: Wed Jun 4 19:54:38 2003 -0400 Collect pattern (a * b) +- (c * d) in generic-arith, because this operation can usually be computed with one rounding in fixed-point (and it possibly exposes a FMA instruction) commit 10d5f543e3ab8e1f2acb3f0bdeef0b196f22d6e0 Author: Matteo Frigo Date: Wed Jun 4 15:11:29 2003 -0400 Generic-arithmetic unparser commit d705a296a095ba4947adbd5e745b1a5ccf39f04a Author: Matteo Frigo Date: Sun Jun 1 09:05:30 2003 -0400 Oops---randomized CSE was using the same random numbers over and over commit fed2aa57dc95873ed83371338d16b9667c84c6f1 Author: Matteo Frigo Date: Sun Jun 1 07:01:17 2003 -0400 Paranoia. commit 4905e1d1498f044f37a6efd2b168c9bfac54a3de Author: Matteo Frigo Date: Sun Jun 1 07:00:54 2003 -0400 Use relative error instead of absolute error, to avoid problems when normalization factors are used. commit 967eecb8b745332dff25a3610a617bd73897abd6 Author: Steven G. Johnson Date: Sat May 31 22:11:28 2003 -0400 slight opt commit 6186c7e4fbaec8d816ce12a74c211890ad8c7fe9 Author: Steven G. Johnson Date: Sat May 31 22:10:45 2003 -0400 slight optimization commit f547dd3851ad0f88d4b79ac5b32af9ecbe727e4e Author: Steven G. Johnson Date: Sat May 31 20:43:31 2003 -0400 *W is const commit 9d9e1ec4a2623c32dd10304f84d3d3854b17b938 Author: Steven G. Johnson Date: Sat May 31 20:41:15 2003 -0400 comment commit 4269fae3e68941b115c0f8855ad3bb58e18c6f89 Author: Steven G. Johnson Date: Thu May 29 21:31:31 2003 -0400 added experimental MDCT commit 3028a550c6acc0fb367dbe26af192714715a09e2 Author: Steven G. Johnson Date: Wed May 28 22:01:37 2003 -0400 altivec (fma) needs simd codlist.c too commit 083e7c5e6bf88cffc3938cb6225551ec5af0f869 Author: Steven G. Johnson Date: Wed May 28 22:00:49 2003 -0400 make sure we include SIMD codlist.c for non-Unix folks commit 08028f9b0217ea164f25bcbf01824df23bb7245e Author: Steven G. Johnson Date: Tue May 27 20:31:25 2003 -0400 noted howmany_rank == 0 is a single transform commit 46b08adb7f478a6f5d51e1fefa6ab99891d3391d Author: Steven G. Johnson Date: Tue May 27 20:02:31 2003 -0400 further stride clarification commit 1264855f26ebf31705c1a6cf6067c3e996c71114 Author: Matteo Frigo Date: Mon May 26 10:21:22 2003 -0400 Removed transposed dftw problems. I now consider transposed dftw a Bad Idea, since it does not apply to the case that it was originally meant for (speed up four-step) and it complicates the implementation of the other thing I want to try (dftw m-slices). commit c198c3ed318d9864a22877795f7bac4f3eaf4d8b Author: Matteo Frigo Date: Mon May 26 07:22:59 2003 -0400 Obsolete comment commit bb96207501d8ca498bde638f46932c2bdd4bec66 Author: Matteo Frigo Date: Sat May 24 15:00:53 2003 -0400 comment commit a3733f2596b8d6d3972edfd942ce61296f2051d1 Author: Matteo Frigo Date: Sat May 24 07:20:35 2003 -0400 Oops---wrong test NO_UGLYP instead of !NO_UGLYP commit d8575658ab032ccf0b6553c4a84af8510d0cae55 Author: Matteo Frigo Date: Sat May 24 07:05:34 2003 -0400 Implemented radix r, where n=r^2 * p commit c2ca438ef110583287ce3c0e8527d4d382ccedde Author: Steven G. Johnson Date: Wed May 21 01:54:32 2003 -0400 xlc seems to properly use fma as well commit 3bf5cca5763573337c68978f05125e6e3eb4fe3d Author: Steven G. Johnson Date: Tue May 20 23:07:43 2003 -0400 print warning if there is no cycle counter commit 00cef912cbb09cf8a1080a3544d45d6bacc44222 Author: Steven G. Johnson Date: Tue May 20 17:32:04 2003 -0400 updated Funda reference commit a331d8a0841d1ccb7c314dbb98749bd77d69e709 Author: Matteo Frigo Date: Mon May 19 20:12:36 2003 -0400 const commit 200b3519c0ee6bd32babf52b352b8f8385da45d1 Author: Matteo Frigo Date: Mon May 19 15:41:09 2003 -0400 Implemented generic dif square transposed (q-style) solver. commit 8c6184ff5904082729018b5c36926b5bc479db14 Author: Matteo Frigo Date: Mon May 19 07:00:36 2003 -0400 applicable() is now a property of the solver (in anticipation of transposed solvers) commit 75102fd59a69589a2b23faeab596f43e7bcdc46c Author: Matteo Frigo Date: Mon May 19 06:33:40 2003 -0400 Slight cleanup commit fb49407efc496838a8f21a6cdfa52ad602b1cdfa Author: Matteo Frigo Date: Sun May 18 13:05:51 2003 -0400 Nothing, really commit df97ba383cdf4cc6a473e4e1ad041b11b8f965af Author: Matteo Frigo Date: Sun May 18 09:05:20 2003 -0400 Moved vector loop inside bytwiddle(), in anticipation of a q-style dftw-dit transposed solver. commit ee4edcdb0d229ae56167c04d4ffc36ee84d32361 Author: Matteo Frigo Date: Sun May 18 08:52:02 2003 -0400 Fixed flops count commit dccb90a9de4bd56cb5dad7340b27302bf2c6d245 Author: Matteo Frigo Date: Sun May 18 08:47:20 2003 -0400 style commit b83dec51bb0e588e9bc8fd4aff422b3e8b20e8ca Author: Matteo Frigo Date: Sun May 18 07:16:34 2003 -0400 Faster inner loop. commit c7bd9e9fc6d14ae3a28012fa0425a14330746572 Author: Matteo Frigo Date: Sat May 17 08:02:38 2003 -0400 Print vector length commit 74d6f9435c81aaafcf08927ebec58dd49ecea003 Author: Matteo Frigo Date: Sat May 17 07:55:33 2003 -0400 Oops commit 668b7018dd6ccd5a7802f71b29eb4e155801ff86 Author: Matteo Frigo Date: Sat May 17 07:50:35 2003 -0400 Allow vl > 1 commit 1919840986e87d611bccc5a1ffbaa469bccaa94a Author: Matteo Frigo Date: Sat May 17 07:01:42 2003 -0400 Radix can be derived from problem---no need to pre-specify it. commit a25c7ceebb0e65bb9b03bd7c30e3b97a29063185 Author: Steven G. Johnson Date: Fri May 16 22:50:50 2003 -0400 fixed comment commit e12663ca110ee43071e2344163161cd2122312c4 Author: Steven G. Johnson Date: Fri May 16 22:48:20 2003 -0400 whoops, gcd should be static commit 96c37858e236a75341bc45eca6e8a9a6180c61c8 Author: Steven G. Johnson Date: Fri May 16 22:40:32 2003 -0400 more unrolling commit ce888f64f2b19c2b1d0a2306c0f5a0bfb282e2cf Author: Matteo Frigo Date: Fri May 16 20:03:48 2003 -0400 Hack to avoid infinite recursion. commit 3be61e5d6f8b5c517feed2417902fc4fc8205180 Author: Steven G. Johnson Date: Fri May 16 19:52:43 2003 -0400 consistency commit 10ef200843d878ba4e88a6e59c53ec89b48cba11 Author: Matteo Frigo Date: Fri May 16 19:45:15 2003 -0400 Wrong comment. commit 9ad0f5ae45b2265e1307381d14a436eb43998c8d Author: Matteo Frigo Date: Fri May 16 19:45:03 2003 -0400 Style. commit bc609cdbd63dfd90154f4f0b07a0fb6c9c1f7354 Author: Steven G. Johnson Date: Fri May 16 18:35:27 2003 -0400 punctuation commit 3c931e88733a24d99d4639165f16f073ff25b35e Author: Steven G. Johnson Date: Fri May 16 18:33:45 2003 -0400 added allzero FAQ commit 4e3c1c97240d8dc8b8595f179994e9999e2a7b98 Author: Steven G. Johnson Date: Fri May 16 18:22:45 2003 -0400 simplification: instead of cldb, just use cldf with inputs/output values swapped commit 734444b8d502c323a04fd2cc6e6fc37d53cd4b04 Author: Matteo Frigo Date: Fri May 16 15:47:17 2003 -0400 Allow more general transform sizes. commit 77d47783b6445a14416a351ffb88c70348bb7ae1 Author: Steven G. Johnson Date: Fri May 16 14:22:37 2003 -0400 slight change commit 7bdfe7acb85d3e5c4081537789ee95c49d852e0c Author: Steven G. Johnson Date: Fri May 16 14:22:05 2003 -0400 MS has __int64 type, not long long (grr) commit 6778b57c92f2c729955fec383ee1ec817559b35f Author: Matteo Frigo Date: Fri May 16 13:34:16 2003 -0400 Fixed printout commit 26cad01414a0e9ec0f5809c3fafdcd0ebeca8eee Author: Matteo Frigo Date: Fri May 16 13:23:00 2003 -0400 Fixed flop count commit 6663b7b12eed820ccad08bbe8a7b13d4e7e96baf Author: Matteo Frigo Date: Fri May 16 13:02:06 2003 -0400 New bluestein solver commit a3444150521abc3e4522064bfbce259979d143dd Author: Matteo Frigo Date: Fri May 16 09:51:05 2003 -0400 Implemented generic radix. commit 105268f348df67714aaff107d5a4817a8ea9069e Author: Matteo Frigo Date: Fri May 16 08:19:38 2003 -0400 Removed conditional branch from inner loop in generic.c commit 44179d747df49fe429ae0108b108e1f28b71023c Author: Matteo Frigo Date: Fri May 16 07:48:28 2003 -0400 Simplified indexing commit 6a89bb8523df8e372f221f4ccdd6fa9e75120ec1 Author: Matteo Frigo Date: Fri May 16 06:53:56 2003 -0400 Better still. commit bc7126aa1f6bca65323f8d367629a9e6ddb18310 Author: Matteo Frigo Date: Fri May 16 06:24:31 2003 -0400 Further improvement of generic solver commit 29931919e62a0633afc7fdfe2738fba18419e30c Author: Matteo Frigo Date: Fri May 16 05:57:07 2003 -0400 Cleanup commit 516d81975ebfabe254800e61a072976ce2493792 Author: Matteo Frigo Date: Fri May 16 05:42:57 2003 -0400 Cleanup commit 9c40e9a8ab92ae76b96b1e57a51a8a46cd4202aa Author: Matteo Frigo Date: Fri May 16 05:31:40 2003 -0400 Generic now only works for odd sized. Added check. commit a4abb5b3c869ca3d4c1b572793c4128935461f2e Author: Matteo Frigo Date: Thu May 15 21:53:25 2003 -0400 Increased GENERIC_MIN_BAD because of new algorithm. commit 60bf38f4720ea99600008e1ad0772e3871cc975d Author: Matteo Frigo Date: Thu May 15 21:40:27 2003 -0400 Much, much better. commit bd0ae8b86cd6e44542f040b89670e8559cb4daef Author: Matteo Frigo Date: Thu May 15 21:25:00 2003 -0400 Still trying to understand why rdft-generic-dit is faster then dft-generic... commit 1ace458103964bbd9cc763efde55b6c64543e072 Author: Matteo Frigo Date: Thu May 15 21:04:33 2003 -0400 Nothing, really commit e68561ce796750faf2eb70606053723da3a3651b Author: Matteo Frigo Date: Thu May 15 20:59:45 2003 -0400 Never be clever for the sake of being clever. commit 3480d0dea4ccb5fa65ea2c8950aea4821cea6e3a Author: Matteo Frigo Date: Thu May 15 20:58:06 2003 -0400 Simplified. generic-dit is gone. The solver is now out-of-place only---buffering is done by the buffered solver. commit 21161d67e447696eda8fb463a6e629fdb9b9a286 Author: Matteo Frigo Date: Thu May 15 19:18:18 2003 -0400 rader-dit is gone. commit 99baac8e16f51413086aa8b35ff5894b43ddce25 Author: Matteo Frigo Date: Thu May 15 19:13:03 2003 -0400 Cast commit 9bc90955564668ef3b897434df873ea7a9e987b7 Author: Matteo Frigo Date: Thu May 15 19:09:07 2003 -0400 Introduced twiddle problem ``dftw''. Changed most other things to deal with this change. commit 57d761eab36c018f98849a04c82df5fdc61db498 Author: Steven G. Johnson Date: Thu May 15 18:47:18 2003 -0400 whoops, X(safe_mulmod) not fftw_safe_mulmod commit ad0f04aa3eead44c4f82e436a20241a8d76fdfba Author: Steven G. Johnson Date: Thu May 15 16:53:16 2003 -0400 add VC++ versions of asm commit 4e67675d7f66ce57718045e8ddf3769ba44f378f Author: Steven G. Johnson Date: Thu May 15 15:03:06 2003 -0400 VC++ reportedly supports the intel intrinsics, but requires __inline instead of __inline__ commit 8d72a4d25a56b9b8c9e918cb462ae7f8429fce9c Author: Steven G. Johnson Date: Thu May 15 14:32:06 2003 -0400 precompute array indices with VC++ commit 7a8ca44fd207d2de8947e6d7dad9b6122d5eacf4 Author: Steven G. Johnson Date: Wed May 14 21:57:39 2003 -0400 added doc note commit e21c7e8fbc7073874cec2c052a810cb8c8bafb6c Author: Steven G. Johnson Date: Wed May 14 19:45:54 2003 -0400 autodetect windows commit 9e15f067241ae365258fdae039f13855799566b5 Author: Steven G. Johnson Date: Wed May 14 15:08:49 2003 -0400 don't bother with #ifdef HAVE_CONFIG_H, since non-Unix users always forget to define it commit 8fd89cce099546af6bb94f83b4e84bc46609708a Author: Steven G. Johnson Date: Tue May 13 16:58:07 2003 -0400 VC++ uses __inline commit 02aaa87cb911b0d6d67ec4f11932357f702aa75f Author: Steven G. Johnson Date: Tue May 13 14:51:26 2003 -0400 added leak question commit 64d02f177161f96e87c02cde6015ceff42ac0bfe Author: Steven G. Johnson Date: Mon May 12 18:26:51 2003 -0400 LARGE_INTEGER needs windows.h (supposedly, there is some problem converting _itnt64 to double...damn MS and their nonstandard types) commit 342ab9adfdb7bf9a5936f1c1d3f0820fa729ab2e Author: Steven G. Johnson Date: Mon May 12 18:22:16 2003 -0400 whoops commit a8cb5339f580f0f4ee0ff02c58f91036b7fef886 Author: Steven G. Johnson Date: Mon May 12 17:16:19 2003 -0400 added 256x256 to canonical list commit 446260f6a6a84986c7eec6e1b3c0eea0f66f759d Author: Matteo Frigo Date: Mon May 12 07:02:06 2003 -0400 Oops... commit 924714d15f5ed063b1fa8a40a3bcd2ebe406e572 Author: Matteo Frigo Date: Sun May 11 11:04:46 2003 -0400 Unrolled loops, changed cutoff commit a8e681a4dd1d8cbae25becdb745926efde43cf3c Author: Matteo Frigo Date: Sun May 11 10:20:04 2003 -0400 Do not multiply strides by 2 twice. commit cba6f4731943edfe50f4bacd9de28d0551593f43 Author: Steven G. Johnson Date: Wed May 7 21:09:43 2003 -0400 added 'make smallcheck' commit 8f61201655a4d04b402d2cec34acb86e89cbe35f Author: Steven G. Johnson Date: Wed May 7 20:46:10 2003 -0400 --without-cycle-counter becomes --with-slow-timer, updated docs commit e77df17a33148f0cd531fcf1bcf371af0b69ced6 Author: Steven G. Johnson Date: Wed May 7 18:05:29 2003 -0400 remove duplicate -openmp check; Sun requires -xopenmp commit 34594ee9cdef76091eff1164f9518e92bf0b855b Author: Steven G. Johnson Date: Wed May 7 17:59:23 2003 -0400 fixed compilation under Sun C++ commit fe5788275ebad911c952662c50694c2f296ae4b2 Author: Matteo Frigo Date: Wed May 7 14:24:46 2003 -0400 Use estimator if cycle counter is unavailable, regardless of the FFTW_MEASURE/ESTIMATE setting. commit 7d2473af5f98cac96951e94c009b96f1d96dcea3 Author: Steven G. Johnson Date: Tue May 6 23:15:34 2003 -0400 _WIN32 (not __WIN32__) is always defined commit a9d4840add6bac6ef0f0f194868861615f52d91c Author: Steven G. Johnson Date: Tue May 6 23:11:52 2003 -0400 minor cleanup commit 33d6678021b7c966f0490729f9839282d88b7e68 Author: Steven G. Johnson Date: Tue May 6 22:50:07 2003 -0400 tentative VC++ stuff, some consolidation commit 4b2c8b1fdf43a442999e9309b1abb5f525b2a5e2 Author: Steven G. Johnson Date: Tue May 6 12:17:56 2003 -0400 made cycle.h more self-contained commit 17b78ccd6e3bf378453a85f671f02c82eb45ced4 Author: Matteo Frigo Date: Tue May 6 08:30:39 2003 -0400 Use ``%'' flag to denote commutative operations. commit b98342890cbb95b8f450d6da8e22637b455fccc0 Author: Steven G. Johnson Date: Mon May 5 20:42:30 2003 -0400 MIT license, brief documentation commit 95f79870876aa95354fecf59c0da025edb8982f9 Author: Steven G. Johnson Date: Mon May 5 20:31:16 2003 -0400 whoops, forgot f77_wisdom.f commit 5cdde47077a07f4aa39487741662e868f5cf11df Author: Matteo Frigo Date: Sun May 4 19:37:09 2003 -0400 Improved speed of accuracy test. commit ff00ccc1b3a83962c18e91ab12740ca63bbb6081 Author: Matteo Frigo Date: Tue Apr 29 11:45:34 2003 -0400 s390 cycle counter commit 2ab4e6e2c3bfdd6882bad8bca36fdc105f742847 Author: Steven G. Johnson Date: Sat Apr 26 12:26:15 2003 -0400 forgot r2r directory commit 990abcc219adbf0759807b9b2c20d80f639dd940 Author: Steven G. Johnson Date: Fri Apr 25 20:52:23 2003 -0400 delete unused files, since they don't compile any more commit a35c9fd7e587b8301131f8e530387dad7e62fc00 Author: Matteo Frigo Date: Thu Apr 24 06:37:41 2003 -0400 Better gcc code generation commit 462d92a21265012fd4fb89326da6bccd69f05406 Author: Steven G. Johnson Date: Wed Apr 23 15:30:50 2003 -0400 ccc is the Compaq C compiler on Linux/alpha commit a97d5f6b3b451179f501157bfe6fafde3481ea0d Author: Steven G. Johnson Date: Wed Apr 23 00:06:03 2003 -0400 whoops commit a53aa0afb9f63f64ee6235e07fd99014f6da32fb Author: Matteo Frigo Date: Sat Apr 19 09:18:25 2003 -0400 ia64 cycle counter with intel compiler. commit 20978a6bcaad9c07b4c969eae56ab29ae092e2bb Author: Matteo Frigo Date: Fri Apr 18 18:27:30 2003 -0400 More gcc bugs. Sigh. commit 1a9f1a74c640a09efbabff5043cc7074b6bfefe0 Author: Matteo Frigo Date: Fri Apr 18 18:01:49 2003 -0400 touch ChangeLog to observe GNU standards commit 4241f03ed18cc0acf61072c70b9c38b1c7dc7c31 Author: Matteo Frigo Date: Fri Apr 18 18:01:12 2003 -0400 We now build ChangeLog automatically at distribution time commit 1797417a3e517b1f9b9931e49797dff29ae760da Author: Matteo Frigo Date: Fri Apr 18 18:00:17 2003 -0400 Automatic ChangeLog hackery commit d76cd97496030b79d3450f2c6de88da7b4458bd6 Author: Steven G. Johnson Date: Fri Apr 18 13:25:26 2003 -0400 plural commit 6c59528dcbd6ce3676a0852bad2d909e68fdeeac Author: Steven G. Johnson Date: Fri Apr 18 13:25:01 2003 -0400 updated commit b37ba8f3518f8be4c0c25dd86f102c9e64527104 Author: Matteo Frigo Date: Fri Apr 18 12:59:41 2003 -0400 Updated commit 61ac8832cb1bd238132fe475c699f12e11232781 Author: Steven G. Johnson Date: Fri Apr 18 11:48:39 2003 -0400 a -> an commit 287a97f4366aeb55745345972896f06e378dcacf Author: Steven G. Johnson Date: Fri Apr 18 11:47:56 2003 -0400 hyphen commit 8791b19a3135636c10dd4e99695799a6ac315870 Author: Steven G. Johnson Date: Fri Apr 18 11:47:28 2003 -0400 comma commit ad823e8abbe446885056a5fc05555e295b7a174f Author: Steven G. Johnson Date: Fri Apr 18 11:46:59 2003 -0400 minor commit 7a09c0d18816d7d1c2ed89a0dfbd62843d3442eb Author: Matteo Frigo Date: Fri Apr 18 10:37:31 2003 -0400 Updated commit 13381c1ba5e8af34a8bb03710aa7741a8040cc07 Author: Matteo Frigo Date: Fri Apr 18 10:14:59 2003 -0400 New script that builds the distributions commit 459fa38ea6735a02e430cb0ecbca2323e07cef57 Author: Matteo Frigo Date: Fri Apr 18 08:51:07 2003 -0400 Oops again commit e57ee39dc083cd4461a548cded7fdc45e57fc74f Author: Matteo Frigo Date: Fri Apr 18 08:39:05 2003 -0400 Oops, forgot -sign 1 commit 396b6bc876a10a52ce9dc68230e2eb25af544f29 Author: Matteo Frigo Date: Fri Apr 18 08:28:25 2003 -0400 Reorganization of simd codelets commit fd7cb9b51d78aabe5b32969758bf472ca0d563ae Author: Matteo Frigo Date: Thu Apr 17 21:21:45 2003 -0400 k7 assembly was not updated after conversion of opcnt from int to double commit 3bf64e2b77975db0ec3c2ad2232fa9dfceeae35f Author: Matteo Frigo Date: Thu Apr 17 19:15:53 2003 -0400 Capital `X' looks bad in all-lowercase plans commit 448802e951a73d406bacc449b5d7eb0ece3bfaf6 Author: Matteo Frigo Date: Thu Apr 17 18:53:29 2003 -0400 Removed redundant inline/noinline codelets commit b72b1f8cdf96f12c6776efdb456cdd6e2eff00b3 Author: Matteo Frigo Date: Thu Apr 17 15:25:50 2003 -0400 New noinline Noinline real codelets commit 4f5ff427ae9e39bda6b17cdc61b9ecfad075f322 Author: Steven G. Johnson Date: Thu Apr 17 15:23:03 2003 -0400 more ideas commit e549828bdf29f4ba5b352f8d54e1d8fb65e86b0b Author: Matteo Frigo Date: Thu Apr 17 13:18:45 2003 -0400 Removed duplicate rules. commit 29889996e0d7e2beec6c4afaf8d06936f0bc8e7a Author: Matteo Frigo Date: Thu Apr 17 10:51:09 2003 -0400 acx_pthread.m4 was not distributed commit 4bcda610088022777266a9d4723e3108871a2382 Author: Matteo Frigo Date: Thu Apr 17 07:21:17 2003 -0400 Oops commit 92ba4bfacade8ab1d4dae0e256cda08e8b22eb3c Author: Matteo Frigo Date: Thu Apr 17 07:07:19 2003 -0400 Both inlined and non-inlined notw codelets. commit bc5fcf6d524989489b08f05b59fd2660b4331765 Author: Matteo Frigo Date: Thu Apr 17 06:44:21 2003 -0400 Initial experiment with both inlined and non-inlined simd codelets. Both are included for now. commit 5586bdcd9d791a373355fae20e4df01e8b51ef32 Author: Matteo Frigo Date: Thu Apr 17 05:57:36 2003 -0400 --enable-fma to build FMA distribution commit 6719f26d34cd9bd0dce5a3d279a06b113cd774a7 Author: Matteo Frigo Date: Wed Apr 16 17:21:53 2003 -0400 Inline SIMD nontwiddle codelets commit 5db4d2ae77f5600008ce54b88e93a0e7fbcc649b Author: Matteo Frigo Date: Wed Apr 16 16:18:29 2003 -0400 Pathetic attempt at saving a couple of registers... commit 7267a94d763a0380970f1b07ee84aad71f138c8f Author: Matteo Frigo Date: Wed Apr 16 15:51:27 2003 -0400 for (i = 0; i < m; ++i) ==> for (i = m; i > 0; --i) No proof of evidence that this is any faster, but just in case... commit 607d75d99f7b0c558d5664cc18b0c6a070d3aa02 Author: Steven G. Johnson Date: Tue Apr 15 15:03:20 2003 -0400 added hack to make sure that codelet loops are preferred to vecloop solvers in the estimator commit a772926574ae28c851b95b9eea8d22d0b244e25e Author: Steven G. Johnson Date: Tue Apr 15 14:53:44 2003 -0400 use double for flops commit 84c191f258a2ee1fde8b39e4a567f48bc84d273a Author: Steven G. Johnson Date: Tue Apr 15 14:51:50 2003 -0400 metrowerks reportedly supports gcc assembly extensions on ppc commit d83a8e3d6b322b293fb8b048bb46c1048faf430b Author: Matteo Frigo Date: Mon Apr 14 15:00:50 2003 -0400 foo_CFLAGS generates some automake junk that breaks the build on Redhat 7.3. Screw it. commit 54128f6b2dd2ea009736debbb0c2eee43c4a0ade Author: Matteo Frigo Date: Mon Apr 14 12:22:59 2003 -0400 Carefully check return status commit 8935dbb4a98ff9a7780aa1ce7a98c656b7284d79 Author: Matteo Frigo Date: Sun Apr 13 16:46:12 2003 -0400 Removed annoying -FMA() expressions. commit 8ce45c13aca6d081f84f8dcb84a973383d3e5ee8 Author: Matteo Frigo Date: Sat Apr 12 14:32:22 2003 -0400 Major fma hackery commit 4a294df6343206e437a3f24ed268a71c9cd9edb0 Author: Matteo Frigo Date: Sat Apr 12 14:25:43 2003 -0400 Slight cleanup commit bda29baca81b8098e25fb1d61d3aa7b882f33ef5 Author: Matteo Frigo Date: Sat Apr 12 10:04:51 2003 -0400 Updated version number commit 80a70d1b1e8e1d7e336301a3f65b26d153ba15a7 Author: Matteo Frigo Date: Sat Apr 12 08:03:07 2003 -0400 Damn autoconf commit 27c1334aec2b8b3bbc7e679761d0ddab53212487 Author: Matteo Frigo Date: Sat Apr 12 07:54:20 2003 -0400 Recognize all 74xx processors commit 7b1c87d4184dace0460075da82295c1fc90e523d Author: Matteo Frigo Date: Sat Apr 12 07:35:17 2003 -0400 Detect 7400 processor. commit ea19ce217870bff790c8e91865228328a08cd769 Author: Matteo Frigo Date: Fri Apr 11 20:42:11 2003 -0400 No need to check for gcc-2.95 commit fb6560399e6c1dc7382fc48209545f4494f87e65 Author: Steven G. Johnson Date: Fri Apr 11 16:14:39 2003 -0400 removed duplicate commit 114c5faeddef91d05efc3af062e58f21879cd77d Author: Matteo Frigo Date: Fri Apr 11 08:45:37 2003 -0400 mflops ==> ``mflops'' commit be09e2c160458ffa571c8c207546d77dd86f1022 Author: Matteo Frigo Date: Fri Apr 11 07:00:53 2003 -0400 Print setup time as well commit a55b833ea382a9dcf478517c83d68d044db4dcaa Author: Matteo Frigo Date: Thu Apr 10 15:36:18 2003 -0400 Enforce pointer equality for in-place problems. commit a920de28e41596cc862a7f7bcc14c27c68920b53 Author: Steven G. Johnson Date: Wed Apr 9 17:47:54 2003 -0400 updated commit 080c6f9a24e129da6a216d44e581ea9c5012df83 Author: Steven G. Johnson Date: Wed Apr 9 14:53:38 2003 -0400 cross-ref fftw-wisdom man page commit 920b026d6c76cc8b23b877e8238cb6cd9f71e7af Author: Matteo Frigo Date: Wed Apr 9 10:13:00 2003 -0400 Undone previous change, committed by mistake. commit c636a6d32d90fda78f15b6e5b0060feeea45b47c Author: Matteo Frigo Date: Wed Apr 9 10:12:24 2003 -0400 Quick and dirty README for bench commit e542ee15fcc1eaa62b3c1ea6a58cc3bc468ff599 Author: Matteo Frigo Date: Wed Apr 9 08:50:25 2003 -0400 Consider additional command-line arguments as problems to be benchmarked. commit 2196cad41694a580854e0f04991d3055d9e365e3 Author: Matteo Frigo Date: Wed Apr 9 08:44:13 2003 -0400 Default report format is now human-readable. Removed unnecessary complexity in benchmark reporting. commit 1ecc5f59fb401c7ecdeb54596088ac86ede3639a Author: Matteo Frigo Date: Wed Apr 9 06:10:40 2003 -0400 Updated for new interleaved/split api. commit 8d3536f2476441b3f7754a1d5298102d6cd14ac2 Author: Steven G. Johnson Date: Wed Apr 9 03:01:03 2003 -0400 updated citation commit aaf6b0e66924b89fb8c927bbcc3bc655512aab13 Author: Matteo Frigo Date: Tue Apr 8 19:35:59 2003 -0400 Time for beta3 commit f5c162435d5334990a7c6f3421191592a831bdd7 Author: Steven G. Johnson Date: Tue Apr 8 17:40:59 2003 -0400 whoops, added commit 14108da14b0738a1c8b036a901971104d5374ef7 Author: Steven G. Johnson Date: Tue Apr 8 17:33:47 2003 -0400 more comparison of different R*DFT types commit 75bbef9a658de085d8c21952d597dbca9e8fa722 Author: Steven G. Johnson Date: Tue Apr 8 16:48:08 2003 -0400 comments commit 0ba9318b95b11d2cb5470ebcf73a1819d31caafb Author: Steven G. Johnson Date: Tue Apr 8 16:19:39 2003 -0400 more accurate DCT-I and DST-I, at the expense of up to a factor of 2 in speed and memory commit 404912f3cee76dd7a6b3928798304c64c55129bf Author: Matteo Frigo Date: Tue Apr 8 05:38:09 2003 -0400 Workaround gcc/sparc bug commit f34586737c6b234a5e2316dddbbe425331d4a5b3 Author: Steven G. Johnson Date: Tue Apr 8 01:34:12 2003 -0400 rumors commit 78f9ac7bedec0c22cb7fc5fccdbb2adc8b8c4f3b Author: Steven G. Johnson Date: Mon Apr 7 18:54:11 2003 -0400 added rdft2 paranoid mode commit 55b03e7ef8ab680294bbbe73059cb32cef82ac65 Author: Steven G. Johnson Date: Mon Apr 7 18:47:37 2003 -0400 added paranoid mode for r2r commit 8361bbcd9549f5c5819b31ce44c2e733e185f1a5 Author: Steven G. Johnson Date: Mon Apr 7 15:10:08 2003 -0400 whoops, sincos is predefined on some systems commit 0e1c9cf517e0d4010af5a35d10d141b2d74f7d0e Author: Matteo Frigo Date: Sat Apr 5 16:50:57 2003 -0500 bp->destroy_input was not initialized commit 7b13a4d8dfbf05ed699e7a1a533e7edf9355e2bc Author: Matteo Frigo Date: Sat Apr 5 09:29:11 2003 -0500 Asserted correctness conditions for tainted pointers. (For now, use CK() while we test. They should be changed into A() at some point.) commit 6b16dfb3e11fcfa65d2064af8ee1c66f0e0ba2a2 Author: Matteo Frigo Date: Sat Apr 5 08:18:23 2003 -0500 Untaint pointers before zero'ing arrays and before hashing commit 1495e7c627b502d282f4fb290357d6fe573e12dd Author: Matteo Frigo Date: Sat Apr 5 07:11:56 2003 -0500 Alignment check did not work with icc, which seems to be confused by the fact that the variable is not used. commit e013a83a5d1fb9e98a7d035baa3e4c5a89f768f1 Author: Matteo Frigo Date: Sat Apr 5 06:41:20 2003 -0500 More paranoid paranoid-check commit 377dd60fc68c62595e87882b3a5caede183a1251 Author: Matteo Frigo Date: Sat Apr 5 06:19:25 2003 -0500 0 == x & 7 parses as (0 == x) & 7, which is wrong commit 109ea5550e130de0017dcdf2665c0872ff776e98 Author: Steven G. Johnson Date: Fri Apr 4 21:35:49 2003 -0500 alignment checks commit 97c940cb1b38b75289d99b2206d77a13db97f6fb Author: Steven G. Johnson Date: Fri Apr 4 21:04:14 2003 -0500 prevent infinite loops in exhaustive planning commit 20a0d16cc3bf2a26e0e5b2ed4f1691c617ab676e Author: Steven G. Johnson Date: Fri Apr 4 20:58:20 2003 -0500 split/unsplit guru interface commit ec77ade6672c46fde379e806bb25f26569eaa874 Author: Matteo Frigo Date: Fri Apr 4 20:39:55 2003 -0500 Need UNTAINT in verifier too. commit 2fd0ede87b74bd6a6b325910eb68ce55ee69ae66 Author: Matteo Frigo Date: Fri Apr 4 19:36:46 2003 -0500 Forgot #if HAVE_SIMD commit 0509bf08e01662eb716c8038093caab6bbbe867e Author: Matteo Frigo Date: Fri Apr 4 19:30:37 2003 -0500 Keep track of two separate taint bits commit dc9c49340a92349a47d46befc4f85937b95911c5 Author: Steven G. Johnson Date: Fri Apr 4 19:16:32 2003 -0500 added NO_SIMD problem flag, made UNALIGNED an API issue (taints input pointers) commit 1b8c8e9e88fa0f39226f2a8a853a07719d4faf40 Author: Steven G. Johnson Date: Fri Apr 4 18:14:14 2003 -0500 bugfix in buffered: wrong pointers passed for cldrest; also use TAINT instead of UNALIGNED in buffered2 commit 99fccbefe65b6e39c9b4e9be40e647facb900b4f Author: Matteo Frigo Date: Fri Apr 4 17:19:51 2003 -0500 Reverted previous change, committed accidentally commit db841c316cbd202532fd5b428396e5f4e9c74cf0 Author: Matteo Frigo Date: Fri Apr 4 17:18:39 2003 -0500 What was I thinking? commit 6235b967442a2150ad1e7100ae39070ecbee0ca9 Author: Matteo Frigo Date: Fri Apr 4 17:18:21 2003 -0500 [empty commit message] commit 4d690f88fdc36870e18b359db01ee23dbc005d72 Author: Steven G. Johnson Date: Fri Apr 4 16:48:32 2003 -0500 added --enable-debug-alignment commit 8890a79f285088b2b04ca1c2db939a582ac0328e Author: Steven G. Johnson Date: Fri Apr 4 16:29:43 2003 -0500 X(taint) prototype, define corresponding function only if HAVE_SIMD commit 3f29be3cc8a63846e725e496ae01474af84ab9fc Author: Matteo Frigo Date: Fri Apr 4 16:15:53 2003 -0500 Initial checkin of tained pointers commit faad01bdd384c083438df8ef016b8a18804cb72a Author: Matteo Frigo Date: Fri Apr 4 13:12:58 2003 -0500 More conservative preservation of alignment commit ac40b45c34f80bd09d25405935c3722528595a97 Author: Steven G. Johnson Date: Thu Apr 3 23:16:27 2003 -0500 plan/execute with aligned stack commit 978b7f409d31dde15736857998dada802a3ef49c Author: Steven G. Johnson Date: Thu Apr 3 15:40:01 2003 -0500 whoops, missed FFTW_MEASURE in fftw3.f commit 3274f607dea4bbf61b89f1d09703054007b28776 Author: Steven G. Johnson Date: Thu Apr 3 13:44:46 2003 -0500 use WITH_ALIGNED_STACK for experimental semaphore stuff, too commit 345df91b8bab3216268ca697850bc00767799265 Author: Matteo Frigo Date: Thu Apr 3 09:04:23 2003 -0500 Removed old file commit a36ad0e14ceb92c16cdc3bec19938c53c69f79a5 Author: Matteo Frigo Date: Thu Apr 3 07:50:43 2003 -0500 Improved stack-alignment hack commit c4f4e2d7d432203a0f99e3a50b29168c3d653a83 Author: Steven G. Johnson Date: Thu Apr 3 02:37:57 2003 -0500 use aligned stack for experimental semaphores, too commit cea2d48a884c03d448b9c688a192081e3e984983 Author: Steven G. Johnson Date: Thu Apr 3 02:17:58 2003 -0500 whoops commit c3bdcb8375c4c5181c0c642b0331d2a7268757f8 Author: Steven G. Johnson Date: Thu Apr 3 01:58:32 2003 -0500 fix(?) for SIMD thread problems commit 13dde386673933410cafa316f241cdc6544ecd65 Author: Steven G. Johnson Date: Wed Apr 2 20:33:12 2003 -0500 noted n=1 REDFT01 case commit 4a2c5556d9c6d080f3c3fee8c87d0aee50c12531 Author: Steven G. Johnson Date: Wed Apr 2 20:32:07 2003 -0500 note about n=2 REDFT00 formula commit ac5fe8c3ecce10f2e79f84279ce6e406db3891e8 Author: Steven G. Johnson Date: Wed Apr 2 20:30:10 2003 -0500 note about undefined REDFT00 commit 4761b3e61b5cf393deeacf6eba73d9f3a35e2d12 Author: Steven G. Johnson Date: Wed Apr 2 20:18:03 2003 -0500 noted n=1 RODFT01 case commit 00cd3721f0f757f691e62c836aff445fece4a9ef Author: Steven G. Johnson Date: Wed Apr 2 20:14:07 2003 -0500 corrected definitions commit 20545fe3112a9aa8bd9529129f24586a66f39f9d Author: Steven G. Johnson Date: Wed Apr 2 19:43:59 2003 -0500 added REODFT_KINDP, fixed nontrivial test for R2HC11 and HC2R11 (not that we support these yet anyway) commit dcd456710f59aea75abb5a4b62ad7b8c8592c28f Author: Steven G. Johnson Date: Wed Apr 2 19:16:54 2003 -0500 size 2 hc2r and dht are equivalent to r2hc commit dd3db55bba543cc4db74f3760716a251892089a9 Author: Steven G. Johnson Date: Wed Apr 2 15:09:08 2003 -0500 noted overwriting in upgrading section commit 4752fd3dcd81d75371bc667be6ab701ee36a24d3 Author: Matteo Frigo Date: Wed Apr 2 05:25:56 2003 -0500 Moved with_aligned_stack to its own file commit 821f37e9c6396afa7dcf22eae25e2ddb56f16218 Author: Matteo Frigo Date: Tue Apr 1 21:11:31 2003 -0500 Fixed comments commit 44b77936443c9dcbba1ccf21d3e90c2426a46e01 Author: Matteo Frigo Date: Tue Apr 1 20:57:39 2003 -0500 Alignment hacks commit 52974f9347f673ccfc5eca68ed2af2e39f0ae148 Author: Steven G. Johnson Date: Tue Apr 1 14:26:48 2003 -0500 phew, no, previous version was okay commit f599fa6d8cb159f0e636411e51f0bd07feca296e Author: Steven G. Johnson Date: Tue Apr 1 14:26:15 2003 -0500 whoops, crap commit da939ebd27d69c1e3693ebf71f81060e816af54e Author: Matteo Frigo Date: Tue Apr 1 08:01:06 2003 -0500 support sse2 in forthcoming gcc-3.3 commit bad66cbc1963d1beecba1205ff4d528026003427 Author: Steven G. Johnson Date: Tue Apr 1 01:17:15 2003 -0500 comment commit bde4d633afdc37f663c31f7aa2c4b3f8673e9607 Author: Steven G. Johnson Date: Tue Apr 1 01:16:46 2003 -0500 noted ac_check_headers commit 0e70968689aed47b11b44eb15752b97e21534366 Author: Steven G. Johnson Date: Tue Apr 1 01:11:31 2003 -0500 comment commit 716a92cca66059e083cc6dc764db18de707a6318 Author: Steven G. Johnson Date: Tue Apr 1 01:06:53 2003 -0500 documented autoconf tests, so that cycle.h can be distributed separately commit a081cb59d2fbd65042f4a1cec68ec04698a03594 Author: Steven G. Johnson Date: Mon Mar 31 22:12:02 2003 -0500 IRIX is all-caps commit c21fee75eade0b3c38780e252bb0dbe24383d2cf Author: Steven G. Johnson Date: Mon Mar 31 22:11:42 2003 -0500 noted Irix fix commit 0521214bc9e4b224ee18c31c165c8971d5d09fc6 Author: Steven G. Johnson Date: Mon Mar 31 22:10:33 2003 -0500 whoops commit e42bd5b20e10661a6cd8228c3b19ee7d8f1a1602 Author: Steven G. Johnson Date: Mon Mar 31 22:04:35 2003 -0500 use ithreads_init so as not to confuse fftw 2 users commit 95c74b4b2e3ef14b347ae7b50fdb455d6a7aa719 Author: Steven G. Johnson Date: Mon Mar 31 22:00:42 2003 -0500 IRIX lossage commit 746ced9c6d3a5d53c7b95090cbb99fcfd0b07344 Author: Steven G. Johnson Date: Mon Mar 31 21:19:20 2003 -0500 check for -openmp (icc) among the OpenMP flags (TODO: make this a separate macro, with a loop instead of repeated checks) commit 589adf0e340eafbabdd43f5beacae6740e9e64a0 Author: Steven G. Johnson Date: Mon Mar 31 17:12:19 2003 -0500 clarification commit 4242c9c8bf63111190cbcccd162a224af036e5af Author: Matteo Frigo Date: Mon Mar 31 17:01:16 2003 -0500 More liberal test for solaris CC commit d5928079a514ffaba5eedc2cd5ce76eb2dd9fa9b Author: Matteo Frigo Date: Mon Mar 31 15:13:33 2003 -0500 Allow x86-64 simd commit b7a2252e112c67968e6695c7ef13e375a04d23d1 Author: Matteo Frigo Date: Mon Mar 31 15:13:21 2003 -0500 Added x86-64 timer code commit 7f0d1b516cd025f72f304fdeb210c563b94bff31 Author: Steven G. Johnson Date: Mon Mar 31 13:10:54 2003 -0500 updated commit 914e74201ca244b441f7f971d2f05aced6aa405c Author: Steven G. Johnson Date: Mon Mar 31 13:07:19 2003 -0500 updated commit 202febc5ba6f89ad6e834d4e36a01caf4bb5fde2 Author: Steven G. Johnson Date: Mon Mar 31 13:05:27 2003 -0500 colon commit 2deea3231269a3e4bcbdfa9498ad253ad1b26a48 Author: Matteo Frigo Date: Mon Mar 31 07:20:20 2003 -0500 Reorganized compiler bugs section (which is growing out of control) commit b4bb5597a0d941eeefe4ec01208c139d37e9fce2 Author: Matteo Frigo Date: Mon Mar 31 07:15:20 2003 -0500 solaris gcc bug appears to be also in 2.95.2 commit efb7874ecc58a7e086abf8428d481a6e19c4e0d7 Author: Matteo Frigo Date: Mon Mar 31 07:13:45 2003 -0500 Workaround works---there is another gcc/sparc bug elsehwere commit 8ab897ec05303f83b56d2e349c3dba59da173ef2 Author: Matteo Frigo Date: Mon Mar 31 07:08:56 2003 -0500 Grrr, workaround does not work. commit 32707cc1247ff03834c3d37fceb57f53e268da65 Author: Matteo Frigo Date: Mon Mar 31 07:02:23 2003 -0500 ADDMOD is now function, which seems to avoid gcc bugs. commit c4c605027021db9b801e3e2695c802ed6e1bc44a Author: Matteo Frigo Date: Sun Mar 30 16:40:26 2003 -0500 Workaround sparc gcc bug commit b77fba4459439cf3d969088c9edb010ab151a893 Author: Steven G. Johnson Date: Sun Mar 30 15:51:59 2003 -0500 note commit 91e398fb21a1c46fac8174a5f2faf0b79548e188 Author: Steven G. Johnson Date: Sun Mar 30 15:34:57 2003 -0500 make non-square UGLY, for now commit 4233309534b8e309bce0dafeeff64c29ac9f4b1c Author: Steven G. Johnson Date: Sun Mar 30 15:33:57 2003 -0500 added -o amnesia to forget_wisdom before each plan commit 055907acc9ab6486266e2601f13e76e768bd990f Author: Matteo Frigo Date: Sun Mar 30 09:41:27 2003 -0500 Report setup time in benchmark commit 6a49d54d587cc678c9a4063e3ed620c998d2602e Author: Steven G. Johnson Date: Sat Mar 29 20:21:15 2003 -0500 comment commit 1ccc921a0398eb08789ac928e28840a524100587 Author: Steven G. Johnson Date: Sat Mar 29 19:11:10 2003 -0500 slight change commit ee2cf222eef51c2ee38a761765c58ed6a2faa35e Author: Matteo Frigo Date: Sat Mar 29 18:46:16 2003 -0500 More relaxed definition of UGLYness commit 2afbef3c1cd1edca0168bc5341dac85de41790ba Author: Steven G. Johnson Date: Sat Mar 29 15:28:01 2003 -0500 no more cvs id strings in header files...I'm tired of having to rebuild everything after a commit commit 6922449e07c11f405107d7e5fc63d7dfb0379b5d Author: Steven G. Johnson Date: Sat Mar 29 15:22:28 2003 -0500 rdft2 stride unification commit 383f9ebcd63e13f756a57f0801b6bdc4080f4887 Author: Steven G. Johnson Date: Sat Mar 29 14:38:23 2003 -0500 preserve in-place-ness commit 4989fd02c94baef2f163547b88f643fcd1172a72 Author: Steven G. Johnson Date: Sat Mar 29 14:23:31 2003 -0500 make nowisdom the default commit a2f08dcbee1508f40df179ca67ed0ddcefd66f37 Author: Matteo Frigo Date: Sat Mar 29 14:13:18 2003 -0500 --verbose in paranoid-check produces too much output. Make it quiet. commit 90cdd14a2b342236ae6e8367d94ad2e29ecadd76 Author: Steven G. Johnson Date: Sat Mar 29 13:45:13 2003 -0500 fixed transpose bugs...need to check ri-ii before deciding whether Ntuple fits commit 71fc37fa553e50623f56a5fb21731833cb2d9dfd Author: Matteo Frigo Date: Sat Mar 29 08:10:40 2003 -0500 try more 2^k commit d4e0d59380ec69df5a4250ebd0f62f002c964e19 Author: Matteo Frigo Date: Sat Mar 29 08:05:41 2003 -0500 MIN_ALIGNMENT was defined after being used, causing crash in sse2. commit 101331222a4ff1189042a5997260a7e171ae1136 Author: Steven G. Johnson Date: Sat Mar 29 03:07:34 2003 -0500 real transposes are currently unused, and are not needed for MPI code either commit 55b24758612593bac4f6b7065d32b33b815eb81f Author: Steven G. Johnson Date: Sat Mar 29 02:58:39 2003 -0500 added general transpose commit c111a90447eb6c3c1a0058a93e97557beaaf1605 Author: Steven G. Johnson Date: Fri Mar 28 22:49:04 2003 -0500 added transposition option commit 90fb8971b2092e9a1fa97b10065683ba8af9247c Author: Steven G. Johnson Date: Fri Mar 28 22:09:22 2003 -0500 yikes, fixed incorrect applicability of transpose plans commit 5776651de7f7152e07630b99ee8445fb004131a1 Author: Steven G. Johnson Date: Fri Mar 28 22:06:14 2003 -0500 in the future, we might want to allow sz->rnk == 0, vecsz->rnk arbitrary to be converted to r2hc (the apply function already should work for this case)...disabled for now, though commit db6988d7af647595db1ef218c039bb2755070b59 Author: Steven G. Johnson Date: Fri Mar 28 19:12:08 2003 -0500 use most_unaligned in rdft2 commit bf69a12d650bc9daee88f41bd0a04bf1abe664c8 Author: Steven G. Johnson Date: Fri Mar 28 19:11:47 2003 -0500 slight change commit b79360114562af8636d8c3da2898cc7ed7df7b98 Author: Steven G. Johnson Date: Fri Mar 28 19:00:21 2003 -0500 output message when checks pass commit 14afb1d94a6d7eb23f853cd2097814989148a9e2 Author: Steven G. Johnson Date: Fri Mar 28 17:21:47 2003 -0500 added ifndef alloca around alloca stuff commit 6cccb2f2fed1d79204cbbb7e1ee44685bf2ed300 Author: Matteo Frigo Date: Fri Mar 28 13:45:50 2003 -0500 Proper alignment in rader commit 643528ab599946750ef668ce19266fe5a0bab5c1 Author: Steven G. Johnson Date: Fri Mar 28 12:43:23 2003 -0500 whitespace commit 8c9af83603806d8d769f21b1222dfe717068f7c6 Author: Steven G. Johnson Date: Fri Mar 28 12:41:39 2003 -0500 whoops, alloca stuff inside HAVE_ALLOCA commit 4f4ed55f3679a721e23cf9cb61e9180646f0f176 Author: Steven G. Johnson Date: Fri Mar 28 12:35:21 2003 -0500 make check can afford to be a little bigger commit e7db3e5ee6c6cbb0c24626dc09c00e23e46e70e7 Author: Steven G. Johnson Date: Fri Mar 28 12:31:32 2003 -0500 use same alloca macrology as configure script commit 3c6ec07b6659b60cfb8e77365e75c872ccacd66d Author: Steven G. Johnson Date: Fri Mar 28 03:05:15 2003 -0500 fallback is no longer needed for mingw commit cbc91a4cae1193e9e27ef5885e1bb37c548c191b Author: Steven G. Johnson Date: Fri Mar 28 02:58:45 2003 -0500 alloca fallback for gcc commit 37a6e5be53d9273006dc360b8dafe2e3e53356e5 Author: Steven G. Johnson Date: Fri Mar 28 02:49:59 2003 -0500 _alloca was added for MinGW, but it causes problems there commit 37adf3eddeb59f98c6d4e8888ddb4208b10fb42a Author: Steven G. Johnson Date: Thu Mar 27 22:06:07 2003 -0500 fixed most_unaligned for split format commit fae5ff2e94558a024ef43a1cd4470f5c68b4de17 Author: Steven G. Johnson Date: Thu Mar 27 19:01:58 2003 -0500 whoops commit 5a2216ff945775dcd769967d2a58125b51c4b3c0 Author: Steven G. Johnson Date: Thu Mar 27 19:00:20 2003 -0500 added pkg-config commit 669fc84978762faee9e8d48a7b852eca22ee4303 Author: Steven G. Johnson Date: Thu Mar 27 15:59:01 2003 -0500 fixed asserts commit 8108cd595625c0e28683a556df95de39588c7fb4 Author: Matteo Frigo Date: Thu Mar 27 15:49:53 2003 -0500 Do not adjust r/i pointers separately. commit d9b6e6ea20bc01290efaabb99405a0b10a3dadca Author: Matteo Frigo Date: Thu Mar 27 15:17:40 2003 -0500 iForgot to add files commit 9c79c521f6c57b91f21a90731610214d9dafca8e Author: Matteo Frigo Date: Thu Mar 27 15:10:41 2003 -0500 Specialized n simd codelets for unit vector stride. commit 7e309fd9c2284d234e3932b6d3a2d5bbcc44c9e0 Author: Matteo Frigo Date: Thu Mar 27 08:22:03 2003 -0500 Changed version number to beta2 commit 513db4fd67e83952d5e510cf7a1eb23fbd6ef2bb Author: Matteo Frigo Date: Thu Mar 27 06:37:07 2003 -0500 Changed alignment requirements for n1 simd codelets. Changed mechanism for detecting lack of alignment. commit be8495756a69c610211f28e3f9a7ff20016eb901 Author: Matteo Frigo Date: Thu Mar 27 04:25:06 2003 -0500 Oops, wrong place for hook commit 575731d46f96f87a35f3a1a37dba70cc153728df Author: Steven G. Johnson Date: Thu Mar 27 02:37:52 2003 -0500 added comments to codelet makefiles, to aid people wanting to generate their own code commit 59245164a590789dd1ed892f910bc43a346b791b Author: Steven G. Johnson Date: Thu Mar 27 01:42:27 2003 -0500 Matteo is also a copyright holder commit c558091f6b4b0f37175f86a623e2f2376da9c01a Author: Steven G. Johnson Date: Thu Mar 27 01:41:08 2003 -0500 FORTRAN is officially Fortran, these days commit d0b28f4043bbc0aae200dd359e7ad52da98bc903 Author: Steven G. Johnson Date: Thu Mar 27 01:40:32 2003 -0500 punctuation commit 46bd3b7c143c2291021d10121a74d23936e3ccdd Author: Steven G. Johnson Date: Thu Mar 27 01:40:14 2003 -0500 don't use "wrapper" commit 9a8eba97546a52e4070f6910534c849009d027d8 Author: Steven G. Johnson Date: Thu Mar 27 01:37:53 2003 -0500 plural commit bc26c4cd5feb70158f734130ef3415ce557e207b Author: Steven G. Johnson Date: Thu Mar 27 01:35:32 2003 -0500 grammar commit 246a46ccdfc59616fdc3234ee8f773f54b9b5260 Author: Steven G. Johnson Date: Thu Mar 27 01:33:35 2003 -0500 better phrasing commit 8d4fae80f2b4558c3ec62f108316187e7dad2b84 Author: Steven G. Johnson Date: Wed Mar 26 22:47:58 2003 -0500 stddef.h should not be needed anymore for this file commit 1080fb42895231d251238b34f4af9458ee7329ec Author: Steven G. Johnson Date: Wed Mar 26 22:13:48 2003 -0500 added comments for Franz mode commit e995cc9d9c287c7681b4e8ff9e97dc57e5dcbdb4 Author: Steven G. Johnson Date: Wed Mar 26 22:11:58 2003 -0500 clarification commit cff23bce9ddaeddc3338be9cfcbc3cd3d57f2370 Author: Steven G. Johnson Date: Wed Mar 26 22:08:51 2003 -0500 commented on FRANZ codelets commit 0b33d349e4eab5f3fce6b0873cf0366e83e51d59 Author: Steven G. Johnson Date: Wed Mar 26 22:06:45 2003 -0500 updated commit 8467b9db21a1c618e7ccaf958299f9edc31bbb42 Author: Steven G. Johnson Date: Wed Mar 26 21:51:15 2003 -0500 disable DIF codelets, since they are never used (apparently) except for some non-power-of-two sizes...improve support for the latter by adding size 3, 5, and 6 q^2 codelets. commit 100f8e1667a8ffdc4ad997bbe4346603e7da122b Author: Steven G. Johnson Date: Wed Mar 26 20:07:11 2003 -0500 DHT has no forward/backward commit 9f5c7271cdd393f08d42a71669c9d3d1686ab641 Author: fftw Date: Wed Mar 26 19:46:12 2003 -0500 added hacky way to use an arbitrary flag commit 59d54e87e6bb971ba93e6b371aad0c3ee5d88d11 Author: Matteo Frigo Date: Wed Mar 26 19:44:31 2003 -0500 Better place to install hook commit b35aa5670a5cf242f215c8281c3c09097c3c740a Author: Steven G. Johnson Date: Wed Mar 26 19:40:28 2003 -0500 noted that the user should run make check if they think FFTW has a bug commit dd17b391f48608fdfe190c514eb865ff891689b9 Author: Matteo Frigo Date: Wed Mar 26 17:31:16 2003 -0500 Oops, what am I thinking commit 47c3588218fccd048fb32989c007dc693f402abc Author: Matteo Frigo Date: Wed Mar 26 17:23:56 2003 -0500 Grrr.... fixed bug in estimator commit a3f8ee308a4b9b1f83d1031991c9f8fdc55b3bc2 Author: Matteo Frigo Date: Wed Mar 26 17:16:19 2003 -0500 Oops---the flop count was right. The estimator is broken elsewhere. commit f2103b394847d39a74d720c5dc18b9f3139fc257 Author: Matteo Frigo Date: Wed Mar 26 14:28:41 2003 -0500 Fixed SIMD estimator commit 442a17b47519435071b0c7373c83cc50f5e4b826 Author: Matteo Frigo Date: Wed Mar 26 07:45:03 2003 -0500 Added twidsq simd codelets commit 9647b9a35046476b0697bb196f6ad80a1c81b763 Author: Steven G. Johnson Date: Tue Mar 25 23:33:03 2003 -0500 gensrc -> genfft commit e668b629605e9193d33403c9c87be52a7d08d134 Author: Steven G. Johnson Date: Tue Mar 25 23:32:16 2003 -0500 newline commit 76eeb4a83c788c638126d62924bbedb833573028 Author: Matteo Frigo Date: Tue Mar 25 19:17:08 2003 -0500 Noted need to add dif simd codelets commit 350bf8c788a6f8a0ec21b7b004ce7a83c163f511 Author: Steven G. Johnson Date: Tue Mar 25 13:03:47 2003 -0500 noted shift commit a5fa31a29076ae51d870e2db210b7f51aa46adbb Author: Steven G. Johnson Date: Tue Mar 25 13:02:47 2003 -0500 clarification commit 276ff68eb803fb179adefc146b05b4f616fd226f Author: Steven G. Johnson Date: Tue Mar 25 12:46:44 2003 -0500 need make after bootstrap commit f48787b41d83d8f21ec8ce19b275eaedf5316484 Author: Steven G. Johnson Date: Tue Mar 25 12:31:49 2003 -0500 slight change commit 7143220a87f7444e90964aadccece0c31bf3830b Author: Steven G. Johnson Date: Tue Mar 25 12:30:56 2003 -0500 libtool is also needed commit 212581eeb5c4011118653b3d8fe433b774bbcd1b Author: Steven G. Johnson Date: Tue Mar 25 12:29:52 2003 -0500 added code generator introduction commit e22b4de0a314136783316cc1acbbc7bf97ca105c Author: Steven G. Johnson Date: Tue Mar 25 11:51:49 2003 -0500 added support for REDFT/RODFT/DHT direct codelets commit cc149df36c0ddc161d91558da702572cd01f99c6 Author: Steven G. Johnson Date: Tue Mar 25 11:29:29 2003 -0500 noted ARM bug; thanks to Jay Treacy commit e313a7fb2e0c1c2524eaed8926b25055a38fb957 Author: Matteo Frigo Date: Tue Mar 25 07:55:54 2003 -0500 bugfix from Stefan commit dc62fc48ad26abb231c697a5a18b5f7ca64ab6fe Author: Steven G. Johnson Date: Mon Mar 24 15:59:08 2003 -0500 slight change commit 878030bb1ea7efd3b1e1dab02601732fd5c90c36 Author: Steven G. Johnson Date: Mon Mar 24 15:58:44 2003 -0500 caveat commit 776dd0aac7bb400bce14f59781f664062d7b4117 Author: Steven G. Johnson Date: Mon Mar 24 15:58:04 2003 -0500 warning about DHT commit 2bd26e46c0f7e3622be81d9922f0089923143c65 Author: Matteo Frigo Date: Mon Mar 24 08:34:14 2003 -0500 Oops commit 597693dba60d0535d890bbb5f161c3a01830a1ea Author: Matteo Frigo Date: Mon Mar 24 08:13:15 2003 -0500 Regression test for p4fftwgel commit 92603541cee018def425427f93dcb3739ab7c0f1 Author: Steven G. Johnson Date: Mon Mar 24 03:09:06 2003 -0500 make check is faster, old tests are in make bigcheck commit 518b188f2a8a30b7cbc2a5c34b335940afa54530 Author: Steven G. Johnson Date: Sat Mar 22 00:41:21 2003 -0500 note commit 82b8e611b2d1a65af695db0ddf0cf306a5804886 Author: Steven G. Johnson Date: Sat Mar 22 00:40:05 2003 -0500 whoops, line wrapping commit b9e7ade930fd2e2de8105a28ff7a8f32a799237f Author: Matteo Frigo Date: Fri Mar 21 15:10:00 2003 -0500 Franz-mode codelets even without SIMD. (disabled) commit bd548cc599b6178d2e1bdbc2c6abc08f276ae386 Author: Matteo Frigo Date: Fri Mar 21 09:09:30 2003 -0500 Bug is in netbsd-1.6, not 1.5 commit 1d1b6b166a1164c2499c4a7e5f9bd9b69f3cf5c3 Author: Matteo Frigo Date: Fri Mar 21 07:45:48 2003 -0500 const cast, should placate c++ compilers. commit 677ff57df2415f59cc701368e26dd23d1c6ec956 Author: Steven G. Johnson Date: Thu Mar 20 18:49:49 2003 -0500 added FAQ on why plans are array-specific commit f0c1a0a7c03bfb68f4559001c2b652aa7a601c0d Author: Steven G. Johnson Date: Thu Mar 20 16:12:56 2003 -0500 comment fix commit 3ac192669e4bbb596cc30adb429179fa58f11387 Author: Steven G. Johnson Date: Thu Mar 20 16:12:15 2003 -0500 noted comparison to NR commit 3f8a990d38ae5f796daa261636120dcb936acb2a Author: Steven G. Johnson Date: Wed Mar 19 20:13:16 2003 -0500 whoops, C99 complex didn't work if complex is a macro (as it is with glibc); thanks to Keh-Cheng Chu for the bug report commit faab1981e35c6596ac99e9c1e8379c77d92155fe Author: Steven G. Johnson Date: Wed Mar 19 16:52:54 2003 -0500 noted in help that --enable-k7 enables 3dnow, and that --enable-3dnow is only a fallback commit 297a4fd9785fe05d2149abf128413bd363fa2dbc Author: Matteo Frigo Date: Wed Mar 19 15:09:52 2003 -0500 New gcc bug. html.refs was not in repository/distribution. commit 8a81ec059d2c1c567c69a45f77d76b8f242c8836 Author: Matteo Frigo Date: Wed Mar 19 10:09:16 2003 -0500 Don't write wisdom if you don't have it. commit 4bc446d3e46e7a8c2f084d425e193e58f9ff76ec Author: Matteo Frigo Date: Tue Mar 18 15:44:41 2003 -0500 Added index entries for DHT. Similarly for DCT, DST commit 4a72bfaf1f333116de1e5e0a154bc87d17c9c234 Author: Steven G. Johnson Date: Tue Mar 18 14:50:04 2003 -0500 execute should not go through C api, for efficiency commit 22f933b01c30e0f68f46f8a73e474a1e8a893360 Author: Matteo Frigo Date: Tue Mar 18 06:14:51 2003 -0500 Renamed FFTW_IODIM, FFTW_R2R_KIND commit e57a38d55f979644a5fecd702c1d4bd105b1eac3 Author: Steven G. Johnson Date: Tue Mar 18 00:30:17 2003 -0500 added rfftwnd.eps to dist, so that transfig is not required for people trying to build other formats (e.g. ps); thanks to Brian Gough for the bug report commit f5713b796921f5e1cfded3ce96e33c6df0d09a8c Author: Steven G. Johnson Date: Mon Mar 17 15:17:59 2003 -0500 pointer to upgrading section from tutorial commit 7e222b6349b8a2bceeb8703d347715fb763efadd Author: Steven G. Johnson Date: Mon Mar 17 14:44:40 2003 -0500 make print_plan and fprint_plan, so that the former can be more easily called from other languages commit f358b64955871f01d87a42a05275f5f0cb5094e5 Author: Steven G. Johnson Date: Mon Mar 17 14:19:10 2003 -0500 whoops, forgot to change equation image links to .png commit ea32f5a93a88f6ddec9185886bbbea43cf8ed067 Author: Matteo Frigo Date: Mon Mar 17 04:15:50 2003 -0500 fixed c++ linkage problems commit de7c276d4b4ab36471c8dcb639d3c522d2cbe7cc Author: Matteo Frigo Date: Mon Mar 17 03:25:17 2003 -0500 Removed ``const'', otherwise c++ link fails commit f3bf675c6d0003e3087d634aab2ef34a6745dcb9 Author: Steven G. Johnson Date: Sun Mar 16 20:24:31 2003 -0500 fixed C++ annoyances: void* casts, and global variables are static by default(?!?) commit 45e54b3f9a8c0b5942cc21c0b2d2f19682d3a7c0 Author: Steven G. Johnson Date: Sun Mar 16 15:29:11 2003 -0500 ranlib bug is in binutils commit a17b7eb5a896ea6d7ca3f73fd7251bfc76de500d Author: Steven G. Johnson Date: Sun Mar 16 15:26:42 2003 -0500 ranlib Irix bug commit f482abd59b1c2afed27292d79bef782b935d0c51 Author: Steven G. Johnson Date: Sun Mar 16 15:13:35 2003 -0500 start with random tests commit a216647a57733c53d3407957caaaf759ed0dd700 Author: Steven G. Johnson Date: Sun Mar 16 15:00:04 2003 -0500 silenced some compiler warnings, eliminated unused variables, and fixed Makefile.am for f77funcs.h commit 2b581243067955d0e82eb7cf487def793b8f66b6 Author: Steven G. Johnson Date: Sun Mar 16 14:55:13 2003 -0500 whoops commit 058c4751ed4a98a52e6a878c78335f4997c60294 Author: Steven G. Johnson Date: Sun Mar 16 14:28:22 2003 -0500 3dnow is float commit 0b50f9ef3433e59f0b9cc7983652a8fe3c361fdf Author: Steven G. Johnson Date: Sun Mar 16 14:27:45 2003 -0500 fixed k7 docs commit 9179a17f1ba8db6733ccb2dbe541aa3f5f59727e Author: Steven G. Johnson Date: Sun Mar 16 14:19:10 2003 -0500 SGI compilers now support inline commit 353d40e964502f46aba99f094c08fd610cb9fc2c Author: Steven G. Johnson Date: Sun Mar 16 14:18:32 2003 -0500 cruft commit 455c3aa4c3e0cc6d5404c78ef12ed70b8751da45 Author: Steven G. Johnson Date: Sun Mar 16 14:15:47 2003 -0500 texinfo doesn't like commas in nodes commit 323b6d34cebb6be520075efaf4eeef0a369a6635 Author: Steven G. Johnson Date: Sun Mar 16 13:52:04 2003 -0500 updated commit 7762fe2f89dc86791560cad9326ece6fbdbceaf7 Author: Steven G. Johnson Date: Sun Mar 16 13:47:44 2003 -0500 f77funcs.c -> f77funcs.h so that people don't try to compile it commit acd3f5b16c0b85acfad30bb086199cc65bc6b326 Author: Steven G. Johnson Date: Sun Mar 16 13:46:11 2003 -0500 minor changes commit 8d4f8a05ac24ce13ba6adea137099c22c6f5362b Author: Steven G. Johnson Date: Sun Mar 16 13:39:24 2003 -0500 updated compiler bug list commit dc84fdefd84cac3bd6ecf521f48ca6cab0ae2b0e Author: Steven G. Johnson Date: Sun Mar 16 13:39:07 2003 -0500 noted how to set CC commit d71b55ff07b10fe5ee5dc24799511bdbb0b3f772 Author: Steven G. Johnson Date: Sun Mar 16 13:01:01 2003 -0500 TODONE commit 454b2a79327b2582f18024204a6ab683d97f9f41 Author: Steven G. Johnson Date: Sun Mar 16 13:00:42 2003 -0500 yikes, bugfix commit e741c61f2ab8b259c217e9e25adbcece21a6be4b Author: Steven G. Johnson Date: Sun Mar 16 10:26:28 2003 -0500 whoops commit f0073024ddb3bb621a4c71fcc7ddb575adf42871 Author: Matteo Frigo Date: Sun Mar 16 09:24:19 2003 -0500 Report SIMD extensions in version string commit 0b40f7e79a8110bd4d2215f9d81a3d100f1e9ecc Author: Steven G. Johnson Date: Sat Mar 15 18:56:11 2003 -0500 more verbose output commit 1310aa1ef6043afa44bc6c8bcc2d7b3bae66190c Author: Steven G. Johnson Date: Sat Mar 15 17:41:25 2003 -0500 a couple of additional non-Unix instructions commit 12cb13aafd73275762b5f2c098c436457b8f9be9 Author: Steven G. Johnson Date: Sat Mar 15 17:15:26 2003 -0500 hyphen commit 7aea3d41ed7a9dde86b14f410caf606a05f15fd5 Author: Steven G. Johnson Date: Sat Mar 15 17:12:29 2003 -0500 softened commit 2a251916b17e7380f33bf556d666781828819789 Author: Steven G. Johnson Date: Sat Mar 15 17:09:44 2003 -0500 added FAQ, used PNGs commit d3669c90789fbfcc99404a8fbd8d90540fae6c52 Author: Steven G. Johnson Date: Sat Mar 15 15:29:43 2003 -0500 great copyright update commit 1b82fbfbe632120cba76c9c6107bd3e1abbe4547 Author: Steven G. Johnson Date: Sat Mar 15 15:14:02 2003 -0500 threads in make check commit a7ebafd6aec670afd0a9d5165893abf7d7413870 Author: Steven G. Johnson Date: Sat Mar 15 15:11:24 2003 -0500 fixed const warnings commit b72d4726555aa5ef40e612f712eaa2190324c89e Author: Steven G. Johnson Date: Sat Mar 15 15:08:25 2003 -0500 make sure spawn_loop size > 1 (it has to be at least > 0 lest we crash, but > 1 is an optimization) commit 8f82cc0405e8d264d1a201e4b65d0e82e5822834 Author: Matteo Frigo Date: Sat Mar 15 14:00:17 2003 -0500 hpux seems to want machine/sys/inline.h as opposed to machine/inline.h. commit 195978c28fbdd1b1ead25d381c9c6af6f71a74fb Author: Steven G. Johnson Date: Sat Mar 15 13:36:56 2003 -0500 Sourceforge is really SourceForge.net, and is run by VA commit 93eaa99ca18255b538bd37c4742ff87898a9350c Author: Steven G. Johnson Date: Sat Mar 15 13:34:05 2003 -0500 comma commit aa16c88c1efdf9283884a6f3c28bda36d54c1cb9 Author: Steven G. Johnson Date: Sat Mar 15 13:31:42 2003 -0500 fixed AMD company name commit fa4887fa3ddccb2e53b50158d92f8cb9da3223f2 Author: Steven G. Johnson Date: Sat Mar 15 13:29:41 2003 -0500 minor changes commit 689f73454e57451cc4ceca48e6c9b3856550cc3f Author: Steven G. Johnson Date: Sat Mar 15 13:13:55 2003 -0500 more emitter->read_char renaming commit 469d7370865e70079d60fc5d2144c477847ff50b Author: Steven G. Johnson Date: Sat Mar 15 13:08:45 2003 -0500 more wisdom docs, noted wisdom utilities commit 69c2e6ee0d6523c9181828e9d918d00390f1b07f Author: Steven G. Johnson Date: Sat Mar 15 11:41:32 2003 -0500 compound adjectives are hyphenated commit 1c816b975a4d35c3296bceb2700bc665c2838788 Author: Steven G. Johnson Date: Sat Mar 15 11:40:30 2003 -0500 fftw does support another type of packed array via r2r commit 4510d672da97fc9273a574d9cad23f807c811192 Author: Steven G. Johnson Date: Sat Mar 15 11:29:12 2003 -0500 write_char/read_char for export/import functions commit 5d042765f68d22c08849f8120b432d637364a95a Author: Steven G. Johnson Date: Sat Mar 15 11:19:19 2003 -0500 comments commit ec4d319ec4d855dd2e5c3521429d77dcba1deffa Author: Matteo Frigo Date: Sat Mar 15 10:08:26 2003 -0500 Enabled randomized-cse commit 85619e6f972e3105691588bba210448ad468726f Author: Matteo Frigo Date: Sat Mar 15 09:47:49 2003 -0500 Changed to 3.0-beta1 commit 6c58169a5ef565ec595054c8a1a3644a119575ad Author: Matteo Frigo Date: Sat Mar 15 09:07:31 2003 -0500 First complete draft commit e014222e1611b0fda35eb4e81010d764371f645a Author: Matteo Frigo Date: Sat Mar 15 08:37:52 2003 -0500 EMITTER is a misnomer commit b4e71cdebd8e08a8e4cb6e4e021c9839b0240220 Author: Matteo Frigo Date: Sat Mar 15 05:50:50 2003 -0500 Revision, wisdom tutorial, acks. commit 304d6a33a960a6867e345b7a2391f580de183901 Author: Steven G. Johnson Date: Fri Mar 14 22:59:04 2003 -0500 noted OpenMP commit f0132ff87cdec8cce3eec22776267630ce5d52a8 Author: Steven G. Johnson Date: Fri Mar 14 22:38:49 2003 -0500 comment commit d32e3536671b2ddf95fd19eefd595903f53369c9 Author: Steven G. Johnson Date: Fri Mar 14 22:38:30 2003 -0500 comments commit 758a708f03680fe53ce46466e344370a9537adfc Author: Steven G. Johnson Date: Fri Mar 14 22:38:05 2003 -0500 reformatting commit f31a618619a119ba5df49807d225f5fef53e2acc Author: Steven G. Johnson Date: Fri Mar 14 22:26:28 2003 -0500 whoops commit 47acccb2c662f75a8b9b082032072bfa154f13e5 Author: Steven G. Johnson Date: Fri Mar 14 22:11:23 2003 -0500 some threads fixes, and added experimental semaphore (pre-thread-spawning) and Linux spinlock support commit b3f95134caa95e434d418ab40f2bb57c07521a33 Author: Steven G. Johnson Date: Fri Mar 14 20:50:46 2003 -0500 whoops commit caedcb4f9b8df5449616654ec8782156a2e63e7f Author: Steven G. Johnson Date: Fri Mar 14 18:23:03 2003 -0500 added note that FFTW_PATIENT will disable threads if they are not beneficial commit 34677912b28d4342f4ac1f84e27ee248d2c9ca71 Author: Steven G. Johnson Date: Fri Mar 14 18:20:44 2003 -0500 made fftw_cleanup* more restrictive, in that we don't want to guarantee that previously created plans will still work (they won't, in the case of threaded plans and fftw_cleanup_threads), and there is no reason to provide such a guarantee anyway. commit 4311c764859ea3a4a45fbb507ff0e131d12a5d44 Author: Matteo Frigo Date: Fri Mar 14 17:23:13 2003 -0500 Moved version.c from kernel/ into api/ commit b79acfd84c9dc9bf6ce933ef72af7aafa01623e4 Author: Matteo Frigo Date: Fri Mar 14 17:19:50 2003 -0500 icc-7.0 requires -openmp commit 36f49567ecc9ec71ab72b760ee70ceb688f51f4c Author: Matteo Frigo Date: Fri Mar 14 14:47:52 2003 -0500 Ensure that one can do make dist given the distribution commit 266bb8c14f0aa494b54fcaf1fd0b517c646d5618 Author: Matteo Frigo Date: Fri Mar 14 14:38:11 2003 -0500 Dist fftw3.pdf, not fftw.pdf commit a79801bc40a8ba8ba6f7b27f78aebb9426010b5a Author: Matteo Frigo Date: Fri Mar 14 14:36:25 2003 -0500 Support -onthreads=%d commit 84c91507e6f7f6a050cc8651c7ee8c017d5d1b2f Author: Steven G. Johnson Date: Fri Mar 14 14:34:21 2003 -0500 comment commit a8ef843faf74d0384c6ee1320b456f6aae56c5b2 Author: Steven G. Johnson Date: Fri Mar 14 14:33:27 2003 -0500 whoops commit 3e4f6ed2ad2fda1dbaa2bb444f81cbf116ab1931 Author: Steven G. Johnson Date: Fri Mar 14 12:32:18 2003 -0500 fftw_real is gone commit b55295b022d814a869b207fea2dbbb79c5091525 Author: Steven G. Johnson Date: Fri Mar 14 12:26:04 2003 -0500 typos commit 469579587defd8532f362c0ca4a2935532bae16a Author: Matteo Frigo Date: Fri Mar 14 06:21:43 2003 -0500 More BENCH_DOC strings commit 18f0d31d803f348a8494ac190b4b9ff8d9be7a97 Author: Matteo Frigo Date: Fri Mar 14 05:58:53 2003 -0500 Fixed xref's commit d39f035994e443ebbc933eae51b3d9116bc50bb4 Author: Matteo Frigo Date: Fri Mar 14 05:38:26 2003 -0500 Revised manual (esp. intro and tutorial), fixed texinfo hackery for figures. commit f0cf0419996f46abb0bdf85068d67c1f88435a87 Author: Steven G. Johnson Date: Wed Mar 12 02:42:33 2003 -0500 redirect users from guru execute to advanced interface, if possible commit d30d60239f8f57975f53876649f04f04458b8d90 Author: Steven G. Johnson Date: Wed Mar 12 02:35:22 2003 -0500 punctuation commit cc3b4e3f2fd1880b0a9ced57de8bc592ac868aab Author: Steven G. Johnson Date: Wed Mar 12 02:28:51 2003 -0500 use correct heading level commit 16e33bb6e9eba6c6ac3a3b5e88192f0937cbc79a Author: Steven G. Johnson Date: Wed Mar 12 02:24:37 2003 -0500 html generation commit 8ea08e261cef0528db1c181268c6aabca6c52e50 Author: Steven G. Johnson Date: Wed Mar 12 01:44:00 2003 -0500 added equation GIFs commit 6b511ad0e8551382fb008d5f7d9d6db7c923f5d7 Author: Steven G. Johnson Date: Wed Mar 12 01:43:27 2003 -0500 punctuation commit b223dbcdf2607d546dcde4593dfeb29740b5a2c3 Author: Steven G. Johnson Date: Wed Mar 12 01:26:46 2003 -0500 punctuation commit 8e6421b39b31952d4cde709e9a7dc68146eeac77 Author: Steven G. Johnson Date: Wed Mar 12 01:25:12 2003 -0500 added multi-dimensional transform definitions commit da7ac31fa42d9b594d9a458bc86b31e326d2631b Author: Steven G. Johnson Date: Wed Mar 12 00:14:03 2003 -0500 slight changes commit 4fa36533cd5df28fb24a7cd7678c4ff3a2b8e1f7 Author: Steven G. Johnson Date: Wed Mar 12 00:06:34 2003 -0500 typo commit 93fdbbd4434ff6db48765645e2af3eb2031caece Author: Steven G. Johnson Date: Tue Mar 11 23:50:43 2003 -0500 added 1d version of What FFTW Really Computes commit 989a15455a04e193bd71a2fe4b1daea5649d0f2d Author: Steven G. Johnson Date: Tue Mar 11 21:17:54 2003 -0500 note in upgrading section about FFTW_PATIENT commit f94fc8414c8477ad076f17bed5a1bffe87557ea9 Author: Steven G. Johnson Date: Tue Mar 11 15:18:39 2003 -0500 added cycle-counter section commit 32e58f9ac101c22551198abe31c5021196f69f0e Author: Steven G. Johnson Date: Tue Mar 11 14:53:44 2003 -0500 more ideas commit 54102c10c10da11afcf1dac0451ce4a1e064be8c Author: Steven G. Johnson Date: Mon Mar 10 17:41:35 2003 -0500 noted that indirect should probably be merged with rank-geq2, to make a rank-split solver commit e93a7d1eda3519a9467a0d1a7af57a176aae195c Author: Steven G. Johnson Date: Fri Mar 7 03:01:52 2003 -0500 added non-Unix installation instructions commit 910a5988b2529e4ebd33372540c9db14626a3e8c Author: Steven G. Johnson Date: Fri Mar 7 02:30:59 2003 -0500 also talk about stack alignment with SSE/SSE2 commit 620f6439ff6d382e7f79fba9735243ffbc4e98d6 Author: Steven G. Johnson Date: Fri Mar 7 02:24:07 2003 -0500 made warning more dire commit 6c49e3a0d90853a504b55ee2bb9e67e6961334c6 Author: Steven G. Johnson Date: Fri Mar 7 02:13:25 2003 -0500 fix commit 076cf960691702683f560140c3c90932f531c802 Author: Steven G. Johnson Date: Fri Mar 7 02:09:55 2003 -0500 number commit abe3e1b3e2ac5d1ce15dd74544550011079f056c Author: Steven G. Johnson Date: Fri Mar 7 02:09:08 2003 -0500 fix commit a43149065f2c521c8ce705f9ac0eeb519899ec2b Author: Steven G. Johnson Date: Fri Mar 7 02:08:01 2003 -0500 minor commit cb19343373774be75d78469cbcd3ac4f0f4a903a Author: Steven G. Johnson Date: Fri Mar 7 02:04:45 2003 -0500 minor fix commit d962180e504c71e46dc5b2f71d2304c254fcdace Author: Steven G. Johnson Date: Fri Mar 7 01:58:15 2003 -0500 cross-ref commit 35ef1ce130da4c0389a2f7cef5eaab36dbd614ae Author: Steven G. Johnson Date: Fri Mar 7 01:57:31 2003 -0500 minor commit 28fe03b9f79a6a80be8cc0d02cfc87e090f408d5 Author: Steven G. Johnson Date: Fri Mar 7 01:53:28 2003 -0500 more installation manual commit 650bf3b91d1fe392906f9aa25faed1707244f4f4 Author: Steven G. Johnson Date: Fri Mar 7 00:43:40 2003 -0500 GNU-lly correct commit aec18000f9851e8985d704ee50f49ea4d17f324e Author: Steven G. Johnson Date: Fri Mar 7 00:38:48 2003 -0500 started installation section commit f7bf8016fa681c46c51385297a58d6dae611862f Author: Steven G. Johnson Date: Fri Mar 7 00:25:02 2003 -0500 added --without-cycle-counter option as a last resort commit e97d01d48d003b290d6d2da7dc53cea35c90357d Author: Steven G. Johnson Date: Fri Mar 7 00:07:12 2003 -0500 macros with () arguments were only standardized in C99, and we don't need them anyway commit 459a56abf74ce71af7c63047b31d39f7befefbb9 Author: Steven G. Johnson Date: Thu Mar 6 23:10:41 2003 -0500 wording commit 20d77f4e2a461fab512a8b5cd0ccd301d42f3673 Author: Steven G. Johnson Date: Thu Mar 6 23:03:03 2003 -0500 parallelism commit 881feeb3f3d4813a30da4baf5d71b8af8ca72d23 Author: Steven G. Johnson Date: Thu Mar 6 23:01:47 2003 -0500 additions to upgrading chapter commit 59f6ac21ab762470d0d4740130fa2131cc3f684e Author: Steven G. Johnson Date: Thu Mar 6 22:39:36 2003 -0500 noted additional humility of FFTW 3 wisdom commit cf933ec73d68c2839a79d1fc53ba1198dc63fe39 Author: Steven G. Johnson Date: Thu Mar 6 22:32:44 2003 -0500 renaming commit 724b52700268a45264d168aaf7a63977a16af8bb Author: Steven G. Johnson Date: Thu Mar 6 22:31:00 2003 -0500 added placeholder for wisdom reference commit c286ee068195c75e012cdf36534aa5f4154b394f Author: Steven G. Johnson Date: Thu Mar 6 22:29:38 2003 -0500 wrote upgrading chapter commit 35c5a163f14e561b45a226dece35564f5773ce69 Author: Steven G. Johnson Date: Thu Mar 6 18:01:10 2003 -0500 slight change commit de53b4abb5481ee319ffcfc4e4b215861d814ed6 Author: Steven G. Johnson Date: Thu Mar 6 18:00:43 2003 -0500 placeholder for upgrade chapter commit a8a06d66b81a785625077d6de4fb8699ee4c718f Author: Steven G. Johnson Date: Thu Mar 6 13:47:49 2003 -0500 whoops commit a128a59973d9f74fa491a56fc22b374ad69a5ebc Author: Steven G. Johnson Date: Thu Mar 6 13:36:38 2003 -0500 strengthed warning about time commit 271819893ab4e7634f8cee294f9c68612ff811f2 Author: Steven G. Johnson Date: Thu Mar 6 13:35:42 2003 -0500 noted -t in example commit 08b64e3b15f7ad163677a348ba8d0a1a62720b07 Author: Steven G. Johnson Date: Thu Mar 6 13:21:03 2003 -0500 pay attention to WINDOWS_F77_MANGLING commit 5428bbf998b549e46c06f6f3e2ed9ff435304631 Author: Steven G. Johnson Date: Thu Mar 6 02:52:30 2003 -0500 punctuation commit 1462402c458e7a21360fcde1e6a5e9a023987747 Author: Steven G. Johnson Date: Thu Mar 6 02:51:02 2003 -0500 index commit 3cfc6a120672eeb46fca1300ba357ef6bff2b1cc Author: Steven G. Johnson Date: Thu Mar 6 02:50:38 2003 -0500 documented C++ usage commit 675b0233f6e57d4aa15fe422acb4c156e2c3692a Author: Steven G. Johnson Date: Thu Mar 6 02:25:32 2003 -0500 got rid of overfull hbox TeX warnings commit a5a689c09a184e7f361240b46f8a74cd5c0bea78 Author: Steven G. Johnson Date: Thu Mar 6 02:20:38 2003 -0500 whoops commit 9bdfa427108e546c8fd707d8bde9151b5cacd81d Author: Steven G. Johnson Date: Thu Mar 6 02:20:13 2003 -0500 noted fftw_iodim split for Fortran guru interface commit ba02448b7f27ddbff45651477c0ca5ea4d28b7bd Author: Steven G. Johnson Date: Thu Mar 6 02:14:21 2003 -0500 added guru reference commit db7990c25a72ecb1a1acddfa63bdd8c38fdaeedf Author: Steven G. Johnson Date: Wed Mar 5 22:56:05 2003 -0500 minor commit 58778ac5172128991fd8e88d4461004a03763596 Author: Steven G. Johnson Date: Wed Mar 5 22:45:31 2003 -0500 use @r{...} for comment text in code examples commit bd4b0411a2a7a9485f83d430455ff5d1571019f8 Author: Steven G. Johnson Date: Wed Mar 5 13:14:04 2003 -0500 eliminate warning commit 87d217e8cd045402dbb4d9a4bc7ac81481edbcf9 Author: Steven G. Johnson Date: Wed Mar 5 13:12:56 2003 -0500 SIMD_CFLAGS only for simd code commit 8346b6688d8e88aa91864685b77de030e8cb2549 Author: Matteo Frigo Date: Wed Mar 5 11:06:41 2003 -0500 Minor changes. commit 181d6c8fbdca0f24c1feb199c9a29edcf2187977 Author: Steven G. Johnson Date: Wed Mar 5 02:13:34 2003 -0500 cross-compiling with MinGW can't detect f77 mangling, so add an option to use what seems to be the most common styles commit 17f9e2aabc5526c6614d7055960c5e7f5fda3720 Author: Steven G. Johnson Date: Tue Mar 4 20:00:31 2003 -0500 comment commit b0715eb2e0f6662e3b3b41adf70799a31c2ab630 Author: Steven G. Johnson Date: Tue Mar 4 20:00:13 2003 -0500 we only use our-malloc-16 on machines where size_t == uintptr_t, so don't bother doing the right thing with the benchmark commit 72d331d4dbb9bf0bed0796e05eaf970a17c2975a Author: Steven G. Johnson Date: Tue Mar 4 19:46:09 2003 -0500 support WITH_OUR_MALLOC16 commit d2ee17676db2b01e1d57b6f6fcebe4c9c8987fff Author: fftw Date: Tue Mar 4 18:50:53 2003 -0500 automatically add -msse etcetera for --enable-sse etcetera commit 0a7cb6363f8effac8a34176c7b31d1dfbe4e71d0 Author: fftw Date: Tue Mar 4 18:24:26 2003 -0500 got rid of const warning commit f27a29dff516ba8bf8bd22a3affe1e881a045389 Author: fftw Date: Tue Mar 4 18:22:48 2003 -0500 missing header commit 58b8d88bdb16fde7d1400c93b1d976af4a29acaf Author: Steven G. Johnson Date: Tue Mar 4 15:55:47 2003 -0500 fixes commit a636d3b26c9ca10c0225bb058035e2f99ae41383 Author: Steven G. Johnson Date: Tue Mar 4 15:53:26 2003 -0500 whoops commit 530bdb066779445d91537bb42fafd03d98d24bd1 Author: Steven G. Johnson Date: Tue Mar 4 02:22:14 2003 -0500 started guru reference commit c44336102065022482f5d8a4eda068247672c05c Author: Steven G. Johnson Date: Tue Mar 4 01:44:09 2003 -0500 use same FFTW_IODIM between precisions commit da6302aba33f0dc74c9da6d7cd4824a6c431c948 Author: Steven G. Johnson Date: Tue Mar 4 00:25:57 2003 -0500 renamed section commit bf45437f266c9ce170d54e87466ba34f41b1937d Author: Steven G. Johnson Date: Tue Mar 4 00:21:49 2003 -0500 no need for "advanced" in subheadings commit 5fb9bd9fe4b93abeb0aa4b00e1ca6e9057da2fbd Author: Steven G. Johnson Date: Tue Mar 4 00:20:05 2003 -0500 typo commit 0127b618539bcb2ddf8634d4bb09c10673ba26a5 Author: Steven G. Johnson Date: Tue Mar 4 00:17:23 2003 -0500 finished advanced interface commit 76aa5434ffee4220caa0b1935d813723d43d55eb Author: Steven G. Johnson Date: Mon Mar 3 23:26:12 2003 -0500 more advance interface docs commit 05a9b164357317a362a1f4e0acb2067faa66910e Author: Steven G. Johnson Date: Mon Mar 3 23:12:09 2003 -0500 fail for win32 commit c49ad63f2fb49af4c81ea1fde51303013e637d7b Author: fftw Date: Mon Mar 3 17:18:48 2003 -0500 shortened help string commit 52ebcb06b186e8f796fdc71ae30d3ac7e9e35017 Author: fftw Date: Mon Mar 3 17:16:17 2003 -0500 fixed cross-refs commit 331a793c80e1bb04018aad92d07791ff432d792e Author: fftw Date: Mon Mar 3 17:07:27 2003 -0500 FFTW_POSSIBLY_UNALIGNED -> simpler FFTW_UNALIGNED in API, added bench option commit 3ba1c479988c55e2f9244fac654f491c5b1c4b78 Author: fftw Date: Mon Mar 3 16:58:07 2003 -0500 whoops commit 62a1622e28fcc9408467bccee64c50f977243b7f Author: fftw Date: Mon Mar 3 16:52:58 2003 -0500 noted assumption commit b6a1f1234fe0834ad8c7a313fc15c710bffafdc6 Author: fftw Date: Mon Mar 3 16:50:33 2003 -0500 provide our own malloc16 routine because of Windows lossage commit 22de7295407d77062d3611d326295950f90d4907 Author: Steven G. Johnson Date: Mon Mar 3 13:28:12 2003 -0500 capitalization commit 5756c9b659e1dda142a21c8c4c8fed00015bf29d Author: Steven G. Johnson Date: Mon Mar 3 13:26:32 2003 -0500 whoops commit 0f92b4f922681df3c6ea4a35bafb8c32907a028d Author: Steven G. Johnson Date: Mon Mar 3 12:55:57 2003 -0500 vertical skip looks better than indenting for setting off short paragraphs commit 767a89f2268461313cb0a3666be311640bb288af Author: Matteo Frigo Date: Mon Mar 3 06:34:09 2003 -0500 Removed franz-mode. Automake was distributing franz files whether franz mode was enabled or not. commit d40ea4ed0a561aa7f85008bb970d07b33010a0eb Author: Steven G. Johnson Date: Mon Mar 3 01:44:00 2003 -0500 made output boundary conditions more prominent; they are important, because they make the different transform types inequivalent in parity commit 909ed5b34a848e505c9a62fcb5b07d346183a43d Author: Steven G. Johnson Date: Mon Mar 3 01:17:28 2003 -0500 clarification commit feb1fc01699f139143e536e1d0f961b904bba74e Author: Steven G. Johnson Date: Mon Mar 3 01:17:07 2003 -0500 typo commit 8d2e91da57095741496a5ae8b809cee8bd01bdb9 Author: Steven G. Johnson Date: Mon Mar 3 01:10:28 2003 -0500 started advanced reference commit 34cc962abf3c75c27328c21fb2c9b053426870f8 Author: Steven G. Johnson Date: Mon Mar 3 00:52:02 2003 -0500 r2r reference commit 6a32d0463a93a19f01e9b13bdc2e0d73857c7eaa Author: Steven G. Johnson Date: Sun Mar 2 23:51:21 2003 -0500 workaround for info formatting bug commit a76009f2fb554d2af97e39f4857b70d26a263bf5 Author: Steven G. Johnson Date: Sun Mar 2 23:47:19 2003 -0500 noted lack of fftw_malloc in Fortran commit 53555b1acdefbc4b092702bcd7defa71dd523ee7 Author: Steven G. Johnson Date: Sun Mar 2 23:42:52 2003 -0500 parallelism commit a84b5314b96882b7495c7d0fbdd91a73f678683f Author: Steven G. Johnson Date: Sun Mar 2 23:39:54 2003 -0500 whoops commit f4b30c1aeb9cadcb0ef3586a40e2a41a6087304f Author: Steven G. Johnson Date: Sun Mar 2 23:33:02 2003 -0500 r2c/c2r reference commit 9afb0869850070a47c3b45df511efdaef0c19292 Author: Steven G. Johnson Date: Sun Mar 2 22:44:10 2003 -0500 table of contents was being included twice commit 9433ef02af21f2e3ee1c5a5e6034a2e5a02663af Author: Steven G. Johnson Date: Sun Mar 2 22:42:29 2003 -0500 minor changes commit 34aaf0acd96dc522e8b71c3844077a7d28149690 Author: Steven G. Johnson Date: Sun Mar 2 21:54:13 2003 -0500 started reference section commit 10afdab4c99f7d367227f61d6ea87e43113379ef Author: Steven G. Johnson Date: Sun Mar 2 19:10:02 2003 -0500 whoops commit da1655a272a6bd0bf3db360605818d3684e01919 Author: Steven G. Johnson Date: Sun Mar 2 19:03:23 2003 -0500 started ref. section commit a3cc56c2b538f79864f787f9480a7da21017624a Author: Steven G. Johnson Date: Sun Mar 2 18:50:58 2003 -0500 fftw_flops takes const plan commit e6c9dd42b944a416f6cca057b2277acb2a00d370 Author: Steven G. Johnson Date: Sun Mar 2 15:54:14 2003 -0500 typo commit 9d97e6245d45d65061499080021f2e0c877803b6 Author: Steven G. Johnson Date: Sun Mar 2 15:52:41 2003 -0500 added "Wisdom of Fortran?" section commit 85f80c144fc9da705ddc7da87d0e437a4125d1db Author: Steven G. Johnson Date: Sun Mar 2 15:50:37 2003 -0500 typo commit 0f4d81b32a7ddf1e011dcc66a7ca3a6f01602aa9 Author: Steven G. Johnson Date: Sun Mar 2 15:49:57 2003 -0500 wording commit 6c6dd67d7f64ce4ab293456c0b4fce7397b4204f Author: Steven G. Johnson Date: Sun Mar 2 15:46:13 2003 -0500 added comments commit d9ecf01ce4b7d0bb1c81de9097941541d96f68d0 Author: Steven G. Johnson Date: Sun Mar 2 15:44:01 2003 -0500 added example file commit 37b6da9ec0958f78193e343ff5adbb7221039698 Author: Steven G. Johnson Date: Sun Mar 2 15:37:32 2003 -0500 don't print out READ WISDOM unless we have commit c476c76dbda1de2cfcfed5db46f8eb6a59ca5eda Author: Steven G. Johnson Date: Sun Mar 2 15:36:28 2003 -0500 EOF is not a space commit 789f94ba726188b22495dffa33536923784cc893 Author: Matteo Frigo Date: Sun Mar 2 09:14:37 2003 -0500 Turn on inline by default commit f76cd82b2e8d570d38aafcd3bc479871a6bfef71 Author: Matteo Frigo Date: Sun Mar 2 07:11:56 2003 -0500 Optionally inline loop in notw codelets commit 4ee60a97aba5df7daa9a1f0f20fc8a18b4caeef9 Author: Steven G. Johnson Date: Sun Mar 2 01:37:41 2003 -0500 updated nodes commit a760bacb99bcb4d1b37deac1a0d03048564f06ae Author: Steven G. Johnson Date: Sun Mar 2 01:37:19 2003 -0500 wrote most of Fortran chapter commit 84b26fd1d2d412fc5dae194fa4f49ea8c5ad803b Author: Steven G. Johnson Date: Sun Mar 2 00:58:37 2003 -0500 citation commit 90c66908b4f24f05f5a77a85d890ef77a5946747 Author: Steven G. Johnson Date: Sun Mar 2 00:57:22 2003 -0500 added parallel FFTW chapter commit 1a89e4fc8d30e58c46d409543e5641d74d82012b Author: Steven G. Johnson Date: Sat Mar 1 20:42:23 2003 -0500 typo commit 125c6e2e61c2977a10fe882134b6daa518d211b6 Author: Steven G. Johnson Date: Sat Mar 1 20:34:38 2003 -0500 added inlining to TODO commit 86f19bdcd118e4f74034a5acf2a9f46ae0dd563b Author: Steven G. Johnson Date: Sat Mar 1 19:36:26 2003 -0500 added K commit c471cfe8ed04c68bd3ba96de578160018676966f Author: Steven G. Johnson Date: Sat Mar 1 19:15:18 2003 -0500 use K for constants commit c9132f12b56356608c7430b1aa8674c57982cf6f Author: Steven G. Johnson Date: Sat Mar 1 19:14:54 2003 -0500 fixed cross-ref commit 2c552e93b7ac76c6ed2cb15d84fb724e71d90901 Author: Steven G. Johnson Date: Sat Mar 1 19:14:16 2003 -0500 whoops commit ffd88e528368512ad6260f9829d093be01b0b8e0 Author: Steven G. Johnson Date: Sat Mar 1 18:50:43 2003 -0500 cleanup commit eb500b0aee97bc247fadc5f14053addd510f8911 Author: Steven G. Johnson Date: Sat Mar 1 18:46:38 2003 -0500 "words of wisdom" by itself is a little too obscure commit c110b9bcf9dc0e3ac3bd0a9dc0aa04a3003808ab Author: Steven G. Johnson Date: Sat Mar 1 18:43:21 2003 -0500 re-added multi-dimensional array stuff commit 3c1809be37bedc7b19bb0ad1645d2d0c55fb24af Author: Steven G. Johnson Date: Sat Mar 1 18:15:22 2003 -0500 added alignment section commit 5ea9d154e8d9b180445e82c228f66dc620435630 Author: Steven G. Johnson Date: Sat Mar 1 16:34:21 2003 -0500 shrunk code commit 969e6184c37360147d4377765e4209f740bbbc63 Author: Steven G. Johnson Date: Fri Feb 28 20:22:00 2003 -0500 slight compression commit 61f49745af277cf662c0b684d812bb937991da02 Author: Steven G. Johnson Date: Fri Feb 28 19:01:20 2003 -0500 style commit 7a450c9741b7d712c4b0647c8348b6f5c16c5b5b Author: Steven G. Johnson Date: Fri Feb 28 18:46:53 2003 -0500 noted not in API commit dcb2c790e6afe7674f917a64a27a5d757de04d54 Author: Steven G. Johnson Date: Fri Feb 28 18:43:14 2003 -0500 more updates commit 9c734e0be5f7e454d53ea076c85b07a1563d12d0 Author: Steven G. Johnson Date: Fri Feb 28 18:38:42 2003 -0500 slight updates commit 3e0a26ba8c35cc39e451dddb4ff538a9b6897853 Author: Steven G. Johnson Date: Fri Feb 28 18:28:58 2003 -0500 great const-ification of apply/solve and print commit 7531ed4ba4a1cd9a4e9caf11c225f930a72efc73 Author: Steven G. Johnson Date: Fri Feb 28 17:51:15 2003 -0500 make fftw_execute take a const plan, to remind the user that it is re-entrant (or should be)... commit 4688736baa020b3ea5f442e36b70d793b431c5c5 Author: Steven G. Johnson Date: Fri Feb 28 17:29:40 2003 -0500 weakening commit 0318454412dbe1cd837ddb068bd343ca6e112011 Author: Steven G. Johnson Date: Fri Feb 28 17:28:48 2003 -0500 note commit 91b816d6c3f80bdb7e0d0116306ae7ffd2c455a8 Author: Steven G. Johnson Date: Fri Feb 28 17:27:10 2003 -0500 footnote about why DHT is provided commit ade0a1b900ff7aad1f0b34334d0aeef444f9c6f1 Author: Steven G. Johnson Date: Fri Feb 28 15:07:03 2003 -0500 index commit 02af64c2431e9ebe1f95750c16596bb16b0130e2 Author: Steven G. Johnson Date: Fri Feb 28 15:05:48 2003 -0500 added DHT tutorial commit 4c0a2b93c6fabdbd47e06a9f6ba76008bcb560a1 Author: Steven G. Johnson Date: Fri Feb 28 14:36:45 2003 -0500 fixed O(n log n) commit fd7ecdadbf64ae5027bac415310c4a98a276db60 Author: Steven G. Johnson Date: Fri Feb 28 14:12:15 2003 -0500 whoops commit ee8d32cc161fa77c6d9566dfb000a80af883f835 Author: Steven G. Johnson Date: Fri Feb 28 14:06:22 2003 -0500 slight improvements commit 0b2ef4ccfd465403919403e5151753a4280f683e Author: Steven G. Johnson Date: Fri Feb 28 00:55:50 2003 -0500 addition commit 22bd399df29e7380522c5bac340a3f04a466fd79 Author: Steven G. Johnson Date: Fri Feb 28 00:54:09 2003 -0500 clarification commit 1b357d49f4d4ee22c59374391be91ddb42813a2d Author: Steven G. Johnson Date: Thu Feb 27 23:49:37 2003 -0500 fix commit 1c30eacc33d5c9d5daf303cfbbc5fa74e6a5bfa4 Author: Steven G. Johnson Date: Thu Feb 27 23:43:56 2003 -0500 slight changes commit 053b9356142e3b05c1ee11800f497813e5c9f119 Author: Steven G. Johnson Date: Thu Feb 27 23:27:48 2003 -0500 added R{E,O}DFTab tutorial commit a793a4024b69b7e4ec4bbbeedb00508845c0cab2 Author: Steven G. Johnson Date: Thu Feb 27 17:24:20 2003 -0500 fixes commit de5b2994a11c8c2b3d1948f43525864b0ac5d265 Author: Steven G. Johnson Date: Thu Feb 27 17:20:42 2003 -0500 fixes commit 027014da3b7f99190c9c1edbe0f6d0c0d15e043a Author: Steven G. Johnson Date: Thu Feb 27 17:11:54 2003 -0500 slight change commit 6359d6080ac4a827218faee02ba1bfe5a5a676bf Author: Steven G. Johnson Date: Thu Feb 27 17:07:45 2003 -0500 documented r2hc/hc2r commit a44e1bc64be97cffdf71bf77dcb526786daa8efe Author: Steven G. Johnson Date: Thu Feb 27 16:19:16 2003 -0500 minor changes commit 7186d1f0701c1507ce6b57f943f0d069c69e09d1 Author: Steven G. Johnson Date: Thu Feb 27 13:54:06 2003 -0500 timed planner and unifying radix-2 butterfly loops are not critical for release commit e22ae82e9d2c007712ae8e8523a2ba4844265b26 Author: Steven G. Johnson Date: Thu Feb 27 13:51:20 2003 -0500 reodft/verify.c no longer exists commit d562aee6ca0c9e2c375d31a2f283ef5188b8819a Author: Steven G. Johnson Date: Thu Feb 27 13:44:19 2003 -0500 optimization: REDFT00 of size 2 is same as R2HC commit 35bca2a3e6f2b887fe4517dfed61eb4cc614f9ff Author: Steven G. Johnson Date: Thu Feb 27 12:35:33 2003 -0500 R{E,O}DFT01 of size-1 is identity commit 3e86434a19f94bd85e576be96fb26b0db8456b7e Author: Steven G. Johnson Date: Thu Feb 27 12:15:10 2003 -0500 minor simplification commit 23aeb956f45a31061c6f0bee5c78119e332e9d20 Author: Steven G. Johnson Date: Thu Feb 27 02:46:31 2003 -0500 fixed add count commit 629bf73abe3666100c7a3cdb795cdf85f1c3467c Author: Steven G. Johnson Date: Thu Feb 27 02:25:04 2003 -0500 whoops commit 46350e9b4b06fc596f73c2e8297276e38871fbcc Author: Steven G. Johnson Date: Thu Feb 27 02:22:03 2003 -0500 another optimization commit 16310c985bad6d32fa0da6362c37fd375822d813 Author: Steven G. Johnson Date: Thu Feb 27 01:43:00 2003 -0500 added op counts commit 870808939ac67893ae3193d1eaf47d6722399743 Author: Steven G. Johnson Date: Thu Feb 27 01:29:32 2003 -0500 cleanup commit e13936e36480509c10d5f8da4806a17a1f2c9d34 Author: Steven G. Johnson Date: Thu Feb 27 01:17:23 2003 -0500 typo in comment commit 32c3d158f7f210901f1c16a8c8cbdfff05024993 Author: Steven G. Johnson Date: Thu Feb 27 01:13:49 2003 -0500 fixed comment commit 6e65b622f4e11f6b75ce19b92715054e01726a87 Author: Steven G. Johnson Date: Thu Feb 27 01:12:05 2003 -0500 use E instead of R commit b2dbcc1af3dac45c5dc937090de39d8c50f79f04 Author: Steven G. Johnson Date: Thu Feb 27 01:05:39 2003 -0500 more unrolling to eliminate if statements in loops, for speedups of 25-40% commit efdfcd1ab423b3b5f4c226859c38fe82ef8d5ee3 Author: Steven G. Johnson Date: Thu Feb 27 00:27:00 2003 -0500 some loop splitting to touch each element of output buf only once and eliminate some conditionals...speeds up by 30-40% commit ac2585fa04303d0a9733f25529a4de770165a96a Author: Steven G. Johnson Date: Wed Feb 26 17:48:26 2003 -0500 comma commit 12f6863d7ba56d03a828d47d95226914f7624343 Author: Steven G. Johnson Date: Wed Feb 26 17:46:17 2003 -0500 pointer to odd case commit b305de27048e5d88018afd557b9853fcfd938e7e Author: Steven G. Johnson Date: Wed Feb 26 17:40:54 2003 -0500 precision -> accuracy (c.f. Kahan) commit 8cce3f1c36041dfd0f3099ccd2b4d07af10ba0ae Author: Steven G. Johnson Date: Wed Feb 26 17:36:13 2003 -0500 added time limit for wisdom generation commit 57f9db2fb5d1498630bc04fa9ce59c0362383dc2 Author: Steven G. Johnson Date: Wed Feb 26 13:24:36 2003 -0500 caps commit 194e3fe2a23b43433042f38567d615508f0219f0 Author: Steven G. Johnson Date: Tue Feb 25 20:56:01 2003 -0500 another note commit 74d5d37f8b5f57257ac2996c1b78cd6e178009b2 Author: Steven G. Johnson Date: Tue Feb 25 20:54:57 2003 -0500 note commit 4c454a521c659245d7d5328a0428abe8e0e65ca0 Author: Steven G. Johnson Date: Tue Feb 25 20:42:08 2003 -0500 added new, more accurate (hopefully) reodft11 algorithms; added --disable-debug-malloc; added --impulse-accuracy-rounds=rounds flags to libbench2 for impulse-response accuracy tests commit 56c91af19d265df468a1c332950285ccc35cadf2 Author: Matteo Frigo Date: Sun Feb 23 14:07:48 2003 -0500 fftw_wisdom.1 is in $builddir, not $srcdir commit afb274d60def917682dcfb6752788ae69feb0e89 Author: Steven G. Johnson Date: Mon Feb 17 03:42:19 2003 -0500 pde commit 57844d17a4a5e42a9b3a6e264d4b9ef96a48b7d7 Author: Steven G. Johnson Date: Mon Feb 17 03:40:19 2003 -0500 consistent number commit da10f4a095936c4a272edf95561177e0ba1e0976 Author: Steven G. Johnson Date: Mon Feb 17 03:39:02 2003 -0500 started r2r doc commit 9339401bc1db11ab2b3ea8332adf2b7f8d2bd39d Author: Steven G. Johnson Date: Mon Feb 17 02:31:51 2003 -0500 rfftwnd commit e9481965be99453d16fce50a2cec8a7189d50e5a Author: Steven G. Johnson Date: Sat Feb 15 17:02:07 2003 -0500 continued commit 387c70c9f598cc84949f9b36c3a7ec3aee478107 Author: Steven G. Johnson Date: Sat Feb 15 15:16:26 2003 -0500 started r2c/c2r docs commit 0df57f98fa114607c9ea5a9e17e8aa4fa92bd0c1 Author: Steven G. Johnson Date: Sat Feb 15 01:12:52 2003 -0500 added r{e,o}dft11 accuracy test commit e24081ffd7a170743a930c91ec251fb1fa590072 Author: Steven G. Johnson Date: Sat Feb 15 00:42:48 2003 -0500 added more r2r accuracy checks commit da37c854fdf95a2cfc3cf2c6ef698ab1ed9e8a70 Author: Matteo Frigo Date: Fri Feb 14 19:19:54 2003 -0500 $< is a GNUism commit 01c0739002308b926e8ed648f93c2b46ef885404 Author: Steven G. Johnson Date: Wed Feb 12 21:02:16 2003 -0500 r2r test cases are in commit e0d1053729fe6e63cfc19bf040c14593ced050c5 Author: Steven G. Johnson Date: Wed Feb 12 21:01:28 2003 -0500 added vector radix to TODO commit eabfd75e1f96eb039ac8ba4f612ad92a5de3f3f2 Author: Steven G. Johnson Date: Wed Feb 12 17:21:33 2003 -0500 fixed cross-ref commit fe1a1f526ac5401ffbb69ddc61b07af2f9c08cfc Author: Steven G. Johnson Date: Wed Feb 12 17:19:56 2003 -0500 shorter synopsis commit 73464a04bcc91f1244cca8812515833da6cad60c Author: Steven G. Johnson Date: Wed Feb 12 12:53:19 2003 -0500 obsolete commit f235c4cdb767ed752563b5a12b609f4a606ae89d Author: Steven G. Johnson Date: Wed Feb 12 12:52:53 2003 -0500 removed old dotens commit 1b45907552bf8c3c7e91e77b9256f904a7dc46db Author: Steven G. Johnson Date: Wed Feb 12 12:52:16 2003 -0500 removed old verify files commit d2baa62fc65ce7b8c09581f2feaacd90466c07e2 Author: Steven G. Johnson Date: Wed Feb 12 12:37:17 2003 -0500 disable threads support by default commit 6fc7d66c60a2e9bfbac7bba821b5329c9fde4b0b Author: Matteo Frigo Date: Wed Feb 12 11:03:28 2003 -0500 Removed old test program commit 948df3a1949a1a5d9d8924a1c51c49d015477b73 Author: Steven G. Johnson Date: Tue Feb 11 22:30:55 2003 -0500 joke commit 06377bf381dbb2e1a05674678924168ee9235d46 Author: Steven G. Johnson Date: Tue Feb 11 22:27:44 2003 -0500 add --help and --version, to be GNU-lly correct commit c59c2fb43df57981f39141efe881ade700dffb3f Author: Steven G. Johnson Date: Tue Feb 11 22:27:18 2003 -0500 whoops commit ecc46199c7967a7164deaa4f6be2ad734eb6c986 Author: Steven G. Johnson Date: Tue Feb 11 22:17:35 2003 -0500 better help commit e73d1cfefcfdffa9a318c184463973e309e1f421 Author: Steven G. Johnson Date: Tue Feb 11 21:47:35 2003 -0500 comma commit dc27e6924a7f6e054e0d542d855d4f62c9545ce2 Author: Steven G. Johnson Date: Tue Feb 11 21:46:12 2003 -0500 formatting commit 06c5acf858b96e548a3d5664252103486c5dbb5e Author: Steven G. Johnson Date: Tue Feb 11 21:45:23 2003 -0500 man pages for tools commit d643ece55b08510928523882ac2213361d1eaf43 Author: Steven G. Johnson Date: Tue Feb 11 19:07:12 2003 -0500 added -V commit ad12cdca62eb5030d1388f12f7278fd1a3eb8a3a Author: Steven G. Johnson Date: Tue Feb 11 18:42:17 2003 -0500 added install-wisdom target commit 83162f468afd0941a99c408ae84e6c35ce43dbb3 Author: Steven G. Johnson Date: Tue Feb 11 18:23:02 2003 -0500 another note commit 52735853d05221978df609981a95f9d89ec03c0a Author: Steven G. Johnson Date: Tue Feb 11 17:32:56 2003 -0500 started r2r accuracy tests (only three kinds covered so far) commit 6fb598e12ddd2e595289c0d399cd7c283425540b Author: Steven G. Johnson Date: Mon Feb 10 22:04:18 2003 -0500 silence warning commit b94eaa910fb2a707a185e743514f009a77663600 Author: Matteo Frigo Date: Mon Feb 10 20:55:20 2003 -0500 gcc bug is now avoided. commit d142433a2935361da613eef685c306e1f86ef8cb Author: Matteo Frigo Date: Mon Feb 10 20:37:54 2003 -0500 Accuracy test commit 3e6c6925a0daf524ddff6ef711ebe2dbf07ebda2 Author: Matteo Frigo Date: Mon Feb 10 07:59:57 2003 -0500 There is no point in precomputing strides for the long-double code, as multiplication by sizeof(long double) cannot be folded into the addressing mode. This change also fixes the gcc-2.95 bug that causes miscompilation of certain codelets. commit 1cdf3be30717cb411fcb7272628ab72dc31ea3d0 Author: Steven G. Johnson Date: Mon Feb 10 02:54:35 2003 -0500 added random r2r tests commit 13fd49dc504be79d65f5c3b254b08572689fcd71 Author: Steven G. Johnson Date: Mon Feb 10 02:44:58 2003 -0500 whoops, bugfix: missing stride for ro10 commit f0926d171845f84e02584361b0a6a9b6c4d68e71 Author: Steven G. Johnson Date: Mon Feb 10 02:21:50 2003 -0500 formatting commit 2ec7cca77de0ed39b104a090158f4f3994f18343 Author: Steven G. Johnson Date: Sun Feb 9 23:24:52 2003 -0500 flop counts for reodft commit 1ec87d09b3698d5c2093d8436ea885225d67191a Author: Steven G. Johnson Date: Sun Feb 9 23:22:15 2003 -0500 declare aligned_main commit 9c3374ad54ec97ed408760b77234ea4980fcd311 Author: Steven G. Johnson Date: Sun Feb 9 20:56:06 2003 -0500 corrected rader op counts commit 6803f88282e3117c77721aff1a96515236b27fb9 Author: Steven G. Johnson Date: Sun Feb 9 20:25:32 2003 -0500 punctuation commit e8cbdde425f97261b79551ea78f87322a4983bf3 Author: Steven G. Johnson Date: Sun Feb 9 20:25:17 2003 -0500 noted need for better estimator commit 156eefce1a365107071ac016b4c818354a98e60b Author: Steven G. Johnson Date: Sun Feb 9 19:58:59 2003 -0500 noted F77 api fix for g77 mangling incompatibility commit e160cbe881f0f509fa09e6eedd76141b439c3ad9 Author: Steven G. Johnson Date: Sun Feb 9 19:30:55 2003 -0500 build f77 header file of constants from fftw3.h commit 370b6e68c535ab81d29047d5fd3a9a48f7e3ebec Author: Steven G. Johnson Date: Sun Feb 9 19:04:53 2003 -0500 updates commit f2c761d6d435ea22fc390b1e388dc0d01a747bd4 Author: Steven G. Johnson Date: Sun Feb 9 19:03:34 2003 -0500 threads f77 api commit b84617e3c6d025d4f13cfa3056ddbdbd5227b961 Author: Steven G. Johnson Date: Sun Feb 9 18:54:00 2003 -0500 finished f77 serial api commit 86446f99fc266c435826ab0f0ca77b48117dd21f Author: Steven G. Johnson Date: Sun Feb 9 18:32:26 2003 -0500 added flops, slight cleanups commit b02c6ea6492b370ac0dde405bc4d899b3b4d4ab7 Author: Matteo Frigo Date: Sun Feb 9 18:11:48 2003 -0500 Oops, forgot #include commit 218af736c45f2ac117c4fe70c79029a7bb26ae33 Author: Matteo Frigo Date: Sun Feb 9 18:08:26 2003 -0500 Removed duplication of stack-alignment code commit 5b5fc6186df8fa5214ae22ebaf84922aab584d90 Author: Steven G. Johnson Date: Sun Feb 9 15:48:15 2003 -0500 allow - to read problems from stdin commit c8e7f4b0b4ed904a7dc8b474f220d17bd061809e Author: Steven G. Johnson Date: Sun Feb 9 15:22:23 2003 -0500 added fftw-wisdom tool commit 216bb0693d91019be789666644d90c1f9afde7a5 Author: Steven G. Johnson Date: Sun Feb 9 15:06:38 2003 -0500 elim. warning commit 8ff159c3583032eb2b661bb50b34d77344f1898e Author: Steven G. Johnson Date: Sun Feb 9 14:24:19 2003 -0500 destroy_input should not contaminate flags of other problems commit c1e578a3c33cee071a10e2f8f49a5dd29f4749ae Author: Steven G. Johnson Date: Sun Feb 9 13:06:11 2003 -0500 updated commit 8b09de262bcd31d3ef04cff36791c389f75b733b Author: Steven G. Johnson Date: Sun Feb 9 13:01:45 2003 -0500 removed overzealous inplace check, which caused problems for rdft2 commit bfb7a5cab2f68265d33dea80716baec602a7c5ef Author: Matteo Frigo Date: Sun Feb 9 08:14:03 2003 -0500 Consistent syntax for RNK_MINFTY tensors commit 0f87db2efc6a2d72c3bb8584c195ee3682e09870 Author: Matteo Frigo Date: Sun Feb 9 07:31:13 2003 -0500 lisply-correct tensor print. We no longer need to parse tensors. commit 14826af57fa8cd1490d3d4d8111e64336ad638a4 Author: Steven G. Johnson Date: Sun Feb 9 03:35:56 2003 -0500 removed completed items commit ba72775e69c1b4ff00b77a37bd0c80312bcc072f Author: Steven G. Johnson Date: Sun Feb 9 03:27:56 2003 -0500 slight renaming commit a96011aa06fe98812ad45afba51a6f1c3ceeab31 Author: Steven G. Johnson Date: Sun Feb 9 03:15:28 2003 -0500 multi-dimensional r2r verifier commit d2c2e3058ab81d087848fdd251e8bb6e92416710 Author: Steven G. Johnson Date: Sun Feb 9 02:40:22 2003 -0500 comments commit 5553af4969fd029313dc53f63201fa9c40acd051 Author: Steven G. Johnson Date: Sun Feb 9 02:38:26 2003 -0500 slight simplification commit 8aa7d693d055305129c2518385e9816529c9a334 Author: Steven G. Johnson Date: Sun Feb 9 02:36:25 2003 -0500 added 1d r2r verifier (triple ugh) commit ef489a80e6559cf2828da23340df129302681dd5 Author: Steven G. Johnson Date: Sat Feb 8 22:23:00 2003 -0500 added vector transforms to random tests commit 826567b4d872cb6920840a850e0a584e0cc015e9 Author: Steven G. Johnson Date: Sat Feb 8 20:59:07 2003 -0500 whoops commit b4d28e3488a63128b0ad1500d2e8b5777eadc8e4 Author: Steven G. Johnson Date: Sat Feb 8 19:52:58 2003 -0500 fixed interaction between dwims for sz/vecsz with rdft2 transforms commit 0c8c54737beaa3db6a119769716f40416ddc7718 Author: Steven G. Johnson Date: Sat Feb 8 19:35:56 2003 -0500 added destroy_input flag/check commit 0e205231678541426c1a0bcd61b0442e7e24ad4a Author: Steven G. Johnson Date: Sat Feb 8 19:11:58 2003 -0500 added rdft2 verifier commit a20a05830b52221eda2f16ab7da1dc80e0e5a050 Author: Steven G. Johnson Date: Sat Feb 8 13:31:14 2003 -0500 an additional check for in-place case commit 6096b268ec7fb9e5c1ad5d41aff355e8f674fd22 Author: Steven G. Johnson Date: Fri Feb 7 17:36:56 2003 -0500 slight fix: hc2r constraints are mostly determined by sub-plan commit 668b0af47a07011aaa3202ee70d3588aeca0ddd9 Author: Steven G. Johnson Date: Fri Feb 7 16:28:55 2003 -0500 make radix2-dft inapplicable to in-place/split case (r == rio, iio >= rio + n/2+1 != r + 1) commit 533f0a1824842664dfe63287e03800c2426b8ba5 Author: Matteo Frigo Date: Tue Feb 4 06:36:29 2003 -0500 Allow plnr->hook to be 0 commit 5c89a91a0efc3714980409d12055f91a2bd33693 Author: Steven G. Johnson Date: Tue Feb 4 03:25:36 2003 -0500 moved dft stuff into verify-dft commit 941da36d0fa1562ef98fd796b05c0a8a94c4ff94 Author: Steven G. Johnson Date: Tue Feb 4 03:25:00 2003 -0500 cruft commit 727cc86ece827ba21a236149c66ef6c7e7890d6b Author: Steven G. Johnson Date: Tue Feb 4 03:18:28 2003 -0500 further unify libbench2 and paranoid verifiers commit 6fb68912913cd9ab647b0206a713470e1bad462b Author: Steven G. Johnson Date: Sun Feb 2 01:45:37 2003 -0500 typo in comment commit 3d1a5701f571ec275672faf3da2d7ea6f1e34b93 Author: Matteo Frigo Date: Sat Feb 1 09:30:03 2003 -0500 Fixed p==2 case commit 723093b36b481e0f742822129f33998ba5acff14 Author: Matteo Frigo Date: Sat Feb 1 09:23:43 2003 -0500 Incorporated new find_generator by Greg Dionne. commit 93a75fda2dad56fbf69030eabdb09af0987e5316 Author: Matteo Frigo Date: Fri Jan 31 20:46:24 2003 -0500 Removed nonportable call to gettext() commit 0d937fc4f0800cdad67d7a6a496c30c67c70b0ae Author: Matteo Frigo Date: Wed Jan 29 19:03:43 2003 -0500 uintptr_t is in in openbsd commit 9ffa4f6b400e1818a4c50a1385d916d501ff16b7 Author: Matteo Frigo Date: Wed Jan 29 15:41:56 2003 -0500 Huge speedups in wisdom I/O. commit 426e786cc0662f3926cd79d3d76b0825a65ff445 Author: Matteo Frigo Date: Tue Jan 28 19:36:51 2003 -0500 Added appropriate warning against likely future bug. commit b254ecc51abc22f1642e0bae9d6d22fbb2efb771 Author: Matteo Frigo Date: Tue Jan 28 19:00:24 2003 -0500 Don't attempt to remove bogus wisdom entries. commit c19570082c79ce6d86613248e700ee17bb3582b8 Author: Matteo Frigo Date: Tue Jan 28 18:16:24 2003 -0500 Fixed a couple of very very very nasty bugs---pointers became invalid after the hash table was relocated. commit 123972fa083c9fb07f18c3ee3a902a79606f5987 Author: Matteo Frigo Date: Tue Jan 28 07:34:10 2003 -0500 Read wisdom at can_do() time, otherwise wisdom is destroyed. commit d1e805e6353a689a61b6aec66a28d568723717fc Author: Matteo Frigo Date: Tue Jan 28 06:54:38 2003 -0500 More conservative inheritance of blessings commit e718fe3fa7a7c4194011493e0bd86b78b222c0b9 Author: Matteo Frigo Date: Tue Jan 28 06:50:20 2003 -0500 Print the same info as it is hashed commit 84199fe5035171395b24754b6f4428513b850e84 Author: Matteo Frigo Date: Tue Jan 28 06:49:48 2003 -0500 Print name of executable when FAILURE commit 3919d8a49fb4779e470deefd35cc3c7fc09c20ce Author: Matteo Frigo Date: Mon Jan 27 06:59:40 2003 -0500 New NO_SEARCH planner flag, which avoids searching altogether. A wisdom entry must lead to a NO_SEARCH-grade plan, or else the wisdom entry is bogus. commit 9534126e49e082098917ef5500133d8ef8a7289a Author: Matteo Frigo Date: Sun Jan 26 20:45:21 2003 -0500 Use cosl()/sinl() when appropriate commit 5cc66fc2964feb54cff148e70280c083715d371f Author: Matteo Frigo Date: Sun Jan 26 16:29:18 2003 -0500 Use null pointers when estimating. The estimator should never time anything. commit 8a54d02af36535be471d8326bf4e061165295320 Author: Steven G. Johnson Date: Sun Jan 26 15:19:01 2003 -0500 note commit 127681d03bd37c45649032138e7c976ec3395c99 Author: Steven G. Johnson Date: Sun Jan 26 15:16:22 2003 -0500 support multiple mangling schemes with g77 commit 757b13e27cfe6317d5c871796c129ec5b693e89b Author: Steven G. Johnson Date: Sun Jan 26 12:58:57 2003 -0500 fixed verbose, made random tests only use selected rank, use rank <= 4, fixed final flush_problems call commit c379edca317112097e76dacd0dfb69c83c319023 Author: Steven G. Johnson Date: Sun Jan 26 12:42:49 2003 -0500 fixed typo (count instead of maxcount) commit 6c0c2a4aac442f27536a584f1e619c69f6aa7ca6 Author: Steven G. Johnson Date: Sun Jan 26 12:12:07 2003 -0500 hypot is no longer used commit 8466e0fb929081a67acbe832ddd155f33ee13734 Author: Steven G. Johnson Date: Sun Jan 26 12:07:43 2003 -0500 check for _alloca (MSVC) commit 34321edf6b705ea8f04c0ac903baf9a2d0239cd2 Author: Steven G. Johnson Date: Sun Jan 26 11:56:53 2003 -0500 slight fix in assert commit c099d12f16d6131750ccde572b7651661b84881b Author: Matteo Frigo Date: Sun Jan 26 11:55:39 2003 -0500 Allocate problem in all cases--- can_do may need correct pointers. commit 93ba509b275e8a2b798b237dae50927c04da5b74 Author: Matteo Frigo Date: Sun Jan 26 11:51:27 2003 -0500 Nastier checks commit 91419140c877e227d804c4cbb18cb89b350527b3 Author: Matteo Frigo Date: Sun Jan 26 11:51:16 2003 -0500 X(use_plan) is a relic. commit 9cc664aacbc213b2cdbca13e686ca9f15f4d89f4 Author: Matteo Frigo Date: Sun Jan 26 09:23:16 2003 -0500 Print full pathname of the bench executable, so that I don't get confused when running multiple tests for different configurations. commit a755e0b1e768f7624d20ba4d564d9b658fc8aa45 Author: Matteo Frigo Date: Sun Jan 26 07:35:46 2003 -0500 Split done() into done() and cleanup(), in order to test multiple problems with the same planner from the command line. commit c9a2310aa41b815190cd73c801d28f6b68635734 Author: Matteo Frigo Date: Sat Jan 25 20:44:49 2003 -0500 Improved readability commit 6a7d0ba4578fa4f1989e521e80cd1504dddb5ff9 Author: Steven G. Johnson Date: Sat Jan 25 19:17:26 2003 -0500 comment commit 7e5332d67aa4dd505518874798560834170c2d1c Author: Steven G. Johnson Date: Sat Jan 25 19:16:53 2003 -0500 added macos9 mpallocatealigned function commit f3bba67e15e3ff2cc63e615a97ef4161af9fbb6d Author: Steven G. Johnson Date: Sat Jan 25 18:59:55 2003 -0500 sometimes __APPLE__ is defined instead of __MACOSX__ commit eb44a626174b2c4b5a3b91799f929f087ab89b90 Author: Steven G. Johnson Date: Sat Jan 25 18:54:39 2003 -0500 macos x malloc is already 16-byte aligned commit e1f4dfe3d7d517b655cbf1d9f34910cf5b91f16f Author: Matteo Frigo Date: Sat Jan 25 13:38:32 2003 -0500 Include because uintptr_t is defined there on solaris. commit 2e0d88fe660fa8d5dd70ac8b4d7ce327b8e3143a Author: Matteo Frigo Date: Sat Jan 25 13:22:59 2003 -0500 Oops---forgot getopt_long commit 8ca5ca5adffa7f09e53fd6876720807c9e09b526 Author: Matteo Frigo Date: Sat Jan 25 13:17:29 2003 -0500 Include default includes when checking for uintptr_t. (Otherwise solaris breaks.) commit b2e7887137a70e836841860650f673a32d8fd0e0 Author: Matteo Frigo Date: Sat Jan 25 12:39:52 2003 -0500 distribute check.pl commit f523570817e6d4e02d1229eb4fae65aa54b39c90 Author: Matteo Frigo Date: Sat Jan 25 12:38:34 2003 -0500 Check split format, too. commit 4cf6b31bc1d606a85ebe86b81538440c32ba16d4 Author: Matteo Frigo Date: Sat Jan 25 11:48:19 2003 -0500 New tests, added make check commit cc595c7702af171d1850e32593ad093a1884fa98 Author: Matteo Frigo Date: Thu Jan 23 08:34:24 2003 -0500 More tests commit 132d24bf7371a5738a8703d6700452432c1ff8d6 Author: Matteo Frigo Date: Tue Jan 21 20:32:12 2003 -0500 Deal with rnk(sz)=-infinity commit dbf5eba2cfe458f7fa1853b8b73ac880f50268d3 Author: Matteo Frigo Date: Tue Jan 21 10:07:16 2003 -0500 Crazy idea commit 538d043b618e83f8c9dd443618e30fad09412560 Author: Matteo Frigo Date: Tue Jan 21 07:14:22 2003 -0500 Test program, still barely worthy of the name. commit bd13e47fca93beafd8c87bd039e4c7f6f9843cc0 Author: Matteo Frigo Date: Mon Jan 20 08:29:21 2003 -0500 Stylistic changes commit 7a7f938bfa0596d8a971476e304a584e80c9af3e Author: Matteo Frigo Date: Mon Jan 20 07:03:38 2003 -0500 Implemented flops api commit 06f9de2ae48e8508332300af57ce4a892d5d7327 Author: Steven G. Johnson Date: Sun Jan 19 14:27:21 2003 -0500 cleanup commit 0004f3c1d04b2dbf2cd0c329464a761f513b17e8 Author: Steven G. Johnson Date: Sun Jan 19 14:14:49 2003 -0500 'v' syntax now defaults to an 'internal' (stride 1) vector, which is a more interesting case and corresponds more closely to the intuitive notion of a 'vector' transform, while '*' does the old 'external' (stride n) vector commit 3ae6aeb8ad69f728e24a22eaff8cb1c2d769dbfd Author: Steven G. Johnson Date: Sun Jan 19 13:55:35 2003 -0500 removed '/' overloading commit 16e5b7c653597353fa972d5da6226e3d1c21f09c Author: Steven G. Johnson Date: Sun Jan 19 13:52:09 2003 -0500 get rid of '*' and ',' synonyms for 'x' in problem parser; there's no need to clutter the namespace with syntax we never use commit fe570b1a3ef49b842a35c74088e1893023c924a3 Author: Matteo Frigo Date: Sun Jan 19 07:28:27 2003 -0500 Signed/unsigned fixes. commit 39087e0b7d51d64ce70403c94042723a27ebd90e Author: Matteo Frigo Date: Sun Jan 19 07:09:54 2003 -0500 Test split arrays. commit f5c448ba8c68ad9343a147b9ee0edddd48101248 Author: Steven G. Johnson Date: Sat Jan 18 23:46:57 2003 -0500 clarification commit ce827c93bd4b66fca1e4c6925c9638fc061f2a9c Author: Steven G. Johnson Date: Sat Jan 18 21:53:18 2003 -0500 caps commit 8e0bc243bfa4a19e901e09af2175220823a29fda Author: Steven G. Johnson Date: Sat Jan 18 21:52:51 2003 -0500 brackets commit 205193db72e85418b6db84064c2d0c417d3622ae Author: Steven G. Johnson Date: Sat Jan 18 21:52:32 2003 -0500 quote commit 53b6dc0784f2f573114f99a64e3c3a3f5c25d144 Author: Steven G. Johnson Date: Sat Jan 18 20:53:11 2003 -0500 referencing commit dc903b262a3cdbfacda95f8cacf08a79b26a3725 Author: Steven G. Johnson Date: Sat Jan 18 20:33:28 2003 -0500 fix commit 34867e8b93f1464aeb74afe7a57e6db29a6bf6ef Author: Steven G. Johnson Date: Sat Jan 18 20:31:41 2003 -0500 slight change commit de2f4e199030747045d6b15f10f81015e6fa77c9 Author: Matteo Frigo Date: Sat Jan 18 20:31:22 2003 -0500 Print errors when --verify. commit a241dce3b13972ae124686d2a73d6845172dca10 Author: Steven G. Johnson Date: Sat Jan 18 20:30:27 2003 -0500 improved description, noted that FFTW_ESTIMATE does not destroy arrays commit de9ad7da59f6f405cb9698340a708c51879074fa Author: Steven G. Johnson Date: Sat Jan 18 20:23:12 2003 -0500 FFTW_DEFAULTS isn't really needed commit 742ec9578cb87f7e8640c998b6455f0c1347cbad Author: Steven G. Johnson Date: Sat Jan 18 20:21:09 2003 -0500 added FFTW_MEASURE synonym for FFTW_DEFAULTS commit 7e4c0117633ecc6c774e5747fb88e5d9b901ade1 Author: Steven G. Johnson Date: Sat Jan 18 20:18:29 2003 -0500 slight change commit 1f5d8e6883c07b8b55b3ccdd76728dba0db83b51 Author: Matteo Frigo Date: Sat Jan 18 20:16:08 2003 -0500 Clearer name commit 72f6ff219f76d5836c974d7739c9deb1fdaae1b1 Author: Matteo Frigo Date: Sat Jan 18 20:13:14 2003 -0500 Completed dft api test commit d98d355d8025c3244f40cb21d3c13fd49b95bb31 Author: Steven G. Johnson Date: Sat Jan 18 20:07:33 2003 -0500 index commit 5abf9be2a1e971d3911958c2f2b0f830c1e94507 Author: Steven G. Johnson Date: Sat Jan 18 20:05:50 2003 -0500 fix commit 6ce8d648e9020903839bb75540e2c7f31c350a77 Author: Steven G. Johnson Date: Sat Jan 18 20:04:11 2003 -0500 parallel structure commit bf5e342ad8e4ec2778b7cb07e9fe04c3b88fef1c Author: Steven G. Johnson Date: Sat Jan 18 20:03:18 2003 -0500 fix commit 27f73ffc519eca88af9d51fccafa9d0a9eaec3d7 Author: Steven G. Johnson Date: Sat Jan 18 20:00:24 2003 -0500 joke commit fc0561411a690340303ab579fe66b5b919e94706 Author: Steven G. Johnson Date: Sat Jan 18 19:59:28 2003 -0500 recommendation to read tutorial in-order commit 23f008eece1d5af1cc9aff5c21f16c4b78626a43 Author: Steven G. Johnson Date: Sat Jan 18 19:54:55 2003 -0500 expanded outline commit dd05ed963f2b3b6248a90c8a28ec92ed0748447f Author: Steven G. Johnson Date: Sat Jan 18 19:35:52 2003 -0500 clarification commit 45f4203a263004153eb30c5e2b6d5fbf7d363ebe Author: Steven G. Johnson Date: Sat Jan 18 19:17:27 2003 -0500 draft complex-dft tutorial commit a1cf23e6204a958c2adb5fa5ad6908ed9ae8d5aa Author: Matteo Frigo Date: Sat Jan 18 17:27:15 2003 -0500 Paranoid mode is back. Fixed dwim to do what I mean. commit 02a981bcb005fd082e832f912a0d6970469af2db Author: Steven G. Johnson Date: Sat Jan 18 17:13:51 2003 -0500 started tut. commit db27392f9335988028063634e188cd6e0329b2db Author: Matteo Frigo Date: Sat Jan 18 16:13:15 2003 -0500 Great renaming, so that we can include both bench-user.h and ifftw.h to implement the paranoid-mode hook. commit 272ce9998c6c2ba1440c85c89adf525029c3713c Author: Matteo Frigo Date: Sat Jan 18 15:41:18 2003 -0500 Trying to tweak the verifier so that I can use it in bench.c for paranoid mode commit 9406410c744c3d040dcf53cab0033e6289315e0d Author: Matteo Frigo Date: Sat Jan 18 10:24:05 2003 -0500 Added stride_factor for complex arrays. commit be5440925e131346debad7cb5c52ec9ccca20838 Author: Matteo Frigo Date: Sat Jan 18 10:02:11 2003 -0500 can_do now calls the planner. commit 1c2aa801bd04200c319430596f26e33c57ade5b7 Author: Matteo Frigo Date: Sat Jan 18 09:59:24 2003 -0500 Call guru api in bench.c commit 5ccc685036846da380536544c08668012a62953a Author: Matteo Frigo Date: Sat Jan 18 08:17:23 2003 -0500 Fixed prototype. commit 6a0efba859963432de8d7ddef8a68615fba215df Author: Matteo Frigo Date: Sat Jan 18 08:14:48 2003 -0500 Attempt to make the signed/unsigned use of flags consistent. commit 6c6caca90a9df0f2f76cae61abf4d5b4108e5a16 Author: Matteo Frigo Date: Sat Jan 18 08:03:07 2003 -0500 Implemented useropt. commit 7165449ca5470fe7104141090f15d804f8fa3d58 Author: Matteo Frigo Date: Sat Jan 18 08:02:05 2003 -0500 The first map_flags pass must be transitive, i.e., always use the latest flags value as opposed to the original value. (I think.) commit b5ff8655a95e88173c98942113dedb0b8f293154 Author: Matteo Frigo Date: Sat Jan 18 07:20:19 2003 -0500 Started working on verifier commit 6c1864f54390a4ba6483dd6f6af716030275af40 Author: Steven G. Johnson Date: Fri Jan 17 14:53:28 2003 -0500 added X(threads_cleanup) commit 53ccbeeb98ce85aeee2dfdc73a79518dd428cabd Author: Matteo Frigo Date: Fri Jan 17 10:35:56 2003 -0500 Use C style for upper and lower array bounds. Free tensors properly. commit 8f979d12529a8cdcbc19773db64b203d396667f3 Author: Matteo Frigo Date: Fri Jan 17 08:50:42 2003 -0500 Fixed ambiguous syntax commit b4a79fbfcd90a04148e114c3dc9ffeec57475b91 Author: Matteo Frigo Date: Fri Jan 17 08:20:57 2003 -0500 Parse minus sign, bugfixes commit 74b7faa7902bc94ba6a2cb2229b29a0ae7fc1ae6 Author: Matteo Frigo Date: Fri Jan 17 08:11:56 2003 -0500 Skeleton libbench2 implemented (probably still buggy) commit e589fb07c231478fcaac2ff1747634bf9f06ea8f Author: Matteo Frigo Date: Fri Jan 17 04:23:37 2003 -0500 Formatting commit ca9524db0c137f154e83a76d36cf935f00674f2e Author: fftw Date: Fri Jan 17 03:15:24 2003 -0500 slight updates commit 9cf580eecfb6efdc94025f0016482c3b39e42d44 Author: Steven G. Johnson Date: Fri Jan 17 01:44:44 2003 -0500 eliminated obsolete uimin/uimax commit 43e7097cd5f50fec4d5cba68968062d735c70118 Author: Steven G. Johnson Date: Fri Jan 17 01:40:10 2003 -0500 threads needs to have its own library, lest all programs linking to libfftw3.so need -lpthread commit dfbd69e73262bfd32e4238660b05e9e66f2d4639 Author: Steven G. Johnson Date: Thu Jan 16 19:53:46 2003 -0500 whoops commit 2270fad47a873f34165771451625eb46b32f8934 Author: Steven G. Johnson Date: Thu Jan 16 19:53:30 2003 -0500 better name commit 90d92f5270d46d9e8f4775937e55433d425a5706 Author: Steven G. Johnson Date: Thu Jan 16 19:52:36 2003 -0500 added more functions commit 3f06842ca4733e7ecabf350ae1e679d52ed7698a Author: Steven G. Johnson Date: Thu Jan 16 16:57:06 2003 -0500 if 'long' is big enough, use it for mulmod in preference to 'long long' commit 66e1948c825bca967b2ad7e6746242e8b23f2b00 Author: Steven G. Johnson Date: Thu Jan 16 14:53:41 2003 -0500 use uintptr_t for pointer alignment arithmetic commit 7eb1f83c40d65241a97769cbd182b979f54b3694 Author: Matteo Frigo Date: Thu Jan 16 07:58:28 2003 -0500 More signed/unsigned cleanup commit 45b331a5c2824f7d0d08df9385910c66db337edf Author: Matteo Frigo Date: Thu Jan 16 07:57:40 2003 -0500 null function pointers are technically nonportable commit 67822e08115a08b056287208aa8db4cf8679eeb5 Author: Matteo Frigo Date: Thu Jan 16 07:17:45 2003 -0500 Free short_options commit 477c8d3241c4d9943d025ae59f9305a0b149231b Author: Matteo Frigo Date: Thu Jan 16 05:48:30 2003 -0500 Oops, forgot STACK_FREE commit 0e20238b7462741468c08db5854a75106766b2ef Author: Matteo Frigo Date: Thu Jan 16 05:40:39 2003 -0500 Do not require memalign() unless HAVE_SIMD commit e1ab6010079824a6d2eba12510455609646681fc Author: Steven G. Johnson Date: Thu Jan 16 01:03:31 2003 -0500 MS VC++ _aligned_malloc commit b60bc7e076569eb05d30aea259d6d6347e6a2da0 Author: Steven G. Johnson Date: Thu Jan 16 00:44:45 2003 -0500 added api fftw_malloc/free commit fce03e8f76d32e4642d3e3abe4ace0d60e5e14f5 Author: Steven G. Johnson Date: Thu Jan 16 00:43:48 2003 -0500 silence warning commit 641795cb961dfc1336f70563c2c7ad1ed3192395 Author: Steven G. Johnson Date: Wed Jan 15 22:39:04 2003 -0500 send error output to stderr commit f6710096b2309498d0d21582380e4edf3f3cc75c Author: Matteo Frigo Date: Wed Jan 15 13:20:35 2003 -0500 Pure paranoia. commit 91f5030882cd7a147a68a99634aa5e2b962998cf Author: Matteo Frigo Date: Wed Jan 15 06:51:34 2003 -0500 Fixed formatting that was messed up by the conversion uint->int. Ensure that iodims etc are kosher. commit e013a32092d6ec5aa0e9f2d9ae6c26d4b8659c6f Author: Steven G. Johnson Date: Wed Jan 15 01:32:18 2003 -0500 added version stamp commit 50b479b4aac66242696e7fd98f58455325526959 Author: Steven G. Johnson Date: Wed Jan 15 01:28:20 2003 -0500 added warning commit a1084fccb8215cfd46c69f6b5eeb7ff22f358d82 Author: Steven G. Johnson Date: Wed Jan 15 01:23:25 2003 -0500 add fftw-wisdom-to-conf to BUILT_SOURCES commit 4b8e34f3dba941ca2f59b9705ee49a9f29951906 Author: Steven G. Johnson Date: Wed Jan 15 01:09:29 2003 -0500 added const commit e3063ad93de5985a0cea8fcc35052dfdd31d3f24 Author: Steven G. Johnson Date: Wed Jan 15 01:04:10 2003 -0500 added wisdom-to-conf commit f1bc153c63191407f4af84ca6641b4153481abca Author: Steven G. Johnson Date: Wed Jan 15 00:23:36 2003 -0500 include type prefix in wisdom preamble commit 564b63e0eb961ab85824847dd4171323d185f2d3 Author: Steven G. Johnson Date: Wed Jan 15 00:02:31 2003 -0500 updates commit eed0a2c1a6165c360b7f87ff1aa77341dc112be5 Author: Steven G. Johnson Date: Tue Jan 14 23:59:26 2003 -0500 check the_plan before printing commit b90c45ecd325b1cbb5821b7d22b7d1003a01e11b Author: Matteo Frigo Date: Tue Jan 14 21:10:25 2003 -0500 Eliminated those unsigned values that would break LP64 machines. commit 2cfc97931df736f5090ba7eec7fa6d13686c6899 Author: Steven G. Johnson Date: Tue Jan 14 15:14:29 2003 -0500 comments commit 3b9adee3905d5c9686dd26e6af706297c57d3e6e Author: Matteo Frigo Date: Tue Jan 14 08:00:08 2003 -0500 Oops commit b8ef56b0756c8db296926946f027105168ac91c9 Author: Matteo Frigo Date: Tue Jan 14 07:59:14 2003 -0500 int/uint confusion commit 4063890615e1ebdd337cd0b6b79e3d8c191f7ac7 Author: Steven G. Johnson Date: Tue Jan 14 02:25:33 2003 -0500 updated introduction and some organization commit 23ce88399655bc3c3f102fb81927f18f964381d7 Author: Steven G. Johnson Date: Tue Jan 14 01:34:46 2003 -0500 whoops commit b165736884413d29ac6ea2d63b7784ebf40c8400 Author: Steven G. Johnson Date: Tue Jan 14 01:33:04 2003 -0500 newline commit 02a1859f44c60c5452b4d9fb3e89ecdac0d57873 Author: Steven G. Johnson Date: Tue Jan 14 00:23:04 2003 -0500 added win32 timer commit d0e64f8319671968827241d6923c1dcc613734ec Author: Steven G. Johnson Date: Tue Jan 14 00:12:21 2003 -0500 sync with kernel/alloc.c commit 1e179069c40aafd83bbaedf588ced907c60d8f7d Author: Steven G. Johnson Date: Tue Jan 14 00:03:20 2003 -0500 handle missing F77_FUNC_ commit d1e7472bbe33eaf99e4464fea7629ea9dc2549d8 Author: Steven G. Johnson Date: Mon Jan 13 17:42:50 2003 -0500 used fint instead of int to make Fortran integer type easier to change commit 2a5dd8f944a6ed354d8245abf6cc67de05ca7457 Author: Steven G. Johnson Date: Mon Jan 13 17:38:56 2003 -0500 slight abbreviation commit 1371e68a5061a7de34681052e5c7f31139752046 Author: Steven G. Johnson Date: Mon Jan 13 17:35:20 2003 -0500 the great lengthening, part I: int -> long in api; mv mktensor-rowmajor to api commit 1011711ec3ed3d1252ee9ea5134e8e18a9925081 Author: Steven G. Johnson Date: Mon Jan 13 15:23:22 2003 -0500 long types commit 2f236bb6f4b8d4b68a2799c59eed45c3fa5d9bef Author: Matteo Frigo Date: Mon Jan 13 04:20:37 2003 -0500 Renamed fftw_malloc -> MALLOC, X(free) -> X(ifree), X(free0) -> X(ifree0), non_fftw_malloc -> NATIVE_MALLOC commit ab8d02fc9ecab18a2639a0167616e782995592eb Author: Steven G. Johnson Date: Mon Jan 13 02:37:22 2003 -0500 added beginning of Fortran interface commit f2c44ba05c22d8cab1e72c5393ba64e97fb4eb57 Author: Steven G. Johnson Date: Mon Jan 13 01:05:29 2003 -0500 add fortran mangling check commit 59c96c1e23c8b4b4830a0f1aa70d8715a57db138 Author: Steven G. Johnson Date: Mon Jan 13 00:33:28 2003 -0500 added guru r2r interface commit d1b297f4a235356f816342e21c1f69617d836a4f Author: Steven G. Johnson Date: Mon Jan 13 00:23:26 2003 -0500 whoops commit 07839004aaa3e10e2493cf14fcaf6c603703ecf6 Author: Steven G. Johnson Date: Mon Jan 13 00:16:20 2003 -0500 added r2r planner commit f0e64dbc84dcd207fcf9ab13bf270707e878b9cb Author: Steven G. Johnson Date: Sun Jan 12 22:58:18 2003 -0500 more long-double checks commit fc870a86543c6166b8f3c689278c3e6c429c8fb5 Author: Steven G. Johnson Date: Sun Jan 12 20:01:51 2003 -0500 slight regrouping commit 889820ff1b5fe4f8ccf4c0f321cbcfc6066facc6 Author: Steven G. Johnson Date: Sun Jan 12 19:58:46 2003 -0500 added joke commit 709357f9b9ad15dbc409491672174b0369364de5 Author: Steven G. Johnson Date: Sun Jan 12 19:53:58 2003 -0500 simplified rdft2 padding commit b724cc6adafccd6b09b69ea2433567634d2b18fc Author: Steven G. Johnson Date: Sun Jan 12 19:02:09 2003 -0500 added comment commit f7e00499811c30295febdd6d70699c0bee9c2260 Author: Steven G. Johnson Date: Sun Jan 12 18:54:49 2003 -0500 use latest api commit 17dfd8a88bbd556d885e59de2c75f6ed10666ee3 Author: Steven G. Johnson Date: Sun Jan 12 18:49:58 2003 -0500 nembed should only be in advanced (many) interface, not basic interface...only a handful of people over the years have ever requested that functionality. commit de10a37b79b7222dff049d7b17a2f52c4d8818f7 Author: Steven G. Johnson Date: Sun Jan 12 18:41:57 2003 -0500 impatient is default; generalize mapping functions using xor trick commit d759ad32aae7ec487f4bacbacd50c36e9b9252de Author: Steven G. Johnson Date: Sun Jan 12 14:39:42 2003 -0500 use NULL nembed to signal padding commit c52303f271b00a8388b368b13e26e492e34e8ac1 Author: Steven G. Johnson Date: Sun Jan 12 14:23:00 2003 -0500 accept NULL nembed commit 130e62b15d813a605dbe261661f3d4f73eefd869 Author: Steven G. Johnson Date: Sun Jan 12 13:57:13 2003 -0500 added execute-dft-r2c/c2r commit 1fd627fbdef40e0158d1880e765131d7316614eb Author: Steven G. Johnson Date: Sun Jan 12 13:43:20 2003 -0500 don't need dft.h commit 116ca5713809a7a18bea146e4e1d2c13679f0570 Author: Steven G. Johnson Date: Sun Jan 12 13:22:14 2003 -0500 tensors are compressed in the problem, duh commit 2ede363d3dc04c22c6d801931c613f8acb365f20 Author: Steven G. Johnson Date: Sun Jan 12 12:45:26 2003 -0500 noted that posix_memalign bug is now fixed, thanks to bug report by yours truly commit aa78a752a45559c3cd10009619c38714715b4bd8 Author: Matteo Frigo Date: Sun Jan 12 12:44:43 2003 -0500 Bug: n[3] instead of n[2]. Bug was propagated by copy-and-paste. Grrr... commit da61449b6d55793e890eaf9246c2cef570656949 Author: Matteo Frigo Date: Sun Jan 12 12:41:43 2003 -0500 Express plan_dft() in terms of plan_many_dft() commit f50b2491505035d8da53cdc2f807f777ab7f2fa2 Author: Steven G. Johnson Date: Sun Jan 12 12:19:53 2003 -0500 whoops commit 91650cec6fb479345ace984c86a3d0bf8dd45fa3 Author: Matteo Frigo Date: Sun Jan 12 06:00:46 2003 -0500 Manual skeleton. commit 58983b0fcb5dde376eef5290f5afeda420bb3516 Author: Steven G. Johnson Date: Sat Jan 11 23:46:34 2003 -0500 added r2c/c2r guru api commit 7ab4791d61cf0a563110b7c4458f092ad3209452 Author: Steven G. Johnson Date: Sat Jan 11 23:42:10 2003 -0500 FFTW_DESTROY_INPUT is default for c2r transforms commit 702d37e3aa20ac9e2007f9415a6c09875dc58eec Author: Steven G. Johnson Date: Sat Jan 11 23:36:26 2003 -0500 added more of r2c/c2r api commit 877b1c30a8de302c16d17618928ea9eeafa1d840 Author: Steven G. Johnson Date: Sat Jan 11 21:09:41 2003 -0500 r2c doesn't have adjustible sign commit d7e17c10e9b94495bf5b8d91ee938bddb15a778f Author: Steven G. Johnson Date: Sat Jan 11 21:07:55 2003 -0500 note that copyright year is out of date commit 414ef3efe34a68c1d1886ebc7bf3696c9888312d Author: Steven G. Johnson Date: Sat Jan 11 21:04:23 2003 -0500 updated api for r2c commit f55aa9fa016782becff68f499151eb9b1142f48c Author: Steven G. Johnson Date: Sat Jan 11 21:00:07 2003 -0500 removed annoying nophys == niphys case commit 5ac383f909dedb9038b26d2534d1c50f831bb622 Author: Steven G. Johnson Date: Sat Jan 11 20:58:13 2003 -0500 added basic r2c/c2r planner commit 67a72b6fe3950808458f9db07fd17fdb5a23f174 Author: Steven G. Johnson Date: Sat Jan 11 19:34:14 2003 -0500 dist should be in terms of complex values commit 38330465de7aa72398c41e080a538abe90f11525 Author: Steven G. Johnson Date: Sat Jan 11 19:14:24 2003 -0500 added plan-with-nthreads commit 250cd26e00612f247ec647a8b1cd12757c6bc2dd Author: Steven G. Johnson Date: Sat Jan 11 19:12:51 2003 -0500 added function to set nthr commit 2a3a928928d0a1a720099f63d46f9c9335e60d07 Author: Steven G. Johnson Date: Sat Jan 11 18:04:57 2003 -0500 slight cleanup commit c70d4ae24f18ed8573746982ded357d7a66e45d7 Author: Steven G. Johnson Date: Sat Jan 11 17:57:29 2003 -0500 whoops commit f6eead982eac7fba05e3e1b211e92218fa75ac0e Author: Steven G. Johnson Date: Sat Jan 11 17:55:39 2003 -0500 maxlen is maximum string length, not including null termination commit 7d5ced1616625f49a8064e213c195cb0dfdfd015 Author: Steven G. Johnson Date: Sat Jan 11 17:50:49 2003 -0500 imprt reverts hashtable on failure commit 8b8397f47fca460f6f4d799ed4d3523dbb9febe4 Author: Steven G. Johnson Date: Sat Jan 11 16:43:54 2003 -0500 slight move commit ef10382faf88c76dbed1b15712ea4385f5c53d60 Author: Steven G. Johnson Date: Sat Jan 11 16:34:56 2003 -0500 stdio.h should be inlcuded outside of extern "C" commit a12b4db5cb652f45955ba8597ac6aad5494d10f2 Author: Steven G. Johnson Date: Sat Jan 11 16:26:35 2003 -0500 added guru planner API commit 6612a3a2112dc6db386ca1a7c9e80e0cdff8060a Author: Steven G. Johnson Date: Sat Jan 11 15:54:57 2003 -0500 added FFTW_FORWARD/BACKWARD commit e9182c7a6103c19c79d2d457aadfd3cbafc8e7f8 Author: Steven G. Johnson Date: Sat Jan 11 15:52:17 2003 -0500 added plan_many_dft commit 93581dbc842eb787a6a1f514d9ae4a3af66da1ae Author: Steven G. Johnson Date: Sat Jan 11 15:44:37 2003 -0500 indenting commit 4b42a448907aaef4bce3be179fe2676f89dc7580 Author: Matteo Frigo Date: Sat Jan 11 14:49:08 2003 -0500 Final \n commit 3174ca24fb957b047983215e5651a9f5db6a6687 Author: Matteo Frigo Date: Sat Jan 11 14:47:31 2003 -0500 Do not compile if not defined(FFTW_DEBUG), in order to avoid unused code in the shared library. commit e011c0ebee3524df3ebfe4c485e34247e5167ffd Author: Matteo Frigo Date: Sat Jan 11 14:45:56 2003 -0500 Implemented print_plan() commit 83d6f1227a7413bf1cee8e8fda10b15569e6391f Author: Steven G. Johnson Date: Sat Jan 11 13:12:01 2003 -0500 changed the OOP-like plan_destroy to the more-grammatical destroy_plan commit 7f9077eb88fc4deb6d8c1457988ad518ab450a92 Author: Steven G. Johnson Date: Sat Jan 11 12:58:04 2003 -0500 added guru execute_dft commit eb4083006537a4a3dc5ee3d202d1bc9c07909a3b Author: Steven G. Johnson Date: Sat Jan 11 12:38:40 2003 -0500 allow for malloc errors in wisdom string, since non-fftw-malloc commit a84ffa432d7e480e83d87090ed763a8ba8deefc1 Author: Steven G. Johnson Date: Sat Jan 11 12:16:05 2003 -0500 cleanup should reset plnr to zero so that fftw can be restarted commit 5c64b4a2e584e795861b00a8c2683f1ab740c5f4 Author: Steven G. Johnson Date: Sat Jan 11 12:13:18 2003 -0500 NO_UGLY is an internal planner flag commit f9e7b4ae52caaf33854eeab2f49cc98c0ed76431 Author: Matteo Frigo Date: Sat Jan 11 11:23:13 2003 -0500 Written 1d api in terms of generic n-d api. The code is less compact but easier to test commit 96c701ecb75dbd0236023c61f59a0cde3f0f330d Author: Matteo Frigo Date: Sat Jan 11 11:07:25 2003 -0500 Added wisdom to header file, made scanners/printer static. stdio.h no longer needed in fftw.h, removed. Probably the printer_file should be reintroduced in a separate file if we ever want to print plans... commit ea3e4d45a1aec83b6e5534b2f0fbdd8271601ebd Author: Matteo Frigo Date: Sat Jan 11 09:49:30 2003 -0500 Implemented more APIs commit 41044feee9e55eb6ff29128fca2ad7458087146b Author: Matteo Frigo Date: Sat Jan 11 09:21:53 2003 -0500 Added cleanup() to API commit 125c89f921354d7d4e18aa61700b2d2ce8704e5f Author: Matteo Frigo Date: Sat Jan 11 09:17:34 2003 -0500 Started new bench.c. I had to rename plan_destroy -> plan_destroy_internal to avoid conflicts with API commit f315b29db425d56e3daffc4a5710b8e6542a0c91 Author: Steven G. Johnson Date: Sat Jan 11 02:45:39 2003 -0500 fix types commit 9419d5287867213b7dec8bbb7e594a3f77157be2 Author: Steven G. Johnson Date: Sat Jan 11 02:13:25 2003 -0500 whoops commit 8a271133e0891ed171ae642860ef03dff81e4bce Author: Steven G. Johnson Date: Sat Jan 11 02:10:50 2003 -0500 added wisdom api commit ce93efb2d481b23ccb261df25cb2021dd38b5668 Author: Steven G. Johnson Date: Sat Jan 11 01:01:17 2003 -0500 grammar commit faefac80f41e5203c91c356619f97c3c1cdf8b13 Author: Steven G. Johnson Date: Sat Jan 11 00:54:54 2003 -0500 slight change commit c01969dd0125889865e81c33fff6f5a0055f71b5 Author: Steven G. Johnson Date: Sat Jan 11 00:52:04 2003 -0500 implemented api/mapflags commit fb2e4c252410f4b6d3e26ce97ca17083ca45a773 Author: Steven G. Johnson Date: Sat Jan 11 00:48:27 2003 -0500 IMPATIENT is an api issue commit 21879988984f816cbd2ff5250d85b004a7dac217 Author: Steven G. Johnson Date: Fri Jan 10 01:57:41 2003 -0500 removed un-needed headers commit d0d8c732879727d7af40c9301c58d661c4c62343 Author: Steven G. Johnson Date: Fri Jan 10 01:56:59 2003 -0500 mkplanner initializes nthr to 1 already commit e0b0e74169c6b2ee44abc01b035e356b5ab1aaa8 Author: Steven G. Johnson Date: Thu Jan 9 18:53:09 2003 -0500 boilerplate commit f90417638448166e44f56b0f7bcc61f0263c40b7 Author: Steven G. Johnson Date: Thu Jan 9 18:16:39 2003 -0500 fold vecloop into r{e,o}dft apply function to share buffer, etcetera commit aa1101d19e86b64d4753f8bf562df5db7ea5de73 Author: Steven G. Johnson Date: Thu Jan 9 18:10:19 2003 -0500 whoops, bugfix in impulse test for vecn > 1 commit d70526c96f5f1959cf5fa3df3e15ff71a8e66487 Author: Steven G. Johnson Date: Thu Jan 9 14:23:51 2003 -0500 bugfix, grr commit 6c4923f6c3d24b14b644dea8a85adaddfb165ef5 Author: Steven G. Johnson Date: Thu Jan 9 14:21:16 2003 -0500 fixed signed-ness enum problem commit 656713c6b00d9f53d81820fd4675fe568a204088 Author: Matteo Frigo Date: Thu Jan 9 14:12:42 2003 -0500 Explicit cast commit 06f32cbb5be575b9880ff2b1a0e4031fa9be68d1 Author: Matteo Frigo Date: Thu Jan 9 13:41:51 2003 -0500 Added configure_planner(). mkplan() behaves properly when plan is null. commit 0c9627b61142790be11d642e3348808cbfa7cd5e Author: Matteo Frigo Date: Thu Jan 9 06:48:53 2003 -0500 More API work commit e21443ac067af4615dc8513d68d880f78801b983 Author: Matteo Frigo Date: Thu Jan 9 05:40:34 2003 -0500 First skeleton of API infrastructure commit 8c1212b04af0632d827194223919a73133593c54 Author: Steven G. Johnson Date: Thu Jan 9 03:19:35 2003 -0500 unsigned strikes again commit d9142b307e261d5d50a1b2086eef9012e2c36602 Author: Steven G. Johnson Date: Thu Jan 9 01:51:45 2003 -0500 put rdft2_inplace_strides and rdft2_tensor_max_index in their own files for tighter linking commit 7c048dc37ce30e18367fc3e84ec7759a2c2f0b7e Author: Steven G. Johnson Date: Thu Jan 9 01:43:13 2003 -0500 added rdft2_tensor_max_index...incorrect use of tensor_max_index was preventing proper loop ordering for rnk > 2 rdft2 commit 561ca9cb4f10d1710cea9126fc0fa63366814127 Author: Steven G. Johnson Date: Thu Jan 9 00:44:45 2003 -0500 arbitrary spltrnk in rdft2 rank-geq2 commit f17e0e00c61ac65b9353fc879a5e33f185bc8f36 Author: Steven G. Johnson Date: Thu Jan 9 00:40:17 2003 -0500 don't mention wisdom when non-verbose commit be48b68a4776f2add565a8ff0b0b0c4b8095518e Author: Steven G. Johnson Date: Thu Jan 9 00:02:35 2003 -0500 bug fix: printing %T should pass tensor *, not tensor ** commit 13e8d5776b965f625f836ffb3ed0541c5ec1c3b4 Author: Steven G. Johnson Date: Wed Jan 8 23:40:48 2003 -0500 correct(?) normalization for rodft00 ... all of the even/odd transforms should be normalized according to the expanded'' DFT of ~twice the length commit 08581922580b63f9723d7bd0da7e341d49b0225a Author: Steven G. Johnson Date: Wed Jan 8 23:18:23 2003 -0500 fixed tests for n=1 commit cd3f97118a39ef25691a86f62df1a53abfe3f15b Author: Steven G. Johnson Date: Wed Jan 8 22:10:08 2003 -0500 fixed bug in vector tests for rdft(2) commit 81b7636d1db4f7c33fa315720b91077f0c189f22 Author: Steven G. Johnson Date: Wed Jan 8 20:12:00 2003 -0500 fixed handling when first rnk-1 dimensions compress to nothing (ugh) commit a1150e27e85473748f0705407bb3858272d25bca Author: Steven G. Johnson Date: Wed Jan 8 20:02:35 2003 -0500 fixed incorrect/missing rdft2 rank-0 handling commit 77ab86cd9b8adef254ae54ee9f5f1355efb82b6f Author: Steven G. Johnson Date: Wed Jan 8 19:49:05 2003 -0500 bug fix: for rnk > 1, must compress rnk-1 dims separately (ugh) commit cba19ba921fb1d660ea71804cd40ba3d14fac750 Author: Steven G. Johnson Date: Wed Jan 8 17:39:14 2003 -0500 added trailing newline commit 925276da406dd1908a70b57c584cab6719dfb44b Author: Steven G. Johnson Date: Wed Jan 8 17:38:02 2003 -0500 updated commit 3740fe7538b1f9e0c2776a305f1c46dde0e12082 Author: Steven G. Johnson Date: Wed Jan 8 16:53:16 2003 -0500 got rid of compiler warning commit 2dfda812b6ddde932b9dd627cfbc2677ec4caeb0 Author: Steven G. Johnson Date: Wed Jan 8 16:49:48 2003 -0500 whoops, test r2hc and not rodft00 by default commit 12f2eb610a61a32de3a2d961676f005a3c7bc0c9 Author: Steven G. Johnson Date: Wed Jan 8 16:46:24 2003 -0500 got rid of real_n...use physical n everywhere in rdft; fixed rdft sz compression; fixed rodft00 verify bug commit 17233aac9a159de06ecf2dec334205094e3e0a03 Author: Matteo Frigo Date: Wed Jan 8 07:20:47 2003 -0500 icc-6.0 bug workaround commit 8490d0c5c383dcfe1910afc3e006557fb7c9aa76 Author: Matteo Frigo Date: Wed Jan 8 04:21:40 2003 -0500 Reclaimed the fftw_real identifier, because I need it for the API commit 82c0ab6a22809a05739960cb8c06c9d14d5e7968 Author: Matteo Frigo Date: Wed Jan 8 04:14:55 2003 -0500 Use recommended AC_OUTPUT syntax commit 38010c2e123c85caeb3c0827f769f304b8f77c87 Author: Matteo Frigo Date: Wed Jan 8 04:00:22 2003 -0500 Removed FFTW(foo) as a synonym for X(foo). This is an API issue. commit 863cf56f79b7eac7b70f307d24f431d71bfbdd52 Author: Steven G. Johnson Date: Tue Jan 7 17:45:52 2003 -0500 get rid of warning commit fbc87e15fead24d239286af63e298620ac46b30b Author: Matteo Frigo Date: Tue Jan 7 16:22:39 2003 -0500 Renamed conflicting files */codelet.h into dft/codelet-dft.h and rdft/codelet-rdft.h commit 683c665e1da6396f9b2c2dc8ecc749b90e666907 Author: Steven G. Johnson Date: Tue Jan 7 16:21:16 2003 -0500 updated commit 13ef7881b1b28e1772271d54f92e7f7d96059c25 Author: Matteo Frigo Date: Tue Jan 7 15:47:24 2003 -0500 Silence warnings commit f35b6c4c226aa4cba7f2a0b30b5493cc94517e13 Author: Steven G. Johnson Date: Tue Jan 7 15:00:14 2003 -0500 fftw2 used spltrnk=1 commit 9a9b9463c83f021eeefa0743fd50b9e11c008103 Author: Matteo Frigo Date: Tue Jan 7 14:32:06 2003 -0500 Silence warning commit 86d050e48df435dd7091a75e4ee9647cc31d65e0 Author: Steven G. Johnson Date: Tue Jan 7 12:13:50 2003 -0500 noted deficiency commit 97269b487afae721bc3efc07d4510284d184500e Author: Matteo Frigo Date: Tue Jan 7 07:18:51 2003 -0500 Strengthened conditions for a problem to be POSSIBLY_UNALIGNED commit d135e51da8af9610080ca861eec8a12f04e33617 Author: Matteo Frigo Date: Tue Jan 7 05:09:42 2003 -0500 Strengthened conditions for a plan to be POSSIBLY_UNALIGNED commit 41d4363cc830c074d8e602a4046fcfb361714aa3 Author: Steven G. Johnson Date: Sun Jan 5 02:43:45 2003 -0500 added copyright todo commit 81f531aeaa2fb148c7f8b5519a792c7e226060dd Author: Steven G. Johnson Date: Sun Jan 5 02:37:31 2003 -0500 modified comment commit e17581aca74e377a94b5506199a6f3c0d95dd218 Author: Steven G. Johnson Date: Sun Jan 5 02:34:36 2003 -0500 fixed comment commit f33e50cd3dedd8472c0b37116e337749dd80efa8 Author: Steven G. Johnson Date: Sun Jan 5 02:31:56 2003 -0500 implemented rdft2 verify commit 352eadf383e28c25c7132ace3c4179e561c54aa8 Author: Steven G. Johnson Date: Sat Jan 4 16:20:42 2003 -0500 fix --enable-single commit 3cd824b965de4c51a977683e83bfaa1f2d8b37ab Author: Steven G. Johnson Date: Wed Oct 23 12:59:12 2002 -0400 slight fixes commit 64f0f3180cf46058053d0a452152f3fb7e4d5363 Author: Steven G. Johnson Date: Wed Oct 23 12:42:39 2002 -0400 typo commit b6cffe0e74206ccd7ae7726181a361bea4d94986 Author: Matteo Frigo Date: Tue Oct 1 09:32:56 2002 -0400 Experimental stuff commit b92e96518b5b9ac3275a6f7194d5e1ec49b36e7d Author: Matteo Frigo Date: Sat Sep 28 13:03:53 2002 -0400 Experimental Franz mode commit fd2ac8fb21fc75eccec5c5352069388b52ab00ea Author: Matteo Frigo Date: Thu Sep 26 15:14:38 2002 -0400 const-correct commit ec5733489ef85cbe78e5253358fdb320be5b2642 Author: Matteo Frigo Date: Thu Sep 26 15:06:38 2002 -0400 Reuse dimcmp routine for other purposes commit 6fa12bfc6f9ac208da72478981473011a292f57d Author: Matteo Frigo Date: Wed Sep 25 07:37:38 2002 -0400 Use tornk1 correctly. commit 97b84fbe4c90ade6b9cad6ac2efba9b6fb305412 Author: Matteo Frigo Date: Wed Sep 25 07:36:38 2002 -0400 Hmm... I thought I had fixed this before... commit 69de6d4b5d66e405c267001886d8a7ae9e84224b Author: Matteo Frigo Date: Tue Sep 24 21:27:49 2002 -0400 Collect more common idioms commit 1bbba9625dca12e70a6e26402ba1a2262b7ca984 Author: Matteo Frigo Date: Tue Sep 24 21:15:57 2002 -0400 Still collecting common idioms... commit 01a7139392f8170c8563510d0c489bfd91687520 Author: Matteo Frigo Date: Tue Sep 24 21:13:00 2002 -0400 More garbage collection. commit 45bb1a6c49ce1569ebc75896da0ed42b0b03ee59 Author: Matteo Frigo Date: Tue Sep 24 21:08:19 2002 -0400 More compact code commit eed5c4ed8045a26be389b99e1492aedc5017f448 Author: Matteo Frigo Date: Tue Sep 24 20:54:43 2002 -0400 Collect common pattern if (foo) free(foo) ==> free0(foo) commit e7d2657d2d3bb77eb2403856e102678d865de742 Author: Matteo Frigo Date: Tue Sep 24 20:08:44 2002 -0400 Collect some common code in */buffered*.c commit 61cd95889228f7a100d853c42e461780fd01dd92 Author: Steven G. Johnson Date: Tue Sep 24 19:39:22 2002 -0400 use STRUCT_HACK #define to determing rdft kind[] allocation commit 337af322b345f45b275182f7bc8f5949794ea140 Author: Steven G. Johnson Date: Tue Sep 24 17:21:09 2002 -0400 report total pcost of measured/estimated plans...epcost is especially useful to estimate the effects of various impatience flags on planning time for large transforms commit 5cbf8b44eabe724a226d58fbeac341b7f3c13e49 Author: Matteo Frigo Date: Mon Sep 23 18:49:10 2002 -0400 Prevent unwanted inlining commit 7342f004be53b759052eaf9a01a9a574dc64631f Author: Matteo Frigo Date: Mon Sep 23 18:37:59 2002 -0400 Space compaction commit d8299eef074631210e64b01453a7602dad45d6b8 Author: Matteo Frigo Date: Mon Sep 23 11:49:32 2002 -0400 Still reducing size commit 5df9269dc8d95153c138fd44e41effd6ed1f58e2 Author: Matteo Frigo Date: Sun Sep 22 16:03:30 2002 -0400 Saved another 5KB by redesigning opcnt protocol. (gasp!) commit 074344d84ab955d0ad7efdc9b58f8414952a0372 Author: Matteo Frigo Date: Sun Sep 22 15:00:59 2002 -0400 More code compression commit 7e2e90935398c3d3d50cc2bbcab66d4b188bf757 Author: Matteo Frigo Date: Sun Sep 22 13:27:46 2002 -0400 Smaller code size. commit 1da75a085efd3348694dafb0905fb59e2c6cee27 Author: Matteo Frigo Date: Sun Sep 22 12:50:36 2002 -0400 Started unification of rader commit e0cb464fbac3602192afd97211885e814674d246 Author: Matteo Frigo Date: Sun Sep 22 12:35:30 2002 -0400 Typo commit 363f9b3b1a6bbf78e371c46a74645b55281ec0ca Author: Matteo Frigo Date: Sun Sep 22 12:25:20 2002 -0400 Changed protocol for destroy_plan so as to save space. commit daf930d4450cc9caa5d528b631f964bfbf16a208 Author: Matteo Frigo Date: Sun Sep 22 11:08:57 2002 -0400 Introduced convenient function X(mkplan_d) commit e74d86afcd19e77f275c86c916449ae2b82310be Author: Matteo Frigo Date: Sun Sep 22 10:21:36 2002 -0400 Split tensor/md5 into separate files to allow independent linking and/or prevent undesidred inlining commit 249329f66447c68d67536d4a868ac589b264a9ff Author: Matteo Frigo Date: Sun Sep 22 09:49:09 2002 -0400 Treat all tensors as dynamically allocated objects. They were dynamically allocated in part anyway, so there is no point in complicating the object code with the clumsy calling conventions for by-value structs. commit 53cf5c7cab96e0657153327e660e787279e77c4f Author: Steven G. Johnson Date: Sat Sep 21 18:24:55 2002 -0400 typo commit e36da5f9b63af8a62dab370b005e2472e5edc33f Author: Matteo Frigo Date: Sat Sep 21 18:10:07 2002 -0400 Avoid generating NaN when n = 0. commit a49b921ea278fcb353b2be6338d04daf3b0a72dc Author: Matteo Frigo Date: Sat Sep 21 18:04:05 2002 -0400 Saved more. commit 2008afba6889d6f2b9d3f00dcbcf0bc9edd8c7c2 Author: Matteo Frigo Date: Sat Sep 21 17:47:36 2002 -0400 Save 1200 bytes of object code. Do not pass structs by value whenever practical, because the calling protocol generates clumsy code. commit 8dbaef7c3531ccca29ae4f52528ed11c5089700d Author: Matteo Frigo Date: Sat Sep 21 12:10:21 2002 -0400 Do not allocate buffers for rader omegas. Let the planner do it if necessary. commit 0cd3107a7fe058042f7e23b73658bacf82d08805 Author: Matteo Frigo Date: Sat Sep 21 12:03:46 2002 -0400 Check rank *before* reading kind[0], which may be undefined if rnk < 1 commit ffab113d0748937a80de8e046d5d971a7cfde97f Author: Matteo Frigo Date: Sat Sep 21 11:48:50 2002 -0400 Second step towards rader unification. commit 054daf75a708d4b060c35b13a48ee8e8b1732cc1 Author: Matteo Frigo Date: Sat Sep 21 11:37:06 2002 -0400 First step towards unification of Rader code commit fc97f7d9567238bd1930e63614352160ff2bc202 Author: Matteo Frigo Date: Sat Sep 21 07:58:11 2002 -0400 Fix ugliness condition for cooley-tukey. commit 28fe4962b2e634dc302c3fba3853b87788b411ad Author: Matteo Frigo Date: Fri Sep 20 16:53:45 2002 -0400 Removed RADER_MIN_GOOD and associated machinery commit dc40093700e7a00e3808b606108137c7ce5cb592 Author: Matteo Frigo Date: Fri Sep 20 14:49:12 2002 -0400 Proper cast commit 2eec2b720ae866f16db023e3815f27875f572a56 Author: Matteo Frigo Date: Fri Sep 20 14:45:54 2002 -0400 Typo commit d55f46a0acf7e75a5c216964aa0016166254876f Author: Matteo Frigo Date: Fri Sep 20 14:38:13 2002 -0400 Implemented NO_LARGE_GENERIC commit 535ecb44b8e4450306cf760afb294431e5595ae6 Author: Matteo Frigo Date: Thu Sep 19 07:48:25 2002 -0400 Consistent macroization of NO_DHT_R2HC commit fe02be9d79515c92b53d929977c270b46a8b7fdd Author: Matteo Frigo Date: Wed Sep 18 21:47:17 2002 -0400 NO_DHT_R2HC is a planner flag, otherwise the EXHAUSTIVE planner loops. commit 20e70850bb3d2cd4590c9bfce7777b8a2f9a80fc Author: Matteo Frigo Date: Wed Sep 18 20:47:31 2002 -0400 Resurrected NO_EXHAUSTIVE commit 4e477d8e68603cc899c8d0104fc6897817fd74d9 Author: Steven G. Johnson Date: Wed Sep 18 19:31:57 2002 -0400 au revoir, score() commit c3f01031fa05a9088d18e643a9b3476fa6a6437d Author: Steven G. Johnson Date: Wed Sep 18 19:31:05 2002 -0400 eliminated unused commit 25e32538394211412f3aac06baa6677ae148ea03 Author: Steven G. Johnson Date: Wed Sep 18 18:28:44 2002 -0400 capitalize and parenthesize SUBSUMES commit 7115ad27ce3a4390e6c81800126315f757abbdb5 Author: Steven G. Johnson Date: Wed Sep 18 18:26:58 2002 -0400 comment commit 3ec48dd0fd8e9cc88fd85a1b7b74f9ec5ef1789d Author: Matteo Frigo Date: Wed Sep 18 18:03:18 2002 -0400 Use flags from wisdom if wisdom is applicable. commit e16b332f900b1872044fe195f7e40ae15e5ed5e6 Author: Matteo Frigo Date: Wed Sep 18 17:16:17 2002 -0400 Removed score() machinery commit bc4041b9adab2d69de986123e38bee24f480eb3a Author: Matteo Frigo Date: Wed Sep 18 14:12:21 2002 -0400 Revised planner hack commit 4f3717ebf2eca24ac5e8017eaf8856bf5270020c Author: Matteo Frigo Date: Wed Sep 18 10:14:41 2002 -0400 Fix warning commit b627b00f2f1adf8c8839b27618ac1765064c0b78 Author: Matteo Frigo Date: Tue Sep 17 17:54:07 2002 -0400 Type qualifiers. commit 48fc716d339ceb08432ab2a6704e79de578ad5fc Author: Matteo Frigo Date: Tue Sep 17 16:17:55 2002 -0400 ESTIMATE is no longer subsumed by everything else. commit 1c6447f56fe864b172d70f8940dc9de74a15a499 Author: Matteo Frigo Date: Tue Sep 17 10:55:15 2002 -0400 NO_BUFFERING is a planner flag, not a problem flag commit 458afba08480c2115e585e898153352ca125ff39 Author: Matteo Frigo Date: Tue Sep 17 09:36:16 2002 -0400 Maintain flags in canonical form. commit 1a01c050c68d0ffa9380d7bd780194509c75a31e Author: Matteo Frigo Date: Tue Sep 17 09:09:57 2002 -0400 In dramatic break with tradition, SUBSUME is now a partial order. I swear. commit 7c1f9aafa59ca2c68f98f1b2f88ca8b029506e09 Author: Matteo Frigo Date: Tue Sep 17 07:29:00 2002 -0400 Added comment commit 8b8f6515fa6ddcc9ac579c80062a9b9aa55917c1 Author: Matteo Frigo Date: Tue Sep 17 07:27:17 2002 -0400 Inverted ESTIMATE flag, renamed USE_SCORE for consistency with the convention that 0 subsumes 1. commit 1fd38e50ba038d57947daa7c999bab9da4a33836 Author: Steven G. Johnson Date: Tue Sep 17 02:50:15 2002 -0400 NO_INDIRECT -> NO_INDIRECT_OP (out-of-place only) commit ff2617c02989df82c99fd064ec298e107afe627e Author: Steven G. Johnson Date: Tue Sep 17 00:40:04 2002 -0400 hpux needs -D_REENTRANT (thanks to Clinton Roy for the bug report) commit ef127fa967046516cf5658be72eb70c93b817120 Author: Matteo Frigo Date: Mon Sep 16 23:54:34 2002 -0400 Oops. commit 70546cbffe33ede1657b54f626e133039ba26528 Author: Matteo Frigo Date: Mon Sep 16 23:44:47 2002 -0400 Yet another attempt at getting the planner right. commit ac2a09b8c81db49fcc9c770b94723577beee286c Author: Matteo Frigo Date: Mon Sep 16 21:56:14 2002 -0400 Better coding. commit a0a3d5520d53b44194f63fe2873207b57a07d544 Author: Matteo Frigo Date: Mon Sep 16 21:51:06 2002 -0400 NO_UGLY is no longer a flag, but a separate planner field that does not interfere with wisdom. commit 69253431765ca3b9cfce2c4a56c846512f6c9968 Author: Matteo Frigo Date: Mon Sep 16 19:04:41 2002 -0400 Did not compile without FFTW_DEBUG commit 43a0347e3daa0fd854eec1d7ded5c6f45ce727e4 Author: Matteo Frigo Date: Mon Sep 16 18:37:06 2002 -0400 Changed scoring mechanism. commit e4f00711d6784b6f4196859738a039ae1f7b9edd Author: Matteo Frigo Date: Mon Sep 16 17:13:45 2002 -0400 Count infeasible plans commit 18299388527442d1dfc7b7a5748da0d987c019bc Author: Matteo Frigo Date: Mon Sep 16 16:36:12 2002 -0400 curse subsumed plans before export commit fb22a4fc3b6fa45b0f46f605ed1c94eba5960359 Author: Steven G. Johnson Date: Mon Sep 16 15:40:46 2002 -0400 removed ESTIMATE_BIT vs. ESTIMATE... ESTIMATE | IMPATIENT is a UI issue commit 548808e1fefe66c9b882d332d70488986e3b073d Author: Steven G. Johnson Date: Mon Sep 16 15:31:39 2002 -0400 cleanup commit 14a42333af6152472262413b8d1a97207a1aff59 Author: Steven G. Johnson Date: Mon Sep 16 15:28:47 2002 -0400 use CONSERVE_MEMORY flag to prevent buffered for large sizes commit 688cb6fee87d6ba5ed14e9e9899ba46c96eaddbb Author: Steven G. Johnson Date: Mon Sep 16 15:16:16 2002 -0400 moved NO_DHT_R2HC back into planner flags: there's no reason we would want this flag to block plan reuse commit b06ee447ad1ee0f95af06c2d91092db1475f44a5 Author: Steven G. Johnson Date: Mon Sep 16 14:59:14 2002 -0400 whoops, commas commit 396a6523178fa8aa79f3b716e6a14577bb83c337 Author: Steven G. Johnson Date: Mon Sep 16 14:58:26 2002 -0400 problem_flags == checked in applicable, planner_flags == checked in score commit b7ef5ad344bcd298e14a30b30bd2d6f2b3c7442f Author: Steven G. Johnson Date: Mon Sep 16 14:53:16 2002 -0400 ESTIMATE should not *include* all impatience flags, even if it subsumes them; some impatience flags, like NO_INDIRECT, might make a problem unsolvable commit 81a60e6002c427a15cbb298654f954c09954c9a4 Author: Steven G. Johnson Date: Mon Sep 16 00:56:29 2002 -0400 quotatio marks commit 0833118f7818c740e7387c607c320e79e088c6be Author: Steven G. Johnson Date: Sun Sep 15 23:55:44 2002 -0400 delete blank line commit 4cbe17440ce2d074a4c0a0d3245d25c63dd469fb Author: Steven G. Johnson Date: Sun Sep 15 23:51:14 2002 -0400 substitution commit 3963051622d435d96083c0d753dcd8f503bac2f5 Author: Steven G. Johnson Date: Sun Sep 15 23:49:50 2002 -0400 note that we are not GNUlly correct commit ec9b8c84419f5dd8cd533eca7b07391696019046 Author: Steven G. Johnson Date: Sun Sep 15 23:41:01 2002 -0400 indenting commit bb5f5581a5d05566bf679da7ed67a2e59e68781e Author: Steven G. Johnson Date: Sun Sep 15 23:37:46 2002 -0400 more jokes commit 0db38cc3a3f7215cdd2e9c308fa9d88c0422024e Author: Steven G. Johnson Date: Sun Sep 15 23:20:14 2002 -0400 NONTHREADED_ICKYP includes nthr > 1 check commit a1900e4f7fdc5bc663fe60ec30d99f342ac06d34 Author: Steven G. Johnson Date: Sun Sep 15 22:56:44 2002 -0400 use md5sig commit 7f2631f48f2874827ca50e2c9ee4d59ddf861ba3 Author: Steven G. Johnson Date: Sun Sep 15 22:55:41 2002 -0400 md5sig typedef commit c83d9aa4a856e2b34011a4285df02dc43937d982 Author: Steven G. Johnson Date: Sun Sep 15 22:35:13 2002 -0400 updated commit 849fd22c4bdead7cab04a20c1b63966946b2355a Author: Steven G. Johnson Date: Sun Sep 15 22:30:26 2002 -0400 partially-ordered impatience commit f811a39af185c82590b34fd1439901b8cbf32d03 Author: Matteo Frigo Date: Sat Sep 14 19:47:56 2002 -0400 Removed all that planner inheritance crap. commit 74cf5ca97fc18b5d64c869c64575c0095f8b81c2 Author: Steven G. Johnson Date: Sat Sep 14 16:35:28 2002 -0400 string.h is used for more than strlen commit 7f974585ea055241b1339d303ffdb472305c7d75 Author: Matteo Frigo Date: Sat Sep 14 12:19:13 2002 -0400 Reduced hashtable size by 1/6 (on 32-bit machines) at the expense of messier planner. commit 8b1efa0ba0e2a490fc04c66900ad41248a55c86c Author: Matteo Frigo Date: Sat Sep 14 08:31:29 2002 -0400 Only print wisdom if verbose > 3 commit abd7a17545150645bb864c140559ba794257a897 Author: Matteo Frigo Date: Sat Sep 14 07:56:56 2002 -0400 Changed syntax of temporaries to avoid shadowing library functions (which is harmless but I hate the warning) commit a120b53fd271fad4f9b879ff6247840764061813 Author: Steven G. Johnson Date: Fri Sep 13 23:07:39 2002 -0400 only add warnings in debug/maintainer mode, and add a few more warning flags; eliminate more warnings; add support for posix_memalign (broken in glibc, grrr) commit 7832eabf884004c42c3e4089fe637e205f47732e Author: Matteo Frigo Date: Fri Sep 13 21:57:50 2002 -0400 Explicit cast commit d5127e37db1b557049933fe9aff91d9c3b0a1dc0 Author: Matteo Frigo Date: Fri Sep 13 21:54:50 2002 -0400 Use double-hashing. This allows a slightly higher load factor at the expense of a messier computation of the hashtable size. commit e689e22a6e1b981c379989a760186035fa18939c Author: Steven G. Johnson Date: Fri Sep 13 17:53:13 2002 -0400 typo commit 1d2a159b2078ef8c6063fad80d7358fd30bd4f9d Author: Matteo Frigo Date: Fri Sep 13 15:36:07 2002 -0400 Slight change in hash table growth functions. commit 9e1d9f0454bc70a807bcdb0f9ff25ed18a7c9903 Author: Matteo Frigo Date: Fri Sep 13 14:58:22 2002 -0400 More statistics. commit e1049bcf9446871a4ed34cebaae6b5b542dbca53 Author: Matteo Frigo Date: Fri Sep 13 10:13:02 2002 -0400 Clearer logic. commit 4dda68614273939d7843da329ab6c8b4da2e7bb6 Author: Matteo Frigo Date: Fri Sep 13 10:11:10 2002 -0400 Oops. commit 223f36a95fc7bd42e3b2d4bac8ad506cc2e4c8d9 Author: Matteo Frigo Date: Fri Sep 13 09:31:46 2002 -0400 Cleaned up commit c78314bd598dfb1c7e54f18fc2d050240910de4c Author: Matteo Frigo Date: Fri Sep 13 09:16:07 2002 -0400 Deal properly with infeasible problems. commit 875f159755baff40c2dd02f462c6bc36c49fbc59 Author: Matteo Frigo Date: Fri Sep 13 07:15:06 2002 -0400 Redundantly initialize hash table to prevent valgrind warnings. commit b7047a11d0b3e41e53e1890f73135be6f2ce3b2c Author: Matteo Frigo Date: Thu Sep 12 19:00:22 2002 -0400 Removed relics from past. commit 4cae827eaf456e69f3155183afd52e4c0216c980 Author: Matteo Frigo Date: Thu Sep 12 18:53:44 2002 -0400 md5hash a problem only once. commit a1ef1699bff263e0141cd43801c7a4ff3431389e Author: Matteo Frigo Date: Thu Sep 12 16:33:49 2002 -0400 Renamed k7 codelets commit c4367d998eeed60d4618b8a5b54162d500b84271 Author: Steven G. Johnson Date: Thu Sep 12 16:32:03 2002 -0400 FORBID_DHT_R2HC -> DHT_R2HC_VERBOTEN for consistency commit 39a9858e45ca542695b9419c09ec6b61b09a6004 Author: Steven G. Johnson Date: Thu Sep 12 16:28:43 2002 -0400 removed obsolete macro commit bd1d1de9ba94e111921e911a49de82fe4ff2d16e Author: Matteo Frigo Date: Thu Sep 12 16:20:39 2002 -0400 Split flags in SIMD code. commit b9fbfffc3ee26e56c4c16448ced8db523670de55 Author: Matteo Frigo Date: Thu Sep 12 16:18:51 2002 -0400 Forgot to fix threads commit 1d3447ab63c27d4bd97beb41882ca34addd0df18 Author: Matteo Frigo Date: Thu Sep 12 16:10:05 2002 -0400 Split flags into planner_flags and problem_flags commit 075ff4047c6c5a98bd268a3bae692df6a9ec1d84 Author: Steven G. Johnson Date: Thu Sep 12 15:46:56 2002 -0400 tetrameter commit ff7f0235dd4460a5e3c332725151037107514954 Author: Matteo Frigo Date: Thu Sep 12 15:11:21 2002 -0400 Overwrite less impatient solutions properly. commit b470f419700398ec14357990abeb69aa6eb4d857 Author: Matteo Frigo Date: Thu Sep 12 11:29:16 2002 -0400 Oops. commit 1ae2a65b2895b51b43f316fa11fcc932ed127ae2 Author: Matteo Frigo Date: Thu Sep 12 10:58:56 2002 -0400 Keep less impatient solution in case of conflict. Paranoid cast to uint in certain places. commit 640b1f4df72028daa4293c304e53af0da7f31c36 Author: Matteo Frigo Date: Thu Sep 12 10:02:51 2002 -0400 Complete reimplementation of planner hash table. commit 40f47f4111154bd1e17e44f87908228ede39af18 Author: Matteo Frigo Date: Thu Sep 12 07:58:45 2002 -0400 planner->cnt was not properly decremented. commit fda67f21284b158043d7ba171a81a933a3891e1a Author: Steven G. Johnson Date: Wed Sep 11 17:52:39 2002 -0400 typo commit ebe84b30659823364a95bfd646512b387bef4629 Author: Matteo Frigo Date: Mon Sep 9 17:10:45 2002 -0400 Simplified commit 230458a658da6fa62fac4ccd66918f38442df00b Author: Matteo Frigo Date: Mon Sep 9 17:03:32 2002 -0400 Always overwrite old wisdom with new, in case the old is corrupt/conclicting. commit 2d91c8d00acc06eb228d7ba5492e8f6ec7ee24d9 Author: Steven G. Johnson Date: Mon Sep 9 16:56:03 2002 -0400 added quote/joke commit 0173e3dc140cec9b061b4cf0cc0a626e41105e1f Author: Matteo Frigo Date: Mon Sep 9 15:04:47 2002 -0400 Completed wisdom import commit dee4de2b0b7a9c60d9501d67e5ab17c5e828e474 Author: Matteo Frigo Date: Mon Sep 9 10:14:22 2002 -0400 Slight cleanup of md5 interface. commit b990a36ddb7c8ec04d248fcbbbc4a135827cf7b2 Author: Matteo Frigo Date: Tue Sep 3 22:32:43 2002 -0400 More consistent protocol between planner and inferior. commit fbf287fea9e51a6eb2a62030a115aea58ef2f630 Author: Matteo Frigo Date: Tue Sep 3 21:08:30 2002 -0400 I can't think of any situation where saving infeasible problems would be desirable. Removed relevant code. commit c194f7f7a34d6909408bcd55e543f4cbf7a60a2f Author: Matteo Frigo Date: Tue Sep 3 20:57:03 2002 -0400 Encoder registrar's names in wisdom. Remove export_conf, since a separate program can now generate it. commit e9a30d633c905ac3eba878af7839ad73f153dd06 Author: Matteo Frigo Date: Tue Sep 3 15:11:06 2002 -0400 Fixed typo commit 51b8ddee6eda85459d68909df089b7e251fb65ca Author: Matteo Frigo Date: Tue Sep 3 14:52:45 2002 -0400 Fixed broken trochaic meter. commit eb531c7d3242141b4603cca8f270d88bab0f48e3 Author: Matteo Frigo Date: Tue Sep 3 09:49:50 2002 -0400 Initialize planner->score. It is correct to leave it uninitialized, but I don't want people to send reports about purify complaining. commit 6a000fc379ab96b4ea9310e76f5391af7d6131a8 Author: Matteo Frigo Date: Tue Sep 3 09:03:46 2002 -0400 More latin silliness commit 3b9fecd5d0365958954cc149251fed5b9ce07ddc Author: Steven G. Johnson Date: Mon Sep 2 17:57:32 2002 -0400 updated commit 28a40bce8ba8e91b240d4f6e7ddcf55b68f05e77 Author: Steven G. Johnson Date: Mon Sep 2 17:33:49 2002 -0400 added clock() getseconds timer commit c004f7f51d23ac8f1c6220ff4a18e83d2e4fe7cf Author: Matteo Frigo Date: Mon Sep 2 16:16:58 2002 -0400 Oops commit 3f227ec57b346fe8688fbf2e08dfcf6cc3c3c955 Author: Matteo Frigo Date: Mon Sep 2 15:58:19 2002 -0400 Experimental INDIRECT_VERBOTEN flag (not used) commit 802f348a8a8c0bd97e15ed827e8092ab358abbe2 Author: Matteo Frigo Date: Mon Sep 2 15:36:21 2002 -0400 Do not allow buffering in children of indirect solvers. commit f081fc5e202141350664e3d9adbe947d5331ab9f Author: Matteo Frigo Date: Mon Sep 2 15:02:11 2002 -0400 Oops commit 08826857c28146b6366770565a9971eb4bdd9505 Author: Matteo Frigo Date: Mon Sep 2 14:32:28 2002 -0400 Hash sizeof(R) as part of wisdom. commit ff803ad22f9a5a980be0ab10aebd18ab210557de Author: Steven G. Johnson Date: Mon Sep 2 13:47:57 2002 -0400 added --enable-float synonym for --enable-single (since with have --enable-long-double) commit 99672a129c0032eeb3c99424d16f16547e1fff5d Author: Matteo Frigo Date: Mon Sep 2 13:46:08 2002 -0400 zerotens is now in its own file, so it does not cause dft to be linked in if only rdft is used. commit fe35f517845e4692b46077037dc40b155c9fa500 Author: Matteo Frigo Date: Mon Sep 2 11:56:37 2002 -0400 Removed unused var. commit 2a4e8a9a1121f4adb852256865ae2b52743d6f40 Author: Matteo Frigo Date: Mon Sep 2 11:55:33 2002 -0400 Split insert() in preparation for wisdom import commit fc1cf1dfa9a72fe8968426c4e694b7336926a03d Author: Matteo Frigo Date: Mon Sep 2 11:46:57 2002 -0400 Moved debugging infrastructure to test directory so that it is not linked into the shared library. commit 07a825b37e850981bc9fa18460538346cc7dd137 Author: Matteo Frigo Date: Mon Sep 2 11:04:54 2002 -0400 Reactivated wisdom export commit dfcc8fd2b155015d11c95762e4384ae139f922c6 Author: Matteo Frigo Date: Sun Sep 1 21:30:58 2002 -0400 Dump errors to stderr, not stdout. commit 44e1a88d3527239c405ca268888b3695c902c807 Author: Matteo Frigo Date: Sun Sep 1 21:26:38 2002 -0400 Removed traverse.c. traverse.c is no longer need for plan blessing. I figured out a way to avoid using it in planner-score.c, so the file is now redundant. commit 1c9ef6ccf7c373d274215c83bbede0fea30682c7 Author: Matteo Frigo Date: Sun Sep 1 19:51:50 2002 -0400 Removed code made obsolete by new MD5 scheme: problem equality tests, scanners, and associated list of problem kinds. commit 726b571dc2949d3369fc09237b41f99ef8ce058b Author: Matteo Frigo Date: Sun Sep 1 19:22:54 2002 -0400 Started md5 implementation commit d89348364f45c8f635356d8ae1047f6cb8508158 Author: Matteo Frigo Date: Sat Aug 31 14:00:04 2002 -0400 Keep track of hit rate commit 971b014a7698a2b54a6dec16b486b7953a1f41e0 Author: Matteo Frigo Date: Sat Aug 31 12:44:04 2002 -0400 Only dump when verbose > 4 commit 018df5704e37547209aeaecadc50811defcf5f2b Author: Matteo Frigo Date: Sat Aug 31 09:55:57 2002 -0400 Debugging infrastructure commit 837cecb7a386caabb14f3b41518083f18b364fc7 Author: Matteo Frigo Date: Sat Aug 31 09:21:48 2002 -0400 Use debug infrastructure to dump planner. commit dfaf407162a50f7a77ef3496dbaf754b27551029 Author: Matteo Frigo Date: Fri Aug 30 21:29:10 2002 -0400 Do not store plans in planner, plus general planner cleanup. commit e74dd299a89b62e7d0e51be3293163325b5044ea Author: Steven G. Johnson Date: Fri Aug 30 18:07:52 2002 -0400 renamed IN_DHT_R2HC to the more general FORBID_DHT_R2HC commit 904ff75e31a716b0c9d97f7e4efd425bd4309c06 Author: Steven G. Johnson Date: Fri Aug 30 18:07:21 2002 -0400 eliminated unused var commit 60f3382238de7ce933a192a923f06657900e4853 Author: Matteo Frigo Date: Fri Aug 30 12:09:48 2002 -0400 Score planner was not working correctly when using wisdom. Fixed. commit 0522d2fcf8b00ec9f1f3cdbd38cfa03d764e1154 Author: Matteo Frigo Date: Fri Aug 30 08:20:48 2002 -0400 Use hash table in debug malloc commit e451f6d74620eefd71304a543d0eb48ab9ea953d Author: Steven G. Johnson Date: Fri Aug 30 02:45:15 2002 -0400 listed some good stuff commit 4d5aeb7a8917367942f1e798bd19d5c3be2feaf1 Author: Steven G. Johnson Date: Fri Aug 30 02:17:30 2002 -0400 timed planner commit 3bb76589f0d1efc68b0a82eeecbdf30748a991fd Author: Steven G. Johnson Date: Fri Aug 30 02:07:00 2002 -0400 fma? commit e231f879031ab34c3a8fa8a8da0fbf642cd88a0b Author: Steven G. Johnson Date: Fri Aug 30 02:05:55 2002 -0400 update commit 1f29de7940d8c29acc6b608deec341cd982cc706 Author: Steven G. Johnson Date: Fri Aug 30 01:31:47 2002 -0400 rader-dht -> dht-rader commit 888439dcea7af2d64300c776238afa7c5c18a372 Author: Steven G. Johnson Date: Fri Aug 30 01:21:37 2002 -0400 add DHT solver, and break up rader-dht and r2hc-hc2r commit df668dba33c6181cb6feb58dba6f649c89d73fe6 Author: Steven G. Johnson Date: Thu Aug 29 23:20:35 2002 -0400 another option commit 213d66b5bcf73e5525c97bb6dc9461808600d483 Author: Steven G. Johnson Date: Thu Aug 29 22:55:29 2002 -0400 generalized indirect solvers for fftw2-like buffering and more commit 795353001d9c3db1d30d5def55b8671cb4eb67b0 Author: Steven G. Johnson Date: Thu Aug 29 18:08:16 2002 -0400 tensor_max_index and tensor_min_stride are now both unsigned commit d59c4e92144b248504c9c01cae0ea5d3f4e0aa60 Author: Steven G. Johnson Date: Thu Aug 29 17:58:35 2002 -0400 added iabs.c, and tensor_min_stride returns min absolute value commit 9247f8665076f514844f3fd4a0478f9988313251 Author: Steven G. Johnson Date: Thu Aug 29 17:31:39 2002 -0400 bug fix in cldrest hc2c/c2hc copy loops commit c8d575230e1d18331b0a1d8fe22d6c8fcfd70ce4 Author: Matteo Frigo Date: Thu Aug 29 13:45:08 2002 -0400 Added things to do. commit c8b62313d7952baa412a1c18427473c010451303 Author: Steven G. Johnson Date: Thu Aug 29 13:10:04 2002 -0400 added automake prereq commit 3a6be2745802ae618bbb0521fe5e54e17e1eecb2 Author: Matteo Frigo Date: Thu Aug 29 08:36:36 2002 -0400 Use indexed addressing commit b27567a442018af7ec7cf782f117bfc9f5476e08 Author: Matteo Frigo Date: Thu Aug 29 08:20:55 2002 -0400 Ooops commit dc19f8daf30d11be571d8856cb10e8d8a9046b6c Author: Matteo Frigo Date: Thu Aug 29 07:45:37 2002 -0400 Oops commit ee4bb3eb29224fe9ef1e5c3c8416256b04a22599 Author: Steven G. Johnson Date: Thu Aug 29 02:32:13 2002 -0400 updates to win32 threads code (ick) commit 6cfa2e60b0f2831342f21a72eb3732bf80366250 Author: Steven G. Johnson Date: Thu Aug 29 01:44:33 2002 -0400 added threaded version commit 8cc323cbe02b7c270bb664e2c8acedce4ddf48b5 Author: Steven G. Johnson Date: Wed Aug 28 19:47:21 2002 -0400 fix make dist commit 43fa7922084aae1bdb1b69aff7109aed56e1f025 Author: Steven G. Johnson Date: Wed Aug 28 15:09:03 2002 -0400 whoops, bugfix for inverse commit b3136883e3f02c5bbb68338e01d134aa2b2a25eb Author: Matteo Frigo Date: Wed Aug 28 14:50:34 2002 -0400 Use C9x convention for naming (fftwf etc.). Removed installable header files since they will be part of the API. commit ed3b5e17a932fe32d1a9397642c6e043eb3dbc40 Author: Steven G. Johnson Date: Tue Aug 27 23:34:00 2002 -0400 allow _1 variants to accept rnk 0 (sz 1) problems commit 004227a749ea2b1990047e29b15826437335239f Author: Steven G. Johnson Date: Tue Aug 27 15:56:09 2002 -0400 updated commit ca46171fc03dc63c4a93462abded63a039acdd82 Author: Matteo Frigo Date: Mon Aug 26 20:14:56 2002 -0400 Loop unroll is useless commit 6528250479b8700bc9082f776d5f3c340080175a Author: Matteo Frigo Date: Mon Aug 26 20:00:41 2002 -0400 Use indexed addressing commit 8f4d60a4abc3e10d5e4e2739119422babe433a1d Author: Matteo Frigo Date: Mon Aug 26 19:46:46 2002 -0400 Use indexed addressing in transpose routines. (Seems to be slightly better on athlon.) commit aac3c6a8800ddbc174774e9eeeb32f054c8af6a6 Author: Steven G. Johnson Date: Mon Aug 26 12:59:44 2002 -0400 added comment about stability commit bdaced931410f8e984ac5c3a833e842d6ffa8965 Author: Matteo Frigo Date: Mon Aug 26 07:43:53 2002 -0400 Approximate opcount commit 46c5151b696b0d6f0ff98f952d8a13283d95877b Author: Matteo Frigo Date: Mon Aug 26 06:38:49 2002 -0400 Finished rdft2 via dft/rdft commit c9122c8dbfcac3ae13893442c0ad348e410b646b Author: Steven G. Johnson Date: Mon Aug 26 00:15:59 2002 -0400 some updates commit b049bb9502bd3c00a3a8f1ff3cfd3c5596e9e7c0 Author: Steven G. Johnson Date: Mon Aug 26 00:05:53 2002 -0400 rdft kind is now per-dimension, added rdft/rank-geq2 commit 32db021f7eabd57af68c88e1e6266589a828df35 Author: Steven G. Johnson Date: Sun Aug 25 22:45:38 2002 -0400 added note commit e174f61162d8e897e181f1ae8f01b5c8ba4122d8 Author: Steven G. Johnson Date: Sun Aug 25 22:28:12 2002 -0400 must zero real sz commit 7881bf396852115443bd3ce1dbdce177d8d64b6e Author: Steven G. Johnson Date: Sun Aug 25 22:06:52 2002 -0400 unified pickdim funcs commit 9b588fdbd50ab5d47cb936aae0569b6f9b54fc4c Author: fftw Date: Sun Aug 25 14:10:55 2002 -0400 silence warnings commit f58dff38cc5c75f8c508e971a1ffd286be572f87 Author: Matteo Frigo Date: Sun Aug 25 13:16:49 2002 -0400 I had to add another planner flag to record whether pointers could become unaligned because of vrank-geq1 solvers (these solvers only plan the first element of a vector problem, but the second element may have a different alignment). This addition is ugly, but I don't see any way around it. commit 3633f42453ac103289d0c471630892680f1b0625 Author: Matteo Frigo Date: Sun Aug 25 10:18:25 2002 -0400 Added thoughts commit fb9c1acef1a3499a8629190b172a1ec0430260b4 Author: Matteo Frigo Date: Sun Aug 25 10:08:59 2002 -0400 Implemented rdft2 via vector rdft + radix2 step commit 8bc1aed075f15afc6de9d82adc44d6ab8b5e50f6 Author: Matteo Frigo Date: Sat Aug 24 17:43:54 2002 -0400 Stylistic changes commit 188add2600049e4313ba1e77c1976b887544ae90 Author: Matteo Frigo Date: Sat Aug 24 11:19:30 2002 -0400 Simplified mktwiddle interface commit 1c91434a8fea606141e28014376c2d2c9937f1b7 Author: Matteo Frigo Date: Sat Aug 24 11:05:08 2002 -0400 Unification of certain vector computations. rdft2-dft is now a vector transform. commit f9311503a90a428a78350116e1bf47c6ffefddcc Author: Matteo Frigo Date: Fri Aug 23 20:21:25 2002 -0400 Intel compiler seems to be still buggy commit de1bb9192bc2740a6fbe70bbac497a1ac34450f0 Author: Matteo Frigo Date: Fri Aug 23 16:07:12 2002 -0400 Streamlined twiddle protocol commit e2a28ed96b7b40db5f3fe6b72852acf550cdaca4 Author: Matteo Frigo Date: Fri Aug 23 13:22:17 2002 -0400 Implemented rdft2 via dft (forward only for now) commit 17d57ef85db79c55dcd5c77260618e798833e1d0 Author: Matteo Frigo Date: Thu Aug 22 11:29:29 2002 -0400 More cleanup of verify commit 0eb03788a7c1c13953638a19182235738511b77a Author: Matteo Frigo Date: Thu Aug 22 11:16:03 2002 -0400 Changed error criterion because old one was too strict commit e97f092fbcd955e08b36522352e0e9b94cfd473e Author: Matteo Frigo Date: Thu Aug 22 11:15:17 2002 -0400 Disable shared commit f611df2e683eea128367d652cf12e2c4a2b3dfb1 Author: Matteo Frigo Date: Thu Aug 22 09:19:12 2002 -0400 Added thoughts commit 6ad63b2082094c83de1fe5a01400423c249564a1 Author: Matteo Frigo Date: Thu Aug 22 09:17:28 2002 -0400 Oops commit 42381c1bcded2b7d1854300ff5de0addfca36575 Author: Matteo Frigo Date: Thu Aug 22 09:11:34 2002 -0400 Do not use inline. Minor changes. commit b9b2448db23e3d5d0d6d10b4bf3fe1858a847a05 Author: Steven G. Johnson Date: Wed Aug 21 16:23:26 2002 -0400 more commented flags commit 1c316981c3ed81a4550d6ba9bfa733745532bcf2 Author: Steven G. Johnson Date: Tue Aug 20 19:44:43 2002 -0400 added DCT-IV and DST-IV commit c15e995b01901e957e42c369b5341b6dbeaf3ac2 Author: Matteo Frigo Date: Tue Aug 20 16:01:36 2002 -0400 Slight improvement in twiddle scheme commit 8ba1ef4db7a3866fae35bd1825a5a5c35ae5673f Author: Steven G. Johnson Date: Tue Aug 20 15:31:54 2002 -0400 name fix commit 5b56bb057b02f41413a9d5436a3d9d1c50a0e7f5 Author: Steven G. Johnson Date: Tue Aug 20 15:16:48 2002 -0400 removed extraneous variable commit d8b1080be08abcfa55a88f38d7bc677d99a9c2d9 Author: Matteo Frigo Date: Tue Aug 20 11:46:29 2002 -0400 Oops commit b500a0d285f5b95a9b88952b8830aa8423be9332 Author: Matteo Frigo Date: Tue Aug 20 08:37:45 2002 -0400 Still playing around commit 7a44ac35618394bd3715c928e9dc0b3a7a149f0e Author: Matteo Frigo Date: Mon Aug 19 19:56:29 2002 -0400 Playing around with addition chain commit ac8dfff733ce38c8b013523ff4e9fc9888456989 Author: Steven G. Johnson Date: Mon Aug 19 19:48:56 2002 -0400 comments commit 175b3b2cec3441b1a0e34d03343e9f581f0e030e Author: Steven G. Johnson Date: Mon Aug 19 19:45:35 2002 -0400 comment fixes commit 0b6386c1ef38de51383f1306c82e7966b0db0d52 Author: Steven G. Johnson Date: Mon Aug 19 19:40:18 2002 -0400 added reodft stuff commit 8d4aef3c5738367c010a1bfd3004c94f73281950 Author: Matteo Frigo Date: Sun Aug 18 19:44:14 2002 -0400 Sync with nbenchfft commit 6ec9197550f61b20dad2a8e238bd00da3bf2cf23 Author: Matteo Frigo Date: Sun Aug 18 16:02:37 2002 -0400 Economy of thought commit 70610d2a45fcebc6b9c3c61e5dd6caa96d292b4f Author: Steven G. Johnson Date: Sat Aug 17 15:52:05 2002 -0400 distribute addchain.c commit 2fea59351e2abaeec9bb1ea094d06097282bdf7e Author: Matteo Frigo Date: Sat Aug 17 14:09:11 2002 -0400 Nothing serious commit c5ef4a2ddba0963a8c9a388edf050c5ee2fbbb00 Author: Matteo Frigo Date: Sat Aug 17 10:47:59 2002 -0400 New twiddle policy (disabled for now) commit bf62c3f3bb4be6257869db7d46f69b694c7a2688 Author: Steven G. Johnson Date: Fri Aug 16 23:44:28 2002 -0400 bug fix for hc2r (must use inverse dft) commit e7434c44d6c3f77e761da5e7a8e850f48c6fb872 Author: Matteo Frigo Date: Fri Aug 16 20:27:10 2002 -0400 New log3 twiddle policy commit d0f1857c45c12d35cbd9fded016c3b7ceac70aa7 Author: Matteo Frigo Date: Fri Aug 16 18:10:33 2002 -0400 More verify cleanup commit c8f750da8aab093581b6be29ff8d781906ca771b Author: Matteo Frigo Date: Fri Aug 16 16:31:19 2002 -0400 Oops commit eae86b4ff97b783a93ebd6f7b0a6352cea48359c Author: Matteo Frigo Date: Fri Aug 16 15:22:36 2002 -0400 Economy of thought (and code) commit d7bdf2e9b5a011b19bb16a9d12d5d763a3196c2a Author: Matteo Frigo Date: Fri Aug 16 14:05:45 2002 -0400 Added comment commit 4bec01a2c6ea089d18a81b8d7d3ce649cbe80fe8 Author: Matteo Frigo Date: Fri Aug 16 12:57:43 2002 -0400 Cleaner rounding algorithm commit d91cc0e5f23a61e226b2a575c23f35c79ec3ca06 Author: Matteo Frigo Date: Fri Aug 16 11:27:43 2002 -0400 Can get away with shorter length in bluestein (I think). commit 56113aa7d008511b8387a1d1652e03d9fd8844e7 Author: Matteo Frigo Date: Fri Aug 16 11:08:09 2002 -0400 Portability improvements commit b58468b7ecd1e0ff7b9a2b1236d64e357627d8a0 Author: Matteo Frigo Date: Fri Aug 16 08:06:31 2002 -0400 Optionally average accuracy test over many rounds commit 21b850aeaafa046e663e6bc5a42a9538c9571180 Author: Matteo Frigo Date: Fri Aug 16 07:50:24 2002 -0400 More accurate formula for trig tables commit ce0241125c235817e2132e938e8c9dcd3166773f Author: Matteo Frigo Date: Fri Aug 16 06:42:02 2002 -0400 Implemented accuracy test for all integers commit db374e203e4d37c399e6b3d877da8cdf192ec649 Author: Matteo Frigo Date: Thu Aug 15 18:54:44 2002 -0400 inv, neg: make static commit 79f1c53641c0cec5612621c1f72726a81d56144e Author: Matteo Frigo Date: Thu Aug 15 17:25:37 2002 -0400 Verify was not complete for real transforms commit c60e8fcdedd600b93f30f098ca49f794375e8377 Author: Matteo Frigo Date: Thu Aug 15 16:30:03 2002 -0400 Oops commit 688a0ef88f8ef6a8d060ad2b04ce51b4d94870b7 Author: Matteo Frigo Date: Thu Aug 15 16:29:16 2002 -0400 Fixed hb codelets commit 8a4d71183c0b5cd62b6f9f53f41cfd68a8b602fc Author: Matteo Frigo Date: Thu Aug 15 14:10:45 2002 -0400 Changed twiddle policy commit 9905db7579db957d5ebc9f472847910d24b65e10 Author: Steven G. Johnson Date: Thu Aug 15 13:32:24 2002 -0400 whoops commit 59c9d170edb4001d3f37b64cfb2d8d48e9cb9b7d Author: Matteo Frigo Date: Thu Aug 15 11:01:04 2002 -0400 No point in libbench being a shared library commit b75824c63085764eb1fbf97b83961eb0411bd969 Author: Matteo Frigo Date: Thu Aug 15 09:48:37 2002 -0400 Moved accuracy test to libbench commit ebac0dde6d77f268c45cfc7ba17230c547e60800 Author: Matteo Frigo Date: Wed Aug 14 19:48:23 2002 -0400 Modified accuracy test commit d5e2c4a63b8f9b195e7812f817cefc61617accf9 Author: Matteo Frigo Date: Wed Aug 14 08:34:26 2002 -0400 Fixes for long double commit 41c23eb9e18add9786e959871bcba7d1ffc44bb6 Author: Matteo Frigo Date: Wed Aug 14 08:17:57 2002 -0400 Normalize input commit d83e36740eaf0cd2d8ffecb25d38ffdcd6412bcb Author: Matteo Frigo Date: Wed Aug 14 07:26:41 2002 -0400 Oops commit 0d312034a4b8a8ece11903c0b81aa4ce57151783 Author: Matteo Frigo Date: Wed Aug 14 07:25:34 2002 -0400 Also compute relative error commit 10c281df8a8195c0eb497cd3b73955807c64f06e Author: Matteo Frigo Date: Wed Aug 14 07:08:20 2002 -0400 Loop over N commit 588a70753f53ef9fe7801fd3c0cd1b1f2c5da7fc Author: Matteo Frigo Date: Wed Aug 14 06:54:50 2002 -0400 simple-minded accuracy test commit b25380fae97bb3af2b38f592f5393b10be1ff430 Author: Steven G. Johnson Date: Wed Aug 14 03:26:06 2002 -0400 whoops commit d32e62c62857ed17fdf4a9fa7ebb12007c8e32bc Author: Matteo Frigo Date: Tue Aug 13 11:42:41 2002 -0400 fma() stuff is too nonportable, removed commit 583c58e086a592a198619d6c36fcb6137b4ea068 Author: Steven G. Johnson Date: Mon Aug 12 14:07:44 2002 -0400 slight fix commit 2be67d85e7e8c8b4db5644bdeca6cfa1a0211959 Author: Steven G. Johnson Date: Mon Aug 12 14:07:18 2002 -0400 use table for rdft_kind_str commit 331ca343e551313e04bc1c88ae8c8cd3e1dfb4fd Author: Steven G. Johnson Date: Mon Aug 12 13:43:08 2002 -0400 slight fixes commit 8bf7bf1145ef67937cf020d64f0e9913aef84d58 Author: Steven G. Johnson Date: Mon Aug 12 13:31:37 2002 -0400 multidimensional rdft2 commit 4457a7cf6bf60cff0e842bfd127e22e7b3de55e5 Author: Steven G. Johnson Date: Sat Aug 10 19:33:23 2002 -0400 use tensor_copy_inplace commit 5e370a1a072a67b940639f311e296a97150acf1c Author: Steven G. Johnson Date: Sat Aug 10 19:32:03 2002 -0400 bugfix, use tensor_copy_inplace commit 92f280c99e002d9ee78e42967ee81bb4fcf84d2b Author: Steven G. Johnson Date: Sat Aug 10 19:30:39 2002 -0400 use tensor_copy_inplace commit 55ee1b50c140e81f41abc05975f01393c8bd4cbd Author: Steven G. Johnson Date: Sat Aug 10 19:28:07 2002 -0400 added tensor_copy_inplace commit ce8083b65d5ae7952d40c253896ae0e6759e73e8 Author: Steven G. Johnson Date: Sat Aug 10 19:25:50 2002 -0400 fixed trig-function table type commit 9b354635204711389328f487a058a54604d58e0a Author: Matteo Frigo Date: Sat Aug 10 14:41:04 2002 -0400 Improved trig scheme commit 466d2a03411d082ab673c73582a08842f12f6846 Author: Matteo Frigo Date: Fri Aug 9 21:05:01 2002 -0400 Allow for testing using long double instead of pari commit 14b243d1d509236a5b19e8783570989cdfda6333 Author: Matteo Frigo Date: Fri Aug 9 20:49:32 2002 -0400 Yet another trig scheme. commit 361e112752a93e14cab74d86d92fccb88686fed1 Author: Matteo Frigo Date: Fri Aug 9 20:38:07 2002 -0400 Yet another scheme commit b3ca7c941515736b0ebd97c7d1195cd736d2b8d8 Author: Matteo Frigo Date: Fri Aug 9 20:31:16 2002 -0400 Careful with overflow commit c1af0a91c6bbcd3482427d1be4a812a0c061d879 Author: Matteo Frigo Date: Fri Aug 9 20:16:23 2002 -0400 Avoid overflow commit f06cb59c469661f10f65f220b91d79e8d98097f7 Author: Matteo Frigo Date: Fri Aug 9 19:26:57 2002 -0400 New(er) trig routines commit ba6e2f6487663745c402856288f95441c6191fe8 Author: Matteo Frigo Date: Fri Aug 9 19:25:44 2002 -0400 Oops commit 267f53395f5e34f83a0664c9405e4d7b17583695 Author: Matteo Frigo Date: Fri Aug 9 18:49:04 2002 -0400 New file commit cc25b36b4369f7fd773b614e416185078bc3e20c Author: Matteo Frigo Date: Fri Aug 9 13:04:00 2002 -0400 Commented about likely gcc bug commit 745572695256ffc140f9b3bd828b561f56bea1a5 Author: Matteo Frigo Date: Fri Aug 9 13:01:49 2002 -0400 Improved accuracy of twiddle factors commit b90ec91c045668caabc583c27da9400331fc34cc Author: Matteo Frigo Date: Thu Aug 8 06:36:23 2002 -0400 Wrong comment commit 01653dbd957c931c5e562c6cdf727c26a4570680 Author: Matteo Frigo Date: Wed Aug 7 17:14:09 2002 -0400 Experimental 3dnow port using gcc, to compare it with Stefan's stuff. commit 9716316af3a8a84ac9888e8b184fad1f8b34279d Author: Matteo Frigo Date: Wed Aug 7 12:58:10 2002 -0400 End of AREF experiment commit 03365b937b905ad6dd6dad3ec0044f010f2cec51 Author: Matteo Frigo Date: Wed Aug 7 07:47:19 2002 -0400 Oops commit 00d1519ee07579c41da9738b4bd0d9e130c252df Author: Matteo Frigo Date: Wed Aug 7 07:46:38 2002 -0400 Pathetic attempt to reduce size of configure script commit 882c809b6257b73377a20a807a20a61f5cc5a655 Author: Matteo Frigo Date: Tue Aug 6 20:38:11 2002 -0400 Changed array syntax for experiments. commit 06bf9f0b7d08eb0a66a07b4b517fede0514a4a2c Author: Matteo Frigo Date: Tue Aug 6 19:58:20 2002 -0400 Fix warning commit dd2b973d27111516233a46e5d44734f2d1cea503 Author: Matteo Frigo Date: Tue Aug 6 13:35:28 2002 -0400 Move nonportable stuff in one place. commit 3a3a36d48074544b746b464bd194f93a371615b9 Author: Matteo Frigo Date: Tue Aug 6 10:32:53 2002 -0400 Economy of thought: I didn't like having two algorithms for removing solutions, both correct. At least now we have the same algorithm copied twice. commit e0cf8fd96853061b2160a99ed871b621a69bacbe Author: Matteo Frigo Date: Tue Aug 6 09:12:21 2002 -0400 Added things to do commit f96ded332986cff7099c0dd6bf2cff07d3e59217 Author: Steven G. Johnson Date: Mon Aug 5 19:54:31 2002 -0400 improved interaction of planner with patience flags commit f37ad7a0a0a7009a6c29c02ff53b06440f12e846 Author: Steven G. Johnson Date: Mon Aug 5 14:17:58 2002 -0400 set up for real-even/odd DFTs, where n is not the size of the data commit 1a2ea854fa6156b907c817752dc47a1c07ef5c2e Author: Steven G. Johnson Date: Sun Aug 4 23:57:51 2002 -0400 DESTROY_INPUT flag commit 18483232ce3afae0412e565222de6c48891700d7 Author: Steven G. Johnson Date: Sun Aug 4 22:50:19 2002 -0400 CLASSIC -> IMPATIENT commit 0fee1c8d39ed87aaab3387028cc3ff4422261a41 Author: Matteo Frigo Date: Sun Aug 4 19:05:43 2002 -0400 Require make maintainer-clean to remove the generator, as opposed to make clean. In this way we can type make clean without regenerating all codelets. commit b633708685610bf42bb69bbe71f31f0fd849aff5 Author: Steven G. Johnson Date: Sun Aug 4 17:34:04 2002 -0400 ESTIMATE plans are not blessed commit 17f106f814fd30121f7fcc2de65cc78f77a6448d Author: Steven G. Johnson Date: Sun Aug 4 17:24:37 2002 -0400 use flags in wisdom commit 342928973eaf98429367ce537b088761c391505c Author: Steven G. Johnson Date: Sun Aug 4 17:03:45 2002 -0400 score now takes plnr, not flags, as arg commit 5ef96008dcfb0e7428716122ea8ea56d0637898a Author: Steven G. Johnson Date: Sun Aug 4 16:37:46 2002 -0400 align initial stack in alignment check, which should now pass for gcc 3.1.1 commit ce14480bda337274a988627272fbe696bcaf5589 Author: Matteo Frigo Date: Sat Aug 3 20:04:57 2002 -0400 Detect ultrasparc (sort of) commit 946e964b908a9fcd9b98345a5f525049b8143cce Author: Steven G. Johnson Date: Sat Aug 3 19:38:17 2002 -0400 added solvtab_rdft_r2r placeholder commit db8c63ea924d244e0c207d514dd425bfab39f2b6 Author: Matteo Frigo Date: Sat Aug 3 19:34:49 2002 -0400 Damn solaris commit 6f4f2a31d28db1040f796b703d9b6c9fd7b4052d Author: Steven G. Johnson Date: Sat Aug 3 17:55:44 2002 -0400 use E extended precision in solvers commit eb1a98695f9827716943ddc0ca00475c2d61d9c2 Author: Steven G. Johnson Date: Sat Aug 3 17:53:29 2002 -0400 an alternative notation for D{C,S}T: DXTio, where i/o are {0,1} according to whether the input/output are shifted, respectively. Alternatively, io is the binary representation of the usual DXT-{I,II,III,IV} nomenclature, minus 1. commit 24b13985e810f08cbef3c5dac739433c5ac0161a Author: Steven G. Johnson Date: Sat Aug 3 17:49:11 2002 -0400 use E extended precision in solvers commit 46b2fc024b187b4356bf6a7977d508a4c4ba22c1 Author: Matteo Frigo Date: Sat Aug 3 15:39:49 2002 -0400 More portability fixes, compiler bugs workarounds, etc. commit ca88f96aed7b0399f4d2199342c5287639e51d3b Author: Matteo Frigo Date: Sat Aug 3 15:09:56 2002 -0400 More portability work commit 3cfd742c2225f91d295d75af9e6ddc46cd4c39f4 Author: Matteo Frigo Date: Sat Aug 3 14:33:40 2002 -0400 Improved portability, removed gnu make dependencies commit ac8aa3edbc9864af3b3e3e8d753cc2388b80732c Author: Matteo Frigo Date: Sat Aug 3 13:48:53 2002 -0400 Remember to thank XXX commit 807dc0e147fedfa044a4ae2a03dbff426e155136 Author: Matteo Frigo Date: Fri Aug 2 17:38:18 2002 -0400 Multiplication on altivec requires FMA with -0.0 to be IEEE754 compliant. commit dfa0ebdb72edd084c82b682b62fffdbd8f9a7611 Author: Matteo Frigo Date: Fri Aug 2 15:26:37 2002 -0400 Allow for extended precision in codelets commit 2eee7899ea3308e919dbeafffeee423dd0c810b5 Author: Matteo Frigo Date: Fri Aug 2 08:52:04 2002 -0400 Shortened names commit 239f0f6f2197b4761abad5f8ac2f1da6736a5ccd Author: Steven G. Johnson Date: Fri Aug 2 03:49:09 2002 -0400 added infrastructure for future r2r transforms commit 4f64527883bd151d5f597abec9870dc9e6d0c8b7 Author: Matteo Frigo Date: Thu Aug 1 21:29:14 2002 -0400 Version info commit 1f6a7039b9fe3a439d6aa9fa83d179fb864ab920 Author: Matteo Frigo Date: Thu Aug 1 21:06:22 2002 -0400 Listened to one customer and added radix-12. Added radix-15 for consistency (whatever that is) commit ece6187a35d44322c45b0fc946187615d8d3bebd Author: Steven G. Johnson Date: Thu Aug 1 19:50:53 2002 -0400 whoops again, fixed the wrong line commit 53c48f4c8eb4f39a1bcea9b47a2cf78c669e2dd2 Author: Steven G. Johnson Date: Thu Aug 1 19:50:16 2002 -0400 whoops commit afb281f39223c26fe968873928fd8ca0c69c1fe7 Author: Steven G. Johnson Date: Thu Aug 1 16:01:15 2002 -0400 use new AC_INIT and add VERSION to wisdom commit 1d4b7a029734d0948b44713fb94429ffd4ce40d4 Author: Steven G. Johnson Date: Thu Aug 1 14:56:45 2002 -0400 mygetR -> getR commit 010ffe455949d901be083a52aeb485e933d4c252 Author: Steven G. Johnson Date: Thu Aug 1 14:56:02 2002 -0400 scanner cleanups: just return 0/1, simplify integer reads commit 052184d84276b884548c95a76e89d5f2ccd124d2 Author: Matteo Frigo Date: Thu Aug 1 08:04:01 2002 -0400 Reverted back to casting pointer to ulong commit c61b1e4aa77a06a1565d2f816bc2b6a22c82f6d8 Author: Matteo Frigo Date: Thu Aug 1 08:03:46 2002 -0400 Cast to unsigned long, not long commit bc2a8794eec9dbdd2eaa2d10070974ab1cdcc3bf Author: Steven G. Johnson Date: Thu Aug 1 03:14:50 2002 -0400 additional comment commit 72bc55e7f202b4d772bc8a50263870f1434becb0 Author: Steven G. Johnson Date: Thu Aug 1 03:12:37 2002 -0400 added comment commit 980a9e749d1361de03ea2256209ee0216942a6aa Author: Steven G. Johnson Date: Thu Aug 1 03:03:18 2002 -0400 added wisdom import commit b9bcf9486c742271f7c9fa64f41791666cf16cb6 Author: Steven G. Johnson Date: Wed Jul 31 23:12:05 2002 -0400 whoops commit 183a8a7311c571981db4ef087608b599de96b062 Author: Steven G. Johnson Date: Wed Jul 31 22:06:46 2002 -0400 use %u for alignment_of commit f9cc3f2e326569214e7ac246b5dacabe10f9f4aa Author: Steven G. Johnson Date: Wed Jul 31 21:47:15 2002 -0400 ptrdiff_t form commit 26346129bd45ff91529e18e5770220025ae5cc8c Author: Matteo Frigo Date: Wed Jul 31 21:33:35 2002 -0400 Cast to avoid warning from C++ compiler commit dc8c0c64365fd7f14a579a730f50107f4c01839e Author: Matteo Frigo Date: Wed Jul 31 18:57:04 2002 -0400 Make problem equality depend on alignments. commit 185babf3691983eb1fc109f4d2864ea80070319f Author: Matteo Frigo Date: Wed Jul 31 15:45:31 2002 -0400 Shorter names commit d0a23f2a7ca0ef90c893e1bc9fe38562bf4b97c4 Author: Matteo Frigo Date: Wed Jul 31 14:38:00 2002 -0400 Oops commit db553c5b6c9be77013e5e6862aecb074abd05daf Author: Matteo Frigo Date: Wed Jul 31 14:37:19 2002 -0400 Fix warning commit 20ce4a31106f745c8765cafa87b94df7b152ba01 Author: Matteo Frigo Date: Wed Jul 31 07:52:53 2002 -0400 Removed silly abstraction barrier. Also, cons() terminology was no longer appropriate. commit 6e519e71ee2bff45a45acc9860e6688b5a2ac0ca Author: Steven G. Johnson Date: Tue Jul 30 22:35:24 2002 -0400 removed register_registrar and solvtab_exec_reverse hacks commit 3bb2201fd6c0b2a0e2e6e1cb07849fc640c23fe4 Author: Steven G. Johnson Date: Tue Jul 30 19:54:41 2002 -0400 register_registrar doesn't search whole solver list (maybe we should change register_solver instead) commit acf987d04a520c14c0d452f2036338e4d89e91a0 Author: Steven G. Johnson Date: Tue Jul 30 19:36:37 2002 -0400 credit commit 1ae9a399e262ce07b3733a11fcb23ea08541bd45 Author: Steven G. Johnson Date: Tue Jul 30 19:34:16 2002 -0400 added HP/UX ia64 support, courtesy of Teresa L. Johnson commit 76ce2ea38a0a18376e316ee3348e8ffd069aebe1 Author: Matteo Frigo Date: Tue Jul 30 13:28:33 2002 -0400 Fixed alignment checks commit 7356d1bc11f552e41d0de8df8fc9e0ef4f83b1a0 Author: Steven G. Johnson Date: Tue Jul 30 01:20:11 2002 -0400 ugh, wisdom id fixes in exprt_conf commit 110cfd3d5abb89da042f3953d99179c04fcb6839 Author: Steven G. Johnson Date: Tue Jul 30 00:41:15 2002 -0400 exprt_registrars -> exprt_conf, added missing SOLVTAB_END commit 331b32dd8322273182a47c852416afaac4f6007b Author: Steven G. Johnson Date: Tue Jul 30 00:36:26 2002 -0400 exprt_registrars should output self-contained configuration commit ddd63d9b49d333a58f352f4f561a6ff1fbe17a5a Author: Steven G. Johnson Date: Mon Jul 29 23:52:07 2002 -0400 added exprt_registrars commit 691ba278639460f94cfd6ff45e14e10007d4f62c Author: Steven G. Johnson Date: Mon Jul 29 23:42:27 2002 -0400 whoops commit ebcd431d564b1f5f86f3bb274ed123971d449415 Author: Matteo Frigo Date: Mon Jul 29 21:24:51 2002 -0400 More stringent requirements on strides for SIMD codelets commit 4fa11627e55b15059ce9b91dce1383c29040f2bb Author: Steven G. Johnson Date: Mon Jul 29 21:05:49 2002 -0400 remove warning commit 30f4b2f2ca2fd97ae591c98d812ec38546a1cd8b Author: Steven G. Johnson Date: Mon Jul 29 20:51:19 2002 -0400 use %td for ptrdiff_t and %T for tensors commit 33c7a10abb7b7b1e3250654481f536b6e33de824 Author: Matteo Frigo Date: Mon Jul 29 16:17:11 2002 -0400 Fix for SIMD commit 1688dda0ec01678ac3d2e16af154c4898a56b568 Author: Matteo Frigo Date: Mon Jul 29 15:40:53 2002 -0400 Missing lfftw_mkstride and lfftw_stride_destroy commit 2e84b7c68c4270593cc2a1c152520b6f55e3c0c9 Author: Matteo Frigo Date: Mon Jul 29 14:34:46 2002 -0400 Implement LDA/STA commit 385b21d8dc7b1c465acbb83b5414caefa80960d2 Author: Matteo Frigo Date: Mon Jul 29 14:19:21 2002 -0400 More SIMD work commit 00e43e5facae3c33e901ca12dd57cf5905c8508d Author: Matteo Frigo Date: Mon Jul 29 13:16:12 2002 -0400 Cleanup commit 6fb8177180cf59f95bc37163f4e8d4c68b1657e8 Author: Steven G. Johnson Date: Mon Jul 29 13:02:38 2002 -0400 update commit 8354486a52f87afe52440aa3316acec7c768ac75 Author: Matteo Frigo Date: Mon Jul 29 12:45:33 2002 -0400 Also check strides in SIMD codelets commit 7b48f56b4e18bd9799c46214829e7b15531e5244 Author: Matteo Frigo Date: Mon Jul 29 11:26:08 2002 -0400 Minor changes, mostly for consistency with the big-endian processor commit ec8f6e4c58d50603587e0bba533ec2086e6174b8 Author: Steven G. Johnson Date: Mon Jul 29 00:50:06 2002 -0400 added comment commit a7cc792884a9acb1d81a346cdab1d9b07e9b2bdf Author: Steven G. Johnson Date: Sun Jul 28 21:19:35 2002 -0400 added code for icc's _mm_malloc (memalign replacement) commit d1398d4a205bae587e99b40049dd6a301c1f58f9 Author: Steven G. Johnson Date: Sun Jul 28 17:33:07 2002 -0400 slight fixes commit 5f21f0a04a242174ff85c63925c88e15e6ff101e Author: Steven G. Johnson Date: Sun Jul 28 16:28:43 2002 -0400 whoops commit d5256b19914cddf9b241ebce04f10042f4837e9b Author: Matteo Frigo Date: Sun Jul 28 16:13:19 2002 -0400 Use vec_xor to change sign commit ec0a29c8d03cbed27c09a96fcb3f022bfc9f647f Author: Steven G. Johnson Date: Sun Jul 28 16:10:59 2002 -0400 added rdft2 commit 516c9c1117a0811dba416bfa1ba20a5c93e91532 Author: Matteo Frigo Date: Sun Jul 28 15:45:54 2002 -0400 Optimized commit 8619a2039d6723004a1fef760203e5d6f33f9469 Author: Matteo Frigo Date: Sun Jul 28 15:11:14 2002 -0400 Changed ALIGNMENT commit 37c7c1fd79835b212e94e565fb3abe0352243919 Author: Matteo Frigo Date: Sun Jul 28 15:09:40 2002 -0400 alignment := 8 commit 27b891f615494d0f1996ed1acf0665eb386704e6 Author: Matteo Frigo Date: Sun Jul 28 14:57:22 2002 -0400 Avoid warning commit fd53f4d5bdfc8c8daf0126fab34270b09cf49aa2 Author: Matteo Frigo Date: Sun Jul 28 14:53:03 2002 -0400 Oops commit 8b749b4406276bb4b4d6e3b4c78486d0fea38fff Author: Matteo Frigo Date: Sun Jul 28 14:50:09 2002 -0400 New altivec experiment commit 87bd001083f039c6728a3a19d03b7e14eac11666 Author: Matteo Frigo Date: Sun Jul 28 13:48:20 2002 -0400 Nothing commit 3a5876fd4582a075560988801d7c958b0ca75a74 Author: Matteo Frigo Date: Sun Jul 28 13:47:50 2002 -0400 Oops commit 551ad6c0e199fcd5fce5defd470ce7d975dcacb4 Author: Matteo Frigo Date: Sun Jul 28 13:44:28 2002 -0400 Nothing commit a13f42aa3c37a3065a08a59220529d5292683ac9 Author: Matteo Frigo Date: Sun Jul 28 10:38:10 2002 -0400 Constants are now in separate file. commit 94226e68396c790ce6bfbbf8db0c299fed32e2f9 Author: Matteo Frigo Date: Sun Jul 28 07:58:37 2002 -0400 More precise comment commit 4009a4d5579eb5520346c956632ef0c2df5273d2 Author: Matteo Frigo Date: Sun Jul 28 07:56:40 2002 -0400 gcc-3.1 bug workaround commit 092830f99bf3fd15390980b4e441d4c7d1a9826c Author: Steven G. Johnson Date: Sun Jul 28 01:39:54 2002 -0400 slight optimization, and exported zerotens functions commit e3797dbb5984f5f1272b452c7005c775badb6fb2 Author: Steven G. Johnson Date: Sun Jul 28 00:54:59 2002 -0400 should be a plan_dft, not a plan_rdft commit ab69981af2f5c80981e7f1432560cbdaae08770c Author: Matteo Frigo Date: Sat Jul 27 21:36:46 2002 -0400 Optimizations. Make it work with vanilla non-Apple gcc. commit 9a7ad02a4bfebfa91a4afe01756023a3a74f5d8b Author: Steven G. Johnson Date: Sat Jul 27 19:20:09 2002 -0400 whoops commit 4aac8a4d98395b964b16b1251d8d52410fc232f8 Author: Steven G. Johnson Date: Sat Jul 27 18:54:01 2002 -0400 added hc2r (dif) commit abe907208a2a3e7ff558b3f12bb0b254768d670a Author: Steven G. Johnson Date: Sat Jul 27 18:31:43 2002 -0400 add hc2r (dif) case commit b933474c3373bdca65dd9cce3b16272c2b197ee8 Author: Matteo Frigo Date: Sat Jul 27 15:09:40 2002 -0400 Altivec port commit 0884acf4e8fc2cd9ec4144877e5a0879bbf779e6 Author: Matteo Frigo Date: Sat Jul 27 15:06:21 2002 -0400 Fixed signed/unsigned bug. commit 11508c3160c5d3a404a58eb143139d9088a213e5 Author: Matteo Frigo Date: Thu Jul 25 20:11:26 2002 -0400 Make rank0 unapplicable to in-place problems. commit 81a49b1e405be525a9ee5476ddfa16e8c70ef702 Author: Steven G. Johnson Date: Thu Jul 25 17:10:52 2002 -0400 only works for r odd commit 2b54747fb0e87bbd03b3c7b04ed1cb752a470796 Author: Matteo Frigo Date: Thu Jul 25 15:30:06 2002 -0400 Reinserted much better timing-avoidance heuristic commit 171716115f0f318397186964ecc341ac9268fd84 Author: Matteo Frigo Date: Thu Jul 25 15:21:13 2002 -0400 Score is now a property of the plan, not of the solver. Revised representation of closures. commit 67c69e319a7ca8ac6c81a45a1d0f6dde9efc2e12 Author: Matteo Frigo Date: Thu Jul 25 06:36:51 2002 -0400 Cosmetic changes. Added hc2r_128.c commit 0a22b8dd9629f62d1a682af581c17d6dc71e244a Author: Steven G. Johnson Date: Thu Jul 25 01:37:53 2002 -0400 added hc2r commit 22bad3aea85c62120134db4652c6ac990c8607e2 Author: Steven G. Johnson Date: Thu Jul 25 00:51:45 2002 -0400 added hc2hc-difbuf commit aac8e9d03008ccbe1c244717e404e283c03eabe1 Author: Steven G. Johnson Date: Thu Jul 25 00:25:06 2002 -0400 added rdft-dif commit 39d632acade375e06e60dc11cd0b693ed29bbf07 Author: Steven G. Johnson Date: Thu Jul 25 00:22:36 2002 -0400 whoops, hc2r must be conjugated to have right sign commit ebc9e7b4083f1d545cc47032a7bffbcc5d5a26ce Author: Steven G. Johnson Date: Wed Jul 24 23:27:45 2002 -0400 slight change commit 6c5a0b11d3a86a418e02108a90472ff19d97bae0 Author: Steven G. Johnson Date: Wed Jul 24 23:24:24 2002 -0400 whoops commit 28adebe469b82ee53e436f33389b459d8707a603 Author: Steven G. Johnson Date: Wed Jul 24 22:46:39 2002 -0400 support hc2r codelets commit af7b3ec85871349e26698fb5edf95c6a1e96bbbf Author: Steven G. Johnson Date: Wed Jul 24 22:01:53 2002 -0400 use vector plan for r/i instead of two separate plans commit b31e3e7d86ef1ab3aa58145768cc801979ba5cd6 Author: Steven G. Johnson Date: Wed Jul 24 20:36:34 2002 -0400 hack to allow rader/generic to work in-place for small prime sizes, instead of always using buffered commit cddf15b3b7c1d3baec98982550f18344c3361216 Author: Steven G. Johnson Date: Wed Jul 24 18:04:41 2002 -0400 added rdft-generic commit 76637f738e056d7e4fcba907ffd4ab52db457fed Author: Steven G. Johnson Date: Wed Jul 24 17:27:34 2002 -0400 fixed add count commit 7c1f6a8f3b35a5034daacc521a10c06424144047 Author: Steven G. Johnson Date: Wed Jul 24 14:52:26 2002 -0400 again commit ab910c9e4a7fc66e0a19e1b9557669e896ac465b Author: Steven G. Johnson Date: Wed Jul 24 14:51:58 2002 -0400 slight fix commit 2169c91de93a2c096765218e2b25e32e6f2d47f0 Author: Steven G. Johnson Date: Wed Jul 24 14:51:07 2002 -0400 fixed comment commit b6ed79694396f04555b0009027b94355c81a4019 Author: Steven G. Johnson Date: Wed Jul 24 14:41:24 2002 -0400 whoops commit 10fabba80f177e1ee4bfca04ac09836c798998ef Author: Steven G. Johnson Date: Wed Jul 24 14:38:15 2002 -0400 added rader-hc2hc commit 3015fea221f119cf88e68c12087c0ca8fbb508a9 Author: Steven G. Johnson Date: Wed Jul 24 00:07:59 2002 -0400 whoops, initialize W commit d48486c4715a0db6bb2653a34d868f5f52732f66 Author: Steven G. Johnson Date: Tue Jul 23 23:03:09 2002 -0400 strides should not be unsigned commit 7d6e7cacd21c97ef1622d681de2543e71ac2171d Author: Steven G. Johnson Date: Tue Jul 23 23:02:08 2002 -0400 more stride sign fixes commit b967fadc107addb8cec4effc1f0e7ae7d6ce1f86 Author: Steven G. Johnson Date: Tue Jul 23 23:01:04 2002 -0400 strides should not be unsigned! commit 0ad85517c669d39fcf0ac6f77e73ed8c2fa80e89 Author: Steven G. Johnson Date: Tue Jul 23 14:55:25 2002 -0400 added comment commit 5d278e1ac3640bc39cd6b7e19aaa5563cd319de4 Author: Steven G. Johnson Date: Tue Jul 23 14:52:04 2002 -0400 another fix to op count commit 9260aed8161a66eb5de14e68c932d11bc113cd56 Author: Steven G. Johnson Date: Tue Jul 23 14:51:01 2002 -0400 whoops commit 3f42b7510d2c0f2b1e7bc34342041f8123667897 Author: Steven G. Johnson Date: Tue Jul 23 14:49:43 2002 -0400 slight fix to op counts commit 1288dec288612070c531c98067255cf3de3d90b1 Author: Steven G. Johnson Date: Tue Jul 23 14:09:19 2002 -0400 added dft-r2hc commit ad4bf834d8b55b38d2766779e5d00c4f61e30dbe Author: Steven G. Johnson Date: Tue Jul 23 02:50:12 2002 -0400 better comment and var. name commit f1ab8ef1b9cf77432f6bb627a5c3ec2f586ebcd9 Author: Steven G. Johnson Date: Tue Jul 23 02:39:11 2002 -0400 fixed tests for hc2r, and added r2hc-hc2r commit d3b91945fd199f6bb99711479972b7074c00b352 Author: Steven G. Johnson Date: Tue Jul 23 00:45:23 2002 -0400 added rader-dht commit 57036068d38970156c0bcf5d4edd72cdb20a09fd Author: Matteo Frigo Date: Mon Jul 22 21:05:12 2002 -0400 Added r2hc_128, what the hell. commit d82c1c99be202e2cc55851a4bd406b4682cb0b4a Author: Matteo Frigo Date: Mon Jul 22 20:48:59 2002 -0400 Added codelets that compute twiddle factors commit f98ad798168562c0da5714110eb0c37972178728 Author: Steven G. Johnson Date: Mon Jul 22 19:57:16 2002 -0400 added rdft-buffered commit 50b0158abe60a9e62698baf54e31623bf29a34f9 Author: Steven G. Johnson Date: Mon Jul 22 19:43:39 2002 -0400 added hc2hc-ditbuf commit 4b7abfd7514cb4d98a0c87746c25fcafe6d263b3 Author: Steven G. Johnson Date: Mon Jul 22 14:29:04 2002 -0400 use STACK_MALLOC (alloca), since generic radix is always small commit d083d389c40a363c4b90a6ca2efd202c52b81713 Author: Steven G. Johnson Date: Mon Jul 22 14:22:43 2002 -0400 small cleanup commit 851d792b2de11df3620f32093e02632f78aeef6e Author: Matteo Frigo Date: Mon Jul 22 07:42:13 2002 -0400 What the hell was I thinking? commit 7237f72026e6aad4325427a52b0fb683ec4b2e0d Author: Matteo Frigo Date: Mon Jul 22 07:37:12 2002 -0400 Reduced code size by using table instead of switch statement. commit f253821d2c79215c87e18cf134e218e02d0235ed Author: Matteo Frigo Date: Mon Jul 22 07:27:06 2002 -0400 Changed hash function to avoid collisions with DFT. commit 602ef947b9122139e2b55fca3e007ca6bcda4bbf Author: Steven G. Johnson Date: Mon Jul 22 01:37:06 2002 -0400 added missing file, whoops commit 6b3144d456eb3c0caee53880e7fe60ddbd2c48d5 Author: Steven G. Johnson Date: Mon Jul 22 01:24:17 2002 -0400 whoops, generate enough twiddles for odd m commit 4738a6cbbc5206c3fdc7b0bf7cdc481609439497 Author: Steven G. Johnson Date: Mon Jul 22 01:10:21 2002 -0400 don't try to verify R2HCII or HC2RIII plans commit ec9e9517ca4ac29008a9b1b8f79f4543ef4ae90a Author: Steven G. Johnson Date: Mon Jul 22 01:05:00 2002 -0400 recursive case now works, I think commit 7ebf4c56ae4cc7861840cb8ee5d8a482c5e3f64a Author: Steven G. Johnson Date: Mon Jul 22 01:04:40 2002 -0400 add extra impulse test for debugging commit 7dacfd5778747c8ae3b803ddf37d7921eeab713f Author: Steven G. Johnson Date: Mon Jul 22 01:02:38 2002 -0400 whoops, multiply ios offset by stride (and rename to ioffset) commit ca3c5bf3c6de8946f1caf40e779487110e5b59ce Author: Steven G. Johnson Date: Mon Jul 22 00:22:02 2002 -0400 whoops commit e40381e0407f8570c076968fb7c7138ffebe9ae2 Author: Steven G. Johnson Date: Sun Jul 21 23:58:14 2002 -0400 whoops commit 030d0f7f52cbc000070d885b815768bfadb86649 Author: Steven G. Johnson Date: Sun Jul 21 23:43:03 2002 -0400 added hc2hc-dit commit c1c28b632b9cc24c542610dbdb3bf424eb740810 Author: Steven G. Johnson Date: Sun Jul 21 23:15:12 2002 -0400 twiddles can be shared with smaller m's commit cbb0b11b1a8aa023f1d81dba688278012451de8e Author: Steven G. Johnson Date: Sun Jul 21 22:34:28 2002 -0400 preparing for recursive rdft... commit 8f48e0e3caf86690c7328cd128cc981364c9026f Author: Steven G. Johnson Date: Sun Jul 21 19:31:22 2002 -0400 slight fix, to match libbench/verify.c commit d9aec187c49dbc272df30d040d4acfc160220b07 Author: Steven G. Johnson Date: Sun Jul 21 18:43:12 2002 -0400 r2hcII has imag parts offset by n-1, not n. We can also allocate fewer strides. commit 00e3acce93c910450482c37155200244adfc51b4 Author: Steven G. Johnson Date: Sun Jul 21 18:27:09 2002 -0400 delete unused var commit 469254570eda6c6851c8c76ac2ce631c7e160d40 Author: Steven G. Johnson Date: Sun Jul 21 02:06:53 2002 -0400 added some rdft solvers commit c267ad079a4ef7cb7a9fdbe4556c89121137be02 Author: Steven G. Johnson Date: Sun Jul 21 01:52:54 2002 -0400 pass identifier in FFTW() through another macro so that the mangled name can itself be a preprocessor symbol commit cf660c4cf10b80d7ec37cd99825c9663738d77e7 Author: Steven G. Johnson Date: Sun Jul 21 01:05:21 2002 -0400 fix in comment commit 14081a9d216ccc757b9ce46631d956f5135628ea Author: Steven G. Johnson Date: Sun Jul 21 00:47:03 2002 -0400 bench tests rdft plans commit 61ed41f792df937cc80b9fa0c643871ab7449968 Author: Steven G. Johnson Date: Sun Jul 21 00:22:14 2002 -0400 make rdft.h and dft.h compatible commit d314a5c84e70803b32075884ee96c0099c695d9a Author: Steven G. Johnson Date: Sun Jul 21 00:12:19 2002 -0400 first-draft rdft verify commit 01403979be858145b4f1f61f03c9f1f98c59587d Author: Steven G. Johnson Date: Sat Jul 20 22:09:15 2002 -0400 got rid of annoying warnings commit 710e4e4687092d0d823f7fe442c0bc981d99a598 Author: Steven G. Johnson Date: Sat Jul 20 22:07:37 2002 -0400 added stub codelet registration for linking purposes commit 9b9775415f67f53910d95e1ac963f1bed389ff9b Author: Steven G. Johnson Date: Sat Jul 20 21:46:03 2002 -0400 basic rdft stuff commit 9c7a553bedb1f7f2fce816ae284d4867ffc0924d Author: Steven G. Johnson Date: Sat Jul 20 21:06:50 2002 -0400 rdft codelets now compile commit fb7686cdfd1674f03c35ed523fcc2d11db157ecb Author: Matteo Frigo Date: Sat Jul 20 18:40:31 2002 -0400 Oops, was generating rdfts instead of hdfts commit f82dcb63a673b994a4677ed9f8d065766c79c31c Author: Matteo Frigo Date: Sat Jul 20 18:25:47 2002 -0400 Added hc2r codelets commit d8127083d80d0f0b9de30e6e3c9ae1b90f92a90d Author: Matteo Frigo Date: Sat Jul 20 17:54:39 2002 -0400 return W in hc2hc codelets commit 354e28470103a92db21d621263a687a6bf437595 Author: Matteo Frigo Date: Sat Jul 20 17:51:06 2002 -0400 Some work on rdft codelets commit 8a7b5a3242b8bd823c8d70e1b04e9492d6d65d43 Author: Matteo Frigo Date: Tue Jul 16 13:55:50 2002 -0400 fix const commit 6a3576889b8683e9ee15d2f95fb76f6fef645667 Author: Matteo Frigo Date: Tue Jul 16 07:00:10 2002 -0400 Separate CFLAGS in codelets. Fix const in certain places. commit 7870f6dff402e655def77265bc4ac0225608f677 Author: Steven G. Johnson Date: Mon Jul 15 21:10:42 2002 -0400 note buffering problem commit 0bd6af885007a6f0561577c521d4999c903f27df Author: Matteo Frigo Date: Mon Jul 15 20:27:51 2002 -0400 Removed unpredictable branch from inner loop commit 8a40f059239add905fa4c0abd6c20a40644559fa Author: Steven G. Johnson Date: Mon Jul 15 19:35:04 2002 -0400 update commit 45bb25aa64ce46c8821c9717770a28a5ab60e492 Author: Steven G. Johnson Date: Mon Jul 15 19:31:39 2002 -0400 optimization commit 526958106e6a43bfc1d4a7cab335fc3df41a7d9f Author: Steven G. Johnson Date: Mon Jul 15 19:28:30 2002 -0400 added generic dit commit aebc10cb69074f33b3370de5aff7bc20b684800b Author: Steven G. Johnson Date: Mon Jul 15 17:03:53 2002 -0400 whoops, mksolver should be static commit e2b6303fa6575e6796c2834f222b77d221e1a77e Author: Matteo Frigo Date: Mon Jul 15 16:46:36 2002 -0400 First implementation of gen_hc2hc, probably still buggy. commit 0105f03531806f86cc1c0e89c18b91947b15cb10 Author: Steven G. Johnson Date: Mon Jul 15 16:40:23 2002 -0400 don't count loading of twiddle factors in ops.other, since it isn't counted for the codelets commit 47f3220441f5da7ee844e0abd36f41f32b4bc17e Author: Steven G. Johnson Date: Mon Jul 15 15:13:19 2002 -0400 plan_destroy puts plan to sleep before deallocating it, to eliminate duplicate free calls in solvers commit 90347b282680ec2b611ededef7ab7272beb2857a Author: Steven G. Johnson Date: Mon Jul 15 15:07:41 2002 -0400 fftw2-like vector recursion flag commit df45d5d1588019d57e80aee326c92e5ffb49715f Author: Matteo Frigo Date: Sun Jul 14 21:01:44 2002 -0400 More jokes commit 5efd22c7d5499d7bad84216e526ec11551fb81f0 Author: Matteo Frigo Date: Sun Jul 14 20:36:01 2002 -0400 Bless plan for testing purposes commit dd45761d063a5473473f44d5800a55b35794c8a6 Author: Matteo Frigo Date: Sun Jul 14 20:35:49 2002 -0400 Canonical linked-list deletion (hope it is right) commit 04cbcbfe2eb171da6ee678d000f1cf7aa2435f5d Author: Steven G. Johnson Date: Sun Jul 14 18:26:19 2002 -0400 use estimating planner for cld_omega commit c2e125a60dc8101c25c1f08debd9a4b1661b1658 Author: Steven G. Johnson Date: Sun Jul 14 18:10:56 2002 -0400 better internal naming commit fac5147b9b14fe2997cde8bbd5a39c956f577eaf Author: Steven G. Johnson Date: Sun Jul 14 18:10:01 2002 -0400 printing should really be fixed now, grrr commit 8dbd07648153ea12bd52c12aad39b58adc479140 Author: Steven G. Johnson Date: Sun Jul 14 17:57:12 2002 -0400 print all distinct child plans commit af0c968546d0c39197e3c7925e21bf1897f1b5ab Author: Steven G. Johnson Date: Sun Jul 14 17:49:21 2002 -0400 whoops commit af601a5405861e68cedd4314f70b677b6c36e640 Author: Steven G. Johnson Date: Sun Jul 14 17:45:54 2002 -0400 whoops, destroy should delete twiddle/omega from list commit a20712e3a4b5d2364f092fd222b540cbc8c2df44 Author: Steven G. Johnson Date: Sun Jul 14 17:33:02 2002 -0400 whoops commit a43e6c2aeb20b20987891fa7461cc6a2898d1785 Author: Steven G. Johnson Date: Sun Jul 14 17:12:14 2002 -0400 added plan_bless and FORGET_ACCURSED commit 8da186b0e85df747bbd0a91db772c869e9b35e3c Author: Steven G. Johnson Date: Sun Jul 14 16:15:43 2002 -0400 further cleanup commit 255479f4ad3175722fe32fd4a2b7cefa858b3b35 Author: Steven G. Johnson Date: Sun Jul 14 16:14:15 2002 -0400 slight cleanup commit d5346f1dfc7154d9a4fbade8fba1dcce90c7bec5 Author: Steven G. Johnson Date: Sun Jul 14 16:09:17 2002 -0400 added traverse_plan via print (ugh) commit 1edcc5b2fb3cf8741aec7b32042453803a1eb145 Author: Steven G. Johnson Date: Sun Jul 14 15:08:29 2002 -0400 added TW_FULL, and additional n parameter for twiddles commit c300c4c9e119ee5e657fe52fa48ce8251321f8a2 Author: Steven G. Johnson Date: Sun Jul 14 15:03:51 2002 -0400 whoops commit 8eb08032b56dac1d0b4200c2a1e17e6a33674395 Author: Steven G. Johnson Date: Sun Jul 14 13:49:20 2002 -0400 save flags before invoking solver mkplan commit 68d1b66d54458eb685bee1a95bd4433421a68f58 Author: Matteo Frigo Date: Sun Jul 14 09:28:37 2002 -0400 [empty commit message] commit 8f335f455b54a35089073c690ecd93c7380c1b95 Author: Steven G. Johnson Date: Sat Jul 13 22:17:29 2002 -0400 added support for UNICOS _rtc() real-time-clock intrinsic function commit e0550777d2519623392bd1678c39e7491fd3d38c Author: Steven G. Johnson Date: Sat Jul 13 22:06:35 2002 -0400 fixed typo: HAVE_TIME_H should include time.h, not sys/time.h commit fcff09d063384ac24b87c16cfed1c246de45623f Author: Steven G. Johnson Date: Sat Jul 13 21:46:02 2002 -0400 support AIX read_real_time timer commit ca89737634be3d5de4851c1f3fbc19d900cf22b0 Author: Steven G. Johnson Date: Sat Jul 13 17:02:51 2002 -0400 use && instead of the (sigh) unportable -a commit 769cf9267df8b75c3e2849a171e995136bacd4af Author: Steven G. Johnson Date: Sat Jul 13 16:38:18 2002 -0400 use AC_HELP_STRING commit 6600ee1ae97f1919117e4c3877092249443c545b Author: Steven G. Johnson Date: Sat Jul 13 16:05:43 2002 -0400 support long-double precision commit d7aff35e1553f8735b56597fd524c5b52d8e475f Author: Steven G. Johnson Date: Sat Jul 13 15:48:10 2002 -0400 whoops whoops commit 9a20964a145ceef9018cf8bf0977be7ba63ecb6e Author: Steven G. Johnson Date: Sat Jul 13 15:47:39 2002 -0400 whoops commit d040c7ef353abc5accf76a9953a26feb2d11fc0c Author: Steven G. Johnson Date: Sat Jul 13 14:13:42 2002 -0400 buffered solver strides have been fixed commit 6bcbee663a0b8b8b23b70a180e1ca12ee1141724 Author: Steven G. Johnson Date: Sat Jul 13 13:48:13 2002 -0400 convention commit 4d3d49e4b447b49a45b803fea4ff4d23a31288a0 Author: Steven G. Johnson Date: Sat Jul 13 12:50:06 2002 -0400 share twiddle arrays in Rader commit 91dbf0b319de38c0b67df70aa4c39ccac0b523da Author: Steven G. Johnson Date: Sat Jul 13 12:48:10 2002 -0400 call done() after verify commit b5b07111cda5f2b5b1130829d05b698575f4a5f8 Author: Steven G. Johnson Date: Fri Jul 12 15:42:04 2002 -0400 output planner time with -v commit b71bd73584d1e960018fbda1b8f078fa4e2ea542 Author: Steven G. Johnson Date: Fri Jul 12 15:40:14 2002 -0400 support double outputs commit e808db8fbfe2b7d4afbabe558d5a757379c49dd8 Author: Steven G. Johnson Date: Fri Jul 12 15:09:19 2002 -0400 removed extraneous parens commit b1ca74610947a0007932fb0eb65c794868f15977 Author: Steven G. Johnson Date: Fri Jul 12 15:08:13 2002 -0400 increase maxbufsz to 64k; makes a big difference for large 2d transforms commit 92dfa732c944f37774e1b4b9b889ba64a7621ccb Author: Matteo Frigo Date: Fri Jul 12 05:59:26 2002 -0400 Fix commit fdae83d7830d6df98ac417066e0c5ea8bc254d35 Author: Steven G. Johnson Date: Fri Jul 12 01:22:38 2002 -0400 fix comment commit 8b316634377ad2f829c26d6d107005638c6ab8ba Author: Steven G. Johnson Date: Fri Jul 12 00:59:29 2002 -0400 fix in comment commit efb8ce2f3a2e518f02245f8eb12425a30bb420c0 Author: Steven G. Johnson Date: Fri Jul 12 00:13:13 2002 -0400 updated commit 21a94bd1783b4cde2728d54932cdf1ecf2485a21 Author: Steven G. Johnson Date: Thu Jul 11 23:39:27 2002 -0400 buffered malloc's buffers commit 2cec064ce9f3fd0ccd891651557a5739409c19c3 Author: Steven G. Johnson Date: Thu Jul 11 23:30:26 2002 -0400 share more code between apply and apply_dit in Rader commit afd5fe37e6c3060145530115b6c2f2d676ddfe37 Author: Matteo Frigo Date: Mon Jul 8 12:30:34 2002 -0400 Polished commit 71ae7be079984537f7770d1b70280b77ad23c722 Author: Matteo Frigo Date: Mon Jul 8 09:47:11 2002 -0400 [empty commit message] commit a5760afe5aed6838383937fad0a3243528ce15fc Author: Matteo Frigo Date: Mon Jul 8 09:42:08 2002 -0400 SIMD/FMA stuff commit ec76a60088a86df970b3cbf4005506ade4570040 Author: Matteo Frigo Date: Mon Jul 8 07:43:51 2002 -0400 Avoid code duplication commit ffce0587abc26960f1bffb08b99f61280176d25b Author: Matteo Frigo Date: Sun Jul 7 20:56:15 2002 -0400 Fixes for FMA+SIMD commit cd1278e203d5014ee55026d00eef0c5cd87091a7 Author: Matteo Frigo Date: Sun Jul 7 20:32:01 2002 -0400 Major changes in SIMD fftw commit 47b31e4f895a8385d29297182fd4ab3cbe756486 Author: Matteo Frigo Date: Fri Jul 5 17:32:09 2002 -0400 Use unpck instructions instead of shuffles commit 8390c4b680fa05c264d6308d70aeb8b69e97b44a Author: Matteo Frigo Date: Fri Jul 5 15:49:14 2002 -0400 Minor tweaks commit 9939d14df8ec43f0f3724eccce6907f723ae7bcd Author: Matteo Frigo Date: Fri Jul 5 15:02:54 2002 -0400 Use score planner commit db780c34509c8cc70385f63815249dbb982371ab Author: Matteo Frigo Date: Fri Jul 5 14:49:59 2002 -0400 Added verifier commit a4c35fbcefcca25a0e31431dcdabb0d44a2bfb98 Author: Matteo Frigo Date: Wed Jul 3 20:32:28 2002 -0400 More simd codelets commit 0b41b3e8c38d89cca050b1b0df6110056a55463e Author: Matteo Frigo Date: Tue Jul 2 16:18:09 2002 -0400 Oops commit 4aa14927597947c2f2c0c38064e92ce29850f3eb Author: Matteo Frigo Date: Tue Jul 2 16:13:24 2002 -0400 Fixed classic mode commit b32c4fa8b6fddb6a4af23a7d2794adb53733fc2d Author: Matteo Frigo Date: Tue Jul 2 15:38:36 2002 -0400 Use LDK for constants so that we can play games. commit 38505faa2d20e4d958d80dce05620dbf20420822 Author: Matteo Frigo Date: Tue Jul 2 13:15:58 2002 -0400 Improved support for fixed strides commit 2c99260bbd5e86361b09120817f07543384fc5e0 Author: Matteo Frigo Date: Tue Jul 2 10:30:58 2002 -0400 Changed accounting of flops commit ae3999cb3d0ec0d5140c9dead499de0cf4318e5e Author: Matteo Frigo Date: Tue Jul 2 08:51:38 2002 -0400 Wrong code in non-fma mode commit 55015bd78bfbcbffb833554b7df558efd1a826cc Author: Matteo Frigo Date: Mon Jul 1 23:17:06 2002 -0400 sse2 stuff commit 021b59968903662e9727f7833c19c613f1b192b7 Author: Matteo Frigo Date: Mon Jul 1 14:05:56 2002 -0400 Identify CPUs for special codelets commit f304f0151ac1ad333b9450a6d78c8cd8f8724d1a Author: Matteo Frigo Date: Mon Jul 1 09:26:42 2002 -0400 Change split problem syntax commit 48a37449fcbd51779bb3dde3f3b8c2d02bbc323d Author: Matteo Frigo Date: Mon Jul 1 09:11:39 2002 -0400 Removed -fma flag commit 38d63d20ab1bd900f5f89914ca7b7f90191d866f Author: Matteo Frigo Date: Sun Jun 30 20:08:26 2002 -0400 Work around gcc bug commit 6b96cea114710b35caa1d65a669f92a46e20f27f Author: Matteo Frigo Date: Sun Jun 30 18:34:06 2002 -0400 New simd stuff commit 5dd26fcef247d6776f4b5b587b15a1b317a97431 Author: Matteo Frigo Date: Sun Jun 30 17:00:09 2002 -0400 Added altivec support commit a81146d09b6762b1c82b9f998720454245ee5e25 Author: Matteo Frigo Date: Sun Jun 30 14:47:47 2002 -0400 Forgot file commit 734f270fddd3a03066eb4323706090f99c82bf9c Author: Matteo Frigo Date: Sun Jun 30 14:37:55 2002 -0400 Progress towards simd implementation commit 05dc59aa43296d7fb6b79edf5decf65e9a100401 Author: Matteo Frigo Date: Tue Jun 25 20:23:29 2002 -0400 Add 128- codelet commit a94cde7e798f7da510f20b25d1a4bd4aea4dc58b Author: Matteo Frigo Date: Sat Jun 22 20:47:28 2002 -0400 More simd changes. Ensure proper stack alignment in k7 codelets. commit c4ef1c53884a0a7fee4587ef0b1f4317ae83a644 Author: Matteo Frigo Date: Sat Jun 22 13:01:33 2002 -0400 Fixed prototypes commit 5383095f6dd63b1db544695c4b0f7f244344d698 Author: Matteo Frigo Date: Sat Jun 22 12:53:26 2002 -0400 Sparc cycle counter requires v9 commit a25e9ee31d663307683038fbf703f0ed60a502e3 Author: Matteo Frigo Date: Sat Jun 22 11:45:48 2002 -0400 Minor fixes commit 01922e916a7428cc0f7f188518720710b101e9e6 Author: Matteo Frigo Date: Sat Jun 22 11:11:46 2002 -0400 Fixed ev67 detection commit ff1e337f46f9f1bbdcc32756bcdaa0e86566df2e Author: Matteo Frigo Date: Sat Jun 22 10:19:11 2002 -0400 Print flops commit 3f2d94c9ccabcf93d7f8b1230c5b2f97ac6cc4d2 Author: Matteo Frigo Date: Sat Jun 22 09:36:47 2002 -0400 Nothing really commit 8843a01025a26b7bb5a3f26b94c46243332cca25 Author: Matteo Frigo Date: Fri Jun 21 22:19:20 2002 -0400 More simd work commit 3a71fc73cd6c17ba8df8d7aad29fa1bf5ad71a96 Author: Matteo Frigo Date: Thu Jun 20 21:22:41 2002 -0400 More simd work commit fcbb846544687bd8c14e9d004746230e8b6a49e2 Author: Matteo Frigo Date: Thu Jun 20 18:51:33 2002 -0400 More simd work commit 4e25f887266e572b5418dcd0bad8db69265ba3b3 Author: Matteo Frigo Date: Thu Jun 20 15:04:37 2002 -0400 Moving towards incorporation of simd stuff commit 8b98bf67f5e3c976f17988acc8c2e063d3848408 Author: Matteo Frigo Date: Wed Jun 19 18:47:55 2002 -0400 Reorganized k7 stuff into own directory commit 30f0669ed255025d8e4bad6adb9737022a7a6b34 Author: Matteo Frigo Date: Wed Jun 19 13:21:13 2002 -0400 Minor experimental stuff commit 51b0055df9ca6f4c9ea05ccbeb4714bd61d8469b Author: Matteo Frigo Date: Wed Jun 19 11:20:29 2002 -0400 Cosmetic changes commit 3309bd7a80d6dfb0da60d6c2a185d5213505b036 Author: fftw Date: Wed Jun 19 01:43:31 2002 -0400 allocate buffers on the fly commit d290c98ff9fdac1646b56bf9fd80913ecf4b42ed Author: Matteo Frigo Date: Tue Jun 18 17:48:41 2002 -0400 Added ct-ditbuf-k7.c . Major changes required in generator. commit c672bc817602bb03a8e6d7b637e78e84dc2f1f1d Author: Matteo Frigo Date: Tue Jun 18 11:55:57 2002 -0400 Nothing, really commit 327d908d5d396a22fa85216b24203c6b709f9379 Author: Matteo Frigo Date: Tue Jun 18 11:19:59 2002 -0400 !SINGLE ==> !K7_MODE (for some reason the contrapositive sounds wrong) commit fc34a6a92882caf5e8346225e9c25a2a1b6fcb2c Author: Matteo Frigo Date: Tue Jun 18 11:07:13 2002 -0400 Buffer is now symmetric wrt forward/backward transform commit 2dfa3580ad3c4d0960a3b18aba5464c6d9cc91c1 Author: Matteo Frigo Date: Tue Jun 18 10:33:58 2002 -0400 Fixed applicable() in indirect.c commit 40e9e2373f160b3760c87aa707e20aa91a30479d Author: Matteo Frigo Date: Tue Jun 18 08:41:18 2002 -0400 Fixed attempt to free() uninitialized pointer. commit 854771dbd1a262127ab6ef87b31e6ec3645dbce4 Author: Matteo Frigo Date: Tue Jun 18 08:13:55 2002 -0400 Added reference counts for awake() commit b65907ea5ba6b84347976ae59688acbc4a982b82 Author: Steven G. Johnson Date: Mon Jun 17 20:49:05 2002 -0400 updated comment commit 11cbdda3f3c4e688e59eb28d43946c6053ce461e Author: Steven G. Johnson Date: Mon Jun 17 20:46:45 2002 -0400 slight update commit df79d1312e84ddc47b65056c2b59fb6eec20915b Author: fftw Date: Mon Jun 17 02:30:16 2002 -0400 moved prime-number stuff into primes.c, so it can be shared with generic codelet and with rfftw rader commit 92c3a4be474dca7d3e4f2076f2d5c61e040e6739 Author: fftw Date: Mon Jun 17 02:01:58 2002 -0400 added comment commit b580b3303f83589a5b87ad88f104ed41b8c641a3 Author: fftw Date: Mon Jun 17 01:39:55 2002 -0400 added rader-dit commit 51f015c4a6df06fa475e6e291d7e0f8a0293a891 Author: fftw Date: Sun Jun 16 23:50:16 2002 -0400 added initial Rader (no DIT yet) commit d45d1bbd591b1c30885baa7b1025ae016b23e937 Author: fftw Date: Sun Jun 16 22:29:51 2002 -0400 don't warn about long long commit 16ad72af138d08605df2fadf239e142ace2680eb Author: Matteo Frigo Date: Sun Jun 16 21:30:42 2002 -0400 Added k7 DIF codelets commit 3728b053dde52469a7deea4654ca0352c40bd6d4 Author: Matteo Frigo Date: Sun Jun 16 19:13:31 2002 -0400 Added stuff to do commit 6eadf663f6d05339378e742de11484c6d46283a5 Author: Matteo Frigo Date: Sun Jun 16 19:05:58 2002 -0400 Handle dual case R = I + 1 commit f6e99e6f7959f1e91fb66ff344ce372c75567e9b Author: Matteo Frigo Date: Sun Jun 16 18:54:31 2002 -0400 Removed useless flag commit d38b5396624973e2ddc9531587b799eaa13f9daf Author: Matteo Frigo Date: Sun Jun 16 18:30:32 2002 -0400 Removed useless file commit 376194067d9decb0e20df12443a0ac64a16f67a2 Author: Matteo Frigo Date: Sun Jun 16 18:30:18 2002 -0400 More k7 work. Switched to runtime CLASSIC mode. commit 331b9479423ed2ad02e75cb8a759241178460ede Author: Steven G. Johnson Date: Sun Jun 16 17:15:18 2002 -0400 spelling commit c278f9c639c8c850773474634acb51104857ad21 Author: Matteo Frigo Date: Sun Jun 16 15:51:44 2002 -0400 Do not compile if not K7_MODE commit 375f66850303af4e6eb9c7263fe5c5c769876065 Author: Matteo Frigo Date: Sun Jun 16 15:35:02 2002 -0400 Do not require K7 definitions to compile commit 684b95447b420d52337cd499589d31dce430ca4a Author: Matteo Frigo Date: Sun Jun 16 08:05:17 2002 -0400 More k7 stuff commit 535c1c74ca2099099e3edbab3aa7acafa02ea7d6 Author: Matteo Frigo Date: Sat Jun 15 18:30:43 2002 -0400 Try to be compatible with automake-1.6 commit ba06164744e743ea9ea8dabb9359dd0e3b830c9c Author: Matteo Frigo Date: Sat Jun 15 18:23:40 2002 -0400 More merging of Stefan's generator with main genfft branch commit fbe6e3f25188363b31caadf2a598be16eeb2a6a7 Author: Matteo Frigo Date: Sat Jun 15 13:51:39 2002 -0400 Slowly merging genfft-k7 with main genfft branch commit 8a567ee1d079e0cb9ba06e66a3f06d9be0eb34b0 Author: Matteo Frigo Date: Fri Jun 14 21:33:02 2002 -0400 Fixed, really commit 6bd3b52c770b109a025d9982098066f91a84c937 Author: Matteo Frigo Date: Fri Jun 14 21:27:12 2002 -0400 Oops... commit 5b6a71f941de2e28005b7821317ec7032c0b9743 Author: Matteo Frigo Date: Fri Jun 14 21:25:34 2002 -0400 Work properly when $(ALL_CODELETS) = "" commit 07399bfca3c51c6e59dc5f7f9a0f1e7600ab2f41 Author: Matteo Frigo Date: Fri Jun 14 21:11:16 2002 -0400 Fixed k7 build machinery commit 015e0a9b55cfef97ad18e06af166a22d02240d09 Author: Matteo Frigo Date: Fri Jun 14 17:42:35 2002 -0400 More work on k7 stuff commit cf8b11788ed23e9e651cbadf1950302c67102f62 Author: Matteo Frigo Date: Fri Jun 14 15:54:29 2002 -0400 More work on k7 stuff commit 8bd8bb064c77d442eb9da44432a3030b2b48f88a Author: Matteo Frigo Date: Fri Jun 14 14:18:15 2002 -0400 Changed my mind again commit 1cfe47c1940dab749e1c63c15bfde4d0eed29495 Author: Matteo Frigo Date: Fri Jun 14 11:53:09 2002 -0400 Removed some useless stuff. commit 315edea8537b5df64d2eb456cd20f12e16c50544 Author: Matteo Frigo Date: Fri Jun 14 11:01:39 2002 -0400 Hmm... commit fb33fef8ea963d61445cfdef5dcf576cbd616cd8 Author: Matteo Frigo Date: Fri Jun 14 10:28:12 2002 -0400 More work in preparation for k7 stuff commit 1efd1ce4b0d507eab8ff1b618a75bb66303b690d Author: Matteo Frigo Date: Fri Jun 14 07:25:28 2002 -0400 Still preparing to include k7 stuff commit 4c4195fba4ee1d73c35a4d8695a714d788a542cb Author: Matteo Frigo Date: Fri Jun 14 07:06:02 2002 -0400 Create .depend commit 4a55307470a6737b48d2cfb10be690ce60b7ac25 Author: Matteo Frigo Date: Fri Jun 14 06:56:15 2002 -0400 Imported Stefan's K7 generator commit 61a7a73d101594961d9b63fb34fb2340635aaf1b Author: Matteo Frigo Date: Thu Jun 13 15:30:41 2002 -0400 Generator for real->halfcomplex and halfcomplex->real codelets commit d43342dc8c0aa4ac1802eff495ca96ac37c9fb22 Author: Matteo Frigo Date: Thu Jun 13 11:54:02 2002 -0400 Improved hash functions, printers commit b26a2a40553deea06a47abfd3ba1f45d54b28cc2 Author: Matteo Frigo Date: Thu Jun 13 11:17:31 2002 -0400 Only regenerate codlist.c in maintainer mode commit 39f18e53eaf87dc26e7001bccb2839fca11d1bf8 Author: Matteo Frigo Date: Thu Jun 13 11:04:24 2002 -0400 Planner can export solution list commit 94b38d1890179198beac49ead7263d4c0a345a59 Author: Matteo Frigo Date: Thu Jun 13 08:59:53 2002 -0400 Fixed for intel compiler commit bd4fbd5827358275240d991af98b0c477a4158cf Author: Matteo Frigo Date: Thu Jun 13 08:48:51 2002 -0400 Revised strategy for constants in codelets commit 9d1d76e5a32774ac2c7d9c43d2594032aeb20933 Author: Matteo Frigo Date: Thu Jun 13 06:21:31 2002 -0400 Enable score planner in classic mode, naive planner in pro mode. commit 81ae7fad037e7325c65da97a3c842c220ea68a2d Author: Matteo Frigo Date: Wed Jun 12 19:18:18 2002 -0400 Report classic/pro commit 2c6576feda81b17c61be5678e97ca1a4db5f7935 Author: Matteo Frigo Date: Wed Jun 12 19:07:48 2002 -0400 Fixed behavior of buffered solver for large buffers. commit e241b59af22b29dd33ada31e9dcfc93eadfc594f Author: Matteo Frigo Date: Wed Jun 12 18:57:19 2002 -0400 Make assumption COST(vector) = length * COST(scalar) in classic mode. commit 2ccbe15f93c5f1bd3bdf6997f28ef1f10a2a5c76 Author: Matteo Frigo Date: Wed Jun 12 18:19:48 2002 -0400 Revised planner implementation in preparation for wisdom. commit 40ff868c177cd7afc80a4b35f2771aadd59ec6a4 Author: Matteo Frigo Date: Wed Jun 12 08:27:36 2002 -0400 Manually hoist loop invariants. commit fca6f800f5e3a40bf20f80e4a2b3da5fe64da13e Author: Matteo Frigo Date: Wed Jun 12 07:47:41 2002 -0400 Revised loop to compile better with gcc -O commit c3e9fb16552f161f337915be9bba97d50aeca6a6 Author: Matteo Frigo Date: Tue Jun 11 17:24:09 2002 -0400 Changed tensor syntax commit c2fb4345e13f698ea829e790e9e50bf5f218d740 Author: Matteo Frigo Date: Tue Jun 11 16:39:45 2002 -0400 Added stuff to do. commit 79b41ae4b2cf1c9d44e386a0fad68531b6fc78a7 Author: Matteo Frigo Date: Tue Jun 11 16:28:14 2002 -0400 Report classic/pro in version number commit e80c221c61c5f10652272a7b86dc61d461ace6d1 Author: Matteo Frigo Date: Tue Jun 11 14:22:49 2002 -0400 Renamed versions into classic/pro commit 49f5d6fa06030e9ee51346622a68b20e37eff60b Author: Matteo Frigo Date: Tue Jun 11 14:06:06 2002 -0400 Revised planners, estimator commit cd4556586e22a54132fe9b13920474a2b404fe21 Author: Matteo Frigo Date: Tue Jun 11 11:45:41 2002 -0400 I don't know what I am doing. commit 2feca6ebca91d4c762240f32f2381be534adb289 Author: Matteo Frigo Date: Tue Jun 11 10:35:52 2002 -0400 Massive revision of estimator commit 9147712cee87c8b5aff5dc3deebcadf1c159ea90 Author: Matteo Frigo Date: Tue Jun 11 07:32:20 2002 -0400 Many changes commit a68d4a6962649cd7996d2320934764d7490bf685 Author: Matteo Frigo Date: Mon Jun 10 21:35:29 2002 -0400 Keep it simple, stupid. commit 6561b587e1de9ea7d45bd354ad1907512733d3e2 Author: Matteo Frigo Date: Mon Jun 10 19:24:28 2002 -0400 Fixed when #undef PRECOMPUTE_ARRAY_INDICES commit dc412dcd137b131ea524fedf1e7012ad405068bc Author: Matteo Frigo Date: Mon Jun 10 17:58:13 2002 -0400 Minor changes commit 9a633a7a248735746b8e8e4ef7675a2c0b2a8330 Author: Matteo Frigo Date: Mon Jun 10 16:30:37 2002 -0400 Added ct-ditbuf.c, many changes everywhere commit c8406b59941a26cd2145a45edbfa10d136572b94 Author: Matteo Frigo Date: Mon Jun 10 10:55:40 2002 -0400 More name mangling commit 0889cbac352d22d2214500301375f43bf584a6f6 Author: Matteo Frigo Date: Mon Jun 10 10:08:27 2002 -0400 Fixed build system for single/double precision commit c27e0f15eeb5336b8eebc47c215d2ba092e68aa9 Author: Matteo Frigo Date: Mon Jun 10 09:04:21 2002 -0400 Massive renaming to support both single and double precision. (Must recompile everything twice). commit a0b9a19548d7a6278163b36fa1d3088a3ae24cd7 Author: Matteo Frigo Date: Mon Jun 10 06:49:55 2002 -0400 Preliminary crude support for vector transforms in benchmark library. commit ac333b5f92707c9dd867b283d23a53bc64ef69dc Author: Matteo Frigo Date: Sun Jun 9 21:11:51 2002 -0400 Wrong cast commit 669c861be6b66386cd4e8bf2763a8beacebe9bf7 Author: Matteo Frigo Date: Sun Jun 9 16:48:54 2002 -0400 Added things to do. commit 9bfab3a2538b271e95b879402f2269d3cb796144 Author: Matteo Frigo Date: Sun Jun 9 16:07:12 2002 -0400 twlen0: make static commit 89e7b8c3a0488bd401d699312ce92ac4fc7ad872 Author: Matteo Frigo Date: Sun Jun 9 15:34:54 2002 -0400 Nothing commit d8b5a396d1befb9c8709b66e032d5babd304779b Author: Matteo Frigo Date: Sun Jun 9 15:30:13 2002 -0400 Forgot break in switch statement. commit 36c2890f3950c1994c94a98ddf84e218064211d1 Author: Matteo Frigo Date: Sun Jun 9 15:27:24 2002 -0400 Fix for c++ compatibility commit 71882a5547eeb7ca329df8033e7b2eddb759bf76 Author: Matteo Frigo Date: Sun Jun 9 15:16:43 2002 -0400 Added printer, changed everything commit fea5c4c9f8ba921c22c6d7388e5babe9c73b8e26 Author: Matteo Frigo Date: Sun Jun 9 11:37:07 2002 -0400 Removed redundant nop solver commit 382dec87c6ce525bdbf78111112feabc152eb8f7 Author: Matteo Frigo Date: Sun Jun 9 11:06:31 2002 -0400 More things to do commit e471ce8b9ea5726e0a4d2b63770ca8e242327e71 Author: Matteo Frigo Date: Sun Jun 9 11:01:41 2002 -0400 Introduced idea of rank -infinity and associated NOP plans commit 408125e1afdc419d7a914c6475dc663199d1e372 Author: Matteo Frigo Date: Sun Jun 9 08:36:27 2002 -0400 Fixed comment commit eb28a96b485229b8d6883369b0cab54fe3df5b3d Author: Matteo Frigo Date: Sun Jun 9 08:20:13 2002 -0400 Removed useless assertions. commit a05a666ec036474e44f87cecc89a76f13249c4cd Author: Matteo Frigo Date: Sun Jun 9 08:19:26 2002 -0400 Don't malloc(0). commit 53cf44cd85d0b4fe9fd486f24c591d66fd4cb9ff Author: Matteo Frigo Date: Sun Jun 9 08:08:13 2002 -0400 Fixed signed/unsigned puns commit e2aae81e2ec289bb799c02ff72172ffe7f34d076 Author: Matteo Frigo Date: Sun Jun 9 07:52:22 2002 -0400 Added buffered.c commit bb2eb63fb1a1ddf2a8c28a918c2108f9b128053a Author: Matteo Frigo Date: Sat Jun 8 16:57:54 2002 -0400 Fixed printout commit 5e196db13b6e3ea78d999f2a2958c7e65c4e5e68 Author: Matteo Frigo Date: Sat Jun 8 16:42:52 2002 -0400 Fixed comment commit 00c4dc36263864bb6edf511df147475ca8049e35 Author: Matteo Frigo Date: Sat Jun 8 16:40:58 2002 -0400 Added vrank3-transpose, renamed vrank0-transpose -> vrank2-transpose commit 70736b265541c54ab69946ad312e21300b0b2f47 Author: Matteo Frigo Date: Sat Jun 8 15:51:46 2002 -0400 Added vrank0-transpose commit 21fa46aab36c0306bf64d254b8718c8eaf96118c Author: Matteo Frigo Date: Sat Jun 8 15:11:09 2002 -0400 Added planner-score.c commit 3700ef27b625aa5dd4e42e19969b9d5d0e546b39 Author: Matteo Frigo Date: Sat Jun 8 11:10:44 2002 -0400 Added indirect.c commit 92cdbe47f89adcf0d5ae5a1535e171fa58b94f0c Author: Matteo Frigo Date: Sat Jun 8 09:34:58 2002 -0400 dif, ditf solvers commit 9382caabae5da944eb93a194ed963efa5b0dcd2a Author: Matteo Frigo Date: Fri Jun 7 18:07:53 2002 -0400 Implemented rank_geq2. Revised build system commit fd9d18f8d6fbe910aa48673bfcfa7b3e9a0ab7e4 Author: Matteo Frigo Date: Fri Jun 7 07:12:25 2002 -0400 Fixed printout commit 3f7d553cb7059f0f68b15004eb6a62a2a396c092 Author: Matteo Frigo Date: Fri Jun 7 07:07:46 2002 -0400 Added rank0. Revised codelet organization. commit 9b91cf2747d0b49d1220a4e1fc494fe72d2a6f64 Author: Matteo Frigo Date: Thu Jun 6 18:03:17 2002 -0400 Added memoization commit 61d0f601e04b8e1f9ff8fe350c4ef61211be20de Author: Matteo Frigo Date: Thu Jun 6 08:07:33 2002 -0400 Added vecloop commit 5995d0f346270a179f5156150ce1997613636f8f Author: Matteo Frigo Date: Wed Jun 5 19:02:56 2002 -0400 First DIT solver/plan commit 70714a1ca5a962fc4acf3c91b5b1dbb2b5518310 Author: Matteo Frigo Date: Wed Jun 5 16:03:44 2002 -0400 More work on ct commit 479c74ce55bbc9e78c22e4ed8506db9c494871c5 Author: Matteo Frigo Date: Wed Jun 5 11:28:09 2002 -0400 Only use cycle counters commit f397307a4a2ecb675736dbfaee0e3cb117ecafc4 Author: Matteo Frigo Date: Tue Jun 4 20:22:23 2002 -0400 Signed/unsigned fixup commit 26f9cec858cb382ad71e3d025ef4258dffce6e0d Author: Matteo Frigo Date: Tue Jun 4 20:03:56 2002 -0400 New file twiddle.c commit e31060a0f867b05a6cde5ade185fa6edc487c43b Author: Matteo Frigo Date: Tue Jun 4 17:49:39 2002 -0400 Made tensor ranks and vector lengths unsigned. Hopefully fixed all places where it matters. commit 33459884b1d822c57fec05becfccd6cf4a5c1efd Author: Matteo Frigo Date: Tue Jun 4 16:28:58 2002 -0400 System is in working state now (but very incomplete) commit 53e8499d5241b1194dcf9d39bb617b45ba00beed Author: Matteo Frigo Date: Mon Jun 3 18:10:12 2002 -0400 Started implementing planners commit 41ec7720369ea65ef877dacee5b43c73d36e9dc0 Author: Matteo Frigo Date: Mon Jun 3 11:44:18 2002 -0400 Imported libbench from the new benchfft. We will use libbench for benchmarking and testing. commit 936272ab0ecd1ecb585f889ef7d8c996a9e9920c Author: Matteo Frigo Date: Mon Jun 3 09:18:46 2002 -0400 Removed useless rand.c commit c0d9815658256675268110505f21e80b6a850e2c Author: Matteo Frigo Date: Mon Jun 3 08:09:05 2002 -0400 Added timer commit 1135ab2472586a6ddfd23571aa9724686470238c Author: Matteo Frigo Date: Sun Jun 2 21:03:51 2002 -0400 Split codelets into standard and inplace commit 8394a2d62bb726c0af14d1ca1bf721d538bbb6ce Author: Matteo Frigo Date: Sun Jun 2 19:49:03 2002 -0400 Many many changes commit 069ab6b1a318b8d31556dafc6b07953becc026ab Author: Matteo Frigo Date: Sun Jun 2 15:00:11 2002 -0400 Fixed anachronism commit 9a85d188479075e3f6d1de0d33c35008578e5b97 Author: Matteo Frigo Date: Sun Jun 2 14:42:32 2002 -0400 Initial revision