/*********************************************************************/ /* Copyright 2009, 2010 The University of Texas at Austin. */ /* All rights reserved. */ /* */ /* Redistribution and use in source and binary forms, with or */ /* without modification, are permitted provided that the following */ /* conditions are met: */ /* */ /* 1. Redistributions of source code must retain the above */ /* copyright notice, this list of conditions and the following */ /* disclaimer. */ /* */ /* 2. Redistributions in binary form must reproduce the above */ /* copyright notice, this list of conditions and the following */ /* disclaimer in the documentation and/or other materials */ /* provided with the distribution. */ /* */ /* THIS SOFTWARE IS PROVIDED BY THE UNIVERSITY OF TEXAS AT */ /* AUSTIN ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, */ /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */ /* DISCLAIMED. IN NO EVENT SHALL THE UNIVERSITY OF TEXAS AT */ /* AUSTIN OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, */ /* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES */ /* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE */ /* GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR */ /* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF */ /* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT */ /* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT */ /* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */ /* POSSIBILITY OF SUCH DAMAGE. */ /* */ /* The views and conclusions contained in the software and */ /* documentation are those of the authors and should not be */ /* interpreted as representing official policies, either expressed */ /* or implied, of The University of Texas at Austin. */ /*********************************************************************/ #define ASSEMBLER #include "common.h" #define M $4 #define A $7 #define LDA $8 #define X $9 #define INCX $10 #define Y $11 #define INCY $5 #define BUFFER $6 #define XX $12 #define YY $13 #define I $14 #define IS $15 #define AO1 $16 #define AO2 $17 #define Y1 $18 #define TEMP $19 #define II INCX #define ALPHA_R $f13 #define ALPHA_I $f14 #define a1 $f0 #define a2 $f1 #define a3 $f2 #define a4 $f3 #define a5 $f4 #define a6 $f5 #define a7 $f6 #define a8 $f7 #define alpha1 $f8 #define alpha2 $f9 #define alpha3 $f10 #define alpha4 $f11 #define x1 $f12 #define x2 $f15 #define x3 $f16 #define x4 $f17 #define xsum1 $f18 #define xsum2 $f19 #define xsum3 $f20 #define xsum4 $f21 #define ysum1 $f22 #define ysum2 $f23 #define ysum3 $f24 #define ysum4 $f25 #ifndef HEMV #define ADD1 NMSUB #define ADD2 MADD #else #define ADD1 MADD #define ADD2 NMSUB #endif PROLOGUE LDARG INCY, 0($sp) LDARG BUFFER, 8($sp) #ifdef __64BIT__ daddiu $sp, $sp, -64 #else daddiu $sp, $sp, -80 #endif SDARG $16, 0($sp) dsll LDA, LDA, ZBASE_SHIFT SDARG $17, 8($sp) dsll INCX, INCX, ZBASE_SHIFT SDARG $18, 16($sp) dsll INCY, INCY, ZBASE_SHIFT SDARG $19, 24($sp) nop sdc1 $f24, 32($sp) sdc1 $f25, 40($sp) #ifndef __64BIT__ sdc1 $f20, 48($sp) sdc1 $f21, 56($sp) sdc1 $f22, 64($sp) sdc1 $f23, 72($sp) #endif blez M, .L999 li IS, 2 * SIZE beq IS, INCX, .L05 move Y1, Y dsra I, M, 2 move XX, X blez I, .L02 move X, BUFFER .align 3 .L01: LD a1, 0 * SIZE(XX) LD a2, 1 * SIZE(XX) daddu XX, XX, INCX LD a3, 0 * SIZE(XX) LD a4, 1 * SIZE(XX) daddu XX, XX, INCX LD a5, 0 * SIZE(XX) LD a6, 1 * SIZE(XX) daddu XX, XX, INCX LD a7, 0 * SIZE(XX) LD a8, 1 * SIZE(XX) daddu XX, XX, INCX ST a1, 0 * SIZE(BUFFER) ST a2, 1 * SIZE(BUFFER) ST a3, 2 * SIZE(BUFFER) ST a4, 3 * SIZE(BUFFER) ST a5, 4 * SIZE(BUFFER) ST a6, 5 * SIZE(BUFFER) ST a7, 6 * SIZE(BUFFER) ST a8, 7 * SIZE(BUFFER) daddiu I, I, -1 bgtz I, .L01 daddiu BUFFER, BUFFER, 8 * SIZE .align 3 .L02: andi I, M, 3 blez I, .L05 NOP .align 3 .L03: LD a1, 0 * SIZE(XX) LD a2, 1 * SIZE(XX) daddu XX, XX, INCX ST a1, 0 * SIZE(BUFFER) ST a2, 1 * SIZE(BUFFER) daddiu I, I, -1 bgtz I, .L03 daddiu BUFFER, BUFFER, 2 * SIZE .align 3 .L05: beq IS, INCY, .L10 daddiu BUFFER, BUFFER, 255 li TEMP, -256 and BUFFER, BUFFER, TEMP dsra I, M, 2 move Y1, BUFFER blez I, .L07 move YY, Y .align 3 .L06: LD a1, 0 * SIZE(YY) LD a2, 1 * SIZE(YY) daddu YY, YY, INCY LD a3, 0 * SIZE(YY) LD a4, 1 * SIZE(YY) daddu YY, YY, INCY LD a5, 0 * SIZE(YY) LD a6, 1 * SIZE(YY) daddu YY, YY, INCY LD a7, 0 * SIZE(YY) LD a8, 1 * SIZE(YY) daddu YY, YY, INCY ST a1, 0 * SIZE(BUFFER) ST a2, 1 * SIZE(BUFFER) ST a3, 2 * SIZE(BUFFER) ST a4, 3 * SIZE(BUFFER) ST a5, 4 * SIZE(BUFFER) ST a6, 5 * SIZE(BUFFER) ST a7, 6 * SIZE(BUFFER) ST a8, 7 * SIZE(BUFFER) daddiu I, I, -1 bgtz I, .L06 daddiu BUFFER, BUFFER, 8 * SIZE .align 3 .L07: andi I, M, 3 blez I, .L10 NOP .align 3 .L08: LD a1, 0 * SIZE(YY) LD a2, 1 * SIZE(YY) daddu YY, YY, INCY ST a1, 0 * SIZE(BUFFER) ST a2, 1 * SIZE(BUFFER) daddiu I, I, -1 bgtz I, .L08 daddiu BUFFER, BUFFER, 2 * SIZE .align 3 .L10: slti TEMP, M, 2 nop bgtz TEMP, .L20 li IS, 0 .align 3 .L11: dsll TEMP, IS, ZBASE_SHIFT nop daddu XX, X, TEMP daddu YY, Y1, TEMP LD alpha1, 0 * SIZE(XX) LD alpha2, 1 * SIZE(XX) LD alpha3, 2 * SIZE(XX) LD alpha4, 3 * SIZE(XX) move AO1, A daddu AO2, A, LDA LD a1, 0 * SIZE(AO1) LD a2, 1 * SIZE(AO1) LD a3, 2 * SIZE(AO1) LD a4, 3 * SIZE(AO1) LD a5, 0 * SIZE(AO2) LD a6, 1 * SIZE(AO2) LD a7, 2 * SIZE(AO2) LD a8, 3 * SIZE(AO2) MUL xsum1, alpha1, a1 daddiu XX, XX, 4 * SIZE MUL xsum2, alpha2, a1 daddiu YY, YY, 4 * SIZE MUL xsum3, alpha1, a3 daddu A, AO2, LDA MUL xsum4, alpha2, a3 daddiu A, A, 4 * SIZE #ifndef HEMV NMSUB xsum1, xsum1, alpha2, a2 MADD xsum2, xsum2, alpha1, a2 #endif NMSUB xsum3, xsum3, alpha2, a4 daddiu AO1, AO1, 4 * SIZE MADD xsum4, xsum4, alpha1, a4 daddiu AO2, AO2, 4 * SIZE MADD xsum1, xsum1, alpha3, a3 MADD xsum2, xsum2, alpha4, a3 MADD xsum3, xsum3, alpha3, a7 MADD xsum4, xsum4, alpha4, a7 ADD1 xsum1, xsum1, alpha4, a4 ADD2 xsum2, xsum2, alpha3, a4 #ifndef HEMV ADD1 xsum3, xsum3, alpha4, a8 ADD2 xsum4, xsum4, alpha3, a8 #endif MOV x1, alpha1 dsubu II, M, IS MOV x2, alpha2 daddiu II, II, - 2 MOV x3, alpha3 dsra I, II, 1 MOV x4, alpha4 nop MUL alpha1, ALPHA_R, alpha1 MUL alpha2, ALPHA_R, alpha2 MUL alpha3, ALPHA_R, alpha3 MUL alpha4, ALPHA_R, alpha4 NMSUB alpha1, alpha1, ALPHA_I, x2 MADD alpha2, alpha2, ALPHA_I, x1 NMSUB alpha3, alpha3, ALPHA_I, x4 MADD alpha4, alpha4, ALPHA_I, x3 blez I, .L15 daddiu I, I, -1 LD x1, 0 * SIZE(XX) LD x2, 1 * SIZE(XX) LD x4, 3 * SIZE(XX) LD a1, 0 * SIZE(AO1) LD a2, 1 * SIZE(AO1) LD a3, 2 * SIZE(AO1) LD a4, 3 * SIZE(AO1) LD a5, 0 * SIZE(AO2) LD a6, 1 * SIZE(AO2) LD a7, 2 * SIZE(AO2) LD a8, 3 * SIZE(AO2) LD ysum1, 0 * SIZE(YY) blez I, .L13 LD ysum2, 1 * SIZE(YY) .align 3 .L12: MADD ysum1, ysum1, alpha1, a1 LD ysum3, 2 * SIZE(YY) MADD ysum2, ysum2, alpha2, a1 LD ysum4, 3 * SIZE(YY) MADD xsum1, xsum1, x1, a1 LD a8, 3 * SIZE(AO2) MADD xsum2, xsum2, x2, a1 LD a1, 4 * SIZE(AO1) MADD ysum3, ysum3, alpha1, a3 LD x3, 2 * SIZE(XX) MADD ysum4, ysum4, alpha2, a3 daddiu I, I, -1 MADD xsum3, xsum3, x1, a5 MADD xsum4, xsum4, x2, a5 NMSUB ysum1, ysum1, alpha2, a2 MADD ysum2, ysum2, alpha1, a2 ADD1 xsum1, xsum1, x2, a2 daddiu AO2, AO2, 4 * SIZE ADD2 xsum2, xsum2, x1, a2 LD a2, 5 * SIZE(AO1) NMSUB ysum3, ysum3, alpha2, a4 MADD ysum4, ysum4, alpha1, a4 ADD1 xsum3, xsum3, x2, a6 LD x2, 5 * SIZE(XX) ADD2 xsum4, xsum4, x1, a6 LD x1, 4 * SIZE(XX) MADD ysum1, ysum1, alpha3, a5 MADD ysum2, ysum2, alpha4, a5 MADD xsum1, xsum1, x3, a3 LD a5, 0 * SIZE(AO2) MADD xsum2, xsum2, x4, a3 LD a3, 6 * SIZE(AO1) MADD ysum3, ysum3, alpha3, a7 MADD ysum4, ysum4, alpha4, a7 MADD xsum3, xsum3, x3, a7 daddiu AO1, AO1, 4 * SIZE MADD xsum4, xsum4, x4, a7 LD a7, 2 * SIZE(AO2) NMSUB ysum1, ysum1, alpha4, a6 daddiu XX, XX, 4 * SIZE MADD ysum2, ysum2, alpha3, a6 LD a6, 1 * SIZE(AO2) ADD1 xsum1, xsum1, x4, a4 daddiu YY, YY, 4 * SIZE ADD2 xsum2, xsum2, x3, a4 LD a4, 3 * SIZE(AO1) NMSUB ysum3, ysum3, alpha4, a8 ST ysum1,-4 * SIZE(YY) MADD ysum4, ysum4, alpha3, a8 ST ysum2,-3 * SIZE(YY) LD ysum1, 0 * SIZE(YY) LD ysum2, 1 * SIZE(YY) ADD1 xsum3, xsum3, x4, a8 LD x4, 3 * SIZE(XX) ADD2 xsum4, xsum4, x3, a8 ST ysum3,-2 * SIZE(YY) bgtz I, .L12 ST ysum4,-1 * SIZE(YY) .align 3 .L13: MADD ysum1, ysum1, alpha1, a1 LD ysum3, 2 * SIZE(YY) MADD ysum2, ysum2, alpha2, a1 LD ysum4, 3 * SIZE(YY) MADD xsum1, xsum1, x1, a1 LD a8, 3 * SIZE(AO2) MADD xsum2, xsum2, x2, a1 LD x3, 2 * SIZE(XX) MADD ysum3, ysum3, alpha1, a3 MADD ysum4, ysum4, alpha2, a3 MADD xsum3, xsum3, x1, a5 MADD xsum4, xsum4, x2, a5 NMSUB ysum1, ysum1, alpha2, a2 MADD ysum2, ysum2, alpha1, a2 ADD1 xsum1, xsum1, x2, a2 ADD2 xsum2, xsum2, x1, a2 NMSUB ysum3, ysum3, alpha2, a4 MADD ysum4, ysum4, alpha1, a4 ADD1 xsum3, xsum3, x2, a6 ADD2 xsum4, xsum4, x1, a6 MADD ysum1, ysum1, alpha3, a5 MADD ysum2, ysum2, alpha4, a5 MADD xsum1, xsum1, x3, a3 MADD xsum2, xsum2, x4, a3 MADD ysum3, ysum3, alpha3, a7 MADD ysum4, ysum4, alpha4, a7 MADD xsum3, xsum3, x3, a7 MADD xsum4, xsum4, x4, a7 NMSUB ysum1, ysum1, alpha4, a6 MADD ysum2, ysum2, alpha3, a6 ADD1 xsum1, xsum1, x4, a4 ADD2 xsum2, xsum2, x3, a4 NMSUB ysum3, ysum3, alpha4, a8 daddiu XX, XX, 4 * SIZE MADD ysum4, ysum4, alpha3, a8 daddiu YY, YY, 4 * SIZE ADD1 xsum3, xsum3, x4, a8 daddiu AO1, AO1, 4 * SIZE ADD2 xsum4, xsum4, x3, a8 daddiu AO2, AO2, 4 * SIZE ST ysum1, -4 * SIZE(YY) ST ysum2, -3 * SIZE(YY) ST ysum3, -2 * SIZE(YY) ST ysum4, -1 * SIZE(YY) .align 3 .L15: andi I, M, 1 NOP blez I, .L16 NOP LD x1, 0 * SIZE(XX) LD x2, 1 * SIZE(XX) LD a1, 0 * SIZE(AO1) LD a2, 1 * SIZE(AO1) LD a3, 2 * SIZE(AO1) LD a4, 3 * SIZE(AO1) LD a5, 0 * SIZE(AO2) LD a6, 1 * SIZE(AO2) LD a7, 2 * SIZE(AO2) LD a8, 3 * SIZE(AO2) LD ysum1, 0 * SIZE(YY) LD ysum2, 1 * SIZE(YY) MADD ysum1, ysum1, alpha1, a1 MADD ysum2, ysum2, alpha2, a1 MADD xsum1, xsum1, x1, a1 MADD xsum2, xsum2, x2, a1 MADD xsum3, xsum3, x1, a5 MADD xsum4, xsum4, x2, a5 NMSUB ysum1, ysum1, alpha2, a2 MADD ysum2, ysum2, alpha1, a2 ADD1 xsum1, xsum1, x2, a2 ADD2 xsum2, xsum2, x1, a2 ADD1 xsum3, xsum3, x2, a6 ADD2 xsum4, xsum4, x1, a6 MADD ysum1, ysum1, alpha3, a5 MADD ysum2, ysum2, alpha4, a5 NMSUB ysum1, ysum1, alpha4, a6 MADD ysum2, ysum2, alpha3, a6 daddiu XX, XX, 2 * SIZE daddiu YY, YY, 2 * SIZE daddiu AO1, AO1, 2 * SIZE daddiu AO2, AO2, 2 * SIZE ST ysum1, -2 * SIZE(YY) ST ysum2, -1 * SIZE(YY) .align 3 .L16: dsll TEMP, IS, ZBASE_SHIFT daddu TEMP, Y1, TEMP LD ysum1, 0 * SIZE(TEMP) LD ysum2, 1 * SIZE(TEMP) LD ysum3, 2 * SIZE(TEMP) LD ysum4, 3 * SIZE(TEMP) MADD ysum1, ysum1, ALPHA_R, xsum1 MADD ysum2, ysum2, ALPHA_I, xsum1 MADD ysum3, ysum3, ALPHA_R, xsum3 MADD ysum4, ysum4, ALPHA_I, xsum3 NMSUB ysum1, ysum1, ALPHA_I, xsum2 MADD ysum2, ysum2, ALPHA_R, xsum2 NMSUB ysum3, ysum3, ALPHA_I, xsum4 MADD ysum4, ysum4, ALPHA_R, xsum4 ST ysum1, 0 * SIZE(TEMP) ST ysum2, 1 * SIZE(TEMP) ST ysum3, 2 * SIZE(TEMP) ST ysum4, 3 * SIZE(TEMP) daddiu TEMP, IS, 4 slt TEMP, M, TEMP beqz TEMP, .L11 daddiu IS, IS, 2 .align 3 .L20: andi TEMP, M, 1 nop blez TEMP, .L900 nop dsll TEMP, IS, ZBASE_SHIFT nop daddu XX, X, TEMP daddu YY, Y1, TEMP LD alpha1, 0 * SIZE(XX) LD alpha2, 1 * SIZE(XX) LD a1, 0 * SIZE(A) LD a2, 1 * SIZE(A) MUL xsum1, alpha1, a1 LD ysum1, 0 * SIZE(YY) MUL xsum2, alpha2, a1 LD ysum2, 1 * SIZE(YY) #ifndef HEMV NMSUB xsum1, xsum1, alpha2, a2 MADD xsum2, xsum2, alpha1, a2 #endif MOV x1, alpha1 MOV x2, alpha2 MUL alpha1, ALPHA_R, alpha1 MUL alpha2, ALPHA_R, alpha2 NMSUB alpha1, alpha1, ALPHA_I, x2 MADD alpha2, alpha2, ALPHA_I, x1 MADD ysum1, ysum1, ALPHA_R, xsum1 MADD ysum2, ysum2, ALPHA_I, xsum1 NMSUB ysum1, ysum1, ALPHA_I, xsum2 MADD ysum2, ysum2, ALPHA_R, xsum2 ST ysum1, 0 * SIZE(YY) ST ysum2, 1 * SIZE(YY) .align 3 .L900: li IS, 2 * SIZE NOP beq INCY, IS, .L999 dsra I, M, 2 blez I, .L905 NOP .align 3 .L902: LD a1, 0 * SIZE(Y1) LD a2, 1 * SIZE(Y1) LD a3, 2 * SIZE(Y1) LD a4, 3 * SIZE(Y1) LD a5, 4 * SIZE(Y1) LD a6, 5 * SIZE(Y1) LD a7, 6 * SIZE(Y1) LD a8, 7 * SIZE(Y1) ST a1, 0 * SIZE(Y) ST a2, 1 * SIZE(Y) daddu Y, Y, INCY ST a3, 0 * SIZE(Y) ST a4, 1 * SIZE(Y) daddu Y, Y, INCY ST a5, 0 * SIZE(Y) ST a6, 1 * SIZE(Y) daddu Y, Y, INCY ST a7, 0 * SIZE(Y) ST a8, 1 * SIZE(Y) daddu Y, Y, INCY daddiu I, I, -1 bgtz I, .L902 daddiu Y1, Y1, 8 * SIZE .align 3 .L905: andi I, M, 3 blez I, .L999 NOP .align 3 .L906: LD a1, 0 * SIZE(Y1) LD a2, 1 * SIZE(Y1) daddiu Y1, Y1, 2 * SIZE ST a1, 0 * SIZE(Y) ST a2, 1 * SIZE(Y) daddiu I, I, -1 bgtz I, .L906 daddu Y, Y, INCY .align 3 .L999: LDARG $16, 0($sp) LDARG $17, 8($sp) LDARG $18, 16($sp) LDARG $19, 24($sp) ldc1 $f24, 32($sp) ldc1 $f25, 40($sp) #ifndef __64BIT__ ldc1 $f20, 48($sp) ldc1 $f21, 56($sp) ldc1 $f22, 64($sp) ldc1 $f23, 72($sp) #endif j $31 #ifdef __64BIT__ daddiu $sp, $sp, 64 #else daddiu $sp, $sp, 80 #endif EPILOGUE