/* BLIS An object-based framework for developing high-performance BLAS-like libraries. Copyright (C) 2014, The University of Texas at Austin Copyright (C) 2018 - 2019, Advanced Micro Devices, Inc. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: - Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. - 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. - Neither the name(s) of the copyright holder(s) nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "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 COPYRIGHT HOLDER 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. */ #include "blis.h" #define FUNCPTR_T gemm_fp typedef void (*FUNCPTR_T) ( doff_t diagoffb, pack_t schema_a, pack_t schema_b, dim_t m, dim_t n, dim_t k, void* alpha, void* a, inc_t cs_a, dim_t pd_a, inc_t ps_a, void* b, inc_t rs_b, dim_t pd_b, inc_t ps_b, void* beta, void* c, inc_t rs_c, inc_t cs_c, cntx_t* cntx, rntm_t* rntm, thrinfo_t* thread ); static FUNCPTR_T GENARRAY(ftypes,trmm_rl_ker_var2rr); // // -- Macrokernel functions for round-robin partitioning ----------------------- // void bli_trmm_rl_ker_var2rr ( obj_t* a, obj_t* b, obj_t* c, cntx_t* cntx, rntm_t* rntm, cntl_t* cntl, thrinfo_t* thread ) { num_t dt_exec = bli_obj_exec_dt( c ); doff_t diagoffb = bli_obj_diag_offset( b ); pack_t schema_a = bli_obj_pack_schema( a ); pack_t schema_b = bli_obj_pack_schema( b ); dim_t m = bli_obj_length( c ); dim_t n = bli_obj_width( c ); dim_t k = bli_obj_width( a ); void* buf_a = bli_obj_buffer_at_off( a ); inc_t cs_a = bli_obj_col_stride( a ); dim_t pd_a = bli_obj_panel_dim( a ); inc_t ps_a = bli_obj_panel_stride( a ); void* buf_b = bli_obj_buffer_at_off( b ); inc_t rs_b = bli_obj_row_stride( b ); dim_t pd_b = bli_obj_panel_dim( b ); inc_t ps_b = bli_obj_panel_stride( b ); void* buf_c = bli_obj_buffer_at_off( c ); inc_t rs_c = bli_obj_row_stride( c ); inc_t cs_c = bli_obj_col_stride( c ); obj_t scalar_a; obj_t scalar_b; void* buf_alpha; void* buf_beta; FUNCPTR_T f; // Detach and multiply the scalars attached to A and B. bli_obj_scalar_detach( a, &scalar_a ); bli_obj_scalar_detach( b, &scalar_b ); bli_mulsc( &scalar_a, &scalar_b ); // Grab the addresses of the internal scalar buffers for the scalar // merged above and the scalar attached to C. buf_alpha = bli_obj_internal_scalar_buffer( &scalar_b ); buf_beta = bli_obj_internal_scalar_buffer( c ); // Index into the type combination array to extract the correct // function pointer. f = ftypes[dt_exec]; // Invoke the function. f( diagoffb, schema_a, schema_b, m, n, k, buf_alpha, buf_a, cs_a, pd_a, ps_a, buf_b, rs_b, pd_b, ps_b, buf_beta, buf_c, rs_c, cs_c, cntx, rntm, thread ); } #undef GENTFUNC #define GENTFUNC( ctype, ch, varname ) \ \ void PASTEMAC(ch,varname) \ ( \ doff_t diagoffb, \ pack_t schema_a, \ pack_t schema_b, \ dim_t m, \ dim_t n, \ dim_t k, \ void* alpha, \ void* a, inc_t cs_a, dim_t pd_a, inc_t ps_a, \ void* b, inc_t rs_b, dim_t pd_b, inc_t ps_b, \ void* beta, \ void* c, inc_t rs_c, inc_t cs_c, \ cntx_t* cntx, \ rntm_t* rntm, \ thrinfo_t* thread \ ) \ { \ const num_t dt = PASTEMAC(ch,type); \ \ /* Alias some constants to simpler names. */ \ const dim_t MR = pd_a; \ const dim_t NR = pd_b; \ const dim_t PACKMR = cs_a; \ const dim_t PACKNR = rs_b; \ \ /* Query the context for the micro-kernel address and cast it to its function pointer type. */ \ PASTECH(ch,gemm_ukr_ft) \ gemm_ukr = bli_cntx_get_l3_vir_ukr_dt( dt, BLIS_GEMM_UKR, cntx ); \ \ /* Temporary C buffer for edge cases. Note that the strides of this temporary buffer are set so that they match the storage of the original C matrix. For example, if C is column-stored, ct will be column-stored as well. */ \ ctype ct[ BLIS_STACK_BUF_MAX_SIZE \ / sizeof( ctype ) ] \ __attribute__((aligned(BLIS_STACK_BUF_ALIGN_SIZE))); \ const bool col_pref = bli_cntx_ukr_prefers_cols_dt( dt, BLIS_GEMM_VIR_UKR, cntx ); \ const inc_t rs_ct = ( col_pref ? 1 : NR ); \ const inc_t cs_ct = ( col_pref ? MR : 1 ); \ \ ctype* restrict one = PASTEMAC(ch,1); \ ctype* restrict zero = PASTEMAC(ch,0); \ ctype* restrict a_cast = a; \ ctype* restrict b_cast = b; \ ctype* restrict c_cast = c; \ ctype* restrict alpha_cast = alpha; \ ctype* restrict beta_cast = beta; \ ctype* restrict b1; \ ctype* restrict c1; \ \ doff_t diagoffb_j; \ dim_t k_full; \ dim_t m_iter, m_left; \ dim_t n_iter, n_left; \ dim_t m_cur; \ dim_t n_cur; \ dim_t k_b1121; \ dim_t off_b1121; \ dim_t i, j; \ inc_t rstep_a; \ inc_t cstep_b; \ inc_t rstep_c, cstep_c; \ inc_t istep_a; \ inc_t istep_b; \ inc_t off_scl; \ inc_t ss_b_num; \ inc_t ss_b_den; \ inc_t ps_b_cur; \ inc_t is_b_cur; \ auxinfo_t aux; \ \ /* Assumptions/assertions: rs_a == 1 cs_a == PACKMR pd_a == MR ps_a == stride to next micro-panel of A rs_b == PACKNR cs_b == 1 pd_b == NR ps_b == stride to next micro-panel of B rs_c == (no assumptions) cs_c == (no assumptions) */ \ \ /* Safety trap: Certain indexing within this macro-kernel does not work as intended if both MR and NR are odd. */ \ if ( ( bli_is_odd( PACKMR ) && bli_is_odd( NR ) ) || \ ( bli_is_odd( PACKNR ) && bli_is_odd( MR ) ) ) bli_abort(); \ \ /* If any dimension is zero, return immediately. */ \ if ( bli_zero_dim3( m, n, k ) ) return; \ \ /* Safeguard: If the current panel of B is entirely above the diagonal, it is implicitly zero. So we do nothing. */ \ if ( bli_is_strictly_above_diag_n( diagoffb, k, n ) ) return; \ \ /* Compute k_full. For all trmm, k_full is simply k. This is needed because some parameter combinations of trmm reduce k to advance past zero regions in the triangular matrix, and when computing the imaginary stride of A (the non-triangular matrix), which is used by 4m1/3m1 implementations, we need this unreduced value of k. */ \ k_full = k; \ \ /* Compute indexing scaling factor for for 4m or 3m. This is needed because one of the packing register blocksizes (PACKMR or PACKNR) is used to index into the micro-panels of the non- triangular matrix when computing with a diagonal-intersecting micro-panel of the triangular matrix. In the case of 4m or 3m, real values are stored in both sub-panels, and so the indexing needs to occur in units of real values. The value computed here is divided into the complex pointer offset to cause the pointer to be advanced by the correct value. */ \ if ( bli_is_4mi_packed( schema_b ) || \ bli_is_3mi_packed( schema_b ) || \ bli_is_rih_packed( schema_b ) ) off_scl = 2; \ else off_scl = 1; \ \ /* Compute the storage stride scaling. Usually this is just 1. However, in the case of interleaved 3m, we need to scale the offset by 3/2. And if we are packing real-only, imag-only, or summed-only, we need to scale the computed panel sizes by 1/2 to compensate for the fact that the pointer arithmetic occurs in terms of complex elements rather than real elements. */ \ if ( bli_is_3mi_packed( schema_b ) ) { ss_b_num = 3; ss_b_den = 2; } \ else if ( bli_is_rih_packed( schema_b ) ) { ss_b_num = 1; ss_b_den = 2; } \ else { ss_b_num = 1; ss_b_den = 1; } \ \ /* If there is a zero region above where the diagonal of B intersects the left edge of the panel, adjust the pointer to A and treat this case as if the diagonal offset were zero. Note that we don't need to adjust the pointer to B since packm would have simply skipped over the region that was not stored. */ \ if ( diagoffb < 0 ) \ { \ j = -diagoffb; \ k = k - j; \ diagoffb = 0; \ a_cast = a_cast + ( j * PACKMR ) / off_scl; \ } \ \ /* If there is a zero region to the right of where the diagonal of B intersects the bottom of the panel, shrink it to prevent "no-op" iterations from executing. */ \ if ( diagoffb + k < n ) \ { \ n = diagoffb + k; \ } \ \ /* Clear the temporary C buffer in case it has any infs or NaNs. */ \ PASTEMAC(ch,set0s_mxn)( MR, NR, \ ct, rs_ct, cs_ct ); \ \ /* Compute number of primary and leftover components of the m and n dimensions. */ \ n_iter = n / NR; \ n_left = n % NR; \ \ m_iter = m / MR; \ m_left = m % MR; \ \ if ( n_left ) ++n_iter; \ if ( m_left ) ++m_iter; \ \ /* Determine some increments used to step through A, B, and C. */ \ rstep_a = ps_a; \ \ cstep_b = ps_b; \ \ rstep_c = rs_c * MR; \ cstep_c = cs_c * NR; \ \ istep_a = PACKMR * k_full; \ istep_b = PACKNR * k; \ \ if ( bli_is_odd( istep_a ) ) istep_a += 1; \ if ( bli_is_odd( istep_b ) ) istep_b += 1; \ \ /* Save the pack schemas of A and B to the auxinfo_t object. */ \ bli_auxinfo_set_schema_a( schema_a, &aux ); \ bli_auxinfo_set_schema_b( schema_b, &aux ); \ \ /* Save the imaginary stride of A to the auxinfo_t object. */ \ bli_auxinfo_set_is_a( istep_a, &aux ); \ \ thrinfo_t* caucus = bli_thrinfo_sub_node( thread ); \ \ dim_t jr_nt = bli_thrinfo_n_way( thread ); \ dim_t jr_tid = bli_thrinfo_work_id( thread ); \ dim_t ir_nt = bli_thrinfo_n_way( caucus ); \ dim_t ir_tid = bli_thrinfo_work_id( caucus ); \ \ dim_t jr_start, jr_end; \ dim_t ir_start, ir_end; \ dim_t jr_inc, ir_inc; \ \ /* Note that we partition the 2nd loop into two regions: the rectangular part of B, and the triangular portion. */ \ dim_t n_iter_rct; \ dim_t n_iter_tri; \ \ if ( bli_is_strictly_below_diag_n( diagoffb, m, n ) ) \ { \ /* If the entire panel of B does not intersect the diagonal, there is no triangular region, and therefore we can skip the second set of loops. */ \ n_iter_rct = n_iter; \ n_iter_tri = 0; \ } \ else \ { \ /* If the panel of B does intersect the diagonal, compute the number of iterations in the rectangular region by dividing NR into the diagonal offset. (There should never be any remainder in this division.) The number of iterations in the triangular (or trapezoidal) region is computed as the remaining number of iterations in the n dimension. */ \ n_iter_rct = diagoffb / NR; \ n_iter_tri = n_iter - n_iter_rct; \ } \ \ /* Use round-robin assignment of micropanels to threads in the 2nd and 1st loops for the initial rectangular region of B (if it exists). */ \ bli_thread_range_jrir_rr( thread, n_iter_rct, 1, FALSE, &jr_start, &jr_end, &jr_inc ); \ bli_thread_range_jrir_rr( caucus, m_iter, 1, FALSE, &ir_start, &ir_end, &ir_inc ); \ \ /* Loop over the n dimension (NR columns at a time). */ \ for ( j = jr_start; j < jr_end; j += jr_inc ) \ { \ ctype* restrict a1; \ ctype* restrict c11; \ ctype* restrict b2; \ \ b1 = b_cast + j * cstep_b; \ c1 = c_cast + j * cstep_c; \ \ n_cur = ( bli_is_not_edge_f( j, n_iter, n_left ) ? NR : n_left ); \ \ /* Initialize our next panel of B to be the current panel of B. */ \ b2 = b1; \ \ { \ /* Save the 4m1/3m1 imaginary stride of B to the auxinfo_t object. */ \ bli_auxinfo_set_is_b( istep_b, &aux ); \ \ /* Loop over the m dimension (MR rows at a time). */ \ for ( i = ir_start; i < ir_end; i += ir_inc ) \ { \ ctype* restrict a2; \ \ a1 = a_cast + i * rstep_a; \ c11 = c1 + i * rstep_c; \ \ m_cur = ( bli_is_not_edge_f( i, m_iter, m_left ) ? MR : m_left ); \ \ /* Compute the addresses of the next panels of A and B. */ \ a2 = bli_trmm_get_next_a_upanel( a1, rstep_a, ir_inc ); \ if ( bli_is_last_iter_rr( i, m_iter, ir_tid, ir_nt ) ) \ { \ a2 = a_cast; \ b2 = bli_trmm_get_next_b_upanel( b1, cstep_b, jr_inc ); \ if ( bli_is_last_iter_rr( j, n_iter, jr_tid, jr_nt ) ) \ b2 = b_cast; \ } \ \ /* Save addresses of next panels of A and B to the auxinfo_t object. */ \ bli_auxinfo_set_next_a( a2, &aux ); \ bli_auxinfo_set_next_b( b2, &aux ); \ \ /* Handle interior and edge cases separately. */ \ if ( m_cur == MR && n_cur == NR ) \ { \ /* Invoke the gemm micro-kernel. */ \ gemm_ukr \ ( \ k, \ alpha_cast, \ a1, \ b1, \ one, \ c11, rs_c, cs_c, \ &aux, \ cntx \ ); \ } \ else \ { \ /* Invoke the gemm micro-kernel. */ \ gemm_ukr \ ( \ k, \ alpha_cast, \ a1, \ b1, \ zero, \ ct, rs_ct, cs_ct, \ &aux, \ cntx \ ); \ \ /* Add the result to the edge of C. */ \ PASTEMAC(ch,adds_mxn)( m_cur, n_cur, \ ct, rs_ct, cs_ct, \ c11, rs_c, cs_c ); \ } \ } \ } \ } \ \ /* If there is no triangular region, then we're done. */ \ if ( n_iter_tri == 0 ) return; \ \ /* Use round-robin assignment of micropanels to threads in the 2nd loop for the remaining triangular region of B (if it exists). NOTE: We don't need to call bli_thread_range_jrir*() here since we employ a hack that calls for each thread to execute every iteration of the jr and ir loops but skip all but the pointer increment for iterations that are not assigned to it. */ \ \ /* Advance the starting b1 and c1 pointers to the positions corresponding to the start of the triangular region of B. */ \ jr_start = n_iter_rct; \ b1 = b_cast + jr_start * cstep_b; \ c1 = c_cast + jr_start * cstep_c; \ \ /* Loop over the n dimension (NR columns at a time). */ \ for ( j = jr_start; j < n_iter; ++j ) \ { \ ctype* restrict a1; \ ctype* restrict c11; \ ctype* restrict b2; \ \ diagoffb_j = diagoffb - ( doff_t )j*NR; \ \ /* Determine the offset to the beginning of the panel that was packed so we can index into the corresponding location in A. Then compute the length of that panel. */ \ off_b1121 = bli_max( -diagoffb_j, 0 ); \ k_b1121 = k - off_b1121; \ \ a1 = a_cast; \ c11 = c1; \ \ n_cur = ( bli_is_not_edge_f( j, n_iter, n_left ) ? NR : n_left ); \ \ /* Initialize our next panel of B to be the current panel of B. */ \ b2 = b1; \ \ /* If the current panel of B intersects the diagonal, scale C by beta. If it is strictly below the diagonal, scale by one. This allows the current macro-kernel to work for both trmm and trmm3. */ \ { \ /* Compute the panel stride for the current diagonal- intersecting micro-panel. */ \ is_b_cur = k_b1121 * PACKNR; \ is_b_cur += ( bli_is_odd( is_b_cur ) ? 1 : 0 ); \ ps_b_cur = ( is_b_cur * ss_b_num ) / ss_b_den; \ \ if ( bli_trmm_my_iter( j, thread ) ) { \ \ /* Save the 4m1/3m1 imaginary stride of B to the auxinfo_t object. */ \ bli_auxinfo_set_is_b( is_b_cur, &aux ); \ \ /* Loop over the m dimension (MR rows at a time). */ \ for ( i = 0; i < m_iter; ++i ) \ { \ if ( bli_trmm_my_iter( i, caucus ) ) { \ \ ctype* restrict a1_i; \ ctype* restrict a2; \ \ m_cur = ( bli_is_not_edge_f( i, m_iter, m_left ) ? MR : m_left ); \ \ a1_i = a1 + ( off_b1121 * PACKMR ) / off_scl; \ \ /* Compute the addresses of the next panels of A and B. */ \ a2 = a1; \ if ( bli_is_last_iter_rr( i, m_iter, 0, 1 ) ) \ { \ a2 = a_cast; \ b2 = b1; \ if ( bli_is_last_iter_rr( j, n_iter, jr_tid, jr_nt ) ) \ b2 = b_cast; \ } \ \ /* Save addresses of next panels of A and B to the auxinfo_t object. */ \ bli_auxinfo_set_next_a( a2, &aux ); \ bli_auxinfo_set_next_b( b2, &aux ); \ \ /* Handle interior and edge cases separately. */ \ if ( m_cur == MR && n_cur == NR ) \ { \ /* Invoke the gemm micro-kernel. */ \ gemm_ukr \ ( \ k_b1121, \ alpha_cast, \ a1_i, \ b1, \ beta_cast, \ c11, rs_c, cs_c, \ &aux, \ cntx \ ); \ } \ else \ { \ /* Copy edge elements of C to the temporary buffer. */ \ PASTEMAC(ch,copys_mxn)( m_cur, n_cur, \ c11, rs_c, cs_c, \ ct, rs_ct, cs_ct ); \ \ /* Invoke the gemm micro-kernel. */ \ gemm_ukr \ ( \ k_b1121, \ alpha_cast, \ a1_i, \ b1, \ beta_cast, \ ct, rs_ct, cs_ct, \ &aux, \ cntx \ ); \ \ /* Copy the result to the edge of C. */ \ PASTEMAC(ch,copys_mxn)( m_cur, n_cur, \ ct, rs_ct, cs_ct, \ c11, rs_c, cs_c ); \ } \ } \ \ a1 += rstep_a; \ c11 += rstep_c; \ } \ } \ \ b1 += ps_b_cur; \ } \ \ c1 += cstep_c; \ } \ \ /*PASTEMAC(ch,fprintm)( stdout, "trmm_rl_ker_var2rr: a1", MR, k_b1121, a1, 1, MR, "%4.1f", "" );*/ \ /*PASTEMAC(ch,fprintm)( stdout, "trmm_rl_ker_var2rr: b1", k_b1121, NR, b1_i, NR, 1, "%4.1f", "" );*/ \ } INSERT_GENTFUNC_BASIC0( trmm_rl_ker_var2rr )