|
- /*********************************************************************/
- /* 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. */
- /*********************************************************************/
-
- #include <stdio.h>
- #include <ctype.h>
- #include "common.h"
-
- const static FLOAT dp1 = 1.;
-
- #ifdef CONJ
- #define GEMM_KERNEL GEMM_KERNEL_R
- #define TRMM_KERNEL_N TRMM_KERNEL_RR
- #define TRMM_KERNEL_T TRMM_KERNEL_RC
- #else
- #define GEMM_KERNEL GEMM_KERNEL_N
- #define TRMM_KERNEL_N TRMM_KERNEL_RN
- #define TRMM_KERNEL_T TRMM_KERNEL_RT
- #endif
-
- #if 0
- #undef GEMM_P
- #undef GEMM_Q
- #undef GEMM_R
-
- #define GEMM_P 8
- #define GEMM_Q 12
- #define GEMM_R 16
- #endif
-
- int CNAME(blas_arg_t *args, BLASLONG *range_m, BLASLONG *range_n, FLOAT *sa, FLOAT *sb, BLASLONG dummy) {
-
- BLASLONG m, n, lda, ldb;
- FLOAT *beta, *a, *b;
-
- BLASLONG ls, is, js;
- BLASLONG min_l, min_i, min_j;
- BLASLONG jjs, min_jj;
- #if !((!defined(UPPER) && !defined(TRANSA)) || (defined(UPPER) && defined(TRANSA)))
- BLASLONG start_ls;
- #endif
-
- m = args -> m;
- n = args -> n;
-
- a = (FLOAT *)args -> a;
- b = (FLOAT *)args -> b;
-
- lda = args -> lda;
- ldb = args -> ldb;
-
- beta = (FLOAT *)args -> beta;
-
- if (range_m) {
- BLASLONG m_from = *(((BLASLONG *)range_m) + 0);
- BLASLONG m_to = *(((BLASLONG *)range_m) + 1);
-
- m = m_to - m_from;
-
- b += m_from * COMPSIZE;
- }
-
- if (beta) {
- #ifndef COMPLEX
- if (beta[0] != ONE)
- GEMM_BETA(m, n, 0, beta[0], NULL, 0, NULL, 0, b, ldb);
- if (beta[0] == ZERO) return 0;
- #else
- if ((beta[0] != ONE) || (beta[1] != ZERO))
- GEMM_BETA(m, n, 0, beta[0], beta[1], NULL, 0, NULL, 0, b, ldb);
- if ((beta[0] == ZERO) && (beta[1] == ZERO)) return 0;
- #endif
- }
-
- #if (!defined(UPPER) && !defined(TRANSA)) || (defined(UPPER) && defined(TRANSA))
-
- for(js = 0; js < n; js += GEMM_R){
- min_j = n - js;
- if (min_j > GEMM_R) min_j = GEMM_R;
-
- for(ls = js; ls < js + min_j; ls += GEMM_Q){
- min_l = js + min_j - ls;
- if (min_l > GEMM_Q) min_l = GEMM_Q;
- min_i = m;
- if (min_i > GEMM_P) min_i = GEMM_P;
-
- GEMM_ITCOPY(min_l, min_i, b + (ls * ldb) * COMPSIZE, ldb, sa);
-
- for(jjs = 0; jjs < ls - js; jjs += min_jj){
- min_jj = ls - js - jjs;
- #if defined(SKYLAKEX) || defined(COOPERLAKE)
- /* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */
- if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;
- #else
- if (min_jj > GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3;
- else
- if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
- #endif
- #ifndef TRANSA
- GEMM_ONCOPY(min_l, min_jj, a + (ls + (js + jjs) * lda) * COMPSIZE, lda, sb + min_l * jjs * COMPSIZE);
- #else
- GEMM_OTCOPY(min_l, min_jj, a + ((js + jjs) + ls * lda) * COMPSIZE, lda, sb + min_l * jjs * COMPSIZE);
- #endif
-
- GEMM_KERNEL(min_i, min_jj, min_l, dp1,
- #ifdef COMPLEX
- ZERO,
- #endif
- sa, sb + min_l * jjs * COMPSIZE,
- b + ((js + jjs) * ldb) * COMPSIZE, ldb);
- }
-
- for(jjs = 0; jjs < min_l; jjs += min_jj){
- min_jj = min_l - jjs;
- #if defined(SKYLAKEX) || defined(COOPERLAKE)
- /* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */
- if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;
- #else
- if (min_jj > GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3;
- else
- if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
- #endif
- #ifndef TRANSA
- TRMM_OLNCOPY(min_l, min_jj, a, lda, ls, ls + jjs, sb + min_l * (ls - js + jjs) * COMPSIZE);
- #else
- TRMM_OUTCOPY(min_l, min_jj, a, lda, ls, ls + jjs, sb + min_l * (ls - js + jjs) * COMPSIZE);
- #endif
-
- TRMM_KERNEL_T(min_i, min_jj, min_l, dp1,
- #ifdef COMPLEX
- ZERO,
- #endif
- sa,
- sb + (ls - js + jjs) * min_l * COMPSIZE,
- b + ((ls + jjs) * ldb) * COMPSIZE, ldb, -jjs);
- }
-
- for(is = min_i; is < m; is += GEMM_P){
- min_i = m - is;
- if (min_i > GEMM_P) min_i = GEMM_P;
-
- GEMM_ITCOPY(min_l, min_i, b + (is + ls * ldb) * COMPSIZE, ldb, sa);
-
- GEMM_KERNEL(min_i, ls - js, min_l, dp1,
- #ifdef COMPLEX
- ZERO,
- #endif
- sa, sb,
- b + (is + js * ldb) * COMPSIZE, ldb);
-
- TRMM_KERNEL_T(min_i, min_l, min_l, dp1,
- #ifdef COMPLEX
- ZERO,
- #endif
- sa,
- sb + (ls - js) * min_l * COMPSIZE,
- b + (is + ls * ldb) * COMPSIZE, ldb, 0);
- }
- }
-
-
- for(ls = js + min_j; ls < n; ls += GEMM_Q){
- min_l = n - ls;
- if (min_l > GEMM_Q) min_l = GEMM_Q;
- min_i = m;
- if (min_i > GEMM_P) min_i = GEMM_P;
-
- GEMM_ITCOPY(min_l, min_i, b + (ls * ldb) * COMPSIZE, ldb, sa);
-
- for(jjs = js; jjs < js + min_j; jjs += min_jj){
- min_jj = min_j + js - jjs;
- #if defined(SKYLAKEX) || defined(COOPERLAKE)
- /* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */
- if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;
- #else
- if (min_jj > GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3;
- else
- if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
- #endif
- #ifndef TRANSA
- GEMM_ONCOPY(min_l, min_jj, a + (ls + jjs * lda) * COMPSIZE, lda, sb + min_l * (jjs - js) * COMPSIZE);
- #else
- GEMM_OTCOPY(min_l, min_jj, a + (jjs + ls * lda) * COMPSIZE, lda, sb + min_l * (jjs - js) * COMPSIZE);
- #endif
-
- GEMM_KERNEL(min_i, min_jj, min_l, dp1,
- #ifdef COMPLEX
- ZERO,
- #endif
- sa, sb + min_l * (jjs - js) * COMPSIZE,
- b + (jjs * ldb) * COMPSIZE, ldb);
- }
-
- for(is = min_i; is < m; is += GEMM_P){
- min_i = m - is;
- if (min_i > GEMM_P) min_i = GEMM_P;
-
- GEMM_ITCOPY(min_l, min_i, b + (is + ls * ldb) * COMPSIZE, ldb, sa);
-
- GEMM_KERNEL(min_i, min_j, min_l, dp1,
- #ifdef COMPLEX
- ZERO,
- #endif
- sa, sb, b + (is + js * ldb) * COMPSIZE, ldb);
- }
- }
- }
-
- #else
- for(js = n; js > 0; js -= GEMM_R){
- min_j = js;
- if (min_j > GEMM_R) min_j = GEMM_R;
-
- start_ls = js - min_j;
- while (start_ls + GEMM_Q < js) start_ls += GEMM_Q;
-
- for(ls = start_ls; ls >= js - min_j; ls -= GEMM_Q){
- min_l = js - ls;
- if (min_l > GEMM_Q) min_l = GEMM_Q;
- min_i = m;
- if (min_i > GEMM_P) min_i = GEMM_P;
-
- GEMM_ITCOPY(min_l, min_i, b + (ls * ldb) * COMPSIZE, ldb, sa);
-
- for(jjs = 0; jjs < min_l; jjs += min_jj){
- min_jj = min_l - jjs;
- #if defined(SKYLAKEX) || defined(COOPERLAKE)
- /* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */
- if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;
- #else
- if (min_jj > GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3;
- else
- if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
- #endif
- #ifndef TRANSA
- TRMM_OUNCOPY(min_l, min_jj, a, lda, ls, ls + jjs, sb + min_l * jjs * COMPSIZE);
- #else
- TRMM_OLTCOPY(min_l, min_jj, a, lda, ls, ls + jjs, sb + min_l * jjs * COMPSIZE);
- #endif
-
- TRMM_KERNEL_N(min_i, min_jj, min_l, dp1,
- #ifdef COMPLEX
- ZERO,
- #endif
- sa,
- sb + min_l * jjs * COMPSIZE,
- b + ((ls + jjs) * ldb) * COMPSIZE, ldb, -jjs);
- }
-
- for(jjs = 0; jjs < js - ls - min_l; jjs += min_jj){
- min_jj = js - ls - min_l - jjs;
- #if defined(SKYLAKEX) || defined(COOPERLAKE)
- /* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */
- if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;
- #else
- if (min_jj > GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3;
- else
- if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
- #endif
- #ifndef TRANSA
- GEMM_ONCOPY(min_l, min_jj, a + (ls + (ls + min_l + jjs) * lda) * COMPSIZE, lda,
- sb + min_l * (min_l + jjs) * COMPSIZE);
- #else
- GEMM_OTCOPY(min_l, min_jj, a + ((ls + min_l + jjs) + ls * lda) * COMPSIZE, lda,
- sb + min_l * (min_l + jjs) * COMPSIZE);
- #endif
-
- GEMM_KERNEL(min_i, min_jj, min_l, dp1,
- #ifdef COMPLEX
- ZERO,
- #endif
- sa,
- sb + min_l * (min_l + jjs) * COMPSIZE,
- b + ((ls + min_l + jjs) * ldb) * COMPSIZE, ldb);
- }
-
- for(is = min_i; is < m; is += GEMM_P){
- min_i = m - is;
- if (min_i > GEMM_P) min_i = GEMM_P;
-
- GEMM_ITCOPY(min_l, min_i, b + (is + ls * ldb) * COMPSIZE, ldb, sa);
-
- TRMM_KERNEL_N(min_i, min_l, min_l, dp1,
- #ifdef COMPLEX
- ZERO,
- #endif
- sa,
- sb,
- b + (is + ls * ldb) * COMPSIZE, ldb, 0);
-
- if (js - ls - min_l > 0) {
- GEMM_KERNEL(min_i, js - ls - min_l, min_l, dp1,
- #ifdef COMPLEX
- ZERO,
- #endif
- sa,
- sb + min_l * min_l * COMPSIZE,
- b + (is + (ls + min_l) * ldb) * COMPSIZE, ldb);
- }
- }
- }
-
- for(ls = 0; ls < js - min_j; ls += GEMM_Q){
- min_l = js - min_j - ls;
- if (min_l > GEMM_Q) min_l = GEMM_Q;
- min_i = m;
- if (min_i > GEMM_P) min_i = GEMM_P;
-
- GEMM_ITCOPY(min_l, min_i, b + (ls * ldb) * COMPSIZE, ldb, sa);
-
- for(jjs = js; jjs < js + min_j; jjs += min_jj){
- min_jj = min_j + js - jjs;
- #if defined(SKYLAKEX) || defined(COOPERLAKE)
- /* the current AVX512 s/d/c/z GEMM kernel requires n>=6*GEMM_UNROLL_N to achieve the best performance */
- if (min_jj >= 6*GEMM_UNROLL_N) min_jj = 6*GEMM_UNROLL_N;
- #else
- if (min_jj > GEMM_UNROLL_N*3) min_jj = GEMM_UNROLL_N*3;
- else
- if (min_jj > GEMM_UNROLL_N) min_jj = GEMM_UNROLL_N;
- #endif
- #ifndef TRANSA
- GEMM_ONCOPY(min_l, min_jj, a + (ls + (jjs - min_j) * lda) * COMPSIZE, lda, sb + min_l * (jjs - js) * COMPSIZE);
- #else
- GEMM_OTCOPY(min_l, min_jj, a + ((jjs - min_j) + ls * lda) * COMPSIZE, lda, sb + min_l * (jjs - js) * COMPSIZE);
- #endif
-
- GEMM_KERNEL(min_i, min_jj, min_l, dp1,
- #ifdef COMPLEX
- ZERO,
- #endif
- sa, sb + min_l * (jjs - js) * COMPSIZE,
- b + ((jjs - min_j) * ldb) * COMPSIZE, ldb);
- }
-
- for(is = min_i; is < m; is += GEMM_P){
- min_i = m - is;
- if (min_i > GEMM_P) min_i = GEMM_P;
-
- GEMM_ITCOPY(min_l, min_i, b + (is + ls * ldb) * COMPSIZE, ldb, sa);
-
- GEMM_KERNEL(min_i, min_j, min_l, dp1,
- #ifdef COMPLEX
- ZERO,
- #endif
- sa, sb, b + (is + (js - min_j) * ldb) * COMPSIZE, ldb);
- }
- }
- }
-
- #endif
-
- return 0;
- }
|