Merge pull request #2361 from wjc404/develop

Optimize AVX2 SGEMM & STRMM

Author: martin-frbg

CONTRIBUTORS.md

@@ -179,3 +179,4 @@ In chronological order:
   * [2019-11-12] AVX512 CGEMM & ZGEMM kernels
   * [2019-12-23] optimize AVX2 CGEMM and ZGEMM
   * [2019-12-30] AVX2 CGEMM3M & ZGEMM3M kernels
+  * [2020-01-07] optimize AVX2 SGEMM and STRMM

kernel/x86_64/KERNEL.HASWELL

@@ -31,11 +31,11 @@ DAXPYKERNEL = daxpy.c
 CAXPYKERNEL = caxpy.c
 ZAXPYKERNEL = zaxpy.c
 
-STRMMKERNEL    =  sgemm_kernel_16x4_haswell.S
-SGEMMKERNEL    =  sgemm_kernel_16x4_haswell.S
+STRMMKERNEL    =  sgemm_kernel_8x4_haswell.c
+SGEMMKERNEL    =  sgemm_kernel_8x4_haswell.c
 SGEMM_BETA     =  sgemm_beta_skylakex.c
-SGEMMINCOPY    =  ../generic/gemm_ncopy_16.c
-SGEMMITCOPY    =  ../generic/gemm_tcopy_16.c
+SGEMMINCOPY    =  ../generic/gemm_ncopy_8.c
+SGEMMITCOPY    =  ../generic/gemm_tcopy_8.c
 SGEMMONCOPY    =  sgemm_ncopy_4_skylakex.c
 SGEMMOTCOPY    =  ../generic/gemm_tcopy_4.c
 SGEMMINCOPYOBJ =  sgemm_incopy$(TSUFFIX).$(SUFFIX)

kernel/x86_64/KERNEL.ZEN

@@ -30,10 +30,10 @@ DAXPYKERNEL = daxpy.c
 CAXPYKERNEL = caxpy.c
 ZAXPYKERNEL = zaxpy.c
 
-STRMMKERNEL    =  sgemm_kernel_16x4_haswell.S
-SGEMMKERNEL    =  sgemm_kernel_16x4_haswell.S
-SGEMMINCOPY    =  ../generic/gemm_ncopy_16.c
-SGEMMITCOPY    =  ../generic/gemm_tcopy_16.c
+STRMMKERNEL    =  sgemm_kernel_8x4_haswell.c
+SGEMMKERNEL    =  sgemm_kernel_8x4_haswell.c
+SGEMMINCOPY    =  ../generic/gemm_ncopy_8.c
+SGEMMITCOPY    =  ../generic/gemm_tcopy_8.c
 SGEMMONCOPY    =  ../generic/gemm_ncopy_4.c
 SGEMMOTCOPY    =  ../generic/gemm_tcopy_4.c
 SGEMMINCOPYOBJ =  sgemm_incopy$(TSUFFIX).$(SUFFIX)