floraachy
/
OpenBLAS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Small Matrix: skip compile in unimplemented data type

This commit is contained in:
Wangyang Guo 2021-08-05 05:46:13 +00:00
parent 44d0032f3b
commit c17d6dacb2
1 changed files with 8 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
interface/gemm.c
View File

@ -105,8 +105,13 @@ static int (*gemm[])(blas_arg_t *, BLASLONG *, BLASLONG *, IFLOAT *, IFLOAT *, B
#endif #endif
}; };
#ifndef GEMM3M #if defined(SMALL_MATRIX_OPT) && !defined(GEMM3M) && !defined(XDOUBLE) && !defined(BFLOAT16)
#ifdef SMALL_MATRIX_OPT #define USE_SMALL_MATRIX_OPT 1
#else
#define USE_SMALL_MATRIX_OPT 0
#endif
#if USE_SMALL_MATRIX_OPT
#ifndef DYNAMIC_ARCH #ifndef DYNAMIC_ARCH
#define SMALL_KERNEL_ADDR(table, idx) ((void *)(table[idx])) #define SMALL_KERNEL_ADDR(table, idx) ((void *)(table[idx]))
#else #else
@ -148,7 +153,6 @@ static size_t zgemm_small_kernel_b0[] = {
#define ZGEMM_SMALL_KERNEL_B0(idx) (int (*)(BLASLONG, BLASLONG, BLASLONG, FLOAT *, BLASLONG, FLOAT , FLOAT, FLOAT *, BLASLONG, FLOAT *, BLASLONG)) SMALL_KERNEL_ADDR(zgemm_small_kernel_b0, (idx)) #define ZGEMM_SMALL_KERNEL_B0(idx) (int (*)(BLASLONG, BLASLONG, BLASLONG, FLOAT *, BLASLONG, FLOAT , FLOAT, FLOAT *, BLASLONG, FLOAT *, BLASLONG)) SMALL_KERNEL_ADDR(zgemm_small_kernel_b0, (idx))
#endif #endif
#endif #endif
#endif
#ifndef CBLAS #ifndef CBLAS
@ -462,8 +466,7 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE TransA, enum CBLAS_TRANS
FUNCTION_PROFILE_START(); FUNCTION_PROFILE_START();
#ifndef GEMM3M #if USE_SMALL_MATRIX_OPT
#ifdef SMALL_MATRIX_OPT
#if !defined(COMPLEX) #if !defined(COMPLEX)
if(GEMM_SMALL_MATRIX_PERMIT(transa, transb, args.m, args.n, args.k, *(FLOAT *)(args.alpha), *(FLOAT *)(args.beta))){ if(GEMM_SMALL_MATRIX_PERMIT(transa, transb, args.m, args.n, args.k, *(FLOAT *)(args.alpha), *(FLOAT *)(args.beta))){
if(*(FLOAT *)(args.beta) == 0.0){ if(*(FLOAT *)(args.beta) == 0.0){
@ -483,7 +486,6 @@ void CNAME(enum CBLAS_ORDER order, enum CBLAS_TRANSPOSE TransA, enum CBLAS_TRANS
return; return;
} }
#endif #endif
#endif
#endif #endif
buffer = (XFLOAT *)blas_memory_alloc(0); buffer = (XFLOAT *)blas_memory_alloc(0);