Merge pull request #446 from grisuthedragon/cblas_matcopy

Add a CBLAS interface for the BLAS extension s/d/c/z*matcopy routines.
This commit is contained in:
Zhang Xianyi 2014-09-10 16:31:31 +08:00
commit 1cba8e7b11
8 changed files with 160 additions and 5 deletions

18
cblas.h
View File

@ -318,6 +318,24 @@ void cblas_caxpby(OPENBLAS_CONST blasint n, OPENBLAS_CONST float *alpha, OPENBLA
void cblas_zaxpby(OPENBLAS_CONST blasint n, OPENBLAS_CONST double *alpha, OPENBLAS_CONST double *x, OPENBLAS_CONST blasint incx,OPENBLAS_CONST double *beta, double *y, OPENBLAS_CONST blasint incy);
void cblas_somatcopy(OPENBLAS_CONST enum CBLAS_ORDER CORDER, OPENBLAS_CONST enum CBLAS_TRANSPOSE CTRANS, OPENBLAS_CONST blasint crows, OPENBLAS_CONST blasint ccols, OPENBLAS_CONST float calpha, OPENBLAS_CONST float *a,
OPENBLAS_CONST blasint clda, float *b, OPENBLAS_CONST blasint cldb);
void cblas_domatcopy(OPENBLAS_CONST enum CBLAS_ORDER CORDER, OPENBLAS_CONST enum CBLAS_TRANSPOSE CTRANS, OPENBLAS_CONST blasint crows, OPENBLAS_CONST blasint ccols, OPENBLAS_CONST double calpha, OPENBLAS_CONST double *a,
OPENBLAS_CONST blasint clda, double *b, OPENBLAS_CONST blasint cldb);
void cblas_comatcopy(OPENBLAS_CONST enum CBLAS_ORDER CORDER, OPENBLAS_CONST enum CBLAS_TRANSPOSE CTRANS, OPENBLAS_CONST blasint crows, OPENBLAS_CONST blasint ccols, OPENBLAS_CONST float* calpha, OPENBLAS_CONST float* a,
OPENBLAS_CONST blasint clda, float*b, OPENBLAS_CONST blasint cldb);
void cblas_zomatcopy(OPENBLAS_CONST enum CBLAS_ORDER CORDER, OPENBLAS_CONST enum CBLAS_TRANSPOSE CTRANS, OPENBLAS_CONST blasint crows, OPENBLAS_CONST blasint ccols, OPENBLAS_CONST double* calpha, OPENBLAS_CONST double* a,
OPENBLAS_CONST blasint clda, double *b, OPENBLAS_CONST blasint cldb);
void cblas_simatcopy(OPENBLAS_CONST enum CBLAS_ORDER CORDER, OPENBLAS_CONST enum CBLAS_TRANSPOSE CTRANS, OPENBLAS_CONST blasint crows, OPENBLAS_CONST blasint ccols, OPENBLAS_CONST float calpha, float *a,
OPENBLAS_CONST blasint clda, OPENBLAS_CONST blasint cldb);
void cblas_dimatcopy(OPENBLAS_CONST enum CBLAS_ORDER CORDER, OPENBLAS_CONST enum CBLAS_TRANSPOSE CTRANS, OPENBLAS_CONST blasint crows, OPENBLAS_CONST blasint ccols, OPENBLAS_CONST double calpha, double *a,
OPENBLAS_CONST blasint clda, OPENBLAS_CONST blasint cldb);
void cblas_cimatcopy(OPENBLAS_CONST enum CBLAS_ORDER CORDER, OPENBLAS_CONST enum CBLAS_TRANSPOSE CTRANS, OPENBLAS_CONST blasint crows, OPENBLAS_CONST blasint ccols, OPENBLAS_CONST float* calpha, float* a,
OPENBLAS_CONST blasint clda, OPENBLAS_CONST blasint cldb);
void cblas_zimatcopy(OPENBLAS_CONST enum CBLAS_ORDER CORDER, OPENBLAS_CONST enum CBLAS_TRANSPOSE CTRANS, OPENBLAS_CONST blasint crows, OPENBLAS_CONST blasint ccols, OPENBLAS_CONST double* calpha, double* a,
OPENBLAS_CONST blasint clda, OPENBLAS_CONST blasint cldb);
#ifdef __cplusplus
}
#endif /* __cplusplus */

View File

@ -306,7 +306,23 @@ void cblas_caxpby(blasint n, float *alpha, float *x, blasint incx,float *beta, f
void cblas_zaxpby(blasint n, double *alpha, double *x, blasint incx,double *beta, double *y, blasint incy);
void cblas_somatcopy( enum CBLAS_ORDER CORDER, enum CBLAS_TRANSPOSE CTRANS, blasint crows, blasint ccols, float calpha, float *a,
blasint clda, float *b, blasint cldb);
void cblas_domatcopy( enum CBLAS_ORDER CORDER, enum CBLAS_TRANSPOSE CTRANS, blasint crows, blasint ccols, double calpha, double *a,
blasint clda, double *b, blasint cldb);
void cblas_comatcopy( enum CBLAS_ORDER CORDER, enum CBLAS_TRANSPOSE CTRANS, blasint crows, blasint ccols, void* calpha, void* a,
blasint clda, void *b, blasint cldb);
void cblas_zomatcopy( enum CBLAS_ORDER CORDER, enum CBLAS_TRANSPOSE CTRANS, blasint crows, blasint ccols, void* calpha, void* a,
blasint clda, void *b, blasint cldb);
void cblas_simatcopy( enum CBLAS_ORDER CORDER, enum CBLAS_TRANSPOSE CTRANS, blasint crows, blasint ccols, float calpha, float *a,
blasint clda, blasint cldb);
void cblas_dimatcopy( enum CBLAS_ORDER CORDER, enum CBLAS_TRANSPOSE CTRANS, blasint crows, blasint ccols, double calpha, double *a,
blasint clda, blasint cldb);
void cblas_cimatcopy( enum CBLAS_ORDER CORDER, enum CBLAS_TRANSPOSE CTRANS, blasint crows, blasint ccols, float* calpha, float* a,
blasint clda, blasint cldb);
void cblas_zimatcopy( enum CBLAS_ORDER CORDER, enum CBLAS_TRANSPOSE CTRANS, blasint crows, blasint ccols, double* calpha, double* a,
blasint clda, blasint cldb);
#ifdef __cplusplus
}
#endif /* __cplusplus */

View File

@ -52,7 +52,9 @@
cblas_zhpr, cblas_zscal, cblas_zswap, cblas_zsymm, cblas_zsyr2k, cblas_zsyrk,
cblas_ztbmv, cblas_ztbsv, cblas_ztpmv, cblas_ztpsv, cblas_ztrmm, cblas_ztrmv, cblas_ztrsm,
cblas_ztrsv, cblas_cdotc_sub, cblas_cdotu_sub, cblas_zdotc_sub, cblas_zdotu_sub,
cblas_saxpby,cblas_daxpby,cblas_caxpby,cblas_zaxpby
cblas_saxpby,cblas_daxpby,cblas_caxpby,cblas_zaxpby,
cblas_somatcopy, cblas_domatcopy, cblas_comatcopy, cblas_zomatcopy,
cblas_simatcopy, cblas_dimatcopy, cblas_cimatcopy, cblas_zimatcopy
);
@exblasobjs = (

View File

@ -267,7 +267,7 @@ CSBLAS2OBJS = \
CSBLAS3OBJS = \
cblas_sgemm.$(SUFFIX) cblas_ssymm.$(SUFFIX) cblas_strmm.$(SUFFIX) cblas_strsm.$(SUFFIX) \
cblas_ssyrk.$(SUFFIX) cblas_ssyr2k.$(SUFFIX)
cblas_ssyrk.$(SUFFIX) cblas_ssyr2k.$(SUFFIX) cblas_somatcopy.$(SUFFIX) cblas_simatcopy.$(SUFFIX)
CDBLAS1OBJS = \
cblas_idamax.$(SUFFIX) cblas_dasum.$(SUFFIX) cblas_daxpy.$(SUFFIX) \
@ -283,7 +283,7 @@ CDBLAS2OBJS = \
CDBLAS3OBJS += \
cblas_dgemm.$(SUFFIX) cblas_dsymm.$(SUFFIX) cblas_dtrmm.$(SUFFIX) cblas_dtrsm.$(SUFFIX) \
cblas_dsyrk.$(SUFFIX) cblas_dsyr2k.$(SUFFIX)
cblas_dsyrk.$(SUFFIX) cblas_dsyr2k.$(SUFFIX) cblas_domatcopy.$(SUFFIX) cblas_dimatcopy.$(SUFFIX)
CCBLAS1OBJS = \
cblas_icamax.$(SUFFIX) cblas_scasum.$(SUFFIX) cblas_caxpy.$(SUFFIX) \
@ -305,7 +305,9 @@ CCBLAS2OBJS = \
CCBLAS3OBJS = \
cblas_cgemm.$(SUFFIX) cblas_csymm.$(SUFFIX) cblas_ctrmm.$(SUFFIX) cblas_ctrsm.$(SUFFIX) \
cblas_csyrk.$(SUFFIX) cblas_csyr2k.$(SUFFIX) \
cblas_chemm.$(SUFFIX) cblas_cherk.$(SUFFIX) cblas_cher2k.$(SUFFIX)
cblas_chemm.$(SUFFIX) cblas_cherk.$(SUFFIX) cblas_cher2k.$(SUFFIX) \
cblas_comatcopy.$(SUFFIX) cblas_cimatcopy.$(SUFFIX)
CZBLAS1OBJS = \
cblas_izamax.$(SUFFIX) cblas_dzasum.$(SUFFIX) cblas_zaxpy.$(SUFFIX) \
@ -327,7 +329,9 @@ CZBLAS2OBJS = \
CZBLAS3OBJS = \
cblas_zgemm.$(SUFFIX) cblas_zsymm.$(SUFFIX) cblas_ztrmm.$(SUFFIX) cblas_ztrsm.$(SUFFIX) \
cblas_zsyrk.$(SUFFIX) cblas_zsyr2k.$(SUFFIX) \
cblas_zhemm.$(SUFFIX) cblas_zherk.$(SUFFIX) cblas_zher2k.$(SUFFIX)
cblas_zhemm.$(SUFFIX) cblas_zherk.$(SUFFIX) cblas_zher2k.$(SUFFIX)\
cblas_zomatcopy.$(SUFFIX) cblas_zimatcopy.$(SUFFIX)
ifndef NO_CBLAS
@ -2035,25 +2039,49 @@ cblas_caxpby.$(SUFFIX) cblas_caxpby.$(PSUFFIX) : zaxpby.c
domatcopy.$(SUFFIX) domatcopy.$(PSUFFIX) : omatcopy.c
$(CC) -c $(CFLAGS) $< -o $(@F)
cblas_domatcopy.$(SUFFIX) cblas_domatcopy.$(PSUFFIX) : omatcopy.c
$(CC) -c $(CFLAGS) -DCBLAS $< -o $(@F)
somatcopy.$(SUFFIX) somatcopy.$(PSUFFIX) : omatcopy.c
$(CC) -c $(CFLAGS) $< -o $(@F)
cblas_somatcopy.$(SUFFIX) cblas_somatcopy.$(PSUFFIX) : omatcopy.c
$(CC) -c $(CFLAGS) -DCBLAS $< -o $(@F)
comatcopy.$(SUFFIX) comatcopy.$(PSUFFIX) : zomatcopy.c
$(CC) -c $(CFLAGS) $< -o $(@F)
cblas_comatcopy.$(SUFFIX) cblas_comatcopy.$(PSUFFIX) : zomatcopy.c
$(CC) -c $(CFLAGS) -DCBLAS $< -o $(@F)
zomatcopy.$(SUFFIX) zomatcopy.$(PSUFFIX) : zomatcopy.c
$(CC) -c $(CFLAGS) $< -o $(@F)
cblas_zomatcopy.$(SUFFIX) cblas_zomatcopy.$(PSUFFIX) : zomatcopy.c
$(CC) -c $(CFLAGS) -DCBLAS $< -o $(@F)
dimatcopy.$(SUFFIX) dimatcopy.$(PSUFFIX) : imatcopy.c
$(CC) -c $(CFLAGS) $< -o $(@F)
cblas_dimatcopy.$(SUFFIX) cblas_dimatcopy.$(PSUFFIX) : imatcopy.c
$(CC) -c $(CFLAGS) -DCBLAS $< -o $(@F)
simatcopy.$(SUFFIX) simatcopy.$(PSUFFIX) : imatcopy.c
$(CC) -c $(CFLAGS) $< -o $(@F)
cblas_simatcopy.$(SUFFIX) cblas_simatcopy.$(PSUFFIX) : imatcopy.c
$(CC) -c $(CFLAGS) -DCBLAS $< -o $(@F)
cimatcopy.$(SUFFIX) cimatcopy.$(PSUFFIX) : zimatcopy.c
$(CC) -c $(CFLAGS) $< -o $(@F)
cblas_cimatcopy.$(SUFFIX) cblas_cimatcopy.$(PSUFFIX) : zimatcopy.c
$(CC) -c $(CFLAGS) -DCBLAS $< -o $(@F)
zimatcopy.$(SUFFIX) zimatcopy.$(PSUFFIX) : zimatcopy.c
$(CC) -c $(CFLAGS) $< -o $(@F)
cblas_zimatcopy.$(SUFFIX) cblas_zimatcopy.$(PSUFFIX) : zimatcopy.c
$(CC) -c $(CFLAGS) -DCBLAS $< -o $(@F)

View File

@ -50,6 +50,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#undef malloc
#undef free
#ifndef CBLAS
void NAME( char* ORDER, char* TRANS, blasint *rows, blasint *cols, FLOAT *alpha, FLOAT *a, blasint *lda, blasint *ldb)
{
@ -71,6 +72,28 @@ void NAME( char* ORDER, char* TRANS, blasint *rows, blasint *cols, FLOAT *alpha,
if ( Trans == 'R' ) trans = BlasNoTrans;
if ( Trans == 'T' ) trans = BlasTrans;
if ( Trans == 'C' ) trans = BlasTrans;
#else
void CNAME( enum CBLAS_ORDER CORDER, enum CBLAS_TRANSPOSE CTRANS, blasint crows, blasint ccols, FLOAT calpha, FLOAT *a, blasint clda, blasint cldb)
{
char Order, Trans;
int order=-1,trans=-1;
blasint info = -1;
FLOAT *b;
size_t msize;
blasint *lda, *ldb, *rows, *cols;
FLOAT *alpha;
if ( CORDER == CblasColMajor) order = BlasColMajor;
if ( CORDER == CblasRowMajor) order = BlasRowMajor;
if ( CTRANS == CblasNoTrans || CTRANS == CblasConjNoTrans) trans = BlasNoTrans;
if ( CTRANS == CblasTrans || CTRANS == CblasConjTrans ) trans = BlasTrans;
rows = &crows;
cols = &ccols;
alpha = &calpha;
lda = &clda;
ldb = &cldb;
#endif
if ( order == BlasColMajor)
{

View File

@ -47,6 +47,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define BlasNoTrans 0
#define BlasTrans 1
#ifndef CBLAS
void NAME( char* ORDER, char* TRANS, blasint *rows, blasint *cols, FLOAT *alpha, FLOAT *a, blasint *lda, FLOAT *b, blasint *ldb)
{
@ -66,7 +67,27 @@ void NAME( char* ORDER, char* TRANS, blasint *rows, blasint *cols, FLOAT *alpha,
if ( Trans == 'R' ) trans = BlasNoTrans;
if ( Trans == 'T' ) trans = BlasTrans;
if ( Trans == 'C' ) trans = BlasTrans;
#else
void CNAME(enum CBLAS_ORDER CORDER, enum CBLAS_TRANSPOSE CTRANS, blasint crows, blasint ccols, FLOAT calpha, FLOAT *a, blasint clda, FLOAT *b, blasint cldb)
{
blasint *rows, *cols, *lda, *ldb;
FLOAT *alpha;
int order=-1,trans=-1;
blasint info = -1;
if ( CORDER == CblasColMajor ) order = BlasColMajor;
if ( CORDER == CblasRowMajor ) order = BlasRowMajor;
if ( CTRANS == CblasNoTrans || CTRANS == CblasConjNoTrans ) trans = BlasNoTrans;
if ( CTRANS == CblasTrans || CTRANS == CblasConjTrans ) trans = BlasTrans;
rows = &crows;
cols = &ccols;
lda = &clda;
ldb = &cldb;
alpha = &calpha;
#endif
if ( order == BlasColMajor)
{
if ( trans == BlasNoTrans && *ldb < *rows ) info = 9;

View File

@ -49,6 +49,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define BlasTransConj 2
#define BlasConj 3
#ifndef CBLAS
void NAME( char* ORDER, char* TRANS, blasint *rows, blasint *cols, FLOAT *alpha, FLOAT *a, blasint *lda, blasint *ldb)
{
@ -71,6 +73,30 @@ void NAME( char* ORDER, char* TRANS, blasint *rows, blasint *cols, FLOAT *alpha,
if ( Trans == 'C' ) trans = BlasTransConj;
if ( Trans == 'R' ) trans = BlasConj;
#else
void CNAME( enum CBLAS_ORDER CORDER, enum CBLAS_TRANSPOSE CTRANS, blasint crows, blasint ccols, FLOAT *alpha, FLOAT *a, blasint clda, blasint cldb)
{
blasint *rows, *cols, *lda, *ldb;
int order=-1,trans=-1;
blasint info = -1;
FLOAT *b;
size_t msize;
if ( CORDER == CblasColMajor ) order = BlasColMajor;
if ( CORDER == CblasRowMajor ) order = BlasRowMajor;
if ( CTRANS == CblasNoTrans) trans = BlasNoTrans;
if ( CTRANS == CblasConjNoTrans ) trans = BlasConj;
if ( CTRANS == CblasTrans) trans = BlasTrans;
if ( CTRANS == CblasConjTrans) trans = BlasTransConj;
rows = &crows;
cols = &ccols;
lda = &clda;
ldb = &cldb;
#endif
if ( order == BlasColMajor)
{
if ( trans == BlasNoTrans && *ldb < *rows ) info = 9;

View File

@ -49,6 +49,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define BlasTransConj 2
#define BlasConj 3
#ifndef CBLAS
void NAME( char* ORDER, char* TRANS, blasint *rows, blasint *cols, FLOAT *alpha, FLOAT *a, blasint *lda, FLOAT *b, blasint *ldb)
{
@ -69,6 +70,26 @@ void NAME( char* ORDER, char* TRANS, blasint *rows, blasint *cols, FLOAT *alpha,
if ( Trans == 'C' ) trans = BlasTransConj;
if ( Trans == 'R' ) trans = BlasConj;
#else
void CNAME(enum CBLAS_ORDER CORDER, enum CBLAS_TRANSPOSE CTRANS, blasint crows, blasint ccols, FLOAT *alpha, FLOAT *a, blasint clda, FLOAT*b, blasint cldb)
{
blasint *rows, *cols, *lda, *ldb;
int order=-1,trans=-1;
blasint info = -1;
if ( CORDER == CblasColMajor ) order = BlasColMajor;
if ( CORDER == CblasRowMajor ) order = BlasRowMajor;
if ( CTRANS == CblasNoTrans) trans = BlasNoTrans;
if ( CTRANS == CblasConjNoTrans ) trans = BlasConj;
if ( CTRANS == CblasTrans) trans = BlasTrans;
if ( CTRANS == CblasConjTrans) trans = BlasTransConj;
rows = &crows;
cols = &ccols;
lda = &clda;
ldb = &cldb;
#endif
if ( order == BlasColMajor)
{
if ( trans == BlasNoTrans && *ldb < *rows ) info = 9;