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

added cgemm_kernel for Piledriver

This commit is contained in:
wernsaar 2013-10-31 08:38:17 +01:00
parent 1cf4b974b2
commit e172b70ea2
3 changed files with 1929 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
kernel/x86_64/KERNEL.PILEDRIVER
View File

@ -27,7 +27,7 @@ DGEMMITCOPYOBJ = dgemm_itcopy$(TSUFFIX).$(SUFFIX)
DGEMMONCOPYOBJ = dgemm_oncopy$(TSUFFIX).$(SUFFIX) DGEMMONCOPYOBJ = dgemm_oncopy$(TSUFFIX).$(SUFFIX)
DGEMMOTCOPYOBJ = dgemm_otcopy$(TSUFFIX).$(SUFFIX) DGEMMOTCOPYOBJ = dgemm_otcopy$(TSUFFIX).$(SUFFIX)
CGEMMKERNEL = cgemm_kernel_4x2_bulldozer.S CGEMMKERNEL = cgemm_kernel_4x2_piledriver.S
CGEMMINCOPY = ../generic/zgemm_ncopy_4.c CGEMMINCOPY = ../generic/zgemm_ncopy_4.c
CGEMMITCOPY = ../generic/zgemm_tcopy_4.c CGEMMITCOPY = ../generic/zgemm_tcopy_4.c
CGEMMONCOPY = ../generic/zgemm_ncopy_2.c CGEMMONCOPY = ../generic/zgemm_ncopy_2.c

1920
kernel/x86_64/cgemm_kernel_4x2_piledriver.S Normal file
View File

File diff suppressed because it is too large Load Diff

12
param.h
View File

@ -348,25 +348,29 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#if defined(ARCH_X86_64) #if defined(ARCH_X86_64)
#define SGEMM_DEFAULT_P 768 #define SGEMM_DEFAULT_P 768
#define DGEMM_DEFAULT_P 384 #define DGEMM_DEFAULT_P 384
#define ZGEMM_DEFAULT_P 384
#define CGEMM_DEFAULT_P 768
#else #else
#define SGEMM_DEFAULT_P 448 #define SGEMM_DEFAULT_P 448
#define DGEMM_DEFAULT_P 480 #define DGEMM_DEFAULT_P 480
#define ZGEMM_DEFAULT_P 112
#define CGEMM_DEFAULT_P 224
#endif #endif
#define QGEMM_DEFAULT_P 112 #define QGEMM_DEFAULT_P 112
#define CGEMM_DEFAULT_P 224
#define ZGEMM_DEFAULT_P 112
#define XGEMM_DEFAULT_P 56 #define XGEMM_DEFAULT_P 56
#if defined(ARCH_X86_64) #if defined(ARCH_X86_64)
#define SGEMM_DEFAULT_Q 192 #define SGEMM_DEFAULT_Q 192
#define DGEMM_DEFAULT_Q 168 #define DGEMM_DEFAULT_Q 168
#define ZGEMM_DEFAULT_Q 168
#define CGEMM_DEFAULT_Q 168
#else #else
#define SGEMM_DEFAULT_Q 224 #define SGEMM_DEFAULT_Q 224
#define DGEMM_DEFAULT_Q 224 #define DGEMM_DEFAULT_Q 224
#define ZGEMM_DEFAULT_Q 224
#define CGEMM_DEFAULT_Q 224
#endif #endif
#define QGEMM_DEFAULT_Q 224 #define QGEMM_DEFAULT_Q 224
#define CGEMM_DEFAULT_Q 224
#define ZGEMM_DEFAULT_Q 224
#define XGEMM_DEFAULT_Q 224 #define XGEMM_DEFAULT_Q 224
#define SGEMM_DEFAULT_R 12288 #define SGEMM_DEFAULT_R 12288