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

added optimized cgemm-kernel for SANDYBRIDGE

This commit is contained in:
wernsaar 2014-06-27 13:40:29 +02:00
parent 578d1b6219
commit 365e8de346
3 changed files with 2287 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
kernel/x86_64/KERNEL.SANDYBRIDGE
View File

@ -21,11 +21,11 @@ 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 = zgemm_kernel_2x4_nehalem.S CGEMMKERNEL = cgemm_kernel_8x2_sandy.S
CGEMMINCOPY = zgemm_ncopy_2.S CGEMMINCOPY = ../generic/zgemm_ncopy_8.c
CGEMMITCOPY = zgemm_tcopy_2.S CGEMMITCOPY = ../generic/zgemm_tcopy_8.c
CGEMMONCOPY = ../generic/zgemm_ncopy_4.c CGEMMONCOPY = ../generic/zgemm_ncopy_2.c
CGEMMOTCOPY = ../generic/zgemm_tcopy_4.c CGEMMOTCOPY = ../generic/zgemm_tcopy_2.c
CGEMMINCOPYOBJ = cgemm_incopy$(TSUFFIX).$(SUFFIX) CGEMMINCOPYOBJ = cgemm_incopy$(TSUFFIX).$(SUFFIX)
CGEMMITCOPYOBJ = cgemm_itcopy$(TSUFFIX).$(SUFFIX) CGEMMITCOPYOBJ = cgemm_itcopy$(TSUFFIX).$(SUFFIX)
CGEMMONCOPYOBJ = cgemm_oncopy$(TSUFFIX).$(SUFFIX) CGEMMONCOPYOBJ = cgemm_oncopy$(TSUFFIX).$(SUFFIX)

2278
kernel/x86_64/cgemm_kernel_8x2_sandy.S Normal file
View File

File diff suppressed because it is too large Load Diff

8
param.h
View File

@ -1111,14 +1111,14 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define SGEMM_DEFAULT_UNROLL_M 16 #define SGEMM_DEFAULT_UNROLL_M 16
#define DGEMM_DEFAULT_UNROLL_M 8 #define DGEMM_DEFAULT_UNROLL_M 8
#define QGEMM_DEFAULT_UNROLL_M 2 #define QGEMM_DEFAULT_UNROLL_M 2
#define CGEMM_DEFAULT_UNROLL_M 2 #define CGEMM_DEFAULT_UNROLL_M 8
#define ZGEMM_DEFAULT_UNROLL_M 4 #define ZGEMM_DEFAULT_UNROLL_M 4
#define XGEMM_DEFAULT_UNROLL_M 1 #define XGEMM_DEFAULT_UNROLL_M 1
#define SGEMM_DEFAULT_UNROLL_N 4 #define SGEMM_DEFAULT_UNROLL_N 4
#define DGEMM_DEFAULT_UNROLL_N 4 #define DGEMM_DEFAULT_UNROLL_N 4
#define QGEMM_DEFAULT_UNROLL_N 2 #define QGEMM_DEFAULT_UNROLL_N 2
#define CGEMM_DEFAULT_UNROLL_N 4 #define CGEMM_DEFAULT_UNROLL_N 2
#define ZGEMM_DEFAULT_UNROLL_N 4 #define ZGEMM_DEFAULT_UNROLL_N 4
#define XGEMM_DEFAULT_UNROLL_N 1 #define XGEMM_DEFAULT_UNROLL_N 1
#endif #endif
@ -1134,7 +1134,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define QGEMM_DEFAULT_P 504 #define QGEMM_DEFAULT_P 504
#define QGEMM_DEFAULT_R qgemm_r #define QGEMM_DEFAULT_R qgemm_r
#define CGEMM_DEFAULT_P 128 #define CGEMM_DEFAULT_P 384
//#define CGEMM_DEFAULT_R cgemm_r //#define CGEMM_DEFAULT_R cgemm_r
#define CGEMM_DEFAULT_R 1024 #define CGEMM_DEFAULT_R 1024
@ -1148,7 +1148,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define SGEMM_DEFAULT_Q 384 #define SGEMM_DEFAULT_Q 384
#define DGEMM_DEFAULT_Q 256 #define DGEMM_DEFAULT_Q 256
#define QGEMM_DEFAULT_Q 128 #define QGEMM_DEFAULT_Q 128
#define CGEMM_DEFAULT_Q 256 #define CGEMM_DEFAULT_Q 192
#define ZGEMM_DEFAULT_Q 192 #define ZGEMM_DEFAULT_Q 192
#define XGEMM_DEFAULT_Q 128 #define XGEMM_DEFAULT_Q 128