replaced sgemm and cgemm kernels because lapack bugs
This commit is contained in:
parent
2424af62fd
commit
5f3b68b4d4
|
@ -1,34 +1,35 @@
|
||||||
SGEMMKERNEL = sgemm_kernel_8x8_sandy.S
|
SGEMMKERNEL = gemm_kernel_4x8_nehalem.S
|
||||||
SGEMMINCOPY =
|
SGEMMINCOPY = gemm_ncopy_4.S
|
||||||
SGEMMITCOPY =
|
SGEMMITCOPY = gemm_tcopy_4.S
|
||||||
SGEMMONCOPY = ../generic/gemm_ncopy_8.c
|
SGEMMONCOPY = ../generic/gemm_ncopy_8.c
|
||||||
SGEMMOTCOPY = ../generic/gemm_tcopy_8.c
|
SGEMMOTCOPY = ../generic/gemm_tcopy_8.c
|
||||||
SGEMMINCOPYOBJ =
|
SGEMMINCOPYOBJ = sgemm_incopy$(TSUFFIX).$(SUFFIX)
|
||||||
SGEMMITCOPYOBJ =
|
SGEMMITCOPYOBJ = sgemm_itcopy$(TSUFFIX).$(SUFFIX)
|
||||||
SGEMMONCOPYOBJ = sgemm_oncopy$(TSUFFIX).$(SUFFIX)
|
SGEMMONCOPYOBJ = sgemm_oncopy$(TSUFFIX).$(SUFFIX)
|
||||||
SGEMMOTCOPYOBJ = sgemm_otcopy$(TSUFFIX).$(SUFFIX)
|
SGEMMOTCOPYOBJ = sgemm_otcopy$(TSUFFIX).$(SUFFIX)
|
||||||
|
|
||||||
|
|
||||||
DGEMMKERNEL = dgemm_kernel_4x8_sandy.S
|
DGEMMKERNEL = dgemm_kernel_4x8_sandy.S
|
||||||
DGEMMINCOPY = ../generic/gemm_ncopy_8.c
|
DGEMMINCOPY = ../generic/gemm_ncopy_8.c
|
||||||
DGEMMITCOPY = ../generic/gemm_tcopy_8.c
|
DGEMMITCOPY = ../generic/gemm_tcopy_8.c
|
||||||
#DGEMMONCOPY = gemm_ncopy_4.S
|
|
||||||
DGEMMONCOPY = ../generic/gemm_ncopy_4.c
|
DGEMMONCOPY = ../generic/gemm_ncopy_4.c
|
||||||
DGEMMOTCOPY = ../generic/gemm_tcopy_4.c
|
DGEMMOTCOPY = ../generic/gemm_tcopy_4.c
|
||||||
#DGEMMOTCOPY = gemm_tcopy_4.S
|
|
||||||
DGEMMINCOPYOBJ = dgemm_incopy$(TSUFFIX).$(SUFFIX)
|
DGEMMINCOPYOBJ = dgemm_incopy$(TSUFFIX).$(SUFFIX)
|
||||||
DGEMMITCOPYOBJ = dgemm_itcopy$(TSUFFIX).$(SUFFIX)
|
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_4x8_sandy.S
|
CGEMMKERNEL = zgemm_kernel_2x4_nehalem.S
|
||||||
CGEMMINCOPY = ../generic/zgemm_ncopy_8_sandy.c
|
CGEMMINCOPY = zgemm_ncopy_2.S
|
||||||
CGEMMITCOPY = ../generic/zgemm_tcopy_8_sandy.c
|
CGEMMITCOPY = zgemm_tcopy_2.S
|
||||||
CGEMMONCOPY = ../generic/zgemm_ncopy_4_sandy.c
|
CGEMMONCOPY = ../generic/zgemm_ncopy_4.c
|
||||||
CGEMMOTCOPY = ../generic/zgemm_tcopy_4_sandy.c
|
CGEMMOTCOPY = ../generic/zgemm_tcopy_4.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)
|
||||||
CGEMMOTCOPYOBJ = cgemm_otcopy$(TSUFFIX).$(SUFFIX)
|
CGEMMOTCOPYOBJ = cgemm_otcopy$(TSUFFIX).$(SUFFIX)
|
||||||
#ZGEMMKERNEL = zgemm_kernel_1x4_nehalem.S
|
|
||||||
|
|
||||||
ZGEMMKERNEL = zgemm_kernel_4x4_sandy.S
|
ZGEMMKERNEL = zgemm_kernel_4x4_sandy.S
|
||||||
ZGEMMINCOPY =
|
ZGEMMINCOPY =
|
||||||
ZGEMMITCOPY =
|
ZGEMMITCOPY =
|
||||||
|
@ -58,6 +59,7 @@ ZGEMMOTCOPYOBJ = zgemm_otcopy$(TSUFFIX).$(SUFFIX)
|
||||||
#ZTRSMKERNEL_LT = ztrsm_kernel_LT_1x4_nehalem.S
|
#ZTRSMKERNEL_LT = ztrsm_kernel_LT_1x4_nehalem.S
|
||||||
#ZTRSMKERNEL_RN = ztrsm_kernel_LT_1x4_nehalem.S
|
#ZTRSMKERNEL_RN = ztrsm_kernel_LT_1x4_nehalem.S
|
||||||
#ZTRSMKERNEL_RT = ztrsm_kernel_RT_1x4_nehalem.S
|
#ZTRSMKERNEL_RT = ztrsm_kernel_RT_1x4_nehalem.S
|
||||||
|
|
||||||
STRSMKERNEL_LN = ../generic/trsm_kernel_LN.c
|
STRSMKERNEL_LN = ../generic/trsm_kernel_LN.c
|
||||||
STRSMKERNEL_LT = ../generic/trsm_kernel_LT.c
|
STRSMKERNEL_LT = ../generic/trsm_kernel_LT.c
|
||||||
STRSMKERNEL_RN = ../generic/trsm_kernel_RN.c
|
STRSMKERNEL_RN = ../generic/trsm_kernel_RN.c
|
||||||
|
|
4
param.h
4
param.h
|
@ -1104,10 +1104,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
||||||
#define ZGEMM_DEFAULT_UNROLL_N 2
|
#define ZGEMM_DEFAULT_UNROLL_N 2
|
||||||
#define XGEMM_DEFAULT_UNROLL_N 1
|
#define XGEMM_DEFAULT_UNROLL_N 1
|
||||||
#else
|
#else
|
||||||
#define SGEMM_DEFAULT_UNROLL_M 8
|
#define SGEMM_DEFAULT_UNROLL_M 4
|
||||||
#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 8
|
#define CGEMM_DEFAULT_UNROLL_M 2
|
||||||
#define ZGEMM_DEFAULT_UNROLL_M 4
|
#define ZGEMM_DEFAULT_UNROLL_M 4
|
||||||
#define XGEMM_DEFAULT_UNROLL_M 1
|
#define XGEMM_DEFAULT_UNROLL_M 1
|
||||||
|
|
||||||
|
|
Loading…
Reference in New Issue