Use latest non-SVE kernels in ARMV8SVE

These are generally better and, in some cases, include threading which helps in the cores we're targeting here.
This commit is contained in:
Chris Sidebottom 2023-07-25 14:05:17 +01:00
parent 7976deff80
commit aea2a4622b
2 changed files with 25 additions and 67 deletions

View File

@ -57,7 +57,7 @@ CAMAXKERNEL = zamax.S
ZAMAXKERNEL = zamax.S ZAMAXKERNEL = zamax.S
SAXPYKERNEL = axpy.S SAXPYKERNEL = axpy.S
DAXPYKERNEL = axpy.S DAXPYKERNEL = daxpy_thunderx2t99.S
CAXPYKERNEL = zaxpy.S CAXPYKERNEL = zaxpy.S
ZAXPYKERNEL = zaxpy.S ZAXPYKERNEL = zaxpy.S
@ -81,45 +81,35 @@ DGEMVTKERNEL = gemv_t.S
CGEMVTKERNEL = zgemv_t.S CGEMVTKERNEL = zgemv_t.S
ZGEMVTKERNEL = zgemv_t.S ZGEMVTKERNEL = zgemv_t.S
SASUMKERNEL = sasum_thunderx2t99.c
DASUMKERNEL = dasum_thunderx2t99.c
CASUMKERNEL = casum_thunderx2t99.c
ZASUMKERNEL = zasum_thunderx2t99.c
SASUMKERNEL = asum.S SCOPYKERNEL = copy_thunderx2t99.c
DASUMKERNEL = asum.S DCOPYKERNEL = copy_thunderx2t99.c
CASUMKERNEL = casum.S CCOPYKERNEL = copy_thunderx2t99.c
ZASUMKERNEL = zasum.S ZCOPYKERNEL = copy_thunderx2t99.c
SCOPYKERNEL = copy.S SSWAPKERNEL = swap_thunderx2t99.S
DCOPYKERNEL = copy.S DSWAPKERNEL = swap_thunderx2t99.S
CCOPYKERNEL = copy.S CSWAPKERNEL = swap_thunderx2t99.S
ZCOPYKERNEL = copy.S ZSWAPKERNEL = swap_thunderx2t99.S
SSWAPKERNEL = swap.S ISAMAXKERNEL = iamax_thunderx2t99.c
DSWAPKERNEL = swap.S IDAMAXKERNEL = iamax_thunderx2t99.c
CSWAPKERNEL = swap.S ICAMAXKERNEL = izamax_thunderx2t99.c
ZSWAPKERNEL = swap.S IZAMAXKERNEL = izamax_thunderx2t99.c
ISAMAXKERNEL = iamax.S SNRM2KERNEL = scnrm2_thunderx2t99.c
IDAMAXKERNEL = iamax.S DNRM2KERNEL = dznrm2_thunderx2t99.c
ICAMAXKERNEL = izamax.S CNRM2KERNEL = scnrm2_thunderx2t99.c
IZAMAXKERNEL = izamax.S ZNRM2KERNEL = dznrm2_thunderx2t99.c
SNRM2KERNEL = nrm2.S DDOTKERNEL = dot.c
DNRM2KERNEL = nrm2.S SDOTKERNEL = dot.c
CNRM2KERNEL = znrm2.S CDOTKERNEL = zdot_thunderx2t99.c
ZNRM2KERNEL = znrm2.S ZDOTKERNEL = zdot_thunderx2t99.c
DDOTKERNEL = dot.S
ifneq ($(C_COMPILER), PGI)
SDOTKERNEL = ../generic/dot.c
else
SDOTKERNEL = dot.S
endif
ifneq ($(C_COMPILER), PGI)
CDOTKERNEL = zdot.S
ZDOTKERNEL = zdot.S
else
CDOTKERNEL = ../arm/zdot.c
ZDOTKERNEL = ../arm/zdot.c
endif
DSDOTKERNEL = dot.S DSDOTKERNEL = dot.S
DGEMM_BETA = dgemm_beta.S DGEMM_BETA = dgemm_beta.S

View File

@ -1,37 +1,5 @@
include $(KERNELDIR)/KERNEL.ARMV8SVE include $(KERNELDIR)/KERNEL.ARMV8SVE
DAXPYKERNEL = daxpy_thunderx2t99.S
SASUMKERNEL = sasum_thunderx2t99.c
DASUMKERNEL = dasum_thunderx2t99.c
CASUMKERNEL = casum_thunderx2t99.c
ZASUMKERNEL = zasum_thunderx2t99.c
SCOPYKERNEL = copy_thunderx2t99.c
DCOPYKERNEL = copy_thunderx2t99.c
CCOPYKERNEL = copy_thunderx2t99.c
ZCOPYKERNEL = copy_thunderx2t99.c
SSWAPKERNEL = swap_thunderx2t99.S
DSWAPKERNEL = swap_thunderx2t99.S
CSWAPKERNEL = swap_thunderx2t99.S
ZSWAPKERNEL = swap_thunderx2t99.S
ISAMAXKERNEL = iamax_thunderx2t99.c
IDAMAXKERNEL = iamax_thunderx2t99.c
ICAMAXKERNEL = izamax_thunderx2t99.c
IZAMAXKERNEL = izamax_thunderx2t99.c
SNRM2KERNEL = scnrm2_thunderx2t99.c
DNRM2KERNEL = dznrm2_thunderx2t99.c
CNRM2KERNEL = scnrm2_thunderx2t99.c
ZNRM2KERNEL = dznrm2_thunderx2t99.c
DDOTKERNEL = dot.c
SDOTKERNEL = dot.c
CDOTKERNEL = zdot_thunderx2t99.c
ZDOTKERNEL = zdot_thunderx2t99.c
CTRSMKERNEL_LN = ../generic/trsm_kernel_LN.c CTRSMKERNEL_LN = ../generic/trsm_kernel_LN.c
CTRSMKERNEL_LT = ../generic/trsm_kernel_LT.c CTRSMKERNEL_LT = ../generic/trsm_kernel_LT.c
CTRSMKERNEL_RN = ../generic/trsm_kernel_RN.c CTRSMKERNEL_RN = ../generic/trsm_kernel_RN.c