Add workaround for NVIDIA HPC mishandling of the asm DOT kernels
This commit is contained in:
parent
43aac5bacc
commit
c2a8ebfe69
|
@ -97,9 +97,18 @@ CNRM2KERNEL = znrm2.S
|
||||||
ZNRM2KERNEL = znrm2.S
|
ZNRM2KERNEL = znrm2.S
|
||||||
|
|
||||||
DDOTKERNEL = dot.S
|
DDOTKERNEL = dot.S
|
||||||
|
ifneq ($(C_COMPILER), PGI)
|
||||||
SDOTKERNEL = ../generic/dot.c
|
SDOTKERNEL = ../generic/dot.c
|
||||||
|
else
|
||||||
|
SDOTKERNEL = dot.S
|
||||||
|
endif
|
||||||
|
ifneq ($(C_COMPILER), PGI)
|
||||||
CDOTKERNEL = zdot.S
|
CDOTKERNEL = zdot.S
|
||||||
ZDOTKERNEL = 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
|
||||||
|
|
Loading…
Reference in New Issue