Skip to content

Commit 49959d4

Browse files
authored
Add workaround for NVIDIA HPC
1 parent 0f27a03 commit 49959d4

File tree

1 file changed

+14
-5
lines changed

1 file changed

+14
-5
lines changed

kernel/arm64/KERNEL.CORTEXA53

Lines changed: 14 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -96,11 +96,20 @@ DNRM2KERNEL = nrm2.S
9696
CNRM2KERNEL = znrm2.S
9797
ZNRM2KERNEL = znrm2.S
9898

99-
DDOTKERNEL = dot.S
100-
SDOTKERNEL = ../generic/dot.c
101-
CDOTKERNEL = zdot.S
102-
ZDOTKERNEL = zdot.S
103-
DSDOTKERNEL = dot.S
99+
ifneq ($(C_COMPILER), PGI)
100+
SDOTKERNEL = ../generic/dot.c
101+
else
102+
SDOTKERNEL = dot.S
103+
endif
104+
DDOTKERNEL = dot.S
105+
ifneq ($(C_COMPILER), PGI)
106+
CDOTKERNEL = zdot.S
107+
ZDOTKERNEL = zdot.S
108+
else
109+
CDOTKERNEL = ../arm/zdot.c
110+
ZDOTKERNEL = ../arm/zdot.c
111+
endif
112+
DSDOTKERNEL = dot.S
104113

105114
DGEMM_BETA = dgemm_beta.S
106115
SGEMM_BETA = sgemm_beta.S

0 commit comments

Comments
 (0)