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

Add workaround for NVIDIA HPC mishandling of the asm DOT kernels

This commit is contained in:
Martin Kroeker 2021-01-12 16:39:35 +01:00 committed by GitHub
parent c2a8ebfe69
commit 0f27a03607
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 9 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
kernel/arm64/KERNEL.CORTEXA57
View File

@ -70,10 +70,19 @@ DCOPYKERNEL = copy.S
CCOPYKERNEL = copy.S
ZCOPYKERNEL = copy.S
ifneq ($(C_COMPILER), PGI)
SDOTKERNEL = ../generic/dot.c
else
SDOTKERNEL = dot.S
endif
DDOTKERNEL = dot.S
ifneq ($(C_COMPILER), PGI)
CDOTKERNEL = zdot.S
ZDOTKERNEL = zdot.S
else
CDOTKERNEL = ../arm/zdot.c
ZDOTKERNEL = ../arm/zdot.c
endif
DSDOTKERNEL = dot.S
SNRM2KERNEL = nrm2.S