Fix to store results in correct order for POWER10 GEMM kernels
authorRajalakshmi Srinivasaraghavan <rajis@linux.ibm.com>
Sat, 25 Jul 2020 04:08:11 +0000 (23:08 -0500)
committerRajalakshmi Srinivasaraghavan <rajis@linux.ibm.com>
Sat, 25 Jul 2020 04:08:11 +0000 (23:08 -0500)
commit9be2688c78e1646e406e425b4c79e6f82db9f94e
treeae3a47d3220df6bfc93f3fe44de49dba613a504c
parent6a2a60038c9804fc89bd7c66771aac2a5691fee9
Fix to store results in correct order for POWER10 GEMM kernels

There is a recent compiler change in __builtin_mma_disassemble_acc() which
affects the order of storing result in POWER10. Also removing new LDFLAG
-mno-power10-stub as it is handled by linker automatically.
Makefile.system
kernel/power/dgemm_kernel_power10.c
kernel/power/sgemm_kernel_power10.c
kernel/power/shgemm_kernel_power10.c