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

Merge pull request #3155 from martin-frbg/issue3152 

Fix recent SGEMM_direct breakage on SkylakeX and Cooperlake
This commit is contained in:
Martin Kroeker 2021-03-19 09:55:31 +01:00 committed by GitHub
parent 1e4b2e98d9 7b294a99fd
commit c4b91bfcf1
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
kernel/x86_64/sgemm_direct_skylakex.c
View File

@ -1,10 +1,10 @@
/* the direct sgemm code written by Arjan van der Ven */ /* the direct sgemm code written by Arjan van der Ven */
#include "common.h"
#if defined(SKYLAKEX) || defined (COOPERLAKE) #if defined(SKYLAKEX) || defined (COOPERLAKE)
#include <immintrin.h> #include <immintrin.h>
#include "common.h"
/* /*
* "Direct sgemm" code. This code operates directly on the inputs and outputs * "Direct sgemm" code. This code operates directly on the inputs and outputs
@ -472,7 +472,7 @@ void CNAME (BLASLONG M, BLASLONG N, BLASLONG K, float * __restrict A, BLASLONG s
} }
} }
#else #else
#include "common.h"
void CNAME (BLASLONG M, BLASLONG N, BLASLONG K, float * __restrict A, BLASLONG strideA, float * __restrict B, BLASLONG strideB , float * __restrict R, BLASLONG strideR) void CNAME (BLASLONG M, BLASLONG N, BLASLONG K, float * __restrict A, BLASLONG strideA, float * __restrict B, BLASLONG strideB , float * __restrict R, BLASLONG strideR)
{} {}
#endif #endif