Skip to content

Commit 334cd24

Browse files
authored
Merge pull request #5348 from hideaki-motoki/issue5343_prefered_size_for_a64fx
Setting `GEMM_PREFERED_SIZE` parameter for `A64FX`
2 parents 4062c10 + bba75d5 commit 334cd24

File tree

1 file changed

+6
-0
lines changed

1 file changed

+6
-0
lines changed

param.h

Lines changed: 6 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -3701,6 +3701,12 @@ is a big desktop or server with abundant cache rather than a phone or embedded d
37013701

37023702
#elif defined(A64FX) // 512-bit SVE
37033703

3704+
#if defined(XDOUBLE) || defined(DOUBLE)
3705+
#define GEMM_PREFERED_SIZE 8
3706+
#else
3707+
#define GEMM_PREFERED_SIZE 16
3708+
#endif
3709+
37043710
/* When all BLAS3 routines are implemeted with SVE, SGEMM_DEFAULT_UNROLL_M should be "sve_vl".
37053711
Until then, just keep it different than DGEMM_DEFAULT_UNROLL_N to keep copy routines in both directions seperated. */
37063712
#define SGEMM_DEFAULT_UNROLL_M 4

0 commit comments

Comments
 (0)