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

RFC : Add half precision gemm for bfloat16 in OpenBLAS

Browse Source 
This patch adds support for bfloat16 data type matrix multiplication kernel.
For architectures that don't support bfloat16, it is defined as unsigned short
(2 bytes).  Default unroll sizes can be changed as per architecture as done for
SGEMM and for now 8 and 4 are used for M and N.  Size of ncopy/tcopy can be
changed as per architecture requirement and for now, size 2 is used.

Added shgemm in kernel/power/KERNEL.POWER9 and tested in powerpc64le and
powerpc64.  For reference, added a small test compare_sgemm_shgemm.c to compare
sgemm and shgemm output.

This patch does not cover OpenBLAS test, benchmark and lapack tests for shgemm.
Complex type implementation can be discussed and added once this is approved.
This commit is contained in:
Rajalakshmi Srinivasaraghavan
2020-04-14 14:55:08 -05:00
parent c861b2a7bd
commit 7eb55504b1
27 changed files with 594 additions and 65 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
common_interface.h
View File

@@ -37,6 +37,9 @@
/*********************************************************************/
#ifndef ASSEMBLER
#ifndef BFLOAT16
typedef unsigned short bfloat16;
#endif
#ifdef __cplusplus
extern "C" {
@@ -469,6 +472,8 @@ void BLASFUNC(xhbmv)(char *, blasint *, blasint *, xdouble *, xdouble *, blasint
/* Level 3 routines */
void BLASFUNC(shgemm)(char *, char *, blasint *, blasint *, blasint *, float *,
bfloat16 *, blasint *, bfloat16 *, blasint *, float *, float *, blasint *);
void BLASFUNC(sgemm)(char *, char *, blasint *, blasint *, blasint *, float *,
float *, blasint *, float *, blasint *, float *, float *, blasint *);
void BLASFUNC(dgemm)(char *, char *, blasint *, blasint *, blasint *, double *,