gaspersic/triton
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Improvements w/ Auto-Tuning and standard benchmarks (#57)

Browse Source 
[PYTHON] Bug-fixes in the auto-tuning module and improvement of the existing API for it
This commit is contained in:
Philippe Tillet
2021-02-03 13:37:21 -08:00
committed by Philippe Tillet
parent ad005d49ac
commit 6fb4800f57
12 changed files with 215 additions and 149 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
python/triton/ops/matmul.c
View File

@@ -83,8 +83,8 @@ __global__ void matmul(TYPE * A __noalias __readonly __aligned(16),
*?(checkc) pc = c;
#else
// accumulate partial result using spin-locks
int *plock = locks + rid;
int *pcount = plock + get_num_programs(0) * get_num_programs(1);
int *plock = locks + pid;
int *pcount = plock + get_num_programs(0);
for(int repeat = 1; repeat == 1; repeat = atomic_cas(plock, 0, 1));
int count = *pcount;
if(count == 0)