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

[OPTIMIZER] Added memory coalescing pass (#31)

Browse Source
This commit is contained in:
Philippe Tillet
2022-07-31 20:59:31 -07:00
committed by GitHub
parent d1593e6ca8
commit 3236642e8f
7 changed files with 180 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
python/examples/copy_strided.py
View File

@@ -5,8 +5,8 @@ import triton.language as tl
# triton kernel
@triton.jit
def kernel(X, stride_xm, stride_xn,
Z, stride_zm, stride_zn,
def kernel(X, stride_xm,
Z, stride_zn,
BLOCK_M: tl.constexpr, BLOCK_N: tl.constexpr):
off_m = tl.arange(0, BLOCK_M)
off_n = tl.arange(0, BLOCK_N)
@@ -15,5 +15,5 @@ def kernel(X, stride_xm, stride_xn,
tl.store(Zs, tl.load(Xs))
ret = triton.compile(kernel, "*fp32,i32,i32,*fp32,i32,i32", constants={"BLOCK_M": 64, "BLOCK_N": 64}, output="ttgir")
ret = triton.compile(kernel, "*fp32,i32,*fp32,i32", constants={"BLOCK_M": 64, "BLOCK_N": 64}, output="ttgir")
print(ret)