mirror of
https://github.com/tile-ai/tilelang.git
synced 2026-10-24 07:04:00 +08:00
Gongen-Ali
4e53b50750
* Fix some bugs on ci and open rocm ci test * fix pre-commit * fix build issue on hip * refactor on hip * pre-commit fix * fix test with pdl * fix pdl test
36 lines
926 B
Python
36 lines
926 B
Python
import torch
|
|
import tilelang
|
|
import tilelang.testing
|
|
from tilelang import language as T
|
|
|
|
|
|
@tilelang.jit(
|
|
pass_configs={
|
|
tilelang.PassConfigKey.TL_DISABLE_WARP_SPECIALIZED: True,
|
|
tilelang.PassConfigKey.TL_DISABLE_TMA_LOWER: True,
|
|
},
|
|
)
|
|
def _cumsum_view_infer_layout(hidden):
|
|
num_tokens = T.dynamic("num_tokens")
|
|
|
|
@T.prim_func
|
|
def buggy_kernel(x: T.Tensor[(num_tokens, hidden), T.float]):
|
|
with T.Kernel(num_tokens, threads=128) as pid:
|
|
smem = T.alloc_shared((hidden,), dtype=T.float32)
|
|
T.copy(x[pid, :], smem)
|
|
T.cumsum(T.view(smem, (1, hidden)), dim=1)
|
|
|
|
return buggy_kernel
|
|
|
|
|
|
@tilelang.testing.requires_cuda
|
|
def test_cumsum_view_infer_layout():
|
|
hidden = 128
|
|
x = torch.randn(1, hidden, device="cuda", dtype=torch.float)
|
|
kernel = _cumsum_view_infer_layout(hidden)
|
|
kernel(x)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
tilelang.testing.main()
|