masahi commented on code in PR #12564:
URL: https://github.com/apache/tvm/pull/12564#discussion_r953313110
##########
python/tvm/tir/tensor_intrin/cuda.py:
##########
@@ -699,18 +697,18 @@ def wmma_sync_desc(a: T.handle, b: T.handle, c: T.handle)
-> None:
@T.prim_func
def wmma_sync_impl(a: T.handle, b: T.handle, c: T.handle) -> None:
A = T.match_buffer(
- a, (m_dim, k_dim), in_dtype, align=128, offset_factor=16,
scope="wmma.matrix_a"
+ a, (m_dim, k_dim), in_dtype, align=64, offset_factor=16,
scope="wmma.matrix_a"
)
B = T.match_buffer(
b,
maybe_swap(k_dim, n_dim),
in_dtype,
- align=128,
+ align=64,
offset_factor=16,
scope="wmma.matrix_b",
)
C = T.match_buffer(
- c, (m_dim, n_dim), out_dtype, align=128, offset_factor=16,
scope="wmma.accumulator"
+ c, (m_dim, n_dim), out_dtype, align=64, offset_factor=16,
scope="wmma.accumulator"
Review Comment:
I think all CUDA code want 128 bit alignment in practice. cc @vinx13
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
