Skip to content

[Bugfix] rocm shared memory issue on MI250 #16901

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
Apr 17, 2024
Merged

[Bugfix] rocm shared memory issue on MI250 #16901

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
c018ec4
[Bugfix] rocm shared memory issue on MI250
LeshengJin Apr 17, 2024
c730bfb
set LOAD_V_SHARED=False
LeshengJin Apr 17, 2024
9cd7a94
upd the comment
LeshengJin Apr 17, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 4 additions & 1 deletion python/tvm/dlight/gpu/gemv.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -469,7 +469,10 @@ def apply(
TS, TR = 2, 64
elif target.kind.name == "rocm":
VEC_C = 4
LOAD_V_SHARED = True
# TODO: set LOAD_V_SHARED = False for now
# rocm might have some issues when load/store of shared do not belong to same data type
# and only works for certain vector lens, our commonly useful vector lens are in 4
LOAD_V_SHARED = False
LOAD_V_VEC = 8
UNROLL = 256
if isinstance(len_S, int):
Expand Down