-
Notifications
You must be signed in to change notification settings - Fork 734
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.
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
[SYCL][ESIMD]Fix lsc_slm_block_load test #12484
Conversation
ESIMD/matrix_transpose_glb.cpp test failures are not related to the change |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I don't understand why this change can fix something.
It is even look wrong to me.
The initialization is done in int-type equivalent units - similarly as the verification is done later. That is done intentionally to avoid bits lost or roundings for FP types.
For example, Vals gets different numbers than 1M*GroupID when it is T
and when it is Tuint
.
ExpectedVal in the verification for the transpose case is using T but the kernel uses Tuint, I assumed that's the problem. |
Ough..., I see now. It seem it is some mess in init/use/verify code here. non-transpose verification has Uint to verify with expected values, but both transposed/non-transposed use the common LSM initialization code. |
AWS-CUDA fails are unrelated to changes in 2 ESIMD E2E tests. Proceeding to merge now. |
No description provided.