-
Notifications
You must be signed in to change notification settings - Fork 35
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
rocPRIM test_warp_load failing with Intel CPU driver #480
Labels
Comments
Still fails with PoCL-CPU (v4.0) and Intel CPU.
Works with Intel iGPU via OpenCL, Intel iGPU via Level Zero and PoCL via the Level0 driver, so not blocking the 1.0 due to this. |
pjaaskel
added a commit
to pjaaskel/chip-spv
that referenced
this issue
Jul 10, 2023
This is done to trigger required subgroup size to warp size when there are warp-id-dependent memory accesses in the kernel. This fixes CHIP-SPV#480 for PoCL-CPU.
pvelesko
added a commit
that referenced
this issue
Jul 11, 2023
This is done to trigger required subgroup size to warp size when there are warp-id-dependent memory accesses in the kernel. This fixes #480 for PoCL-CPU. Co-authored-by: Paulius Velesko <[email protected]>
This still fails with Intel OpenCL CPU, so keeping it open. |
pjaaskel
changed the title
rocPRIM test_warp_load failing with CPU drivers
rocPRIM test_warp_load failing with Intel CPU driver
Jul 11, 2023
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Both Intel's and PoCL's CPU drivers produce wrong results with this quite simple test. As it's very short test it's something to investigate and could affect other kernels too. Maybe assumes warp lock-step semantics or multiple writers to same destination to succeed or such.
The text was updated successfully, but these errors were encountered: