Skip to content

Conversation

againull
Copy link
Contributor

Existing Sample Changes

Description

Currently there is an out-of-bound access in last iteration of the main loop in iso3dfd kernel (5_GPU_optimized sample):


prev_acc is accessed out-of-bounds in the last iteration.
To avoid this, don't prepare unnecessary data for the next iteration if the current iteration is the last one.

Fixes Issue#
CMPLRLLVM-69572

Type of change

  • Bug fix (non-breaking change which fixes an issue)

How Has This Been Tested?

Verified on
Intel(R) Data Center GPU Max 1550
with gpu driver 25.05.32567.17
OS: Ubuntu 22.04

intel/llvm compiler: intel/llvm@e8c8555

Commands:
clang++ -fsycl 5_GPU_optimized.cpp -o 5_GPU_optimized
./5_GPU_optimized 256 256 256 100 256 16 16 verify

Hangs without the fix, passes verification with the fix.

@againull
Copy link
Contributor Author

@IgorOchocki Could you please take a look at this PR and approve/merge if looks ok. Please feel free to redirect if I should ask someone else.

@againull
Copy link
Contributor Author

Can this PR be merged please?

@IgorOchocki IgorOchocki merged commit 44b82a8 into oneapi-src:main Aug 20, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants