ESIMD/mask_expand_load.cpp failure in post-commit

[bader]

Describe the bug

ESIMD/mask_expand_load.cpp has failed in post-commit for NFC change - ac77ad8.

Relevant part of the log:

FAIL: SYCL :: ESIMD/mask_expand_load.cpp (787 of 2221)
******************** TEST 'SYCL :: ESIMD/mask_expand_load.cpp' FAILED ********************
Exit Code: 1

Command Output (stdout):
--
# RUN: at line 9
D:/github/_work/llvm/llvm/install/bin/clang++.exe  -Werror  -Wno-error=deprecated-declarations -fsycl -fsycl-targets=spir64  D:\github\_work\llvm\llvm\llvm\sycl\test-e2e\ESIMD\mask_expand_load.cpp -fsycl-device-code-split=per_kernel -o D:\github\_work\llvm\llvm\build-e2e\ESIMD\Output\mask_expand_load.cpp.tmp.out
# executed command: D:/github/_work/llvm/llvm/install/bin/clang++.exe -Werror -Wno-error=deprecated-declarations -fsycl -fsycl-targets=spir64 'D:\github\_work\llvm\llvm\llvm\sycl\test-e2e\ESIMD\mask_expand_load.cpp' -fsycl-device-code-split=per_kernel -o 'D:\github\_work\llvm\llvm\build-e2e\ESIMD\Output\mask_expand_load.cpp.tmp.out'
# RUN: at line 10
env ONEAPI_DEVICE_SELECTOR=level_zero:gpu  D:\github\_work\llvm\llvm\build-e2e\ESIMD\Output\mask_expand_load.cpp.tmp.out
# executed command: env ONEAPI_DEVICE_SELECTOR=level_zero:gpu 'D:\github\_work\llvm\llvm\build-e2e\ESIMD\Output\mask_expand_load.cpp.tmp.out'
# .---command stdout------------
# | Running on Intel(R) Iris(R) Xe Graphics, driver=[1.3.29803]
# | mask_compress_load: error at I = 1: 0 != [61](https://github.com/intel/llvm/actions/runs/10672162525/job/29580079013#step:12:62)1 != 611
# | mask_compress_load: error at I = 1: 0 != 611 != 0
# `-----------------------------
# error: command failed with exit status: 1

Link to the full log containing steps to reproduce and environment information - https://github.com/intel/llvm/actions/runs/10672162525/job/29580079013.

@intel/dpcpp-esimd-reviewers, any ideas?

To reproduce

No response

Environment

No response

Additional context

No response

[sarnex]

@fineg74 Mind taking a look at this? Seems it's on Win Gen12. Maybe there is a test or driver issue? Thanks

[fineg74]

Is there any difference between post-commit and pre-commit environments ? It passes in pre-commit pretty consistently and doesn't use any hw/os specific operations i.e. it is emulation using common and basic API that is expected to work everywhere

[sarnex]

@fineg74 For Windows Gen12 pre and postcommit should be the same environment as far as I know.

[fineg74]

It sounds like it is a random issue somehow related to the environment. If I remember correctly addc/subb tests were passing on pre-commit but failing on post-commit also on Windows

[sarnex]

I can't reproduce this even using our flaky test script which almost always works, and we haven't seen this in a while, so I'll close this as not reproducible and if it happens again I'll use the failing runner itself directly to try to reproduce.

[sarnex]

Nevermind, I found the problem and have a fix.