Test #537 failure upon building Highway-1.0.4 with GCCcore-12.3.0 on x86_64

[monfortm]

Trying to build Highway 1.0.4 fails on Zen4 CPU. 1 test (#537) is failing with the following log

537/1187 Test  #537: HwyConvertTestGroup/HwyConvertTest.TestAllTruncate/AVX3_ZEN4  # GetParam() = 64 ........................................Subprocess aborted***Exception:   0.20 sec
Running main() from [...]/googletest/1.13.0/GCCcore-12.3.0/googletest-1.13.0/googletest/src/gtest_main.cc
Note: Google Test filter = HwyConvertTestGroup/HwyConvertTest.TestAllTruncate/AVX3_ZEN4
[==========] Running 1 test from 1 test suite.
[----------] Global test environment set-up.
[----------] 1 test from HwyConvertTestGroup/HwyConvertTest
[ RUN      ] HwyConvertTestGroup/HwyConvertTest.TestAllTruncate/AVX3_ZEN4


u8x16 expect [0+ ->]:
  0x00,0x01,0x02,0x03,0x04,0x05,0x06,
u8x16 actual [0+ ->]:
  0x00,0x1A,0x02,0x1A,0x04,0x1A,0x06,
Abort at [...]/Highway/1.0.4/GCCcore-12.3.0/highway-1.0.4/hwy/tests/convert_test.cc:423: AVX3_ZEN4, u8x16 lane 1 mismatch: expected '0x01', got '0x1A'.

[...]

99% tests passed, 1 tests failed out of 1187

Total Test time (real) =  30.58 sec

The following tests FAILED:
        537 - HwyConvertTestGroup/HwyConvertTest.TestAllTruncate/AVX3_ZEN4  # GetParam() = 64 (Subprocess aborted)
Errors while running CTest

cc: @jpecar

[jan-wassenberg]

Thanks for letting us know. 1.04 is pretty old. Are you able to update to 1.0.7 or HEAD and see if that is also affected?

[monfortm]

Hello, thanks for the prompt answer. The thing is we would need this specific version, as it's part of a dependency tree we have for a fairly recent software release...

[monfortm]

OK, we have moved from 1.0.4 to 1.0.7 and it solved the issue. Thanks for the heads up!

edit: I leave you with the choice to close the issue or keep it open, cheers

[jan-wassenberg]

Ah, I have been working on repro-ing the issue with SDE. Glad to hear it is solved for you, closing :)

[monfortm]

Ah, I have been working on repro-ing the issue with SDE. Glad to hear it is solved for you, closing :)

Oh, I'm sorry, hope I did not waste your time 😳 thx again

[jan-wassenberg]

No worries, was anyway good to set up SDE, it's easier than going to a separate machine.

[boegel]

We're hitting the same problem with Highway 1.0.4 (see EESSI/software-layer#613 (comment)) , and I'm wondering if you happen to have any pointers on where the fix for this may have been made in between Highway 1.0.4 and 1.0.7...
We're hoping to pinpoint the fix, so we can backport a patch to Highway 1.0.4 rather than having to jump to 1.0.7 or newer.