Open BruceForstall opened 2 months ago
Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch See info in area-owners.md if you want to be subscribed.
@BruceForstall PTAL.
@BruceForstall #107371 or #107403 might have exposed this
@BruceForstall #107371 or #107403 might have exposed this
I can repro the failure both before and after these PRs. (I tried 42f3dceae23a5c7acb5d3cb8e0419e1195911956, just before the first of those)
I verified this repros on .NET 9.
@BruceForstall I'm not able to repro this using a mainline build or the build from the failing CI run. I'm not sure if this is CPU-sensitive or not, but out of curiosity, what hardware did you repro this on?
@amanasifkhalid Here's my repro:
set DOTNET_TieredCompilation=0
set DOTNET_JitStress=2be
set CORE_ROOT=c:\gh\runtime\artifacts\tests\coreclr\windows.x86.Checked\Tests\Core_Root
c:\gh\runtime\artifacts\tests\coreclr\windows.x86.Checked\Tests\Core_Root\corerun.exe C:\gh\runtime\artifacts\tests\coreclr\windows.x64.Checked\managed\Compilation\Compilation\Compilation.dll
(I only build the tests for x64 Checked, but that shouldn't matter.)
It does NOT fail if you use windows.x86.Debug. I'm testing on an Intel DevBox (AVX-512 capable).
This failure only exists with JitOptRepeat
(which is enabled by the stress mode on this test). Setting set DOTNET_JitEnableOptRepeat=0
resolves the failure.
Since OptRepeat is not enabled in .NET 9, this is a .NET 10 issue.
@kunalspathak, it is related to loop alignment. PTAL.
pipeline: runtime-coreclr jitstress-random job: coreclr windows x86 Checked jitstress_random_1 test: managed/Compilation/Compilation/Compilation.cmd
https://dev.azure.com/dnceng-public/public/_build/results?buildId=801287&view=ms.vss-test-web.build-test-results-tab
@dotnet/jit-contrib