-
Notifications
You must be signed in to change notification settings - Fork 4.9k
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
Zero-sized diffs in smoke_tests.nativeaot #112756
Comments
Tagging subscribers to this area: @agocke, @MichalStrehovsky, @jkotas |
That PR has change in RISC-V JIT for emitting unwinding codes. I'm not surprised that it triggered the pipelines that generate those diffs. |
It was just an example. The point is that we see spurious diffs in smoke_tests.nativeaot. Here is another example: https://dev.azure.com/dnceng-public/public/_build/results?buildId=957764&view=ms.vss-build-web.run-extensions-tab |
That PR also made change in the jit directory (src/coreclr/jit/CMakeLists.txt), so that's also not too surprising to me. 🤷♂️ |
The surprise is not that superpmi-diffs is running, the surprise is that smoke_tests.nativeaot is showing up as having diffs when the PRs do not change anything about what is being emitted. |
Output
The problem looks to be that coredistools fails to disassemble
@tannergooding is this a newly emitted instruction, and are we emitting the right bytes here? If so we probably need to update coredistools. |
That's an AVX10.2 instruction and the bytes look correct. But I'm confused as to why NAOT is emitting it in the first place as AVX10.2 is off by default for the JIT and we presumably shouldn't have NAOT tests targeting "all possible ISAs" |
We have a couple tests that enable extra ISAs under src\tests\nativeaot\SmokeTests\HardwareIntrinsics. This one looks to be:
So maybe it's the test that enables |
The problem is that avx10v2 is optimistically enabled any time avx512f is enabled: runtime/src/coreclr/tools/Common/InstructionSetHelpers.cs Lines 230 to 234 in 306ced2
Not sure if that's the right behavior or not. |
It could be if there's a bug. However, that shouldn't be causing That is given the The VM should then pass both the
We have this split because the For NAOT this should simply be querying |
Then this just looks like a bug in SPMI's implementation of runtime/src/coreclr/tools/superpmi/superpmi/icorjitinfo.cpp Lines 1529 to 1533 in 306ced2
|
We're currently getting zero-sized diffs in the smoke_tests.nativeaot collection, even on changes without any expected JIT diffs.
Example: https://dev.azure.com/dnceng-public/public/_build/results?buildId=957900&view=ms.vss-build-web.run-extensions-tab
cc @dotnet/jit-contrib
The text was updated successfully, but these errors were encountered: