Fix AMX detection to request OS XSAVE permission on Linux

[alexreinking]

On Linux, the AMX XTILEDATA state component (XCR0 bit 18) is allocated lazily: the kernel leaves it disabled until a thread explicitly requests it via arch_prctl(ARCH_REQ_XCOMP_PERM). Without this call, xgetbv() reports AMX as OS-disabled even on hardware that fully supports it, and code generated for AVX512_SapphireRapids can fault at runtime. Request the permission and gate the feature on the resulting XCR0 bits in both the AOT runtime feature-detection module and the compiler's own host-target detection in Target.cpp.

Fixes #9191

Breaking changes

None

Checklist

• Commits include AI attribution where applicable (see Code of Conduct)

[alexreinking]

@sanvila -- Can you test this patch, please?

[abadams]

Approving because it looks right, but agree that it would be good to have it tested by the original reporter before merging.

[codecov]

Codecov Report

❌ Patch coverage is 75.00000% with 1 line in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (main@47e0ce8). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
src/LLVM_Runtime_Linker.cpp	75.00%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #9199   +/-   ##
=======================================
  Coverage        ?   69.66%           
=======================================
  Files           ?      255           
  Lines           ?    78567           
  Branches        ?    18787           
=======================================
  Hits            ?    54734           
  Misses          ?    18191           
  Partials        ?     5642           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[sanvila]

Hi. Sorry, the patch does not apply cleanly on top of the current Debian package, so I can't test this easily (I can only test Debian source packages, either from Debian unstable, or imported from some other place).

Maybe the Debian maintainer @LebedevRI has some idea about what we could do here.

[LebedevRI]

Hi. Sorry, the patch does not apply cleanly on top of the current Debian package, so I can't test this easily (I can only test Debian source packages, either from Debian unstable, or imported from some other place).

Maybe the Debian maintainer @LebedevRI has some idea about what we could do here.

That's practically expected given that last release was almost a year ago.
One would need to manually backport the patch, e.g. by manually applying the actual textual diff.

[alexreinking]

I'll see how far I can get in WSL on my Sapphire Rapids workstation

[alexreinking]

Tested locally on my Sapphire Rapids workstation (Xeon w5-2455X) via WSL2. On main, I reproduced the SIGILL in correctness_tiled_matmul and on this branch, the test passes.

Without arch_prctl(ARCH_REQ_XCOMP_PERM), tileloadd faults even though XCR0 already reports the AMX bits as set. XCR0 reflects system-wide hardware/kernel capability, rather than current-process permission, so the arch_prctl call is necessary (harmless if already granted). After the call, tileloadd succeeds.