Ensure that fgMorphFieldToSimdGetElement doesn't call gtNewSimdGetElementNode if the ISA is unsupported #66853
Conversation
…mentNode if the ISA is unsupported
|
CC. @kunalspathak, @BruceForstall Will kick off the outerloop ISA jobs after JIT formatting and a couple other changes come back as "passing" to avoid unnecessary CI churn. |
maryamariyan
added
the
area-CodeGen-coreclr
|
Tagging subscribers to this area: @JulieLeeMSFT Issue DetailsThis resolves #66205 Noting that this is the "simple" fix. The change I discussed in the issue is more complex/involved and will require a few other changes/refactorings to be possible, so I opted to do the simple thing first to unblock CI. I'll log an issue tracking the more complex change.
|
|
/azp run runtime-coreclr jitstress-isas-x86, runtime-coreclr jitstress-isas-arm, runtime-coreclr outerloop |
|
Azure Pipelines successfully started running 3 pipeline(s). |
… support on Arm64
|
/azp run runtime-coreclr jitstress-isas-x86, runtime-coreclr jitstress-isas-arm, runtime-coreclr outerloop |
|
Azure Pipelines successfully started running 3 pipeline(s). |
|
/azp run runtime-coreclr jitstress-isas-x86, runtime-coreclr jitstress-isas-arm, runtime-coreclr outerloop |
|
Azure Pipelines successfully started running 3 pipeline(s). |
|
ARM64 only had an unrelated failure in LSRA. There was another related failure in x86/x64 that should be fixed now (was checking for Vector128 rather than for baseline support) |
|
Remaining ARM failure is #66921, currently assigned out to @kunalspathak R2R-CG2 failures are likewise pre-existing/unrelated |
…mentNode if the ISA is unsupported (dotnet#66853) * Ensure that fgMorphFieldToSimdGetElement doesn't call gtNewSimdGetElementNode if the ISA is unsupported * Ensure fgMorphFieldToSimdGetElement is correctly checking for AdvSimd support on Arm64 * Ensure impExpandHalfConstEqualsSIMD checks for baseline simd support
ghost
locked as resolved and limited conversation to collaborators
This resolves #66205
Noting that this is the "simple" fix. The change I discussed in the issue is more complex/involved and will require a few other changes/refactorings to be possible, so I opted to do the simple thing first to unblock CI. I'll log an issue tracking the more complex change.