SPMI: Add fallback for notifyInstructionSetUsage
#112925
Merged
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
I noticed that
notifyInstructionSetUsage
misses are pretty common, particularly because we query for instruction sets for some very basic instructions (e.g.GT_RSZ
/GT_RSH
). So changes that introduce new kinds of nodes now seem to be much more likely to miss.The best solution would probably be to add another set of ISAs that is passed to the JIT so that
notifyInstructionSetUsage
can be purely a "result" type of JIT-EE call, but this change at least gets us back to what we had before with an improvement to when we actually know the answer.As an example, I started getting lots of misses on win-x64 for #112740 after I merged the previous change. That turns out to be because that PR produces
GT_RSZ
which ends reportingBMI2
usage, and that now results in a miss.cc @dotnet/jit-contrib