JIT: Produce correctly typed IR in TLS helper expansion #108836
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The TLS helper expansion would produce `ADD(long, int)` trees like ``` ▌ STORE_LCL_VAR long V01 rat0 └──▌ ADD long ├──▌ ADD long │ ├──▌ LCL_VAR long V02 rat1 │ └──▌ CNS_INT int 72 $41 └──▌ CNS_INT long 192 ``` which is not legal IR. Fix that by inserting a cast. Also do some more aggressive folding to get rid of these casts in the normal cases, and to fold some constant arithmeticf in other cases.
dotnet-issue-labeler
bot
added
the
area-CodeGen-coreclr
|
cc @dotnet/jit-contrib PTAL @kunalspathak |
There was a problem hiding this comment.
Added a question.
I did verify the difference between my local disasm vs. what jit-diff is showing. The reason I didn't see diff locally was because on a standalone app, the tlsindex value was low that was getting removed because we were able to use an addressing mode to load it. However, in jit-diffs, we get higher tls index value which cannot be expressed using addressing mode and hence couldn't be eliminated.
| @@ -1060,6 +1068,14 @@ bool Compiler::fgExpandThreadLocalAccessForCall(BasicBlock** pBlock, Statement* | |||
| // Create tree to "threadStaticBlockValue = threadStaticBlockBase[typeIndex]" | |||
| typeThreadStaticBlockIndexValue = gtNewOperNode(GT_MUL, TYP_INT, gtCloneExpr(typeThreadStaticBlockIndexValue), | |||
| gtNewIconNode(TARGET_POINTER_SIZE, TYP_INT)); | |||
| // Usually a constant, try to fold | |||
| typeThreadStaticBlockIndexValue = gtFoldExpr(typeThreadStaticBlockIndexValue); | |||
There was a problem hiding this comment.
should this be or do we need to fold expression before casting it as well?
#ifdef TARGET_X64
typeThreadStaticBlockIndexValue = gtNewCastNode...
#endif
typeThreadStaticBlockIndexValue = gtFoldExpr(...There was a problem hiding this comment.
We need to fold it before as well since gtFoldExpr is not recursive.
kunalspathak
approved these changes
Oct 15, 2024
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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.
The TLS helper expansion would produce
ADD(long, int)trees likewhich is not legal IR. Fix that by inserting a cast. Also do some more aggressive folding to get rid of these casts in the normal cases, and to fold some constant arithmetic in other cases.
Noticed this while investigating #108830.