JIT: Remove CallArg::m_tmpNum
#104429
Merged
JIT: Remove CallArg::m_tmpNum
#104429
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
Before this change the JIT has two places where it may introduce temps during call args morphing: 1. Directly during `fgMorphArgs` as part of making struct copies for some struct args, like implicit byref args and other struct args requiring to be of `LCL_VAR` shape 2. During `EvalArgsToTemps` as part of making sure evaluation order is maintained while we get the call into the right shape with registers To make this work we have `CallArg::m_tmpNum` that communicates the local from 1 to 2; the responsibility of creating the actual `LCL_VAR` node to put in the late argument list was left to `EvalArgsToTemps` while `fgMorphArgs` would just change the early setup node to a store into the temporary. This PR changes it so that 1 directly introduces the right late node when it creates its temporary. That is, it puts the call argument into the right shape immediately and avoids relying on `EvalArgsToTemps` to create the local node in the late list. The benefit of that is that we no longer need to communicate the temporary as part of every `CallArg`. However, the motivation here is something else: as part of dotnet#104370 we may have arguments that need multiple temporaries to copy, so having things working in this way introduces complications for that.
dotnet-issue-labeler
bot
added
the
area-CodeGen-coreclr
|
Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch |
jakobbotsch
commented
Jul 4, 2024
Comment on lines
-877
to
-883
|
|
||
| if (argx == nullptr) | ||
| { | ||
| // Should only happen if remorphing in which case we do not need to | ||
| // make a decision about temps. | ||
| continue; | ||
| } |
There was a problem hiding this comment.
ArgsComplete is not called during remorphing, so the code should be dead.
jakobbotsch
commented
Jul 4, 2024
| @@ -3264,29 +3222,31 @@ void Compiler::fgMakeOutgoingStructArgCopy(GenTreeCall* call, CallArg* arg) | |||
| assert(!fgGlobalMorph); | |||
| } | |||
|
|
|||
| call->gtArgs.SetNeedsTemp(arg); | |||
There was a problem hiding this comment.
This was previously done in ArgsComplete (the code in the GTF_ASG case), but it makes more sense just to do it immediately when we introduce the temp.
This was referenced Jul 4, 2024
Open
|
/azp run runtime-coreclr superpmi-diffs |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
Ping @EgorBo |
EgorBo
approved these changes
Jul 10, 2024
matouskozak
added a commit
to matouskozak/runtime
that referenced
this pull request
Jul 11, 2024
This reverts commit 7029008.
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 JIT currently has two places where it may introduce temps during call args morphing:
fgMorphArgsas part of making struct copies for some struct args, like implicit byref args and other struct args requiring to be ofLCL_VARshapeEvalArgsToTempsas part of making sure evaluation order is maintained while we get the call into the right shape with registersTo make this work we have
CallArg::m_tmpNumthat communicates the local from 1 to 2; the responsibility of creating the actualLCL_VARnode to put in the late argument list was left toEvalArgsToTempswhilefgMorphArgswould just change the early setup node to a store into the temporary.This PR changes it so that 1 directly introduces the right late node when it creates its temporary. That is, it puts the call argument into the right shape immediately and avoids relying on
EvalArgsToTempsto create the local node in the late list.The benefit of that is that we no longer need to communicate the temporary as part of every
CallArg. However, the motivation here is something else: as part of #104370 we may have arguments that need multiple temporaries to copy, so having things working in this way introduces complications for that.No diffs are expected.