JIT: Add separate ABI classifiers for x86/win-x64/SysV-x64/ARM64 #100276
Conversation
This file contains hidden or 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
ghost
added
the
area-CodeGen-coreclr
jakobbotsch
changed the title
| bool HasRetBuff = false; | ||
| }; | ||
|
|
||
| class X86Classifier |
There was a problem hiding this comment.
What do you think about having target-specific classifiers in target-specific files (targetXYZ.cpp, which we already have but don't really use)?
There was a problem hiding this comment.
I like that idea. They'll probably have to stay as being defined in abi.h, but I'll try moving the classification Implementation into those target specific files.
This was referenced Mar 26, 2024
For HVAs the element size can be > 8, but should at most be 8-byte aligned.
|
cc @dotnet/jit-contrib PTAL @AndyAyersMS No diffs. Some TP regressions, with some MinOpts outliers in the collections that don't have a lot of MinOpts collections. |
1 task
AndyAyersMS
approved these changes
Mar 29, 2024
matouskozak
pushed a commit
to matouskozak/runtime
that referenced
this pull request
Apr 30, 2024
…net#100276) While trying to implement struct support for Swift reverse pinvokes I ended up having to factor the ABI classification out of `lvaInitUserArgs` to be able to invoke it multiple times for structs. In the end that led me to do the work to split out the ABI classification into a different classifier for each ABI. ARM32, LA64 and RV64 still need to be split out. We still don't actually use this information for anything, but it's asserted that it is correct. We are going to be using it in the backend for Swift reverse pinvokes, and I expect to move more and more things to use it and get completely rid of the ABI classification within `lvaInitUserArgs` (and in the long run hopefully `CallArgs::AddFinalArgsAndDetermineABIInfo`). Another todo: this classification should be where we determine whether something is passed implicit by reference instead of doing that in `lvaSetStruct`/`getArgTypeForStruct`.
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.
While trying to implement struct support for Swift reverse pinvokes I ended up having to factor the ABI classification out of
lvaInitUserArgsto be able to invoke it multiple times for structs. In the end that led me to do the work to split out the ABI classification into a different classifier for each ABI.ARM32, LA64 and RV64 still need to be split out.
We still don't actually use this information for anything, but it's asserted that it is correct. We are going to be using it in the backend for Swift reverse pinvokes, and I expect to move more and more things to use it and get completely rid of the ABI classification within
lvaInitUserArgs(and in the long run hopefullyCallArgs::AddFinalArgsAndDetermineABIInfo).Another todo: this classification should be where we determine whether something is passed implicit by reference instead of doing that in
lvaSetStruct/getArgTypeForStruct.