[Xtensa] Fix FP mul-sub fusion (LLVM-276) #76
Closed
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
`performMADD_MSUBCombine` treats `x * y ± z` and `z ± x * y` interchangeably which is wrong for `msub.s` which expects the latter. Added a check to determine that the orientation is correct, and if not, negate the result.
github-actions
bot
changed the title
|
Correction: MIPS does in fact have Planning to add some llc/filechecker codegen tests for this fix |
1. Prefer `fneg.s` to `l32r ar, 0x80000000; xor ar, as, ar` when `wfr/rfr` are used anyway 2. Add Patterns for fma/madd/msub to automatically generate msub when it's the better choice 3. XtensaISelLowering.cpp: Rely on LLVM to lower FMA to madd.s/msub.s instead of hardcoding 4. Add float-fma.ll with various fused multiply add/subtract permutations
espressif-bot
added
Status: In Progress
espressif-bot
added
Status: In Review
and removed
Status: In Progress
andreisfr
reviewed
Aug 17, 2023
espressif-bot
added
Status: Opened
Status: In Progress
|
Closed with 19e97af |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
performMADD_MSUBCombinetreatsx * y ± zandz ± x * yinterchangeably which is wrong formsub.swhich expects the latter. Added a check to determine that the orientation is correct, and if not, negate the result.Seems like the original code was inspired by the MIPS code, which only has fused-multiply-add, and the order of the operands is irrelevant in that case.
This is my first PR to LLVM, so keep this in mind while reviewing, because I don't want to introduce a regression.
Related to (and hopefully fixes) esp-rs/rust#180, esp-rs/rust#185.