merge main into amd-staging #755
Merged
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
…while`. (llvm#170630) The lowering fails when `fir.iterate_while` has no final value; this patch fixes the bug.
clock_conversion.h implements convert_clock which shifts a timestamp from one clock domain to another. It naturally does not depend on any OS specific interface. Making it generic will allow common use.
Change `getSupportedMetadata` to return `SmallVector<unsigned>` instead of `ArrayRef<unsigned>` and make the list non-static. This ensures metadata identifiers are correctly obtained per LLVM context, preventing incorrect import when multiple contexts are used (for metadata like vector hints or work group sizes which have non-static IDs).
…lvm#169950) Builds up on the solution proposed for llvm#169491 and llvm#169924 and applies it for SDIV exact as well. Almost a carbon copy of UDIV exact solution from llvm#169949.
…dexToSPIRV.cpp (NFC)
…t) pattern. (llvm#166935) Fixes: llvm#152797 alive2: https://alive2.llvm.org/ce/z/h8HYTo godbolt: https://godbolt.org/z/Mqzs9W8q4
…n cast function pointer (llvm#170502) Part of llvm#170476 When check equal of type, we need to ignore ParenType
Return the input chain at the callsite, not the entry node chain. Presumably this could cause issues somewhere.
Just an attempt to simplify some checks, remove extra calls and reorder checks to make code simpler and faster Reviewers: RKSimon, hiraditya Reviewed By: hiraditya Pull Request: llvm#170382
…transform (llvm#170562) Type legalization can promote constant operands. The MULHU optimization `mulhu x, (1 << c) -> x >> (bitwidth - c)` was failing when constants were promoted because: 1. `isConstantOrConstantVector` check rejected promoted constants 2. `BuildLogBase2` -> `takeInexpensiveLog2` -> `matchUnaryPredicate` rejected promoted constants This fixes both by adding `AllowTruncation=true`, following the pattern from the recent UDIV fix (llvm#169491).
llvm#170351) Use GetOmpObjectList instead of extracting the object list by hand.
See discussion on llvm#169061
Addresses this warning:
```
/home/fedora/src/llvm-project/main/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp:3358
79: warning: suggest parentheses around ‘&&’ within ‘||’ [-Wparentheses]
3358 | (getImm(I.getOperand(4), MRI) && foldImm(I.getOperand(4), MRI) == 0) &&
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~
3359 | "Cannot translate GEP to OpAccessChain. First index must be 0.");
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
```
This was introduced in llvm#169559
`VmpaIntID` was used in variable definition of `VmpaIntID`.
```
/home/fedora/src/llvm-project/main/llvm/lib/Target/Hexagon/HexagonGenWideningVecInstr.cpp: In member function ‘bool {anonymous}::HexagonGenWideningVecInstr::replaceWithVmpaIntrinsic(llvm::Instruction*, OPInfo*)’:
/home/fedora/src/llvm-project/main/llvm/lib/Target/Hexagon/HexagonGenWideningVecInstr.cpp:843:41: warning: operation on ‘VmpaIntID’ may be undefined [-Wsequence-point]
843 | (NewResEltSize == 16) ? VmpaIntID = Intrinsic::hexagon_V6_vmpabus_128B
| ~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
```
This PR re-lands llvm#156830 This PR aims at fixing the nvdsl examples which got a bit out of sync not being tested in the CI. The fixed bugs were related to the following PRs: - move to nanobind llvm#118583 - split gpu module initialization llvm#135478 - gpu dialect python API change llvm#163883
…eck prefixes (llvm#170662) Add missing common AVX prefixes Remove FALLBACK prefixes - now that the update script will tell us when there are clashes
Moves the insertion of SPIRV::ASSIGN_TYPE for type-folded instructions from the PreLegalizer to the PostLegalizer. This allows the legalizer to do more because the ASSIGN_TYPE instructions do not get in the way.
…sed (llvm#170531) llvm#170525 Based on our [RFC](https://discourse.llvm.org/t/rfc-libc-wctype-header-implementation/88941/10), we are starting the implementation of the wctype header with C.UTF8 support. This implementation will increase the binary size by approximately ~70KB, so this PR introduces a flag so that this is configurable, allowing the user to choose between a simple ASCII implementation or the full Unicode one.
We need to cat all files that start with comments now that premerge writes out files containing comments without that exact name.
Collaborator
Author
ronlieb
approved these changes
Dec 4, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
20 participants
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.
No description provided.