[all-commits] [llvm/llvm-project] 23bc21: [AMDGPU] Canonicalize num_records to its actual wi...
Krzysztof Drewniak via All-commits
all-commits at lists.llvm.org
Thu Oct 1 13:47:31 PDT 2026
Branch: refs/heads/main
Home: https://github.com/llvm/llvm-project
Commit: 23bc2182f84a42e3bc108d5115bd272faed545c5
https://github.com/llvm/llvm-project/commit/23bc2182f84a42e3bc108d5115bd272faed545c5
Author: Krzysztof Drewniak <Krzysztof.Drewniak at amd.com>
Date: 2026-10-01 (Thu, 01 Oct 2026)
Changed paths:
M clang/test/CodeGenOpenCL/builtins-amdgcn-make-buffer-rsrc.cl
M llvm/lib/Target/AMDGPU/AMDGPUInstCombineIntrinsic.cpp
M llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
M llvm/test/Transforms/InstCombine/AMDGPU/make-buffer-rsrc-num-records.ll
Log Message:
-----------
[AMDGPU] Canonicalize num_records to its actual width in InstCombine (#217068)
This PR adds code to InstCombineIntrinsic to change the width of the
num_recods field (by extension or truncation) to the correct width for
the target triple (if a concrete enough target triple has been set) so
that LLVM IR-level optimizations can see the lack of demand for the
high bits, for example.
Assisted by Claude, which also found those buffer lowering edge cases
To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications
More information about the All-commits
mailing list