projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 82cf72a) | patch
AMDGPU: Create alloca wide load/store with explicit alignment
authorMatt Arsenault <Matthew.Arsenault@amd.com>
Tue, 3 Jan 2023 15:37:08 +0000 (10:37 -0500)
committerMatt Arsenault <Matthew.Arsenault@amd.com>
Tue, 3 Jan 2023 16:29:18 +0000 (11:29 -0500)
commit687e0e205e5e95ff157aa5977164c155fb424f23
tree7fd4dff6f39b7aea78ad0a31b22c5ff6d075f477tree | snapshot
parent82cf72a83b0823b9426f280f8620daab67c221e4commit | diff
AMDGPU: Create alloca wide load/store with explicit alignment

This was introducing transient UB by using the default alignment of a
larger vector type.
llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp diff | blob | history
llvm/test/CodeGen/AMDGPU/promote-alloca-array-aggregate.ll diff | blob | history
llvm/test/CodeGen/AMDGPU/promote-alloca-pointer-array.ll diff | blob | history
Domain: System / Toolchain;