projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 4de9a75) | patch
AMDGPU: Skip folding REG_SEQUENCE if found unknown regclasses for its users
authorChristudasan Devadasan <Christudasan.Devadasan@amd.com>
Wed, 2 Mar 2022 06:46:29 +0000 (12:16 +0530)
committerChristudasan Devadasan <Christudasan.Devadasan@amd.com>
Tue, 8 Mar 2022 04:41:57 +0000 (10:11 +0530)
commit0d849b8249e8d25484cab3e6aaba3cbcddc408dd
tree3799352d26cab5909b115c2aa533099131278b9btree | snapshot
parent4de9a752d6affb7427d53c50bb64c94c3014b3f0commit | diff
AMDGPU: Skip folding REG_SEQUENCE if found unknown regclasses for its users

Use TII::getRegClass to return a valid regclass or a nullptr
if the RC is unknown for a given OpIdx. This fixes a potential
crash occurred while getting the RC from a variadic instruction.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D120813
llvm/lib/Target/AMDGPU/SIFoldOperands.cpp diff | blob | history
llvm/test/CodeGen/AMDGPU/skip-fold-regsequence.mir [new file with mode: 0644] blob
Domain: System / Toolchain;