projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: bcdaccf) | patch
[AMDGPU] GCNDPPCombine: don't shrink V_ADD_CO_U32 if carry out is used
authorJay Foad <jay.foad@amd.com>
Mon, 19 Apr 2021 13:48:20 +0000 (14:48 +0100)
committerJay Foad <jay.foad@amd.com>
Tue, 20 Apr 2021 08:17:52 +0000 (09:17 +0100)
commitb22721f01a580ceed64923f528e1b7f3d66a12a9
treea4ae9a5f0e2abe335652aeec73ba4f7dd1cf977atree | snapshot
parentbcdaccfe3466dcce889596599de4a6188726a84ecommit | diff
[AMDGPU] GCNDPPCombine: don't shrink V_ADD_CO_U32 if carry out is used

Don't shrink VOP3 instructions if there are any uses of a carry-out
operand, because the shrunken form of the instruction would write the
carry-out to vcc instead of to a virtual register.

Differential Revision: https://reviews.llvm.org/D100760
llvm/lib/Target/AMDGPU/GCNDPPCombine.cpp diff | blob | history
llvm/test/CodeGen/AMDGPU/dpp_combine.mir diff | blob | history
Domain: System / Toolchain;