AMDGPU: Swap order of operands in fadd/fsub combine
ClosedPublic
Actions

Authored by arsenm on Dec 18 2016, 1:51 PM.

Details

Reviewers

• tstellarAMD
nhaehnle

Summary

FMA is canonicalized to constant in the middle operand. Do
the same so fmad matches and avoid an extra combine step.

arsenm retitled this revision from to AMDGPU: Swap order of operands in fadd/fsub combine .

arsenm updated this object.

arsenm added a reviewer: nhaehnle.

arsenm added a subscriber: llvm-commits.

LGTM.

This revision is now accepted and ready to land.Dec 19 2016, 2:21 PM

r290313

			Path	Packages
M			lib/Target/AMDGPU/SIISelLowering.cpp (8 lines)
M			test/CodeGen/AMDGPU/fmul-2-combine-multi-use.ll (16 lines)
M			test/CodeGen/AMDGPU/fmuladd.f16.ll (4 lines)
M			test/CodeGen/AMDGPU/fmuladd.f32.ll (4 lines)

Diff	ID	Base	Description	Created	Lint	Unit
Base			Base
Diff 1	81892			Dec 18 2016, 1:51 PM	★	★