This is an archive of the discontinued LLVM Phabricator instance.

AMDGPU/SI: Add test for constant folding
ClosedPublic

Authored by axeldavy on Aug 11 2015, 11:56 AM.

Reviewers

• tstellarAMD
arsenm

Summary

This test currently fails

axeldavy retitled this revision from to AMDGPU/SI: Add test for constant folding.

axeldavy updated this object.

axeldavy added a subscriber: llvm-commits.

I think CHECK-NOT v_mov is overly general, and needs to check for a moved immediate. You also definitely don't want strict-whitspace

Improved the pattern matching

LGTM

This revision is now accepted and ready to land.Aug 17 2015, 8:41 AM

r246167. I renamed the test file and changed the commit message a bit

Path

Size

test/

CodeGen/

AMDGPU/

17 lines

				; XFAIL: *
				; RUN: llc -march=amdgcn -mcpu=SI -verify-machineinstrs < %s \| FileCheck %s

				; CHECK-LABEL: {{^}}main:
				; CHECK-NOT: v_mov_b32_e32 v{{[0-9]+}}, 0xbf4353f8

				define void @main(float) #0 {
				main_body:
				%1 = fmul float %0, 0x3FE86A7F00000000
				%2 = fmul float %0, 0xBFE86A7F00000000
				call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %1, float %1, float %2, float %2)
				ret void
				}

				declare void @llvm.SI.export(i32, i32, i32, i32, i32, float, float, float, float)

				attributes #0 = { "ShaderType"="1" }