diff options
| author | Matt Arsenault <Matthew.Arsenault@amd.com> | 2013-12-19 05:32:55 +0000 |
|---|---|---|
| committer | Matt Arsenault <Matthew.Arsenault@amd.com> | 2013-12-19 05:32:55 +0000 |
| commit | e2c28ad84f45fd19e21a7995a87f124dccc8974d (patch) | |
| tree | f68de17938869968162b554f09dc3d6aea67d553 /test/CodeGen/R600/array-ptr-calc-i32.ll | |
| parent | 251b9217c77fd7ef8a75734688f147d8bd073def (diff) | |
| download | llvm-e2c28ad84f45fd19e21a7995a87f124dccc8974d.tar.gz llvm-e2c28ad84f45fd19e21a7995a87f124dccc8974d.tar.bz2 llvm-e2c28ad84f45fd19e21a7995a87f124dccc8974d.tar.xz | |
R600/SI: Make private pointers be 32-bit.
Different sized address spaces should theoretically work
most of the time now, and since 64-bit add is currently
disabled, using more 32-bit pointers fixes some cases.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@197659 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen/R600/array-ptr-calc-i32.ll')
| -rw-r--r-- | test/CodeGen/R600/array-ptr-calc-i32.ll | 31 |
1 files changed, 31 insertions, 0 deletions
diff --git a/test/CodeGen/R600/array-ptr-calc-i32.ll b/test/CodeGen/R600/array-ptr-calc-i32.ll new file mode 100644 index 0000000000..cb2a1c80a1 --- /dev/null +++ b/test/CodeGen/R600/array-ptr-calc-i32.ll @@ -0,0 +1,31 @@ +; RUN: llc -march=r600 -mcpu=SI < %s | FileCheck -check-prefix=SI %s + +declare i32 @llvm.SI.tid() nounwind readnone +declare void @llvm.AMDGPU.barrier.local() nounwind noduplicate + +; The required pointer calculations for the alloca'd actually requires +; an add and won't be folded into the addressing, which fails with a +; 64-bit pointer add. This should work since private pointers should +; be 32-bits. + +; SI-LABEL: @test_private_array_ptr_calc: +; SI: V_ADD_I32_e32 [[PTRREG:v[0-9]+]] +; SI: V_MOVRELD_B32_e32 {{v[0-9]+}}, [[PTRREG]] +define void @test_private_array_ptr_calc(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %inA, i32 addrspace(1)* noalias %inB) { + %alloca = alloca [4 x i32], i32 4, align 16 + %tid = call i32 @llvm.SI.tid() readnone + %a_ptr = getelementptr i32 addrspace(1)* %inA, i32 %tid + %b_ptr = getelementptr i32 addrspace(1)* %inB, i32 %tid + %a = load i32 addrspace(1)* %a_ptr + %b = load i32 addrspace(1)* %b_ptr + %result = add i32 %a, %b + %alloca_ptr = getelementptr [4 x i32]* %alloca, i32 1, i32 %b + store i32 %result, i32* %alloca_ptr, align 4 + ; Dummy call + call void @llvm.AMDGPU.barrier.local() nounwind noduplicate + %reload = load i32* %alloca_ptr, align 4 + %out_ptr = getelementptr i32 addrspace(1)* %out, i32 %tid + store i32 %reload, i32 addrspace(1)* %out_ptr, align 4 + ret void +} + |