aboutsummaryrefslogtreecommitdiffstats
path: root/test/CodeGen/R600/array-ptr-calc-i32.ll
diff options
context:
space:
mode:
authorStephen Hines <srhines@google.com>2014-12-01 14:51:49 -0800
committerStephen Hines <srhines@google.com>2014-12-02 16:08:10 -0800
commit37ed9c199ca639565f6ce88105f9e39e898d82d0 (patch)
tree8fb36d3910e3ee4c4e1b7422f4f017108efc52f5 /test/CodeGen/R600/array-ptr-calc-i32.ll
parentd2327b22152ced7bc46dc629fc908959e8a52d03 (diff)
downloadexternal_llvm-37ed9c199ca639565f6ce88105f9e39e898d82d0.zip
external_llvm-37ed9c199ca639565f6ce88105f9e39e898d82d0.tar.gz
external_llvm-37ed9c199ca639565f6ce88105f9e39e898d82d0.tar.bz2
Update aosp/master LLVM for rebase to r222494.
Change-Id: Ic787f5e0124df789bd26f3f24680f45e678eef2d
Diffstat (limited to 'test/CodeGen/R600/array-ptr-calc-i32.ll')
-rw-r--r--test/CodeGen/R600/array-ptr-calc-i32.ll16
1 files changed, 12 insertions, 4 deletions
diff --git a/test/CodeGen/R600/array-ptr-calc-i32.ll b/test/CodeGen/R600/array-ptr-calc-i32.ll
index 3230353..84d3540 100644
--- a/test/CodeGen/R600/array-ptr-calc-i32.ll
+++ b/test/CodeGen/R600/array-ptr-calc-i32.ll
@@ -1,4 +1,5 @@
-; RUN: llc -verify-machineinstrs -march=r600 -mcpu=SI < %s | FileCheck -check-prefix=SI %s
+; RUN: llc -verify-machineinstrs -march=r600 -mcpu=SI -mattr=-promote-alloca < %s | FileCheck -check-prefix=SI-ALLOCA -check-prefix=SI %s
+; RUN: llc -verify-machineinstrs -march=r600 -mcpu=SI -mattr=+promote-alloca < %s | FileCheck -check-prefix=SI-PROMOTE -check-prefix=SI %s
declare i32 @llvm.SI.tid() nounwind readnone
declare void @llvm.AMDGPU.barrier.local() nounwind noduplicate
@@ -8,14 +9,21 @@ declare void @llvm.AMDGPU.barrier.local() nounwind noduplicate
; 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-LABEL: {{^}}test_private_array_ptr_calc:
+
+; FIXME: We end up with zero argument for ADD, because
+; SIRegisterInfo::eliminateFrameIndex() blindly replaces the frame index
+; with the appropriate offset. We should fold this into the store.
+; SI-ALLOCA: v_add_i32_e32 [[PTRREG:v[0-9]+]], 0, v{{[0-9]+}}
+; SI-ALLOCA: buffer_store_dword {{v[0-9]+}}, [[PTRREG]], s[{{[0-9]+:[0-9]+}}]
;
; FIXME: The AMDGPUPromoteAlloca pass should be able to convert this
; alloca to a vector. It currently fails because it does not know how
; to interpret:
; getelementptr [4 x i32]* %alloca, i32 1, i32 %b
-; SI: DS_WRITE_B32 {{v[0-9]+}}, [[PTRREG]]
+
+; SI-PROMOTE: v_add_i32_e32 [[PTRREG:v[0-9]+]], 16
+; SI-PROMOTE: ds_write_b32 [[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