RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-19 11:02:59 +02:00
Code Issues Projects Releases Wiki Activity

AMDGPU/R600: Fold global address operand

Browse Source 
Reviewers: tstellard

Subscribers: arsenm

Differential Revision: http://reviews.llvm.org/D19793

llvm-svn: 269480
This commit is contained in:
Jan Vesely 2016-05-13 20:39:31 +00:00
parent 8960f3da2c
commit 2161d1ce12
2 changed files with 11 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
lib/Target/AMDGPU/R600ISelLowering.cpp
View File

@ -2250,6 +2250,13 @@ FoldOperand(SDNode *ParentNode, unsigned SrcIdx, SDValue &Src, SDValue &Neg,
Src = DAG.getRegister(AMDGPU::ALU_CONST, MVT::f32); Src = DAG.getRegister(AMDGPU::ALU_CONST, MVT::f32);
return true; return true;
} }
case AMDGPU::MOV_IMM_GLOBAL_ADDR:
// Check if the Imm slot is used. Taken from below.
if (cast<ConstantSDNode>(Imm)->getZExtValue())
return false;
Imm = Src.getOperand(0);
Src = DAG.getRegister(AMDGPU::ALU_LITERAL_X, MVT::i32);
return true;
case AMDGPU::MOV_IMM_I32: case AMDGPU::MOV_IMM_I32:
case AMDGPU::MOV_IMM_F32: { case AMDGPU::MOV_IMM_F32: {
unsigned ImmReg = AMDGPU::ALU_LITERAL_X; unsigned ImmReg = AMDGPU::ALU_LITERAL_X;

4
test/CodeGen/AMDGPU/gv-const-addrspace.ll
View File

@ -14,6 +14,7 @@
; EG: VTX_READ_32 ; EG: VTX_READ_32
; EG: @float_gv ; EG: @float_gv
; EG-NOT: MOVA_INT ; EG-NOT: MOVA_INT
; EG-NOT: MOV
define void @float(float addrspace(1)* %out, i32 %index) { define void @float(float addrspace(1)* %out, i32 %index) {
entry: entry:
%0 = getelementptr inbounds [5 x float], [5 x float] addrspace(2)* @float_gv, i32 0, i32 %index %0 = getelementptr inbounds [5 x float], [5 x float] addrspace(2)* @float_gv, i32 0, i32 %index
@ -31,6 +32,7 @@ entry:
; EG: VTX_READ_32 ; EG: VTX_READ_32
; EG: @i32_gv ; EG: @i32_gv
; EG-NOT: MOVA_INT ; EG-NOT: MOVA_INT
; EG-NOT: MOV
define void @i32(i32 addrspace(1)* %out, i32 %index) { define void @i32(i32 addrspace(1)* %out, i32 %index) {
entry: entry:
%0 = getelementptr inbounds [5 x i32], [5 x i32] addrspace(2)* @i32_gv, i32 0, i32 %index %0 = getelementptr inbounds [5 x i32], [5 x i32] addrspace(2)* @i32_gv, i32 0, i32 %index
@ -50,6 +52,7 @@ entry:
; EG: VTX_READ_32 ; EG: VTX_READ_32
; EG: @struct_foo_gv ; EG: @struct_foo_gv
; EG-NOT: MOVA_INT ; EG-NOT: MOVA_INT
; EG-NOT: MOV
define void @struct_foo_gv_load(i32 addrspace(1)* %out, i32 %index) { define void @struct_foo_gv_load(i32 addrspace(1)* %out, i32 %index) {
%gep = getelementptr inbounds [1 x %struct.foo], [1 x %struct.foo] addrspace(2)* @struct_foo_gv, i32 0, i32 0, i32 1, i32 %index %gep = getelementptr inbounds [1 x %struct.foo], [1 x %struct.foo] addrspace(2)* @struct_foo_gv, i32 0, i32 0, i32 1, i32 %index
%load = load i32, i32 addrspace(2)* %gep, align 4 %load = load i32, i32 addrspace(2)* %gep, align 4
@ -68,6 +71,7 @@ define void @struct_foo_gv_load(i32 addrspace(1)* %out, i32 %index) {
; EG: VTX_READ_32 ; EG: VTX_READ_32
; EG: @array_v1_gv ; EG: @array_v1_gv
; EG-NOT: MOVA_INT ; EG-NOT: MOVA_INT
; EG-NOT: MOV
define void @array_v1_gv_load(<1 x i32> addrspace(1)* %out, i32 %index) { define void @array_v1_gv_load(<1 x i32> addrspace(1)* %out, i32 %index) {
%gep = getelementptr inbounds [4 x <1 x i32>], [4 x <1 x i32>] addrspace(2)* @array_v1_gv, i32 0, i32 %index %gep = getelementptr inbounds [4 x <1 x i32>], [4 x <1 x i32>] addrspace(2)* @array_v1_gv, i32 0, i32 %index
%load = load <1 x i32>, <1 x i32> addrspace(2)* %gep, align 4 %load = load <1 x i32>, <1 x i32> addrspace(2)* %gep, align 4