RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-20 03:23:01 +02:00
Code Issues Projects Releases Wiki Activity

AMDGPU: Handle or in multi-use shl ptr combine

Browse Source 
llvm-svn: 318223
This commit is contained in:
Matt Arsenault 2017-11-14 23:46:42 +00:00
parent 3cc4cef33b
commit 66ee956d8d
2 changed files with 36 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
lib/Target/AMDGPU/SIISelLowering.cpp
View File

@ -5199,8 +5199,8 @@ SDValue SITargetLowering::performSHLPtrCombine(SDNode *N,
// We only do this to handle cases where it's profitable when there are
// multiple uses of the add, so defer to the standard combine.
// TODO: Support or
if (N0.getOpcode() != ISD::ADD || N0->hasOneUse())
if ((N0.getOpcode() != ISD::ADD && N0.getOpcode() != ISD::OR) ||
N0->hasOneUse())
return SDValue();
const ConstantSDNode *CN1 = dyn_cast<ConstantSDNode>(N1);

34
test/CodeGen/AMDGPU/shl_add_ptr.ll
View File

@ -384,5 +384,39 @@ define void @shl_add_ptr_combine_2use_both_max_private_offset(i16 zeroext %idx.a
ret void
}
; GCN-LABEL: {{^}}shl_or_ptr_combine_2use_lds:
; GCN: v_lshlrev_b32_e32 [[SCALE0:v[0-9]+]], 3, v0
; GCN: ds_write_b32 [[SCALE0]], v{{[0-9]+}} offset:32
; GCN: v_lshlrev_b32_e32 [[SCALE1:v[0-9]+]], 4, v0
; GCN: ds_write_b32 [[SCALE1]], v{{[0-9]+}} offset:64
define void @shl_or_ptr_combine_2use_lds(i32 %idx) #0 {
%idx.add = or i32 %idx, 4
%shl0 = shl i32 %idx.add, 3
%shl1 = shl i32 %idx.add, 4
%ptr0 = inttoptr i32 %shl0 to i32 addrspace(3)*
%ptr1 = inttoptr i32 %shl1 to i32 addrspace(3)*
store volatile i32 9, i32 addrspace(3)* %ptr0
store volatile i32 10, i32 addrspace(3)* %ptr1
ret void
}
; GCN-LABEL: {{^}}shl_or_ptr_combine_2use_max_lds_offset:
; GCN-DAG: v_lshlrev_b32_e32 [[SCALE0:v[0-9]+]], 3, v0
; GCN-DAG: v_lshlrev_b32_e32 [[SCALE1:v[0-9]+]], 4, v0
; GCN-DAG: ds_write_b32 [[SCALE0]], v{{[0-9]+}} offset:65528
; GCN-DAG: v_or_b32_e32 [[ADD1:v[0-9]+]], 0x1fff0, [[SCALE1]]
; GCN: ds_write_b32 [[ADD1]], v{{[0-9]+$}}
define void @shl_or_ptr_combine_2use_max_lds_offset(i32 %idx) #0 {
%idx.add = or i32 %idx, 8191
%shl0 = shl i32 %idx.add, 3
%shl1 = shl i32 %idx.add, 4
%ptr0 = inttoptr i32 %shl0 to i32 addrspace(3)*
%ptr1 = inttoptr i32 %shl1 to i32 addrspace(3)*
store volatile i32 9, i32 addrspace(3)* %ptr0
store volatile i32 10, i32 addrspace(3)* %ptr1
ret void
}
attributes #0 = { nounwind }
attributes #1 = { nounwind readnone }