maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

AMDGPU/GlobalISel: Disallow VGPR->SCC copies 

This fixes using scalar adds when only the carry in is a VGPR
using greedy regbankselect.

llvm-svn: 350593
This commit is contained in:
Matt Arsenault 2019-01-08 01:13:20 +00:00
parent 68c668a5f3
commit ae6f1e07fc
5 changed files with 24 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
llvm/lib/Target/AMDGPU/AMDGPURegisterBankInfo.cpp
View File

@ -74,13 +74,19 @@ unsigned AMDGPURegisterBankInfo::copyCost(const RegisterBank &Dst,
const RegisterBank &Src,
unsigned Size) const {
if (Dst.getID() == AMDGPU::SGPRRegBankID &&
Src.getID() == AMDGPU::VGPRRegBankID)
Src.getID() == AMDGPU::VGPRRegBankID) {
// For boolean values, we can do a v_cmp to "copy" a VGPR to VCC.
if (Size == 1)
return 1;
return std::numeric_limits<unsigned>::max();
}
// SGPRRegBank with size 1 is actually vcc or another 64-bit sgpr written by
// the valu.
if (Size == 1 && Dst.getID() == AMDGPU::SCCRegBankID &&
Src.getID() == AMDGPU::SGPRRegBankID)
(Src.getID() == AMDGPU::SGPRRegBankID ||
Src.getID() == AMDGPU::VGPRRegBankID))
return std::numeric_limits<unsigned>::max();
return RegisterBankInfo::copyCost(Dst, Src, Size);

6
llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sadde.mir
View File

@ -84,8 +84,10 @@ body: |
; GREEDY: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1
; GREEDY: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
; GREEDY: [[TRUNC:%[0-9]+]]:vgpr(s1) = G_TRUNC [[COPY2]](s32)
; GREEDY: [[COPY3:%[0-9]+]]:scc(s1) = COPY [[TRUNC]](s1)
; GREEDY: [[SADDE:%[0-9]+]]:sgpr(s32), [[SADDE1:%[0-9]+]]:scc(s1) = G_SADDE [[COPY]], [[COPY1]], [[COPY3]]
; GREEDY: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY]](s32)
; GREEDY: [[COPY4:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32)
; GREEDY: [[COPY5:%[0-9]+]]:sgpr(s1) = COPY [[TRUNC]](s1)
; GREEDY: [[SADDE:%[0-9]+]]:vgpr(s32), [[SADDE1:%[0-9]+]]:sgpr(s1) = G_SADDE [[COPY3]], [[COPY4]], [[COPY5]]
%0:_(s32) = COPY $sgpr0
%1:_(s32) = COPY $sgpr1
%2:_(s32) = COPY $vgpr0

6
llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-ssube.mir
View File

@ -84,8 +84,10 @@ body: |
; GREEDY: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1
; GREEDY: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
; GREEDY: [[TRUNC:%[0-9]+]]:vgpr(s1) = G_TRUNC [[COPY2]](s32)
; GREEDY: [[COPY3:%[0-9]+]]:scc(s1) = COPY [[TRUNC]](s1)
; GREEDY: [[SSUBE:%[0-9]+]]:sgpr(s32), [[SSUBE1:%[0-9]+]]:scc(s1) = G_SSUBE [[COPY]], [[COPY1]], [[COPY3]]
; GREEDY: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY]](s32)
; GREEDY: [[COPY4:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32)
; GREEDY: [[COPY5:%[0-9]+]]:sgpr(s1) = COPY [[TRUNC]](s1)
; GREEDY: [[SSUBE:%[0-9]+]]:vgpr(s32), [[SSUBE1:%[0-9]+]]:sgpr(s1) = G_SSUBE [[COPY3]], [[COPY4]], [[COPY5]]
%0:_(s32) = COPY $sgpr0
%1:_(s32) = COPY $sgpr1
%2:_(s32) = COPY $vgpr0

6
llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-uadde.mir
View File

@ -83,8 +83,10 @@ body: |
; GREEDY: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1
; GREEDY: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
; GREEDY: [[TRUNC:%[0-9]+]]:vgpr(s1) = G_TRUNC [[COPY2]](s32)
; GREEDY: [[COPY3:%[0-9]+]]:scc(s1) = COPY [[TRUNC]](s1)
; GREEDY: [[UADDE:%[0-9]+]]:sgpr(s32), [[UADDE1:%[0-9]+]]:scc(s1) = G_UADDE [[COPY]], [[COPY1]], [[COPY3]]
; GREEDY: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY]](s32)
; GREEDY: [[COPY4:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32)
; GREEDY: [[COPY5:%[0-9]+]]:sgpr(s1) = COPY [[TRUNC]](s1)
; GREEDY: [[UADDE:%[0-9]+]]:vgpr(s32), [[UADDE1:%[0-9]+]]:sgpr(s1) = G_UADDE [[COPY3]], [[COPY4]], [[COPY5]]
%0:_(s32) = COPY $sgpr0
%1:_(s32) = COPY $sgpr1
%2:_(s32) = COPY $vgpr0

6
llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-usube.mir
View File

@ -84,8 +84,10 @@ body: |
; GREEDY: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1
; GREEDY: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr0
; GREEDY: [[TRUNC:%[0-9]+]]:vgpr(s1) = G_TRUNC [[COPY2]](s32)
; GREEDY: [[COPY3:%[0-9]+]]:scc(s1) = COPY [[TRUNC]](s1)
; GREEDY: [[USUBE:%[0-9]+]]:sgpr(s32), [[USUBE1:%[0-9]+]]:scc(s1) = G_USUBE [[COPY]], [[COPY1]], [[COPY3]]
; GREEDY: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY]](s32)
; GREEDY: [[COPY4:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32)
; GREEDY: [[COPY5:%[0-9]+]]:sgpr(s1) = COPY [[TRUNC]](s1)
; GREEDY: [[USUBE:%[0-9]+]]:vgpr(s32), [[USUBE1:%[0-9]+]]:sgpr(s1) = G_USUBE [[COPY3]], [[COPY4]], [[COPY5]]
%0:_(s32) = COPY $sgpr0
%1:_(s32) = COPY $sgpr1
%2:_(s32) = COPY $vgpr0