diff options
| author | Tom Stellard <tstellar@redhat.com> | 2019-05-15 05:35:34 +0000 |
|---|---|---|
| committer | Tom Stellard <tstellar@redhat.com> | 2019-05-15 05:35:34 +0000 |
| commit | 352b154f1912df8e1afa2e2c79eb318a084abaf2 (patch) | |
| tree | 428121ae458578674ef5c4de808d7c9fb410d0ba | |
| parent | 195f8d2e4c9fb163e9ccf7aa392c64b1b47da294 (diff) | |
| download | llvm-352b154f1912df8e1afa2e2c79eb318a084abaf2.tar.gz | |
Merging r359883:
------------------------------------------------------------------------
r359883 | arsenm | 2019-05-03 06:42:56 -0700 (Fri, 03 May 2019) | 6 lines
AMDGPU: Fix incorrect commute with sub when folding immediates
When a fold of an immediate into a sub/subrev required shrinking the
instruction, the wrong VOP2 opcode was used. This was using the VOP2
equivalent of the original instruction, not the commuted instruction
with the inverted opcode.
------------------------------------------------------------------------
git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_80@360752 91177308-0d34-0410-b5e6-96231b3b80d8
| -rw-r--r-- | lib/Target/AMDGPU/SIFoldOperands.cpp | 5 | ||||
| -rw-r--r-- | test/CodeGen/AMDGPU/fold-immediate-operand-shrink.mir | 16 |
2 files changed, 12 insertions, 9 deletions
diff --git a/lib/Target/AMDGPU/SIFoldOperands.cpp b/lib/Target/AMDGPU/SIFoldOperands.cpp index f4e866958369..4e29b73f6fac 100644 --- a/lib/Target/AMDGPU/SIFoldOperands.cpp +++ b/lib/Target/AMDGPU/SIFoldOperands.cpp @@ -357,7 +357,10 @@ static bool tryAddToFoldList(SmallVectorImpl<FoldCandidate> &FoldList, assert(MI->getOperand(1).isDef()); - int Op32 = AMDGPU::getVOPe32(Opc); + // Make sure to get the 32-bit version of the commuted opcode. + unsigned MaybeCommutedOpc = MI->getOpcode(); + int Op32 = AMDGPU::getVOPe32(MaybeCommutedOpc); + FoldList.push_back(FoldCandidate(MI, CommuteOpNo, OpToFold, true, Op32)); return true; diff --git a/test/CodeGen/AMDGPU/fold-immediate-operand-shrink.mir b/test/CodeGen/AMDGPU/fold-immediate-operand-shrink.mir index 847c2b720cd4..e4ea36f4b1fb 100644 --- a/test/CodeGen/AMDGPU/fold-immediate-operand-shrink.mir +++ b/test/CodeGen/AMDGPU/fold-immediate-operand-shrink.mir @@ -250,8 +250,8 @@ body: | ; GCN-LABEL: name: shrink_scalar_imm_vgpr_v_sub_i32_e64_no_carry_out_use ; GCN: [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0 = S_MOV_B32 12345 ; GCN: [[DEF:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF - ; GCN: [[V_SUBREV_I32_e32_:%[0-9]+]]:vgpr_32 = V_SUBREV_I32_e32 [[S_MOV_B32_]], [[DEF]], implicit-def $vcc, implicit $exec - ; GCN: S_ENDPGM implicit [[V_SUBREV_I32_e32_]] + ; GCN: [[V_SUB_I32_e32_:%[0-9]+]]:vgpr_32 = V_SUB_I32_e32 [[S_MOV_B32_]], [[DEF]], implicit-def $vcc, implicit $exec + ; GCN: S_ENDPGM implicit [[V_SUB_I32_e32_]] %0:sreg_32_xm0 = S_MOV_B32 12345 %1:vgpr_32 = IMPLICIT_DEF %2:vgpr_32, %3:sreg_64 = V_SUB_I32_e64 %0, %1, implicit $exec @@ -269,8 +269,8 @@ body: | ; GCN-LABEL: name: shrink_vgpr_scalar_imm_v_sub_i32_e64_no_carry_out_use ; GCN: [[DEF:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF ; GCN: [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0 = S_MOV_B32 12345 - ; GCN: [[V_SUB_I32_e32_:%[0-9]+]]:vgpr_32 = V_SUB_I32_e32 [[S_MOV_B32_]], [[DEF]], implicit-def $vcc, implicit $exec - ; GCN: S_ENDPGM implicit [[V_SUB_I32_e32_]] + ; GCN: [[V_SUBREV_I32_e32_:%[0-9]+]]:vgpr_32 = V_SUBREV_I32_e32 [[S_MOV_B32_]], [[DEF]], implicit-def $vcc, implicit $exec + ; GCN: S_ENDPGM implicit [[V_SUBREV_I32_e32_]] %0:vgpr_32 = IMPLICIT_DEF %1:sreg_32_xm0 = S_MOV_B32 12345 %2:vgpr_32, %3:sreg_64 = V_SUB_I32_e64 %0, %1, implicit $exec @@ -288,8 +288,8 @@ body: | ; GCN-LABEL: name: shrink_scalar_imm_vgpr_v_subrev_i32_e64_no_carry_out_use ; GCN: [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0 = S_MOV_B32 12345 ; GCN: [[DEF:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF - ; GCN: [[V_SUB_I32_e32_:%[0-9]+]]:vgpr_32 = V_SUB_I32_e32 [[S_MOV_B32_]], [[DEF]], implicit-def $vcc, implicit $exec - ; GCN: S_ENDPGM implicit [[V_SUB_I32_e32_]] + ; GCN: [[V_SUBREV_I32_e32_:%[0-9]+]]:vgpr_32 = V_SUBREV_I32_e32 [[S_MOV_B32_]], [[DEF]], implicit-def $vcc, implicit $exec + ; GCN: S_ENDPGM implicit [[V_SUBREV_I32_e32_]] %0:sreg_32_xm0 = S_MOV_B32 12345 %1:vgpr_32 = IMPLICIT_DEF %2:vgpr_32, %3:sreg_64 = V_SUBREV_I32_e64 %0, %1, implicit $exec @@ -307,8 +307,8 @@ body: | ; GCN-LABEL: name: shrink_vgpr_scalar_imm_v_subrev_i32_e64_no_carry_out_use ; GCN: [[DEF:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF ; GCN: [[S_MOV_B32_:%[0-9]+]]:sreg_32_xm0 = S_MOV_B32 12345 - ; GCN: [[V_SUBREV_I32_e32_:%[0-9]+]]:vgpr_32 = V_SUBREV_I32_e32 [[S_MOV_B32_]], [[DEF]], implicit-def $vcc, implicit $exec - ; GCN: S_ENDPGM implicit [[V_SUBREV_I32_e32_]] + ; GCN: [[V_SUB_I32_e32_:%[0-9]+]]:vgpr_32 = V_SUB_I32_e32 [[S_MOV_B32_]], [[DEF]], implicit-def $vcc, implicit $exec + ; GCN: S_ENDPGM implicit [[V_SUB_I32_e32_]] %0:vgpr_32 = IMPLICIT_DEF %1:sreg_32_xm0 = S_MOV_B32 12345 %2:vgpr_32, %3:sreg_64 = V_SUBREV_I32_e64 %0, %1, implicit $exec |