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

AMDGPU/GlobalISel: Handle VOP3NoMods

This commit is contained in:
Matt Arsenault 2020-01-26 00:25:35 -05:00
parent d309b4ebe4
commit e60d658260
5 changed files with 35 additions and 16 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
llvm/lib/Target/AMDGPU/AMDGPUGISel.td
View File

@ -31,6 +31,10 @@ def gi_vop3mods :
GIComplexOperandMatcher<s32, "selectVOP3Mods">,
GIComplexPatternEquiv<VOP3Mods>;
def gi_vop3_no_mods :
GIComplexOperandMatcher<s32, "selectVOP3NoMods">,
GIComplexPatternEquiv<VOP3NoMods>;
def gi_vop3mods_nnan :
GIComplexOperandMatcher<s32, "selectVOP3Mods_nnan">,
GIComplexPatternEquiv<VOP3Mods_nnan>;

12
llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp
View File

@ -2218,6 +2218,18 @@ AMDGPUInstructionSelector::selectVOP3Mods(MachineOperand &Root) const {
}};
}
InstructionSelector::ComplexRendererFns
AMDGPUInstructionSelector::selectVOP3NoMods(MachineOperand &Root) const {
Register Reg = Root.getReg();
const MachineInstr *Def = getDefIgnoringCopies(Reg, *MRI);
if (Def && (Def->getOpcode() == AMDGPU::G_FNEG ||
Def->getOpcode() == AMDGPU::G_FABS))
return {};
return {{
[=](MachineInstrBuilder &MIB) { MIB.addReg(Reg); },
}};
}
InstructionSelector::ComplexRendererFns
AMDGPUInstructionSelector::selectVOP3Mods_nnan(MachineOperand &Root) const {
Register Src;

3
llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.h
View File

@ -142,6 +142,9 @@ private:
selectVOP3OMods(MachineOperand &Root) const;
InstructionSelector::ComplexRendererFns
selectVOP3Mods(MachineOperand &Root) const;
ComplexRendererFns selectVOP3NoMods(MachineOperand &Root) const;
InstructionSelector::ComplexRendererFns
selectVOP3Mods_nnan(MachineOperand &Root) const;

24
llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-fma.s32.mir
View File

@ -23,15 +23,15 @@ body: |
; GFX9-DL: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GFX9-DL: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; GFX9-DL: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX9-DL: [[V_FMA_F32_:%[0-9]+]]:vgpr_32 = V_FMA_F32 0, [[COPY]], 0, [[COPY1]], 0, [[COPY2]], 0, 0, implicit $exec
; GFX9-DL: S_ENDPGM 0, implicit [[V_FMA_F32_]]
; GFX9-DL: [[V_FMAC_F32_e64_:%[0-9]+]]:vgpr_32 = V_FMAC_F32_e64 0, [[COPY]], 0, [[COPY1]], 0, [[COPY2]], 0, 0, implicit $exec
; GFX9-DL: S_ENDPGM 0, implicit [[V_FMAC_F32_e64_]]
; GFX10-LABEL: name: fma_f32
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX10: [[V_FMA_F32_:%[0-9]+]]:vgpr_32 = V_FMA_F32 0, [[COPY]], 0, [[COPY1]], 0, [[COPY2]], 0, 0, implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_FMA_F32_]]
; GFX10: [[V_FMAC_F32_e64_:%[0-9]+]]:vgpr_32 = V_FMAC_F32_e64 0, [[COPY]], 0, [[COPY1]], 0, [[COPY2]], 0, 0, implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_FMAC_F32_e64_]]
%0:vgpr(s32) = COPY $vgpr0
%1:vgpr(s32) = COPY $vgpr1
%2:vgpr(s32) = COPY $vgpr2
@ -60,15 +60,15 @@ body: |
; GFX9-DL: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GFX9-DL: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; GFX9-DL: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX9-DL: [[V_FMA_F32_:%[0-9]+]]:vgpr_32 = V_FMA_F32 1, [[COPY]], 0, [[COPY1]], 0, [[COPY2]], 0, 0, implicit $exec
; GFX9-DL: S_ENDPGM 0, implicit [[V_FMA_F32_]]
; GFX9-DL: [[V_FMAC_F32_e64_:%[0-9]+]]:vgpr_32 = V_FMAC_F32_e64 1, [[COPY]], 0, [[COPY1]], 0, [[COPY2]], 0, 0, implicit $exec
; GFX9-DL: S_ENDPGM 0, implicit [[V_FMAC_F32_e64_]]
; GFX10-LABEL: name: fma_f32_fneg_src0
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX10: [[V_FMA_F32_:%[0-9]+]]:vgpr_32 = V_FMA_F32 1, [[COPY]], 0, [[COPY1]], 0, [[COPY2]], 0, 0, implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_FMA_F32_]]
; GFX10: [[V_FMAC_F32_e64_:%[0-9]+]]:vgpr_32 = V_FMAC_F32_e64 1, [[COPY]], 0, [[COPY1]], 0, [[COPY2]], 0, 0, implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_FMAC_F32_e64_]]
%0:vgpr(s32) = COPY $vgpr0
%1:vgpr(s32) = COPY $vgpr1
%2:vgpr(s32) = COPY $vgpr2
@ -98,15 +98,15 @@ body: |
; GFX9-DL: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GFX9-DL: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; GFX9-DL: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX9-DL: [[V_FMA_F32_:%[0-9]+]]:vgpr_32 = V_FMA_F32 0, [[COPY]], 1, [[COPY1]], 0, [[COPY2]], 0, 0, implicit $exec
; GFX9-DL: S_ENDPGM 0, implicit [[V_FMA_F32_]]
; GFX9-DL: [[V_FMAC_F32_e64_:%[0-9]+]]:vgpr_32 = V_FMAC_F32_e64 0, [[COPY]], 1, [[COPY1]], 0, [[COPY2]], 0, 0, implicit $exec
; GFX9-DL: S_ENDPGM 0, implicit [[V_FMAC_F32_e64_]]
; GFX10-LABEL: name: fma_f32_fneg_src1
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX10: [[V_FMA_F32_:%[0-9]+]]:vgpr_32 = V_FMA_F32 0, [[COPY]], 1, [[COPY1]], 0, [[COPY2]], 0, 0, implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_FMA_F32_]]
; GFX10: [[V_FMAC_F32_e64_:%[0-9]+]]:vgpr_32 = V_FMAC_F32_e64 0, [[COPY]], 1, [[COPY1]], 0, [[COPY2]], 0, 0, implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_FMAC_F32_e64_]]
%0:vgpr(s32) = COPY $vgpr0
%1:vgpr(s32) = COPY $vgpr1
%2:vgpr(s32) = COPY $vgpr2

8
llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-fmad.s32.mir
View File

@ -16,15 +16,15 @@ body: |
; GFX6: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GFX6: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; GFX6: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX6: [[V_MAD_F32_:%[0-9]+]]:vgpr_32 = V_MAD_F32 0, [[COPY]], 0, [[COPY1]], 0, [[COPY2]], 0, 0, implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_MAD_F32_]]
; GFX6: [[V_MAC_F32_e64_:%[0-9]+]]:vgpr_32 = V_MAC_F32_e64 0, [[COPY]], 0, [[COPY1]], 0, [[COPY2]], 0, 0, implicit $exec
; GFX6: S_ENDPGM 0, implicit [[V_MAC_F32_e64_]]
; GFX10-LABEL: name: fmad_f32
; GFX10: $vcc_hi = IMPLICIT_DEF
; GFX10: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0
; GFX10: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1
; GFX10: [[COPY2:%[0-9]+]]:vgpr_32 = COPY $vgpr2
; GFX10: [[V_MAD_F32_:%[0-9]+]]:vgpr_32 = V_MAD_F32 0, [[COPY]], 0, [[COPY1]], 0, [[COPY2]], 0, 0, implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_MAD_F32_]]
; GFX10: [[V_MAC_F32_e64_:%[0-9]+]]:vgpr_32 = V_MAC_F32_e64 0, [[COPY]], 0, [[COPY1]], 0, [[COPY2]], 0, 0, implicit $exec
; GFX10: S_ENDPGM 0, implicit [[V_MAC_F32_e64_]]
%0:vgpr(s32) = COPY $vgpr0
%1:vgpr(s32) = COPY $vgpr1
%2:vgpr(s32) = COPY $vgpr2