Index: lib/Target/AMDGPU/SIInstrInfo.td =================================================================== --- lib/Target/AMDGPU/SIInstrInfo.td +++ lib/Target/AMDGPU/SIInstrInfo.td @@ -932,28 +932,30 @@ class getVALUDstForVT { RegisterOperand ret = !if(!eq(VT.Size, 32), VOPDstOperand, !if(!eq(VT.Size, 64), VOPDstOperand, - VOPDstOperand)); // else VT == i1 + !if(!eq(VT.Size, 16), VOPDstOperand, + VOPDstOperand))); // else VT == i1 } // Returns the register class to use for source 0 of VOP[12C] // instructions for the given VT. class getVOPSrc0ForVT { - RegisterOperand ret = !if(!eq(VT.Size, 32), VSrc_32, VSrc_64); + RegisterOperand ret = !if(!eq(VT.Size, 64), VSrc_64, VSrc_32); } // Returns the register class to use for source 1 of VOP[12C] for the // given VT. class getVOPSrc1ForVT { - RegisterClass ret = !if(!eq(VT.Size, 32), VGPR_32, VReg_64); + RegisterClass ret = !if(!eq(VT.Size, 64), VReg_64, VGPR_32); } // Returns the register class to use for sources of VOP3 instructions for the // given VT. class getVOP3SrcForVT { - RegisterOperand ret = !if(!eq(VT.Size, 32), VCSrc_32, VCSrc_64); + RegisterOperand ret = !if(!eq(VT.Size, 64), VCSrc_64, VCSrc_32); } // Returns 1 if the source arguments have modifiers, 0 if they do not. +// XXX - do f16 instructions? class hasModifiers { bit ret = !if(!eq(SrcVT.Value, f32.Value), 1, !if(!eq(SrcVT.Value, f64.Value), 1, 0)); @@ -1060,12 +1062,12 @@ // FIXME: I think these F16/I16 profiles will need to use f16/i16 types in order // for the instruction patterns to work. -def VOP_F16_F16 : VOPProfile <[f32, f32, untyped, untyped]>; -def VOP_F16_I16 : VOPProfile <[f32, i32, untyped, untyped]>; -def VOP_I16_F16 : VOPProfile <[i32, f32, untyped, untyped]>; +def VOP_F16_F16 : VOPProfile <[f16, f16, untyped, untyped]>; +def VOP_F16_I16 : VOPProfile <[f16, i32, untyped, untyped]>; +def VOP_I16_F16 : VOPProfile <[i32, f16, untyped, untyped]>; -def VOP_F16_F16_F16 : VOPProfile <[f32, f32, f32, untyped]>; -def VOP_F16_F16_I16 : VOPProfile <[f32, f32, i32, untyped]>; +def VOP_F16_F16_F16 : VOPProfile <[f16, f16, f16, untyped]>; +def VOP_F16_F16_I16 : VOPProfile <[f16, f16, i32, untyped]>; def VOP_I16_I16_I16 : VOPProfile <[i32, i32, i32, untyped]>; def VOP_F32_F32 : VOPProfile <[f32, f32, untyped, untyped]>;