diff options
Diffstat (limited to 'llvm/lib/Target/AMDGPU/VOP2Instructions.td')
| -rw-r--r-- | llvm/lib/Target/AMDGPU/VOP2Instructions.td | 19 |
1 files changed, 15 insertions, 4 deletions
diff --git a/llvm/lib/Target/AMDGPU/VOP2Instructions.td b/llvm/lib/Target/AMDGPU/VOP2Instructions.td index f136a43..c001c5d 100644 --- a/llvm/lib/Target/AMDGPU/VOP2Instructions.td +++ b/llvm/lib/Target/AMDGPU/VOP2Instructions.td @@ -503,6 +503,7 @@ def VOP_MAC_F16_t16 : VOP_MAC <f16> { dpp8:$dpp8, Dpp8FI:$fi); let Src2Mod = FP32InputMods; // dummy unused modifiers let Src2RC64 = VGPRSrc_32; // stub argument + let Src1ModVOP3DPP = getSrcModVOP3DPP<Src1VT, 1/*IsFake16*/>.ret; } def VOP_MAC_F32 : VOP_MAC <f32>; let HasExtDPP = 0, HasExt32BitDPP = 0 in @@ -618,7 +619,7 @@ class VOP2e_SGPR<list<ValueType> ArgVT> : VOPProfile<ArgVT> { let AsmVOP3Base = "$vdst, $src0_modifiers, $src1_modifiers, $src2"; let Outs32 = (outs DstRC:$vdst); - let Outs64 = (outs DstRC:$vdst); + let Outs64 = (outs DstRC64:$vdst); // Suppress src2 implied by type since the 32-bit encoding uses an // implicit VCC use. @@ -652,7 +653,7 @@ class VOP2e_SGPR<list<ValueType> ArgVT> : VOPProfile<ArgVT> { dpp8:$dpp8, Dpp8FI:$fi); let Src0ModVOP3DPP = FPVRegInputMods; - let Src1ModVOP3DPP = FPVRegInputMods; + let Src1ModVOP3DPP = FP32VCSrcInputMods; let HasExt = 1; let HasExtDPP = 1; @@ -662,7 +663,17 @@ class VOP2e_SGPR<list<ValueType> ArgVT> : VOPProfile<ArgVT> { } def VOP2e_I32_I32_I32_I1 : VOP2e_SGPR<[i32, i32, i32, i1]>; -def VOP2e_I16_I16_I16_I1 : VOP2e_SGPR<[i16, i16, i16, i1]>; +def VOP2e_I16_I16_I16_I1_fake16 : VOP2e_SGPR<[i16, i16, i16, i1]> { + let IsTrue16 = 1; + let DstRC64 = getVALUDstForVT<DstVT>.ret; + + let Src0Mod = getSrcMod<f16>.ret; + let Src1Mod = getSrcMod<f16>.ret; + + let Src0VOP3DPP = VGPRSrc_32; + let Src1VOP3DPP = getVOP3DPPSrcForVT<Src1VT>.ret; + let Src1ModVOP3DPP = getSrcModVOP3DPP<f16, 1/*IsFake16*/>.ret; +} def VOP_READLANE : VOPProfile<[i32, i32, i32, untyped]> { let Outs32 = (outs SReg_32:$vdst); @@ -703,7 +714,7 @@ def VOP_WRITELANE : VOPProfile<[i32, i32, i32, i32]> { //===----------------------------------------------------------------------===// let SubtargetPredicate = isGFX11Plus in -defm V_CNDMASK_B16 : VOP2eInst <"v_cndmask_b16", VOP2e_I16_I16_I16_I1>; +defm V_CNDMASK_B16 : VOP2eInst <"v_cndmask_b16", VOP2e_I16_I16_I16_I1_fake16>; defm V_CNDMASK_B32 : VOP2eInst_VOPD <"v_cndmask_b32", VOP2e_I32_I32_I32_I1, 0x9, "v_cndmask_b32">; let SubtargetPredicate = HasMadMacF32Insts, isReMaterializable = 1 in def V_MADMK_F32 : VOP2_Pseudo <"v_madmk_f32", VOP_MADMK_F32, []>; |