//===-- VOPInstructions.td - Vector Instruction Definitions ---------------===// // // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. // See https://llvm.org/LICENSE.txt for license information. // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // dummies for outer let class LetDummies { bit TRANS; bit ReadsModeReg; bit mayRaiseFPException; bit isCommutable; bit isConvertibleToThreeAddress; bit isMoveImm; bit isReMaterializable; bit isAsCheapAsAMove; bit VOPAsmPrefer32Bit; bit FPDPRounding; Predicate SubtargetPredicate; string Constraints; string DisableEncoding; list SchedRW; list Uses; list Defs; } class VOP { string OpName = opName; } class VOPAnyCommon pattern> : InstSI { let mayLoad = 0; let mayStore = 0; let hasSideEffects = 0; let UseNamedOperandTable = 1; let VALU = 1; let Uses = !if(ReadsModeReg, [MODE, EXEC], [EXEC]); } class VOP_Pseudo pattern> : InstSI , VOP , SIMCInstr { let isPseudo = 1; let isCodeGenOnly = 1; let UseNamedOperandTable = 1; string Mnemonic = opName; VOPProfile Pfl = P; string AsmOperands; } class VOP3Common pattern = [], bit HasMods = 0, bit VOP3Only = 0> : VOPAnyCommon { // Using complex patterns gives VOP3 patterns a very high complexity rating, // but standalone patterns are almost always preferred, so we need to adjust the // priority lower. The goal is to use a high number to reduce complexity to // zero (or less than zero). let AddedComplexity = -1000; let VOP3 = 1; let AsmVariantName = AMDGPUAsmVariants.VOP3; let AsmMatchConverter = !if(HasMods, "cvtVOP3", ""); let isCodeGenOnly = 0; int Size = 8; // Because SGPRs may be allowed if there are multiple operands, we // need a post-isel hook to insert copies in order to avoid // violating constant bus requirements. let hasPostISelHook = 1; } class VOP3_Pseudo pattern = [], bit VOP3Only = 0, bit isVOP3P = 0, bit isVop3OpSel = 0> : VOP_Pseudo { let VOP3_OPSEL = isVop3OpSel; let IsPacked = P.IsPacked; let IsMAI = P.IsMAI; let AsmOperands = !if(isVop3OpSel, P.AsmVOP3OpSel, !if(!and(isVOP3P, P.IsPacked), P.AsmVOP3P, P.Asm64)); let Size = 8; let mayLoad = 0; let mayStore = 0; let hasSideEffects = 0; // Because SGPRs may be allowed if there are multiple operands, we // need a post-isel hook to insert copies in order to avoid // violating constant bus requirements. let hasPostISelHook = 1; // Using complex patterns gives VOP3 patterns a very high complexity rating, // but standalone patterns are almost always preferred, so we need to adjust the // priority lower. The goal is to use a high number to reduce complexity to // zero (or less than zero). let AddedComplexity = -1000; let VOP3 = 1; let VALU = 1; let FPClamp = P.HasFPClamp; let IntClamp = P.HasIntClamp; let ClampLo = P.HasClampLo; let ClampHi = P.HasClampHi; let ReadsModeReg = !or(isFloatType.ret, isFloatType.ret); let mayRaiseFPException = ReadsModeReg; let Uses = !if(ReadsModeReg, [MODE, EXEC], [EXEC]); let AsmVariantName = AMDGPUAsmVariants.VOP3; let AsmMatchConverter = !if(isVOP3P, "cvtVOP3P", !if(!or(P.HasModifiers, P.HasOMod, P.HasIntClamp), "cvtVOP3", "")); } class VOP3P_Pseudo pattern = []> : VOP3_Pseudo { let VOP3P = 1; } class VOP3_Real : InstSI , SIMCInstr { let isPseudo = 0; let isCodeGenOnly = 0; let UseNamedOperandTable = 1; let Constraints = ps.Constraints; let DisableEncoding = ps.DisableEncoding; // copy relevant pseudo op flags let SubtargetPredicate = ps.SubtargetPredicate; let OtherPredicates = ps.OtherPredicates; let AsmMatchConverter = ps.AsmMatchConverter; let AsmVariantName = ps.AsmVariantName; let Constraints = ps.Constraints; let DisableEncoding = ps.DisableEncoding; let TSFlags = ps.TSFlags; let UseNamedOperandTable = ps.UseNamedOperandTable; let Uses = ps.Uses; let Defs = ps.Defs; VOPProfile Pfl = ps.Pfl; } // XXX - Is there any reason to distinguish this from regular VOP3 // here? class VOP3P_Real : VOP3_Real; class VOP3a : Enc64 { bits<4> src0_modifiers; bits<9> src0; bits<3> src1_modifiers; bits<9> src1; bits<3> src2_modifiers; bits<9> src2; bits<1> clamp; bits<2> omod; let Inst{8} = !if(P.HasSrc0Mods, src0_modifiers{1}, 0); let Inst{9} = !if(P.HasSrc1Mods, src1_modifiers{1}, 0); let Inst{10} = !if(P.HasSrc2Mods, src2_modifiers{1}, 0); let Inst{31-26} = 0x34; //encoding let Inst{40-32} = !if(P.HasSrc0, src0, 0); let Inst{49-41} = !if(P.HasSrc1, src1, 0); let Inst{58-50} = !if(P.HasSrc2, src2, 0); let Inst{60-59} = !if(P.HasOMod, omod, 0); let Inst{61} = !if(P.HasSrc0Mods, src0_modifiers{0}, 0); let Inst{62} = !if(P.HasSrc1Mods, src1_modifiers{0}, 0); let Inst{63} = !if(P.HasSrc2Mods, src2_modifiers{0}, 0); } class VOP3a_gfx6_gfx7 op, VOPProfile p> : VOP3a

{ let Inst{11} = !if(p.HasClamp, clamp{0}, 0); let Inst{25-17} = op; } class VOP3a_gfx10 op, VOPProfile p> : VOP3a

{ let Inst{15} = !if(p.HasClamp, clamp{0}, 0); let Inst{25-16} = op; let Inst{31-26} = 0x35; } class VOP3a_vi op, VOPProfile P> : VOP3a

{ let Inst{25-16} = op; let Inst{15} = !if(P.HasClamp, clamp{0}, 0); } class VOP3e_gfx6_gfx7 op, VOPProfile p> : VOP3a_gfx6_gfx7 { bits<8> vdst; let Inst{7-0} = !if(p.EmitDst, vdst{7-0}, 0); } class VOP3e_gfx10 op, VOPProfile p> : VOP3a_gfx10 { bits<8> vdst; let Inst{7-0} = !if(p.EmitDst, vdst{7-0}, 0); } class VOP3e_vi op, VOPProfile P> : VOP3a_vi { bits<8> vdst; let Inst{7-0} = !if(P.EmitDst, vdst{7-0}, 0); } class VOP3OpSel_gfx9 op, VOPProfile P> : VOP3e_vi { let Inst{11} = !if(P.HasSrc0, src0_modifiers{2}, 0); let Inst{12} = !if(P.HasSrc1, src1_modifiers{2}, 0); let Inst{13} = !if(P.HasSrc2, src2_modifiers{2}, 0); let Inst{14} = !if(P.HasDst, src0_modifiers{3}, 0); } class VOP3OpSel_gfx10 op, VOPProfile p> : VOP3e_gfx10 { let Inst{11} = !if(p.HasSrc0, src0_modifiers{2}, 0); let Inst{12} = !if(p.HasSrc1, src1_modifiers{2}, 0); let Inst{13} = !if(p.HasSrc2, src2_modifiers{2}, 0); let Inst{14} = !if(p.HasDst, src0_modifiers{3}, 0); } // NB: For V_INTERP* opcodes, src0 is encoded as src1 and vice versa class VOP3Interp_vi op, VOPProfile P> : VOP3e_vi { bits<2> attrchan; bits<6> attr; bits<1> high; let Inst{8} = 0; // No modifiers for src0 let Inst{61} = 0; let Inst{9} = !if(P.HasSrc0Mods, src0_modifiers{1}, 0); let Inst{62} = !if(P.HasSrc0Mods, src0_modifiers{0}, 0); let Inst{37-32} = attr; let Inst{39-38} = attrchan; let Inst{40} = !if(P.HasHigh, high, 0); let Inst{49-41} = src0; } class VOP3Interp_gfx10 op, VOPProfile p> : VOP3e_gfx10 { bits<6> attr; bits<2> attrchan; bits<1> high; let Inst{8} = 0; let Inst{9} = !if(p.HasSrc0Mods, src0_modifiers{1}, 0); let Inst{37-32} = attr; let Inst{39-38} = attrchan; let Inst{40} = !if(p.HasHigh, high, 0); let Inst{49-41} = src0; let Inst{61} = 0; let Inst{62} = !if(p.HasSrc0Mods, src0_modifiers{0}, 0); } class VOP3be : Enc64 { bits<8> vdst; bits<2> src0_modifiers; bits<9> src0; bits<2> src1_modifiers; bits<9> src1; bits<2> src2_modifiers; bits<9> src2; bits<7> sdst; bits<2> omod; let Inst{7-0} = vdst; let Inst{14-8} = sdst; let Inst{31-26} = 0x34; //encoding let Inst{40-32} = !if(P.HasSrc0, src0, 0); let Inst{49-41} = !if(P.HasSrc1, src1, 0); let Inst{58-50} = !if(P.HasSrc2, src2, 0); let Inst{60-59} = !if(P.HasOMod, omod, 0); let Inst{61} = !if(P.HasSrc0Mods, src0_modifiers{0}, 0); let Inst{62} = !if(P.HasSrc1Mods, src1_modifiers{0}, 0); let Inst{63} = !if(P.HasSrc2Mods, src2_modifiers{0}, 0); } class VOP3Pe op, VOPProfile P> : Enc64 { bits<8> vdst; // neg, neg_hi, op_sel put in srcN_modifiers bits<4> src0_modifiers; bits<9> src0; bits<4> src1_modifiers; bits<9> src1; bits<4> src2_modifiers; bits<9> src2; bits<1> clamp; let Inst{7-0} = vdst; let Inst{8} = !if(P.HasSrc0Mods, src0_modifiers{1}, 0); // neg_hi src0 let Inst{9} = !if(P.HasSrc1Mods, src1_modifiers{1}, 0); // neg_hi src1 let Inst{10} = !if(P.HasSrc2Mods, src2_modifiers{1}, 0); // neg_hi src2 let Inst{11} = !if(!and(P.HasSrc0, P.HasOpSel), src0_modifiers{2}, 0); // op_sel(0) let Inst{12} = !if(!and(P.HasSrc1, P.HasOpSel), src1_modifiers{2}, 0); // op_sel(1) let Inst{13} = !if(!and(P.HasSrc2, P.HasOpSel), src2_modifiers{2}, 0); // op_sel(2) let Inst{14} = !if(!and(P.HasSrc2, P.HasOpSel), src2_modifiers{3}, 0); // op_sel_hi(2) let Inst{15} = !if(P.HasClamp, clamp{0}, 0); let Inst{22-16} = op; let Inst{31-23} = 0x1a7; //encoding let Inst{40-32} = !if(P.HasSrc0, src0, 0); let Inst{49-41} = !if(P.HasSrc1, src1, 0); let Inst{58-50} = !if(P.HasSrc2, src2, 0); let Inst{59} = !if(!and(P.HasSrc0, P.HasOpSel), src0_modifiers{3}, 0); // op_sel_hi(0) let Inst{60} = !if(!and(P.HasSrc1, P.HasOpSel), src1_modifiers{3}, 0); // op_sel_hi(1) let Inst{61} = !if(P.HasSrc0Mods, src0_modifiers{0}, 0); // neg (lo) let Inst{62} = !if(P.HasSrc1Mods, src1_modifiers{0}, 0); // neg (lo) let Inst{63} = !if(P.HasSrc2Mods, src2_modifiers{0}, 0); // neg (lo) } class VOP3Pe_MAI op, VOPProfile P> : Enc64 { bits<8> vdst; bits<10> src0; bits<10> src1; bits<9> src2; bits<3> blgp; bits<3> cbsz; bits<4> abid; bits<1> clamp; let Inst{7-0} = vdst; let Inst{10-8} = !if(P.HasSrc1, cbsz, 0); let Inst{14-11} = !if(P.HasSrc1, abid, 0); let Inst{15} = !if(P.HasClamp, clamp{0}, 0); let Inst{22-16} = op; let Inst{31-23} = 0x1a7; //encoding let Inst{40-32} = !if(P.HasSrc0, src0{8-0}, 0); let Inst{49-41} = !if(P.HasSrc1, src1{8-0}, 0); let Inst{58-50} = !if(P.HasSrc2, src2, 0); let Inst{59} = !if(P.HasSrc0, src0{9}, 0); // acc(0) let Inst{60} = !if(P.HasSrc1, src1{9}, 0); // acc(1) let Inst{63-61} = !if(P.HasSrc1, blgp, 0); } class VOP3Pe_gfx10 op, VOPProfile P> : VOP3Pe { let Inst{31-23} = 0x198; //encoding } class VOP3be_gfx6_gfx7 op, VOPProfile p> : VOP3be

{ let Inst{25-17} = op; } class VOP3be_gfx10 op, VOPProfile p> : VOP3be

{ bits<1> clamp; let Inst{15} = !if(p.HasClamp, clamp{0}, 0); let Inst{25-16} = op; let Inst{31-26} = 0x35; } class VOP3be_vi op, VOPProfile P> : VOP3be

{ bits<1> clamp; let Inst{25-16} = op; let Inst{15} = !if(P.HasClamp, clamp{0}, 0); } def SDWA { // sdwa_sel int BYTE_0 = 0; int BYTE_1 = 1; int BYTE_2 = 2; int BYTE_3 = 3; int WORD_0 = 4; int WORD_1 = 5; int DWORD = 6; // dst_unused int UNUSED_PAD = 0; int UNUSED_SEXT = 1; int UNUSED_PRESERVE = 2; } class VOP_SDWAe : Enc64 { bits<8> src0; bits<3> src0_sel; bits<2> src0_modifiers; // float: {abs,neg}, int {sext} bits<3> src1_sel; bits<2> src1_modifiers; bits<3> dst_sel; bits<2> dst_unused; bits<1> clamp; let Inst{39-32} = !if(P.HasSrc0, src0{7-0}, 0); let Inst{42-40} = !if(P.EmitDst, dst_sel{2-0}, 0); let Inst{44-43} = !if(P.EmitDst, dst_unused{1-0}, 0); let Inst{45} = !if(P.HasSDWAClamp, clamp{0}, 0); let Inst{50-48} = !if(P.HasSrc0, src0_sel{2-0}, 0); let Inst{51} = !if(P.HasSrc0IntMods, src0_modifiers{0}, 0); let Inst{53-52} = !if(P.HasSrc0FloatMods, src0_modifiers{1-0}, 0); let Inst{58-56} = !if(P.HasSrc1, src1_sel{2-0}, 0); let Inst{59} = !if(P.HasSrc1IntMods, src1_modifiers{0}, 0); let Inst{61-60} = !if(P.HasSrc1FloatMods, src1_modifiers{1-0}, 0); } // GFX9 adds two features to SDWA: // 1. Add 3 fields to the SDWA microcode word: S0, S1 and OMOD. // a. S0 and S1 indicate that source 0 and 1 respectively are SGPRs rather // than VGPRs (at most 1 can be an SGPR); // b. OMOD is the standard output modifier (result *2, *4, /2) // 2. Add a new version of the SDWA microcode word for VOPC: SDWAB. This // replaces OMOD and the dest fields with SD and SDST (SGPR destination) // field. // a. When SD=1, the SDST is used as the destination for the compare result; // b. When SD=0, VCC is used. // // In GFX9, V_MAC_F16, V_MAC_F32 opcodes cannot be used with SDWA // gfx9 SDWA basic encoding class VOP_SDWA9e : Enc64 { bits<9> src0; // {src0_sgpr{0}, src0{7-0}} bits<3> src0_sel; bits<2> src0_modifiers; // float: {abs,neg}, int {sext} bits<3> src1_sel; bits<2> src1_modifiers; bits<1> src1_sgpr; let Inst{39-32} = !if(P.HasSrc0, src0{7-0}, 0); let Inst{50-48} = !if(P.HasSrc0, src0_sel{2-0}, 0); let Inst{51} = !if(P.HasSrc0IntMods, src0_modifiers{0}, 0); let Inst{53-52} = !if(P.HasSrc0FloatMods, src0_modifiers{1-0}, 0); let Inst{55} = !if(P.HasSrc0, src0{8}, 0); let Inst{58-56} = !if(P.HasSrc1, src1_sel{2-0}, 0); let Inst{59} = !if(P.HasSrc1IntMods, src1_modifiers{0}, 0); let Inst{61-60} = !if(P.HasSrc1FloatMods, src1_modifiers{1-0}, 0); let Inst{63} = 0; // src1_sgpr - should be specified in subclass } // gfx9 SDWA-A class VOP_SDWA9Ae : VOP_SDWA9e

{ bits<3> dst_sel; bits<2> dst_unused; bits<1> clamp; bits<2> omod; let Inst{42-40} = !if(P.EmitDst, dst_sel{2-0}, 0); let Inst{44-43} = !if(P.EmitDst, dst_unused{1-0}, 0); let Inst{45} = !if(P.HasSDWAClamp, clamp{0}, 0); let Inst{47-46} = !if(P.HasSDWAOMod, omod{1-0}, 0); } // gfx9 SDWA-B class VOP_SDWA9Be : VOP_SDWA9e

{ bits<8> sdst; // {vcc_sdst{0}, sdst{6-0}} let Inst{46-40} = !if(P.EmitDst, sdst{6-0}, ?); let Inst{47} = !if(P.EmitDst, sdst{7}, 0); } class VOP_SDWA_Pseudo pattern=[]> : InstSI , VOP , SIMCInstr { let isPseudo = 1; let isCodeGenOnly = 1; let UseNamedOperandTable = 1; string Mnemonic = opName; string AsmOperands = P.AsmSDWA; string AsmOperands9 = P.AsmSDWA9; let Size = 8; let mayLoad = 0; let mayStore = 0; let hasSideEffects = 0; let VALU = 1; let SDWA = 1; let ReadsModeReg = !or(isFloatType.ret, isFloatType.ret); let mayRaiseFPException = ReadsModeReg; let Uses = !if(ReadsModeReg, [MODE, EXEC], [EXEC]); let SubtargetPredicate = HasSDWA; let AssemblerPredicate = HasSDWA; let AsmVariantName = !if(P.HasExtSDWA, AMDGPUAsmVariants.SDWA, AMDGPUAsmVariants.Disable); let DecoderNamespace = "SDWA"; VOPProfile Pfl = P; } class VOP_SDWA_Real : InstSI , SIMCInstr { let isPseudo = 0; let isCodeGenOnly = 0; let Defs = ps.Defs; let Uses = ps.Uses; let SchedRW = ps.SchedRW; let hasSideEffects = ps.hasSideEffects; let Constraints = ps.Constraints; let DisableEncoding = ps.DisableEncoding; // Copy relevant pseudo op flags let SubtargetPredicate = ps.SubtargetPredicate; let AssemblerPredicate = ps.AssemblerPredicate; let AsmMatchConverter = ps.AsmMatchConverter; let AsmVariantName = ps.AsmVariantName; let UseNamedOperandTable = ps.UseNamedOperandTable; let DecoderNamespace = ps.DecoderNamespace; let Constraints = ps.Constraints; let DisableEncoding = ps.DisableEncoding; let TSFlags = ps.TSFlags; } class Base_VOP_SDWA9_Real : InstSI { let isPseudo = 0; let isCodeGenOnly = 0; let Defs = ps.Defs; let Uses = ps.Uses; let SchedRW = ps.SchedRW; let hasSideEffects = ps.hasSideEffects; let Constraints = ps.Constraints; let DisableEncoding = ps.DisableEncoding; let SubtargetPredicate = HasSDWA9; let AssemblerPredicate = HasSDWA9; let AsmVariantName = !if(ps.Pfl.HasExtSDWA9, AMDGPUAsmVariants.SDWA9, AMDGPUAsmVariants.Disable); let DecoderNamespace = "SDWA9"; // Copy relevant pseudo op flags let AsmMatchConverter = ps.AsmMatchConverter; let UseNamedOperandTable = ps.UseNamedOperandTable; let Constraints = ps.Constraints; let DisableEncoding = ps.DisableEncoding; let TSFlags = ps.TSFlags; } class VOP_SDWA9_Real : Base_VOP_SDWA9_Real , SIMCInstr ; class Base_VOP_SDWA10_Real : Base_VOP_SDWA9_Real { let SubtargetPredicate = HasSDWA10; let AssemblerPredicate = HasSDWA10; let DecoderNamespace = "SDWA10"; } class VOP_SDWA10_Real : Base_VOP_SDWA10_Real, SIMCInstr; class VOP_DPPe : Enc64 { bits<2> src0_modifiers; bits<8> src0; bits<2> src1_modifiers; bits<9> dpp_ctrl; bits<1> bound_ctrl; bits<4> bank_mask; bits<4> row_mask; bit fi; let Inst{39-32} = !if(P.HasSrc0, src0{7-0}, 0); let Inst{48-40} = dpp_ctrl; let Inst{50} = !if(IsDPP16, fi, ?); let Inst{51} = bound_ctrl; let Inst{52} = !if(P.HasSrc0Mods, src0_modifiers{0}, 0); // src0_neg let Inst{53} = !if(P.HasSrc0Mods, src0_modifiers{1}, 0); // src0_abs let Inst{54} = !if(P.HasSrc1Mods, src1_modifiers{0}, 0); // src1_neg let Inst{55} = !if(P.HasSrc1Mods, src1_modifiers{1}, 0); // src1_abs let Inst{59-56} = bank_mask; let Inst{63-60} = row_mask; } class VOP_DPP_Pseudo pattern=[]> : InstSI , VOP , SIMCInstr { let isPseudo = 1; let isCodeGenOnly = 1; let mayLoad = 0; let mayStore = 0; let hasSideEffects = 0; let UseNamedOperandTable = 1; let VALU = 1; let DPP = 1; let Size = 8; let ReadsModeReg = !or(isFloatType.ret, isFloatType.ret); let mayRaiseFPException = ReadsModeReg; let Uses = !if(ReadsModeReg, [MODE, EXEC], [EXEC]); let isConvergent = 1; string Mnemonic = OpName; string AsmOperands = P.AsmDPP; let AsmMatchConverter = !if(P.HasModifiers, "cvtDPP", ""); let SubtargetPredicate = HasDPP; let AssemblerPredicate = HasDPP; let AsmVariantName = !if(P.HasExtDPP, AMDGPUAsmVariants.DPP, AMDGPUAsmVariants.Disable); let Constraints = !if(P.NumSrcArgs, P.TieRegDPP # " = $vdst", ""); let DisableEncoding = !if(P.NumSrcArgs, P.TieRegDPP, ""); let DecoderNamespace = "DPP"; VOPProfile Pfl = P; } class VOP_DPP_Real : InstSI , SIMCInstr { let isPseudo = 0; let isCodeGenOnly = 0; let Defs = ps.Defs; let Uses = ps.Uses; let SchedRW = ps.SchedRW; let hasSideEffects = ps.hasSideEffects; let Constraints = ps.Constraints; let DisableEncoding = ps.DisableEncoding; // Copy relevant pseudo op flags let isConvergent = ps.isConvergent; let SubtargetPredicate = ps.SubtargetPredicate; let AssemblerPredicate = ps.AssemblerPredicate; let AsmMatchConverter = ps.AsmMatchConverter; let AsmVariantName = ps.AsmVariantName; let UseNamedOperandTable = ps.UseNamedOperandTable; let DecoderNamespace = ps.DecoderNamespace; let Constraints = ps.Constraints; let DisableEncoding = ps.DisableEncoding; let TSFlags = ps.TSFlags; } class VOP_DPP : InstSI , VOP_DPPe { let mayLoad = 0; let mayStore = 0; let hasSideEffects = 0; let UseNamedOperandTable = 1; let VALU = 1; let DPP = 1; let Size = 8; let AsmMatchConverter = !if(P.HasModifiers, "cvtDPP", ""); let SubtargetPredicate = HasDPP; let AssemblerPredicate = HasDPP; let AsmVariantName = !if(P.HasExtDPP, AMDGPUAsmVariants.DPP, AMDGPUAsmVariants.Disable); let Constraints = !if(P.NumSrcArgs, P.TieRegDPP # " = $vdst", ""); let DisableEncoding = !if(P.NumSrcArgs, P.TieRegDPP, ""); let DecoderNamespace = "DPP"; } class VOP_DPP8e : Enc64 { bits<8> src0; bits<24> dpp8; bits<9> fi; let Inst{39-32} = !if(P.HasSrc0, src0{7-0}, 0); let Inst{63-40} = dpp8{23-0}; } class VOP_DPP8 : InstSI, VOP_DPP8e

{ let mayLoad = 0; let mayStore = 0; let hasSideEffects = 0; let UseNamedOperandTable = 1; let VALU = 1; let DPP = 1; let Size = 8; let AsmMatchConverter = "cvtDPP8"; let SubtargetPredicate = HasDPP8; let AssemblerPredicate = HasDPP8; let AsmVariantName = !if(P.HasExt, AMDGPUAsmVariants.DPP, AMDGPUAsmVariants.Disable); let Constraints = !if(P.NumSrcArgs, P.TieRegDPP # " = $vdst", ""); let DisableEncoding = !if(P.NumSrcArgs, P.TieRegDPP, ""); } def DPP8Mode { int FI_0 = 0xE9; int FI_1 = 0xEA; } class getNumNodeArgs { SDNode N = !cast(Op); SDTypeProfile TP = N.TypeProfile; int ret = TP.NumOperands; } class getDivergentFrag { int NumSrcArgs = getNumNodeArgs.ret; PatFrag ret = PatFrag < !if(!eq(NumSrcArgs, 1), (ops node:$src0), !if(!eq(NumSrcArgs, 2), (ops node:$src0, node:$src1), (ops node:$src0, node:$src1, node:$src2))), !if(!eq(NumSrcArgs, 1), (Op $src0), !if(!eq(NumSrcArgs, 2), (Op $src0, $src1), (Op $src0, $src1, $src2))), [{ return N->isDivergent(); }] >; } class VOPPatGen { PatFrag Operator = getDivergentFrag < Op >.ret; dag Ins = !foreach(tmp, P.Ins32, !subst(ins, Operator, !subst(P.Src0RC32, P.Src0VT, !subst(P.Src1RC32, P.Src1VT, tmp)))); dag Outs = !foreach(tmp, P.Outs32, !subst(outs, set, !subst(P.DstRC, P.DstVT, tmp))); list ret = [!con(Outs, (set Ins))]; } class VOPPatOrNull { list ret = !if(!ne(P.NeedPatGen,PatGenMode.NoPattern), VOPPatGen.ret, []); } class DivergentFragOrOp { SDPatternOperator ret = !if(!eq(P.NeedPatGen,PatGenMode.Pattern), !if(!isa(Op), getDivergentFrag.ret, Op), Op); } class getVSrcOp { RegisterOperand ret = !if(!eq(vt.Size, 32), VSrc_b32, VSrc_b16); } // Class for binary integer operations with the clamp bit set for saturation // TODO: Add sub with negated inline constant pattern. class VOPBinOpClampPat : GCNPat<(node vt:$src0, vt:$src1), (inst getVSrcOp.ret:$src0, getVSrcOp.ret:$src1, DSTCLAMP.ENABLE) >; include "VOPCInstructions.td" include "VOP1Instructions.td" include "VOP2Instructions.td" include "VOP3Instructions.td" include "VOP3PInstructions.td"