| //===-- MIMGInstructions.td - MIMG Instruction Defintions -----------------===// |
| // |
| // 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 |
| // |
| //===----------------------------------------------------------------------===// |
| |
| // MIMG-specific encoding families to distinguish between semantically |
| // equivalent machine instructions with different encoding. |
| // |
| // - MIMGEncGfx6: encoding introduced with gfx6 (obsoleted for atomics in gfx8) |
| // - MIMGEncGfx8: encoding introduced with gfx8 for atomics |
| // - MIMGEncGfx10Default: gfx default (non-NSA) encoding |
| // - MIMGEncGfx10NSA: gfx10 NSA encoding |
| class MIMGEncoding; |
| |
| def MIMGEncGfx6 : MIMGEncoding; |
| def MIMGEncGfx8 : MIMGEncoding; |
| def MIMGEncGfx10Default : MIMGEncoding; |
| def MIMGEncGfx10NSA : MIMGEncoding; |
| |
| def MIMGEncoding : GenericEnum { |
| let FilterClass = "MIMGEncoding"; |
| } |
| |
| // Represent an ISA-level opcode, independent of the encoding and the |
| // vdata/vaddr size. |
| class MIMGBaseOpcode : PredicateControl { |
| MIMGBaseOpcode BaseOpcode = !cast<MIMGBaseOpcode>(NAME); |
| bit Store = 0; |
| bit Atomic = 0; |
| bit AtomicX2 = 0; // (f)cmpswap |
| bit Sampler = 0; |
| bit Gather4 = 0; |
| bits<8> NumExtraArgs = 0; |
| bit Gradients = 0; |
| bit Coordinates = 1; |
| bit LodOrClampOrMip = 0; |
| bit HasD16 = 0; |
| } |
| |
| def MIMGBaseOpcode : GenericEnum { |
| let FilterClass = "MIMGBaseOpcode"; |
| } |
| |
| def MIMGBaseOpcodesTable : GenericTable { |
| let FilterClass = "MIMGBaseOpcode"; |
| let CppTypeName = "MIMGBaseOpcodeInfo"; |
| let Fields = ["BaseOpcode", "Store", "Atomic", "AtomicX2", "Sampler", "Gather4", |
| "NumExtraArgs", "Gradients", "Coordinates", "LodOrClampOrMip", |
| "HasD16"]; |
| GenericEnum TypeOf_BaseOpcode = MIMGBaseOpcode; |
| |
| let PrimaryKey = ["BaseOpcode"]; |
| let PrimaryKeyName = "getMIMGBaseOpcodeInfo"; |
| } |
| |
| def MIMGDim : GenericEnum { |
| let FilterClass = "AMDGPUDimProps"; |
| } |
| |
| def MIMGDimInfoTable : GenericTable { |
| let FilterClass = "AMDGPUDimProps"; |
| let CppTypeName = "MIMGDimInfo"; |
| let Fields = ["Dim", "NumCoords", "NumGradients", "DA", "Encoding", "AsmSuffix"]; |
| GenericEnum TypeOf_Dim = MIMGDim; |
| |
| let PrimaryKey = ["Dim"]; |
| let PrimaryKeyName = "getMIMGDimInfo"; |
| } |
| |
| def getMIMGDimInfoByEncoding : SearchIndex { |
| let Table = MIMGDimInfoTable; |
| let Key = ["Encoding"]; |
| } |
| |
| def getMIMGDimInfoByAsmSuffix : SearchIndex { |
| let Table = MIMGDimInfoTable; |
| let Key = ["AsmSuffix"]; |
| } |
| |
| class mimg <bits<8> si_gfx10, bits<8> vi = si_gfx10> { |
| field bits<8> SI_GFX10 = si_gfx10; |
| field bits<8> VI = vi; |
| } |
| |
| class MIMGLZMapping<MIMGBaseOpcode l, MIMGBaseOpcode lz> { |
| MIMGBaseOpcode L = l; |
| MIMGBaseOpcode LZ = lz; |
| } |
| |
| def MIMGLZMappingTable : GenericTable { |
| let FilterClass = "MIMGLZMapping"; |
| let CppTypeName = "MIMGLZMappingInfo"; |
| let Fields = ["L", "LZ"]; |
| GenericEnum TypeOf_L = MIMGBaseOpcode; |
| GenericEnum TypeOf_LZ = MIMGBaseOpcode; |
| |
| let PrimaryKey = ["L"]; |
| let PrimaryKeyName = "getMIMGLZMappingInfo"; |
| } |
| |
| class MIMGMIPMapping<MIMGBaseOpcode mip, MIMGBaseOpcode nonmip> { |
| MIMGBaseOpcode MIP = mip; |
| MIMGBaseOpcode NONMIP = nonmip; |
| } |
| |
| def MIMGMIPMappingTable : GenericTable { |
| let FilterClass = "MIMGMIPMapping"; |
| let CppTypeName = "MIMGMIPMappingInfo"; |
| let Fields = ["MIP", "NONMIP"]; |
| GenericEnum TypeOf_MIP = MIMGBaseOpcode; |
| GenericEnum TypeOf_NONMIP = MIMGBaseOpcode; |
| |
| let PrimaryKey = ["MIP"]; |
| let PrimaryKeyName = "getMIMGMIPMappingInfo"; |
| } |
| |
| class MIMG_Base <dag outs, string dns = ""> |
| : InstSI <outs, (ins), "", []> { |
| |
| let VM_CNT = 1; |
| let EXP_CNT = 1; |
| let MIMG = 1; |
| let Uses = [EXEC]; |
| let mayLoad = 1; |
| let mayStore = 0; |
| let SchedRW = [WriteVMEM]; |
| let UseNamedOperandTable = 1; |
| let hasSideEffects = 0; // XXX ???? |
| |
| let DecoderNamespace = dns; |
| let isAsmParserOnly = !if(!eq(dns,""), 1, 0); |
| let usesCustomInserter = 1; |
| } |
| |
| class MIMG <dag outs, string dns = ""> |
| : MIMG_Base <outs, dns> { |
| |
| let hasPostISelHook = 1; |
| let AsmMatchConverter = "cvtMIMG"; |
| |
| Instruction Opcode = !cast<Instruction>(NAME); |
| MIMGBaseOpcode BaseOpcode; |
| MIMGEncoding MIMGEncoding; |
| bits<8> VDataDwords; |
| bits<8> VAddrDwords; |
| } |
| |
| def MIMGInfoTable : GenericTable { |
| let FilterClass = "MIMG"; |
| let CppTypeName = "MIMGInfo"; |
| let Fields = ["Opcode", "BaseOpcode", "MIMGEncoding", "VDataDwords", "VAddrDwords"]; |
| GenericEnum TypeOf_BaseOpcode = MIMGBaseOpcode; |
| GenericEnum TypeOf_MIMGEncoding = MIMGEncoding; |
| |
| let PrimaryKey = ["BaseOpcode", "MIMGEncoding", "VDataDwords", "VAddrDwords"]; |
| let PrimaryKeyName = "getMIMGOpcodeHelper"; |
| } |
| |
| def getMIMGInfo : SearchIndex { |
| let Table = MIMGInfoTable; |
| let Key = ["Opcode"]; |
| } |
| |
| // This is a separate class so that TableGen memoizes the computations. |
| class MIMGNSAHelper<int num_addrs> { |
| list<string> AddrAsmNames = |
| !foldl([]<string>, [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11], lhs, i, |
| !if(!lt(i, num_addrs), !listconcat(lhs, ["vaddr"#!size(lhs)]), lhs)); |
| dag AddrIns = !dag(ins, !foreach(arg, AddrAsmNames, VGPR_32), AddrAsmNames); |
| string AddrAsm = "[" # !foldl("$" # !head(AddrAsmNames), !tail(AddrAsmNames), lhs, rhs, |
| lhs # ", $" # rhs) # "]"; |
| |
| int NSA = !if(!le(num_addrs, 1), ?, |
| !if(!le(num_addrs, 5), 1, |
| !if(!le(num_addrs, 9), 2, |
| !if(!le(num_addrs, 13), 3, ?)))); |
| } |
| |
| // Base class of all pre-gfx10 MIMG instructions. |
| class MIMG_gfx6789<bits<8> op, dag outs, string dns = ""> |
| : MIMG<outs, dns>, MIMGe_gfx6789<op> { |
| let SubtargetPredicate = isGFX6GFX7GFX8GFX9; |
| let AssemblerPredicate = isGFX6GFX7GFX8GFX9; |
| |
| let MIMGEncoding = MIMGEncGfx6; |
| |
| let d16 = !if(BaseOpcode.HasD16, ?, 0); |
| } |
| |
| // Base class of all non-NSA gfx10 MIMG instructions. |
| class MIMG_gfx10<int op, dag outs, string dns = ""> |
| : MIMG<outs, dns>, MIMGe_gfx10<op> { |
| let SubtargetPredicate = isGFX10Plus; |
| let AssemblerPredicate = isGFX10Plus; |
| |
| let MIMGEncoding = MIMGEncGfx10Default; |
| |
| let d16 = !if(BaseOpcode.HasD16, ?, 0); |
| let nsa = 0; |
| } |
| |
| // Base class for all NSA MIMG instructions. Note that 1-dword addresses always |
| // use non-NSA variants. |
| class MIMG_nsa_gfx10<int op, dag outs, int num_addrs, string dns=""> |
| : MIMG<outs, dns>, MIMGe_gfx10<op> { |
| let SubtargetPredicate = isGFX10Plus; |
| let AssemblerPredicate = isGFX10Plus; |
| |
| let MIMGEncoding = MIMGEncGfx10NSA; |
| |
| MIMGNSAHelper nsah = MIMGNSAHelper<num_addrs>; |
| dag AddrIns = nsah.AddrIns; |
| string AddrAsm = nsah.AddrAsm; |
| |
| let d16 = !if(BaseOpcode.HasD16, ?, 0); |
| let nsa = nsah.NSA; |
| } |
| |
| class MIMG_NoSampler_Helper <bits<8> op, string asm, |
| RegisterClass dst_rc, |
| RegisterClass addr_rc, |
| string dns=""> |
| : MIMG_gfx6789 <op, (outs dst_rc:$vdata), dns> { |
| let InOperandList = !con((ins addr_rc:$vaddr, SReg_256:$srsrc, |
| DMask:$dmask, UNorm:$unorm, GLC:$glc, SLC:$slc, |
| R128A16:$r128, TFE:$tfe, LWE:$lwe, DA:$da), |
| !if(BaseOpcode.HasD16, (ins D16:$d16), (ins))); |
| let AsmString = asm#" $vdata, $vaddr, $srsrc$dmask$unorm$glc$slc$r128$tfe$lwe$da" |
| #!if(BaseOpcode.HasD16, "$d16", ""); |
| } |
| |
| class MIMG_NoSampler_gfx10<int op, string opcode, |
| RegisterClass DataRC, RegisterClass AddrRC, |
| string dns=""> |
| : MIMG_gfx10<op, (outs DataRC:$vdata), dns> { |
| let InOperandList = !con((ins AddrRC:$vaddr0, SReg_256:$srsrc, DMask:$dmask, |
| Dim:$dim, UNorm:$unorm, DLC:$dlc, GLC:$glc, |
| SLC:$slc, R128A16:$r128, TFE:$tfe, LWE:$lwe), |
| !if(BaseOpcode.HasD16, (ins D16:$d16), (ins))); |
| let AsmString = opcode#" $vdata, $vaddr0, $srsrc$dmask$dim$unorm$dlc$glc$slc$r128$tfe$lwe" |
| #!if(BaseOpcode.HasD16, "$d16", ""); |
| } |
| |
| class MIMG_NoSampler_nsa_gfx10<int op, string opcode, |
| RegisterClass DataRC, int num_addrs, |
| string dns=""> |
| : MIMG_nsa_gfx10<op, (outs DataRC:$vdata), num_addrs, dns> { |
| let InOperandList = !con(AddrIns, |
| (ins SReg_256:$srsrc, DMask:$dmask, |
| Dim:$dim, UNorm:$unorm, DLC:$dlc, GLC:$glc, |
| SLC:$slc, R128A16:$r128, TFE:$tfe, LWE:$lwe), |
| !if(BaseOpcode.HasD16, (ins D16:$d16), (ins))); |
| let AsmString = opcode#" $vdata, "#AddrAsm#", $srsrc$dmask$dim$unorm$dlc$glc$slc$r128$tfe$lwe" |
| #!if(BaseOpcode.HasD16, "$d16", ""); |
| } |
| |
| multiclass MIMG_NoSampler_Src_Helper <bits<8> op, string asm, |
| RegisterClass dst_rc, |
| bit enableDisasm> { |
| let ssamp = 0 in { |
| let VAddrDwords = 1 in { |
| def _V1 : MIMG_NoSampler_Helper <op, asm, dst_rc, VGPR_32, |
| !if(enableDisasm, "AMDGPU", "")>; |
| def _V1_gfx10 : MIMG_NoSampler_gfx10<op, asm, dst_rc, VGPR_32, |
| !if(enableDisasm, "AMDGPU", "")>; |
| } |
| |
| let VAddrDwords = 2 in { |
| def _V2 : MIMG_NoSampler_Helper <op, asm, dst_rc, VReg_64>; |
| def _V2_gfx10 : MIMG_NoSampler_gfx10<op, asm, dst_rc, VReg_64>; |
| def _V2_nsa_gfx10 : MIMG_NoSampler_nsa_gfx10<op, asm, dst_rc, 2>; |
| } |
| |
| let VAddrDwords = 3 in { |
| def _V3 : MIMG_NoSampler_Helper <op, asm, dst_rc, VReg_96>; |
| def _V3_gfx10 : MIMG_NoSampler_gfx10<op, asm, dst_rc, VReg_96>; |
| def _V3_nsa_gfx10 : MIMG_NoSampler_nsa_gfx10<op, asm, dst_rc, 3>; |
| } |
| |
| let VAddrDwords = 4 in { |
| def _V4 : MIMG_NoSampler_Helper <op, asm, dst_rc, VReg_128>; |
| def _V4_gfx10 : MIMG_NoSampler_gfx10<op, asm, dst_rc, VReg_128>; |
| def _V4_nsa_gfx10 : MIMG_NoSampler_nsa_gfx10<op, asm, dst_rc, 4, |
| !if(enableDisasm, "AMDGPU", "")>; |
| } |
| } |
| } |
| |
| multiclass MIMG_NoSampler <bits<8> op, string asm, bit has_d16, bit mip = 0, |
| bit isResInfo = 0> { |
| def "" : MIMGBaseOpcode { |
| let Coordinates = !if(isResInfo, 0, 1); |
| let LodOrClampOrMip = mip; |
| let HasD16 = has_d16; |
| } |
| |
| let BaseOpcode = !cast<MIMGBaseOpcode>(NAME), |
| mayLoad = !if(isResInfo, 0, 1) in { |
| let VDataDwords = 1 in |
| defm _V1 : MIMG_NoSampler_Src_Helper <op, asm, VGPR_32, 1>; |
| let VDataDwords = 2 in |
| defm _V2 : MIMG_NoSampler_Src_Helper <op, asm, VReg_64, 0>; |
| let VDataDwords = 3 in |
| defm _V3 : MIMG_NoSampler_Src_Helper <op, asm, VReg_96, 0>; |
| let VDataDwords = 4 in |
| defm _V4 : MIMG_NoSampler_Src_Helper <op, asm, VReg_128, 0>; |
| let VDataDwords = 5 in |
| defm _V5 : MIMG_NoSampler_Src_Helper <op, asm, VReg_160, 0>; |
| } |
| } |
| |
| class MIMG_Store_Helper <bits<8> op, string asm, |
| RegisterClass data_rc, |
| RegisterClass addr_rc, |
| string dns = ""> |
| : MIMG_gfx6789<op, (outs), dns> { |
| let InOperandList = !con((ins data_rc:$vdata, addr_rc:$vaddr, SReg_256:$srsrc, |
| DMask:$dmask, UNorm:$unorm, GLC:$glc, SLC:$slc, |
| R128A16:$r128, TFE:$tfe, LWE:$lwe, DA:$da), |
| !if(BaseOpcode.HasD16, (ins D16:$d16), (ins))); |
| let AsmString = asm#" $vdata, $vaddr, $srsrc$dmask$unorm$glc$slc$r128$tfe$lwe$da" |
| #!if(BaseOpcode.HasD16, "$d16", ""); |
| } |
| |
| class MIMG_Store_gfx10<int op, string opcode, |
| RegisterClass DataRC, RegisterClass AddrRC, |
| string dns=""> |
| : MIMG_gfx10<op, (outs), dns> { |
| let InOperandList = !con((ins DataRC:$vdata, AddrRC:$vaddr0, SReg_256:$srsrc, |
| DMask:$dmask, Dim:$dim, UNorm:$unorm, DLC:$dlc, |
| GLC:$glc, SLC:$slc, R128A16:$r128, TFE:$tfe, LWE:$lwe), |
| !if(BaseOpcode.HasD16, (ins D16:$d16), (ins))); |
| let AsmString = opcode#" $vdata, $vaddr0, $srsrc$dmask$dim$unorm$dlc$glc$slc$r128$tfe$lwe" |
| #!if(BaseOpcode.HasD16, "$d16", ""); |
| } |
| |
| class MIMG_Store_nsa_gfx10<int op, string opcode, |
| RegisterClass DataRC, int num_addrs, |
| string dns=""> |
| : MIMG_nsa_gfx10<op, (outs), num_addrs, dns> { |
| let InOperandList = !con((ins DataRC:$vdata), |
| AddrIns, |
| (ins SReg_256:$srsrc, DMask:$dmask, |
| Dim:$dim, UNorm:$unorm, DLC:$dlc, GLC:$glc, |
| SLC:$slc, R128A16:$r128, TFE:$tfe, LWE:$lwe), |
| !if(BaseOpcode.HasD16, (ins D16:$d16), (ins))); |
| let AsmString = opcode#" $vdata, "#AddrAsm#", $srsrc$dmask$dim$unorm$dlc$glc$slc$r128$tfe$lwe" |
| #!if(BaseOpcode.HasD16, "$d16", ""); |
| } |
| |
| multiclass MIMG_Store_Addr_Helper <int op, string asm, |
| RegisterClass data_rc, |
| bit enableDisasm> { |
| let mayLoad = 0, mayStore = 1, hasSideEffects = 0, hasPostISelHook = 0, |
| DisableWQM = 1, ssamp = 0 in { |
| let VAddrDwords = 1 in { |
| def _V1 : MIMG_Store_Helper <op, asm, data_rc, VGPR_32, |
| !if(enableDisasm, "AMDGPU", "")>; |
| def _V1_gfx10 : MIMG_Store_gfx10 <op, asm, data_rc, VGPR_32, |
| !if(enableDisasm, "AMDGPU", "")>; |
| } |
| let VAddrDwords = 2 in { |
| def _V2 : MIMG_Store_Helper <op, asm, data_rc, VReg_64>; |
| def _V2_gfx10 : MIMG_Store_gfx10 <op, asm, data_rc, VReg_64>; |
| def _V2_nsa_gfx10 : MIMG_Store_nsa_gfx10 <op, asm, data_rc, 2>; |
| } |
| let VAddrDwords = 3 in { |
| def _V3 : MIMG_Store_Helper <op, asm, data_rc, VReg_96>; |
| def _V3_gfx10 : MIMG_Store_gfx10 <op, asm, data_rc, VReg_96>; |
| def _V3_nsa_gfx10 : MIMG_Store_nsa_gfx10 <op, asm, data_rc, 3>; |
| } |
| let VAddrDwords = 4 in { |
| def _V4 : MIMG_Store_Helper <op, asm, data_rc, VReg_128>; |
| def _V4_gfx10 : MIMG_Store_gfx10 <op, asm, data_rc, VReg_128>; |
| def _V4_nsa_gfx10 : MIMG_Store_nsa_gfx10 <op, asm, data_rc, 4, |
| !if(enableDisasm, "AMDGPU", "")>; |
| } |
| } |
| } |
| |
| multiclass MIMG_Store <bits<8> op, string asm, bit has_d16, bit mip = 0> { |
| def "" : MIMGBaseOpcode { |
| let Store = 1; |
| let LodOrClampOrMip = mip; |
| let HasD16 = has_d16; |
| } |
| |
| let BaseOpcode = !cast<MIMGBaseOpcode>(NAME) in { |
| let VDataDwords = 1 in |
| defm _V1 : MIMG_Store_Addr_Helper <op, asm, VGPR_32, 1>; |
| let VDataDwords = 2 in |
| defm _V2 : MIMG_Store_Addr_Helper <op, asm, VReg_64, 0>; |
| let VDataDwords = 3 in |
| defm _V3 : MIMG_Store_Addr_Helper <op, asm, VReg_96, 0>; |
| let VDataDwords = 4 in |
| defm _V4 : MIMG_Store_Addr_Helper <op, asm, VReg_128, 0>; |
| } |
| } |
| |
| class MIMG_Atomic_gfx6789_base <bits<8> op, string asm, RegisterClass data_rc, |
| RegisterClass addr_rc, string dns=""> |
| : MIMG_gfx6789 <op, (outs data_rc:$vdst), dns> { |
| let Constraints = "$vdst = $vdata"; |
| let AsmMatchConverter = "cvtMIMGAtomic"; |
| |
| let InOperandList = (ins data_rc:$vdata, addr_rc:$vaddr, SReg_256:$srsrc, |
| DMask:$dmask, UNorm:$unorm, GLC:$glc, SLC:$slc, |
| R128A16:$r128, TFE:$tfe, LWE:$lwe, DA:$da); |
| let AsmString = asm#" $vdst, $vaddr, $srsrc$dmask$unorm$glc$slc$r128$tfe$lwe$da"; |
| } |
| |
| class MIMG_Atomic_si<mimg op, string asm, RegisterClass data_rc, |
| RegisterClass addr_rc, bit enableDasm = 0> |
| : MIMG_Atomic_gfx6789_base<op.SI_GFX10, asm, data_rc, addr_rc, |
| !if(enableDasm, "GFX6GFX7", "")> { |
| let AssemblerPredicate = isGFX6GFX7; |
| } |
| |
| class MIMG_Atomic_vi<mimg op, string asm, RegisterClass data_rc, |
| RegisterClass addr_rc, bit enableDasm = 0> |
| : MIMG_Atomic_gfx6789_base<op.VI, asm, data_rc, addr_rc, !if(enableDasm, "GFX8", "")> { |
| let AssemblerPredicate = isGFX8GFX9; |
| let MIMGEncoding = MIMGEncGfx8; |
| } |
| |
| class MIMG_Atomic_gfx10<mimg op, string opcode, |
| RegisterClass DataRC, RegisterClass AddrRC, |
| bit enableDisasm = 0> |
| : MIMG_gfx10<!cast<int>(op.SI_GFX10), (outs DataRC:$vdst), |
| !if(enableDisasm, "AMDGPU", "")> { |
| let Constraints = "$vdst = $vdata"; |
| let AsmMatchConverter = "cvtMIMGAtomic"; |
| |
| let InOperandList = (ins DataRC:$vdata, AddrRC:$vaddr0, SReg_256:$srsrc, |
| DMask:$dmask, Dim:$dim, UNorm:$unorm, DLC:$dlc, |
| GLC:$glc, SLC:$slc, R128A16:$r128, TFE:$tfe, LWE:$lwe); |
| let AsmString = opcode#" $vdst, $vaddr0, $srsrc$dmask$dim$unorm$dlc$glc$slc$r128$tfe$lwe"; |
| } |
| |
| class MIMG_Atomic_nsa_gfx10<mimg op, string opcode, |
| RegisterClass DataRC, int num_addrs, |
| bit enableDisasm = 0> |
| : MIMG_nsa_gfx10<!cast<int>(op.SI_GFX10), (outs DataRC:$vdst), num_addrs, |
| !if(enableDisasm, "AMDGPU", "")> { |
| let Constraints = "$vdst = $vdata"; |
| let AsmMatchConverter = "cvtMIMGAtomic"; |
| |
| let InOperandList = !con((ins DataRC:$vdata), |
| AddrIns, |
| (ins SReg_256:$srsrc, DMask:$dmask, |
| Dim:$dim, UNorm:$unorm, DLC:$dlc, GLC:$glc, |
| SLC:$slc, R128A16:$r128, TFE:$tfe, LWE:$lwe)); |
| let AsmString = opcode#" $vdata, "#AddrAsm#", $srsrc$dmask$dim$unorm$dlc$glc$slc$r128$tfe$lwe"; |
| } |
| |
| multiclass MIMG_Atomic_Addr_Helper_m <mimg op, string asm, |
| RegisterClass data_rc, |
| bit enableDasm = 0> { |
| let hasSideEffects = 1, // FIXME: remove this |
| mayLoad = 1, mayStore = 1, hasPostISelHook = 0, DisableWQM = 1, |
| ssamp = 0 in { |
| let VAddrDwords = 1 in { |
| def _V1_si : MIMG_Atomic_si <op, asm, data_rc, VGPR_32, enableDasm>; |
| def _V1_vi : MIMG_Atomic_vi <op, asm, data_rc, VGPR_32, enableDasm>; |
| def _V1_gfx10 : MIMG_Atomic_gfx10 <op, asm, data_rc, VGPR_32, enableDasm>; |
| } |
| let VAddrDwords = 2 in { |
| def _V2_si : MIMG_Atomic_si <op, asm, data_rc, VReg_64, 0>; |
| def _V2_vi : MIMG_Atomic_vi <op, asm, data_rc, VReg_64, 0>; |
| def _V2_gfx10 : MIMG_Atomic_gfx10 <op, asm, data_rc, VReg_64, 0>; |
| def _V2_nsa_gfx10 : MIMG_Atomic_nsa_gfx10 <op, asm, data_rc, 2, 0>; |
| } |
| let VAddrDwords = 3 in { |
| def _V3_si : MIMG_Atomic_si <op, asm, data_rc, VReg_96, 0>; |
| def _V3_vi : MIMG_Atomic_vi <op, asm, data_rc, VReg_96, 0>; |
| def _V3_gfx10 : MIMG_Atomic_gfx10 <op, asm, data_rc, VReg_96, 0>; |
| def _V3_nsa_gfx10 : MIMG_Atomic_nsa_gfx10 <op, asm, data_rc, 3, 0>; |
| } |
| let VAddrDwords = 4 in { |
| def _V4_si : MIMG_Atomic_si <op, asm, data_rc, VReg_128, 0>; |
| def _V4_vi : MIMG_Atomic_vi <op, asm, data_rc, VReg_128, 0>; |
| def _V4_gfx10 : MIMG_Atomic_gfx10 <op, asm, data_rc, VReg_128, 0>; |
| def _V4_nsa_gfx10 : MIMG_Atomic_nsa_gfx10 <op, asm, data_rc, 4, enableDasm>; |
| } |
| } |
| } |
| |
| multiclass MIMG_Atomic <mimg op, string asm, bit isCmpSwap = 0> { // 64-bit atomics |
| def "" : MIMGBaseOpcode { |
| let Atomic = 1; |
| let AtomicX2 = isCmpSwap; |
| } |
| |
| let BaseOpcode = !cast<MIMGBaseOpcode>(NAME) in { |
| // _V* variants have different dst size, but the size is encoded implicitly, |
| // using dmask and tfe. Only 32-bit variant is registered with disassembler. |
| // Other variants are reconstructed by disassembler using dmask and tfe. |
| let VDataDwords = !if(isCmpSwap, 2, 1) in |
| defm _V1 : MIMG_Atomic_Addr_Helper_m <op, asm, !if(isCmpSwap, VReg_64, VGPR_32), 1>; |
| let VDataDwords = !if(isCmpSwap, 4, 2) in |
| defm _V2 : MIMG_Atomic_Addr_Helper_m <op, asm, !if(isCmpSwap, VReg_128, VReg_64)>; |
| } |
| } |
| |
| class MIMG_Sampler_Helper <bits<8> op, string asm, RegisterClass dst_rc, |
| RegisterClass src_rc, string dns=""> |
| : MIMG_gfx6789 <op, (outs dst_rc:$vdata), dns> { |
| let InOperandList = !con((ins src_rc:$vaddr, SReg_256:$srsrc, SReg_128:$ssamp, |
| DMask:$dmask, UNorm:$unorm, GLC:$glc, SLC:$slc, |
| R128A16:$r128, TFE:$tfe, LWE:$lwe, DA:$da), |
| !if(BaseOpcode.HasD16, (ins D16:$d16), (ins))); |
| let AsmString = asm#" $vdata, $vaddr, $srsrc, $ssamp$dmask$unorm$glc$slc$r128$tfe$lwe$da" |
| #!if(BaseOpcode.HasD16, "$d16", ""); |
| } |
| |
| class MIMG_Sampler_gfx10<int op, string opcode, |
| RegisterClass DataRC, RegisterClass AddrRC, |
| string dns=""> |
| : MIMG_gfx10<op, (outs DataRC:$vdata), dns> { |
| let InOperandList = !con((ins AddrRC:$vaddr0, SReg_256:$srsrc, SReg_128:$ssamp, |
| DMask:$dmask, Dim:$dim, UNorm:$unorm, DLC:$dlc, |
| GLC:$glc, SLC:$slc, R128A16:$r128, TFE:$tfe, LWE:$lwe), |
| !if(BaseOpcode.HasD16, (ins D16:$d16), (ins))); |
| let AsmString = opcode#" $vdata, $vaddr0, $srsrc, $ssamp$dmask$dim$unorm" |
| #"$dlc$glc$slc$r128$tfe$lwe" |
| #!if(BaseOpcode.HasD16, "$d16", ""); |
| } |
| |
| class MIMG_Sampler_nsa_gfx10<int op, string opcode, |
| RegisterClass DataRC, int num_addrs, |
| string dns=""> |
| : MIMG_nsa_gfx10<op, (outs DataRC:$vdata), num_addrs, dns> { |
| let InOperandList = !con(AddrIns, |
| (ins SReg_256:$srsrc, SReg_128:$ssamp, DMask:$dmask, |
| Dim:$dim, UNorm:$unorm, DLC:$dlc, GLC:$glc, |
| SLC:$slc, R128A16:$r128, TFE:$tfe, LWE:$lwe), |
| !if(BaseOpcode.HasD16, (ins D16:$d16), (ins))); |
| let AsmString = opcode#" $vdata, "#AddrAsm#", $srsrc, $ssamp$dmask$dim$unorm" |
| #"$dlc$glc$slc$r128$tfe$lwe" |
| #!if(BaseOpcode.HasD16, "$d16", ""); |
| } |
| |
| class MIMGAddrSize<int dw, bit enable_disasm> { |
| int NumWords = dw; |
| |
| RegisterClass RegClass = !if(!le(NumWords, 0), ?, |
| !if(!eq(NumWords, 1), VGPR_32, |
| !if(!eq(NumWords, 2), VReg_64, |
| !if(!eq(NumWords, 3), VReg_96, |
| !if(!eq(NumWords, 4), VReg_128, |
| !if(!le(NumWords, 8), VReg_256, |
| !if(!le(NumWords, 16), VReg_512, ?))))))); |
| |
| // Whether the instruction variant with this vaddr size should be enabled for |
| // the auto-generated disassembler. |
| bit Disassemble = enable_disasm; |
| } |
| |
| // Return whether x is in lst. |
| class isIntInList<int x, list<int> lst> { |
| bit ret = !foldl(0, lst, lhs, y, !or(lhs, !eq(x, y))); |
| } |
| |
| // Return whether a value inside the range [min, max] (endpoints inclusive) |
| // is in the given list. |
| class isRangeInList<int min, int max, list<int> lst> { |
| bit ret = !foldl(0, lst, lhs, y, !or(lhs, !and(!le(min, y), !le(y, max)))); |
| } |
| |
| class MIMGAddrSizes_tmp<list<MIMGAddrSize> lst, int min> { |
| list<MIMGAddrSize> List = lst; |
| int Min = min; |
| } |
| |
| class MIMG_Sampler_AddrSizes<AMDGPUSampleVariant sample> { |
| // List of all possible numbers of address words, taking all combinations of |
| // A16 and image dimension into account (note: no MSAA, since this is for |
| // sample/gather ops). |
| list<int> AllNumAddrWords = |
| !foreach(dw, !if(sample.Gradients, |
| !if(!eq(sample.LodOrClamp, ""), |
| [2, 3, 4, 5, 6, 7, 9], |
| [2, 3, 4, 5, 7, 8, 10]), |
| !if(!eq(sample.LodOrClamp, ""), |
| [1, 2, 3], |
| [1, 2, 3, 4])), |
| !add(dw, !size(sample.ExtraAddrArgs))); |
| |
| // Generate machine instructions based on possible register classes for the |
| // required numbers of address words. The disassembler defaults to the |
| // smallest register class. |
| list<MIMGAddrSize> MachineInstrs = |
| !foldl(MIMGAddrSizes_tmp<[], 0>, [1, 2, 3, 4, 8, 16], lhs, dw, |
| !if(isRangeInList<lhs.Min, dw, AllNumAddrWords>.ret, |
| MIMGAddrSizes_tmp< |
| !listconcat(lhs.List, [MIMGAddrSize<dw, !empty(lhs.List)>]), |
| !if(!eq(dw, 3), 3, !add(dw, 1))>, // we still need _V4 for codegen w/ 3 dwords |
| lhs)).List; |
| |
| // For NSA, generate machine instructions for all possible numbers of words |
| // except 1 (which is already covered by the non-NSA case). |
| // The disassembler defaults to the largest number of arguments among the |
| // variants with the same number of NSA words, and custom code then derives |
| // the exact variant based on the sample variant and the image dimension. |
| list<MIMGAddrSize> NSAInstrs = |
| !foldl([]<MIMGAddrSize>, [[12, 11, 10], [9, 8, 7, 6], [5, 4, 3, 2]], prev, nsa_group, |
| !listconcat(prev, |
| !foldl([]<MIMGAddrSize>, nsa_group, lhs, dw, |
| !if(isIntInList<dw, AllNumAddrWords>.ret, |
| !listconcat(lhs, [MIMGAddrSize<dw, !empty(lhs)>]), |
| lhs)))); |
| } |
| |
| multiclass MIMG_Sampler_Src_Helper <bits<8> op, string asm, |
| AMDGPUSampleVariant sample, RegisterClass dst_rc, |
| bit enableDisasm = 0> { |
| foreach addr = MIMG_Sampler_AddrSizes<sample>.MachineInstrs in { |
| let VAddrDwords = addr.NumWords in { |
| def _V # addr.NumWords |
| : MIMG_Sampler_Helper <op, asm, dst_rc, addr.RegClass, |
| !if(!and(enableDisasm, addr.Disassemble), "AMDGPU", "")>; |
| def _V # addr.NumWords # _gfx10 |
| : MIMG_Sampler_gfx10 <op, asm, dst_rc, addr.RegClass, |
| !if(!and(enableDisasm, addr.Disassemble), "AMDGPU", "")>; |
| } |
| } |
| |
| foreach addr = MIMG_Sampler_AddrSizes<sample>.NSAInstrs in { |
| let VAddrDwords = addr.NumWords in { |
| def _V # addr.NumWords # _nsa_gfx10 |
| : MIMG_Sampler_nsa_gfx10<op, asm, dst_rc, addr.NumWords, |
| !if(!and(enableDisasm, addr.Disassemble), "AMDGPU", "")>; |
| } |
| } |
| } |
| |
| class MIMG_Sampler_BaseOpcode<AMDGPUSampleVariant sample> |
| : MIMGBaseOpcode { |
| let Sampler = 1; |
| let NumExtraArgs = !size(sample.ExtraAddrArgs); |
| let Gradients = sample.Gradients; |
| let LodOrClampOrMip = !ne(sample.LodOrClamp, ""); |
| } |
| |
| multiclass MIMG_Sampler <bits<8> op, AMDGPUSampleVariant sample, bit wqm = 0, |
| bit isGetLod = 0, |
| string asm = "image_sample"#sample.LowerCaseMod> { |
| def "" : MIMG_Sampler_BaseOpcode<sample> { |
| let HasD16 = !if(isGetLod, 0, 1); |
| } |
| |
| let BaseOpcode = !cast<MIMGBaseOpcode>(NAME), WQM = wqm, |
| mayLoad = !if(isGetLod, 0, 1) in { |
| let VDataDwords = 1 in |
| defm _V1 : MIMG_Sampler_Src_Helper<op, asm, sample, VGPR_32, 1>; |
| let VDataDwords = 2 in |
| defm _V2 : MIMG_Sampler_Src_Helper<op, asm, sample, VReg_64>; |
| let VDataDwords = 3 in |
| defm _V3 : MIMG_Sampler_Src_Helper<op, asm, sample, VReg_96>; |
| let VDataDwords = 4 in |
| defm _V4 : MIMG_Sampler_Src_Helper<op, asm, sample, VReg_128>; |
| let VDataDwords = 5 in |
| defm _V5 : MIMG_Sampler_Src_Helper<op, asm, sample, VReg_160>; |
| } |
| } |
| |
| multiclass MIMG_Sampler_WQM <bits<8> op, AMDGPUSampleVariant sample> |
| : MIMG_Sampler<op, sample, 1>; |
| |
| multiclass MIMG_Gather <bits<8> op, AMDGPUSampleVariant sample, bit wqm = 0, |
| string asm = "image_gather4"#sample.LowerCaseMod> { |
| def "" : MIMG_Sampler_BaseOpcode<sample> { |
| let HasD16 = 1; |
| let Gather4 = 1; |
| } |
| |
| let BaseOpcode = !cast<MIMGBaseOpcode>(NAME), WQM = wqm, |
| Gather4 = 1, hasPostISelHook = 0 in { |
| let VDataDwords = 2 in |
| defm _V2 : MIMG_Sampler_Src_Helper<op, asm, sample, VReg_64>; /* for packed D16 only */ |
| let VDataDwords = 4 in |
| defm _V4 : MIMG_Sampler_Src_Helper<op, asm, sample, VReg_128, 1>; |
| let VDataDwords = 5 in |
| defm _V5 : MIMG_Sampler_Src_Helper<op, asm, sample, VReg_160>; |
| } |
| } |
| |
| multiclass MIMG_Gather_WQM <bits<8> op, AMDGPUSampleVariant sample> |
| : MIMG_Gather<op, sample, 1>; |
| |
| //===----------------------------------------------------------------------===// |
| // MIMG Instructions |
| //===----------------------------------------------------------------------===// |
| defm IMAGE_LOAD : MIMG_NoSampler <0x00000000, "image_load", 1>; |
| defm IMAGE_LOAD_MIP : MIMG_NoSampler <0x00000001, "image_load_mip", 1, 1>; |
| defm IMAGE_LOAD_PCK : MIMG_NoSampler <0x00000002, "image_load_pck", 0>; |
| defm IMAGE_LOAD_PCK_SGN : MIMG_NoSampler <0x00000003, "image_load_pck_sgn", 0>; |
| defm IMAGE_LOAD_MIP_PCK : MIMG_NoSampler <0x00000004, "image_load_mip_pck", 0, 1>; |
| defm IMAGE_LOAD_MIP_PCK_SGN : MIMG_NoSampler <0x00000005, "image_load_mip_pck_sgn", 0, 1>; |
| defm IMAGE_STORE : MIMG_Store <0x00000008, "image_store", 1>; |
| defm IMAGE_STORE_MIP : MIMG_Store <0x00000009, "image_store_mip", 1, 1>; |
| defm IMAGE_STORE_PCK : MIMG_Store <0x0000000a, "image_store_pck", 0>; |
| defm IMAGE_STORE_MIP_PCK : MIMG_Store <0x0000000b, "image_store_mip_pck", 0, 1>; |
| |
| defm IMAGE_GET_RESINFO : MIMG_NoSampler <0x0000000e, "image_get_resinfo", 0, 1, 1>; |
| |
| defm IMAGE_ATOMIC_SWAP : MIMG_Atomic <mimg<0x0f, 0x10>, "image_atomic_swap">; |
| defm IMAGE_ATOMIC_CMPSWAP : MIMG_Atomic <mimg<0x10, 0x11>, "image_atomic_cmpswap", 1>; |
| defm IMAGE_ATOMIC_ADD : MIMG_Atomic <mimg<0x11, 0x12>, "image_atomic_add">; |
| defm IMAGE_ATOMIC_SUB : MIMG_Atomic <mimg<0x12, 0x13>, "image_atomic_sub">; |
| //def IMAGE_ATOMIC_RSUB : MIMG_NoPattern_ <"image_atomic_rsub", 0x00000013>; -- not on VI |
| defm IMAGE_ATOMIC_SMIN : MIMG_Atomic <mimg<0x14>, "image_atomic_smin">; |
| defm IMAGE_ATOMIC_UMIN : MIMG_Atomic <mimg<0x15>, "image_atomic_umin">; |
| defm IMAGE_ATOMIC_SMAX : MIMG_Atomic <mimg<0x16>, "image_atomic_smax">; |
| defm IMAGE_ATOMIC_UMAX : MIMG_Atomic <mimg<0x17>, "image_atomic_umax">; |
| defm IMAGE_ATOMIC_AND : MIMG_Atomic <mimg<0x18>, "image_atomic_and">; |
| defm IMAGE_ATOMIC_OR : MIMG_Atomic <mimg<0x19>, "image_atomic_or">; |
| defm IMAGE_ATOMIC_XOR : MIMG_Atomic <mimg<0x1a>, "image_atomic_xor">; |
| defm IMAGE_ATOMIC_INC : MIMG_Atomic <mimg<0x1b>, "image_atomic_inc">; |
| defm IMAGE_ATOMIC_DEC : MIMG_Atomic <mimg<0x1c>, "image_atomic_dec">; |
| //let FPAtomic = 1 in { |
| //def IMAGE_ATOMIC_FCMPSWAP : MIMG_NoPattern_ <"image_atomic_fcmpswap", 0x0000001d, 1>; -- not on VI |
| //def IMAGE_ATOMIC_FMIN : MIMG_NoPattern_ <"image_atomic_fmin", 0x0000001e>; -- not on VI |
| //def IMAGE_ATOMIC_FMAX : MIMG_NoPattern_ <"image_atomic_fmax", 0x0000001f>; -- not on VI |
| //} // End let FPAtomic = 1 |
| defm IMAGE_SAMPLE : MIMG_Sampler_WQM <0x00000020, AMDGPUSample>; |
| defm IMAGE_SAMPLE_CL : MIMG_Sampler_WQM <0x00000021, AMDGPUSample_cl>; |
| defm IMAGE_SAMPLE_D : MIMG_Sampler <0x00000022, AMDGPUSample_d>; |
| defm IMAGE_SAMPLE_D_CL : MIMG_Sampler <0x00000023, AMDGPUSample_d_cl>; |
| defm IMAGE_SAMPLE_L : MIMG_Sampler <0x00000024, AMDGPUSample_l>; |
| defm IMAGE_SAMPLE_B : MIMG_Sampler_WQM <0x00000025, AMDGPUSample_b>; |
| defm IMAGE_SAMPLE_B_CL : MIMG_Sampler_WQM <0x00000026, AMDGPUSample_b_cl>; |
| defm IMAGE_SAMPLE_LZ : MIMG_Sampler <0x00000027, AMDGPUSample_lz>; |
| defm IMAGE_SAMPLE_C : MIMG_Sampler_WQM <0x00000028, AMDGPUSample_c>; |
| defm IMAGE_SAMPLE_C_CL : MIMG_Sampler_WQM <0x00000029, AMDGPUSample_c_cl>; |
| defm IMAGE_SAMPLE_C_D : MIMG_Sampler <0x0000002a, AMDGPUSample_c_d>; |
| defm IMAGE_SAMPLE_C_D_CL : MIMG_Sampler <0x0000002b, AMDGPUSample_c_d_cl>; |
| defm IMAGE_SAMPLE_C_L : MIMG_Sampler <0x0000002c, AMDGPUSample_c_l>; |
| defm IMAGE_SAMPLE_C_B : MIMG_Sampler_WQM <0x0000002d, AMDGPUSample_c_b>; |
| defm IMAGE_SAMPLE_C_B_CL : MIMG_Sampler_WQM <0x0000002e, AMDGPUSample_c_b_cl>; |
| defm IMAGE_SAMPLE_C_LZ : MIMG_Sampler <0x0000002f, AMDGPUSample_c_lz>; |
| defm IMAGE_SAMPLE_O : MIMG_Sampler_WQM <0x00000030, AMDGPUSample_o>; |
| defm IMAGE_SAMPLE_CL_O : MIMG_Sampler_WQM <0x00000031, AMDGPUSample_cl_o>; |
| defm IMAGE_SAMPLE_D_O : MIMG_Sampler <0x00000032, AMDGPUSample_d_o>; |
| defm IMAGE_SAMPLE_D_CL_O : MIMG_Sampler <0x00000033, AMDGPUSample_d_cl_o>; |
| defm IMAGE_SAMPLE_L_O : MIMG_Sampler <0x00000034, AMDGPUSample_l_o>; |
| defm IMAGE_SAMPLE_B_O : MIMG_Sampler_WQM <0x00000035, AMDGPUSample_b_o>; |
| defm IMAGE_SAMPLE_B_CL_O : MIMG_Sampler_WQM <0x00000036, AMDGPUSample_b_cl_o>; |
| defm IMAGE_SAMPLE_LZ_O : MIMG_Sampler <0x00000037, AMDGPUSample_lz_o>; |
| defm IMAGE_SAMPLE_C_O : MIMG_Sampler_WQM <0x00000038, AMDGPUSample_c_o>; |
| defm IMAGE_SAMPLE_C_CL_O : MIMG_Sampler_WQM <0x00000039, AMDGPUSample_c_cl_o>; |
| defm IMAGE_SAMPLE_C_D_O : MIMG_Sampler <0x0000003a, AMDGPUSample_c_d_o>; |
| defm IMAGE_SAMPLE_C_D_CL_O : MIMG_Sampler <0x0000003b, AMDGPUSample_c_d_cl_o>; |
| defm IMAGE_SAMPLE_C_L_O : MIMG_Sampler <0x0000003c, AMDGPUSample_c_l_o>; |
| defm IMAGE_SAMPLE_C_B_CL_O : MIMG_Sampler_WQM <0x0000003e, AMDGPUSample_c_b_cl_o>; |
| defm IMAGE_SAMPLE_C_B_O : MIMG_Sampler_WQM <0x0000003d, AMDGPUSample_c_b_o>; |
| defm IMAGE_SAMPLE_C_LZ_O : MIMG_Sampler <0x0000003f, AMDGPUSample_c_lz_o>; |
| defm IMAGE_GATHER4 : MIMG_Gather_WQM <0x00000040, AMDGPUSample>; |
| defm IMAGE_GATHER4_CL : MIMG_Gather_WQM <0x00000041, AMDGPUSample_cl>; |
| defm IMAGE_GATHER4_L : MIMG_Gather <0x00000044, AMDGPUSample_l>; |
| defm IMAGE_GATHER4_B : MIMG_Gather_WQM <0x00000045, AMDGPUSample_b>; |
| defm IMAGE_GATHER4_B_CL : MIMG_Gather_WQM <0x00000046, AMDGPUSample_b_cl>; |
| defm IMAGE_GATHER4_LZ : MIMG_Gather <0x00000047, AMDGPUSample_lz>; |
| defm IMAGE_GATHER4_C : MIMG_Gather_WQM <0x00000048, AMDGPUSample_c>; |
| defm IMAGE_GATHER4_C_CL : MIMG_Gather_WQM <0x00000049, AMDGPUSample_c_cl>; |
| defm IMAGE_GATHER4_C_L : MIMG_Gather <0x0000004c, AMDGPUSample_c_l>; |
| defm IMAGE_GATHER4_C_B : MIMG_Gather_WQM <0x0000004d, AMDGPUSample_c_b>; |
| defm IMAGE_GATHER4_C_B_CL : MIMG_Gather_WQM <0x0000004e, AMDGPUSample_c_b_cl>; |
| defm IMAGE_GATHER4_C_LZ : MIMG_Gather <0x0000004f, AMDGPUSample_c_lz>; |
| defm IMAGE_GATHER4_O : MIMG_Gather_WQM <0x00000050, AMDGPUSample_o>; |
| defm IMAGE_GATHER4_CL_O : MIMG_Gather_WQM <0x00000051, AMDGPUSample_cl_o>; |
| defm IMAGE_GATHER4_L_O : MIMG_Gather <0x00000054, AMDGPUSample_l_o>; |
| defm IMAGE_GATHER4_B_O : MIMG_Gather_WQM <0x00000055, AMDGPUSample_b_o>; |
| defm IMAGE_GATHER4_B_CL_O : MIMG_Gather <0x00000056, AMDGPUSample_b_cl_o>; |
| defm IMAGE_GATHER4_LZ_O : MIMG_Gather <0x00000057, AMDGPUSample_lz_o>; |
| defm IMAGE_GATHER4_C_O : MIMG_Gather_WQM <0x00000058, AMDGPUSample_c_o>; |
| defm IMAGE_GATHER4_C_CL_O : MIMG_Gather_WQM <0x00000059, AMDGPUSample_c_cl_o>; |
| defm IMAGE_GATHER4_C_L_O : MIMG_Gather <0x0000005c, AMDGPUSample_c_l_o>; |
| defm IMAGE_GATHER4_C_B_O : MIMG_Gather_WQM <0x0000005d, AMDGPUSample_c_b_o>; |
| defm IMAGE_GATHER4_C_B_CL_O : MIMG_Gather_WQM <0x0000005e, AMDGPUSample_c_b_cl_o>; |
| defm IMAGE_GATHER4_C_LZ_O : MIMG_Gather <0x0000005f, AMDGPUSample_c_lz_o>; |
| |
| defm IMAGE_GET_LOD : MIMG_Sampler <0x00000060, AMDGPUSample, 1, 1, "image_get_lod">; |
| |
| defm IMAGE_SAMPLE_CD : MIMG_Sampler <0x00000068, AMDGPUSample_cd>; |
| defm IMAGE_SAMPLE_CD_CL : MIMG_Sampler <0x00000069, AMDGPUSample_cd_cl>; |
| defm IMAGE_SAMPLE_C_CD : MIMG_Sampler <0x0000006a, AMDGPUSample_c_cd>; |
| defm IMAGE_SAMPLE_C_CD_CL : MIMG_Sampler <0x0000006b, AMDGPUSample_c_cd_cl>; |
| defm IMAGE_SAMPLE_CD_O : MIMG_Sampler <0x0000006c, AMDGPUSample_cd_o>; |
| defm IMAGE_SAMPLE_CD_CL_O : MIMG_Sampler <0x0000006d, AMDGPUSample_cd_cl_o>; |
| defm IMAGE_SAMPLE_C_CD_O : MIMG_Sampler <0x0000006e, AMDGPUSample_c_cd_o>; |
| defm IMAGE_SAMPLE_C_CD_CL_O : MIMG_Sampler <0x0000006f, AMDGPUSample_c_cd_cl_o>; |
| //def IMAGE_RSRC256 : MIMG_NoPattern_RSRC256 <"image_rsrc256", 0x0000007e>; |
| //def IMAGE_SAMPLER : MIMG_NoPattern_ <"image_sampler", 0x0000007f>; |
| |
| /********** ========================================= **********/ |
| /********** Table of dimension-aware image intrinsics **********/ |
| /********** ========================================= **********/ |
| |
| class ImageDimIntrinsicInfo<AMDGPUImageDimIntrinsic I> { |
| Intrinsic Intr = I; |
| MIMGBaseOpcode BaseOpcode = !cast<MIMGBaseOpcode>(!strconcat("IMAGE_", I.P.OpMod)); |
| AMDGPUDimProps Dim = I.P.Dim; |
| } |
| |
| def ImageDimIntrinsicTable : GenericTable { |
| let FilterClass = "ImageDimIntrinsicInfo"; |
| let Fields = ["Intr", "BaseOpcode", "Dim"]; |
| GenericEnum TypeOf_BaseOpcode = MIMGBaseOpcode; |
| GenericEnum TypeOf_Dim = MIMGDim; |
| |
| let PrimaryKey = ["Intr"]; |
| let PrimaryKeyName = "getImageDimIntrinsicInfo"; |
| let PrimaryKeyEarlyOut = 1; |
| } |
| |
| foreach intr = !listconcat(AMDGPUImageDimIntrinsics, |
| AMDGPUImageDimAtomicIntrinsics) in { |
| def : ImageDimIntrinsicInfo<intr>; |
| } |
| |
| // L to LZ Optimization Mapping |
| def : MIMGLZMapping<IMAGE_SAMPLE_L, IMAGE_SAMPLE_LZ>; |
| def : MIMGLZMapping<IMAGE_SAMPLE_C_L, IMAGE_SAMPLE_C_LZ>; |
| def : MIMGLZMapping<IMAGE_SAMPLE_L_O, IMAGE_SAMPLE_LZ_O>; |
| def : MIMGLZMapping<IMAGE_SAMPLE_C_L_O, IMAGE_SAMPLE_C_LZ_O>; |
| def : MIMGLZMapping<IMAGE_GATHER4_L, IMAGE_GATHER4_LZ>; |
| def : MIMGLZMapping<IMAGE_GATHER4_C_L, IMAGE_GATHER4_C_LZ>; |
| def : MIMGLZMapping<IMAGE_GATHER4_L_O, IMAGE_GATHER4_LZ_O>; |
| def : MIMGLZMapping<IMAGE_GATHER4_C_L_O, IMAGE_GATHER4_C_LZ_O>; |
| |
| // MIP to NONMIP Optimization Mapping |
| def : MIMGMIPMapping<IMAGE_LOAD_MIP, IMAGE_LOAD>; |
| def : MIMGMIPMapping<IMAGE_STORE_MIP, IMAGE_STORE>; |
