| 1 | //===-- X86TileConfig.cpp - Tile Register Configure----------------------===// |
| 2 | // |
| 3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 4 | // See https://llvm.org/LICENSE.txt for license information. |
| 5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 6 | // |
| 7 | //===----------------------------------------------------------------------===// |
| 8 | // |
| 9 | /// \file Pass to config the shape of AMX physical registers |
| 10 | /// AMX register need to be configured before use. In X86PreTileConfig pass |
| 11 | /// the pldtilecfg instruction is inserted, however at that time we don't |
| 12 | /// know the shape of each physical tile registers, because the register |
| 13 | /// allocation is not done yet. This pass runs after egister allocation |
| 14 | /// pass. It collects the shape information of each physical tile register |
| 15 | /// and store the shape in the stack slot that is allocated for load config |
| 16 | /// to tile config register. |
| 17 | // |
| 18 | //===----------------------------------------------------------------------===// |
| 19 | |
| 20 | #include "X86.h" |
| 21 | #include "X86InstrBuilder.h" |
| 22 | #include "X86MachineFunctionInfo.h" |
| 23 | #include "X86Subtarget.h" |
| 24 | #include "llvm/CodeGen/LiveIntervals.h" |
| 25 | #include "llvm/CodeGen/MachineFrameInfo.h" |
| 26 | #include "llvm/CodeGen/MachineFunctionPass.h" |
| 27 | #include "llvm/CodeGen/MachineInstr.h" |
| 28 | #include "llvm/CodeGen/MachineRegisterInfo.h" |
| 29 | #include "llvm/CodeGen/Passes.h" |
| 30 | #include "llvm/CodeGen/TargetInstrInfo.h" |
| 31 | #include "llvm/CodeGen/TargetRegisterInfo.h" |
| 32 | #include "llvm/CodeGen/TileShapeInfo.h" |
| 33 | #include "llvm/CodeGen/VirtRegMap.h" |
| 34 | #include "llvm/InitializePasses.h" |
| 35 | |
| 36 | using namespace llvm; |
| 37 | |
| 38 | #define DEBUG_TYPE "x86-tile-config" |
| 39 | |
| 40 | namespace { |
| 41 | |
| 42 | struct X86TileConfigLegacy : public MachineFunctionPass { |
| 43 | |
| 44 | X86TileConfigLegacy() : MachineFunctionPass(ID) {} |
| 45 | |
| 46 | /// Return the pass name. |
| 47 | StringRef getPassName() const override { return "Tile Register Configure" ; } |
| 48 | |
| 49 | /// X86TileConfig analysis usage. |
| 50 | void getAnalysisUsage(AnalysisUsage &AU) const override { |
| 51 | AU.setPreservesAll(); |
| 52 | AU.addRequired<VirtRegMapWrapperLegacy>(); |
| 53 | AU.addRequired<LiveIntervalsWrapperPass>(); |
| 54 | MachineFunctionPass::getAnalysisUsage(AU); |
| 55 | } |
| 56 | |
| 57 | /// Perform register allocation. |
| 58 | bool runOnMachineFunction(MachineFunction &MF) override; |
| 59 | |
| 60 | MachineFunctionProperties getRequiredProperties() const override { |
| 61 | return MachineFunctionProperties().setNoPHIs(); |
| 62 | } |
| 63 | |
| 64 | static char ID; |
| 65 | }; |
| 66 | |
| 67 | } // end anonymous namespace |
| 68 | |
| 69 | char X86TileConfigLegacy::ID = 0; |
| 70 | |
| 71 | INITIALIZE_PASS_BEGIN(X86TileConfigLegacy, DEBUG_TYPE, |
| 72 | "Tile Register Configure" , false, false) |
| 73 | INITIALIZE_PASS_DEPENDENCY(VirtRegMapWrapperLegacy) |
| 74 | INITIALIZE_PASS_END(X86TileConfigLegacy, DEBUG_TYPE, "Tile Register Configure" , |
| 75 | false, false) |
| 76 | |
| 77 | static bool tileConfig(MachineFunction &MF, |
| 78 | llvm::function_ref<LiveIntervals *()> GetLIs, |
| 79 | llvm::function_ref<VirtRegMap *()> GetVRM) { |
| 80 | X86MachineFunctionInfo *X86FI = MF.getInfo<X86MachineFunctionInfo>(); |
| 81 | // Early exit in the common case of non-AMX code. |
| 82 | if (X86FI->getAMXProgModel() != AMXProgModelEnum::ManagedRA) |
| 83 | return false; |
| 84 | |
| 85 | const X86Subtarget &ST = MF.getSubtarget<X86Subtarget>(); |
| 86 | const X86RegisterInfo *TRI = ST.getRegisterInfo(); |
| 87 | const TargetInstrInfo *TII = ST.getInstrInfo(); |
| 88 | MachineRegisterInfo &MRI = MF.getRegInfo(); |
| 89 | LiveIntervals &LIS = *GetLIs(); |
| 90 | VirtRegMap &VRM = *GetVRM(); |
| 91 | |
| 92 | if (VRM.isShapeMapEmpty()) |
| 93 | return false; |
| 94 | |
| 95 | int SS = INT_MAX; |
| 96 | for (MachineBasicBlock &MBB : MF) { |
| 97 | for (MachineInstr &MI : MBB) { |
| 98 | if (MI.getOpcode() == X86::PLDTILECFGV) { |
| 99 | SS = MI.getOperand(i: 0).getIndex(); |
| 100 | break; |
| 101 | } |
| 102 | } |
| 103 | if (SS != INT_MAX) |
| 104 | break; |
| 105 | } |
| 106 | // Didn't find PLDTILECFGV, just return false; |
| 107 | if (SS == INT_MAX) |
| 108 | return false; |
| 109 | |
| 110 | // Try to find a point to insert MIs for constant shapes. |
| 111 | // Here we are leveraging the palette id inserted in PreRA pass. |
| 112 | unsigned ConstPos = 0; |
| 113 | MachineInstr *ConstMI = nullptr; |
| 114 | for (MachineInstr &MI : MF.front()) { |
| 115 | if (MI.getOpcode() == X86::MOV8mi && SS == MI.getOperand(i: 0).getIndex()) { |
| 116 | ConstMI = &MI; |
| 117 | break; |
| 118 | } |
| 119 | ++ConstPos; |
| 120 | } |
| 121 | assert(ConstMI && "Cannot find an insertion point" ); |
| 122 | |
| 123 | unsigned AMXRegNum = TRI->getRegClass(i: X86::TILERegClassID)->getNumRegs(); |
| 124 | SmallVector<Register, 8> Phys2Virt(AMXRegNum, 0); |
| 125 | for (unsigned I = 0, E = MRI.getNumVirtRegs(); I != E; ++I) { |
| 126 | Register VirtReg = Register::index2VirtReg(Index: I); |
| 127 | if (MRI.reg_nodbg_empty(RegNo: VirtReg)) |
| 128 | continue; |
| 129 | if (!TRI->isTileRegisterClass(RC: MRI.getRegClass(Reg: VirtReg))) |
| 130 | continue; |
| 131 | MCRegister PhysReg = VRM.getPhys(virtReg: VirtReg); |
| 132 | if (!PhysReg) |
| 133 | continue; |
| 134 | unsigned Index = PhysReg - X86::TMM0; |
| 135 | if (!Phys2Virt[Index]) |
| 136 | Phys2Virt[Index] = VirtReg; |
| 137 | } |
| 138 | |
| 139 | // Fill in the shape of each tile physical register. |
| 140 | for (unsigned I = 0; I < AMXRegNum; ++I) { |
| 141 | if (!Phys2Virt[I]) |
| 142 | continue; |
| 143 | DebugLoc DL; |
| 144 | bool IsRow = true; |
| 145 | MachineInstr *NewMI = nullptr; |
| 146 | ShapeT Shape = VRM.getShape(virtReg: Phys2Virt[I]); |
| 147 | for (auto &R : {Shape.getRow()->getReg(), Shape.getCol()->getReg()}) { |
| 148 | // Here is the data format for the tile config. |
| 149 | // 0 palette |
| 150 | // 1 start_row |
| 151 | // 2-15 reserved, must be zero |
| 152 | // 16-17 tile0.colsb Tile 0 bytes per row. |
| 153 | // 18-19 tile1.colsb Tile 1 bytes per row. |
| 154 | // 20-21 tile2.colsb Tile 2 bytes per row. |
| 155 | // ... (sequence continues) |
| 156 | // 30-31 tile7.colsb Tile 7 bytes per row. |
| 157 | // 32-47 reserved, must be zero |
| 158 | // 48 tile0.rows Tile 0 rows. |
| 159 | // 49 tile1.rows Tile 1 rows. |
| 160 | // 50 tile2.rows Tile 2 rows. |
| 161 | // ... (sequence continues) |
| 162 | // 55 tile7.rows Tile 7 rows. |
| 163 | // 56-63 reserved, must be zero |
| 164 | int64_t Imm = INT64_MAX; |
| 165 | int Offset = IsRow ? 48 + I : 16 + I * 2; |
| 166 | for (auto &DefMI : MRI.def_instructions(Reg: R)) { |
| 167 | MachineBasicBlock &MBB = *DefMI.getParent(); |
| 168 | if (DefMI.isMoveImmediate()) { |
| 169 | if (Imm != INT64_MAX) { |
| 170 | // FIXME: We should handle this case in future. |
| 171 | assert(Imm == DefMI.getOperand(1).getImm() && |
| 172 | "Cannot initialize with different shapes" ); |
| 173 | continue; |
| 174 | } |
| 175 | if (DefMI.getOperand(i: 1).isImm()) { |
| 176 | Imm = DefMI.getOperand(i: 1).getImm(); |
| 177 | } else { |
| 178 | assert(DefMI.getOpcode() == X86::MOV32r0 && |
| 179 | "The opcode is assumed to be MOV32r0 if the operand is not " |
| 180 | "immediate." ); |
| 181 | Imm = 0; |
| 182 | } |
| 183 | |
| 184 | NewMI = addFrameReference( |
| 185 | MIB: BuildMI(BB&: MF.front(), I: ++ConstMI->getIterator(), MIMD: DL, |
| 186 | MCID: TII->get(Opcode: IsRow ? X86::MOV8mi : X86::MOV16mi)), |
| 187 | FI: SS, Offset) |
| 188 | .addImm(Val: Imm); |
| 189 | ConstMI = NewMI; |
| 190 | LIS.InsertMachineInstrInMaps(MI&: *NewMI); |
| 191 | } else { |
| 192 | unsigned SubIdx = IsRow ? X86::sub_8bit : X86::sub_16bit; |
| 193 | unsigned RegSize = TRI->getRegSizeInBits(RC: *MRI.getRegClass(Reg: R)); |
| 194 | if ((IsRow && RegSize == 8) || (!IsRow && RegSize == 16)) |
| 195 | SubIdx = 0; |
| 196 | auto Iter = DefMI.getIterator(); |
| 197 | if (&MBB == &MF.front() && |
| 198 | (unsigned)std::distance(first: MBB.instr_begin(), last: Iter) < ConstPos) |
| 199 | Iter = ConstMI->getIterator(); |
| 200 | NewMI = addFrameReference( |
| 201 | MIB: BuildMI(BB&: MBB, I: ++Iter, MIMD: DL, |
| 202 | MCID: TII->get(Opcode: IsRow ? X86::MOV8mr : X86::MOV16mr)), |
| 203 | FI: SS, Offset) |
| 204 | .addReg(RegNo: R, Flags: {}, SubReg: SubIdx); |
| 205 | SlotIndex SIdx = LIS.InsertMachineInstrInMaps(MI&: *NewMI); |
| 206 | LIS.extendToIndices(LR&: LIS.getInterval(Reg: R), Indices: {SIdx.getRegSlot()}); |
| 207 | } |
| 208 | } |
| 209 | IsRow = false; |
| 210 | } |
| 211 | } |
| 212 | return true; |
| 213 | } |
| 214 | |
| 215 | FunctionPass *llvm::createX86TileConfigLegacyPass() { |
| 216 | return new X86TileConfigLegacy(); |
| 217 | } |
| 218 | |
| 219 | bool X86TileConfigLegacy::runOnMachineFunction(MachineFunction &MF) { |
| 220 | return tileConfig( |
| 221 | MF, |
| 222 | GetLIs: [this]() { return &getAnalysis<LiveIntervalsWrapperPass>().getLIS(); }, |
| 223 | GetVRM: [this]() { return &getAnalysis<VirtRegMapWrapperLegacy>().getVRM(); }); |
| 224 | } |
| 225 | |
| 226 | PreservedAnalyses X86TileConfigPass::run(MachineFunction &MF, |
| 227 | MachineFunctionAnalysisManager &MFAM) { |
| 228 | bool Changed = tileConfig( |
| 229 | MF, GetLIs: [&MFAM, &MF]() { return &MFAM.getResult<LiveIntervalsAnalysis>(IR&: MF); }, |
| 230 | GetVRM: [&MFAM, &MF]() { return &MFAM.getResult<VirtRegMapAnalysis>(IR&: MF); }); |
| 231 | return Changed ? getMachineFunctionPassPreservedAnalyses() |
| 232 | .preserveSet<CFGAnalyses>() |
| 233 | : PreservedAnalyses::all(); |
| 234 | } |
| 235 | |