//===- AArch64LegalizerInfo.cpp ----------------------------------*- C++ -*-==// // // 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 // //===----------------------------------------------------------------------===// /// \file /// This file implements the targeting of the Machinelegalizer class for /// AArch64. /// \todo This should be generated by TableGen. //===----------------------------------------------------------------------===// #include "AArch64LegalizerInfo.h" #include "AArch64RegisterBankInfo.h" #include "AArch64Subtarget.h" #include "llvm/ADT/STLExtras.h" #include "llvm/CodeGen/GlobalISel/GenericMachineInstrs.h" #include "llvm/CodeGen/GlobalISel/LegalizerHelper.h" #include "llvm/CodeGen/GlobalISel/LegalizerInfo.h" #include "llvm/CodeGen/GlobalISel/MIPatternMatch.h" #include "llvm/CodeGen/GlobalISel/MachineIRBuilder.h" #include "llvm/CodeGen/GlobalISel/Utils.h" #include "llvm/CodeGen/MachineInstr.h" #include "llvm/CodeGen/MachineRegisterInfo.h" #include "llvm/CodeGen/TargetOpcodes.h" #include "llvm/CodeGen/ValueTypes.h" #include "llvm/IR/DerivedTypes.h" #include "llvm/IR/Intrinsics.h" #include "llvm/IR/IntrinsicsAArch64.h" #include "llvm/IR/Type.h" #include "llvm/Support/MathExtras.h" #include #define DEBUG_TYPE "aarch64-legalinfo" using namespace llvm; using namespace LegalizeActions; using namespace LegalizeMutations; using namespace LegalityPredicates; using namespace MIPatternMatch; AArch64LegalizerInfo::AArch64LegalizerInfo(const AArch64Subtarget &ST) : ST(&ST) { using namespace TargetOpcode; const LLT p0 = LLT::pointer(0, 64); const LLT s8 = LLT::scalar(8); const LLT s16 = LLT::scalar(16); const LLT s32 = LLT::scalar(32); const LLT s64 = LLT::scalar(64); const LLT s128 = LLT::scalar(128); const LLT v16s8 = LLT::fixed_vector(16, 8); const LLT v8s8 = LLT::fixed_vector(8, 8); const LLT v4s8 = LLT::fixed_vector(4, 8); const LLT v2s8 = LLT::fixed_vector(2, 8); const LLT v8s16 = LLT::fixed_vector(8, 16); const LLT v4s16 = LLT::fixed_vector(4, 16); const LLT v2s16 = LLT::fixed_vector(2, 16); const LLT v2s32 = LLT::fixed_vector(2, 32); const LLT v4s32 = LLT::fixed_vector(4, 32); const LLT v2s64 = LLT::fixed_vector(2, 64); const LLT v2p0 = LLT::fixed_vector(2, p0); const LLT nxv16s8 = LLT::scalable_vector(16, s8); const LLT nxv8s16 = LLT::scalable_vector(8, s16); const LLT nxv4s32 = LLT::scalable_vector(4, s32); const LLT nxv2s64 = LLT::scalable_vector(2, s64); std::initializer_list PackedVectorAllTypeList = {/* Begin 128bit types */ v16s8, v8s16, v4s32, v2s64, v2p0, /* End 128bit types */ /* Begin 64bit types */ v8s8, v4s16, v2s32}; std::initializer_list ScalarAndPtrTypesList = {s8, s16, s32, s64, p0}; SmallVector PackedVectorAllTypesVec(PackedVectorAllTypeList); SmallVector ScalarAndPtrTypesVec(ScalarAndPtrTypesList); const TargetMachine &TM = ST.getTargetLowering()->getTargetMachine(); // FIXME: support subtargets which have neon/fp-armv8 disabled. if (!ST.hasNEON() || !ST.hasFPARMv8()) { getLegacyLegalizerInfo().computeTables(); return; } // Some instructions only support s16 if the subtarget has full 16-bit FP // support. const bool HasFP16 = ST.hasFullFP16(); const LLT &MinFPScalar = HasFP16 ? s16 : s32; const bool HasCSSC = ST.hasCSSC(); const bool HasRCPC3 = ST.hasRCPC3(); getActionDefinitionsBuilder( {G_IMPLICIT_DEF, G_FREEZE, G_CONSTANT_FOLD_BARRIER}) .legalFor({p0, s8, s16, s32, s64}) .legalFor(PackedVectorAllTypeList) .widenScalarToNextPow2(0) .clampScalar(0, s8, s64) .moreElementsToNextPow2(0) .widenVectorEltsToVectorMinSize(0, 64) .clampNumElements(0, v8s8, v16s8) .clampNumElements(0, v4s16, v8s16) .clampNumElements(0, v2s32, v4s32) .clampNumElements(0, v2s64, v2s64); getActionDefinitionsBuilder(G_PHI) .legalFor({p0, s16, s32, s64}) .legalFor(PackedVectorAllTypeList) .widenScalarToNextPow2(0) .clampScalar(0, s16, s64) // Maximum: sN * k = 128 .clampMaxNumElements(0, s8, 16) .clampMaxNumElements(0, s16, 8) .clampMaxNumElements(0, s32, 4) .clampMaxNumElements(0, s64, 2) .clampMaxNumElements(0, p0, 2); getActionDefinitionsBuilder(G_BSWAP) .legalFor({s32, s64, v4s16, v8s16, v2s32, v4s32, v2s64}) .widenScalarOrEltToNextPow2(0, 16) .clampScalar(0, s32, s64) .clampNumElements(0, v4s16, v8s16) .clampNumElements(0, v2s32, v4s32) .clampNumElements(0, v2s64, v2s64) .moreElementsToNextPow2(0); getActionDefinitionsBuilder({G_ADD, G_SUB, G_MUL, G_AND, G_OR, G_XOR}) .legalFor({s32, s64, v2s32, v2s64, v4s32, v4s16, v8s16, v16s8, v8s8}) .widenScalarToNextPow2(0) .clampScalar(0, s32, s64) .clampMaxNumElements(0, s8, 16) .clampMaxNumElements(0, s16, 8) .clampNumElements(0, v2s32, v4s32) .clampNumElements(0, v2s64, v2s64) .minScalarOrEltIf( [=](const LegalityQuery &Query) { return Query.Types[0].getNumElements() <= 2; }, 0, s32) .minScalarOrEltIf( [=](const LegalityQuery &Query) { return Query.Types[0].getNumElements() <= 4; }, 0, s16) .minScalarOrEltIf( [=](const LegalityQuery &Query) { return Query.Types[0].getNumElements() <= 16; }, 0, s8) .moreElementsToNextPow2(0); getActionDefinitionsBuilder({G_SHL, G_ASHR, G_LSHR}) .customIf([=](const LegalityQuery &Query) { const auto &SrcTy = Query.Types[0]; const auto &AmtTy = Query.Types[1]; return !SrcTy.isVector() && SrcTy.getSizeInBits() == 32 && AmtTy.getSizeInBits() == 32; }) .legalFor({ {s32, s32}, {s32, s64}, {s64, s64}, {v8s8, v8s8}, {v16s8, v16s8}, {v4s16, v4s16}, {v8s16, v8s16}, {v2s32, v2s32}, {v4s32, v4s32}, {v2s64, v2s64}, }) .widenScalarToNextPow2(0) .clampScalar(1, s32, s64) .clampScalar(0, s32, s64) .clampNumElements(0, v8s8, v16s8) .clampNumElements(0, v4s16, v8s16) .clampNumElements(0, v2s32, v4s32) .clampNumElements(0, v2s64, v2s64) .moreElementsToNextPow2(0) .minScalarSameAs(1, 0); getActionDefinitionsBuilder(G_PTR_ADD) .legalFor({{p0, s64}, {v2p0, v2s64}}) .clampScalarOrElt(1, s64, s64) .clampNumElements(0, v2p0, v2p0); getActionDefinitionsBuilder(G_PTRMASK).legalFor({{p0, s64}}); getActionDefinitionsBuilder({G_SDIV, G_UDIV}) .legalFor({s32, s64}) .libcallFor({s128}) .clampScalar(0, s32, s64) .widenScalarToNextPow2(0) .scalarize(0); getActionDefinitionsBuilder({G_SREM, G_UREM, G_SDIVREM, G_UDIVREM}) .lowerFor({s8, s16, s32, s64, v2s64, v4s32, v2s32}) .widenScalarOrEltToNextPow2(0) .clampScalarOrElt(0, s32, s64) .clampNumElements(0, v2s32, v4s32) .clampNumElements(0, v2s64, v2s64) .moreElementsToNextPow2(0); getActionDefinitionsBuilder({G_SMULO, G_UMULO}) .widenScalarToNextPow2(0, /*Min = */ 32) .clampScalar(0, s32, s64) .lower(); getActionDefinitionsBuilder({G_SMULH, G_UMULH}) .legalFor({s64, v8s16, v16s8, v4s32}) .lower(); auto &MinMaxActions = getActionDefinitionsBuilder( {G_SMIN, G_SMAX, G_UMIN, G_UMAX}); if (HasCSSC) MinMaxActions .legalFor({s32, s64, v8s8, v16s8, v4s16, v8s16, v2s32, v4s32}) // Making clamping conditional on CSSC extension as without legal types we // lower to CMP which can fold one of the two sxtb's we'd otherwise need // if we detect a type smaller than 32-bit. .minScalar(0, s32); else MinMaxActions .legalFor({v8s8, v16s8, v4s16, v8s16, v2s32, v4s32}); MinMaxActions .clampNumElements(0, v8s8, v16s8) .clampNumElements(0, v4s16, v8s16) .clampNumElements(0, v2s32, v4s32) // FIXME: This sholdn't be needed as v2s64 types are going to // be expanded anyway, but G_ICMP doesn't support splitting vectors yet .clampNumElements(0, v2s64, v2s64) .lower(); getActionDefinitionsBuilder( {G_SADDE, G_SSUBE, G_UADDE, G_USUBE, G_SADDO, G_SSUBO, G_UADDO, G_USUBO}) .legalFor({{s32, s32}, {s64, s32}}) .clampScalar(0, s32, s64) .clampScalar(1, s32, s64) .widenScalarToNextPow2(0); getActionDefinitionsBuilder({G_FADD, G_FSUB, G_FMUL, G_FDIV, G_FMA, G_FNEG, G_FABS, G_FSQRT, G_FMAXNUM, G_FMINNUM, G_FMAXIMUM, G_FMINIMUM, G_FCEIL, G_FFLOOR, G_FRINT, G_FNEARBYINT, G_INTRINSIC_TRUNC, G_INTRINSIC_ROUND, G_INTRINSIC_ROUNDEVEN}) .legalFor({MinFPScalar, s32, s64, v2s32, v4s32, v2s64}) .legalIf([=](const LegalityQuery &Query) { const auto &Ty = Query.Types[0]; return (Ty == v8s16 || Ty == v4s16) && HasFP16; }) .libcallFor({s128}) .minScalarOrElt(0, MinFPScalar) .clampNumElements(0, v4s16, v8s16) .clampNumElements(0, v2s32, v4s32) .clampNumElements(0, v2s64, v2s64) .moreElementsToNextPow2(0); getActionDefinitionsBuilder(G_FREM) .libcallFor({s32, s64}) .minScalar(0, s32) .scalarize(0); getActionDefinitionsBuilder({G_INTRINSIC_LRINT, G_INTRINSIC_LLRINT}) .legalFor({{s64, MinFPScalar}, {s64, s32}, {s64, s64}}) .libcallFor({{s64, s128}}) .minScalarOrElt(1, MinFPScalar); getActionDefinitionsBuilder( {G_FCOS, G_FSIN, G_FPOW, G_FLOG, G_FLOG2, G_FLOG10, G_FTAN, G_FEXP, G_FEXP2, G_FEXP10, G_FACOS, G_FASIN, G_FATAN, G_FCOSH, G_FSINH, G_FTANH}) // We need a call for these, so we always need to scalarize. .scalarize(0) // Regardless of FP16 support, widen 16-bit elements to 32-bits. .minScalar(0, s32) .libcallFor({s32, s64}); getActionDefinitionsBuilder(G_FPOWI) .scalarize(0) .minScalar(0, s32) .libcallFor({{s32, s32}, {s64, s32}}); getActionDefinitionsBuilder(G_INSERT) .legalIf(all(typeInSet(0, {s32, s64, p0}), typeInSet(1, {s8, s16, s32}), smallerThan(1, 0))) .widenScalarToNextPow2(0) .clampScalar(0, s32, s64) .widenScalarToNextPow2(1) .minScalar(1, s8) .maxScalarIf(typeInSet(0, {s32}), 1, s16) .maxScalarIf(typeInSet(0, {s64, p0}), 1, s32); getActionDefinitionsBuilder(G_EXTRACT) .legalIf(all(typeInSet(0, {s16, s32, s64, p0}), typeInSet(1, {s32, s64, s128, p0}), smallerThan(0, 1))) .widenScalarToNextPow2(1) .clampScalar(1, s32, s128) .widenScalarToNextPow2(0) .minScalar(0, s16) .maxScalarIf(typeInSet(1, {s32}), 0, s16) .maxScalarIf(typeInSet(1, {s64, p0}), 0, s32) .maxScalarIf(typeInSet(1, {s128}), 0, s64); for (unsigned Op : {G_SEXTLOAD, G_ZEXTLOAD}) { auto &Actions = getActionDefinitionsBuilder(Op); if (Op == G_SEXTLOAD) Actions.lowerIf(atomicOrderingAtLeastOrStrongerThan(0, AtomicOrdering::Unordered)); // Atomics have zero extending behavior. Actions .legalForTypesWithMemDesc({{s32, p0, s8, 8}, {s32, p0, s16, 8}, {s32, p0, s32, 8}, {s64, p0, s8, 2}, {s64, p0, s16, 2}, {s64, p0, s32, 4}, {s64, p0, s64, 8}, {p0, p0, s64, 8}, {v2s32, p0, s64, 8}}) .widenScalarToNextPow2(0) .clampScalar(0, s32, s64) // TODO: We could support sum-of-pow2's but the lowering code doesn't know // how to do that yet. .unsupportedIfMemSizeNotPow2() // Lower anything left over into G_*EXT and G_LOAD .lower(); } auto IsPtrVecPred = [=](const LegalityQuery &Query) { const LLT &ValTy = Query.Types[0]; return ValTy.isPointerVector() && ValTy.getAddressSpace() == 0; }; auto &LoadActions = getActionDefinitionsBuilder(G_LOAD); auto &StoreActions = getActionDefinitionsBuilder(G_STORE); if (ST.hasSVE()) { LoadActions.legalForTypesWithMemDesc({ // 128 bit base sizes {nxv16s8, p0, nxv16s8, 8}, {nxv8s16, p0, nxv8s16, 8}, {nxv4s32, p0, nxv4s32, 8}, {nxv2s64, p0, nxv2s64, 8}, }); // TODO: Add nxv2p0. Consider bitcastIf. // See #92130 // https://github.com/llvm/llvm-project/pull/92130#discussion_r1616888461 StoreActions.legalForTypesWithMemDesc({ // 128 bit base sizes {nxv16s8, p0, nxv16s8, 8}, {nxv8s16, p0, nxv8s16, 8}, {nxv4s32, p0, nxv4s32, 8}, {nxv2s64, p0, nxv2s64, 8}, }); } LoadActions .customIf([=](const LegalityQuery &Query) { return HasRCPC3 && Query.Types[0] == s128 && Query.MMODescrs[0].Ordering == AtomicOrdering::Acquire; }) .customIf([=](const LegalityQuery &Query) { return Query.Types[0] == s128 && Query.MMODescrs[0].Ordering != AtomicOrdering::NotAtomic; }) .legalForTypesWithMemDesc({{s8, p0, s8, 8}, {s16, p0, s16, 8}, {s32, p0, s32, 8}, {s64, p0, s64, 8}, {p0, p0, s64, 8}, {s128, p0, s128, 8}, {v8s8, p0, s64, 8}, {v16s8, p0, s128, 8}, {v4s16, p0, s64, 8}, {v8s16, p0, s128, 8}, {v2s32, p0, s64, 8}, {v4s32, p0, s128, 8}, {v2s64, p0, s128, 8}}) // These extends are also legal .legalForTypesWithMemDesc( {{s32, p0, s8, 8}, {s32, p0, s16, 8}, {s64, p0, s32, 8}}) .widenScalarToNextPow2(0, /* MinSize = */ 8) .clampMaxNumElements(0, s8, 16) .clampMaxNumElements(0, s16, 8) .clampMaxNumElements(0, s32, 4) .clampMaxNumElements(0, s64, 2) .clampMaxNumElements(0, p0, 2) .lowerIfMemSizeNotByteSizePow2() .clampScalar(0, s8, s64) .narrowScalarIf( [=](const LegalityQuery &Query) { // Clamp extending load results to 32-bits. return Query.Types[0].isScalar() && Query.Types[0] != Query.MMODescrs[0].MemoryTy && Query.Types[0].getSizeInBits() > 32; }, changeTo(0, s32)) // TODO: Use BITCAST for v2i8, v2i16 after G_TRUNC gets sorted out .bitcastIf(typeInSet(0, {v4s8}), [=](const LegalityQuery &Query) { const LLT VecTy = Query.Types[0]; return std::pair(0, LLT::scalar(VecTy.getSizeInBits())); }) .customIf(IsPtrVecPred) .scalarizeIf(typeInSet(0, {v2s16, v2s8}), 0); StoreActions .customIf([=](const LegalityQuery &Query) { return HasRCPC3 && Query.Types[0] == s128 && Query.MMODescrs[0].Ordering == AtomicOrdering::Release; }) .customIf([=](const LegalityQuery &Query) { return Query.Types[0] == s128 && Query.MMODescrs[0].Ordering != AtomicOrdering::NotAtomic; }) .legalForTypesWithMemDesc( {{s8, p0, s8, 8}, {s16, p0, s8, 8}, // truncstorei8 from s16 {s32, p0, s8, 8}, // truncstorei8 from s32 {s64, p0, s8, 8}, // truncstorei8 from s64 {s16, p0, s16, 8}, {s32, p0, s16, 8}, // truncstorei16 from s32 {s64, p0, s16, 8}, // truncstorei16 from s64 {s32, p0, s8, 8}, {s32, p0, s16, 8}, {s32, p0, s32, 8}, {s64, p0, s64, 8}, {s64, p0, s32, 8}, // truncstorei32 from s64 {p0, p0, s64, 8}, {s128, p0, s128, 8}, {v16s8, p0, s128, 8}, {v8s8, p0, s64, 8}, {v4s16, p0, s64, 8}, {v8s16, p0, s128, 8}, {v2s32, p0, s64, 8}, {v4s32, p0, s128, 8}, {v2s64, p0, s128, 8}}) .clampScalar(0, s8, s64) .lowerIf([=](const LegalityQuery &Query) { return Query.Types[0].isScalar() && Query.Types[0] != Query.MMODescrs[0].MemoryTy; }) // Maximum: sN * k = 128 .clampMaxNumElements(0, s8, 16) .clampMaxNumElements(0, s16, 8) .clampMaxNumElements(0, s32, 4) .clampMaxNumElements(0, s64, 2) .clampMaxNumElements(0, p0, 2) .lowerIfMemSizeNotPow2() // TODO: Use BITCAST for v2i8, v2i16 after G_TRUNC gets sorted out .bitcastIf(typeInSet(0, {v4s8}), [=](const LegalityQuery &Query) { const LLT VecTy = Query.Types[0]; return std::pair(0, LLT::scalar(VecTy.getSizeInBits())); }) .customIf(IsPtrVecPred) .scalarizeIf(typeInSet(0, {v2s16, v2s8}), 0); getActionDefinitionsBuilder(G_INDEXED_STORE) // Idx 0 == Ptr, Idx 1 == Val // TODO: we can implement legalizations but as of now these are // generated in a very specific way. .legalForTypesWithMemDesc({ {p0, s8, s8, 8}, {p0, s16, s16, 8}, {p0, s32, s8, 8}, {p0, s32, s16, 8}, {p0, s32, s32, 8}, {p0, s64, s64, 8}, {p0, p0, p0, 8}, {p0, v8s8, v8s8, 8}, {p0, v16s8, v16s8, 8}, {p0, v4s16, v4s16, 8}, {p0, v8s16, v8s16, 8}, {p0, v2s32, v2s32, 8}, {p0, v4s32, v4s32, 8}, {p0, v2s64, v2s64, 8}, {p0, v2p0, v2p0, 8}, {p0, s128, s128, 8}, }) .unsupported(); auto IndexedLoadBasicPred = [=](const LegalityQuery &Query) { LLT LdTy = Query.Types[0]; LLT PtrTy = Query.Types[1]; if (!llvm::is_contained(PackedVectorAllTypesVec, LdTy) && !llvm::is_contained(ScalarAndPtrTypesVec, LdTy) && LdTy != s128) return false; if (PtrTy != p0) return false; return true; }; getActionDefinitionsBuilder(G_INDEXED_LOAD) .unsupportedIf( atomicOrderingAtLeastOrStrongerThan(0, AtomicOrdering::Unordered)) .legalIf(IndexedLoadBasicPred) .unsupported(); getActionDefinitionsBuilder({G_INDEXED_SEXTLOAD, G_INDEXED_ZEXTLOAD}) .unsupportedIf( atomicOrderingAtLeastOrStrongerThan(0, AtomicOrdering::Unordered)) .legalIf(all(typeInSet(0, {s16, s32, s64}), LegalityPredicate([=](const LegalityQuery &Q) { LLT LdTy = Q.Types[0]; LLT PtrTy = Q.Types[1]; LLT MemTy = Q.MMODescrs[0].MemoryTy; if (PtrTy != p0) return false; if (LdTy == s16) return MemTy == s8; if (LdTy == s32) return MemTy == s8 || MemTy == s16; if (LdTy == s64) return MemTy == s8 || MemTy == s16 || MemTy == s32; return false; }))) .unsupported(); // Constants getActionDefinitionsBuilder(G_CONSTANT) .legalFor({p0, s8, s16, s32, s64}) .widenScalarToNextPow2(0) .clampScalar(0, s8, s64); getActionDefinitionsBuilder(G_FCONSTANT) .legalIf([=](const LegalityQuery &Query) { const auto &Ty = Query.Types[0]; if (HasFP16 && Ty == s16) return true; return Ty == s32 || Ty == s64 || Ty == s128; }) .clampScalar(0, MinFPScalar, s128); // FIXME: fix moreElementsToNextPow2 getActionDefinitionsBuilder(G_ICMP) .legalFor({{s32, s32}, {s32, s64}, {s32, p0}}) .widenScalarOrEltToNextPow2(1) .clampScalar(1, s32, s64) .clampScalar(0, s32, s32) .minScalarEltSameAsIf( [=](const LegalityQuery &Query) { const LLT &Ty = Query.Types[0]; const LLT &SrcTy = Query.Types[1]; return Ty.isVector() && !SrcTy.isPointerVector() && Ty.getElementType() != SrcTy.getElementType(); }, 0, 1) .minScalarOrEltIf( [=](const LegalityQuery &Query) { return Query.Types[1] == v2s16; }, 1, s32) .minScalarOrEltIf( [=](const LegalityQuery &Query) { return Query.Types[1] == v2p0; }, 0, s64) .moreElementsToNextPow2(1) .clampNumElements(1, v8s8, v16s8) .clampNumElements(1, v4s16, v8s16) .clampNumElements(1, v2s32, v4s32) .clampNumElements(1, v2s64, v2s64) .customIf(isVector(0)); getActionDefinitionsBuilder(G_FCMP) .legalFor({{s32, MinFPScalar}, {s32, s32}, {s32, s64}, {v4s32, v4s32}, {v2s32, v2s32}, {v2s64, v2s64}}) .legalIf([=](const LegalityQuery &Query) { const auto &Ty = Query.Types[1]; return (Ty == v8s16 || Ty == v4s16) && Ty == Query.Types[0] && HasFP16; }) .widenScalarOrEltToNextPow2(1) .clampScalar(0, s32, s32) .clampScalarOrElt(1, MinFPScalar, s64) .minScalarEltSameAsIf( [=](const LegalityQuery &Query) { const LLT &Ty = Query.Types[0]; const LLT &SrcTy = Query.Types[1]; return Ty.isVector() && !SrcTy.isPointerVector() && Ty.getElementType() != SrcTy.getElementType(); }, 0, 1) .clampNumElements(1, v4s16, v8s16) .clampNumElements(1, v2s32, v4s32) .clampMaxNumElements(1, s64, 2) .moreElementsToNextPow2(1); // Extensions auto ExtLegalFunc = [=](const LegalityQuery &Query) { unsigned DstSize = Query.Types[0].getSizeInBits(); // Handle legal vectors using legalFor if (Query.Types[0].isVector()) return false; if (DstSize < 8 || DstSize >= 128 || !isPowerOf2_32(DstSize)) return false; // Extending to a scalar s128 needs narrowing. const LLT &SrcTy = Query.Types[1]; // Make sure we fit in a register otherwise. Don't bother checking that // the source type is below 128 bits. We shouldn't be allowing anything // through which is wider than the destination in the first place. unsigned SrcSize = SrcTy.getSizeInBits(); if (SrcSize < 8 || !isPowerOf2_32(SrcSize)) return false; return true; }; getActionDefinitionsBuilder({G_ZEXT, G_SEXT, G_ANYEXT}) .legalIf(ExtLegalFunc) .legalFor({{v2s64, v2s32}, {v4s32, v4s16}, {v8s16, v8s8}}) .clampScalar(0, s64, s64) // Just for s128, others are handled above. .moreElementsToNextPow2(0) .clampMaxNumElements(1, s8, 8) .clampMaxNumElements(1, s16, 4) .clampMaxNumElements(1, s32, 2) // Tries to convert a large EXTEND into two smaller EXTENDs .lowerIf([=](const LegalityQuery &Query) { return (Query.Types[0].getScalarSizeInBits() > Query.Types[1].getScalarSizeInBits() * 2) && Query.Types[0].isVector() && (Query.Types[1].getScalarSizeInBits() == 8 || Query.Types[1].getScalarSizeInBits() == 16); }) .clampMinNumElements(1, s8, 8) .clampMinNumElements(1, s16, 4); getActionDefinitionsBuilder(G_TRUNC) .legalFor({{v2s32, v2s64}, {v4s16, v4s32}, {v8s8, v8s16}}) .moreElementsToNextPow2(0) .clampMaxNumElements(0, s8, 8) .clampMaxNumElements(0, s16, 4) .clampMaxNumElements(0, s32, 2) .minScalarOrEltIf( [=](const LegalityQuery &Query) { return Query.Types[0].isVector(); }, 0, s8) .lowerIf([=](const LegalityQuery &Query) { LLT DstTy = Query.Types[0]; LLT SrcTy = Query.Types[1]; return DstTy.isVector() && SrcTy.getSizeInBits() > 128 && DstTy.getScalarSizeInBits() * 2 <= SrcTy.getScalarSizeInBits(); }) .clampMinNumElements(0, s8, 8) .clampMinNumElements(0, s16, 4) .alwaysLegal(); getActionDefinitionsBuilder(G_SEXT_INREG) .legalFor({s32, s64}) .legalFor(PackedVectorAllTypeList) .maxScalar(0, s64) .clampNumElements(0, v8s8, v16s8) .clampNumElements(0, v4s16, v8s16) .clampNumElements(0, v2s32, v4s32) .clampMaxNumElements(0, s64, 2) .lower(); // FP conversions getActionDefinitionsBuilder(G_FPTRUNC) .legalFor( {{s16, s32}, {s16, s64}, {s32, s64}, {v4s16, v4s32}, {v2s32, v2s64}}) .libcallFor({{s16, s128}, {s32, s128}, {s64, s128}}) .clampNumElements(0, v4s16, v4s16) .clampNumElements(0, v2s32, v2s32) .scalarize(0); getActionDefinitionsBuilder(G_FPEXT) .legalFor( {{s32, s16}, {s64, s16}, {s64, s32}, {v4s32, v4s16}, {v2s64, v2s32}}) .libcallFor({{s128, s64}, {s128, s32}, {s128, s16}}) .clampNumElements(0, v4s32, v4s32) .clampNumElements(0, v2s64, v2s64) .scalarize(0); // Conversions getActionDefinitionsBuilder({G_FPTOSI, G_FPTOUI}) .legalFor({{s32, s32}, {s64, s32}, {s32, s64}, {s64, s64}, {v2s64, v2s64}, {v4s32, v4s32}, {v2s32, v2s32}}) .legalIf([=](const LegalityQuery &Query) { return HasFP16 && (Query.Types[1] == s16 || Query.Types[1] == v4s16 || Query.Types[1] == v8s16) && (Query.Types[0] == s32 || Query.Types[0] == s64 || Query.Types[0] == v4s16 || Query.Types[0] == v8s16); }) .scalarizeIf(scalarOrEltWiderThan(0, 64), 0) .scalarizeIf(scalarOrEltWiderThan(1, 64), 1) // The range of a fp16 value fits into an i17, so we can lower the width // to i64. .narrowScalarIf( [=](const LegalityQuery &Query) { return Query.Types[1] == s16 && Query.Types[0].getSizeInBits() > 64; }, changeTo(0, s64)) .moreElementsToNextPow2(0) .widenScalarOrEltToNextPow2OrMinSize(0) .minScalar(0, s32) .widenScalarOrEltToNextPow2OrMinSize(1, /*MinSize=*/HasFP16 ? 16 : 32) .widenScalarIf( [=](const LegalityQuery &Query) { return Query.Types[0].getScalarSizeInBits() <= 64 && Query.Types[0].getScalarSizeInBits() > Query.Types[1].getScalarSizeInBits(); }, LegalizeMutations::changeElementSizeTo(1, 0)) .widenScalarIf( [=](const LegalityQuery &Query) { return Query.Types[1].getScalarSizeInBits() <= 64 && Query.Types[0].getScalarSizeInBits() < Query.Types[1].getScalarSizeInBits(); }, LegalizeMutations::changeElementSizeTo(0, 1)) .clampNumElements(0, v4s16, v8s16) .clampNumElements(0, v2s32, v4s32) .clampMaxNumElements(0, s64, 2) .libcallFor( {{s32, s128}, {s64, s128}, {s128, s128}, {s128, s32}, {s128, s64}}); getActionDefinitionsBuilder({G_SITOFP, G_UITOFP}) .legalFor({{s32, s32}, {s64, s32}, {s32, s64}, {s64, s64}, {v2s64, v2s64}, {v4s32, v4s32}, {v2s32, v2s32}}) .legalIf([=](const LegalityQuery &Query) { return HasFP16 && (Query.Types[0] == s16 || Query.Types[0] == v4s16 || Query.Types[0] == v8s16) && (Query.Types[1] == s32 || Query.Types[1] == s64 || Query.Types[1] == v4s16 || Query.Types[1] == v8s16); }) .scalarizeIf(scalarOrEltWiderThan(1, 64), 1) .scalarizeIf(scalarOrEltWiderThan(0, 64), 0) .moreElementsToNextPow2(1) .widenScalarOrEltToNextPow2OrMinSize(1) .minScalar(1, s32) .widenScalarOrEltToNextPow2OrMinSize(0, /*MinSize=*/HasFP16 ? 16 : 32) .widenScalarIf( [=](const LegalityQuery &Query) { return Query.Types[1].getScalarSizeInBits() <= 64 && Query.Types[0].getScalarSizeInBits() < Query.Types[1].getScalarSizeInBits(); }, LegalizeMutations::changeElementSizeTo(0, 1)) .widenScalarIf( [=](const LegalityQuery &Query) { return Query.Types[0].getScalarSizeInBits() <= 64 && Query.Types[0].getScalarSizeInBits() > Query.Types[1].getScalarSizeInBits(); }, LegalizeMutations::changeElementSizeTo(1, 0)) .clampNumElements(0, v4s16, v8s16) .clampNumElements(0, v2s32, v4s32) .clampMaxNumElements(0, s64, 2) .libcallFor({{s16, s128}, {s32, s128}, {s64, s128}, {s128, s128}, {s128, s32}, {s128, s64}}); // Control-flow getActionDefinitionsBuilder(G_BRCOND) .legalFor({s32}) .clampScalar(0, s32, s32); getActionDefinitionsBuilder(G_BRINDIRECT).legalFor({p0}); getActionDefinitionsBuilder(G_SELECT) .legalFor({{s32, s32}, {s64, s32}, {p0, s32}}) .widenScalarToNextPow2(0) .clampScalar(0, s32, s64) .clampScalar(1, s32, s32) .minScalarEltSameAsIf(all(isVector(0), isVector(1)), 1, 0) .lowerIf(isVector(0)); // Pointer-handling getActionDefinitionsBuilder(G_FRAME_INDEX).legalFor({p0}); if (TM.getCodeModel() == CodeModel::Small) getActionDefinitionsBuilder(G_GLOBAL_VALUE).custom(); else getActionDefinitionsBuilder(G_GLOBAL_VALUE).legalFor({p0}); getActionDefinitionsBuilder(G_PTRAUTH_GLOBAL_VALUE) .legalIf(all(typeIs(0, p0), typeIs(1, p0))); getActionDefinitionsBuilder(G_PTRTOINT) .legalFor({{s64, p0}, {v2s64, v2p0}}) .widenScalarToNextPow2(0, 64) .clampScalar(0, s64, s64); getActionDefinitionsBuilder(G_INTTOPTR) .unsupportedIf([&](const LegalityQuery &Query) { return Query.Types[0].getSizeInBits() != Query.Types[1].getSizeInBits(); }) .legalFor({{p0, s64}, {v2p0, v2s64}}); // Casts for 32 and 64-bit width type are just copies. // Same for 128-bit width type, except they are on the FPR bank. getActionDefinitionsBuilder(G_BITCAST) // Keeping 32-bit instructions legal to prevent regression in some tests .legalForCartesianProduct({s32, v2s16, v4s8}) .legalForCartesianProduct({s64, v8s8, v4s16, v2s32}) .legalForCartesianProduct({s128, v16s8, v8s16, v4s32, v2s64, v2p0}) .lowerIf([=](const LegalityQuery &Query) { return Query.Types[0].isVector() != Query.Types[1].isVector(); }) .moreElementsToNextPow2(0) .clampNumElements(0, v8s8, v16s8) .clampNumElements(0, v4s16, v8s16) .clampNumElements(0, v2s32, v4s32) .lower(); getActionDefinitionsBuilder(G_VASTART).legalFor({p0}); // va_list must be a pointer, but most sized types are pretty easy to handle // as the destination. getActionDefinitionsBuilder(G_VAARG) .customForCartesianProduct({s8, s16, s32, s64, p0}, {p0}) .clampScalar(0, s8, s64) .widenScalarToNextPow2(0, /*Min*/ 8); getActionDefinitionsBuilder(G_ATOMIC_CMPXCHG_WITH_SUCCESS) .lowerIf( all(typeInSet(0, {s8, s16, s32, s64, s128}), typeIs(2, p0))); LegalityPredicate UseOutlineAtomics = [&ST](const LegalityQuery &Query) { return ST.outlineAtomics() && !ST.hasLSE(); }; getActionDefinitionsBuilder(G_ATOMIC_CMPXCHG) .legalIf(all(typeInSet(0, {s32, s64}), typeIs(1, p0), predNot(UseOutlineAtomics))) .customIf(all(typeIs(0, s128), predNot(UseOutlineAtomics))) .customIf([UseOutlineAtomics](const LegalityQuery &Query) { return Query.Types[0].getSizeInBits() == 128 && !UseOutlineAtomics(Query); }) .libcallIf(all(typeInSet(0, {s8, s16, s32, s64, s128}), typeIs(1, p0), UseOutlineAtomics)) .clampScalar(0, s32, s64); getActionDefinitionsBuilder({G_ATOMICRMW_XCHG, G_ATOMICRMW_ADD, G_ATOMICRMW_SUB, G_ATOMICRMW_AND, G_ATOMICRMW_OR, G_ATOMICRMW_XOR}) .legalIf(all(typeInSet(0, {s32, s64}), typeIs(1, p0), predNot(UseOutlineAtomics))) .libcallIf(all(typeInSet(0, {s8, s16, s32, s64}), typeIs(1, p0), UseOutlineAtomics)) .clampScalar(0, s32, s64); // Do not outline these atomics operations, as per comment in // AArch64ISelLowering.cpp's shouldExpandAtomicRMWInIR(). getActionDefinitionsBuilder( {G_ATOMICRMW_MIN, G_ATOMICRMW_MAX, G_ATOMICRMW_UMIN, G_ATOMICRMW_UMAX}) .legalIf(all(typeInSet(0, {s32, s64}), typeIs(1, p0))) .clampScalar(0, s32, s64); getActionDefinitionsBuilder(G_BLOCK_ADDR).legalFor({p0}); // Merge/Unmerge for (unsigned Op : {G_MERGE_VALUES, G_UNMERGE_VALUES}) { unsigned BigTyIdx = Op == G_MERGE_VALUES ? 0 : 1; unsigned LitTyIdx = Op == G_MERGE_VALUES ? 1 : 0; getActionDefinitionsBuilder(Op) .widenScalarToNextPow2(LitTyIdx, 8) .widenScalarToNextPow2(BigTyIdx, 32) .clampScalar(LitTyIdx, s8, s64) .clampScalar(BigTyIdx, s32, s128) .legalIf([=](const LegalityQuery &Q) { switch (Q.Types[BigTyIdx].getSizeInBits()) { case 32: case 64: case 128: break; default: return false; } switch (Q.Types[LitTyIdx].getSizeInBits()) { case 8: case 16: case 32: case 64: return true; default: return false; } }); } getActionDefinitionsBuilder(G_EXTRACT_VECTOR_ELT) .unsupportedIf([=](const LegalityQuery &Query) { const LLT &EltTy = Query.Types[1].getElementType(); return Query.Types[0] != EltTy; }) .minScalar(2, s64) .customIf([=](const LegalityQuery &Query) { const LLT &VecTy = Query.Types[1]; return VecTy == v2s16 || VecTy == v4s16 || VecTy == v8s16 || VecTy == v4s32 || VecTy == v2s64 || VecTy == v2s32 || VecTy == v8s8 || VecTy == v16s8 || VecTy == v2p0; }) .minScalarOrEltIf( [=](const LegalityQuery &Query) { // We want to promote to to if that wouldn't // cause the total vec size to be > 128b. return Query.Types[1].getNumElements() <= 2; }, 0, s64) .minScalarOrEltIf( [=](const LegalityQuery &Query) { return Query.Types[1].getNumElements() <= 4; }, 0, s32) .minScalarOrEltIf( [=](const LegalityQuery &Query) { return Query.Types[1].getNumElements() <= 8; }, 0, s16) .minScalarOrEltIf( [=](const LegalityQuery &Query) { return Query.Types[1].getNumElements() <= 16; }, 0, s8) .minScalarOrElt(0, s8) // Worst case, we need at least s8. .moreElementsToNextPow2(1) .clampMaxNumElements(1, s64, 2) .clampMaxNumElements(1, s32, 4) .clampMaxNumElements(1, s16, 8) .clampMaxNumElements(1, s8, 16) .clampMaxNumElements(1, p0, 2); getActionDefinitionsBuilder(G_INSERT_VECTOR_ELT) .legalIf( typeInSet(0, {v16s8, v8s8, v8s16, v4s16, v4s32, v2s32, v2s64, v2p0})) .moreElementsToNextPow2(0) .widenVectorEltsToVectorMinSize(0, 64) .clampNumElements(0, v8s8, v16s8) .clampNumElements(0, v4s16, v8s16) .clampNumElements(0, v2s32, v4s32) .clampMaxNumElements(0, s64, 2) .clampMaxNumElements(0, p0, 2); getActionDefinitionsBuilder(G_BUILD_VECTOR) .legalFor({{v8s8, s8}, {v16s8, s8}, {v4s16, s16}, {v8s16, s16}, {v2s32, s32}, {v4s32, s32}, {v2p0, p0}, {v2s64, s64}}) .clampNumElements(0, v4s32, v4s32) .clampNumElements(0, v2s64, v2s64) .minScalarOrElt(0, s8) .widenVectorEltsToVectorMinSize(0, 64) .minScalarSameAs(1, 0); getActionDefinitionsBuilder(G_BUILD_VECTOR_TRUNC).lower(); getActionDefinitionsBuilder(G_CTLZ) .legalForCartesianProduct( {s32, s64, v8s8, v16s8, v4s16, v8s16, v2s32, v4s32}) .scalarize(1) .widenScalarToNextPow2(1, /*Min=*/32) .clampScalar(1, s32, s64) .scalarSameSizeAs(0, 1); getActionDefinitionsBuilder(G_CTLZ_ZERO_UNDEF).lower(); // TODO: Custom lowering for v2s32, v4s32, v2s64. getActionDefinitionsBuilder(G_BITREVERSE) .legalFor({s32, s64, v8s8, v16s8}) .widenScalarToNextPow2(0, /*Min = */ 32) .clampScalar(0, s32, s64); getActionDefinitionsBuilder(G_CTTZ_ZERO_UNDEF).lower(); getActionDefinitionsBuilder(G_CTTZ) .lowerIf(isVector(0)) .widenScalarToNextPow2(1, /*Min=*/32) .clampScalar(1, s32, s64) .scalarSameSizeAs(0, 1) .legalIf([=](const LegalityQuery &Query) { return (HasCSSC && typeInSet(0, {s32, s64})(Query)); }) .customIf([=](const LegalityQuery &Query) { return (!HasCSSC && typeInSet(0, {s32, s64})(Query)); }); getActionDefinitionsBuilder(G_SHUFFLE_VECTOR) .legalIf([=](const LegalityQuery &Query) { const LLT &DstTy = Query.Types[0]; const LLT &SrcTy = Query.Types[1]; // For now just support the TBL2 variant which needs the source vectors // to be the same size as the dest. if (DstTy != SrcTy) return false; return llvm::is_contained( {v2s64, v2p0, v2s32, v4s32, v4s16, v16s8, v8s8, v8s16}, DstTy); }) // G_SHUFFLE_VECTOR can have scalar sources (from 1 x s vectors), we // just want those lowered into G_BUILD_VECTOR .lowerIf([=](const LegalityQuery &Query) { return !Query.Types[1].isVector(); }) .moreElementsIf( [](const LegalityQuery &Query) { return Query.Types[0].isVector() && Query.Types[1].isVector() && Query.Types[0].getNumElements() > Query.Types[1].getNumElements(); }, changeTo(1, 0)) .moreElementsToNextPow2(0) .moreElementsIf( [](const LegalityQuery &Query) { return Query.Types[0].isVector() && Query.Types[1].isVector() && Query.Types[0].getNumElements() < Query.Types[1].getNumElements(); }, changeTo(0, 1)) .widenScalarOrEltToNextPow2OrMinSize(0, 8) .clampNumElements(0, v8s8, v16s8) .clampNumElements(0, v4s16, v8s16) .clampNumElements(0, v4s32, v4s32) .clampNumElements(0, v2s64, v2s64); getActionDefinitionsBuilder(G_CONCAT_VECTORS) .legalFor({{v4s32, v2s32}, {v8s16, v4s16}, {v16s8, v8s8}}) .bitcastIf( [=](const LegalityQuery &Query) { return Query.Types[0].getSizeInBits() <= 128 && Query.Types[1].getSizeInBits() <= 64; }, [=](const LegalityQuery &Query) { const LLT DstTy = Query.Types[0]; const LLT SrcTy = Query.Types[1]; return std::pair( 0, DstTy.changeElementSize(SrcTy.getSizeInBits()) .changeElementCount( DstTy.getElementCount().divideCoefficientBy( SrcTy.getNumElements()))); }); getActionDefinitionsBuilder(G_JUMP_TABLE).legalFor({p0}); getActionDefinitionsBuilder(G_BRJT).legalFor({{p0, s64}}); getActionDefinitionsBuilder(G_DYN_STACKALLOC).custom(); getActionDefinitionsBuilder({G_STACKSAVE, G_STACKRESTORE}).lower(); if (ST.hasMOPS()) { // G_BZERO is not supported. Currently it is only emitted by // PreLegalizerCombiner for G_MEMSET with zero constant. getActionDefinitionsBuilder(G_BZERO).unsupported(); getActionDefinitionsBuilder(G_MEMSET) .legalForCartesianProduct({p0}, {s64}, {s64}) .customForCartesianProduct({p0}, {s8}, {s64}) .immIdx(0); // Inform verifier imm idx 0 is handled. getActionDefinitionsBuilder({G_MEMCPY, G_MEMMOVE}) .legalForCartesianProduct({p0}, {p0}, {s64}) .immIdx(0); // Inform verifier imm idx 0 is handled. // G_MEMCPY_INLINE does not have a tailcall immediate getActionDefinitionsBuilder(G_MEMCPY_INLINE) .legalForCartesianProduct({p0}, {p0}, {s64}); } else { getActionDefinitionsBuilder({G_BZERO, G_MEMCPY, G_MEMMOVE, G_MEMSET}) .libcall(); } // FIXME: Legal vector types are only legal with NEON. auto &ABSActions = getActionDefinitionsBuilder(G_ABS); if (HasCSSC) ABSActions .legalFor({s32, s64}); ABSActions.legalFor(PackedVectorAllTypeList) .customIf([=](const LegalityQuery &Q) { // TODO: Fix suboptimal codegen for 128+ bit types. LLT SrcTy = Q.Types[0]; return SrcTy.isScalar() && SrcTy.getSizeInBits() < 128; }) .widenScalarIf( [=](const LegalityQuery &Query) { return Query.Types[0] == v4s8; }, [=](const LegalityQuery &Query) { return std::make_pair(0, v4s16); }) .widenScalarIf( [=](const LegalityQuery &Query) { return Query.Types[0] == v2s16; }, [=](const LegalityQuery &Query) { return std::make_pair(0, v2s32); }) .clampNumElements(0, v8s8, v16s8) .clampNumElements(0, v4s16, v8s16) .clampNumElements(0, v2s32, v4s32) .clampNumElements(0, v2s64, v2s64) .moreElementsToNextPow2(0) .lower(); // For fadd reductions we have pairwise operations available. We treat the // usual legal types as legal and handle the lowering to pairwise instructions // later. getActionDefinitionsBuilder(G_VECREDUCE_FADD) .legalFor({{s32, v2s32}, {s32, v4s32}, {s64, v2s64}}) .legalIf([=](const LegalityQuery &Query) { const auto &Ty = Query.Types[1]; return (Ty == v4s16 || Ty == v8s16) && HasFP16; }) .minScalarOrElt(0, MinFPScalar) .clampMaxNumElements(1, s64, 2) .clampMaxNumElements(1, s32, 4) .clampMaxNumElements(1, s16, 8) .lower(); // For fmul reductions we need to split up into individual operations. We // clamp to 128 bit vectors then to 64bit vectors to produce a cascade of // smaller types, followed by scalarizing what remains. getActionDefinitionsBuilder(G_VECREDUCE_FMUL) .minScalarOrElt(0, MinFPScalar) .clampMaxNumElements(1, s64, 2) .clampMaxNumElements(1, s32, 4) .clampMaxNumElements(1, s16, 8) .clampMaxNumElements(1, s32, 2) .clampMaxNumElements(1, s16, 4) .scalarize(1) .lower(); getActionDefinitionsBuilder({G_VECREDUCE_SEQ_FADD, G_VECREDUCE_SEQ_FMUL}) .scalarize(2) .lower(); getActionDefinitionsBuilder(G_VECREDUCE_ADD) .legalFor({{s8, v16s8}, {s8, v8s8}, {s16, v8s16}, {s16, v4s16}, {s32, v4s32}, {s32, v2s32}, {s64, v2s64}}) .clampMaxNumElements(1, s64, 2) .clampMaxNumElements(1, s32, 4) .clampMaxNumElements(1, s16, 8) .clampMaxNumElements(1, s8, 16) .lower(); getActionDefinitionsBuilder({G_VECREDUCE_FMIN, G_VECREDUCE_FMAX, G_VECREDUCE_FMINIMUM, G_VECREDUCE_FMAXIMUM}) .legalFor({{s32, v4s32}, {s32, v2s32}, {s64, v2s64}}) .legalIf([=](const LegalityQuery &Query) { const auto &Ty = Query.Types[1]; return Query.Types[0] == s16 && (Ty == v8s16 || Ty == v4s16) && HasFP16; }) .minScalarOrElt(0, MinFPScalar) .clampMaxNumElements(1, s64, 2) .clampMaxNumElements(1, s32, 4) .clampMaxNumElements(1, s16, 8) .lower(); getActionDefinitionsBuilder(G_VECREDUCE_MUL) .clampMaxNumElements(1, s32, 2) .clampMaxNumElements(1, s16, 4) .clampMaxNumElements(1, s8, 8) .scalarize(1) .lower(); getActionDefinitionsBuilder( {G_VECREDUCE_SMIN, G_VECREDUCE_SMAX, G_VECREDUCE_UMIN, G_VECREDUCE_UMAX}) .legalFor({{s8, v8s8}, {s8, v16s8}, {s16, v4s16}, {s16, v8s16}, {s32, v2s32}, {s32, v4s32}}) .moreElementsIf( [=](const LegalityQuery &Query) { return Query.Types[1].isVector() && Query.Types[1].getElementType() != s8 && Query.Types[1].getNumElements() & 1; }, LegalizeMutations::moreElementsToNextPow2(1)) .clampMaxNumElements(1, s64, 2) .clampMaxNumElements(1, s32, 4) .clampMaxNumElements(1, s16, 8) .clampMaxNumElements(1, s8, 16) .scalarize(1) .lower(); getActionDefinitionsBuilder( {G_VECREDUCE_OR, G_VECREDUCE_AND, G_VECREDUCE_XOR}) // Try to break down into smaller vectors as long as they're at least 64 // bits. This lets us use vector operations for some parts of the // reduction. .fewerElementsIf( [=](const LegalityQuery &Q) { LLT SrcTy = Q.Types[1]; if (SrcTy.isScalar()) return false; if (!isPowerOf2_32(SrcTy.getNumElements())) return false; // We can usually perform 64b vector operations. return SrcTy.getSizeInBits() > 64; }, [=](const LegalityQuery &Q) { LLT SrcTy = Q.Types[1]; return std::make_pair(1, SrcTy.divide(2)); }) .scalarize(1) .lower(); // TODO: Update this to correct handling when adding AArch64/SVE support. getActionDefinitionsBuilder(G_VECTOR_COMPRESS).lower(); getActionDefinitionsBuilder({G_FSHL, G_FSHR}) .customFor({{s32, s32}, {s32, s64}, {s64, s64}}) .lower(); getActionDefinitionsBuilder(G_ROTR) .legalFor({{s32, s64}, {s64, s64}}) .customIf([=](const LegalityQuery &Q) { return Q.Types[0].isScalar() && Q.Types[1].getScalarSizeInBits() < 64; }) .lower(); getActionDefinitionsBuilder(G_ROTL).lower(); getActionDefinitionsBuilder({G_SBFX, G_UBFX}) .customFor({{s32, s32}, {s64, s64}}); auto always = [=](const LegalityQuery &Q) { return true; }; auto &CTPOPActions = getActionDefinitionsBuilder(G_CTPOP); if (HasCSSC) CTPOPActions .legalFor({{s32, s32}, {s64, s64}, {v8s8, v8s8}, {v16s8, v16s8}}) .customFor({{s128, s128}, {v2s64, v2s64}, {v2s32, v2s32}, {v4s32, v4s32}, {v4s16, v4s16}, {v8s16, v8s16}}); else CTPOPActions .legalFor({{v8s8, v8s8}, {v16s8, v16s8}}) .customFor({{s32, s32}, {s64, s64}, {s128, s128}, {v2s64, v2s64}, {v2s32, v2s32}, {v4s32, v4s32}, {v4s16, v4s16}, {v8s16, v8s16}}); CTPOPActions .clampScalar(0, s32, s128) .widenScalarToNextPow2(0) .minScalarEltSameAsIf(always, 1, 0) .maxScalarEltSameAsIf(always, 1, 0); getActionDefinitionsBuilder({G_UADDSAT, G_SADDSAT, G_USUBSAT, G_SSUBSAT}) .legalFor({v2s64, v2s32, v4s32, v4s16, v8s16, v8s8, v16s8}) .clampNumElements(0, v8s8, v16s8) .clampNumElements(0, v4s16, v8s16) .clampNumElements(0, v2s32, v4s32) .clampMaxNumElements(0, s64, 2) .moreElementsToNextPow2(0) .lower(); // TODO: Libcall support for s128. // TODO: s16 should be legal with full FP16 support. getActionDefinitionsBuilder({G_LROUND, G_LLROUND}) .legalFor({{s64, s32}, {s64, s64}}); // TODO: Custom legalization for mismatched types. getActionDefinitionsBuilder(G_FCOPYSIGN) .moreElementsIf( [](const LegalityQuery &Query) { return Query.Types[0].isScalar(); }, [=](const LegalityQuery &Query) { const LLT Ty = Query.Types[0]; return std::pair(0, LLT::fixed_vector(Ty == s16 ? 4 : 2, Ty)); }) .lower(); getActionDefinitionsBuilder(G_FMAD).lower(); // Access to floating-point environment. getActionDefinitionsBuilder({G_GET_FPENV, G_SET_FPENV, G_RESET_FPENV, G_GET_FPMODE, G_SET_FPMODE, G_RESET_FPMODE}) .libcall(); getActionDefinitionsBuilder(G_IS_FPCLASS).lower(); getActionDefinitionsBuilder(G_PREFETCH).custom(); getActionDefinitionsBuilder({G_SCMP, G_UCMP}).lower(); getLegacyLegalizerInfo().computeTables(); verify(*ST.getInstrInfo()); } bool AArch64LegalizerInfo::legalizeCustom( LegalizerHelper &Helper, MachineInstr &MI, LostDebugLocObserver &LocObserver) const { MachineIRBuilder &MIRBuilder = Helper.MIRBuilder; MachineRegisterInfo &MRI = *MIRBuilder.getMRI(); GISelChangeObserver &Observer = Helper.Observer; switch (MI.getOpcode()) { default: // No idea what to do. return false; case TargetOpcode::G_VAARG: return legalizeVaArg(MI, MRI, MIRBuilder); case TargetOpcode::G_LOAD: case TargetOpcode::G_STORE: return legalizeLoadStore(MI, MRI, MIRBuilder, Observer); case TargetOpcode::G_SHL: case TargetOpcode::G_ASHR: case TargetOpcode::G_LSHR: return legalizeShlAshrLshr(MI, MRI, MIRBuilder, Observer); case TargetOpcode::G_GLOBAL_VALUE: return legalizeSmallCMGlobalValue(MI, MRI, MIRBuilder, Observer); case TargetOpcode::G_SBFX: case TargetOpcode::G_UBFX: return legalizeBitfieldExtract(MI, MRI, Helper); case TargetOpcode::G_FSHL: case TargetOpcode::G_FSHR: return legalizeFunnelShift(MI, MRI, MIRBuilder, Observer, Helper); case TargetOpcode::G_ROTR: return legalizeRotate(MI, MRI, Helper); case TargetOpcode::G_CTPOP: return legalizeCTPOP(MI, MRI, Helper); case TargetOpcode::G_ATOMIC_CMPXCHG: return legalizeAtomicCmpxchg128(MI, MRI, Helper); case TargetOpcode::G_CTTZ: return legalizeCTTZ(MI, Helper); case TargetOpcode::G_BZERO: case TargetOpcode::G_MEMCPY: case TargetOpcode::G_MEMMOVE: case TargetOpcode::G_MEMSET: return legalizeMemOps(MI, Helper); case TargetOpcode::G_EXTRACT_VECTOR_ELT: return legalizeExtractVectorElt(MI, MRI, Helper); case TargetOpcode::G_DYN_STACKALLOC: return legalizeDynStackAlloc(MI, Helper); case TargetOpcode::G_PREFETCH: return legalizePrefetch(MI, Helper); case TargetOpcode::G_ABS: return Helper.lowerAbsToCNeg(MI); case TargetOpcode::G_ICMP: return legalizeICMP(MI, MRI, MIRBuilder); } llvm_unreachable("expected switch to return"); } bool AArch64LegalizerInfo::legalizeFunnelShift(MachineInstr &MI, MachineRegisterInfo &MRI, MachineIRBuilder &MIRBuilder, GISelChangeObserver &Observer, LegalizerHelper &Helper) const { assert(MI.getOpcode() == TargetOpcode::G_FSHL || MI.getOpcode() == TargetOpcode::G_FSHR); // Keep as G_FSHR if shift amount is a G_CONSTANT, else use generic // lowering Register ShiftNo = MI.getOperand(3).getReg(); LLT ShiftTy = MRI.getType(ShiftNo); auto VRegAndVal = getIConstantVRegValWithLookThrough(ShiftNo, MRI); // Adjust shift amount according to Opcode (FSHL/FSHR) // Convert FSHL to FSHR LLT OperationTy = MRI.getType(MI.getOperand(0).getReg()); APInt BitWidth(ShiftTy.getSizeInBits(), OperationTy.getSizeInBits(), false); // Lower non-constant shifts and leave zero shifts to the optimizer. if (!VRegAndVal || VRegAndVal->Value.urem(BitWidth) == 0) return (Helper.lowerFunnelShiftAsShifts(MI) == LegalizerHelper::LegalizeResult::Legalized); APInt Amount = VRegAndVal->Value.urem(BitWidth); Amount = MI.getOpcode() == TargetOpcode::G_FSHL ? BitWidth - Amount : Amount; // If the instruction is G_FSHR, has a 64-bit G_CONSTANT for shift amount // in the range of 0 <-> BitWidth, it is legal if (ShiftTy.getSizeInBits() == 64 && MI.getOpcode() == TargetOpcode::G_FSHR && VRegAndVal->Value.ult(BitWidth)) return true; // Cast the ShiftNumber to a 64-bit type auto Cast64 = MIRBuilder.buildConstant(LLT::scalar(64), Amount.zext(64)); if (MI.getOpcode() == TargetOpcode::G_FSHR) { Observer.changingInstr(MI); MI.getOperand(3).setReg(Cast64.getReg(0)); Observer.changedInstr(MI); } // If Opcode is FSHL, remove the FSHL instruction and create a FSHR // instruction else if (MI.getOpcode() == TargetOpcode::G_FSHL) { MIRBuilder.buildInstr(TargetOpcode::G_FSHR, {MI.getOperand(0).getReg()}, {MI.getOperand(1).getReg(), MI.getOperand(2).getReg(), Cast64.getReg(0)}); MI.eraseFromParent(); } return true; } bool AArch64LegalizerInfo::legalizeICMP(MachineInstr &MI, MachineRegisterInfo &MRI, MachineIRBuilder &MIRBuilder) const { Register DstReg = MI.getOperand(0).getReg(); Register SrcReg1 = MI.getOperand(2).getReg(); Register SrcReg2 = MI.getOperand(3).getReg(); LLT DstTy = MRI.getType(DstReg); LLT SrcTy = MRI.getType(SrcReg1); // Check the vector types are legal if (DstTy.getScalarSizeInBits() != SrcTy.getScalarSizeInBits() || DstTy.getNumElements() != SrcTy.getNumElements() || (DstTy.getSizeInBits() != 64 && DstTy.getSizeInBits() != 128)) return false; // Lowers G_ICMP NE => G_ICMP EQ to allow better pattern matching for // following passes CmpInst::Predicate Pred = (CmpInst::Predicate)MI.getOperand(1).getPredicate(); if (Pred != CmpInst::ICMP_NE) return true; Register CmpReg = MIRBuilder .buildICmp(CmpInst::ICMP_EQ, MRI.getType(DstReg), SrcReg1, SrcReg2) .getReg(0); MIRBuilder.buildNot(DstReg, CmpReg); MI.eraseFromParent(); return true; } bool AArch64LegalizerInfo::legalizeRotate(MachineInstr &MI, MachineRegisterInfo &MRI, LegalizerHelper &Helper) const { // To allow for imported patterns to match, we ensure that the rotate amount // is 64b with an extension. Register AmtReg = MI.getOperand(2).getReg(); LLT AmtTy = MRI.getType(AmtReg); (void)AmtTy; assert(AmtTy.isScalar() && "Expected a scalar rotate"); assert(AmtTy.getSizeInBits() < 64 && "Expected this rotate to be legal"); auto NewAmt = Helper.MIRBuilder.buildZExt(LLT::scalar(64), AmtReg); Helper.Observer.changingInstr(MI); MI.getOperand(2).setReg(NewAmt.getReg(0)); Helper.Observer.changedInstr(MI); return true; } bool AArch64LegalizerInfo::legalizeSmallCMGlobalValue( MachineInstr &MI, MachineRegisterInfo &MRI, MachineIRBuilder &MIRBuilder, GISelChangeObserver &Observer) const { assert(MI.getOpcode() == TargetOpcode::G_GLOBAL_VALUE); // We do this custom legalization to convert G_GLOBAL_VALUE into target ADRP + // G_ADD_LOW instructions. // By splitting this here, we can optimize accesses in the small code model by // folding in the G_ADD_LOW into the load/store offset. auto &GlobalOp = MI.getOperand(1); // Don't modify an intrinsic call. if (GlobalOp.isSymbol()) return true; const auto* GV = GlobalOp.getGlobal(); if (GV->isThreadLocal()) return true; // Don't want to modify TLS vars. auto &TM = ST->getTargetLowering()->getTargetMachine(); unsigned OpFlags = ST->ClassifyGlobalReference(GV, TM); if (OpFlags & AArch64II::MO_GOT) return true; auto Offset = GlobalOp.getOffset(); Register DstReg = MI.getOperand(0).getReg(); auto ADRP = MIRBuilder.buildInstr(AArch64::ADRP, {LLT::pointer(0, 64)}, {}) .addGlobalAddress(GV, Offset, OpFlags | AArch64II::MO_PAGE); // Set the regclass on the dest reg too. MRI.setRegClass(ADRP.getReg(0), &AArch64::GPR64RegClass); // MO_TAGGED on the page indicates a tagged address. Set the tag now. We do so // by creating a MOVK that sets bits 48-63 of the register to (global address // + 0x100000000 - PC) >> 48. The additional 0x100000000 offset here is to // prevent an incorrect tag being generated during relocation when the // global appears before the code section. Without the offset, a global at // `0x0f00'0000'0000'1000` (i.e. at `0x1000` with tag `0xf`) that's referenced // by code at `0x2000` would result in `0x0f00'0000'0000'1000 - 0x2000 = // 0x0eff'ffff'ffff'f000`, meaning the tag would be incorrectly set to `0xe` // instead of `0xf`. // This assumes that we're in the small code model so we can assume a binary // size of <= 4GB, which makes the untagged PC relative offset positive. The // binary must also be loaded into address range [0, 2^48). Both of these // properties need to be ensured at runtime when using tagged addresses. if (OpFlags & AArch64II::MO_TAGGED) { assert(!Offset && "Should not have folded in an offset for a tagged global!"); ADRP = MIRBuilder.buildInstr(AArch64::MOVKXi, {LLT::pointer(0, 64)}, {ADRP}) .addGlobalAddress(GV, 0x100000000, AArch64II::MO_PREL | AArch64II::MO_G3) .addImm(48); MRI.setRegClass(ADRP.getReg(0), &AArch64::GPR64RegClass); } MIRBuilder.buildInstr(AArch64::G_ADD_LOW, {DstReg}, {ADRP}) .addGlobalAddress(GV, Offset, OpFlags | AArch64II::MO_PAGEOFF | AArch64II::MO_NC); MI.eraseFromParent(); return true; } bool AArch64LegalizerInfo::legalizeIntrinsic(LegalizerHelper &Helper, MachineInstr &MI) const { Intrinsic::ID IntrinsicID = cast(MI).getIntrinsicID(); switch (IntrinsicID) { case Intrinsic::vacopy: { unsigned PtrSize = ST->isTargetILP32() ? 4 : 8; unsigned VaListSize = (ST->isTargetDarwin() || ST->isTargetWindows()) ? PtrSize : ST->isTargetILP32() ? 20 : 32; MachineFunction &MF = *MI.getMF(); auto Val = MF.getRegInfo().createGenericVirtualRegister( LLT::scalar(VaListSize * 8)); MachineIRBuilder MIB(MI); MIB.buildLoad(Val, MI.getOperand(2), *MF.getMachineMemOperand(MachinePointerInfo(), MachineMemOperand::MOLoad, VaListSize, Align(PtrSize))); MIB.buildStore(Val, MI.getOperand(1), *MF.getMachineMemOperand(MachinePointerInfo(), MachineMemOperand::MOStore, VaListSize, Align(PtrSize))); MI.eraseFromParent(); return true; } case Intrinsic::get_dynamic_area_offset: { MachineIRBuilder &MIB = Helper.MIRBuilder; MIB.buildConstant(MI.getOperand(0).getReg(), 0); MI.eraseFromParent(); return true; } case Intrinsic::aarch64_mops_memset_tag: { assert(MI.getOpcode() == TargetOpcode::G_INTRINSIC_W_SIDE_EFFECTS); // Anyext the value being set to 64 bit (only the bottom 8 bits are read by // the instruction). MachineIRBuilder MIB(MI); auto &Value = MI.getOperand(3); Register ExtValueReg = MIB.buildAnyExt(LLT::scalar(64), Value).getReg(0); Value.setReg(ExtValueReg); return true; } case Intrinsic::aarch64_prefetch: { MachineIRBuilder MIB(MI); auto &AddrVal = MI.getOperand(1); int64_t IsWrite = MI.getOperand(2).getImm(); int64_t Target = MI.getOperand(3).getImm(); int64_t IsStream = MI.getOperand(4).getImm(); int64_t IsData = MI.getOperand(5).getImm(); unsigned PrfOp = (IsWrite << 4) | // Load/Store bit (!IsData << 3) | // IsDataCache bit (Target << 1) | // Cache level bits (unsigned)IsStream; // Stream bit MIB.buildInstr(AArch64::G_AARCH64_PREFETCH).addImm(PrfOp).add(AddrVal); MI.eraseFromParent(); return true; } case Intrinsic::aarch64_neon_uaddv: case Intrinsic::aarch64_neon_saddv: case Intrinsic::aarch64_neon_umaxv: case Intrinsic::aarch64_neon_smaxv: case Intrinsic::aarch64_neon_uminv: case Intrinsic::aarch64_neon_sminv: { MachineIRBuilder MIB(MI); MachineRegisterInfo &MRI = *MIB.getMRI(); bool IsSigned = IntrinsicID == Intrinsic::aarch64_neon_saddv || IntrinsicID == Intrinsic::aarch64_neon_smaxv || IntrinsicID == Intrinsic::aarch64_neon_sminv; auto OldDst = MI.getOperand(0).getReg(); auto OldDstTy = MRI.getType(OldDst); LLT NewDstTy = MRI.getType(MI.getOperand(2).getReg()).getElementType(); if (OldDstTy == NewDstTy) return true; auto NewDst = MRI.createGenericVirtualRegister(NewDstTy); Helper.Observer.changingInstr(MI); MI.getOperand(0).setReg(NewDst); Helper.Observer.changedInstr(MI); MIB.setInsertPt(MIB.getMBB(), ++MIB.getInsertPt()); MIB.buildExtOrTrunc(IsSigned ? TargetOpcode::G_SEXT : TargetOpcode::G_ZEXT, OldDst, NewDst); return true; } case Intrinsic::aarch64_neon_uaddlp: case Intrinsic::aarch64_neon_saddlp: { MachineIRBuilder MIB(MI); unsigned Opc = IntrinsicID == Intrinsic::aarch64_neon_uaddlp ? AArch64::G_UADDLP : AArch64::G_SADDLP; MIB.buildInstr(Opc, {MI.getOperand(0)}, {MI.getOperand(2)}); MI.eraseFromParent(); return true; } case Intrinsic::aarch64_neon_uaddlv: case Intrinsic::aarch64_neon_saddlv: { MachineIRBuilder MIB(MI); MachineRegisterInfo &MRI = *MIB.getMRI(); unsigned Opc = IntrinsicID == Intrinsic::aarch64_neon_uaddlv ? AArch64::G_UADDLV : AArch64::G_SADDLV; Register DstReg = MI.getOperand(0).getReg(); Register SrcReg = MI.getOperand(2).getReg(); LLT DstTy = MRI.getType(DstReg); LLT MidTy, ExtTy; if (DstTy.isScalar() && DstTy.getScalarSizeInBits() <= 32) { MidTy = LLT::fixed_vector(4, 32); ExtTy = LLT::scalar(32); } else { MidTy = LLT::fixed_vector(2, 64); ExtTy = LLT::scalar(64); } Register MidReg = MIB.buildInstr(Opc, {MidTy}, {SrcReg})->getOperand(0).getReg(); Register ZeroReg = MIB.buildConstant(LLT::scalar(64), 0)->getOperand(0).getReg(); Register ExtReg = MIB.buildInstr(AArch64::G_EXTRACT_VECTOR_ELT, {ExtTy}, {MidReg, ZeroReg}) .getReg(0); if (DstTy.getScalarSizeInBits() < 32) MIB.buildTrunc(DstReg, ExtReg); else MIB.buildCopy(DstReg, ExtReg); MI.eraseFromParent(); return true; } case Intrinsic::aarch64_neon_smax: case Intrinsic::aarch64_neon_smin: case Intrinsic::aarch64_neon_umax: case Intrinsic::aarch64_neon_umin: case Intrinsic::aarch64_neon_fmax: case Intrinsic::aarch64_neon_fmin: case Intrinsic::aarch64_neon_fmaxnm: case Intrinsic::aarch64_neon_fminnm: { MachineIRBuilder MIB(MI); if (IntrinsicID == Intrinsic::aarch64_neon_smax) MIB.buildSMax(MI.getOperand(0), MI.getOperand(2), MI.getOperand(3)); else if (IntrinsicID == Intrinsic::aarch64_neon_smin) MIB.buildSMin(MI.getOperand(0), MI.getOperand(2), MI.getOperand(3)); else if (IntrinsicID == Intrinsic::aarch64_neon_umax) MIB.buildUMax(MI.getOperand(0), MI.getOperand(2), MI.getOperand(3)); else if (IntrinsicID == Intrinsic::aarch64_neon_umin) MIB.buildUMin(MI.getOperand(0), MI.getOperand(2), MI.getOperand(3)); else if (IntrinsicID == Intrinsic::aarch64_neon_fmax) MIB.buildInstr(TargetOpcode::G_FMAXIMUM, {MI.getOperand(0)}, {MI.getOperand(2), MI.getOperand(3)}); else if (IntrinsicID == Intrinsic::aarch64_neon_fmin) MIB.buildInstr(TargetOpcode::G_FMINIMUM, {MI.getOperand(0)}, {MI.getOperand(2), MI.getOperand(3)}); else if (IntrinsicID == Intrinsic::aarch64_neon_fmaxnm) MIB.buildInstr(TargetOpcode::G_FMAXNUM, {MI.getOperand(0)}, {MI.getOperand(2), MI.getOperand(3)}); else if (IntrinsicID == Intrinsic::aarch64_neon_fminnm) MIB.buildInstr(TargetOpcode::G_FMINNUM, {MI.getOperand(0)}, {MI.getOperand(2), MI.getOperand(3)}); MI.eraseFromParent(); return true; } case Intrinsic::vector_reverse: // TODO: Add support for vector_reverse return false; } return true; } bool AArch64LegalizerInfo::legalizeShlAshrLshr( MachineInstr &MI, MachineRegisterInfo &MRI, MachineIRBuilder &MIRBuilder, GISelChangeObserver &Observer) const { assert(MI.getOpcode() == TargetOpcode::G_ASHR || MI.getOpcode() == TargetOpcode::G_LSHR || MI.getOpcode() == TargetOpcode::G_SHL); // If the shift amount is a G_CONSTANT, promote it to a 64 bit type so the // imported patterns can select it later. Either way, it will be legal. Register AmtReg = MI.getOperand(2).getReg(); auto VRegAndVal = getIConstantVRegValWithLookThrough(AmtReg, MRI); if (!VRegAndVal) return true; // Check the shift amount is in range for an immediate form. int64_t Amount = VRegAndVal->Value.getSExtValue(); if (Amount > 31) return true; // This will have to remain a register variant. auto ExtCst = MIRBuilder.buildConstant(LLT::scalar(64), Amount); Observer.changingInstr(MI); MI.getOperand(2).setReg(ExtCst.getReg(0)); Observer.changedInstr(MI); return true; } static void matchLDPSTPAddrMode(Register Root, Register &Base, int &Offset, MachineRegisterInfo &MRI) { Base = Root; Offset = 0; Register NewBase; int64_t NewOffset; if (mi_match(Root, MRI, m_GPtrAdd(m_Reg(NewBase), m_ICst(NewOffset))) && isShiftedInt<7, 3>(NewOffset)) { Base = NewBase; Offset = NewOffset; } } // FIXME: This should be removed and replaced with the generic bitcast legalize // action. bool AArch64LegalizerInfo::legalizeLoadStore( MachineInstr &MI, MachineRegisterInfo &MRI, MachineIRBuilder &MIRBuilder, GISelChangeObserver &Observer) const { assert(MI.getOpcode() == TargetOpcode::G_STORE || MI.getOpcode() == TargetOpcode::G_LOAD); // Here we just try to handle vector loads/stores where our value type might // have pointer elements, which the SelectionDAG importer can't handle. To // allow the existing patterns for s64 to fire for p0, we just try to bitcast // the value to use s64 types. // Custom legalization requires the instruction, if not deleted, must be fully // legalized. In order to allow further legalization of the inst, we create // a new instruction and erase the existing one. Register ValReg = MI.getOperand(0).getReg(); const LLT ValTy = MRI.getType(ValReg); if (ValTy == LLT::scalar(128)) { AtomicOrdering Ordering = (*MI.memoperands_begin())->getSuccessOrdering(); bool IsLoad = MI.getOpcode() == TargetOpcode::G_LOAD; bool IsLoadAcquire = IsLoad && Ordering == AtomicOrdering::Acquire; bool IsStoreRelease = !IsLoad && Ordering == AtomicOrdering::Release; bool IsRcpC3 = ST->hasLSE2() && ST->hasRCPC3() && (IsLoadAcquire || IsStoreRelease); LLT s64 = LLT::scalar(64); unsigned Opcode; if (IsRcpC3) { Opcode = IsLoad ? AArch64::LDIAPPX : AArch64::STILPX; } else { // For LSE2, loads/stores should have been converted to monotonic and had // a fence inserted after them. assert(Ordering == AtomicOrdering::Monotonic || Ordering == AtomicOrdering::Unordered); assert(ST->hasLSE2() && "ldp/stp not single copy atomic without +lse2"); Opcode = IsLoad ? AArch64::LDPXi : AArch64::STPXi; } MachineInstrBuilder NewI; if (IsLoad) { NewI = MIRBuilder.buildInstr(Opcode, {s64, s64}, {}); MIRBuilder.buildMergeLikeInstr( ValReg, {NewI->getOperand(0), NewI->getOperand(1)}); } else { auto Split = MIRBuilder.buildUnmerge(s64, MI.getOperand(0)); NewI = MIRBuilder.buildInstr( Opcode, {}, {Split->getOperand(0), Split->getOperand(1)}); } if (IsRcpC3) { NewI.addUse(MI.getOperand(1).getReg()); } else { Register Base; int Offset; matchLDPSTPAddrMode(MI.getOperand(1).getReg(), Base, Offset, MRI); NewI.addUse(Base); NewI.addImm(Offset / 8); } NewI.cloneMemRefs(MI); constrainSelectedInstRegOperands(*NewI, *ST->getInstrInfo(), *MRI.getTargetRegisterInfo(), *ST->getRegBankInfo()); MI.eraseFromParent(); return true; } if (!ValTy.isPointerVector() || ValTy.getElementType().getAddressSpace() != 0) { LLVM_DEBUG(dbgs() << "Tried to do custom legalization on wrong load/store"); return false; } unsigned PtrSize = ValTy.getElementType().getSizeInBits(); const LLT NewTy = LLT::vector(ValTy.getElementCount(), PtrSize); auto &MMO = **MI.memoperands_begin(); MMO.setType(NewTy); if (MI.getOpcode() == TargetOpcode::G_STORE) { auto Bitcast = MIRBuilder.buildBitcast(NewTy, ValReg); MIRBuilder.buildStore(Bitcast.getReg(0), MI.getOperand(1), MMO); } else { auto NewLoad = MIRBuilder.buildLoad(NewTy, MI.getOperand(1), MMO); MIRBuilder.buildBitcast(ValReg, NewLoad); } MI.eraseFromParent(); return true; } bool AArch64LegalizerInfo::legalizeVaArg(MachineInstr &MI, MachineRegisterInfo &MRI, MachineIRBuilder &MIRBuilder) const { MachineFunction &MF = MIRBuilder.getMF(); Align Alignment(MI.getOperand(2).getImm()); Register Dst = MI.getOperand(0).getReg(); Register ListPtr = MI.getOperand(1).getReg(); LLT PtrTy = MRI.getType(ListPtr); LLT IntPtrTy = LLT::scalar(PtrTy.getSizeInBits()); const unsigned PtrSize = PtrTy.getSizeInBits() / 8; const Align PtrAlign = Align(PtrSize); auto List = MIRBuilder.buildLoad( PtrTy, ListPtr, *MF.getMachineMemOperand(MachinePointerInfo(), MachineMemOperand::MOLoad, PtrTy, PtrAlign)); MachineInstrBuilder DstPtr; if (Alignment > PtrAlign) { // Realign the list to the actual required alignment. auto AlignMinus1 = MIRBuilder.buildConstant(IntPtrTy, Alignment.value() - 1); auto ListTmp = MIRBuilder.buildPtrAdd(PtrTy, List, AlignMinus1.getReg(0)); DstPtr = MIRBuilder.buildMaskLowPtrBits(PtrTy, ListTmp, Log2(Alignment)); } else DstPtr = List; LLT ValTy = MRI.getType(Dst); uint64_t ValSize = ValTy.getSizeInBits() / 8; MIRBuilder.buildLoad( Dst, DstPtr, *MF.getMachineMemOperand(MachinePointerInfo(), MachineMemOperand::MOLoad, ValTy, std::max(Alignment, PtrAlign))); auto Size = MIRBuilder.buildConstant(IntPtrTy, alignTo(ValSize, PtrAlign)); auto NewList = MIRBuilder.buildPtrAdd(PtrTy, DstPtr, Size.getReg(0)); MIRBuilder.buildStore(NewList, ListPtr, *MF.getMachineMemOperand(MachinePointerInfo(), MachineMemOperand::MOStore, PtrTy, PtrAlign)); MI.eraseFromParent(); return true; } bool AArch64LegalizerInfo::legalizeBitfieldExtract( MachineInstr &MI, MachineRegisterInfo &MRI, LegalizerHelper &Helper) const { // Only legal if we can select immediate forms. // TODO: Lower this otherwise. return getIConstantVRegValWithLookThrough(MI.getOperand(2).getReg(), MRI) && getIConstantVRegValWithLookThrough(MI.getOperand(3).getReg(), MRI); } bool AArch64LegalizerInfo::legalizeCTPOP(MachineInstr &MI, MachineRegisterInfo &MRI, LegalizerHelper &Helper) const { // When there is no integer popcount instruction (FEAT_CSSC isn't available), // it can be more efficiently lowered to the following sequence that uses // AdvSIMD registers/instructions as long as the copies to/from the AdvSIMD // registers are cheap. // FMOV D0, X0 // copy 64-bit int to vector, high bits zero'd // CNT V0.8B, V0.8B // 8xbyte pop-counts // ADDV B0, V0.8B // sum 8xbyte pop-counts // UMOV X0, V0.B[0] // copy byte result back to integer reg // // For 128 bit vector popcounts, we lower to the following sequence: // cnt.16b v0, v0 // v8s16, v4s32, v2s64 // uaddlp.8h v0, v0 // v8s16, v4s32, v2s64 // uaddlp.4s v0, v0 // v4s32, v2s64 // uaddlp.2d v0, v0 // v2s64 // // For 64 bit vector popcounts, we lower to the following sequence: // cnt.8b v0, v0 // v4s16, v2s32 // uaddlp.4h v0, v0 // v4s16, v2s32 // uaddlp.2s v0, v0 // v2s32 MachineIRBuilder &MIRBuilder = Helper.MIRBuilder; Register Dst = MI.getOperand(0).getReg(); Register Val = MI.getOperand(1).getReg(); LLT Ty = MRI.getType(Val); unsigned Size = Ty.getSizeInBits(); assert(Ty == MRI.getType(Dst) && "Expected src and dst to have the same type!"); if (ST->hasCSSC() && Ty.isScalar() && Size == 128) { LLT s64 = LLT::scalar(64); auto Split = MIRBuilder.buildUnmerge(s64, Val); auto CTPOP1 = MIRBuilder.buildCTPOP(s64, Split->getOperand(0)); auto CTPOP2 = MIRBuilder.buildCTPOP(s64, Split->getOperand(1)); auto Add = MIRBuilder.buildAdd(s64, CTPOP1, CTPOP2); MIRBuilder.buildZExt(Dst, Add); MI.eraseFromParent(); return true; } if (!ST->hasNEON() || MI.getMF()->getFunction().hasFnAttribute(Attribute::NoImplicitFloat)) { // Use generic lowering when custom lowering is not possible. return Ty.isScalar() && (Size == 32 || Size == 64) && Helper.lowerBitCount(MI) == LegalizerHelper::LegalizeResult::Legalized; } // Pre-conditioning: widen Val up to the nearest vector type. // s32,s64,v4s16,v2s32 -> v8i8 // v8s16,v4s32,v2s64 -> v16i8 LLT VTy = Size == 128 ? LLT::fixed_vector(16, 8) : LLT::fixed_vector(8, 8); if (Ty.isScalar()) { assert((Size == 32 || Size == 64 || Size == 128) && "Expected only 32, 64, or 128 bit scalars!"); if (Size == 32) { Val = MIRBuilder.buildZExt(LLT::scalar(64), Val).getReg(0); } } Val = MIRBuilder.buildBitcast(VTy, Val).getReg(0); // Count bits in each byte-sized lane. auto CTPOP = MIRBuilder.buildCTPOP(VTy, Val); // Sum across lanes. if (ST->hasDotProd() && Ty.isVector() && Ty.getNumElements() >= 2 && Ty.getScalarSizeInBits() != 16) { LLT Dt = Ty == LLT::fixed_vector(2, 64) ? LLT::fixed_vector(4, 32) : Ty; auto Zeros = MIRBuilder.buildConstant(Dt, 0); auto Ones = MIRBuilder.buildConstant(VTy, 1); MachineInstrBuilder Sum; if (Ty == LLT::fixed_vector(2, 64)) { auto UDOT = MIRBuilder.buildInstr(AArch64::G_UDOT, {Dt}, {Zeros, Ones, CTPOP}); Sum = MIRBuilder.buildInstr(AArch64::G_UADDLP, {Ty}, {UDOT}); } else if (Ty == LLT::fixed_vector(4, 32)) { Sum = MIRBuilder.buildInstr(AArch64::G_UDOT, {Dt}, {Zeros, Ones, CTPOP}); } else if (Ty == LLT::fixed_vector(2, 32)) { Sum = MIRBuilder.buildInstr(AArch64::G_UDOT, {Dt}, {Zeros, Ones, CTPOP}); } else { llvm_unreachable("unexpected vector shape"); } Sum->getOperand(0).setReg(Dst); MI.eraseFromParent(); return true; } Register HSum = CTPOP.getReg(0); unsigned Opc; SmallVector HAddTys; if (Ty.isScalar()) { Opc = Intrinsic::aarch64_neon_uaddlv; HAddTys.push_back(LLT::scalar(32)); } else if (Ty == LLT::fixed_vector(8, 16)) { Opc = Intrinsic::aarch64_neon_uaddlp; HAddTys.push_back(LLT::fixed_vector(8, 16)); } else if (Ty == LLT::fixed_vector(4, 32)) { Opc = Intrinsic::aarch64_neon_uaddlp; HAddTys.push_back(LLT::fixed_vector(8, 16)); HAddTys.push_back(LLT::fixed_vector(4, 32)); } else if (Ty == LLT::fixed_vector(2, 64)) { Opc = Intrinsic::aarch64_neon_uaddlp; HAddTys.push_back(LLT::fixed_vector(8, 16)); HAddTys.push_back(LLT::fixed_vector(4, 32)); HAddTys.push_back(LLT::fixed_vector(2, 64)); } else if (Ty == LLT::fixed_vector(4, 16)) { Opc = Intrinsic::aarch64_neon_uaddlp; HAddTys.push_back(LLT::fixed_vector(4, 16)); } else if (Ty == LLT::fixed_vector(2, 32)) { Opc = Intrinsic::aarch64_neon_uaddlp; HAddTys.push_back(LLT::fixed_vector(4, 16)); HAddTys.push_back(LLT::fixed_vector(2, 32)); } else llvm_unreachable("unexpected vector shape"); MachineInstrBuilder UADD; for (LLT HTy : HAddTys) { UADD = MIRBuilder.buildIntrinsic(Opc, {HTy}).addUse(HSum); HSum = UADD.getReg(0); } // Post-conditioning. if (Ty.isScalar() && (Size == 64 || Size == 128)) MIRBuilder.buildZExt(Dst, UADD); else UADD->getOperand(0).setReg(Dst); MI.eraseFromParent(); return true; } bool AArch64LegalizerInfo::legalizeAtomicCmpxchg128( MachineInstr &MI, MachineRegisterInfo &MRI, LegalizerHelper &Helper) const { MachineIRBuilder &MIRBuilder = Helper.MIRBuilder; LLT s64 = LLT::scalar(64); auto Addr = MI.getOperand(1).getReg(); auto DesiredI = MIRBuilder.buildUnmerge({s64, s64}, MI.getOperand(2)); auto NewI = MIRBuilder.buildUnmerge({s64, s64}, MI.getOperand(3)); auto DstLo = MRI.createGenericVirtualRegister(s64); auto DstHi = MRI.createGenericVirtualRegister(s64); MachineInstrBuilder CAS; if (ST->hasLSE()) { // We have 128-bit CASP instructions taking XSeqPair registers, which are // s128. We need the merge/unmerge to bracket the expansion and pair up with // the rest of the MIR so we must reassemble the extracted registers into a // 128-bit known-regclass one with code like this: // // %in1 = REG_SEQUENCE Lo, Hi ; One for each input // %out = CASP %in1, ... // %OldLo = G_EXTRACT %out, 0 // %OldHi = G_EXTRACT %out, 64 auto Ordering = (*MI.memoperands_begin())->getMergedOrdering(); unsigned Opcode; switch (Ordering) { case AtomicOrdering::Acquire: Opcode = AArch64::CASPAX; break; case AtomicOrdering::Release: Opcode = AArch64::CASPLX; break; case AtomicOrdering::AcquireRelease: case AtomicOrdering::SequentiallyConsistent: Opcode = AArch64::CASPALX; break; default: Opcode = AArch64::CASPX; break; } LLT s128 = LLT::scalar(128); auto CASDst = MRI.createGenericVirtualRegister(s128); auto CASDesired = MRI.createGenericVirtualRegister(s128); auto CASNew = MRI.createGenericVirtualRegister(s128); MIRBuilder.buildInstr(TargetOpcode::REG_SEQUENCE, {CASDesired}, {}) .addUse(DesiredI->getOperand(0).getReg()) .addImm(AArch64::sube64) .addUse(DesiredI->getOperand(1).getReg()) .addImm(AArch64::subo64); MIRBuilder.buildInstr(TargetOpcode::REG_SEQUENCE, {CASNew}, {}) .addUse(NewI->getOperand(0).getReg()) .addImm(AArch64::sube64) .addUse(NewI->getOperand(1).getReg()) .addImm(AArch64::subo64); CAS = MIRBuilder.buildInstr(Opcode, {CASDst}, {CASDesired, CASNew, Addr}); MIRBuilder.buildExtract({DstLo}, {CASDst}, 0); MIRBuilder.buildExtract({DstHi}, {CASDst}, 64); } else { // The -O0 CMP_SWAP_128 is friendlier to generate code for because LDXP/STXP // can take arbitrary registers so it just has the normal GPR64 operands the // rest of AArch64 is expecting. auto Ordering = (*MI.memoperands_begin())->getMergedOrdering(); unsigned Opcode; switch (Ordering) { case AtomicOrdering::Acquire: Opcode = AArch64::CMP_SWAP_128_ACQUIRE; break; case AtomicOrdering::Release: Opcode = AArch64::CMP_SWAP_128_RELEASE; break; case AtomicOrdering::AcquireRelease: case AtomicOrdering::SequentiallyConsistent: Opcode = AArch64::CMP_SWAP_128; break; default: Opcode = AArch64::CMP_SWAP_128_MONOTONIC; break; } auto Scratch = MRI.createVirtualRegister(&AArch64::GPR64RegClass); CAS = MIRBuilder.buildInstr(Opcode, {DstLo, DstHi, Scratch}, {Addr, DesiredI->getOperand(0), DesiredI->getOperand(1), NewI->getOperand(0), NewI->getOperand(1)}); } CAS.cloneMemRefs(MI); constrainSelectedInstRegOperands(*CAS, *ST->getInstrInfo(), *MRI.getTargetRegisterInfo(), *ST->getRegBankInfo()); MIRBuilder.buildMergeLikeInstr(MI.getOperand(0), {DstLo, DstHi}); MI.eraseFromParent(); return true; } bool AArch64LegalizerInfo::legalizeCTTZ(MachineInstr &MI, LegalizerHelper &Helper) const { MachineIRBuilder &MIRBuilder = Helper.MIRBuilder; MachineRegisterInfo &MRI = *MIRBuilder.getMRI(); LLT Ty = MRI.getType(MI.getOperand(1).getReg()); auto BitReverse = MIRBuilder.buildBitReverse(Ty, MI.getOperand(1)); MIRBuilder.buildCTLZ(MI.getOperand(0).getReg(), BitReverse); MI.eraseFromParent(); return true; } bool AArch64LegalizerInfo::legalizeMemOps(MachineInstr &MI, LegalizerHelper &Helper) const { MachineIRBuilder &MIRBuilder = Helper.MIRBuilder; // Tagged version MOPSMemorySetTagged is legalised in legalizeIntrinsic if (MI.getOpcode() == TargetOpcode::G_MEMSET) { // Anyext the value being set to 64 bit (only the bottom 8 bits are read by // the instruction). auto &Value = MI.getOperand(1); Register ExtValueReg = MIRBuilder.buildAnyExt(LLT::scalar(64), Value).getReg(0); Value.setReg(ExtValueReg); return true; } return false; } bool AArch64LegalizerInfo::legalizeExtractVectorElt( MachineInstr &MI, MachineRegisterInfo &MRI, LegalizerHelper &Helper) const { assert(MI.getOpcode() == TargetOpcode::G_EXTRACT_VECTOR_ELT); auto VRegAndVal = getIConstantVRegValWithLookThrough(MI.getOperand(2).getReg(), MRI); if (VRegAndVal) return true; return Helper.lowerExtractInsertVectorElt(MI) != LegalizerHelper::LegalizeResult::UnableToLegalize; } bool AArch64LegalizerInfo::legalizeDynStackAlloc( MachineInstr &MI, LegalizerHelper &Helper) const { MachineFunction &MF = *MI.getParent()->getParent(); MachineIRBuilder &MIRBuilder = Helper.MIRBuilder; MachineRegisterInfo &MRI = *MIRBuilder.getMRI(); // If stack probing is not enabled for this function, use the default // lowering. if (!MF.getFunction().hasFnAttribute("probe-stack") || MF.getFunction().getFnAttribute("probe-stack").getValueAsString() != "inline-asm") { Helper.lowerDynStackAlloc(MI); return true; } Register Dst = MI.getOperand(0).getReg(); Register AllocSize = MI.getOperand(1).getReg(); Align Alignment = assumeAligned(MI.getOperand(2).getImm()); assert(MRI.getType(Dst) == LLT::pointer(0, 64) && "Unexpected type for dynamic alloca"); assert(MRI.getType(AllocSize) == LLT::scalar(64) && "Unexpected type for dynamic alloca"); LLT PtrTy = MRI.getType(Dst); Register SPReg = Helper.getTargetLowering().getStackPointerRegisterToSaveRestore(); Register SPTmp = Helper.getDynStackAllocTargetPtr(SPReg, AllocSize, Alignment, PtrTy); auto NewMI = MIRBuilder.buildInstr(AArch64::PROBED_STACKALLOC_DYN, {}, {SPTmp}); MRI.setRegClass(NewMI.getReg(0), &AArch64::GPR64commonRegClass); MIRBuilder.setInsertPt(*NewMI->getParent(), NewMI); MIRBuilder.buildCopy(Dst, SPTmp); MI.eraseFromParent(); return true; } bool AArch64LegalizerInfo::legalizePrefetch(MachineInstr &MI, LegalizerHelper &Helper) const { MachineIRBuilder &MIB = Helper.MIRBuilder; auto &AddrVal = MI.getOperand(0); int64_t IsWrite = MI.getOperand(1).getImm(); int64_t Locality = MI.getOperand(2).getImm(); int64_t IsData = MI.getOperand(3).getImm(); bool IsStream = Locality == 0; if (Locality != 0) { assert(Locality <= 3 && "Prefetch locality out-of-range"); // The locality degree is the opposite of the cache speed. // Put the number the other way around. // The encoding starts at 0 for level 1 Locality = 3 - Locality; } unsigned PrfOp = (IsWrite << 4) | (!IsData << 3) | (Locality << 1) | IsStream; MIB.buildInstr(AArch64::G_AARCH64_PREFETCH).addImm(PrfOp).add(AddrVal); MI.eraseFromParent(); return true; }