blob: 23f05eaad944873433217f9ac3e0910fdd35d51e [file] [log] [blame]
//===-- AArch64ISelLowering.cpp - AArch64 DAG Lowering Implementation ----===//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
// This file implements the AArch64TargetLowering class.
#include "AArch64ISelLowering.h"
#include "AArch64CallingConvention.h"
#include "AArch64ExpandImm.h"
#include "AArch64MachineFunctionInfo.h"
#include "AArch64PerfectShuffle.h"
#include "AArch64RegisterInfo.h"
#include "AArch64Subtarget.h"
#include "MCTargetDesc/AArch64AddressingModes.h"
#include "Utils/AArch64BaseInfo.h"
#include "llvm/ADT/APFloat.h"
#include "llvm/ADT/APInt.h"
#include "llvm/ADT/ArrayRef.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SmallSet.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/ADT/StringRef.h"
#include "llvm/ADT/StringSwitch.h"
#include "llvm/ADT/Triple.h"
#include "llvm/ADT/Twine.h"
#include "llvm/Analysis/VectorUtils.h"
#include "llvm/CodeGen/CallingConvLower.h"
#include "llvm/CodeGen/MachineBasicBlock.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineInstr.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineMemOperand.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/RuntimeLibcalls.h"
#include "llvm/CodeGen/SelectionDAG.h"
#include "llvm/CodeGen/SelectionDAGNodes.h"
#include "llvm/CodeGen/TargetCallingConv.h"
#include "llvm/CodeGen/TargetInstrInfo.h"
#include "llvm/CodeGen/ValueTypes.h"
#include "llvm/IR/Attributes.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/DebugLoc.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/GetElementPtrTypeIterator.h"
#include "llvm/IR/GlobalValue.h"
#include "llvm/IR/IRBuilder.h"
#include "llvm/IR/Instruction.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/Intrinsics.h"
#include "llvm/IR/IntrinsicsAArch64.h"
#include "llvm/IR/Module.h"
#include "llvm/IR/OperandTraits.h"
#include "llvm/IR/PatternMatch.h"
#include "llvm/IR/Type.h"
#include "llvm/IR/Use.h"
#include "llvm/IR/Value.h"
#include "llvm/MC/MCRegisterInfo.h"
#include "llvm/Support/Casting.h"
#include "llvm/Support/CodeGen.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Compiler.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/KnownBits.h"
#include "llvm/Support/MachineValueType.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Target/TargetOptions.h"
#include <algorithm>
#include <bitset>
#include <cassert>
#include <cctype>
#include <cstdint>
#include <cstdlib>
#include <iterator>
#include <limits>
#include <tuple>
#include <utility>
#include <vector>
using namespace llvm;
using namespace llvm::PatternMatch;
#define DEBUG_TYPE "aarch64-lower"
STATISTIC(NumTailCalls, "Number of tail calls");
STATISTIC(NumShiftInserts, "Number of vector shift inserts");
STATISTIC(NumOptimizedImms, "Number of times immediates were optimized");
static cl::opt<bool>
EnableAArch64SlrGeneration("aarch64-shift-insert-generation", cl::Hidden,
cl::desc("Allow AArch64 SLI/SRI formation"),
// FIXME: The necessary dtprel relocations don't seem to be supported
// well in the GNU bfd and gold linkers at the moment. Therefore, by
// default, for now, fall back to GeneralDynamic code generation.
cl::opt<bool> EnableAArch64ELFLocalDynamicTLSGeneration(
"aarch64-elf-ldtls-generation", cl::Hidden,
cl::desc("Allow AArch64 Local Dynamic TLS code generation"),
static cl::opt<bool>
EnableOptimizeLogicalImm("aarch64-enable-logical-imm", cl::Hidden,
cl::desc("Enable AArch64 logical imm instruction "
/// Value type used for condition codes.
static const MVT MVT_CC = MVT::i32;
AArch64TargetLowering::AArch64TargetLowering(const TargetMachine &TM,
const AArch64Subtarget &STI)
: TargetLowering(TM), Subtarget(&STI) {
// AArch64 doesn't have comparisons which set GPRs or setcc instructions, so
// we have to make something up. Arbitrarily, choose ZeroOrOne.
// When comparing vectors the result sets the different elements in the
// vector to all-one or all-zero.
// Set up the register classes.
addRegisterClass(MVT::i32, &AArch64::GPR32allRegClass);
addRegisterClass(MVT::i64, &AArch64::GPR64allRegClass);
if (Subtarget->hasFPARMv8()) {
addRegisterClass(MVT::f16, &AArch64::FPR16RegClass);
addRegisterClass(MVT::f32, &AArch64::FPR32RegClass);
addRegisterClass(MVT::f64, &AArch64::FPR64RegClass);
addRegisterClass(MVT::f128, &AArch64::FPR128RegClass);
if (Subtarget->hasNEON()) {
addRegisterClass(MVT::v16i8, &AArch64::FPR8RegClass);
addRegisterClass(MVT::v8i16, &AArch64::FPR16RegClass);
// Someone set us up the NEON.
if (Subtarget->hasSVE()) {
// Add legal sve predicate types
addRegisterClass(MVT::nxv2i1, &AArch64::PPRRegClass);
addRegisterClass(MVT::nxv4i1, &AArch64::PPRRegClass);
addRegisterClass(MVT::nxv8i1, &AArch64::PPRRegClass);
addRegisterClass(MVT::nxv16i1, &AArch64::PPRRegClass);
// Add legal sve data types
addRegisterClass(MVT::nxv16i8, &AArch64::ZPRRegClass);
addRegisterClass(MVT::nxv8i16, &AArch64::ZPRRegClass);
addRegisterClass(MVT::nxv4i32, &AArch64::ZPRRegClass);
addRegisterClass(MVT::nxv2i64, &AArch64::ZPRRegClass);
addRegisterClass(MVT::nxv2f16, &AArch64::ZPRRegClass);
addRegisterClass(MVT::nxv4f16, &AArch64::ZPRRegClass);
addRegisterClass(MVT::nxv8f16, &AArch64::ZPRRegClass);
addRegisterClass(MVT::nxv2f32, &AArch64::ZPRRegClass);
addRegisterClass(MVT::nxv4f32, &AArch64::ZPRRegClass);
addRegisterClass(MVT::nxv2f64, &AArch64::ZPRRegClass);
for (auto VT : { MVT::nxv16i8, MVT::nxv8i16, MVT::nxv4i32, MVT::nxv2i64 }) {
setOperationAction(ISD::SADDSAT, VT, Legal);
setOperationAction(ISD::UADDSAT, VT, Legal);
setOperationAction(ISD::SSUBSAT, VT, Legal);
setOperationAction(ISD::USUBSAT, VT, Legal);
setOperationAction(ISD::SMAX, VT, Legal);
setOperationAction(ISD::UMAX, VT, Legal);
setOperationAction(ISD::SMIN, VT, Legal);
setOperationAction(ISD::UMIN, VT, Legal);
for (auto VT :
{ MVT::nxv2i8, MVT::nxv2i16, MVT::nxv2i32, MVT::nxv2i64, MVT::nxv4i8,
MVT::nxv4i16, MVT::nxv4i32, MVT::nxv8i8, MVT::nxv8i16 })
setOperationAction(ISD::SIGN_EXTEND_INREG, VT, Legal);
// Compute derived properties from the register classes
// Provide all sorts of operation actions
setOperationAction(ISD::GlobalAddress, MVT::i64, Custom);
setOperationAction(ISD::GlobalTLSAddress, MVT::i64, Custom);
setOperationAction(ISD::SETCC, MVT::i32, Custom);
setOperationAction(ISD::SETCC, MVT::i64, Custom);
setOperationAction(ISD::SETCC, MVT::f16, Custom);
setOperationAction(ISD::SETCC, MVT::f32, Custom);
setOperationAction(ISD::SETCC, MVT::f64, Custom);
setOperationAction(ISD::STRICT_FSETCC, MVT::f16, Custom);
setOperationAction(ISD::STRICT_FSETCC, MVT::f32, Custom);
setOperationAction(ISD::STRICT_FSETCC, MVT::f64, Custom);
setOperationAction(ISD::STRICT_FSETCCS, MVT::f16, Custom);
setOperationAction(ISD::STRICT_FSETCCS, MVT::f32, Custom);
setOperationAction(ISD::STRICT_FSETCCS, MVT::f64, Custom);
setOperationAction(ISD::BITREVERSE, MVT::i32, Legal);
setOperationAction(ISD::BITREVERSE, MVT::i64, Legal);
setOperationAction(ISD::BRCOND, MVT::Other, Expand);
setOperationAction(ISD::BR_CC, MVT::i32, Custom);
setOperationAction(ISD::BR_CC, MVT::i64, Custom);
setOperationAction(ISD::BR_CC, MVT::f16, Custom);
setOperationAction(ISD::BR_CC, MVT::f32, Custom);
setOperationAction(ISD::BR_CC, MVT::f64, Custom);
setOperationAction(ISD::SELECT, MVT::i32, Custom);
setOperationAction(ISD::SELECT, MVT::i64, Custom);
setOperationAction(ISD::SELECT, MVT::f16, Custom);
setOperationAction(ISD::SELECT, MVT::f32, Custom);
setOperationAction(ISD::SELECT, MVT::f64, Custom);
setOperationAction(ISD::SELECT_CC, MVT::i32, Custom);
setOperationAction(ISD::SELECT_CC, MVT::i64, Custom);
setOperationAction(ISD::SELECT_CC, MVT::f16, Custom);
setOperationAction(ISD::SELECT_CC, MVT::f32, Custom);
setOperationAction(ISD::SELECT_CC, MVT::f64, Custom);
setOperationAction(ISD::BR_JT, MVT::Other, Custom);
setOperationAction(ISD::JumpTable, MVT::i64, Custom);
setOperationAction(ISD::SHL_PARTS, MVT::i64, Custom);
setOperationAction(ISD::SRA_PARTS, MVT::i64, Custom);
setOperationAction(ISD::SRL_PARTS, MVT::i64, Custom);
setOperationAction(ISD::FREM, MVT::f32, Expand);
setOperationAction(ISD::FREM, MVT::f64, Expand);
setOperationAction(ISD::FREM, MVT::f80, Expand);
setOperationAction(ISD::BUILD_PAIR, MVT::i64, Expand);
// Custom lowering hooks are needed for XOR
// to fold it into CSINC/CSINV.
setOperationAction(ISD::XOR, MVT::i32, Custom);
setOperationAction(ISD::XOR, MVT::i64, Custom);
// Virtually no operation on f128 is legal, but LLVM can't expand them when
// there's a valid register class, so we need custom operations in most cases.
setOperationAction(ISD::FABS, MVT::f128, Expand);
setOperationAction(ISD::FADD, MVT::f128, Custom);
setOperationAction(ISD::FCOPYSIGN, MVT::f128, Expand);
setOperationAction(ISD::FCOS, MVT::f128, Expand);
setOperationAction(ISD::FDIV, MVT::f128, Custom);
setOperationAction(ISD::FMA, MVT::f128, Expand);
setOperationAction(ISD::FMUL, MVT::f128, Custom);
setOperationAction(ISD::FNEG, MVT::f128, Expand);
setOperationAction(ISD::FPOW, MVT::f128, Expand);
setOperationAction(ISD::FREM, MVT::f128, Expand);
setOperationAction(ISD::FRINT, MVT::f128, Expand);
setOperationAction(ISD::FSIN, MVT::f128, Expand);
setOperationAction(ISD::FSINCOS, MVT::f128, Expand);
setOperationAction(ISD::FSQRT, MVT::f128, Expand);
setOperationAction(ISD::FSUB, MVT::f128, Custom);
setOperationAction(ISD::FTRUNC, MVT::f128, Expand);
setOperationAction(ISD::SETCC, MVT::f128, Custom);
setOperationAction(ISD::STRICT_FSETCC, MVT::f128, Custom);
setOperationAction(ISD::STRICT_FSETCCS, MVT::f128, Custom);
setOperationAction(ISD::BR_CC, MVT::f128, Custom);
setOperationAction(ISD::SELECT, MVT::f128, Custom);
setOperationAction(ISD::SELECT_CC, MVT::f128, Custom);
setOperationAction(ISD::FP_EXTEND, MVT::f128, Custom);
// Lowering for many of the conversions is actually specified by the non-f128
// type. The LowerXXX function will be trivial when f128 isn't involved.
setOperationAction(ISD::FP_TO_SINT, MVT::i32, Custom);
setOperationAction(ISD::FP_TO_SINT, MVT::i64, Custom);
setOperationAction(ISD::FP_TO_SINT, MVT::i128, Custom);
setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::i32, Custom);
setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::i64, Custom);
setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::i128, Custom);
setOperationAction(ISD::FP_TO_UINT, MVT::i32, Custom);
setOperationAction(ISD::FP_TO_UINT, MVT::i64, Custom);
setOperationAction(ISD::FP_TO_UINT, MVT::i128, Custom);
setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::i32, Custom);
setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::i64, Custom);
setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::i128, Custom);
setOperationAction(ISD::SINT_TO_FP, MVT::i32, Custom);
setOperationAction(ISD::SINT_TO_FP, MVT::i64, Custom);
setOperationAction(ISD::SINT_TO_FP, MVT::i128, Custom);
setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::i32, Custom);
setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::i64, Custom);
setOperationAction(ISD::STRICT_SINT_TO_FP, MVT::i128, Custom);
setOperationAction(ISD::UINT_TO_FP, MVT::i32, Custom);
setOperationAction(ISD::UINT_TO_FP, MVT::i64, Custom);
setOperationAction(ISD::UINT_TO_FP, MVT::i128, Custom);
setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::i32, Custom);
setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::i64, Custom);
setOperationAction(ISD::STRICT_UINT_TO_FP, MVT::i128, Custom);
setOperationAction(ISD::FP_ROUND, MVT::f32, Custom);
setOperationAction(ISD::FP_ROUND, MVT::f64, Custom);
setOperationAction(ISD::STRICT_FP_ROUND, MVT::f32, Custom);
setOperationAction(ISD::STRICT_FP_ROUND, MVT::f64, Custom);
// Variable arguments.
setOperationAction(ISD::VASTART, MVT::Other, Custom);
setOperationAction(ISD::VAARG, MVT::Other, Custom);
setOperationAction(ISD::VACOPY, MVT::Other, Custom);
setOperationAction(ISD::VAEND, MVT::Other, Expand);
// Variable-sized objects.
setOperationAction(ISD::STACKSAVE, MVT::Other, Expand);
setOperationAction(ISD::STACKRESTORE, MVT::Other, Expand);
if (Subtarget->isTargetWindows())
setOperationAction(ISD::DYNAMIC_STACKALLOC, MVT::i64, Custom);
setOperationAction(ISD::DYNAMIC_STACKALLOC, MVT::i64, Expand);
// Constant pool entries
setOperationAction(ISD::ConstantPool, MVT::i64, Custom);
// BlockAddress
setOperationAction(ISD::BlockAddress, MVT::i64, Custom);
// Add/Sub overflow ops with MVT::Glues are lowered to NZCV dependences.
setOperationAction(ISD::ADDC, MVT::i32, Custom);
setOperationAction(ISD::ADDE, MVT::i32, Custom);
setOperationAction(ISD::SUBC, MVT::i32, Custom);
setOperationAction(ISD::SUBE, MVT::i32, Custom);
setOperationAction(ISD::ADDC, MVT::i64, Custom);
setOperationAction(ISD::ADDE, MVT::i64, Custom);
setOperationAction(ISD::SUBC, MVT::i64, Custom);
setOperationAction(ISD::SUBE, MVT::i64, Custom);
// AArch64 lacks both left-rotate and popcount instructions.
setOperationAction(ISD::ROTL, MVT::i32, Expand);
setOperationAction(ISD::ROTL, MVT::i64, Expand);
for (MVT VT : MVT::fixedlen_vector_valuetypes()) {
setOperationAction(ISD::ROTL, VT, Expand);
setOperationAction(ISD::ROTR, VT, Expand);
// AArch64 doesn't have {U|S}MUL_LOHI.
setOperationAction(ISD::UMUL_LOHI, MVT::i64, Expand);
setOperationAction(ISD::SMUL_LOHI, MVT::i64, Expand);
setOperationAction(ISD::CTPOP, MVT::i32, Custom);
setOperationAction(ISD::CTPOP, MVT::i64, Custom);
setOperationAction(ISD::SDIVREM, MVT::i32, Expand);
setOperationAction(ISD::SDIVREM, MVT::i64, Expand);
for (MVT VT : MVT::fixedlen_vector_valuetypes()) {
setOperationAction(ISD::SDIVREM, VT, Expand);
setOperationAction(ISD::UDIVREM, VT, Expand);
setOperationAction(ISD::SREM, MVT::i32, Expand);
setOperationAction(ISD::SREM, MVT::i64, Expand);
setOperationAction(ISD::UDIVREM, MVT::i32, Expand);
setOperationAction(ISD::UDIVREM, MVT::i64, Expand);
setOperationAction(ISD::UREM, MVT::i32, Expand);
setOperationAction(ISD::UREM, MVT::i64, Expand);
// Custom lower Add/Sub/Mul with overflow.
setOperationAction(ISD::SADDO, MVT::i32, Custom);
setOperationAction(ISD::SADDO, MVT::i64, Custom);
setOperationAction(ISD::UADDO, MVT::i32, Custom);
setOperationAction(ISD::UADDO, MVT::i64, Custom);
setOperationAction(ISD::SSUBO, MVT::i32, Custom);
setOperationAction(ISD::SSUBO, MVT::i64, Custom);
setOperationAction(ISD::USUBO, MVT::i32, Custom);
setOperationAction(ISD::USUBO, MVT::i64, Custom);
setOperationAction(ISD::SMULO, MVT::i32, Custom);
setOperationAction(ISD::SMULO, MVT::i64, Custom);
setOperationAction(ISD::UMULO, MVT::i32, Custom);
setOperationAction(ISD::UMULO, MVT::i64, Custom);
setOperationAction(ISD::FSIN, MVT::f32, Expand);
setOperationAction(ISD::FSIN, MVT::f64, Expand);
setOperationAction(ISD::FCOS, MVT::f32, Expand);
setOperationAction(ISD::FCOS, MVT::f64, Expand);
setOperationAction(ISD::FPOW, MVT::f32, Expand);
setOperationAction(ISD::FPOW, MVT::f64, Expand);
setOperationAction(ISD::FCOPYSIGN, MVT::f64, Custom);
setOperationAction(ISD::FCOPYSIGN, MVT::f32, Custom);
if (Subtarget->hasFullFP16())
setOperationAction(ISD::FCOPYSIGN, MVT::f16, Custom);
setOperationAction(ISD::FCOPYSIGN, MVT::f16, Promote);
setOperationAction(ISD::FREM, MVT::f16, Promote);
setOperationAction(ISD::FREM, MVT::v4f16, Expand);
setOperationAction(ISD::FREM, MVT::v8f16, Expand);
setOperationAction(ISD::FPOW, MVT::f16, Promote);
setOperationAction(ISD::FPOW, MVT::v4f16, Expand);
setOperationAction(ISD::FPOW, MVT::v8f16, Expand);
setOperationAction(ISD::FPOWI, MVT::f16, Promote);
setOperationAction(ISD::FPOWI, MVT::v4f16, Expand);
setOperationAction(ISD::FPOWI, MVT::v8f16, Expand);
setOperationAction(ISD::FCOS, MVT::f16, Promote);
setOperationAction(ISD::FCOS, MVT::v4f16, Expand);
setOperationAction(ISD::FCOS, MVT::v8f16, Expand);
setOperationAction(ISD::FSIN, MVT::f16, Promote);
setOperationAction(ISD::FSIN, MVT::v4f16, Expand);
setOperationAction(ISD::FSIN, MVT::v8f16, Expand);
setOperationAction(ISD::FSINCOS, MVT::f16, Promote);
setOperationAction(ISD::FSINCOS, MVT::v4f16, Expand);
setOperationAction(ISD::FSINCOS, MVT::v8f16, Expand);
setOperationAction(ISD::FEXP, MVT::f16, Promote);
setOperationAction(ISD::FEXP, MVT::v4f16, Expand);
setOperationAction(ISD::FEXP, MVT::v8f16, Expand);
setOperationAction(ISD::FEXP2, MVT::f16, Promote);
setOperationAction(ISD::FEXP2, MVT::v4f16, Expand);
setOperationAction(ISD::FEXP2, MVT::v8f16, Expand);
setOperationAction(ISD::FLOG, MVT::f16, Promote);
setOperationAction(ISD::FLOG, MVT::v4f16, Expand);
setOperationAction(ISD::FLOG, MVT::v8f16, Expand);
setOperationAction(ISD::FLOG2, MVT::f16, Promote);
setOperationAction(ISD::FLOG2, MVT::v4f16, Expand);
setOperationAction(ISD::FLOG2, MVT::v8f16, Expand);
setOperationAction(ISD::FLOG10, MVT::f16, Promote);
setOperationAction(ISD::FLOG10, MVT::v4f16, Expand);
setOperationAction(ISD::FLOG10, MVT::v8f16, Expand);
if (!Subtarget->hasFullFP16()) {
setOperationAction(ISD::SELECT, MVT::f16, Promote);
setOperationAction(ISD::SELECT_CC, MVT::f16, Promote);
setOperationAction(ISD::SETCC, MVT::f16, Promote);
setOperationAction(ISD::BR_CC, MVT::f16, Promote);
setOperationAction(ISD::FADD, MVT::f16, Promote);
setOperationAction(ISD::FSUB, MVT::f16, Promote);
setOperationAction(ISD::FMUL, MVT::f16, Promote);
setOperationAction(ISD::FDIV, MVT::f16, Promote);
setOperationAction(ISD::FMA, MVT::f16, Promote);
setOperationAction(ISD::FNEG, MVT::f16, Promote);
setOperationAction(ISD::FABS, MVT::f16, Promote);
setOperationAction(ISD::FCEIL, MVT::f16, Promote);
setOperationAction(ISD::FSQRT, MVT::f16, Promote);
setOperationAction(ISD::FFLOOR, MVT::f16, Promote);
setOperationAction(ISD::FNEARBYINT, MVT::f16, Promote);
setOperationAction(ISD::FRINT, MVT::f16, Promote);
setOperationAction(ISD::FROUND, MVT::f16, Promote);
setOperationAction(ISD::FTRUNC, MVT::f16, Promote);
setOperationAction(ISD::FMINNUM, MVT::f16, Promote);
setOperationAction(ISD::FMAXNUM, MVT::f16, Promote);
setOperationAction(ISD::FMINIMUM, MVT::f16, Promote);
setOperationAction(ISD::FMAXIMUM, MVT::f16, Promote);
// promote v4f16 to v4f32 when that is known to be safe.
setOperationAction(ISD::FADD, MVT::v4f16, Promote);
setOperationAction(ISD::FSUB, MVT::v4f16, Promote);
setOperationAction(ISD::FMUL, MVT::v4f16, Promote);
setOperationAction(ISD::FDIV, MVT::v4f16, Promote);
AddPromotedToType(ISD::FADD, MVT::v4f16, MVT::v4f32);
AddPromotedToType(ISD::FSUB, MVT::v4f16, MVT::v4f32);
AddPromotedToType(ISD::FMUL, MVT::v4f16, MVT::v4f32);
AddPromotedToType(ISD::FDIV, MVT::v4f16, MVT::v4f32);
setOperationAction(ISD::FABS, MVT::v4f16, Expand);
setOperationAction(ISD::FNEG, MVT::v4f16, Expand);
setOperationAction(ISD::FROUND, MVT::v4f16, Expand);
setOperationAction(ISD::FMA, MVT::v4f16, Expand);
setOperationAction(ISD::SETCC, MVT::v4f16, Expand);
setOperationAction(ISD::BR_CC, MVT::v4f16, Expand);
setOperationAction(ISD::SELECT, MVT::v4f16, Expand);
setOperationAction(ISD::SELECT_CC, MVT::v4f16, Expand);
setOperationAction(ISD::FTRUNC, MVT::v4f16, Expand);
setOperationAction(ISD::FCOPYSIGN, MVT::v4f16, Expand);
setOperationAction(ISD::FFLOOR, MVT::v4f16, Expand);
setOperationAction(ISD::FCEIL, MVT::v4f16, Expand);
setOperationAction(ISD::FRINT, MVT::v4f16, Expand);
setOperationAction(ISD::FNEARBYINT, MVT::v4f16, Expand);
setOperationAction(ISD::FSQRT, MVT::v4f16, Expand);
setOperationAction(ISD::FABS, MVT::v8f16, Expand);
setOperationAction(ISD::FADD, MVT::v8f16, Expand);
setOperationAction(ISD::FCEIL, MVT::v8f16, Expand);
setOperationAction(ISD::FCOPYSIGN, MVT::v8f16, Expand);
setOperationAction(ISD::FDIV, MVT::v8f16, Expand);
setOperationAction(ISD::FFLOOR, MVT::v8f16, Expand);
setOperationAction(ISD::FMA, MVT::v8f16, Expand);
setOperationAction(ISD::FMUL, MVT::v8f16, Expand);
setOperationAction(ISD::FNEARBYINT, MVT::v8f16, Expand);
setOperationAction(ISD::FNEG, MVT::v8f16, Expand);
setOperationAction(ISD::FROUND, MVT::v8f16, Expand);
setOperationAction(ISD::FRINT, MVT::v8f16, Expand);
setOperationAction(ISD::FSQRT, MVT::v8f16, Expand);
setOperationAction(ISD::FSUB, MVT::v8f16, Expand);
setOperationAction(ISD::FTRUNC, MVT::v8f16, Expand);
setOperationAction(ISD::SETCC, MVT::v8f16, Expand);
setOperationAction(ISD::BR_CC, MVT::v8f16, Expand);
setOperationAction(ISD::SELECT, MVT::v8f16, Expand);
setOperationAction(ISD::SELECT_CC, MVT::v8f16, Expand);
setOperationAction(ISD::FP_EXTEND, MVT::v8f16, Expand);
// AArch64 has implementations of a lot of rounding-like FP operations.
for (MVT Ty : {MVT::f32, MVT::f64}) {
setOperationAction(ISD::FFLOOR, Ty, Legal);
setOperationAction(ISD::FNEARBYINT, Ty, Legal);
setOperationAction(ISD::FCEIL, Ty, Legal);
setOperationAction(ISD::FRINT, Ty, Legal);
setOperationAction(ISD::FTRUNC, Ty, Legal);
setOperationAction(ISD::FROUND, Ty, Legal);
setOperationAction(ISD::FMINNUM, Ty, Legal);
setOperationAction(ISD::FMAXNUM, Ty, Legal);
setOperationAction(ISD::FMINIMUM, Ty, Legal);
setOperationAction(ISD::FMAXIMUM, Ty, Legal);
setOperationAction(ISD::LROUND, Ty, Legal);
setOperationAction(ISD::LLROUND, Ty, Legal);
setOperationAction(ISD::LRINT, Ty, Legal);
setOperationAction(ISD::LLRINT, Ty, Legal);
if (Subtarget->hasFullFP16()) {
setOperationAction(ISD::FNEARBYINT, MVT::f16, Legal);
setOperationAction(ISD::FFLOOR, MVT::f16, Legal);
setOperationAction(ISD::FCEIL, MVT::f16, Legal);
setOperationAction(ISD::FRINT, MVT::f16, Legal);
setOperationAction(ISD::FTRUNC, MVT::f16, Legal);
setOperationAction(ISD::FROUND, MVT::f16, Legal);
setOperationAction(ISD::FMINNUM, MVT::f16, Legal);
setOperationAction(ISD::FMAXNUM, MVT::f16, Legal);
setOperationAction(ISD::FMINIMUM, MVT::f16, Legal);
setOperationAction(ISD::FMAXIMUM, MVT::f16, Legal);
setOperationAction(ISD::PREFETCH, MVT::Other, Custom);
setOperationAction(ISD::FLT_ROUNDS_, MVT::i32, Custom);
setOperationAction(ISD::ATOMIC_CMP_SWAP, MVT::i128, Custom);
setOperationAction(ISD::ATOMIC_LOAD_SUB, MVT::i32, Custom);
setOperationAction(ISD::ATOMIC_LOAD_SUB, MVT::i64, Custom);
setOperationAction(ISD::ATOMIC_LOAD_AND, MVT::i32, Custom);
setOperationAction(ISD::ATOMIC_LOAD_AND, MVT::i64, Custom);
// 128-bit loads and stores can be done without expanding
setOperationAction(ISD::LOAD, MVT::i128, Custom);
setOperationAction(ISD::STORE, MVT::i128, Custom);
// Lower READCYCLECOUNTER using an mrs from PMCCNTR_EL0.
// This requires the Performance Monitors extension.
if (Subtarget->hasPerfMon())
setOperationAction(ISD::READCYCLECOUNTER, MVT::i64, Legal);
if (getLibcallName(RTLIB::SINCOS_STRET_F32) != nullptr &&
getLibcallName(RTLIB::SINCOS_STRET_F64) != nullptr) {
// Issue __sincos_stret if available.
setOperationAction(ISD::FSINCOS, MVT::f64, Custom);
setOperationAction(ISD::FSINCOS, MVT::f32, Custom);
} else {
setOperationAction(ISD::FSINCOS, MVT::f64, Expand);
setOperationAction(ISD::FSINCOS, MVT::f32, Expand);
if (Subtarget->getTargetTriple().isOSMSVCRT()) {
// MSVCRT doesn't have powi; fall back to pow
setLibcallName(RTLIB::POWI_F32, nullptr);
setLibcallName(RTLIB::POWI_F64, nullptr);
// Make floating-point constants legal for the large code model, so they don't
// become loads from the constant pool.
if (Subtarget->isTargetMachO() && TM.getCodeModel() == CodeModel::Large) {
setOperationAction(ISD::ConstantFP, MVT::f32, Legal);
setOperationAction(ISD::ConstantFP, MVT::f64, Legal);
// AArch64 does not have floating-point extending loads, i1 sign-extending
// load, floating-point truncating stores, or v2i32->v2i16 truncating store.
for (MVT VT : MVT::fp_valuetypes()) {
setLoadExtAction(ISD::EXTLOAD, VT, MVT::f16, Expand);
setLoadExtAction(ISD::EXTLOAD, VT, MVT::f32, Expand);
setLoadExtAction(ISD::EXTLOAD, VT, MVT::f64, Expand);
setLoadExtAction(ISD::EXTLOAD, VT, MVT::f80, Expand);
for (MVT VT : MVT::integer_valuetypes())
setLoadExtAction(ISD::SEXTLOAD, VT, MVT::i1, Expand);
setTruncStoreAction(MVT::f32, MVT::f16, Expand);
setTruncStoreAction(MVT::f64, MVT::f32, Expand);
setTruncStoreAction(MVT::f64, MVT::f16, Expand);
setTruncStoreAction(MVT::f128, MVT::f80, Expand);
setTruncStoreAction(MVT::f128, MVT::f64, Expand);
setTruncStoreAction(MVT::f128, MVT::f32, Expand);
setTruncStoreAction(MVT::f128, MVT::f16, Expand);
setOperationAction(ISD::BITCAST, MVT::i16, Custom);
setOperationAction(ISD::BITCAST, MVT::f16, Custom);
// Indexed loads and stores are supported.
for (unsigned im = (unsigned)ISD::PRE_INC;
im != (unsigned)ISD::LAST_INDEXED_MODE; ++im) {
setIndexedLoadAction(im, MVT::i8, Legal);
setIndexedLoadAction(im, MVT::i16, Legal);
setIndexedLoadAction(im, MVT::i32, Legal);
setIndexedLoadAction(im, MVT::i64, Legal);
setIndexedLoadAction(im, MVT::f64, Legal);
setIndexedLoadAction(im, MVT::f32, Legal);
setIndexedLoadAction(im, MVT::f16, Legal);
setIndexedStoreAction(im, MVT::i8, Legal);
setIndexedStoreAction(im, MVT::i16, Legal);
setIndexedStoreAction(im, MVT::i32, Legal);
setIndexedStoreAction(im, MVT::i64, Legal);
setIndexedStoreAction(im, MVT::f64, Legal);
setIndexedStoreAction(im, MVT::f32, Legal);
setIndexedStoreAction(im, MVT::f16, Legal);
// Trap.
setOperationAction(ISD::TRAP, MVT::Other, Legal);
if (Subtarget->isTargetWindows())
setOperationAction(ISD::DEBUGTRAP, MVT::Other, Legal);
// We combine OR nodes for bitfield operations.
// Try to create BICs for vector ANDs.
// Vector add and sub nodes may conceal a high-half opportunity.
// Also, try to fold ADD into CSINC/CSINV..
if (Subtarget->supportsAddressTopByteIgnored())
// In case of strict alignment, avoid an excessive number of byte wide stores.
MaxStoresPerMemsetOptSize = 8;
MaxStoresPerMemset = Subtarget->requiresStrictAlign()
? MaxStoresPerMemsetOptSize : 32;
MaxGluedStoresPerMemcpy = 4;
MaxStoresPerMemcpyOptSize = 4;
MaxStoresPerMemcpy = Subtarget->requiresStrictAlign()
? MaxStoresPerMemcpyOptSize : 16;
MaxStoresPerMemmoveOptSize = MaxStoresPerMemmove = 4;
MaxLoadsPerMemcmpOptSize = 4;
MaxLoadsPerMemcmp = Subtarget->requiresStrictAlign()
? MaxLoadsPerMemcmpOptSize : 8;
EnableExtLdPromotion = true;
// Set required alignment.
// Set preferred alignments.
setPrefLoopAlignment(Align(1ULL << STI.getPrefLoopLogAlignment()));
setPrefFunctionAlignment(Align(1ULL << STI.getPrefFunctionLogAlignment()));
// Only change the limit for entries in a jump table if specified by
// the sub target, but not at the command line.
unsigned MaxJT = STI.getMaximumJumpTableSize();
if (MaxJT && getMaximumJumpTableSize() == UINT_MAX)
setOperationAction(ISD::INTRINSIC_WO_CHAIN, MVT::Other, Custom);
if (Subtarget->hasNEON()) {
// FIXME: v1f64 shouldn't be legal if we can avoid it, because it leads to
// silliness like this:
setOperationAction(ISD::FABS, MVT::v1f64, Expand);
setOperationAction(ISD::FADD, MVT::v1f64, Expand);
setOperationAction(ISD::FCEIL, MVT::v1f64, Expand);
setOperationAction(ISD::FCOPYSIGN, MVT::v1f64, Expand);
setOperationAction(ISD::FCOS, MVT::v1f64, Expand);
setOperationAction(ISD::FDIV, MVT::v1f64, Expand);
setOperationAction(ISD::FFLOOR, MVT::v1f64, Expand);
setOperationAction(ISD::FMA, MVT::v1f64, Expand);
setOperationAction(ISD::FMUL, MVT::v1f64, Expand);
setOperationAction(ISD::FNEARBYINT, MVT::v1f64, Expand);
setOperationAction(ISD::FNEG, MVT::v1f64, Expand);
setOperationAction(ISD::FPOW, MVT::v1f64, Expand);
setOperationAction(ISD::FREM, MVT::v1f64, Expand);
setOperationAction(ISD::FROUND, MVT::v1f64, Expand);
setOperationAction(ISD::FRINT, MVT::v1f64, Expand);
setOperationAction(ISD::FSIN, MVT::v1f64, Expand);
setOperationAction(ISD::FSINCOS, MVT::v1f64, Expand);
setOperationAction(ISD::FSQRT, MVT::v1f64, Expand);
setOperationAction(ISD::FSUB, MVT::v1f64, Expand);
setOperationAction(ISD::FTRUNC, MVT::v1f64, Expand);
setOperationAction(ISD::SETCC, MVT::v1f64, Expand);
setOperationAction(ISD::BR_CC, MVT::v1f64, Expand);
setOperationAction(ISD::SELECT, MVT::v1f64, Expand);
setOperationAction(ISD::SELECT_CC, MVT::v1f64, Expand);
setOperationAction(ISD::FP_EXTEND, MVT::v1f64, Expand);
setOperationAction(ISD::FP_TO_SINT, MVT::v1i64, Expand);
setOperationAction(ISD::FP_TO_UINT, MVT::v1i64, Expand);
setOperationAction(ISD::SINT_TO_FP, MVT::v1i64, Expand);
setOperationAction(ISD::UINT_TO_FP, MVT::v1i64, Expand);
setOperationAction(ISD::FP_ROUND, MVT::v1f64, Expand);
setOperationAction(ISD::MUL, MVT::v1i64, Expand);
// AArch64 doesn't have a direct vector ->f32 conversion instructions for
// elements smaller than i32, so promote the input to i32 first.
setOperationPromotedToType(ISD::UINT_TO_FP, MVT::v4i8, MVT::v4i32);
setOperationPromotedToType(ISD::SINT_TO_FP, MVT::v4i8, MVT::v4i32);
// i8 vector elements also need promotion to i32 for v8i8
setOperationPromotedToType(ISD::SINT_TO_FP, MVT::v8i8, MVT::v8i32);
setOperationPromotedToType(ISD::UINT_TO_FP, MVT::v8i8, MVT::v8i32);
// Similarly, there is no direct i32 -> f64 vector conversion instruction.
setOperationAction(ISD::SINT_TO_FP, MVT::v2i32, Custom);
setOperationAction(ISD::UINT_TO_FP, MVT::v2i32, Custom);
setOperationAction(ISD::SINT_TO_FP, MVT::v2i64, Custom);
setOperationAction(ISD::UINT_TO_FP, MVT::v2i64, Custom);
// Or, direct i32 -> f16 vector conversion. Set it so custom, so the
// conversion happens in two steps: v4i32 -> v4f32 -> v4f16
setOperationAction(ISD::SINT_TO_FP, MVT::v4i32, Custom);
setOperationAction(ISD::UINT_TO_FP, MVT::v4i32, Custom);
if (Subtarget->hasFullFP16()) {
setOperationAction(ISD::SINT_TO_FP, MVT::v4i16, Custom);
setOperationAction(ISD::UINT_TO_FP, MVT::v4i16, Custom);
setOperationAction(ISD::SINT_TO_FP, MVT::v8i16, Custom);
setOperationAction(ISD::UINT_TO_FP, MVT::v8i16, Custom);
} else {
// when AArch64 doesn't have fullfp16 support, promote the input
// to i32 first.
setOperationPromotedToType(ISD::UINT_TO_FP, MVT::v4i16, MVT::v4i32);
setOperationPromotedToType(ISD::SINT_TO_FP, MVT::v4i16, MVT::v4i32);
setOperationPromotedToType(ISD::SINT_TO_FP, MVT::v8i16, MVT::v8i32);
setOperationPromotedToType(ISD::UINT_TO_FP, MVT::v8i16, MVT::v8i32);
setOperationAction(ISD::CTLZ, MVT::v1i64, Expand);
setOperationAction(ISD::CTLZ, MVT::v2i64, Expand);
// AArch64 doesn't have MUL.2d:
setOperationAction(ISD::MUL, MVT::v2i64, Expand);
// Custom handling for some quad-vector types to detect MULL.
setOperationAction(ISD::MUL, MVT::v8i16, Custom);
setOperationAction(ISD::MUL, MVT::v4i32, Custom);
setOperationAction(ISD::MUL, MVT::v2i64, Custom);
for (MVT VT : { MVT::v8i8, MVT::v4i16, MVT::v2i32,
MVT::v16i8, MVT::v8i16, MVT::v4i32, MVT::v2i64 }) {
// Vector reductions
setOperationAction(ISD::VECREDUCE_ADD, VT, Custom);
setOperationAction(ISD::VECREDUCE_SMAX, VT, Custom);
setOperationAction(ISD::VECREDUCE_SMIN, VT, Custom);
setOperationAction(ISD::VECREDUCE_UMAX, VT, Custom);
setOperationAction(ISD::VECREDUCE_UMIN, VT, Custom);
// Saturates
setOperationAction(ISD::SADDSAT, VT, Legal);
setOperationAction(ISD::UADDSAT, VT, Legal);
setOperationAction(ISD::SSUBSAT, VT, Legal);
setOperationAction(ISD::USUBSAT, VT, Legal);
for (MVT VT : { MVT::v4f16, MVT::v2f32,
MVT::v8f16, MVT::v4f32, MVT::v2f64 }) {
setOperationAction(ISD::VECREDUCE_FMAX, VT, Custom);
setOperationAction(ISD::VECREDUCE_FMIN, VT, Custom);
setOperationAction(ISD::ANY_EXTEND, MVT::v4i32, Legal);
setTruncStoreAction(MVT::v2i32, MVT::v2i16, Expand);
// Likewise, narrowing and extending vector loads/stores aren't handled
// directly.
for (MVT VT : MVT::fixedlen_vector_valuetypes()) {
setOperationAction(ISD::SIGN_EXTEND_INREG, VT, Expand);
if (VT == MVT::v16i8 || VT == MVT::v8i16 || VT == MVT::v4i32) {
setOperationAction(ISD::MULHS, VT, Legal);
setOperationAction(ISD::MULHU, VT, Legal);
} else {
setOperationAction(ISD::MULHS, VT, Expand);
setOperationAction(ISD::MULHU, VT, Expand);
setOperationAction(ISD::SMUL_LOHI, VT, Expand);
setOperationAction(ISD::UMUL_LOHI, VT, Expand);
setOperationAction(ISD::BSWAP, VT, Expand);
setOperationAction(ISD::CTTZ, VT, Expand);
for (MVT InnerVT : MVT::fixedlen_vector_valuetypes()) {
setTruncStoreAction(VT, InnerVT, Expand);
setLoadExtAction(ISD::SEXTLOAD, VT, InnerVT, Expand);
setLoadExtAction(ISD::ZEXTLOAD, VT, InnerVT, Expand);
setLoadExtAction(ISD::EXTLOAD, VT, InnerVT, Expand);
// AArch64 has implementations of a lot of rounding-like FP operations.
for (MVT Ty : {MVT::v2f32, MVT::v4f32, MVT::v2f64}) {
setOperationAction(ISD::FFLOOR, Ty, Legal);
setOperationAction(ISD::FNEARBYINT, Ty, Legal);
setOperationAction(ISD::FCEIL, Ty, Legal);
setOperationAction(ISD::FRINT, Ty, Legal);
setOperationAction(ISD::FTRUNC, Ty, Legal);
setOperationAction(ISD::FROUND, Ty, Legal);
if (Subtarget->hasFullFP16()) {
for (MVT Ty : {MVT::v4f16, MVT::v8f16}) {
setOperationAction(ISD::FFLOOR, Ty, Legal);
setOperationAction(ISD::FNEARBYINT, Ty, Legal);
setOperationAction(ISD::FCEIL, Ty, Legal);
setOperationAction(ISD::FRINT, Ty, Legal);
setOperationAction(ISD::FTRUNC, Ty, Legal);
setOperationAction(ISD::FROUND, Ty, Legal);
setTruncStoreAction(MVT::v4i16, MVT::v4i8, Custom);
if (Subtarget->hasSVE()) {
// FIXME: Add custom lowering of MLOAD to handle different passthrus (not a
// splat of 0 or undef) once vector selects supported in SVE codegen. See
// D68877 for more details.
for (MVT VT : MVT::integer_scalable_vector_valuetypes()) {
if (isTypeLegal(VT))
setOperationAction(ISD::SPLAT_VECTOR, VT, Custom);
setOperationAction(ISD::INTRINSIC_WO_CHAIN, MVT::i8, Custom);
setOperationAction(ISD::INTRINSIC_WO_CHAIN, MVT::i16, Custom);
PredictableSelectIsExpensive = Subtarget->predictableSelectIsExpensive();
void AArch64TargetLowering::addTypeForNEON(MVT VT, MVT PromotedBitwiseVT) {
assert(VT.isVector() && "VT should be a vector type");
if (VT.isFloatingPoint()) {
MVT PromoteTo = EVT(VT).changeVectorElementTypeToInteger().getSimpleVT();
setOperationPromotedToType(ISD::LOAD, VT, PromoteTo);
setOperationPromotedToType(ISD::STORE, VT, PromoteTo);
// Mark vector float intrinsics as expand.
if (VT == MVT::v2f32 || VT == MVT::v4f32 || VT == MVT::v2f64) {
setOperationAction(ISD::FSIN, VT, Expand);
setOperationAction(ISD::FCOS, VT, Expand);
setOperationAction(ISD::FPOW, VT, Expand);
setOperationAction(ISD::FLOG, VT, Expand);
setOperationAction(ISD::FLOG2, VT, Expand);
setOperationAction(ISD::FLOG10, VT, Expand);
setOperationAction(ISD::FEXP, VT, Expand);
setOperationAction(ISD::FEXP2, VT, Expand);
// But we do support custom-lowering for FCOPYSIGN.
setOperationAction(ISD::FCOPYSIGN, VT, Custom);
setOperationAction(ISD::EXTRACT_VECTOR_ELT, VT, Custom);
setOperationAction(ISD::INSERT_VECTOR_ELT, VT, Custom);
setOperationAction(ISD::BUILD_VECTOR, VT, Custom);
setOperationAction(ISD::VECTOR_SHUFFLE, VT, Custom);
setOperationAction(ISD::EXTRACT_SUBVECTOR, VT, Custom);
setOperationAction(ISD::SRA, VT, Custom);
setOperationAction(ISD::SRL, VT, Custom);
setOperationAction(ISD::SHL, VT, Custom);
setOperationAction(ISD::OR, VT, Custom);
setOperationAction(ISD::SETCC, VT, Custom);
setOperationAction(ISD::CONCAT_VECTORS, VT, Legal);
setOperationAction(ISD::SELECT, VT, Expand);
setOperationAction(ISD::SELECT_CC, VT, Expand);
setOperationAction(ISD::VSELECT, VT, Expand);
for (MVT InnerVT : MVT::all_valuetypes())
setLoadExtAction(ISD::EXTLOAD, InnerVT, VT, Expand);
// CNT supports only B element sizes, then use UADDLP to widen.
if (VT != MVT::v8i8 && VT != MVT::v16i8)
setOperationAction(ISD::CTPOP, VT, Custom);
setOperationAction(ISD::UDIV, VT, Expand);
setOperationAction(ISD::SDIV, VT, Expand);
setOperationAction(ISD::UREM, VT, Expand);
setOperationAction(ISD::SREM, VT, Expand);
setOperationAction(ISD::FREM, VT, Expand);
setOperationAction(ISD::FP_TO_SINT, VT, Custom);
setOperationAction(ISD::FP_TO_UINT, VT, Custom);
if (!VT.isFloatingPoint())
setOperationAction(ISD::ABS, VT, Legal);
// [SU][MIN|MAX] are available for all NEON types apart from i64.
if (!VT.isFloatingPoint() && VT != MVT::v2i64 && VT != MVT::v1i64)
for (unsigned Opcode : {ISD::SMIN, ISD::SMAX, ISD::UMIN, ISD::UMAX})
setOperationAction(Opcode, VT, Legal);
// F[MIN|MAX][NUM|NAN] are available for all FP NEON types.
if (VT.isFloatingPoint() &&
(VT.getVectorElementType() != MVT::f16 || Subtarget->hasFullFP16()))
for (unsigned Opcode :
setOperationAction(Opcode, VT, Legal);
if (Subtarget->isLittleEndian()) {
for (unsigned im = (unsigned)ISD::PRE_INC;
im != (unsigned)ISD::LAST_INDEXED_MODE; ++im) {
setIndexedLoadAction(im, VT, Legal);
setIndexedStoreAction(im, VT, Legal);
void AArch64TargetLowering::addDRTypeForNEON(MVT VT) {
addRegisterClass(VT, &AArch64::FPR64RegClass);
addTypeForNEON(VT, MVT::v2i32);
void AArch64TargetLowering::addQRTypeForNEON(MVT VT) {
addRegisterClass(VT, &AArch64::FPR128RegClass);
addTypeForNEON(VT, MVT::v4i32);
EVT AArch64TargetLowering::getSetCCResultType(const DataLayout &, LLVMContext &,
EVT VT) const {
if (!VT.isVector())
return MVT::i32;
return VT.changeVectorElementTypeToInteger();
static bool optimizeLogicalImm(SDValue Op, unsigned Size, uint64_t Imm,
const APInt &Demanded,
TargetLowering::TargetLoweringOpt &TLO,
unsigned NewOpc) {
uint64_t OldImm = Imm, NewImm, Enc;
uint64_t Mask = ((uint64_t)(-1LL) >> (64 - Size)), OrigMask = Mask;
// Return if the immediate is already all zeros, all ones, a bimm32 or a
// bimm64.
if (Imm == 0 || Imm == Mask ||
AArch64_AM::isLogicalImmediate(Imm & Mask, Size))
return false;
unsigned EltSize = Size;
uint64_t DemandedBits = Demanded.getZExtValue();
// Clear bits that are not demanded.
Imm &= DemandedBits;
while (true) {
// The goal here is to set the non-demanded bits in a way that minimizes
// the number of switching between 0 and 1. In order to achieve this goal,
// we set the non-demanded bits to the value of the preceding demanded bits.
// For example, if we have an immediate 0bx10xx0x1 ('x' indicates a
// non-demanded bit), we copy bit0 (1) to the least significant 'x',
// bit2 (0) to 'xx', and bit6 (1) to the most significant 'x'.
// The final result is 0b11000011.
uint64_t NonDemandedBits = ~DemandedBits;
uint64_t InvertedImm = ~Imm & DemandedBits;
uint64_t RotatedImm =
((InvertedImm << 1) | (InvertedImm >> (EltSize - 1) & 1)) &
uint64_t Sum = RotatedImm + NonDemandedBits;
bool Carry = NonDemandedBits & ~Sum & (1ULL << (EltSize - 1));
uint64_t Ones = (Sum + Carry) & NonDemandedBits;
NewImm = (Imm | Ones) & Mask;
// If NewImm or its bitwise NOT is a shifted mask, it is a bitmask immediate
// or all-ones or all-zeros, in which case we can stop searching. Otherwise,
// we halve the element size and continue the search.
if (isShiftedMask_64(NewImm) || isShiftedMask_64(~(NewImm | ~Mask)))
// We cannot shrink the element size any further if it is 2-bits.
if (EltSize == 2)
return false;
EltSize /= 2;
Mask >>= EltSize;
uint64_t Hi = Imm >> EltSize, DemandedBitsHi = DemandedBits >> EltSize;
// Return if there is mismatch in any of the demanded bits of Imm and Hi.
if (((Imm ^ Hi) & (DemandedBits & DemandedBitsHi) & Mask) != 0)
return false;
// Merge the upper and lower halves of Imm and DemandedBits.
Imm |= Hi;
DemandedBits |= DemandedBitsHi;
// Replicate the element across the register width.
while (EltSize < Size) {
NewImm |= NewImm << EltSize;
EltSize *= 2;
assert(((OldImm ^ NewImm) & Demanded.getZExtValue()) == 0 &&
"demanded bits should never be altered");
assert(OldImm != NewImm && "the new imm shouldn't be equal to the old imm");
// Create the new constant immediate node.
EVT VT = Op.getValueType();
SDLoc DL(Op);
SDValue New;
// If the new constant immediate is all-zeros or all-ones, let the target
// independent DAG combine optimize this node.
if (NewImm == 0 || NewImm == OrigMask) {
New = TLO.DAG.getNode(Op.getOpcode(), DL, VT, Op.getOperand(0),
TLO.DAG.getConstant(NewImm, DL, VT));
// Otherwise, create a machine node so that target independent DAG combine
// doesn't undo this optimization.
} else {
Enc = AArch64_AM::encodeLogicalImmediate(NewImm, Size);
SDValue EncConst = TLO.DAG.getTargetConstant(Enc, DL, VT);
New = SDValue(
TLO.DAG.getMachineNode(NewOpc, DL, VT, Op.getOperand(0), EncConst), 0);
return TLO.CombineTo(Op, New);
bool AArch64TargetLowering::targetShrinkDemandedConstant(
SDValue Op, const APInt &Demanded, TargetLoweringOpt &TLO) const {
// Delay this optimization to as late as possible.
if (!TLO.LegalOps)
return false;
if (!EnableOptimizeLogicalImm)
return false;
EVT VT = Op.getValueType();
if (VT.isVector())
return false;
unsigned Size = VT.getSizeInBits();
assert((Size == 32 || Size == 64) &&
"i32 or i64 is expected after legalization.");
// Exit early if we demand all bits.
if (Demanded.countPopulation() == Size)
return false;
unsigned NewOpc;
switch (Op.getOpcode()) {
return false;
case ISD::AND:
NewOpc = Size == 32 ? AArch64::ANDWri : AArch64::ANDXri;
case ISD::OR:
NewOpc = Size == 32 ? AArch64::ORRWri : AArch64::ORRXri;
case ISD::XOR:
NewOpc = Size == 32 ? AArch64::EORWri : AArch64::EORXri;
ConstantSDNode *C = dyn_cast<ConstantSDNode>(Op.getOperand(1));
if (!C)
return false;
uint64_t Imm = C->getZExtValue();
return optimizeLogicalImm(Op, Size, Imm, Demanded, TLO, NewOpc);
/// computeKnownBitsForTargetNode - Determine which of the bits specified in
/// Mask are known to be either zero or one and return them Known.
void AArch64TargetLowering::computeKnownBitsForTargetNode(
const SDValue Op, KnownBits &Known,
const APInt &DemandedElts, const SelectionDAG &DAG, unsigned Depth) const {
switch (Op.getOpcode()) {
case AArch64ISD::CSEL: {
KnownBits Known2;
Known = DAG.computeKnownBits(Op->getOperand(0), Depth + 1);
Known2 = DAG.computeKnownBits(Op->getOperand(1), Depth + 1);
Known.Zero &= Known2.Zero;
Known.One &= Known2.One;
case AArch64ISD::LOADgot:
case AArch64ISD::ADDlow: {
if (!Subtarget->isTargetILP32())
// In ILP32 mode all valid pointers are in the low 4GB of the address-space.
Known.Zero = APInt::getHighBitsSet(64, 32);
ConstantSDNode *CN = cast<ConstantSDNode>(Op->getOperand(1));
Intrinsic::ID IntID = static_cast<Intrinsic::ID>(CN->getZExtValue());
switch (IntID) {
default: return;
case Intrinsic::aarch64_ldaxr:
case Intrinsic::aarch64_ldxr: {
unsigned BitWidth = Known.getBitWidth();
EVT VT = cast<MemIntrinsicSDNode>(Op)->getMemoryVT();
unsigned MemBits = VT.getScalarSizeInBits();
Known.Zero |= APInt::getHighBitsSet(BitWidth, BitWidth - MemBits);
unsigned IntNo = cast<ConstantSDNode>(Op.getOperand(0))->getZExtValue();
switch (IntNo) {
case Intrinsic::aarch64_neon_umaxv:
case Intrinsic::aarch64_neon_uminv: {
// Figure out the datatype of the vector operand. The UMINV instruction
// will zero extend the result, so we can mark as known zero all the
// bits larger than the element datatype. 32-bit or larget doesn't need
// this as those are legal types and will be handled by isel directly.
MVT VT = Op.getOperand(1).getValueType().getSimpleVT();
unsigned BitWidth = Known.getBitWidth();
if (VT == MVT::v8i8 || VT == MVT::v16i8) {
assert(BitWidth >= 8 && "Unexpected width!");
APInt Mask = APInt::getHighBitsSet(BitWidth, BitWidth - 8);
Known.Zero |= Mask;
} else if (VT == MVT::v4i16 || VT == MVT::v8i16) {
assert(BitWidth >= 16 && "Unexpected width!");
APInt Mask = APInt::getHighBitsSet(BitWidth, BitWidth - 16);
Known.Zero |= Mask;
} break;
MVT AArch64TargetLowering::getScalarShiftAmountTy(const DataLayout &DL,
EVT) const {
return MVT::i64;
bool AArch64TargetLowering::allowsMisalignedMemoryAccesses(
EVT VT, unsigned AddrSpace, unsigned Align, MachineMemOperand::Flags Flags,
bool *Fast) const {
if (Subtarget->requiresStrictAlign())
return false;
if (Fast) {
// Some CPUs are fine with unaligned stores except for 128-bit ones.
*Fast = !Subtarget->isMisaligned128StoreSlow() || VT.getStoreSize() != 16 ||
// See comments in performSTORECombine() for more details about
// these conditions.
// Code that uses clang vector extensions can mark that it
// wants unaligned accesses to be treated as fast by
// underspecifying alignment to be 1 or 2.
Align <= 2 ||
// Disregard v2i64. Memcpy lowering produces those and splitting
// them regresses performance on micro-benchmarks and olden/bh.
VT == MVT::v2i64;
return true;
// Same as above but handling LLTs instead.
bool AArch64TargetLowering::allowsMisalignedMemoryAccesses(
LLT Ty, unsigned AddrSpace, unsigned Align, MachineMemOperand::Flags Flags,
bool *Fast) const {
if (Subtarget->requiresStrictAlign())
return false;
if (Fast) {
// Some CPUs are fine with unaligned stores except for 128-bit ones.
*Fast = !Subtarget->isMisaligned128StoreSlow() ||
Ty.getSizeInBytes() != 16 ||
// See comments in performSTORECombine() for more details about
// these conditions.
// Code that uses clang vector extensions can mark that it
// wants unaligned accesses to be treated as fast by
// underspecifying alignment to be 1 or 2.
Align <= 2 ||
// Disregard v2i64. Memcpy lowering produces those and splitting
// them regresses performance on micro-benchmarks and olden/bh.
Ty == LLT::vector(2, 64);
return true;
FastISel *
AArch64TargetLowering::createFastISel(FunctionLoweringInfo &funcInfo,
const TargetLibraryInfo *libInfo) const {
return AArch64::createFastISel(funcInfo, libInfo);
const char *AArch64TargetLowering::getTargetNodeName(unsigned Opcode) const {
switch ((AArch64ISD::NodeType)Opcode) {
case AArch64ISD::FIRST_NUMBER: break;
case AArch64ISD::CALL: return "AArch64ISD::CALL";
case AArch64ISD::ADRP: return "AArch64ISD::ADRP";
case AArch64ISD::ADR: return "AArch64ISD::ADR";
case AArch64ISD::ADDlow: return "AArch64ISD::ADDlow";
case AArch64ISD::LOADgot: return "AArch64ISD::LOADgot";
case AArch64ISD::RET_FLAG: return "AArch64ISD::RET_FLAG";
case AArch64ISD::BRCOND: return "AArch64ISD::BRCOND";
case AArch64ISD::CSEL: return "AArch64ISD::CSEL";
case AArch64ISD::FCSEL: return "AArch64ISD::FCSEL";
case AArch64ISD::CSINV: return "AArch64ISD::CSINV";
case AArch64ISD::CSNEG: return "AArch64ISD::CSNEG";
case AArch64ISD::CSINC: return "AArch64ISD::CSINC";
case AArch64ISD::ADC: return "AArch64ISD::ADC";
case AArch64ISD::SBC: return "AArch64ISD::SBC";
case AArch64ISD::ADDS: return "AArch64ISD::ADDS";
case AArch64ISD::SUBS: return "AArch64ISD::SUBS";
case AArch64ISD::ADCS: return "AArch64ISD::ADCS";
case AArch64ISD::SBCS: return "AArch64ISD::SBCS";
case AArch64ISD::ANDS: return "AArch64ISD::ANDS";
case AArch64ISD::CCMP: return "AArch64ISD::CCMP";
case AArch64ISD::CCMN: return "AArch64ISD::CCMN";
case AArch64ISD::FCCMP: return "AArch64ISD::FCCMP";
case AArch64ISD::FCMP: return "AArch64ISD::FCMP";
case AArch64ISD::STRICT_FCMP: return "AArch64ISD::STRICT_FCMP";
case AArch64ISD::STRICT_FCMPE: return "AArch64ISD::STRICT_FCMPE";
case AArch64ISD::DUP: return "AArch64ISD::DUP";
case AArch64ISD::DUPLANE8: return "AArch64ISD::DUPLANE8";
case AArch64ISD::DUPLANE16: return "AArch64ISD::DUPLANE16";
case AArch64ISD::DUPLANE32: return "AArch64ISD::DUPLANE32";
case AArch64ISD::DUPLANE64: return "AArch64ISD::DUPLANE64";
case AArch64ISD::MOVI: return "AArch64ISD::MOVI";
case AArch64ISD::MOVIshift: return "AArch64ISD::MOVIshift";
case AArch64ISD::MOVIedit: return "AArch64ISD::MOVIedit";
case AArch64ISD::MOVImsl: return "AArch64ISD::MOVImsl";
case AArch64ISD::FMOV: return "AArch64ISD::FMOV";
case AArch64ISD::MVNIshift: return "AArch64ISD::MVNIshift";
case AArch64ISD::MVNImsl: return "AArch64ISD::MVNImsl";
case AArch64ISD::BICi: return "AArch64ISD::BICi";
case AArch64ISD::ORRi: return "AArch64ISD::ORRi";
case AArch64ISD::BSL: return "AArch64ISD::BSL";
case AArch64ISD::NEG: return "AArch64ISD::NEG";
case AArch64ISD::EXTR: return "AArch64ISD::EXTR";
case AArch64ISD::ZIP1: return "AArch64ISD::ZIP1";
case AArch64ISD::ZIP2: return "AArch64ISD::ZIP2";
case AArch64ISD::UZP1: return "AArch64ISD::UZP1";
case AArch64ISD::UZP2: return "AArch64ISD::UZP2";
case AArch64ISD::TRN1: return "AArch64ISD::TRN1";
case AArch64ISD::TRN2: return "AArch64ISD::TRN2";
case AArch64ISD::REV16: return "AArch64ISD::REV16";
case AArch64ISD::REV32: return "AArch64ISD::REV32";
case AArch64ISD::REV64: return "AArch64ISD::REV64";
case AArch64ISD::EXT: return "AArch64ISD::EXT";
case AArch64ISD::VSHL: return "AArch64ISD::VSHL";
case AArch64ISD::VLSHR: return "AArch64ISD::VLSHR";
case AArch64ISD::VASHR: return "AArch64ISD::VASHR";
case AArch64ISD::CMEQ: return "AArch64ISD::CMEQ";
case AArch64ISD::CMGE: return "AArch64ISD::CMGE";
case AArch64ISD::CMGT: return "AArch64ISD::CMGT";
case AArch64ISD::CMHI: return "AArch64ISD::CMHI";
case AArch64ISD::CMHS: return "AArch64ISD::CMHS";
case AArch64ISD::FCMEQ: return "AArch64ISD::FCMEQ";
case AArch64ISD::FCMGE: return "AArch64ISD::FCMGE";
case AArch64ISD::FCMGT: return "AArch64ISD::FCMGT";
case AArch64ISD::CMEQz: return "AArch64ISD::CMEQz";
case AArch64ISD::CMGEz: return "AArch64ISD::CMGEz";
case AArch64ISD::CMGTz: return "AArch64ISD::CMGTz";
case AArch64ISD::CMLEz: return "AArch64ISD::CMLEz";
case AArch64ISD::CMLTz: return "AArch64ISD::CMLTz";
case AArch64ISD::FCMEQz: return "AArch64ISD::FCMEQz";
case AArch64ISD::FCMGEz: return "AArch64ISD::FCMGEz";
case AArch64ISD::FCMGTz: return "AArch64ISD::FCMGTz";
case AArch64ISD::FCMLEz: return "AArch64ISD::FCMLEz";
case AArch64ISD::FCMLTz: return "AArch64ISD::FCMLTz";
case AArch64ISD::SADDV: return "AArch64ISD::SADDV";
case AArch64ISD::UADDV: return "AArch64ISD::UADDV";
case AArch64ISD::SMINV: return "AArch64ISD::SMINV";
case AArch64ISD::UMINV: return "AArch64ISD::UMINV";
case AArch64ISD::SMAXV: return "AArch64ISD::SMAXV";
case AArch64ISD::UMAXV: return "AArch64ISD::UMAXV";
case AArch64ISD::SMAXV_PRED: return "AArch64ISD::SMAXV_PRED";
case AArch64ISD::UMAXV_PRED: return "AArch64ISD::UMAXV_PRED";
case AArch64ISD::SMINV_PRED: return "AArch64ISD::SMINV_PRED";
case AArch64ISD::UMINV_PRED: return "AArch64ISD::UMINV_PRED";
case AArch64ISD::ORV_PRED: return "AArch64ISD::ORV_PRED";
case AArch64ISD::EORV_PRED: return "AArch64ISD::EORV_PRED";
case AArch64ISD::ANDV_PRED: return "AArch64ISD::ANDV_PRED";
case AArch64ISD::CLASTA_N: return "AArch64ISD::CLASTA_N";
case AArch64ISD::CLASTB_N: return "AArch64ISD::CLASTB_N";
case AArch64ISD::LASTA: return "AArch64ISD::LASTA";
case AArch64ISD::LASTB: return "AArch64ISD::LASTB";
case AArch64ISD::REV: return "AArch64ISD::REV";
case AArch64ISD::TBL: return "AArch64ISD::TBL";
case AArch64ISD::NOT: return "AArch64ISD::NOT";
case AArch64ISD::BIT: return "AArch64ISD::BIT";
case AArch64ISD::CBZ: return "AArch64ISD::CBZ";
case AArch64ISD::CBNZ: return "AArch64ISD::CBNZ";
case AArch64ISD::TBZ: return "AArch64ISD::TBZ";
case AArch64ISD::TBNZ: return "AArch64ISD::TBNZ";
case AArch64ISD::TC_RETURN: return "AArch64ISD::TC_RETURN";
case AArch64ISD::PREFETCH: return "AArch64ISD::PREFETCH";
case AArch64ISD::SITOF: return "AArch64ISD::SITOF";
case AArch64ISD::UITOF: return "AArch64ISD::UITOF";
case AArch64ISD::NVCAST: return "AArch64ISD::NVCAST";
case AArch64ISD::SQSHL_I: return "AArch64ISD::SQSHL_I";
case AArch64ISD::UQSHL_I: return "AArch64ISD::UQSHL_I";
case AArch64ISD::SRSHR_I: return "AArch64ISD::SRSHR_I";
case AArch64ISD::URSHR_I: return "AArch64ISD::URSHR_I";
case AArch64ISD::SQSHLU_I: return "AArch64ISD::SQSHLU_I";
case AArch64ISD::WrapperLarge: return "AArch64ISD::WrapperLarge";
case AArch64ISD::LD2post: return "AArch64ISD::LD2post";
case AArch64ISD::LD3post: return "AArch64ISD::LD3post";
case AArch64ISD::LD4post: return "AArch64ISD::LD4post";
case AArch64ISD::ST2post: return "AArch64ISD::ST2post";
case AArch64ISD::ST3post: return "AArch64ISD::ST3post";
case AArch64ISD::ST4post: return "AArch64ISD::ST4post";
case AArch64ISD::LD1x2post: return "AArch64ISD::LD1x2post";
case AArch64ISD::LD1x3post: return "AArch64ISD::LD1x3post";
case AArch64ISD::LD1x4post: return "AArch64ISD::LD1x4post";
case AArch64ISD::ST1x2post: return "AArch64ISD::ST1x2post";
case AArch64ISD::ST1x3post: return "AArch64ISD::ST1x3post";
case AArch64ISD::ST1x4post: return "AArch64ISD::ST1x4post";
case AArch64ISD::LD1DUPpost: return "AArch64ISD::LD1DUPpost";
case AArch64ISD::LD2DUPpost: return "AArch64ISD::LD2DUPpost";
case AArch64ISD::LD3DUPpost: return "AArch64ISD::LD3DUPpost";
case AArch64ISD::LD4DUPpost: return "AArch64ISD::LD4DUPpost";
case AArch64ISD::LD1LANEpost: return "AArch64ISD::LD1LANEpost";
case AArch64ISD::LD2LANEpost: return "AArch64ISD::LD2LANEpost";
case AArch64ISD::LD3LANEpost: return "AArch64ISD::LD3LANEpost";
case AArch64ISD::LD4LANEpost: return "AArch64ISD::LD4LANEpost";
case AArch64ISD::ST2LANEpost: return "AArch64ISD::ST2LANEpost";
case AArch64ISD::ST3LANEpost: return "AArch64ISD::ST3LANEpost";
case AArch64ISD::ST4LANEpost: return "AArch64ISD::ST4LANEpost";
case AArch64ISD::SMULL: return "AArch64ISD::SMULL";
case AArch64ISD::UMULL: return "AArch64ISD::UMULL";
case AArch64ISD::FRECPE: return "AArch64ISD::FRECPE";
case AArch64ISD::FRECPS: return "AArch64ISD::FRECPS";
case AArch64ISD::FRSQRTE: return "AArch64ISD::FRSQRTE";
case AArch64ISD::FRSQRTS: return "AArch64ISD::FRSQRTS";
case AArch64ISD::STG: return "AArch64ISD::STG";
case AArch64ISD::STZG: return "AArch64ISD::STZG";
case AArch64ISD::ST2G: return "AArch64ISD::ST2G";
case AArch64ISD::STZ2G: return "AArch64ISD::STZ2G";
case AArch64ISD::SUNPKHI: return "AArch64ISD::SUNPKHI";
case AArch64ISD::SUNPKLO: return "AArch64ISD::SUNPKLO";
case AArch64ISD::UUNPKHI: return "AArch64ISD::UUNPKHI";
case AArch64ISD::UUNPKLO: return "AArch64ISD::UUNPKLO";
case AArch64ISD::INSR: return "AArch64ISD::INSR";
case AArch64ISD::PTEST: return "AArch64ISD::PTEST";
case AArch64ISD::PTRUE: return "AArch64ISD::PTRUE";
case AArch64ISD::GLD1: return "AArch64ISD::GLD1";
case AArch64ISD::GLD1_SCALED: return "AArch64ISD::GLD1_SCALED";
case AArch64ISD::GLD1_SXTW: return "AArch64ISD::GLD1_SXTW";
case AArch64ISD::GLD1_UXTW: return "AArch64ISD::GLD1_UXTW";
case AArch64ISD::GLD1_SXTW_SCALED: return "AArch64ISD::GLD1_SXTW_SCALED";
case AArch64ISD::GLD1_UXTW_SCALED: return "AArch64ISD::GLD1_UXTW_SCALED";
case AArch64ISD::GLD1_IMM: return "AArch64ISD::GLD1_IMM";
case AArch64ISD::GLD1S: return "AArch64ISD::GLD1S";
case AArch64ISD::GLD1S_SCALED: return "AArch64ISD::GLD1S_SCALED";
case AArch64ISD::GLD1S_SXTW: return "AArch64ISD::GLD1S_SXTW";
case AArch64ISD::GLD1S_UXTW: return "AArch64ISD::GLD1S_UXTW";
case AArch64ISD::GLD1S_IMM: return "AArch64ISD::GLD1S_IMM";
case AArch64ISD::SST1: return "AArch64ISD::SST1";
case AArch64ISD::SST1_SCALED: return "AArch64ISD::SST1_SCALED";
case AArch64ISD::SST1_SXTW: return "AArch64ISD::SST1_SXTW";
case AArch64ISD::SST1_UXTW: return "AArch64ISD::SST1_UXTW";
case AArch64ISD::SST1_SXTW_SCALED: return "AArch64ISD::SST1_SXTW_SCALED";
case AArch64ISD::SST1_UXTW_SCALED: return "AArch64ISD::SST1_UXTW_SCALED";
case AArch64ISD::SST1_IMM: return "AArch64ISD::SST1_IMM";
case AArch64ISD::LDP: return "AArch64ISD::LDP";
case AArch64ISD::STP: return "AArch64ISD::STP";
return nullptr;
MachineBasicBlock *
AArch64TargetLowering::EmitF128CSEL(MachineInstr &MI,
MachineBasicBlock *MBB) const {
// We materialise the F128CSEL pseudo-instruction as some control flow and a
// phi node:
// OrigBB:
// [... previous instrs leading to comparison ...]
// TrueBB
// b EndBB
// TrueBB:
// ; Fallthrough
// EndBB:
// Dest = PHI [IfTrue, TrueBB], [IfFalse, OrigBB]
MachineFunction *MF = MBB->getParent();
const TargetInstrInfo *TII = Subtarget->getInstrInfo();
const BasicBlock *LLVM_BB = MBB->getBasicBlock();
DebugLoc DL = MI.getDebugLoc();
MachineFunction::iterator It = ++MBB->getIterator();
Register DestReg = MI.getOperand(0).getReg();
Register IfTrueReg = MI.getOperand(1).getReg();
Register IfFalseReg = MI.getOperand(2).getReg();
unsigned CondCode = MI.getOperand(3).getImm();
bool NZCVKilled = MI.getOperand(4).isKill();
MachineBasicBlock *TrueBB = MF->CreateMachineBasicBlock(LLVM_BB);
MachineBasicBlock *EndBB = MF->CreateMachineBasicBlock(LLVM_BB);
MF->insert(It, TrueBB);
MF->insert(It, EndBB);
// Transfer rest of current basic-block to EndBB
EndBB->splice(EndBB->begin(), MBB, std::next(MachineBasicBlock::iterator(MI)),
BuildMI(MBB, DL, TII->get(AArch64::Bcc)).addImm(CondCode).addMBB(TrueBB);
BuildMI(MBB, DL, TII->get(AArch64::B)).addMBB(EndBB);
// TrueBB falls through to the end.
if (!NZCVKilled) {
BuildMI(*EndBB, EndBB->begin(), DL, TII->get(AArch64::PHI), DestReg)
return EndBB;
MachineBasicBlock *AArch64TargetLowering::EmitLoweredCatchRet(
MachineInstr &MI, MachineBasicBlock *BB) const {
BB->getParent()->getFunction().getPersonalityFn())) &&
"SEH does not use catchret!");
return BB;
MachineBasicBlock *AArch64TargetLowering::EmitLoweredCatchPad(
MachineInstr &MI, MachineBasicBlock *BB) const {
return BB;
MachineBasicBlock *AArch64TargetLowering::EmitInstrWithCustomInserter(
MachineInstr &MI, MachineBasicBlock *BB) const {
switch (MI.getOpcode()) {
#ifndef NDEBUG
llvm_unreachable("Unexpected instruction for custom inserter!");
case AArch64::F128CSEL:
return EmitF128CSEL(MI, BB);
case TargetOpcode::STACKMAP:
case TargetOpcode::PATCHPOINT:
return emitPatchPoint(MI, BB);
case AArch64::CATCHRET:
return EmitLoweredCatchRet(MI, BB);
case AArch64::CATCHPAD:
return EmitLoweredCatchPad(MI, BB);
// AArch64 Lowering private implementation.
// Lowering Code
/// changeIntCCToAArch64CC - Convert a DAG integer condition code to an AArch64
/// CC
static AArch64CC::CondCode changeIntCCToAArch64CC(ISD::CondCode CC) {
switch (CC) {
llvm_unreachable("Unknown condition code!");
case ISD::SETNE:
return AArch64CC::NE;
case ISD::SETEQ:
return AArch64CC::EQ;
case ISD::SETGT:
return AArch64CC::GT;
case ISD::SETGE:
return AArch64CC::GE;
case ISD::SETLT:
return AArch64CC::LT;
case ISD::SETLE:
return AArch64CC::LE;
return AArch64CC::HI;
return AArch64CC::HS;
return AArch64CC::LO;
return AArch64CC::LS;
/// changeFPCCToAArch64CC - Convert a DAG fp condition code to an AArch64 CC.
static void changeFPCCToAArch64CC(ISD::CondCode CC,
AArch64CC::CondCode &CondCode,
AArch64CC::CondCode &CondCode2) {
CondCode2 = AArch64CC::AL;
switch (CC) {
llvm_unreachable("Unknown FP condition!");
case ISD::SETEQ:
CondCode = AArch64CC::EQ;
case ISD::SETGT:
CondCode = AArch64CC::GT;
case ISD::SETGE:
CondCode = AArch64CC::GE;
CondCode = AArch64CC::MI;
CondCode = AArch64CC::LS;
CondCode = AArch64CC::MI;
CondCode2 = AArch64CC::GT;
case ISD::SETO:
CondCode = AArch64CC::VC;
case ISD::SETUO:
CondCode = AArch64CC::VS;
CondCode = AArch64CC::EQ;
CondCode2 = AArch64CC::VS;
CondCode = AArch64CC::HI;
CondCode = AArch64CC::PL;
case ISD::SETLT:
CondCode = AArch64CC::LT;
case ISD::SETLE:
CondCode = AArch64CC::LE;
case ISD::SETNE:
CondCode = AArch64CC::NE;
/// Convert a DAG fp condition code to an AArch64 CC.
/// This differs from changeFPCCToAArch64CC in that it returns cond codes that
/// should be AND'ed instead of OR'ed.
static void changeFPCCToANDAArch64CC(ISD::CondCode CC,
AArch64CC::CondCode &CondCode,
AArch64CC::CondCode &CondCode2) {
CondCode2 = AArch64CC::AL;
switch (CC) {
changeFPCCToAArch64CC(CC, CondCode, CondCode2);
assert(CondCode2 == AArch64CC::AL);
// (a one b)
// == ((a olt b) || (a ogt b))
// == ((a ord b) && (a une b))
CondCode = AArch64CC::VC;
CondCode2 = AArch64CC::NE;
// (a ueq b)
// == ((a uno b) || (a oeq b))
// == ((a ule b) && (a uge b))
CondCode = AArch64CC::PL;
CondCode2 = AArch64CC::LE;
/// changeVectorFPCCToAArch64CC - Convert a DAG fp condition code to an AArch64
/// CC usable with the vector instructions. Fewer operations are available
/// without a real NZCV register, so we have to use less efficient combinations
/// to get the same effect.
static void changeVectorFPCCToAArch64CC(ISD::CondCode CC,
AArch64CC::CondCode &CondCode,
AArch64CC::CondCode &CondCode2,
bool &Invert) {
Invert = false;
switch (CC) {
// Mostly the scalar mappings work fine.
changeFPCCToAArch64CC(CC, CondCode, CondCode2);
case ISD::SETUO:
Invert = true;
case ISD::SETO:
CondCode = AArch64CC::MI;
CondCode2 = AArch64CC::GE;
// All of the compare-mask comparisons are ordered, but we can switch
// between the two by a double inversion. E.g. ULE == !OGT.
Invert = true;
changeFPCCToAArch64CC(getSetCCInverse(CC, /* FP inverse */ MVT::f32),
CondCode, CondCode2);
static bool isLegalArithImmed(uint64_t C) {
// Matches AArch64DAGToDAGISel::SelectArithImmed().
bool IsLegal = (C >> 12 == 0) || ((C & 0xFFFULL) == 0 && C >> 24 == 0);
LLVM_DEBUG(dbgs() << "Is imm " << C
<< " legal: " << (IsLegal ? "yes\n" : "no\n"));
return IsLegal;
// Can a (CMP op1, (sub 0, op2) be turned into a CMN instruction on
// the grounds that "op1 - (-op2) == op1 + op2" ? Not always, the C and V flags
// can be set differently by this operation. It comes down to whether
// "SInt(~op2)+1 == SInt(~op2+1)" (and the same for UInt). If they are then
// everything is fine. If not then the optimization is wrong. Thus general
// comparisons are only valid if op2 != 0.
// So, finally, the only LLVM-native comparisons that don't mention C and V
// are SETEQ and SETNE. They're the only ones we can safely use CMN for in
// the absence of information about op2.
static bool isCMN(SDValue Op, ISD::CondCode CC) {
return Op.getOpcode() == ISD::SUB && isNullConstant(Op.getOperand(0)) &&
static SDValue emitStrictFPComparison(SDValue LHS, SDValue RHS, const SDLoc &dl,
SelectionDAG &DAG, SDValue Chain,
bool IsSignaling) {
EVT VT = LHS.getValueType();
assert(VT != MVT::f128);
assert(VT != MVT::f16 && "Lowering of strict fp16 not yet implemented");
unsigned Opcode =
return DAG.getNode(Opcode, dl, {VT, MVT::Other}, {Chain, LHS, RHS});
static SDValue emitComparison(SDValue LHS, SDValue RHS, ISD::CondCode CC,
const SDLoc &dl, SelectionDAG &DAG) {
EVT VT = LHS.getValueType();
const bool FullFP16 =
static_cast<const AArch64Subtarget &>(DAG.getSubtarget()).hasFullFP16();
if (VT.isFloatingPoint()) {
assert(VT != MVT::f128);
if (VT == MVT::f16 && !FullFP16) {
LHS = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f32, LHS);
RHS = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f32, RHS);
VT = MVT::f32;
return DAG.getNode(AArch64ISD::FCMP, dl, VT, LHS, RHS);
// The CMP instruction is just an alias for SUBS, and representing it as
// SUBS means that it's possible to get CSE with subtract operations.
// A later phase can perform the optimization of setting the destination
// register to WZR/XZR if it ends up being unused.
unsigned Opcode = AArch64ISD::SUBS;
if (isCMN(RHS, CC)) {
// Can we combine a (CMP op1, (sub 0, op2) into a CMN instruction ?
Opcode = AArch64ISD::ADDS;
RHS = RHS.getOperand(1);
} else if (isCMN(LHS, CC)) {
// As we are looking for EQ/NE compares, the operands can be commuted ; can
// we combine a (CMP (sub 0, op1), op2) into a CMN instruction ?
Opcode = AArch64ISD::ADDS;
LHS = LHS.getOperand(1);
} else if (LHS.getOpcode() == ISD::AND && isNullConstant(RHS) &&
!isUnsignedIntSetCC(CC)) {
// Similarly, (CMP (and X, Y), 0) can be implemented with a TST
// (a.k.a. ANDS) except that the flags are only guaranteed to work for one
// of the signed comparisons.
Opcode = AArch64ISD::ANDS;
RHS = LHS.getOperand(1);
LHS = LHS.getOperand(0);
return DAG.getNode(Opcode, dl, DAG.getVTList(VT, MVT_CC), LHS, RHS)
/// \defgroup AArch64CCMP CMP;CCMP matching
/// These functions deal with the formation of CMP;CCMP;... sequences.
/// The CCMP/CCMN/FCCMP/FCCMPE instructions allow the conditional execution of
/// a comparison. They set the NZCV flags to a predefined value if their
/// predicate is false. This allows to express arbitrary conjunctions, for
/// example "cmp 0 (and (setCA (cmp A)) (setCB (cmp B)))"
/// expressed as:
/// cmp A
/// ccmp B, inv(CB), CA
/// check for CB flags
/// This naturally lets us implement chains of AND operations with SETCC
/// operands. And we can even implement some other situations by transforming
/// them:
/// - We can implement (NEG SETCC) i.e. negating a single comparison by
/// negating the flags used in a CCMP/FCCMP operations.
/// - We can negate the result of a whole chain of CMP/CCMP/FCCMP operations
/// by negating the flags we test for afterwards. i.e.
/// NEG (CMP CCMP CCCMP ...) can be implemented.
/// - Note that we can only ever negate all previously processed results.
/// What we can not implement by flipping the flags to test is a negation
/// of two sub-trees (because the negation affects all sub-trees emitted so
/// far, so the 2nd sub-tree we emit would also affect the first).
/// With those tools we can implement some OR operations:
/// - (OR (SETCC A) (SETCC B)) can be implemented via:
/// - After transforming OR to NEG/AND combinations we may be able to use NEG
/// elimination rules from earlier to implement the whole thing as a
/// CCMP/FCCMP chain.
/// As complete example:
/// or (or (setCA (cmp A)) (setCB (cmp B)))
/// (and (setCC (cmp C)) (setCD (cmp D)))"
/// can be reassociated to:
/// or (and (setCC (cmp C)) setCD (cmp D))
// (or (setCA (cmp A)) (setCB (cmp B)))
/// can be transformed to:
/// not (and (not (and (setCC (cmp C)) (setCD (cmp D))))
/// (and (not (setCA (cmp A)) (not (setCB (cmp B))))))"
/// which can be implemented as:
/// cmp C
/// ccmp D, inv(CD), CC
/// ccmp A, CA, inv(CD)
/// ccmp B, CB, inv(CA)
/// check for CB flags
/// A counterexample is "or (and A B) (and C D)" which translates to
/// not (and (not (and (not A) (not B))) (not (and (not C) (not D)))), we
/// can only implement 1 of the inner (not) operations, but not both!
/// @{
/// Create a conditional comparison; Use CCMP, CCMN or FCCMP as appropriate.
static SDValue emitConditionalComparison(SDValue LHS, SDValue RHS,
ISD::CondCode CC, SDValue CCOp,
AArch64CC::CondCode Predicate,
AArch64CC::CondCode OutCC,
const SDLoc &DL, SelectionDAG &DAG) {
unsigned Opcode = 0;
const bool FullFP16 =
static_cast<const AArch64Subtarget &>(DAG.getSubtarget()).hasFullFP16();
if (LHS.getValueType().isFloatingPoint()) {
assert(LHS.getValueType() != MVT::f128);
if (LHS.getValueType() == MVT::f16 && !FullFP16) {
LHS = DAG.getNode(ISD::FP_EXTEND, DL, MVT::f32, LHS);
RHS = DAG.getNode(ISD::FP_EXTEND, DL, MVT::f32, RHS);
Opcode = AArch64ISD::FCCMP;
} else if (RHS.getOpcode() == ISD::SUB) {
SDValue SubOp0 = RHS.getOperand(0);
if (isNullConstant(SubOp0) && (CC == ISD::SETEQ || CC == ISD::SETNE)) {
// See emitComparison() on why we can only do this for SETEQ and SETNE.
Opcode = AArch64ISD::CCMN;
RHS = RHS.getOperand(1);
if (Opcode == 0)
Opcode = AArch64ISD::CCMP;
SDValue Condition = DAG.getConstant(Predicate, DL, MVT_CC);
AArch64CC::CondCode InvOutCC = AArch64CC::getInvertedCondCode(OutCC);
unsigned NZCV = AArch64CC::getNZCVToSatisfyCondCode(InvOutCC);
SDValue NZCVOp = DAG.getConstant(NZCV, DL, MVT::i32);
return DAG.getNode(Opcode, DL, MVT_CC, LHS, RHS, NZCVOp, Condition, CCOp);
/// Returns true if @p Val is a tree of AND/OR/SETCC operations that can be
/// expressed as a conjunction. See \ref AArch64CCMP.
/// \param CanNegate Set to true if we can negate the whole sub-tree just by
/// changing the conditions on the SETCC tests.
/// (this means we can call emitConjunctionRec() with
/// Negate==true on this sub-tree)
/// \param MustBeFirst Set to true if this subtree needs to be negated and we
/// cannot do the negation naturally. We are required to
/// emit the subtree first in this case.
/// \param WillNegate Is true if are called when the result of this
/// subexpression must be negated. This happens when the
/// outer expression is an OR. We can use this fact to know
/// that we have a double negation (or (or ...) ...) that
/// can be implemented for free.
static bool canEmitConjunction(const SDValue Val, bool &CanNegate,
bool &MustBeFirst, bool WillNegate,
unsigned Depth = 0) {
if (!Val.hasOneUse())
return false;
unsigned Opcode = Val->getOpcode();
if (Opcode == ISD::SETCC) {
if (Val->getOperand(0).getValueType() == MVT::f128)
return false;
CanNegate = true;
MustBeFirst = false;
return true;
// Protect against exponential runtime and stack overflow.
if (Depth > 6)
return false;
if (Opcode == ISD::AND || Opcode == ISD::OR) {
bool IsOR = Opcode == ISD::OR;
SDValue O0 = Val->getOperand(0);
SDValue O1 = Val->getOperand(1);
bool CanNegateL;
bool MustBeFirstL;
if (!canEmitConjunction(O0, CanNegateL, MustBeFirstL, IsOR, Depth+1))
return false;
bool CanNegateR;
bool MustBeFirstR;
if (!canEmitConjunction(O1, CanNegateR, MustBeFirstR, IsOR, Depth+1))
return false;
if (MustBeFirstL && MustBeFirstR)
return false;
if (IsOR) {
// For an OR expression we need to be able to naturally negate at least
// one side or we cannot do the transformation at all.
if (!CanNegateL && !CanNegateR)
return false;
// If we the result of the OR will be negated and we can naturally negate
// the leafs, then this sub-tree as a whole negates naturally.
CanNegate = WillNegate && CanNegateL && CanNegateR;
// If we cannot naturally negate the whole sub-tree, then this must be
// emitted first.
MustBeFirst = !CanNegate;
} else {
assert(Opcode == ISD::AND && "Must be OR or AND");
// We cannot naturally negate an AND operation.
CanNegate = false;
MustBeFirst = MustBeFirstL || MustBeFirstR;
return true;
return false;
/// Emit conjunction or disjunction tree with the CMP/FCMP followed by a chain
/// of CCMP/CFCMP ops. See @ref AArch64CCMP.
/// Tries to transform the given i1 producing node @p Val to a series compare
/// and conditional compare operations. @returns an NZCV flags producing node
/// and sets @p OutCC to the flags that should be tested or returns SDValue() if
/// transformation was not possible.
/// \p Negate is true if we want this sub-tree being negated just by changing
/// SETCC conditions.
static SDValue emitConjunctionRec(SelectionDAG &DAG, SDValue Val,
AArch64CC::CondCode &OutCC, bool Negate, SDValue CCOp,
AArch64CC::CondCode Predicate) {
// We're at a tree leaf, produce a conditional comparison operation.
unsigned Opcode = Val->getOpcode();
if (Opcode == ISD::SETCC) {
SDValue LHS = Val->getOperand(0);
SDValue RHS = Val->getOperand(1);
ISD::CondCode CC = cast<CondCodeSDNode>(Val->getOperand(2))->get();
bool isInteger = LHS.getValueType().isInteger();
if (Negate)
CC = getSetCCInverse(CC, LHS.getValueType());
SDLoc DL(Val);
// Determine OutCC and handle FP special case.
if (isInteger) {
OutCC = changeIntCCToAArch64CC(CC);
} else {
AArch64CC::CondCode ExtraCC;
changeFPCCToANDAArch64CC(CC, OutCC, ExtraCC);
// Some floating point conditions can't be tested with a single condition
// code. Construct an additional comparison in this case.
if (ExtraCC != AArch64CC::AL) {
SDValue ExtraCmp;
if (!CCOp.getNode())
ExtraCmp = emitComparison(LHS, RHS, CC, DL, DAG);
ExtraCmp = emitConditionalComparison(LHS, RHS, CC, CCOp, Predicate,
ExtraCC, DL, DAG);
CCOp = ExtraCmp;
Predicate = ExtraCC;
// Produce a normal comparison if we are first in the chain
if (!CCOp)
return emitComparison(LHS, RHS, CC, DL, DAG);
// Otherwise produce a ccmp.
return emitConditionalComparison(LHS, RHS, CC, CCOp, Predicate, OutCC, DL,
assert(Val->hasOneUse() && "Valid conjunction/disjunction tree");
bool IsOR = Opcode == ISD::OR;
SDValue LHS = Val->getOperand(0);
bool CanNegateL;
bool MustBeFirstL;
bool ValidL = canEmitConjunction(LHS, CanNegateL, MustBeFirstL, IsOR);
assert(ValidL && "Valid conjunction/disjunction tree");
SDValue RHS = Val->getOperand(1);
bool CanNegateR;
bool MustBeFirstR;
bool ValidR = canEmitConjunction(RHS, CanNegateR, MustBeFirstR, IsOR);
assert(ValidR && "Valid conjunction/disjunction tree");
// Swap sub-tree that must come first to the right side.
if (MustBeFirstL) {
assert(!MustBeFirstR && "Valid conjunction/disjunction tree");
std::swap(LHS, RHS);
std::swap(CanNegateL, CanNegateR);
std::swap(MustBeFirstL, MustBeFirstR);
bool NegateR;
bool NegateAfterR;
bool NegateL;
bool NegateAfterAll;
if (Opcode == ISD::OR) {
// Swap the sub-tree that we can negate naturally to the left.
if (!CanNegateL) {
assert(CanNegateR && "at least one side must be negatable");
assert(!MustBeFirstR && "invalid conjunction/disjunction tree");
std::swap(LHS, RHS);
NegateR = false;
NegateAfterR = true;
} else {
// Negate the left sub-tree if possible, otherwise negate the result.
NegateR = CanNegateR;
NegateAfterR = !CanNegateR;
NegateL = true;
NegateAfterAll = !Negate;
} else {
assert(Opcode == ISD::AND && "Valid conjunction/disjunction tree");
assert(!Negate && "Valid conjunction/disjunction tree");
NegateL = false;
NegateR = false;
NegateAfterR = false;
NegateAfterAll = false;
// Emit sub-trees.
AArch64CC::CondCode RHSCC;
SDValue CmpR = emitConjunctionRec(DAG, RHS, RHSCC, NegateR, CCOp, Predicate);
if (NegateAfterR)
RHSCC = AArch64CC::getInvertedCondCode(RHSCC);
SDValue CmpL = emitConjunctionRec(DAG, LHS, OutCC, NegateL, CmpR, RHSCC);
if (NegateAfterAll)
OutCC = AArch64CC::getInvertedCondCode(OutCC);
return CmpL;
/// Emit expression as a conjunction (a series of CCMP/CFCMP ops).
/// In some cases this is even possible with OR operations in the expression.
/// See \ref AArch64CCMP.
/// \see emitConjunctionRec().
static SDValue emitConjunction(SelectionDAG &DAG, SDValue Val,
AArch64CC::CondCode &OutCC) {
bool DummyCanNegate;
bool DummyMustBeFirst;
if (!canEmitConjunction(Val, DummyCanNegate, DummyMustBeFirst, false))
return SDValue();
return emitConjunctionRec(DAG, Val, OutCC, false, SDValue(), AArch64CC::AL);
/// @}
/// Returns how profitable it is to fold a comparison's operand's shift and/or
/// extension operations.
static unsigned getCmpOperandFoldingProfit(SDValue Op) {
auto isSupportedExtend = [&](SDValue V) {
if (V.getOpcode() == ISD::SIGN_EXTEND_INREG)
return true;
if (V.getOpcode() == ISD::AND)
if (ConstantSDNode *MaskCst = dyn_cast<ConstantSDNode>(V.getOperand(1))) {
uint64_t Mask = MaskCst->getZExtValue();
return (Mask == 0xFF || Mask == 0xFFFF || Mask == 0xFFFFFFFF);
return false;
if (!Op.hasOneUse())
return 0;
if (isSupportedExtend(Op))
return 1;
unsigned Opc = Op.getOpcode();
if (Opc == ISD::SHL || Opc == ISD::SRL || Opc == ISD::SRA)
if (ConstantSDNode *ShiftCst = dyn_cast<ConstantSDNode>(Op.getOperand(1))) {
uint64_t Shift = ShiftCst->getZExtValue();
if (isSupportedExtend(Op.getOperand(0)))
return (Shift <= 4) ? 2 : 1;
EVT VT = Op.getValueType();
if ((VT == MVT::i32 && Shift <= 31) || (VT == MVT::i64 && Shift <= 63))
return 1;
return 0;
static SDValue getAArch64Cmp(SDValue LHS, SDValue RHS, ISD::CondCode CC,
SDValue &AArch64cc, SelectionDAG &DAG,
const SDLoc &dl) {
if (ConstantSDNode *RHSC = dyn_cast<ConstantSDNode>(RHS.getNode())) {
EVT VT = RHS.getValueType();
uint64_t C = RHSC->getZExtValue();
if (!isLegalArithImmed(C)) {
// Constant does not fit, try adjusting it by one?
switch (CC) {
case ISD::SETLT:
case ISD::SETGE:
if ((VT == MVT::i32 && C != 0x80000000 &&
isLegalArithImmed((uint32_t)(C - 1))) ||
(VT == MVT::i64 && C != 0x80000000ULL &&
isLegalArithImmed(C - 1ULL))) {
C = (VT == MVT::i32) ? (uint32_t)(C - 1) : C - 1;
RHS = DAG.getConstant(C, dl, VT);
if ((VT == MVT::i32 && C != 0 &&
isLegalArithImmed((uint32_t)(C - 1))) ||
(VT == MVT::i64 && C != 0ULL && isLegalArithImmed(C - 1ULL))) {
C = (VT == MVT::i32) ? (uint32_t)(C - 1) : C - 1;
RHS = DAG.getConstant(C, dl, VT);
case ISD::SETLE:
case ISD::SETGT:
if ((VT == MVT::i32 && C != INT32_MAX &&
isLegalArithImmed((uint32_t)(C + 1))) ||
(VT == MVT::i64 && C != INT64_MAX &&
isLegalArithImmed(C + 1ULL))) {
C = (VT == MVT::i32) ? (uint32_t)(C + 1) : C + 1;
RHS = DAG.getConstant(C, dl, VT);
if ((VT == MVT::i32 && C != UINT32_MAX &&
isLegalArithImmed((uint32_t)(C + 1))) ||
(VT == MVT::i64 && C != UINT64_MAX &&
isLegalArithImmed(C + 1ULL))) {
C = (VT == MVT::i32) ? (uint32_t)(C + 1) : C + 1;
RHS = DAG.getConstant(C, dl, VT);
// Comparisons are canonicalized so that the RHS operand is simpler than the
// LHS one, the extreme case being when RHS is an immediate. However, AArch64
// can fold some shift+extend operations on the RHS operand, so swap the
// operands if that can be done.
// For example:
// lsl w13, w11, #1
// cmp w13, w12
// can be turned into:
// cmp w12, w11, lsl #1
if (!isa<ConstantSDNode>(RHS) ||
!isLegalArithImmed(cast<ConstantSDNode>(RHS)->getZExtValue())) {
SDValue TheLHS = isCMN(LHS, CC) ? LHS.getOperand(1) : LHS;
if (getCmpOperandFoldingProfit(TheLHS) > getCmpOperandFoldingProfit(RHS)) {
std::swap(LHS, RHS);
CC = ISD::getSetCCSwappedOperands(CC);
SDValue Cmp;
AArch64CC::CondCode AArch64CC;
if ((CC == ISD::SETEQ || CC == ISD::SETNE) && isa<ConstantSDNode>(RHS)) {
const ConstantSDNode *RHSC = cast<ConstantSDNode>(RHS);
// The imm operand of ADDS is an unsigned immediate, in the range 0 to 4095.
// For the i8 operand, the largest immediate is 255, so this can be easily
// encoded in the compare instruction. For the i16 operand, however, the
// largest immediate cannot be encoded in the compare.
// Therefore, use a sign extending load and cmn to avoid materializing the
// -1 constant. For example,
// movz w1, #65535
// ldrh w0, [x0, #0]
// cmp w0, w1
// >
// ldrsh w0, [x0, #0]
// cmn w0, #1
// Fundamental, we're relying on the property that (zext LHS) == (zext RHS)
// if and only if (sext LHS) == (sext RHS). The checks are in place to
// ensure both the LHS and RHS are truly zero extended and to make sure the
// transformation is profitable.
if ((RHSC->getZExtValue() >> 16 == 0) && isa<LoadSDNode>(LHS) &&
cast<LoadSDNode>(LHS)->getExtensionType() == ISD::ZEXTLOAD &&
cast<LoadSDNode>(LHS)->getMemoryVT() == MVT::i16 &&
LHS.getNode()->hasNUsesOfValue(1, 0)) {
int16_t ValueofRHS = cast<ConstantSDNode>(RHS)->getZExtValue();
if (ValueofRHS < 0 && isLegalArithImmed(-ValueofRHS)) {
SDValue SExt =
DAG.getNode(ISD::SIGN_EXTEND_INREG, dl, LHS.getValueType(), LHS,
Cmp = emitComparison(SExt, DAG.getConstant(ValueofRHS, dl,
CC, dl, DAG);
AArch64CC = changeIntCCToAArch64CC(CC);
if (!Cmp && (RHSC->isNullValue() || RHSC->isOne())) {
if ((Cmp = emitConjunction(DAG, LHS, AArch64CC))) {
if ((CC == ISD::SETNE) ^ RHSC->isNullValue())
AArch64CC = AArch64CC::getInvertedCondCode(AArch64CC);
if (!Cmp) {
Cmp = emitComparison(LHS, RHS, CC, dl, DAG);
AArch64CC = changeIntCCToAArch64CC(CC);
AArch64cc = DAG.getConstant(AArch64CC, dl, MVT_CC);
return Cmp;
static std::pair<SDValue, SDValue>
getAArch64XALUOOp(AArch64CC::CondCode &CC, SDValue Op, SelectionDAG &DAG) {
assert((Op.getValueType() == MVT::i32 || Op.getValueType() == MVT::i64) &&
"Unsupported value type");
SDValue Value, Overflow;
SDLoc DL(Op);
SDValue LHS = Op.getOperand(0);
SDValue RHS = Op.getOperand(1);
unsigned Opc = 0;
switch (Op.getOpcode()) {
llvm_unreachable("Unknown overflow instruction!");
case ISD::SADDO:
Opc = AArch64ISD::ADDS;
CC = AArch64CC::VS;
case ISD::UADDO:
Opc = AArch64ISD::ADDS;
CC = AArch64CC::HS;
case ISD::SSUBO:
Opc = AArch64ISD::SUBS;
CC = AArch64CC::VS;
case ISD::USUBO:
Opc = AArch64ISD::SUBS;
CC = AArch64CC::LO;
// Multiply needs a little bit extra work.
case ISD::SMULO:
case ISD::UMULO: {
CC = AArch64CC::NE;
bool IsSigned = Op.getOpcode() == ISD::SMULO;
if (Op.getValueType() == MVT::i32) {
unsigned ExtendOpc = IsSigned ? ISD::SIGN_EXTEND : ISD::ZERO_EXTEND;
// For a 32 bit multiply with overflow check we want the instruction
// selector to generate a widening multiply (SMADDL/UMADDL). For that we
// need to generate the following pattern:
// (i64 add 0, (i64 mul (i64 sext|zext i32 %a), (i64 sext|zext i32 %b))
LHS = DAG.getNode(ExtendOpc, DL, MVT::i64, LHS);
RHS = DAG.getNode(ExtendOpc, DL, MVT::i64, RHS);
SDValue Mul = DAG.getNode(ISD::MUL, DL, MVT::i64, LHS, RHS);
SDValue Add = DAG.getNode(ISD::ADD, DL, MVT::i64, Mul,
DAG.getConstant(0, DL, MVT::i64));
// On AArch64 the upper 32 bits are always zero extended for a 32 bit
// operation. We need to clear out the upper 32 bits, because we used a
// widening multiply that wrote all 64 bits. In the end this should be a
// noop.
Value = DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, Add);
if (IsSigned) {
// The signed overflow check requires more than just a simple check for
// any bit set in the upper 32 bits of the result. These bits could be
// just the sign bits of a negative number. To perform the overflow
// check we have to arithmetic shift right the 32nd bit of the result by
// 31 bits. Then we compare the result to the upper 32 bits.
SDValue UpperBits = DAG.getNode(ISD::SRL, DL, MVT::i64, Add,
DAG.getConstant(32, DL, MVT::i64));
UpperBits = DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, UpperBits);
SDValue LowerBits = DAG.getNode(ISD::SRA, DL, MVT::i32, Value,
DAG.getConstant(31, DL, MVT::i64));
// It is important that LowerBits is last, otherwise the arithmetic
// shift will not be folded into the compare (SUBS).
SDVTList VTs = DAG.getVTList(MVT::i32, MVT::i32);
Overflow = DAG.getNode(AArch64ISD::SUBS, DL, VTs, UpperBits, LowerBits)
} else {
// The overflow check for unsigned multiply is easy. We only need to
// check if any of the upper 32 bits are set. This can be done with a
// CMP (shifted register). For that we need to generate the following
// pattern:
// (i64 AArch64ISD::SUBS i64 0, (i64 srl i64 %Mul, i64 32)
SDValue UpperBits = DAG.getNode(ISD::SRL, DL, MVT::i64, Mul,
DAG.getConstant(32, DL, MVT::i64));
SDVTList VTs = DAG.getVTList(MVT::i64, MVT::i32);
Overflow =
DAG.getNode(AArch64ISD::SUBS, DL, VTs,
DAG.getConstant(0, DL, MVT::i64),
assert(Op.getValueType() == MVT::i64 && "Expected an i64 value type");
// For the 64 bit multiply
Value = DAG.getNode(ISD::MUL, DL, MVT::i64, LHS, RHS);
if (IsSigned) {
SDValue UpperBits = DAG.getNode(ISD::MULHS, DL, MVT::i64, LHS, RHS);
SDValue LowerBits = DAG.getNode(ISD::SRA, DL, MVT::i64, Value,
DAG.getConstant(63, DL, MVT::i64));
// It is important that LowerBits is last, otherwise the arithmetic
// shift will not be folded into the compare (SUBS).
SDVTList VTs = DAG.getVTList(MVT::i64, MVT::i32);
Overflow = DAG.getNode(AArch64ISD::SUBS, DL, VTs, UpperBits, LowerBits)
} else {
SDValue UpperBits = DAG.getNode(ISD::MULHU, DL, MVT::i64, LHS, RHS);
SDVTList VTs = DAG.getVTList(MVT::i64, MVT::i32);
Overflow =
DAG.getNode(AArch64ISD::SUBS, DL, VTs,
DAG.getConstant(0, DL, MVT::i64),
} // switch (...)
if (Opc) {
SDVTList VTs = DAG.getVTList(Op->getValueType(0), MVT::i32);
// Emit the AArch64 operation with overflow check.
Value = DAG.getNode(Opc, DL, VTs, LHS, RHS);
Overflow = Value.getValue(1);
return std::make_pair(Value, Overflow);
SDValue AArch64TargetLowering::LowerF128Call(SDValue Op, SelectionDAG &DAG,
RTLIB::Libcall Call) const {
bool IsStrict = Op->isStrictFPOpcode();
unsigned Offset = IsStrict ? 1 : 0;
SDValue Chain = IsStrict ? Op.getOperand(0) : SDValue();
SmallVector<SDValue, 2> Ops(Op->op_begin() + Offset, Op->op_end());
MakeLibCallOptions CallOptions;
SDValue Result;
SDLoc dl(Op);
std::tie(Result, Chain) = makeLibCall(DAG, Call, Op.getValueType(), Ops,
CallOptions, dl, Chain);
return IsStrict ? DAG.getMergeValues({Result, Chain}, dl) : Result;
// Returns true if the given Op is the overflow flag result of an overflow
// intrinsic operation.
static bool isOverflowIntrOpRes(SDValue Op) {
unsigned Opc = Op.getOpcode();
return (Op.getResNo() == 1 &&
(Opc == ISD::SADDO || Opc == ISD::UADDO || Opc == ISD::SSUBO ||
Opc == ISD::USUBO || Opc == ISD::SMULO || Opc == ISD::UMULO));
static SDValue LowerXOR(SDValue Op, SelectionDAG &DAG) {
SDValue Sel = Op.getOperand(0);
SDValue Other = Op.getOperand(1);
SDLoc dl(Sel);
// If the operand is an overflow checking operation, invert the condition
// code and kill the Not operation. I.e., transform:
// (xor (overflow_op_bool, 1))
// -->
// (csel 1, 0, invert(cc), overflow_op_bool)
// ... which later gets transformed to just a cset instruction with an
// inverted condition code, rather than a cset + eor sequence.
if (isOneConstant(Other) && isOverflowIntrOpRes(Sel)) {
// Only lower legal XALUO ops.
if (!DAG.getTargetLoweringInfo().isTypeLegal(Sel->getValueType(0)))
return SDValue();
SDValue TVal = DAG.getConstant(1, dl, MVT::i32);
SDValue FVal = DAG.getConstant(0, dl, MVT::i32);
AArch64CC::CondCode CC;
SDValue Value, Overflow;
std::tie(Value, Overflow) = getAArch64XALUOOp(CC, Sel.getValue(0), DAG);
SDValue CCVal = DAG.getConstant(getInvertedCondCode(CC), dl, MVT::i32);
return DAG.getNode(AArch64ISD::CSEL, dl, Op.getValueType(), TVal, FVal,
CCVal, Overflow);
// If neither operand is a SELECT_CC, give up.
if (Sel.getOpcode() != ISD::SELECT_CC)
std::swap(Sel, Other);
if (Sel.getOpcode() != ISD::SELECT_CC)
return Op;
// The folding we want to perform is:
// (xor x, (select_cc a, b, cc, 0, -1) )
// -->
// (csel x, (xor x, -1), cc ...)
// The latter will get matched to a CSINV instruction.
ISD::CondCode CC = cast<CondCodeSDNode>(Sel.getOperand(4))->get();
SDValue LHS = Sel.getOperand(0);
SDValue RHS = Sel.getOperand(1);
SDValue TVal = Sel.getOperand(2);
SDValue FVal = Sel.getOperand(3);
// FIXME: This could be generalized to non-integer comparisons.
if (LHS.getValueType() != MVT::i32 && LHS.getValueType() != MVT::i64)
return Op;
ConstantSDNode *CFVal = dyn_cast<ConstantSDNode>(FVal);
ConstantSDNode *CTVal = dyn_cast<ConstantSDNode>(TVal);
// The values aren't constants, this isn't the pattern we're looking for.
if (!CFVal || !CTVal)
return Op;
// We can commute the SELECT_CC by inverting the condition. This
// might be needed to make this fit into a CSINV pattern.
if (CTVal->isAllOnesValue() && CFVal->isNullValue()) {
std::swap(TVal, FVal);
std::swap(CTVal, CFVal);
CC = ISD::getSetCCInverse(CC, LHS.getValueType());
// If the constants line up, perform the transform!
if (CTVal->isNullValue() && CFVal->isAllOnesValue()) {
SDValue CCVal;
SDValue Cmp = getAArch64Cmp(LHS, RHS, CC, CCVal, DAG, dl);
FVal = Other;
TVal = DAG.getNode(ISD::XOR, dl, Other.getValueType(), Other,
DAG.getConstant(-1ULL, dl, Other.getValueType()));
return DAG.getNode(AArch64ISD::CSEL, dl, Sel.getValueType(), FVal, TVal,
CCVal, Cmp);
return Op;
static SDValue LowerADDC_ADDE_SUBC_SUBE(SDValue Op, SelectionDAG &DAG) {
EVT VT = Op.getValueType();
// Let legalize expand this if it isn't a legal type yet.
if (!DAG.getTargetLoweringInfo().isTypeLegal(VT))
return SDValue();
SDVTList VTs = DAG.getVTList(VT, MVT::i32);
unsigned Opc;
bool ExtraOp = false;
switch (Op.getOpcode()) {
llvm_unreachable("Invalid code");
case ISD::ADDC:
Opc = AArch64ISD::ADDS;
case ISD::SUBC:
Opc = AArch64ISD::SUBS;
case ISD::ADDE: