#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
+#include "llvm/CodeGen/MachineJumpTableInfo.h"
#include "llvm/CodeGen/MachineModuleInfo.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/SelectionDAG.h"
{ RTLIB::ADD_F64, "__aeabi_dadd", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
{ RTLIB::DIV_F64, "__aeabi_ddiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
{ RTLIB::MUL_F64, "__aeabi_dmul", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- // FIXME: double __aeabi_drsub(double x, double y) (rsub)
{ RTLIB::SUB_F64, "__aeabi_dsub", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
// Double-precision floating-point comparison helper functions
// RTABI chapter 4.1.2, Table 3
- // FIXME: void __aeabi_cdcmpeq(double, double)
- // FIXME: void __aeabi_cdcmple(double, double)
- // FIXME: void __aeabi_cdrcmple(double, double)
{ RTLIB::OEQ_F64, "__aeabi_dcmpeq", CallingConv::ARM_AAPCS, ISD::SETNE },
{ RTLIB::UNE_F64, "__aeabi_dcmpeq", CallingConv::ARM_AAPCS, ISD::SETEQ },
{ RTLIB::OLT_F64, "__aeabi_dcmplt", CallingConv::ARM_AAPCS, ISD::SETNE },
{ RTLIB::ADD_F32, "__aeabi_fadd", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
{ RTLIB::DIV_F32, "__aeabi_fdiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
{ RTLIB::MUL_F32, "__aeabi_fmul", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- // FIXME: void __aeabi_frsub(float x, float y)
{ RTLIB::SUB_F32, "__aeabi_fsub", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
// Single-precision floating-point comparison helper functions
// RTABI chapter 4.1.2, Table 5
- // FIXME: void __aeabi_cfcmpeq(float, float)
- // FIXME: void __aeabi_cfcmple(float, float)
- // FIXME: void __aeabi_cfrcmple(float, float)
{ RTLIB::OEQ_F32, "__aeabi_fcmpeq", CallingConv::ARM_AAPCS, ISD::SETNE },
{ RTLIB::UNE_F32, "__aeabi_fcmpeq", CallingConv::ARM_AAPCS, ISD::SETEQ },
{ RTLIB::OLT_F32, "__aeabi_fcmplt", CallingConv::ARM_AAPCS, ISD::SETNE },
// Conversions between floating types.
// RTABI chapter 4.1.2, Table 7
{ RTLIB::FPROUND_F64_F32, "__aeabi_d2f", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- { RTLIB::FPEXT_F32_F64, "__aeabi_f2d", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- // FIXME: float __aeabi_f2f(short)
- // FIXME: float __aeabi_h2f_alt(short)
- // FIXME: short __aeabi_f2h(float)
- // FIXME: short __aeabi_f2h_alt(float)
{ RTLIB::FPROUND_F64_F16, "__aeabi_d2h", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- // FIXME: short __aeabi_d2h_alt(double)
+ { RTLIB::FPEXT_F32_F64, "__aeabi_f2d", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
// Integer to floating-point conversions.
// RTABI chapter 4.1.2, Table 8
// Long long helper functions
// RTABI chapter 4.2, Table 9
- { RTLIB::MUL_I64, "__aeabi_lmul", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- // FIXME: __aeabi_ldivmod is SDIVREM not SDIV; we should custom lower this
- { RTLIB::SDIV_I64, "__aeabi_ldivmod", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- { RTLIB::SDIVREM_I64, "__aeabi_ldivmod", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- // FIXME: __aeabi_uldivmod is UDIVREM not UDIV; we should custom lower this
- { RTLIB::UDIV_I64, "__aeabi_uldivmod", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- { RTLIB::UDIVREM_I64, "__aeabi_uldivmod", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- { RTLIB::SHL_I64, "__aeabi_llsl", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- { RTLIB::SRL_I64, "__aeabi_llsr", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- { RTLIB::SRA_I64, "__aeabi_lasr", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- // FIXME: int __aeabi_lcmp(long long, long long)
- // FIXME: int __aeabi_ulcmp(unsigned long long, unsigned long long)
+ { RTLIB::MUL_I64, "__aeabi_lmul", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+ { RTLIB::SHL_I64, "__aeabi_llsl", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+ { RTLIB::SRL_I64, "__aeabi_llsr", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+ { RTLIB::SRA_I64, "__aeabi_lasr", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
// Integer division functions
// RTABI chapter 4.3.1
- { RTLIB::SDIV_I8, "__aeabi_idiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- { RTLIB::SDIV_I16, "__aeabi_idiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- { RTLIB::SDIV_I32, "__aeabi_idiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- { RTLIB::UDIV_I8, "__aeabi_uidiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- { RTLIB::UDIV_I16, "__aeabi_uidiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- { RTLIB::UDIV_I32, "__aeabi_uidiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- { RTLIB::SDIVREM_I8, "__aeabi_idivmod", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- { RTLIB::SDIVREM_I16, "__aeabi_idivmod", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- { RTLIB::SDIVREM_I32, "__aeabi_idivmod", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- { RTLIB::UDIVREM_I8, "__aeabi_uidivmod", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- { RTLIB::UDIVREM_I16, "__aeabi_uidivmod", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- { RTLIB::UDIVREM_I32, "__aeabi_uidivmod", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+ { RTLIB::SDIV_I8, "__aeabi_idiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+ { RTLIB::SDIV_I16, "__aeabi_idiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+ { RTLIB::SDIV_I32, "__aeabi_idiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+ { RTLIB::SDIV_I64, "__aeabi_ldivmod", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+ { RTLIB::UDIV_I8, "__aeabi_uidiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+ { RTLIB::UDIV_I16, "__aeabi_uidiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+ { RTLIB::UDIV_I32, "__aeabi_uidiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
+ { RTLIB::UDIV_I64, "__aeabi_uldivmod", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
// Memory operations
// RTABI chapter 4.3.4
- // FIXME: void __aeabi_memcpy8(void *, const void *, size_t)
- // FIXME: void __aeabi_memcpy4(void *, const void *, size_t)
{ RTLIB::MEMCPY, "__aeabi_memcpy", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- // FIXME: void __aeabi_memmove8(void *, const void *, size_t)
- // FIXME: void __aeabi_memmove4(void *, const void *, size_t)
{ RTLIB::MEMMOVE, "__aeabi_memmove", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- // FIXME: void __aeabi_memset8(void *, size_t, int)
- // FIXME: void __aeabi_memset4(void *, size_t, int)
{ RTLIB::MEMSET, "__aeabi_memset", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
- // FIXME: void __aeabi_memclr8(void *, size_t)
- // FIXME: void __aeabi_memclr4(void *, size_t)
- // FIXME: void __aeabi_memclr(void *, size_t)
};
for (const auto &LC : LibraryCalls) {
if (LC.Cond != ISD::SETCC_INVALID)
setCmpLibcallCC(LC.Op, LC.Cond);
}
-
- setOperationAction(ISD::SDIVREM, MVT::i32, Custom);
- setOperationAction(ISD::UDIVREM, MVT::i32, Custom);
}
if (Subtarget->isTargetWindows()) {
}
// FIXME: Also set divmod for SREM on EABI
- setOperationAction(ISD::SREM, MVT::i32, Expand);
- setOperationAction(ISD::UREM, MVT::i32, Expand);
- if (!Subtarget->isTargetAEABI()) {
+ setOperationAction(ISD::SREM, MVT::i32, Expand);
+ setOperationAction(ISD::UREM, MVT::i32, Expand);
+ // Register based DivRem for AEABI (RTABI 4.2)
+ if (Subtarget->isTargetAEABI()) {
+ setLibcallName(RTLIB::SDIVREM_I8, "__aeabi_idivmod");
+ setLibcallName(RTLIB::SDIVREM_I16, "__aeabi_idivmod");
+ setLibcallName(RTLIB::SDIVREM_I32, "__aeabi_idivmod");
+ setLibcallName(RTLIB::SDIVREM_I64, "__aeabi_ldivmod");
+ setLibcallName(RTLIB::UDIVREM_I8, "__aeabi_uidivmod");
+ setLibcallName(RTLIB::UDIVREM_I16, "__aeabi_uidivmod");
+ setLibcallName(RTLIB::UDIVREM_I32, "__aeabi_uidivmod");
+ setLibcallName(RTLIB::UDIVREM_I64, "__aeabi_uldivmod");
+
+ setLibcallCallingConv(RTLIB::SDIVREM_I8, CallingConv::ARM_AAPCS);
+ setLibcallCallingConv(RTLIB::SDIVREM_I16, CallingConv::ARM_AAPCS);
+ setLibcallCallingConv(RTLIB::SDIVREM_I32, CallingConv::ARM_AAPCS);
+ setLibcallCallingConv(RTLIB::SDIVREM_I64, CallingConv::ARM_AAPCS);
+ setLibcallCallingConv(RTLIB::UDIVREM_I8, CallingConv::ARM_AAPCS);
+ setLibcallCallingConv(RTLIB::UDIVREM_I16, CallingConv::ARM_AAPCS);
+ setLibcallCallingConv(RTLIB::UDIVREM_I32, CallingConv::ARM_AAPCS);
+ setLibcallCallingConv(RTLIB::UDIVREM_I64, CallingConv::ARM_AAPCS);
+
+ setOperationAction(ISD::SDIVREM, MVT::i32, Custom);
+ setOperationAction(ISD::UDIVREM, MVT::i32, Custom);
+ } else {
setOperationAction(ISD::SDIVREM, MVT::i32, Expand);
setOperationAction(ISD::UDIVREM, MVT::i32, Expand);
}
setOperationAction(ISD::DYNAMIC_STACKALLOC, MVT::i32, Expand);
// ARMv6 Thumb1 (except for CPUs that support dmb / dsb) and earlier use
- // the default expansion.
- if (Subtarget->hasAnyDataBarrier() && !Subtarget->isThumb1Only()) {
+ // the default expansion. If we are targeting a single threaded system,
+ // then set them all for expand so we can lower them later into their
+ // non-atomic form.
+ if (TM.Options.ThreadModel == ThreadModel::Single)
+ setOperationAction(ISD::ATOMIC_FENCE, MVT::Other, Expand);
+ else if (Subtarget->hasAnyDataBarrier() && !Subtarget->isThumb1Only()) {
// ATOMIC_FENCE needs custom lowering; the others should have been expanded
// to ldrex/strex loops already.
setOperationAction(ISD::ATOMIC_FENCE, MVT::Other, Custom);
ConstantSDNode *OrdN = cast<ConstantSDNode>(Op.getOperand(1));
AtomicOrdering Ord = static_cast<AtomicOrdering>(OrdN->getZExtValue());
- unsigned Domain = ARM_MB::ISH;
+ ARM_MB::MemBOpt Domain = ARM_MB::ISH;
if (Subtarget->isMClass()) {
// Only a full system barrier exists in the M-class architectures.
Domain = ARM_MB::SY;
}
CCInfo.rewindByValRegsInfo();
lastInsIndex = -1;
- if (isVarArg) {
+ if (isVarArg && MFI->hasVAStart()) {
unsigned ExtraArgRegsSize;
unsigned ExtraArgRegsSaveSize;
computeRegArea(CCInfo, MF, CCInfo.getInRegsParamsCount(), 0,
}
// varargs
- if (isVarArg)
+ if (isVarArg && MFI->hasVAStart())
VarArgStyleRegisters(CCInfo, DAG, dl, Chain,
CCInfo.getNextStackOffset(),
TotalArgRegsSaveSize);
// N.B. the order the invoke BBs are processed in doesn't matter here.
const MCPhysReg *SavedRegs = RI.getCalleeSavedRegs(MF);
SmallVector<MachineBasicBlock*, 64> MBBLPads;
- for (SmallPtrSet<MachineBasicBlock*, 64>::iterator
- I = InvokeBBs.begin(), E = InvokeBBs.end(); I != E; ++I) {
- MachineBasicBlock *BB = *I;
+ for (MachineBasicBlock *BB : InvokeBBs) {
// Remove the landing pad successor from the invoke block and replace it
// with the new dispatch block.
Tys[n] = VecTy;
Tys[n++] = MVT::i32;
Tys[n] = MVT::Other;
- SDVTList SDTys = DAG.getVTList(ArrayRef<EVT>(Tys, NumResultVecs+2));
+ SDVTList SDTys = DAG.getVTList(makeArrayRef(Tys, NumResultVecs+2));
SmallVector<SDValue, 8> Ops;
Ops.push_back(N->getOperand(0)); // incoming chain
Ops.push_back(N->getOperand(AddrOpIdx));
for (n = 0; n < NumVecs; ++n)
Tys[n] = VT;
Tys[n] = MVT::Other;
- SDVTList SDTys = DAG.getVTList(ArrayRef<EVT>(Tys, NumVecs+1));
+ SDVTList SDTys = DAG.getVTList(makeArrayRef(Tys, NumVecs+1));
SDValue Ops[] = { VLD->getOperand(0), VLD->getOperand(2) };
MemIntrinsicSDNode *VLDMemInt = cast<MemIntrinsicSDNode>(VLD);
SDValue VLDDup = DAG.getMemIntrinsicNode(NewOpc, SDLoc(VLD), SDTys,
return true;
}
-bool ARMTargetLowering::shouldExpandAtomicInIR(Instruction *Inst) const {
- // Loads and stores less than 64-bits are already atomic; ones above that
- // are doomed anyway, so defer to the default libcall and blame the OS when
- // things go wrong. Cortex M doesn't have ldrexd/strexd though, so don't emit
- // anything for those.
- bool IsMClass = Subtarget->isMClass();
- if (StoreInst *SI = dyn_cast<StoreInst>(Inst)) {
- unsigned Size = SI->getValueOperand()->getType()->getPrimitiveSizeInBits();
- return Size == 64 && !IsMClass;
- } else if (LoadInst *LI = dyn_cast<LoadInst>(Inst)) {
- return LI->getType()->getPrimitiveSizeInBits() == 64 && !IsMClass;
+static void makeDMB(IRBuilder<> &Builder, ARM_MB::MemBOpt Domain) {
+ Module *M = Builder.GetInsertBlock()->getParent()->getParent();
+ Function *DMB = llvm::Intrinsic::getDeclaration(M, Intrinsic::arm_dmb);
+ Constant *CDomain = Builder.getInt32(Domain);
+ Builder.CreateCall(DMB, CDomain);
+}
+
+// Based on http://www.cl.cam.ac.uk/~pes20/cpp/cpp0xmappings.html
+void ARMTargetLowering::emitLeadingFence(IRBuilder<> &Builder,
+ AtomicOrdering Ord, bool IsStore,
+ bool IsLoad) const {
+ if (!getInsertFencesForAtomic())
+ return;
+
+ switch (Ord) {
+ case NotAtomic:
+ case Unordered:
+ llvm_unreachable("Invalid fence: unordered/non-atomic");
+ case Monotonic:
+ case Acquire:
+ return; // Nothing to do
+ case SequentiallyConsistent:
+ if (!IsStore)
+ return; // Nothing to do
+ /*FALLTHROUGH*/
+ case Release:
+ case AcquireRelease:
+ if (Subtarget->isSwift())
+ makeDMB(Builder, ARM_MB::ISHST);
+ // FIXME: add a comment with a link to documentation justifying this.
+ else
+ makeDMB(Builder, ARM_MB::ISH);
+ return;
}
+}
+
+void ARMTargetLowering::emitTrailingFence(IRBuilder<> &Builder,
+ AtomicOrdering Ord, bool IsStore,
+ bool IsLoad) const {
+ if (!getInsertFencesForAtomic())
+ return;
+
+ switch (Ord) {
+ case NotAtomic:
+ case Unordered:
+ llvm_unreachable("Invalid fence: unordered/not-atomic");
+ case Monotonic:
+ case Release:
+ return; // Nothing to do
+ case Acquire:
+ case AcquireRelease:
+ case SequentiallyConsistent:
+ makeDMB(Builder, ARM_MB::ISH);
+ return;
+ }
+}
+
+// Loads and stores less than 64-bits are already atomic; ones above that
+// are doomed anyway, so defer to the default libcall and blame the OS when
+// things go wrong. Cortex M doesn't have ldrexd/strexd though, so don't emit
+// anything for those.
+bool ARMTargetLowering::shouldExpandAtomicStoreInIR(StoreInst *SI) const {
+ unsigned Size = SI->getValueOperand()->getType()->getPrimitiveSizeInBits();
+ return (Size == 64) && !Subtarget->isMClass();
+}
+
+// Loads and stores less than 64-bits are already atomic; ones above that
+// are doomed anyway, so defer to the default libcall and blame the OS when
+// things go wrong. Cortex M doesn't have ldrexd/strexd though, so don't emit
+// anything for those.
+bool ARMTargetLowering::shouldExpandAtomicLoadInIR(LoadInst *LI) const {
+ unsigned Size = LI->getType()->getPrimitiveSizeInBits();
+ return (Size == 64) && !Subtarget->isMClass();
+}
- // For the real atomic operations, we have ldrex/strex up to 32 bits,
- // and up to 64 bits on the non-M profiles
- unsigned AtomicLimit = IsMClass ? 32 : 64;
- return Inst->getType()->getPrimitiveSizeInBits() <= AtomicLimit;
+// For the real atomic operations, we have ldrex/strex up to 32 bits,
+// and up to 64 bits on the non-M profiles
+bool ARMTargetLowering::shouldExpandAtomicRMWInIR(AtomicRMWInst *AI) const {
+ unsigned Size = AI->getType()->getPrimitiveSizeInBits();
+ return Size <= (Subtarget->isMClass() ? 32 : 64);
}
// This has so far only been implemented for MachO.
projects / oota-llvm.git / blobdiff