summaryrefslogtreecommitdiff
path: root/lib/Target/ARM
diff options
context:
space:
mode:
Diffstat (limited to 'lib/Target/ARM')
-rw-r--r--lib/Target/ARM/ARMBaseRegisterInfo.cpp1
-rw-r--r--lib/Target/ARM/ARMISelLowering.cpp59
-rw-r--r--lib/Target/ARM/ARMInstrInfo.td10
-rw-r--r--lib/Target/ARM/ARMRegisterInfo.td25
-rw-r--r--lib/Target/ARM/ARMSubtarget.cpp21
-rw-r--r--lib/Target/ARM/ARMTargetMachine.cpp2
-rw-r--r--lib/Target/ARM/Disassembler/ARMDisassembler.cpp18
-rw-r--r--lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.cpp2
8 files changed, 96 insertions, 42 deletions
diff --git a/lib/Target/ARM/ARMBaseRegisterInfo.cpp b/lib/Target/ARM/ARMBaseRegisterInfo.cpp
index b0d34a76b0..4de5b4f41c 100644
--- a/lib/Target/ARM/ARMBaseRegisterInfo.cpp
+++ b/lib/Target/ARM/ARMBaseRegisterInfo.cpp
@@ -94,6 +94,7 @@ getReservedRegs(const MachineFunction &MF) const {
Reserved.set(ARM::SP);
Reserved.set(ARM::PC);
Reserved.set(ARM::FPSCR);
+ Reserved.set(ARM::APSR_NZCV);
if (TFI->hasFP(MF))
Reserved.set(FramePtr);
if (hasBasePointer(MF))
diff --git a/lib/Target/ARM/ARMISelLowering.cpp b/lib/Target/ARM/ARMISelLowering.cpp
index fd77732364..432e3eefb1 100644
--- a/lib/Target/ARM/ARMISelLowering.cpp
+++ b/lib/Target/ARM/ARMISelLowering.cpp
@@ -2147,7 +2147,7 @@ bool ARMTargetLowering::isUsedByReturnOnly(SDNode *N, SDValue &Chain) const {
Copy = *Copy->use_begin();
if (Copy->getOpcode() != ISD::CopyToReg || !Copy->hasNUsesOfValue(1, 0))
return false;
- Chain = Copy->getOperand(0);
+ TCChain = Copy->getOperand(0);
} else {
return false;
}
@@ -5257,6 +5257,23 @@ static bool isZeroExtended(SDNode *N, SelectionDAG &DAG) {
return false;
}
+static EVT getExtensionTo64Bits(const EVT &OrigVT) {
+ if (OrigVT.getSizeInBits() >= 64)
+ return OrigVT;
+
+ assert(OrigVT.isSimple() && "Expecting a simple value type");
+
+ MVT::SimpleValueType OrigSimpleTy = OrigVT.getSimpleVT().SimpleTy;
+ switch (OrigSimpleTy) {
+ default: llvm_unreachable("Unexpected Vector Type");
+ case MVT::v2i8:
+ case MVT::v2i16:
+ return MVT::v2i32;
+ case MVT::v4i8:
+ return MVT::v4i16;
+ }
+}
+
/// AddRequiredExtensionForVMULL - Add a sign/zero extension to extend the total
/// value size to 64 bits. We need a 64-bit D register as an operand to VMULL.
/// We insert the required extension here to get the vector to fill a D register.
@@ -5272,18 +5289,8 @@ static SDValue AddRequiredExtensionForVMULL(SDValue N, SelectionDAG &DAG,
return N;
// Must extend size to at least 64 bits to be used as an operand for VMULL.
- MVT::SimpleValueType OrigSimpleTy = OrigTy.getSimpleVT().SimpleTy;
- EVT NewVT;
- switch (OrigSimpleTy) {
- default: llvm_unreachable("Unexpected Orig Vector Type");
- case MVT::v2i8:
- case MVT::v2i16:
- NewVT = MVT::v2i32;
- break;
- case MVT::v4i8:
- NewVT = MVT::v4i16;
- break;
- }
+ EVT NewVT = getExtensionTo64Bits(OrigTy);
+
return DAG.getNode(ExtOpcode, N->getDebugLoc(), NewVT, N);
}
@@ -5293,22 +5300,22 @@ static SDValue AddRequiredExtensionForVMULL(SDValue N, SelectionDAG &DAG,
/// reach a total size of 64 bits. We have to add the extension separately
/// because ARM does not have a sign/zero extending load for vectors.
static SDValue SkipLoadExtensionForVMULL(LoadSDNode *LD, SelectionDAG& DAG) {
- SDValue NonExtendingLoad =
- DAG.getLoad(LD->getMemoryVT(), LD->getDebugLoc(), LD->getChain(),
+ EVT ExtendedTy = getExtensionTo64Bits(LD->getMemoryVT());
+
+ // The load already has the right type.
+ if (ExtendedTy == LD->getMemoryVT())
+ return DAG.getLoad(LD->getMemoryVT(), LD->getDebugLoc(), LD->getChain(),
LD->getBasePtr(), LD->getPointerInfo(), LD->isVolatile(),
LD->isNonTemporal(), LD->isInvariant(),
LD->getAlignment());
- unsigned ExtOp = 0;
- switch (LD->getExtensionType()) {
- default: llvm_unreachable("Unexpected LoadExtType");
- case ISD::EXTLOAD:
- case ISD::SEXTLOAD: ExtOp = ISD::SIGN_EXTEND; break;
- case ISD::ZEXTLOAD: ExtOp = ISD::ZERO_EXTEND; break;
- }
- MVT::SimpleValueType MemType = LD->getMemoryVT().getSimpleVT().SimpleTy;
- MVT::SimpleValueType ExtType = LD->getValueType(0).getSimpleVT().SimpleTy;
- return AddRequiredExtensionForVMULL(NonExtendingLoad, DAG,
- MemType, ExtType, ExtOp);
+
+ // We need to create a zextload/sextload. We cannot just create a load
+ // followed by a zext/zext node because LowerMUL is also run during normal
+ // operation legalization where we can't create illegal types.
+ return DAG.getExtLoad(LD->getExtensionType(), LD->getDebugLoc(), ExtendedTy,
+ LD->getChain(), LD->getBasePtr(), LD->getPointerInfo(),
+ LD->getMemoryVT(), LD->isVolatile(),
+ LD->isNonTemporal(), LD->getAlignment());
}
/// SkipExtensionForVMULL - For a node that is a SIGN_EXTEND, ZERO_EXTEND,
diff --git a/lib/Target/ARM/ARMInstrInfo.td b/lib/Target/ARM/ARMInstrInfo.td
index 1bd174e341..89f92a589d 100644
--- a/lib/Target/ARM/ARMInstrInfo.td
+++ b/lib/Target/ARM/ARMInstrInfo.td
@@ -4636,11 +4636,11 @@ def : ARMInstAlias<"mcr${p} $cop, $opc1, $Rt, $CRn, $CRm",
(MCR p_imm:$cop, imm0_7:$opc1, GPR:$Rt, c_imm:$CRn,
c_imm:$CRm, 0, pred:$p)>;
def MRC : MovRCopro<"mrc", 1 /* from coprocessor to ARM core register */,
- (outs GPR:$Rt),
+ (outs GPRwithAPSR:$Rt),
(ins p_imm:$cop, imm0_7:$opc1, c_imm:$CRn, c_imm:$CRm,
imm0_7:$opc2), []>;
def : ARMInstAlias<"mrc${p} $cop, $opc1, $Rt, $CRn, $CRm",
- (MRC GPR:$Rt, p_imm:$cop, imm0_7:$opc1, c_imm:$CRn,
+ (MRC GPRwithAPSR:$Rt, p_imm:$cop, imm0_7:$opc1, c_imm:$CRn,
c_imm:$CRm, 0, pred:$p)>;
def : ARMPat<(int_arm_mrc imm:$cop, imm:$opc1, imm:$CRn, imm:$CRm, imm:$opc2),
@@ -4650,7 +4650,7 @@ class MovRCopro2<string opc, bit direction, dag oops, dag iops,
list<dag> pattern>
: ABXI<0b1110, oops, iops, NoItinerary,
!strconcat(opc, "\t$cop, $opc1, $Rt, $CRn, $CRm, $opc2"), pattern> {
- let Inst{31-28} = 0b1111;
+ let Inst{31-24} = 0b11111110;
let Inst{20} = direction;
let Inst{4} = 1;
@@ -4679,11 +4679,11 @@ def : ARMInstAlias<"mcr2$ $cop, $opc1, $Rt, $CRn, $CRm",
(MCR2 p_imm:$cop, imm0_7:$opc1, GPR:$Rt, c_imm:$CRn,
c_imm:$CRm, 0)>;
def MRC2 : MovRCopro2<"mrc2", 1 /* from coprocessor to ARM core register */,
- (outs GPR:$Rt),
+ (outs GPRwithAPSR:$Rt),
(ins p_imm:$cop, imm0_7:$opc1, c_imm:$CRn, c_imm:$CRm,
imm0_7:$opc2), []>;
def : ARMInstAlias<"mrc2$ $cop, $opc1, $Rt, $CRn, $CRm",
- (MRC2 GPR:$Rt, p_imm:$cop, imm0_7:$opc1, c_imm:$CRn,
+ (MRC2 GPRwithAPSR:$Rt, p_imm:$cop, imm0_7:$opc1, c_imm:$CRn,
c_imm:$CRm, 0)>;
def : ARMV5TPat<(int_arm_mrc2 imm:$cop, imm:$opc1, imm:$CRn,
diff --git a/lib/Target/ARM/ARMRegisterInfo.td b/lib/Target/ARM/ARMRegisterInfo.td
index b0f576bc2b..85743d8d5a 100644
--- a/lib/Target/ARM/ARMRegisterInfo.td
+++ b/lib/Target/ARM/ARMRegisterInfo.td
@@ -157,12 +157,15 @@ def Q15 : ARMReg<15, "q15", [D30, D31]>;
// Current Program Status Register.
// We model fpscr with two registers: FPSCR models the control bits and will be
-// reserved. FPSCR_NZCV models the flag bits and will be unreserved.
-def CPSR : ARMReg<0, "cpsr">;
-def APSR : ARMReg<1, "apsr">;
-def SPSR : ARMReg<2, "spsr">;
-def FPSCR : ARMReg<3, "fpscr">;
-def FPSCR_NZCV : ARMReg<3, "fpscr_nzcv"> {
+// reserved. FPSCR_NZCV models the flag bits and will be unreserved. APSR_NZCV
+// models the APSR when it's accessed by some special instructions. In such cases
+// it has the same encoding as PC.
+def CPSR : ARMReg<0, "cpsr">;
+def APSR : ARMReg<1, "apsr">;
+def APSR_NZCV : ARMReg<15, "apsr_nzcv">;
+def SPSR : ARMReg<2, "spsr">;
+def FPSCR : ARMReg<3, "fpscr">;
+def FPSCR_NZCV : ARMReg<3, "fpscr_nzcv"> {
let Aliases = [FPSCR];
}
def ITSTATE : ARMReg<4, "itstate">;
@@ -207,6 +210,16 @@ def GPRnopc : RegisterClass<"ARM", [i32], 32, (sub GPR, PC)> {
}];
}
+// GPRs without the PC but with APSR. Some instructions allow accessing the
+// APSR, while actually encoding PC in the register field. This is usefull
+// for assembly and disassembly only.
+def GPRwithAPSR : RegisterClass<"ARM", [i32], 32, (add GPR, APSR_NZCV)> {
+ let AltOrders = [(add LR, GPRnopc), (trunc GPRnopc, 8)];
+ let AltOrderSelect = [{
+ return 1 + MF.getTarget().getSubtarget<ARMSubtarget>().isThumb1Only();
+ }];
+}
+
// GPRsp - Only the SP is legal. Used by Thumb1 instructions that want the
// implied SP argument list.
// FIXME: It would be better to not use this at all and refactor the
diff --git a/lib/Target/ARM/ARMSubtarget.cpp b/lib/Target/ARM/ARMSubtarget.cpp
index 8653c462f0..9ff0d61481 100644
--- a/lib/Target/ARM/ARMSubtarget.cpp
+++ b/lib/Target/ARM/ARMSubtarget.cpp
@@ -162,10 +162,23 @@ void ARMSubtarget::resetSubtargetFeatures(StringRef CPU, StringRef FS) {
if (!isThumb() || hasThumb2())
PostRAScheduler = true;
- // v6+ may or may not support unaligned mem access depending on the system
- // configuration.
- if (!StrictAlign && hasV6Ops() && isTargetDarwin())
- AllowsUnalignedMem = true;
+ if (!StrictAlign) {
+ // Assume pre-ARMv6 doesn't support unaligned accesses.
+ //
+ // ARMv6 may or may not support unaligned accesses depending on the
+ // SCTLR.U bit, which is architecture-specific. We assume ARMv6
+ // Darwin targets support unaligned accesses, and others don't.
+ //
+ // ARMv7 always has SCTLR.U set to 1, but it has a new SCTLR.A bit
+ // which raises an alignment fault on unaligned accesses. Linux
+ // defaults this bit to 0 and handles it as a system-wide (not
+ // per-process) setting. It is therefore safe to assume that ARMv7+
+ // targets support unaligned accesses.
+ //
+ // The above behavior is consistent with GCC.
+ if (hasV7Ops() || (hasV6Ops() && isTargetDarwin()))
+ AllowsUnalignedMem = true;
+ }
// NEON f32 ops are non-IEEE 754 compliant. Darwin is ok with it by default.
uint64_t Bits = getFeatureBits();
diff --git a/lib/Target/ARM/ARMTargetMachine.cpp b/lib/Target/ARM/ARMTargetMachine.cpp
index 42c7d2c437..17c52c94a0 100644
--- a/lib/Target/ARM/ARMTargetMachine.cpp
+++ b/lib/Target/ARM/ARMTargetMachine.cpp
@@ -85,6 +85,7 @@ ARMTargetMachine::ARMTargetMachine(const Target &T, StringRef TT,
TLInfo(*this),
TSInfo(*this),
FrameLowering(Subtarget) {
+ initAsmInfo();
if (!Subtarget.hasARMOps())
report_fatal_error("CPU: '" + Subtarget.getCPUString() + "' does not "
"support ARM mode execution!");
@@ -117,6 +118,7 @@ ThumbTargetMachine::ThumbTargetMachine(const Target &T, StringRef TT,
FrameLowering(Subtarget.hasThumb2()
? new ARMFrameLowering(Subtarget)
: (ARMFrameLowering*)new Thumb1FrameLowering(Subtarget)) {
+ initAsmInfo();
}
namespace {
diff --git a/lib/Target/ARM/Disassembler/ARMDisassembler.cpp b/lib/Target/ARM/Disassembler/ARMDisassembler.cpp
index ac937f3534..d2896377cc 100644
--- a/lib/Target/ARM/Disassembler/ARMDisassembler.cpp
+++ b/lib/Target/ARM/Disassembler/ARMDisassembler.cpp
@@ -156,6 +156,9 @@ static DecodeStatus DecodeGPRRegisterClass(MCInst &Inst, unsigned RegNo,
static DecodeStatus DecodeGPRnopcRegisterClass(MCInst &Inst,
unsigned RegNo, uint64_t Address,
const void *Decoder);
+static DecodeStatus DecodeGPRwithAPSRRegisterClass(MCInst &Inst,
+ unsigned RegNo, uint64_t Address,
+ const void *Decoder);
static DecodeStatus DecodetGPRRegisterClass(MCInst &Inst, unsigned RegNo,
uint64_t Address, const void *Decoder);
static DecodeStatus DecodetcGPRRegisterClass(MCInst &Inst, unsigned RegNo,
@@ -920,6 +923,21 @@ DecodeGPRnopcRegisterClass(MCInst &Inst, unsigned RegNo,
return S;
}
+static DecodeStatus
+DecodeGPRwithAPSRRegisterClass(MCInst &Inst, unsigned RegNo,
+ uint64_t Address, const void *Decoder) {
+ DecodeStatus S = MCDisassembler::Success;
+
+ if (RegNo == 15)
+ {
+ Inst.addOperand(MCOperand::CreateReg(ARM::APSR_NZCV));
+ return MCDisassembler::Success;
+ }
+
+ Check(S, DecodeGPRRegisterClass(Inst, RegNo, Address, Decoder));
+ return S;
+}
+
static DecodeStatus DecodetGPRRegisterClass(MCInst &Inst, unsigned RegNo,
uint64_t Address, const void *Decoder) {
if (RegNo > 7)
diff --git a/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.cpp b/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.cpp
index 57239f8011..b858fff546 100644
--- a/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.cpp
+++ b/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.cpp
@@ -159,7 +159,7 @@ static MCRegisterInfo *createARMMCRegisterInfo(StringRef Triple) {
return X;
}
-static MCAsmInfo *createARMMCAsmInfo(StringRef TT) {
+static MCAsmInfo *createARMMCAsmInfo(const MCRegisterInfo &MRI, StringRef TT) {
Triple TheTriple(TT);
if (TheTriple.isOSDarwin())
generated by cgit v1.2.3 (git 2.25.1) at 2024-04-29 02:22:30 +0000