[ARM] Add the non-MVE instructions in Arm v8.1-M.

This adds support for the new family of conditional selection /
increment / negation instructions; the low-overhead branch
instructions (e.g. BF, WLS, DLS); the CLRM instruction to zero a whole
list of registers at once; the new VMRS/VMSR and VLDR/VSTR
instructions to get data in and out of 8.1-M system registers,
particularly including the new VPR register used by MVE vector
predication.

To support this, we also add a register name 'zr' (used by the CSEL
family to force one of the inputs to the constant 0), and operand
types for lists of registers that are also allowed to include APSR or
VPR (used by CLRM). The VLDR/VSTR instructions also need some new
addressing modes.

The low-overhead branch instructions exist in their own separate
architecture extension, which we treat as enabled by default, but you
can say -mattr=-lob or equivalent to turn it off.

Reviewers: dmgreen, samparker, SjoerdMeijer, t.p.northover

Reviewed By: samparker

Subscribers: miyuki, javed.absar, kristof.beyls, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D62667

llvm-svn: 362953

7 files changed, 313 insertions, 40 deletions

diff --git a/llvm/lib/Target/ARM/MCTargetDesc/ARMAsmBackend.cpp b/llvm/lib/Target/ARM/MCTargetDesc/ARMAsmBackend.cpp
index 538bc2594c3..d1c891bb3be 100644
--- a/llvm/lib/Target/ARM/MCTargetDesc/ARMAsmBackend.cpp
+++ b/llvm/lib/Target/ARM/MCTargetDesc/ARMAsmBackend.cpp
@@ -29,6 +29,7 @@
 #include "llvm/MC/MCSectionMachO.h"
 #include "llvm/MC/MCSubtargetInfo.h"
 #include "llvm/MC/MCValue.h"
+#include "llvm/MC/MCAsmLayout.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/EndianStream.h"
 #include "llvm/Support/ErrorHandling.h"
@@ -104,6 +105,13 @@ const MCFixupKindInfo &ARMAsmBackend::getFixupKindInfo(MCFixupKind Kind) const {
       {"fixup_t2_movw_lo16", 0, 20, 0},
       {"fixup_arm_mod_imm", 0, 12, 0},
       {"fixup_t2_so_imm", 0, 26, 0},
+      {"fixup_bf_branch", 0, 32, MCFixupKindInfo::FKF_IsPCRel},
+      {"fixup_bf_target", 0, 32, MCFixupKindInfo::FKF_IsPCRel},
+      {"fixup_bfl_target", 0, 32, MCFixupKindInfo::FKF_IsPCRel},
+      {"fixup_bfc_target", 0, 32, MCFixupKindInfo::FKF_IsPCRel},
+      {"fixup_bfcsel_else_target", 0, 32, 0},
+      {"fixup_wls", 0, 32, MCFixupKindInfo::FKF_IsPCRel},
+      {"fixup_le", 0, 32, MCFixupKindInfo::FKF_IsPCRel}
   };
   const static MCFixupKindInfo InfosBE[ARM::NumTargetFixupKinds] = {
       // This table *must* be in the order that the fixup_* kinds are defined in
@@ -155,6 +163,13 @@ const MCFixupKindInfo &ARMAsmBackend::getFixupKindInfo(MCFixupKind Kind) const {
       {"fixup_t2_movw_lo16", 12, 20, 0},
       {"fixup_arm_mod_imm", 20, 12, 0},
       {"fixup_t2_so_imm", 26, 6, 0},
+      {"fixup_bf_branch", 0, 32, MCFixupKindInfo::FKF_IsPCRel},
+      {"fixup_bf_target", 0, 32, MCFixupKindInfo::FKF_IsPCRel},
+      {"fixup_bfl_target", 0, 32, MCFixupKindInfo::FKF_IsPCRel},
+      {"fixup_bfc_target", 0, 32, MCFixupKindInfo::FKF_IsPCRel},
+      {"fixup_bfcsel_else_target", 0, 32, 0},
+      {"fixup_wls", 0, 32, MCFixupKindInfo::FKF_IsPCRel},
+      {"fixup_le", 0, 32, MCFixupKindInfo::FKF_IsPCRel}
   };
 
   if (Kind < FirstTargetFixupKind)
@@ -256,6 +271,56 @@ const char *ARMAsmBackend::reasonForFixupRelaxation(const MCFixup &Fixup,
       return "will be converted to nop";
     break;
   }
+  case ARM::fixup_bf_branch: {
+    int64_t Offset = int64_t(Value) - 4;
+    if (Offset < 0 || Offset > 30)
+      return "out of range pc-relative fixup value";
+    break;
+  }
+  case ARM::fixup_bf_target: {
+    int64_t Offset = int64_t(Value) - 4;
+    if (Offset > 65534 || Offset < -65536)
+      return "out of range pc-relative fixup value";
+    break;
+  }
+  case ARM::fixup_bfl_target: {
+    int64_t Offset = int64_t(Value) - 4;
+    if (Offset > 262142 || Offset < -262144)
+      return "out of range pc-relative fixup value";
+    break;
+  }
+  case ARM::fixup_bfc_target: {
+    int64_t Offset = int64_t(Value) - 4;
+    if (Offset > 4094 || Offset < -4096)
+      return "out of range pc-relative fixup value";
+    break;
+  }
+  case ARM::fixup_bfcsel_else_target: {
+    if (Value != 2 && Value != 4)
+      return "out of range label-relative fixup value";
+    break;
+  }
+  case ARM::fixup_wls: {
+    uint64_t Offset = Value - 4;
+    if (Offset > 4094)
+      return "out of range pc-relative fixup value";
+    break;
+  }
+  case ARM::fixup_le: {
+    // The offset field in the LE and LETP instructions is an 11-bit
+    // value shifted left by 2 (i.e. 0,2,4,...,4094), and it is
+    // interpreted as a negative offset from the value read from pc,
+    // i.e. from instruction_address+4.
+    //
+    // So an LE instruction can in principle address the instruction
+    // immediately after itself, or (not very usefully) the address
+    // half way through the 4-byte LE.
+    uint64_t Offset = -Value + 4;
+    if (Offset > 4094)
+      return "out of range pc-relative fixup value";
+    break;
+  }
+
   default:
     llvm_unreachable("Unexpected fixup kind in reasonForFixupRelaxation()!");
   }
@@ -760,6 +825,60 @@ unsigned ARMAsmBackend::adjustFixupValue(const MCAssembler &Asm,
     EncValue |= (Value & 0xff);
     return swapHalfWords(EncValue, Endian == support::little);
   }
+  case ARM::fixup_bf_branch: {
+    const char *FixupDiagnostic = reasonForFixupRelaxation(Fixup, Value);
+    if (FixupDiagnostic) {
+      Ctx.reportError(Fixup.getLoc(), FixupDiagnostic);
+      return 0;
+    }
+    uint32_t out = (((Value - 4) >> 1) & 0xf) << 23;
+    return swapHalfWords(out, Endian == support::little);
+  }
+  case ARM::fixup_bf_target:
+  case ARM::fixup_bfl_target:
+  case ARM::fixup_bfc_target: {
+    const char *FixupDiagnostic = reasonForFixupRelaxation(Fixup, Value);
+    if (FixupDiagnostic) {
+      Ctx.reportError(Fixup.getLoc(), FixupDiagnostic);
+      return 0;
+    }
+    uint32_t out = 0;
+    uint32_t HighBitMask = (Kind == ARM::fixup_bf_target ? 0xf800 :
+                            Kind == ARM::fixup_bfl_target ? 0x3f800 : 0x800);
+    out |= (((Value - 4) >> 1) & 0x1) << 11;
+    out |= (((Value - 4) >> 1) & 0x7fe);
+    out |= (((Value - 4) >> 1) & HighBitMask) << 5;
+    return swapHalfWords(out, Endian == support::little);
+  }
+  case ARM::fixup_bfcsel_else_target: {
+    // If this is a fixup of a branch future's else target then it should be a
+    // constant MCExpr representing the distance between the branch targetted
+    // and the instruction after that same branch.
+    Value = Target.getConstant();
+
+    const char *FixupDiagnostic = reasonForFixupRelaxation(Fixup, Value);
+    if (FixupDiagnostic) {
+      Ctx.reportError(Fixup.getLoc(), FixupDiagnostic);
+      return 0;
+    }
+    uint32_t out = ((Value >> 2) & 1) << 17;
+    return swapHalfWords(out, Endian == support::little);
+  }
+  case ARM::fixup_wls:
+  case ARM::fixup_le: {
+    const char *FixupDiagnostic = reasonForFixupRelaxation(Fixup, Value);
+    if (FixupDiagnostic) {
+      Ctx.reportError(Fixup.getLoc(), FixupDiagnostic);
+      return 0;
+    }
+    uint64_t real_value = Value - 4;
+    uint32_t out = 0;
+    if (Kind == ARM::fixup_le)
+      real_value = -real_value;
+    out |= ((real_value >> 1) & 0x1) << 11;
+    out |= ((real_value >> 1) & 0x7fe);
+    return swapHalfWords(out, Endian == support::little);
+  }
   }
 }
 
@@ -854,6 +973,13 @@ static unsigned getFixupKindNumBytes(unsigned Kind) {
   case ARM::fixup_t2_movt_hi16:
   case ARM::fixup_t2_movw_lo16:
   case ARM::fixup_t2_so_imm:
+  case ARM::fixup_bf_branch:
+  case ARM::fixup_bf_target:
+  case ARM::fixup_bfl_target:
+  case ARM::fixup_bfc_target:
+  case ARM::fixup_bfcsel_else_target:
+  case ARM::fixup_wls:
+  case ARM::fixup_le:
     return 4;
 
   case FK_SecRel_2:
@@ -910,6 +1036,13 @@ static unsigned getFixupKindContainerSizeBytes(unsigned Kind) {
   case ARM::fixup_t2_movw_lo16:
   case ARM::fixup_arm_mod_imm:
   case ARM::fixup_t2_so_imm:
+  case ARM::fixup_bf_branch:
+  case ARM::fixup_bf_target:
+  case ARM::fixup_bfl_target:
+  case ARM::fixup_bfc_target:
+  case ARM::fixup_bfcsel_else_target:
+  case ARM::fixup_wls:
+  case ARM::fixup_le:
     // Instruction size is 4 bytes.
     return 4;
   }
diff --git a/llvm/lib/Target/ARM/MCTargetDesc/ARMELFObjectWriter.cpp b/llvm/lib/Target/ARM/MCTargetDesc/ARMELFObjectWriter.cpp
index 1d51a9696cd..fda19eea1de 100644
--- a/llvm/lib/Target/ARM/MCTargetDesc/ARMELFObjectWriter.cpp
+++ b/llvm/lib/Target/ARM/MCTargetDesc/ARMELFObjectWriter.cpp
@@ -137,6 +137,12 @@ unsigned ARMELFObjectWriter::GetRelocTypeInner(const MCValue &Target,
       default:
         return ELF::R_ARM_THM_CALL;
       }
+    case ARM::fixup_bf_target:
+      return ELF::R_ARM_THM_BF16;
+    case ARM::fixup_bfc_target:
+      return ELF::R_ARM_THM_BF12;
+    case ARM::fixup_bfl_target:
+      return ELF::R_ARM_THM_BF18;
     }
   }
   switch ((unsigned)Fixup.getKind()) {
diff --git a/llvm/lib/Target/ARM/MCTargetDesc/ARMFixupKinds.h b/llvm/lib/Target/ARM/MCTargetDesc/ARMFixupKinds.h
index 533bf6d4df9..bdf04a208b2 100644
--- a/llvm/lib/Target/ARM/MCTargetDesc/ARMFixupKinds.h
+++ b/llvm/lib/Target/ARM/MCTargetDesc/ARMFixupKinds.h
@@ -103,6 +103,15 @@ enum Fixups {
   // Fixup for Thumb2 8-bit rotated operand
   fixup_t2_so_imm,
 
+  // Fixups for Branch Future.
+  fixup_bf_branch,
+  fixup_bf_target,
+  fixup_bfl_target,
+  fixup_bfc_target,
+  fixup_bfcsel_else_target,
+  fixup_wls,
+  fixup_le,
+
   // Marker
   LastTargetFixupKind,
   NumTargetFixupKinds = LastTargetFixupKind - FirstTargetFixupKind
diff --git a/llvm/lib/Target/ARM/MCTargetDesc/ARMInstPrinter.cpp b/llvm/lib/Target/ARM/MCTargetDesc/ARMInstPrinter.cpp
index ec5fd16cd7d..271959c38de 100644
--- a/llvm/lib/Target/ARM/MCTargetDesc/ARMInstPrinter.cpp
+++ b/llvm/lib/Target/ARM/MCTargetDesc/ARMInstPrinter.cpp
@@ -771,11 +771,13 @@ void ARMInstPrinter::printPKHASRShiftImm(const MCInst *MI, unsigned OpNum,
 void ARMInstPrinter::printRegisterList(const MCInst *MI, unsigned OpNum,
                                        const MCSubtargetInfo &STI,
                                        raw_ostream &O) {
-  assert(std::is_sorted(MI->begin() + OpNum, MI->end(),
-                        [&](const MCOperand &LHS, const MCOperand &RHS) {
-                          return MRI.getEncodingValue(LHS.getReg()) <
-                                 MRI.getEncodingValue(RHS.getReg());
-                        }));
+  if (MI->getOpcode() != ARM::t2CLRM) {
+    assert(std::is_sorted(MI->begin() + OpNum, MI->end(),
+                          [&](const MCOperand &LHS, const MCOperand &RHS) {
+                            return MRI.getEncodingValue(LHS.getReg()) <
+                                   MRI.getEncodingValue(RHS.getReg());
+                          }));
+  }
 
   O << "{";
   for (unsigned i = OpNum, e = MI->getNumOperands(); i != e; ++i) {
@@ -930,6 +932,15 @@ void ARMInstPrinter::printPredicateOperand(const MCInst *MI, unsigned OpNum,
     O << ARMCondCodeToString(CC);
 }
 
+void ARMInstPrinter::printMandatoryRestrictedPredicateOperand(
+    const MCInst *MI, unsigned OpNum, const MCSubtargetInfo &STI,
+    raw_ostream &O) {
+  if ((ARMCC::CondCodes)MI->getOperand(OpNum).getImm() == ARMCC::HS)
+    O << "cs";
+  else
+    printMandatoryPredicateOperand(MI, OpNum, STI, O);
+}
+
 void ARMInstPrinter::printMandatoryPredicateOperand(const MCInst *MI,
                                                     unsigned OpNum,
                                                     const MCSubtargetInfo &STI,
@@ -938,6 +949,14 @@ void ARMInstPrinter::printMandatoryPredicateOperand(const MCInst *MI,
   O << ARMCondCodeToString(CC);
 }
 
+void ARMInstPrinter::printMandatoryInvertedPredicateOperand(const MCInst *MI,
+                                                            unsigned OpNum,
+                                                            const MCSubtargetInfo &STI,
+                                                            raw_ostream &O) {
+  ARMCC::CondCodes CC = (ARMCC::CondCodes)MI->getOperand(OpNum).getImm();
+  O << ARMCondCodeToString(ARMCC::getOppositeCondition(CC));
+}
+
 void ARMInstPrinter::printSBitModifierOperand(const MCInst *MI, unsigned OpNum,
                                               const MCSubtargetInfo &STI,
                                               raw_ostream &O) {
diff --git a/llvm/lib/Target/ARM/MCTargetDesc/ARMInstPrinter.h b/llvm/lib/Target/ARM/MCTargetDesc/ARMInstPrinter.h
index d07daa7bd15..ea65cde81a9 100644
--- a/llvm/lib/Target/ARM/MCTargetDesc/ARMInstPrinter.h
+++ b/llvm/lib/Target/ARM/MCTargetDesc/ARMInstPrinter.h
@@ -170,6 +170,13 @@ public:
   void printMandatoryPredicateOperand(const MCInst *MI, unsigned OpNum,
                                       const MCSubtargetInfo &STI,
                                       raw_ostream &O);
+  void printMandatoryRestrictedPredicateOperand(const MCInst *MI,
+                                                unsigned OpNum,
+                                                const MCSubtargetInfo &STI,
+                                                raw_ostream &O);
+  void printMandatoryInvertedPredicateOperand(const MCInst *MI, unsigned OpNum,
+                                      const MCSubtargetInfo &STI,
+                                      raw_ostream &O);
   void printSBitModifierOperand(const MCInst *MI, unsigned OpNum,
                                 const MCSubtargetInfo &STI, raw_ostream &O);
   void printRegisterList(const MCInst *MI, unsigned OpNum,
diff --git a/llvm/lib/Target/ARM/MCTargetDesc/ARMMCCodeEmitter.cpp b/llvm/lib/Target/ARM/MCTargetDesc/ARMMCCodeEmitter.cpp
index acc00c70c02..e84fe35ee20 100644
--- a/llvm/lib/Target/ARM/MCTargetDesc/ARMMCCodeEmitter.cpp
+++ b/llvm/lib/Target/ARM/MCTargetDesc/ARMMCCodeEmitter.cpp
@@ -49,7 +49,7 @@ namespace {
 
 class ARMMCCodeEmitter : public MCCodeEmitter {
   const MCInstrInfo &MCII;
-  const MCContext &CTX;
+  MCContext &CTX;
   bool IsLittleEndian;
 
 public:
@@ -180,18 +180,24 @@ public:
                                    SmallVectorImpl<MCFixup> &Fixups,
                                    const MCSubtargetInfo &STI) const;
 
+  /// getT2AddrModeImm7s4OpValue - Return encoding info for 'reg +/- imm7<<2'
+  /// operand.
+  uint32_t getT2AddrModeImm7s4OpValue(const MCInst &MI, unsigned OpIdx,
+                                      SmallVectorImpl<MCFixup> &Fixups,
+                                      const MCSubtargetInfo &STI) const;
+
   /// getT2AddrModeImm0_1020s4OpValue - Return encoding info for 'reg + imm8<<2'
   /// operand.
   uint32_t getT2AddrModeImm0_1020s4OpValue(const MCInst &MI, unsigned OpIdx,
                                    SmallVectorImpl<MCFixup> &Fixups,
                                    const MCSubtargetInfo &STI) const;
 
-  /// getT2Imm8s4OpValue - Return encoding info for '+/- imm8<<2'
+  /// getT2ScaledImmOpValue - Return encoding info for '+/- immX<<Y'
   /// operand.
-  uint32_t getT2Imm8s4OpValue(const MCInst &MI, unsigned OpIdx,
-                              SmallVectorImpl<MCFixup> &Fixups,
-                              const MCSubtargetInfo &STI) const;
-
+  template<unsigned Bits, unsigned Shift>
+  uint32_t getT2ScaledImmOpValue(const MCInst &MI, unsigned OpIdx,
+                                 SmallVectorImpl<MCFixup> &Fixups,
+                                 const MCSubtargetInfo &STI) const;
 
   /// getLdStSORegOpValue - Return encoding info for 'reg +/- reg shop imm'
   /// operand as needed by load/store instructions.
@@ -416,6 +422,15 @@ public:
   void encodeInstruction(const MCInst &MI, raw_ostream &OS,
                          SmallVectorImpl<MCFixup> &Fixups,
                          const MCSubtargetInfo &STI) const override;
+
+  template <bool isNeg, ARM::Fixups fixup>
+  uint32_t getBFTargetOpValue(const MCInst &MI, unsigned OpIdx,
+                              SmallVectorImpl<MCFixup> &Fixups,
+                              const MCSubtargetInfo &STI) const;
+
+  uint32_t getBFAfterTargetOpValue(const MCInst &MI, unsigned OpIdx,
+                                   SmallVectorImpl<MCFixup> &Fixups,
+                                   const MCSubtargetInfo &STI) const;
 };
 
 } // end anonymous namespace
@@ -894,12 +909,11 @@ getAddrModeImm12OpValue(const MCInst &MI, unsigned OpIdx,
   return Binary;
 }
 
-/// getT2Imm8s4OpValue - Return encoding info for
-/// '+/- imm8<<2' operand.
+template<unsigned Bits, unsigned Shift>
 uint32_t ARMMCCodeEmitter::
-getT2Imm8s4OpValue(const MCInst &MI, unsigned OpIdx,
-                   SmallVectorImpl<MCFixup> &Fixups,
-                   const MCSubtargetInfo &STI) const {
+getT2ScaledImmOpValue(const MCInst &MI, unsigned OpIdx,
+                      SmallVectorImpl<MCFixup> &Fixups,
+                      const MCSubtargetInfo &STI) const {
   // FIXME: The immediate operand should have already been encoded like this
   // before ever getting here. The encoder method should just need to combine
   // the MI operands for the register and the offset into a single
@@ -907,25 +921,23 @@ getT2Imm8s4OpValue(const MCInst &MI, unsigned OpIdx,
   // style, unfortunately. As-is, we can't represent the distinct encoding
   // for #-0.
 
-  // {8}    = (U)nsigned (add == '1', sub == '0')
-  // {7-0}  = imm8
-  int32_t Imm8 = MI.getOperand(OpIdx).getImm();
-  bool isAdd = Imm8 >= 0;
+  // {Bits}    = (U)nsigned (add == '1', sub == '0')
+  // {(Bits-1)-0}  = immediate
+  int32_t Imm = MI.getOperand(OpIdx).getImm();
+  bool isAdd = Imm >= 0;
 
   // Immediate is always encoded as positive. The 'U' bit controls add vs sub.
-  if (Imm8 < 0)
-    Imm8 = -(uint32_t)Imm8;
+  if (Imm < 0)
+    Imm = -(uint32_t)Imm;
 
-  // Scaled by 4.
-  Imm8 /= 4;
+  Imm >>= Shift;
 
-  uint32_t Binary = Imm8 & 0xff;
+  uint32_t Binary = Imm & ((1U << Bits) - 1);
   // Immediate is always encoded as positive. The 'U' bit controls add vs sub.
   if (isAdd)
-    Binary |= (1 << 8);
+    Binary |= (1U << Bits);
   return Binary;
 }
-
 /// getT2AddrModeImm8s4OpValue - Return encoding info for
 /// 'reg +/- imm8<<2' operand.
 uint32_t ARMMCCodeEmitter::
@@ -967,6 +979,33 @@ getT2AddrModeImm8s4OpValue(const MCInst &MI, unsigned OpIdx,
   return Binary;
 }
 
+/// getT2AddrModeImm7s4OpValue - Return encoding info for
+/// 'reg +/- imm7<<2' operand.
+uint32_t
+ARMMCCodeEmitter::getT2AddrModeImm7s4OpValue(const MCInst &MI, unsigned OpIdx,
+                                             SmallVectorImpl<MCFixup> &Fixups,
+                                             const MCSubtargetInfo &STI) const {
+  // {11-8} = reg
+  // {7}    = (A)dd (add == '1', sub == '0')
+  // {6-0}  = imm7
+  unsigned Reg, Imm7;
+  // If The first operand isn't a register, we have a label reference.
+  bool isAdd = EncodeAddrModeOpValues(MI, OpIdx, Reg, Imm7, Fixups, STI);
+
+  // FIXME: The immediate operand should have already been encoded like this
+  // before ever getting here. The encoder method should just need to combine
+  // the MI operands for the register and the offset into a single
+  // representation for the complex operand in the .td file. This isn't just
+  // style, unfortunately. As-is, we can't represent the distinct encoding
+  // for #-0.
+  uint32_t Binary = (Imm7 >> 2) & 0xff;
+  // Immediate is always encoded as positive. The 'A' bit controls add vs sub.
+  if (isAdd)
+    Binary |= (1 << 7);
+  Binary |= (Reg << 8);
+  return Binary;
+}
+
 /// getT2AddrModeImm0_1020s4OpValue - Return encoding info for
 /// 'reg + imm8<<2' operand.
 uint32_t ARMMCCodeEmitter::
@@ -1499,7 +1538,7 @@ unsigned ARMMCCodeEmitter::
 getRegisterListOpValue(const MCInst &MI, unsigned Op,
                        SmallVectorImpl<MCFixup> &Fixups,
                        const MCSubtargetInfo &STI) const {
-  // VLDM/VSTM:
+  // VLDM/VSTM/VSCCLRM:
   //   {12-8} = Vd
   //   {7-0}  = Number of registers
   //
@@ -1508,28 +1547,40 @@ getRegisterListOpValue(const MCInst &MI, unsigned Op,
   unsigned Reg = MI.getOperand(Op).getReg();
   bool SPRRegs = ARMMCRegisterClasses[ARM::SPRRegClassID].contains(Reg);
   bool DPRRegs = ARMMCRegisterClasses[ARM::DPRRegClassID].contains(Reg);
+  bool CLRMRegs = MI.getOpcode() == ARM::t2CLRM;
 
   unsigned Binary = 0;
 
   if (SPRRegs || DPRRegs) {
-    // VLDM/VSTM
+    // VLDM/VSTM/VSCCLRM
     unsigned RegNo = CTX.getRegisterInfo()->getEncodingValue(Reg);
     unsigned NumRegs = (MI.getNumOperands() - Op) & 0xff;
     Binary |= (RegNo & 0x1f) << 8;
+
+    // Ignore VPR
+    if (MI.getOpcode() == ARM::VSCCLRMD || MI.getOpcode() == ARM::VSCCLRMS)
+      --NumRegs;
     if (SPRRegs)
       Binary |= NumRegs;
     else
       Binary |= NumRegs * 2;
   } else {
     const MCRegisterInfo &MRI = *CTX.getRegisterInfo();
-    assert(std::is_sorted(MI.begin() + Op, MI.end(),
-                          [&](const MCOperand &LHS, const MCOperand &RHS) {
-                            return MRI.getEncodingValue(LHS.getReg()) <
-                                   MRI.getEncodingValue(RHS.getReg());
-                          }));
+    if (!CLRMRegs) {
+      assert(std::is_sorted(MI.begin() + Op, MI.end(),
+                            [&](const MCOperand &LHS, const MCOperand &RHS) {
+                              return MRI.getEncodingValue(LHS.getReg()) <
+                                     MRI.getEncodingValue(RHS.getReg());
+                            }));
+    }
 
     for (unsigned I = Op, E = MI.getNumOperands(); I < E; ++I) {
-      unsigned RegNo = MRI.getEncodingValue(MI.getOperand(I).getReg());
+      unsigned RegNo;
+      if (CLRMRegs && MI.getOperand(I).getReg() == ARM::APSR) {
+        RegNo = 15;
+      } else {
+        RegNo = MRI.getEncodingValue(MI.getOperand(I).getReg());
+      }
       Binary |= 1 << RegNo;
     }
   }
@@ -1675,6 +1726,39 @@ encodeInstruction(const MCInst &MI, raw_ostream &OS,
   ++MCNumEmitted;  // Keep track of the # of mi's emitted.
 }
 
+template <bool isNeg, ARM::Fixups fixup>
+uint32_t
+ARMMCCodeEmitter::getBFTargetOpValue(const MCInst &MI, unsigned OpIdx,
+                                     SmallVectorImpl<MCFixup> &Fixups,
+                                     const MCSubtargetInfo &STI) const {
+  const MCOperand MO = MI.getOperand(OpIdx);
+  if (MO.isExpr())
+    return ::getBranchTargetOpValue(MI, OpIdx, fixup, Fixups, STI);
+  return isNeg ? -(MO.getImm() >> 1) : (MO.getImm() >> 1);
+}
+
+uint32_t
+ARMMCCodeEmitter::getBFAfterTargetOpValue(const MCInst &MI, unsigned OpIdx,
+                                          SmallVectorImpl<MCFixup> &Fixups,
+                                          const MCSubtargetInfo &STI) const {
+  const MCOperand MO = MI.getOperand(OpIdx);
+  const MCOperand BranchMO = MI.getOperand(0);
+
+  if (MO.isExpr()) {
+    assert(BranchMO.isExpr());
+    const MCExpr *DiffExpr = MCBinaryExpr::createSub(
+        MO.getExpr(), BranchMO.getExpr(), CTX);
+    MCFixupKind Kind = MCFixupKind(ARM::fixup_bfcsel_else_target);
+    Fixups.push_back(llvm::MCFixup::create(0, DiffExpr, Kind, MI.getLoc()));
+    return 0;
+  }
+
+  assert(MO.isImm() && BranchMO.isImm());
+  int Diff = MO.getImm() - BranchMO.getImm();
+  assert(Diff == 4 || Diff == 2);
+
+  return Diff == 4;
+}
 #include "ARMGenMCCodeEmitter.inc"
 
 MCCodeEmitter *llvm::createARMLEMCCodeEmitter(const MCInstrInfo &MCII,
diff --git a/llvm/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.cpp b/llvm/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.cpp
index 24491b35fa5..90022a8d88a 100644
--- a/llvm/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.cpp
+++ b/llvm/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.cpp
@@ -277,14 +277,29 @@ class ThumbMCInstrAnalysis : public ARMMCInstrAnalysis {
 public:
   ThumbMCInstrAnalysis(const MCInstrInfo *Info) : ARMMCInstrAnalysis(Info) {}
 
-  bool evaluateBranch(const MCInst &Inst, uint64_t Addr,
-                      uint64_t Size, uint64_t &Target) const override {
+  bool evaluateBranch(const MCInst &Inst, uint64_t Addr, uint64_t Size,
+                      uint64_t &Target) const override {
+    unsigned OpId;
+    switch (Inst.getOpcode()) {
+    default:
+      OpId = 0;
+      break;
+    case ARM::t2WLS:
+    case ARM::t2LEUpdate:
+      OpId = 2;
+      break;
+    case ARM::t2LE:
+      OpId = 1;
+      break;
+    }
+
     // We only handle PCRel branches for now.
-    if (Info->get(Inst.getOpcode()).OpInfo[0].OperandType!=MCOI::OPERAND_PCREL)
+    if (Info->get(Inst.getOpcode()).OpInfo[OpId].OperandType !=
+        MCOI::OPERAND_PCREL)
       return false;
 
-    int64_t Imm = Inst.getOperand(0).getImm();
-    Target = Addr+Imm+4; // In Thumb mode the PC is always off by 4 bytes.
+    // In Thumb mode the PC is always off by 4 bytes.
+    Target = Addr + Inst.getOperand(OpId).getImm() + 4;
     return true;
   }
 };