X-Git-Url: http://demsky.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FTarget%2FARM%2FARMInstrThumb.td;h=ae7a5c00bd74da3f3de34e9c65cd8a4ec863b926;hb=9ad0f4907b3ba0916a8b6cdb95d298d2ddb7d405;hp=056cbfce74ad7ed45825f69e24868822afa76b2b;hpb=aa8d1b80ff35a4370a75b9d08a3e94f19e0d3dae;p=oota-llvm.git

diff --git a/lib/Target/ARM/ARMInstrThumb.td b/lib/Target/ARM/ARMInstrThumb.td
index 056cbfce74a..ae7a5c00bd7 100644
--- a/lib/Target/ARM/ARMInstrThumb.td
+++ b/lib/Target/ARM/ARMInstrThumb.td
@@ -1,4 +1,4 @@
-//===- ARMInstrThumb.td - Thumb support for ARM ------------*- tablegen -*-===//
+//===-- ARMInstrThumb.td - Thumb support for ARM -----------*- tablegen -*-===//
 //
 //                     The LLVM Compiler Infrastructure
 //
@@ -19,25 +19,27 @@ def ARMtcall : SDNode<"ARMISD::tCALL", SDT_ARMcall,
                       [SDNPHasChain, SDNPOptInGlue, SDNPOutGlue,
                        SDNPVariadic]>;
 
-def imm_neg_XFORM : SDNodeXForm<imm, [{
-  return CurDAG->getTargetConstant(-(int)N->getZExtValue(), MVT::i32);
+def imm_sr_XFORM: SDNodeXForm<imm, [{
+  unsigned Imm = N->getZExtValue();
+  return CurDAG->getTargetConstant((Imm == 32 ? 0 : Imm), MVT::i32);
 }]>;
+def ThumbSRImmAsmOperand: AsmOperandClass { let Name = "ImmThumbSR"; }
+def imm_sr : Operand<i32>, PatLeaf<(imm), [{
+  uint64_t Imm = N->getZExtValue();
+  return Imm > 0 && Imm <= 32;
+}], imm_sr_XFORM> {
+  let PrintMethod = "printThumbSRImm";
+  let ParserMatchClass = ThumbSRImmAsmOperand;
+}
+
 def imm_comp_XFORM : SDNodeXForm<imm, [{
   return CurDAG->getTargetConstant(~((uint32_t)N->getZExtValue()), MVT::i32);
 }]>;
 
-/// imm0_7 predicate - True if the 32-bit immediate is in the range [0,7].
-def imm0_7 : ImmLeaf<i32, [{
-  return Imm >= 0 && Imm < 8;
-}]>;
 def imm0_7_neg : PatLeaf<(i32 imm), [{
   return (uint32_t)-N->getZExtValue() < 8;
 }], imm_neg_XFORM>;
 
-def imm0_255_asmoperand : AsmOperandClass { let Name = "Imm0_255"; }
-def imm0_255 : Operand<i32>, ImmLeaf<i32, [{ return Imm >= 0 && Imm < 256; }]> {
-  let ParserMatchClass = imm0_255_asmoperand;
-}
 def imm0_255_comp : PatLeaf<(i32 imm), [{
   return ~((uint32_t)N->getZExtValue()) < 256;
 }]>;
@@ -73,120 +75,155 @@ def t_adrlabel : Operand<i32> {
 }
 
 // Scaled 4 immediate.
-def t_imm_s4 : Operand<i32> {
+def t_imm0_1020s4_asmoperand: AsmOperandClass { let Name = "Imm0_1020s4"; }
+def t_imm0_1020s4 : Operand<i32> {
+  let PrintMethod = "printThumbS4ImmOperand";
+  let ParserMatchClass = t_imm0_1020s4_asmoperand;
+  let OperandType = "OPERAND_IMMEDIATE";
+}
+
+def t_imm0_508s4_asmoperand: AsmOperandClass { let Name = "Imm0_508s4"; }
+def t_imm0_508s4 : Operand<i32> {
   let PrintMethod = "printThumbS4ImmOperand";
+  let ParserMatchClass = t_imm0_508s4_asmoperand;
+  let OperandType = "OPERAND_IMMEDIATE";
+}
+// Alias use only, so no printer is necessary.
+def t_imm0_508s4_neg_asmoperand: AsmOperandClass { let Name = "Imm0_508s4Neg"; }
+def t_imm0_508s4_neg : Operand<i32> {
+  let ParserMatchClass = t_imm0_508s4_neg_asmoperand;
+  let OperandType = "OPERAND_IMMEDIATE";
 }
 
 // Define Thumb specific addressing modes.
 
+let OperandType = "OPERAND_PCREL" in {
 def t_brtarget : Operand<OtherVT> {
   let EncoderMethod = "getThumbBRTargetOpValue";
+  let DecoderMethod = "DecodeThumbBROperand";
 }
 
 def t_bcctarget : Operand<i32> {
   let EncoderMethod = "getThumbBCCTargetOpValue";
+  let DecoderMethod = "DecodeThumbBCCTargetOperand";
 }
 
 def t_cbtarget : Operand<i32> {
   let EncoderMethod = "getThumbCBTargetOpValue";
+  let DecoderMethod = "DecodeThumbCmpBROperand";
 }
 
 def t_bltarget : Operand<i32> {
   let EncoderMethod = "getThumbBLTargetOpValue";
+  let DecoderMethod = "DecodeThumbBLTargetOperand";
 }
 
 def t_blxtarget : Operand<i32> {
   let EncoderMethod = "getThumbBLXTargetOpValue";
+  let DecoderMethod = "DecodeThumbBLXOffset";
 }
-
-def MemModeRegThumbAsmOperand : AsmOperandClass {
-  let Name = "MemModeRegThumb";
-  let SuperClasses = [];
-}
-
-def MemModeImmThumbAsmOperand : AsmOperandClass {
-  let Name = "MemModeImmThumb";
-  let SuperClasses = [];
 }
 
 // t_addrmode_rr := reg + reg
 //
+def t_addrmode_rr_asm_operand : AsmOperandClass { let Name = "MemThumbRR"; }
 def t_addrmode_rr : Operand<i32>,
                     ComplexPattern<i32, 2, "SelectThumbAddrModeRR", []> {
   let EncoderMethod = "getThumbAddrModeRegRegOpValue";
   let PrintMethod = "printThumbAddrModeRROperand";
+  let DecoderMethod = "DecodeThumbAddrModeRR";
+  let ParserMatchClass = t_addrmode_rr_asm_operand;
   let MIOperandInfo = (ops tGPR:$base, tGPR:$offsreg);
 }
 
 // t_addrmode_rrs := reg + reg
 //
+// We use separate scaled versions because the Select* functions need
+// to explicitly check for a matching constant and return false here so that
+// the reg+imm forms will match instead. This is a horrible way to do that,
+// as it forces tight coupling between the methods, but it's how selectiondag
+// currently works.
 def t_addrmode_rrs1 : Operand<i32>,
                       ComplexPattern<i32, 2, "SelectThumbAddrModeRI5S1", []> {
   let EncoderMethod = "getThumbAddrModeRegRegOpValue";
   let PrintMethod = "printThumbAddrModeRROperand";
+  let DecoderMethod = "DecodeThumbAddrModeRR";
+  let ParserMatchClass = t_addrmode_rr_asm_operand;
   let MIOperandInfo = (ops tGPR:$base, tGPR:$offsreg);
-  let ParserMatchClass = MemModeRegThumbAsmOperand;
 }
 def t_addrmode_rrs2 : Operand<i32>,
                       ComplexPattern<i32, 2, "SelectThumbAddrModeRI5S2", []> {
   let EncoderMethod = "getThumbAddrModeRegRegOpValue";
+  let DecoderMethod = "DecodeThumbAddrModeRR";
   let PrintMethod = "printThumbAddrModeRROperand";
+  let ParserMatchClass = t_addrmode_rr_asm_operand;
   let MIOperandInfo = (ops tGPR:$base, tGPR:$offsreg);
-  let ParserMatchClass = MemModeRegThumbAsmOperand;
 }
 def t_addrmode_rrs4 : Operand<i32>,
                       ComplexPattern<i32, 2, "SelectThumbAddrModeRI5S4", []> {
   let EncoderMethod = "getThumbAddrModeRegRegOpValue";
+  let DecoderMethod = "DecodeThumbAddrModeRR";
   let PrintMethod = "printThumbAddrModeRROperand";
+  let ParserMatchClass = t_addrmode_rr_asm_operand;
   let MIOperandInfo = (ops tGPR:$base, tGPR:$offsreg);
-  let ParserMatchClass = MemModeRegThumbAsmOperand;
 }
 
 // t_addrmode_is4 := reg + imm5 * 4
 //
+def t_addrmode_is4_asm_operand : AsmOperandClass { let Name = "MemThumbRIs4"; }
 def t_addrmode_is4 : Operand<i32>,
                      ComplexPattern<i32, 2, "SelectThumbAddrModeImm5S4", []> {
   let EncoderMethod = "getAddrModeISOpValue";
+  let DecoderMethod = "DecodeThumbAddrModeIS";
   let PrintMethod = "printThumbAddrModeImm5S4Operand";
+  let ParserMatchClass = t_addrmode_is4_asm_operand;
   let MIOperandInfo = (ops tGPR:$base, i32imm:$offsimm);
-  let ParserMatchClass = MemModeImmThumbAsmOperand;
 }
 
 // t_addrmode_is2 := reg + imm5 * 2
 //
+def t_addrmode_is2_asm_operand : AsmOperandClass { let Name = "MemThumbRIs2"; }
 def t_addrmode_is2 : Operand<i32>,
                      ComplexPattern<i32, 2, "SelectThumbAddrModeImm5S2", []> {
   let EncoderMethod = "getAddrModeISOpValue";
+  let DecoderMethod = "DecodeThumbAddrModeIS";
   let PrintMethod = "printThumbAddrModeImm5S2Operand";
+  let ParserMatchClass = t_addrmode_is2_asm_operand;
   let MIOperandInfo = (ops tGPR:$base, i32imm:$offsimm);
-  let ParserMatchClass = MemModeImmThumbAsmOperand;
 }
 
 // t_addrmode_is1 := reg + imm5
 //
+def t_addrmode_is1_asm_operand : AsmOperandClass { let Name = "MemThumbRIs1"; }
 def t_addrmode_is1 : Operand<i32>,
                      ComplexPattern<i32, 2, "SelectThumbAddrModeImm5S1", []> {
   let EncoderMethod = "getAddrModeISOpValue";
+  let DecoderMethod = "DecodeThumbAddrModeIS";
   let PrintMethod = "printThumbAddrModeImm5S1Operand";
+  let ParserMatchClass = t_addrmode_is1_asm_operand;
   let MIOperandInfo = (ops tGPR:$base, i32imm:$offsimm);
-  let ParserMatchClass = MemModeImmThumbAsmOperand;
 }
 
 // t_addrmode_sp := sp + imm8 * 4
 //
+// FIXME: This really shouldn't have an explicit SP operand at all. It should
+// be implicit, just like in the instruction encoding itself.
+def t_addrmode_sp_asm_operand : AsmOperandClass { let Name = "MemThumbSPI"; }
 def t_addrmode_sp : Operand<i32>,
                     ComplexPattern<i32, 2, "SelectThumbAddrModeSP", []> {
   let EncoderMethod = "getAddrModeThumbSPOpValue";
+  let DecoderMethod = "DecodeThumbAddrModeSP";
   let PrintMethod = "printThumbAddrModeSPOperand";
+  let ParserMatchClass = t_addrmode_sp_asm_operand;
   let MIOperandInfo = (ops GPR:$base, i32imm:$offsimm);
-  let ParserMatchClass = MemModeImmThumbAsmOperand;
 }
 
 // t_addrmode_pc := <label> => pc + imm8 * 4
 //
 def t_addrmode_pc : Operand<i32> {
   let EncoderMethod = "getAddrModePCOpValue";
-  let ParserMatchClass = MemModeImmThumbAsmOperand;
+  let DecoderMethod = "DecodeThumbAddrModePC";
+  let PrintMethod = "printThumbLdrLabelOperand";
 }
 
 //===----------------------------------------------------------------------===//
@@ -208,68 +245,56 @@ def tADJCALLSTACKDOWN :
             Requires<[IsThumb, IsThumb1Only]>;
 }
 
-// T1Disassembly - A simple class to make encoding some disassembly patterns
-// easier and less verbose.
-class T1Disassembly<bits<2> op1, bits<8> op2>
+class T1SystemEncoding<bits<8> opc>
   : T1Encoding<0b101111> {
-  let Inst{9-8} = op1;
-  let Inst{7-0} = op2;
+  let Inst{9-8} = 0b11;
+  let Inst{7-0} = opc;
 }
 
-def tNOP : T1pI<(outs), (ins), NoItinerary, "nop", "",
-                [/* For disassembly only; pattern left blank */]>,
-           T1Disassembly<0b11, 0x00>; // A8.6.110
+def tNOP : T1pI<(outs), (ins), NoItinerary, "nop", "", []>,
+           T1SystemEncoding<0x00>, // A8.6.110
+        Requires<[IsThumb2]>;
 
-def tYIELD : T1pI<(outs), (ins), NoItinerary, "yield", "",
-                  [/* For disassembly only; pattern left blank */]>,
-           T1Disassembly<0b11, 0x10>; // A8.6.410
+def tYIELD : T1pI<(outs), (ins), NoItinerary, "yield", "", []>,
+           T1SystemEncoding<0x10>, // A8.6.410
+           Requires<[IsThumb2]>;
 
-def tWFE : T1pI<(outs), (ins), NoItinerary, "wfe", "",
-                [/* For disassembly only; pattern left blank */]>,
-           T1Disassembly<0b11, 0x20>; // A8.6.408
+def tWFE : T1pI<(outs), (ins), NoItinerary, "wfe", "", []>,
+           T1SystemEncoding<0x20>, // A8.6.408
+           Requires<[IsThumb2]>;
 
-def tWFI : T1pI<(outs), (ins), NoItinerary, "wfi", "",
-                [/* For disassembly only; pattern left blank */]>,
-           T1Disassembly<0b11, 0x30>; // A8.6.409
+def tWFI : T1pI<(outs), (ins), NoItinerary, "wfi", "", []>,
+           T1SystemEncoding<0x30>, // A8.6.409
+           Requires<[IsThumb2]>;
 
-def tSEV : T1pI<(outs), (ins), NoItinerary, "sev", "",
-                [/* For disassembly only; pattern left blank */]>,
-           T1Disassembly<0b11, 0x40>; // A8.6.157
+def tSEV : T1pI<(outs), (ins), NoItinerary, "sev", "", []>,
+           T1SystemEncoding<0x40>, // A8.6.157
+           Requires<[IsThumb2]>;
 
-// The i32imm operand $val can be used by a debugger to store more information
+// The imm operand $val can be used by a debugger to store more information
 // about the breakpoint.
-def tBKPT : T1I<(outs), (ins i32imm:$val), NoItinerary, "bkpt\t$val",
-                [/* For disassembly only; pattern left blank */]>,
-           T1Disassembly<0b10, {?,?,?,?,?,?,?,?}> {
+def tBKPT : T1I<(outs), (ins imm0_255:$val), NoItinerary, "bkpt\t$val",
+                []>,
+           T1Encoding<0b101111> {
+  let Inst{9-8} = 0b10;
   // A8.6.22
   bits<8> val;
   let Inst{7-0} = val;
 }
 
-def tSETENDBE : T1I<(outs), (ins), NoItinerary, "setend\tbe",
-                    [/* For disassembly only; pattern left blank */]>,
-                T1Encoding<0b101101> {
-  // A8.6.156
-  let Inst{9-5} = 0b10010;
-  let Inst{4}   = 1;
-  let Inst{3}   = 1;            // Big-Endian
-  let Inst{2-0} = 0b000;
-}
-
-def tSETENDLE : T1I<(outs), (ins), NoItinerary, "setend\tle",
-                    [/* For disassembly only; pattern left blank */]>,
-                T1Encoding<0b101101> {
+def tSETEND : T1I<(outs), (ins setend_op:$end), NoItinerary, "setend\t$end",
+                  []>, T1Encoding<0b101101> {
+  bits<1> end;
   // A8.6.156
   let Inst{9-5} = 0b10010;
   let Inst{4}   = 1;
-  let Inst{3}   = 0;            // Little-Endian
+  let Inst{3}   = end;
   let Inst{2-0} = 0b000;
 }
 
 // Change Processor State is a system instruction -- for disassembly only.
 def tCPS : T1I<(outs), (ins imod_op:$imod, iflags_op:$iflags),
-                NoItinerary, "cps$imod $iflags",
-                [/* For disassembly only; pattern left blank */]>,
+                NoItinerary, "cps$imod $iflags", []>,
            T1Misc<0b0110011> {
   // A8.6.38 & B6.1.1
   bit imod;
@@ -278,6 +303,7 @@ def tCPS : T1I<(outs), (ins imod_op:$imod, iflags_op:$iflags),
   let Inst{4}   = imod;
   let Inst{3}   = 0;
   let Inst{2-0} = iflags;
+  let DecoderMethod = "DecodeThumbCPS";
 }
 
 // For both thumb1 and thumb2.
@@ -291,70 +317,75 @@ def tPICADD : TIt<(outs GPR:$dst), (ins GPR:$lhs, pclabel:$cp), IIC_iALUr, "",
   let Inst{2-0} = dst;
 }
 
-// PC relative add (ADR).
-def tADDrPCi : T1I<(outs tGPR:$dst), (ins t_imm_s4:$rhs), IIC_iALUi,
-                   "add\t$dst, pc, $rhs", []>,
-               T1Encoding<{1,0,1,0,0,?}> {
-  // A6.2 & A8.6.10
-  bits<3> dst;
-  bits<8> rhs;
-  let Inst{10-8} = dst;
-  let Inst{7-0}  = rhs;
-}
-
 // ADD <Rd>, sp, #<imm8>
-// This is rematerializable, which is particularly useful for taking the
-// address of locals.
-let isReMaterializable = 1 in
-def tADDrSPi : T1I<(outs tGPR:$dst), (ins GPR:$sp, t_imm_s4:$rhs), IIC_iALUi,
-                   "add\t$dst, $sp, $rhs", []>,
+// FIXME: This should not be marked as having side effects, and it should be
+// rematerializable. Clearing the side effect bit causes miscompilations,
+// probably because the instruction can be moved around.
+def tADDrSPi : T1pI<(outs tGPR:$dst), (ins GPRsp:$sp, t_imm0_1020s4:$imm),
+                    IIC_iALUi, "add", "\t$dst, $sp, $imm", []>,
                T1Encoding<{1,0,1,0,1,?}> {
   // A6.2 & A8.6.8
   bits<3> dst;
-  bits<8> rhs;
+  bits<8> imm;
   let Inst{10-8} = dst;
-  let Inst{7-0}  = rhs;
+  let Inst{7-0}  = imm;
+  let DecoderMethod = "DecodeThumbAddSpecialReg";
 }
 
 // ADD sp, sp, #<imm7>
-def tADDspi : TIt<(outs GPR:$dst), (ins GPR:$lhs, t_imm_s4:$rhs), IIC_iALUi,
-                  "add\t$dst, $rhs", []>,
+def tADDspi : T1pIt<(outs GPRsp:$Rdn), (ins GPRsp:$Rn, t_imm0_508s4:$imm),
+                     IIC_iALUi, "add", "\t$Rdn, $imm", []>,
               T1Misc<{0,0,0,0,0,?,?}> {
   // A6.2.5 & A8.6.8
-  bits<7> rhs;
-  let Inst{6-0} = rhs;
+  bits<7> imm;
+  let Inst{6-0} = imm;
+  let DecoderMethod = "DecodeThumbAddSPImm";
 }
 
 // SUB sp, sp, #<imm7>
 // FIXME: The encoding and the ASM string don't match up.
-def tSUBspi : TIt<(outs GPR:$dst), (ins GPR:$lhs, t_imm_s4:$rhs), IIC_iALUi,
-                  "sub\t$dst, $rhs", []>,
+def tSUBspi : T1pIt<(outs GPRsp:$Rdn), (ins GPRsp:$Rn, t_imm0_508s4:$imm),
+                    IIC_iALUi, "sub", "\t$Rdn, $imm", []>,
               T1Misc<{0,0,0,0,1,?,?}> {
   // A6.2.5 & A8.6.214
-  bits<7> rhs;
-  let Inst{6-0} = rhs;
+  bits<7> imm;
+  let Inst{6-0} = imm;
+  let DecoderMethod = "DecodeThumbAddSPImm";
 }
 
+def : tInstAlias<"add${p} sp, $imm",
+                 (tSUBspi SP, t_imm0_508s4_neg:$imm, pred:$p)>;
+def : tInstAlias<"add${p} sp, sp, $imm",
+                 (tSUBspi SP, t_imm0_508s4_neg:$imm, pred:$p)>;
+
+// Can optionally specify SP as a three operand instruction.
+def : tInstAlias<"add${p} sp, sp, $imm",
+                 (tADDspi SP, t_imm0_508s4:$imm, pred:$p)>;
+def : tInstAlias<"sub${p} sp, sp, $imm",
+                 (tSUBspi SP, t_imm0_508s4:$imm, pred:$p)>;
+
 // ADD <Rm>, sp
-def tADDrSP : TIt<(outs GPR:$dst), (ins GPR:$lhs, GPR:$rhs), IIC_iALUr,
-                  "add\t$dst, $rhs", []>,
+def tADDrSP : T1pI<(outs GPR:$Rdn), (ins GPRsp:$sp, GPR:$Rn), IIC_iALUr,
+                   "add", "\t$Rdn, $sp, $Rn", []>,
               T1Special<{0,0,?,?}> {
   // A8.6.9 Encoding T1
-  bits<4> dst;
-  let Inst{7}   = dst{3};
+  bits<4> Rdn;
+  let Inst{7}   = Rdn{3};
   let Inst{6-3} = 0b1101;
-  let Inst{2-0} = dst{2-0};
+  let Inst{2-0} = Rdn{2-0};
+  let DecoderMethod = "DecodeThumbAddSPReg";
 }
 
 // ADD sp, <Rm>
-def tADDspr : TIt<(outs GPR:$dst), (ins GPR:$lhs, GPR:$rhs), IIC_iALUr,
-                  "add\t$dst, $rhs", []>,
+def tADDspr : T1pIt<(outs GPRsp:$Rdn), (ins GPRsp:$Rn, GPR:$Rm), IIC_iALUr,
+                  "add", "\t$Rdn, $Rm", []>,
               T1Special<{0,0,?,?}> {
   // A8.6.9 Encoding T2
-  bits<4> dst;
+  bits<4> Rm;
   let Inst{7} = 1;
-  let Inst{6-3} = dst;
+  let Inst{6-3} = Rm;
   let Inst{2-0} = 0b101;
+  let DecoderMethod = "DecodeThumbAddSPReg";
 }
 
 //===----------------------------------------------------------------------===//
@@ -369,16 +400,17 @@ let isBranch = 1, isTerminator = 1, isBarrier = 1, isIndirectBranch = 1 in {
     bits<4> Rm;
     let Inst{6-3} = Rm;
     let Inst{2-0} = 0b000;
+    let Unpredictable{2-0} = 0b111;
   }
 }
 
 let isReturn = 1, isTerminator = 1, isBarrier = 1 in {
-  def tBX_RET : tPseudoExpand<(outs), (ins pred:$p), Size2Bytes, IIC_Br,
+  def tBX_RET : tPseudoExpand<(outs), (ins pred:$p), 2, IIC_Br,
                    [(ARMretflag)], (tBX LR, pred:$p)>;
 
   // Alternative return instruction used by vararg functions.
   def tBX_RET_vararg : tPseudoExpand<(outs), (ins tGPR:$Rm, pred:$p),
-                   Size2Bytes, IIC_Br, [],
+                   2, IIC_Br, [],
                    (tBX GPR:$Rm, pred:$p)>;
 }
 
@@ -386,41 +418,41 @@ let isReturn = 1, isTerminator = 1, isBarrier = 1 in {
 // prevent stack-pointer assignments that appear immediately before calls from
 // potentially appearing dead.
 let isCall = 1,
-  // On non-Darwin platforms R9 is callee-saved.
-  Defs = [R0,  R1,  R2,  R3,  R12, LR, QQQQ0, QQQQ2, QQQQ3, CPSR, FPSCR],
-  Uses = [SP] in {
+  Defs = [LR], Uses = [SP] in {
   // Also used for Thumb2
   def tBL  : TIx2<0b11110, 0b11, 1,
-                  (outs), (ins t_bltarget:$func, variable_ops), IIC_Br,
-                  "bl\t$func",
+                  (outs), (ins pred:$p, t_bltarget:$func), IIC_Br,
+                  "bl${p}\t$func",
                   [(ARMtcall tglobaladdr:$func)]>,
-             Requires<[IsThumb, IsNotDarwin]> {
-    bits<21> func;
+             Requires<[IsThumb]> {
+    bits<24> func;
+    let Inst{26} = func{23};
     let Inst{25-16} = func{20-11};
-    let Inst{13} = 1;
-    let Inst{11} = 1;
+    let Inst{13} = func{22};
+    let Inst{11} = func{21};
     let Inst{10-0} = func{10-0};
   }
 
   // ARMv5T and above, also used for Thumb2
   def tBLXi : TIx2<0b11110, 0b11, 0,
-                   (outs), (ins t_blxtarget:$func, variable_ops), IIC_Br,
-                   "blx\t$func",
+                 (outs), (ins pred:$p, t_blxtarget:$func), IIC_Br,
+                   "blx${p}\t$func",
                    [(ARMcall tglobaladdr:$func)]>,
-              Requires<[IsThumb, HasV5T, IsNotDarwin]> {
-    bits<21> func;
+              Requires<[IsThumb, HasV5T]> {
+    bits<24> func;
+    let Inst{26} = func{23};
     let Inst{25-16} = func{20-11};
-    let Inst{13} = 1;
-    let Inst{11} = 1;
+    let Inst{13} = func{22};
+    let Inst{11} = func{21};
     let Inst{10-1} = func{10-1};
     let Inst{0} = 0; // func{0} is assumed zero
   }
 
   // Also used for Thumb2
-  def tBLXr : TI<(outs), (ins GPR:$func, variable_ops), IIC_Br,
-                  "blx\t$func",
+  def tBLXr : TI<(outs), (ins pred:$p, GPR:$func), IIC_Br,
+                  "blx${p}\t$func",
                   [(ARMtcall GPR:$func)]>,
-              Requires<[IsThumb, HasV5T, IsNotDarwin]>,
+              Requires<[IsThumb, HasV5T]>,
               T1Special<{1,1,1,?}> { // A6.2.3 & A8.6.24;
     bits<4> func;
     let Inst{6-3} = func;
@@ -428,68 +460,16 @@ let isCall = 1,
   }
 
   // ARMv4T
-  def tBX_CALL : tPseudoInst<(outs), (ins tGPR:$func, variable_ops),
-                  Size4Bytes, IIC_Br,
+  def tBX_CALL : tPseudoInst<(outs), (ins tGPR:$func),
+                  4, IIC_Br,
                   [(ARMcall_nolink tGPR:$func)]>,
-            Requires<[IsThumb, IsThumb1Only, IsNotDarwin]>;
-}
-
-let isCall = 1,
-  // On Darwin R9 is call-clobbered.
-  // R7 is marked as a use to prevent frame-pointer assignments from being
-  // moved above / below calls.
-  Defs = [R0,  R1,  R2,  R3,  R9,  R12, LR, QQQQ0, QQQQ2, QQQQ3, CPSR, FPSCR],
-  Uses = [R7, SP] in {
-  // Also used for Thumb2
-  def tBLr9 : TIx2<0b11110, 0b11, 1,
-                   (outs), (ins pred:$p, t_bltarget:$func, variable_ops),
-                   IIC_Br, "bl${p}\t$func",
-                   [(ARMtcall tglobaladdr:$func)]>,
-              Requires<[IsThumb, IsDarwin]> {
-    bits<21> func;
-    let Inst{25-16} = func{20-11};
-    let Inst{13} = 1;
-    let Inst{11} = 1;
-    let Inst{10-0} = func{10-0};
-  }
-
-  // ARMv5T and above, also used for Thumb2
-  def tBLXi_r9 : TIx2<0b11110, 0b11, 0,
-                      (outs), (ins pred:$p, t_blxtarget:$func, variable_ops),
-                      IIC_Br, "blx${p}\t$func",
-                      [(ARMcall tglobaladdr:$func)]>,
-                 Requires<[IsThumb, HasV5T, IsDarwin]> {
-    bits<21> func;
-    let Inst{25-16} = func{20-11};
-    let Inst{13} = 1;
-    let Inst{11} = 1;
-    let Inst{10-1} = func{10-1};
-    let Inst{0} = 0; // func{0} is assumed zero
-  }
-
-  // Also used for Thumb2
-  def tBLXr_r9 : TI<(outs), (ins pred:$p, GPR:$func, variable_ops), IIC_Br,
-                    "blx${p}\t$func",
-                    [(ARMtcall GPR:$func)]>,
-                 Requires<[IsThumb, HasV5T, IsDarwin]>,
-                 T1Special<{1,1,1,?}> {
-    // A6.2.3 & A8.6.24
-    bits<4> func;
-    let Inst{6-3} = func;
-    let Inst{2-0} = 0b000;
-  }
-
-  // ARMv4T
-  def tBXr9_CALL : tPseudoInst<(outs), (ins tGPR:$func, variable_ops),
-                   Size4Bytes, IIC_Br,
-                   [(ARMcall_nolink tGPR:$func)]>,
-              Requires<[IsThumb, IsThumb1Only, IsDarwin]>;
+            Requires<[IsThumb, IsThumb1Only]>;
 }
 
 let isBranch = 1, isTerminator = 1, isBarrier = 1 in {
   let isPredicable = 1 in
-  def tB   : T1I<(outs), (ins t_brtarget:$target), IIC_Br,
-                 "b\t$target", [(br bb:$target)]>,
+  def tB   : T1pI<(outs), (ins t_brtarget:$target), IIC_Br,
+                 "b", "\t$target", [(br bb:$target)]>,
              T1Encoding<{1,1,1,0,0,?}> {
     bits<11> target;
     let Inst{10-0} = target;
@@ -499,12 +479,12 @@ let isBranch = 1, isTerminator = 1, isBarrier = 1 in {
   // Just a pseudo for a tBL instruction. Needed to let regalloc know about
   // the clobber of LR.
   let Defs = [LR] in
-  def tBfar : tPseudoExpand<(outs), (ins t_bltarget:$target),
-                          Size4Bytes, IIC_Br, [], (tBL t_bltarget:$target)>;
+  def tBfar : tPseudoExpand<(outs), (ins t_bltarget:$target, pred:$p),
+                          4, IIC_Br, [], (tBL pred:$p, t_bltarget:$target)>;
 
   def tBR_JTr : tPseudoInst<(outs),
                       (ins tGPR:$target, i32imm:$jt, i32imm:$id),
-                      SizeSpecial, IIC_Br,
+                      0, IIC_Br,
                       [(ARMbrjt tGPR:$target, tjumptable:$jt, imm:$id)]> {
     list<Predicate> Predicates = [IsThumb, IsThumb1Only];
   }
@@ -523,63 +503,33 @@ let isBranch = 1, isTerminator = 1 in
   let Inst{7-0} = target;
 }
 
-// Compare and branch on zero / non-zero
-let isBranch = 1, isTerminator = 1 in {
-  def tCBZ  : T1I<(outs), (ins tGPR:$Rn, t_cbtarget:$target), IIC_Br,
-                  "cbz\t$Rn, $target", []>,
-              T1Misc<{0,0,?,1,?,?,?}> {
-    // A8.6.27
-    bits<6> target;
-    bits<3> Rn;
-    let Inst{9}   = target{5};
-    let Inst{7-3} = target{4-0};
-    let Inst{2-0} = Rn;
-  }
-
-  def tCBNZ : T1I<(outs), (ins tGPR:$cmp, t_cbtarget:$target), IIC_Br,
-                  "cbnz\t$cmp, $target", []>,
-              T1Misc<{1,0,?,1,?,?,?}> {
-    // A8.6.27
-    bits<6> target;
-    bits<3> Rn;
-    let Inst{9}   = target{5};
-    let Inst{7-3} = target{4-0};
-    let Inst{2-0} = Rn;
-  }
-}
-
 // Tail calls
 let isCall = 1, isTerminator = 1, isReturn = 1, isBarrier = 1 in {
-  // Darwin versions.
-  let Defs = [R0, R1, R2, R3, R9, R12, QQQQ0, QQQQ2, QQQQ3, PC],
-      Uses = [SP] in {
-    // tTAILJMPd: Darwin version uses a Thumb2 branch (no Thumb1 tail calls
-    // on Darwin), so it's in ARMInstrThumb2.td.
-    def tTAILJMPr : tPseudoExpand<(outs), (ins tcGPR:$dst, variable_ops),
-                     Size4Bytes, IIC_Br, [],
+  // IOS versions.
+  let Uses = [SP] in {
+    def tTAILJMPr : tPseudoExpand<(outs), (ins tcGPR:$dst),
+                     4, IIC_Br, [],
                      (tBX GPR:$dst, (ops 14, zero_reg))>,
-                     Requires<[IsThumb, IsDarwin]>;
+                     Requires<[IsThumb]>;
   }
-  // Non-Darwin versions (the difference is R9).
-  let Defs = [R0, R1, R2, R3, R12, QQQQ0, QQQQ2, QQQQ3, PC],
-      Uses = [SP] in {
-    def tTAILJMPdND : tPseudoExpand<(outs), (ins t_brtarget:$dst, variable_ops),
-                   Size4Bytes, IIC_Br, [],
-                   (tB t_brtarget:$dst)>,
-                 Requires<[IsThumb, IsNotDarwin]>;
-    def tTAILJMPrND : tPseudoExpand<(outs), (ins tcGPR:$dst, variable_ops),
-                     Size4Bytes, IIC_Br, [],
-                     (tBX GPR:$dst, (ops 14, zero_reg))>,
-                     Requires<[IsThumb, IsNotDarwin]>;
+  // tTAILJMPd: IOS version uses a Thumb2 branch (no Thumb1 tail calls
+  // on IOS), so it's in ARMInstrThumb2.td.
+  // Non-IOS version:
+  let Uses = [SP] in {
+    def tTAILJMPdND : tPseudoExpand<(outs),
+                   (ins t_brtarget:$dst, pred:$p),
+                   4, IIC_Br, [],
+                   (tB t_brtarget:$dst, pred:$p)>,
+                 Requires<[IsThumb, IsNotIOS]>;
   }
 }
 
 
-// A8.6.218 Supervisor Call (Software Interrupt) -- for disassembly only
+// A8.6.218 Supervisor Call (Software Interrupt)
 // A8.6.16 B: Encoding T1
 // If Inst{11-8} == 0b1111 then SEE SVC
 let isCall = 1, Uses = [SP] in
-def tSVC : T1pI<(outs), (ins i32imm:$imm), IIC_Br,
+def tSVC : T1pI<(outs), (ins imm0_255:$imm), IIC_Br,
                 "svc", "\t$imm", []>, Encoding16 {
   bits<8> imm;
   let Inst{15-12} = 0b1101;
@@ -654,17 +604,17 @@ defm tLDRH : thumb_ld_rr_ri_enc<0b101, 0b1000, t_addrmode_rrs2,
 
 let AddedComplexity = 10 in
 def tLDRSB :                    // A8.6.80
-  T1pILdStEncode<0b011, (outs tGPR:$dst), (ins t_addrmode_rr:$addr),
+  T1pILdStEncode<0b011, (outs tGPR:$Rt), (ins t_addrmode_rr:$addr),
                  AddrModeT1_1, IIC_iLoad_bh_r,
-                 "ldrsb", "\t$dst, $addr",
-                 [(set tGPR:$dst, (sextloadi8 t_addrmode_rr:$addr))]>;
+                 "ldrsb", "\t$Rt, $addr",
+                 [(set tGPR:$Rt, (sextloadi8 t_addrmode_rr:$addr))]>;
 
 let AddedComplexity = 10 in
 def tLDRSH :                    // A8.6.84
-  T1pILdStEncode<0b111, (outs tGPR:$dst), (ins t_addrmode_rr:$addr),
+  T1pILdStEncode<0b111, (outs tGPR:$Rt), (ins t_addrmode_rr:$addr),
                  AddrModeT1_2, IIC_iLoad_bh_r,
-                 "ldrsh", "\t$dst, $addr",
-                 [(set tGPR:$dst, (sextloadi16 t_addrmode_rr:$addr))]>;
+                 "ldrsh", "\t$Rt, $addr",
+                 [(set tGPR:$Rt, (sextloadi16 t_addrmode_rr:$addr))]>;
 
 let canFoldAsLoad = 1 in
 def tLDRspi : T1pIs<(outs tGPR:$Rt), (ins t_addrmode_sp:$addr), IIC_iLoad_i,
@@ -678,8 +628,8 @@ def tLDRspi : T1pIs<(outs tGPR:$Rt), (ins t_addrmode_sp:$addr), IIC_iLoad_i,
 }
 
 // Load tconstpool
-// FIXME: Use ldr.n to work around a Darwin assembler bug.
-let canFoldAsLoad = 1, isReMaterializable = 1 in
+// FIXME: Use ldr.n to work around a darwin assembler bug.
+let canFoldAsLoad = 1, isReMaterializable = 1, isCodeGenOnly = 1 in
 def tLDRpci : T1pIs<(outs tGPR:$Rt), (ins t_addrmode_pc:$addr), IIC_iLoad_i,
                   "ldr", ".n\t$Rt, $addr",
                   [(set tGPR:$Rt, (load (ARMWrapper tconstpool:$addr)))]>,
@@ -692,10 +642,9 @@ def tLDRpci : T1pIs<(outs tGPR:$Rt), (ins t_addrmode_pc:$addr), IIC_iLoad_i,
 }
 
 // FIXME: Remove this entry when the above ldr.n workaround is fixed.
-// For disassembly use only.
-def tLDRpciDIS : T1pIs<(outs tGPR:$Rt), (ins t_addrmode_pc:$addr), IIC_iLoad_i,
-                       "ldr", "\t$Rt, $addr",
-                       [/* disassembly only */]>,
+// For assembly/disassembly use only.
+def tLDRpciASM : T1pIs<(outs tGPR:$Rt), (ins t_addrmode_pc:$addr), IIC_iLoad_i,
+                       "ldr", "\t$Rt, $addr", []>,
                  T1Encoding<{0,1,0,0,1,?}> {
   // A6.2 & A8.6.59
   bits<3> Rt;
@@ -737,42 +686,53 @@ def tSTRspi : T1pIs<(outs), (ins tGPR:$Rt, t_addrmode_sp:$addr), IIC_iStore_i,
 //  Load / store multiple Instructions.
 //
 
-multiclass thumb_ldst_mult<string asm, InstrItinClass itin,
-                           InstrItinClass itin_upd, bits<6> T1Enc,
-                           bit L_bit> {
-  def IA :
-    T1I<(outs), (ins GPR:$Rn, pred:$p, reglist:$regs, variable_ops),
-        itin, !strconcat(asm, "ia${p}\t$Rn, $regs"), []>,
-       T1Encoding<T1Enc> {
-    bits<3> Rn;
-    bits<8> regs;
-    let Inst{10-8} = Rn;
-    let Inst{7-0}  = regs;
-  }
-  def IA_UPD :
-    T1It<(outs GPR:$wb), (ins GPR:$Rn, pred:$p, reglist:$regs, variable_ops),
-         itin_upd, !strconcat(asm, "ia${p}\t$Rn!, $regs"), "$Rn = $wb", []>,
-        T1Encoding<T1Enc> {
-    bits<3> Rn;
-    bits<8> regs;
-    let Inst{10-8} = Rn;
-    let Inst{7-0}  = regs;
-  }
-}
-
 // These require base address to be written back or one of the loaded regs.
 let neverHasSideEffects = 1 in {
 
 let mayLoad = 1, hasExtraDefRegAllocReq = 1 in
-defm tLDM : thumb_ldst_mult<"ldm", IIC_iLoad_m, IIC_iLoad_mu,
-                            {1,1,0,0,1,?}, 1>;
-
+def tLDMIA : T1I<(outs), (ins tGPR:$Rn, pred:$p, reglist:$regs, variable_ops),
+        IIC_iLoad_m, "ldm${p}\t$Rn, $regs", []>, T1Encoding<{1,1,0,0,1,?}> {
+  bits<3> Rn;
+  bits<8> regs;
+  let Inst{10-8} = Rn;
+  let Inst{7-0}  = regs;
+}
+
+// Writeback version is just a pseudo, as there's no encoding difference.
+// Writeback happens iff the base register is not in the destination register
+// list.
+def tLDMIA_UPD :
+    InstTemplate<AddrModeNone, 0, IndexModeNone, Pseudo, GenericDomain,
+                 "$Rn = $wb", IIC_iLoad_mu>,
+    PseudoInstExpansion<(tLDMIA tGPR:$Rn, pred:$p, reglist:$regs)> {
+  let Size = 2;
+  let OutOperandList = (outs GPR:$wb);
+  let InOperandList = (ins GPR:$Rn, pred:$p, reglist:$regs, variable_ops);
+  let Pattern = [];
+  let isCodeGenOnly = 1;
+  let isPseudo = 1;
+  list<Predicate> Predicates = [IsThumb];
+}
+
+// There is no non-writeback version of STM for Thumb.
 let mayStore = 1, hasExtraSrcRegAllocReq = 1 in
-defm tSTM : thumb_ldst_mult<"stm", IIC_iStore_m, IIC_iStore_mu,
-                            {1,1,0,0,0,?}, 0>;
+def tSTMIA_UPD : Thumb1I<(outs GPR:$wb),
+                         (ins tGPR:$Rn, pred:$p, reglist:$regs, variable_ops),
+                         AddrModeNone, 2, IIC_iStore_mu,
+                         "stm${p}\t$Rn!, $regs", "$Rn = $wb", []>,
+                     T1Encoding<{1,1,0,0,0,?}> {
+  bits<3> Rn;
+  bits<8> regs;
+  let Inst{10-8} = Rn;
+  let Inst{7-0}  = regs;
+}
 
 } // neverHasSideEffects
 
+def : InstAlias<"ldm${p} $Rn!, $regs",
+                (tLDMIA tGPR:$Rn, pred:$p, reglist:$regs)>,
+        Requires<[IsThumb, IsThumb1Only]>;
+
 let mayLoad = 1, Uses = [SP], Defs = [SP], hasExtraDefRegAllocReq = 1 in
 def tPOP : T1I<(outs), (ins pred:$p, reglist:$regs, variable_ops),
                IIC_iPop,
@@ -877,7 +837,7 @@ def tADC :                      // A8.6.2
 
 // Add immediate
 def tADDi3 :                    // A8.6.4 T1
-  T1sIGenEncodeImm<0b01110, (outs tGPR:$Rd), (ins tGPR:$Rm, i32imm:$imm3),
+  T1sIGenEncodeImm<0b01110, (outs tGPR:$Rd), (ins tGPR:$Rm, imm0_7:$imm3),
                    IIC_iALUi,
                    "add", "\t$Rd, $Rm, $imm3",
                    [(set tGPR:$Rd, (add tGPR:$Rm, imm0_7:$imm3))]> {
@@ -886,8 +846,8 @@ def tADDi3 :                    // A8.6.4 T1
 }
 
 def tADDi8 :                    // A8.6.4 T2
-  T1sItGenEncodeImm<{1,1,0,?,?}, (outs tGPR:$Rdn), (ins tGPR:$Rn, i32imm:$imm8),
-                    IIC_iALUi,
+  T1sItGenEncodeImm<{1,1,0,?,?}, (outs tGPR:$Rdn),
+                    (ins tGPR:$Rn, imm0_255:$imm8), IIC_iALUi,
                     "add", "\t$Rdn, $imm8",
                     [(set tGPR:$Rdn, (add tGPR:$Rn, imm8_255:$imm8))]>;
 
@@ -921,10 +881,10 @@ def tAND :                      // A8.6.12
 
 // ASR immediate
 def tASRri :                    // A8.6.14
-  T1sIGenEncodeImm<{0,1,0,?,?}, (outs tGPR:$Rd), (ins tGPR:$Rm, i32imm:$imm5),
+  T1sIGenEncodeImm<{0,1,0,?,?}, (outs tGPR:$Rd), (ins tGPR:$Rm, imm_sr:$imm5),
                    IIC_iMOVsi,
                    "asr", "\t$Rd, $Rm, $imm5",
-                   [(set tGPR:$Rd, (sra tGPR:$Rm, (i32 imm:$imm5)))]> {
+                   [(set tGPR:$Rd, (sra tGPR:$Rm, (i32 imm_sr:$imm5)))]> {
   bits<5> imm5;
   let Inst{10-6} = imm5;
 }
@@ -963,7 +923,7 @@ def tCMNz :                     // A8.6.33
 
 // CMP immediate
 let isCompare = 1, Defs = [CPSR] in {
-def tCMPi8 : T1pI<(outs), (ins tGPR:$Rn, i32imm:$imm8), IIC_iCMPi,
+def tCMPi8 : T1pI<(outs), (ins tGPR:$Rn, imm0_255:$imm8), IIC_iCMPi,
                   "cmp", "\t$Rn, $imm8",
                   [(ARMcmp tGPR:$Rn, imm0_255:$imm8)]>,
              T1General<{1,0,1,?,?}> {
@@ -1004,7 +964,7 @@ def tEOR :                      // A8.6.45
 
 // LSL immediate
 def tLSLri :                    // A8.6.88
-  T1sIGenEncodeImm<{0,0,0,?,?}, (outs tGPR:$Rd), (ins tGPR:$Rm, i32imm:$imm5),
+  T1sIGenEncodeImm<{0,0,0,?,?}, (outs tGPR:$Rd), (ins tGPR:$Rm, imm0_31:$imm5),
                    IIC_iMOVsi,
                    "lsl", "\t$Rd, $Rm, $imm5",
                    [(set tGPR:$Rd, (shl tGPR:$Rm, (i32 imm:$imm5)))]> {
@@ -1021,10 +981,10 @@ def tLSLrr :                    // A8.6.89
 
 // LSR immediate
 def tLSRri :                    // A8.6.90
-  T1sIGenEncodeImm<{0,0,1,?,?}, (outs tGPR:$Rd), (ins tGPR:$Rm, i32imm:$imm5),
+  T1sIGenEncodeImm<{0,0,1,?,?}, (outs tGPR:$Rd), (ins tGPR:$Rm, imm_sr:$imm5),
                    IIC_iMOVsi,
                    "lsr", "\t$Rd, $Rm, $imm5",
-                   [(set tGPR:$Rd, (srl tGPR:$Rm, (i32 imm:$imm5)))]> {
+                   [(set tGPR:$Rd, (srl tGPR:$Rm, (i32 imm_sr:$imm5)))]> {
   bits<5> imm5;
   let Inst{10-6} = imm5;
 }
@@ -1048,12 +1008,16 @@ def tMOVi8 : T1sI<(outs tGPR:$Rd), (ins imm0_255:$imm8), IIC_iMOVi,
   let Inst{10-8} = Rd;
   let Inst{7-0}  = imm8;
 }
+// Because we have an explicit tMOVSr below, we need an alias to handle
+// the immediate "movs" form here. Blech.
+def : tInstAlias <"movs $Rdn, $imm",
+                 (tMOVi8 tGPR:$Rdn, CPSR, imm0_255:$imm, 14, 0)>;
 
 // A7-73: MOV(2) - mov setting flag.
 
 let neverHasSideEffects = 1 in {
 def tMOVr : Thumb1pI<(outs GPR:$Rd), (ins GPR:$Rm), AddrModeNone,
-                      Size2Bytes, IIC_iMOVr,
+                      2, IIC_iMOVr,
                       "mov", "\t$Rd, $Rm", "", []>,
                   T1Special<{1,0,?,?}> {
   // A8.6.97
@@ -1078,10 +1042,19 @@ def tMOVSr      : T1I<(outs tGPR:$Rd), (ins tGPR:$Rm), IIC_iMOVr,
 // Multiply register
 let isCommutable = 1 in
 def tMUL :                      // A8.6.105 T1
-  T1sItDPEncode<0b1101, (outs tGPR:$Rdn), (ins tGPR:$Rn, tGPR:$Rm),
-                IIC_iMUL32,
-                "mul", "\t$Rdn, $Rm, $Rdn",
-                [(set tGPR:$Rdn, (mul tGPR:$Rn, tGPR:$Rm))]>;
+  Thumb1sI<(outs tGPR:$Rd), (ins tGPR:$Rn, tGPR:$Rm), AddrModeNone, 2,
+           IIC_iMUL32, "mul", "\t$Rd, $Rn, $Rm", "$Rm = $Rd",
+           [(set tGPR:$Rd, (mul tGPR:$Rn, tGPR:$Rm))]>,
+      T1DataProcessing<0b1101> {
+  bits<3> Rd;
+  bits<3> Rn;
+  let Inst{5-3} = Rn;
+  let Inst{2-0} = Rd;
+  let AsmMatchConverter = "cvtThumbMultiply";
+}
+
+def :tInstAlias<"mul${s}${p} $Rdm, $Rn", (tMUL tGPR:$Rdm, s_cc_out:$s, tGPR:$Rn,
+                                               pred:$p)>;
 
 // Move inverse register
 def tMVN :                      // A8.6.107
@@ -1143,7 +1116,7 @@ def tSBC :                      // A8.6.151
 
 // Subtract immediate
 def tSUBi3 :                    // A8.6.210 T1
-  T1sIGenEncodeImm<0b01111, (outs tGPR:$Rd), (ins tGPR:$Rm, i32imm:$imm3),
+  T1sIGenEncodeImm<0b01111, (outs tGPR:$Rd), (ins tGPR:$Rm, imm0_7:$imm3),
                    IIC_iALUi,
                    "sub", "\t$Rd, $Rm, $imm3",
                    [(set tGPR:$Rd, (add tGPR:$Rm, imm0_7_neg:$imm3))]> {
@@ -1152,8 +1125,8 @@ def tSUBi3 :                    // A8.6.210 T1
 }
 
 def tSUBi8 :                    // A8.6.210 T2
-  T1sItGenEncodeImm<{1,1,1,?,?}, (outs tGPR:$Rdn), (ins tGPR:$Rn, i32imm:$imm8),
-                    IIC_iALUi,
+  T1sItGenEncodeImm<{1,1,1,?,?}, (outs tGPR:$Rdn),
+                    (ins tGPR:$Rn, imm0_255:$imm8), IIC_iALUi,
                     "sub", "\t$Rdn, $imm8",
                     [(set tGPR:$Rdn, (add tGPR:$Rn, imm8_255_neg:$imm8))]>;
 
@@ -1164,8 +1137,6 @@ def tSUBrr :                    // A8.6.212
                 "sub", "\t$Rd, $Rn, $Rm",
                 [(set tGPR:$Rd, (sub tGPR:$Rn, tGPR:$Rm))]>;
 
-// TODO: A7-96: STMIA - store multiple.
-
 // Sign-extend byte
 def tSXTB :                     // A8.6.222
   T1pIMiscEncode<{0,0,1,0,0,1,?}, (outs tGPR:$Rd), (ins tGPR:$Rm),
@@ -1217,115 +1188,23 @@ let usesCustomInserter = 1 in  // Expanded after instruction selection.
 // assembler.
 
 def tADR : T1I<(outs tGPR:$Rd), (ins t_adrlabel:$addr, pred:$p),
-               IIC_iALUi, "adr{$p}\t$Rd, #$addr", []>,
+               IIC_iALUi, "adr{$p}\t$Rd, $addr", []>,
                T1Encoding<{1,0,1,0,0,?}> {
   bits<3> Rd;
   bits<8> addr;
   let Inst{10-8} = Rd;
   let Inst{7-0} = addr;
+  let DecoderMethod = "DecodeThumbAddSpecialReg";
 }
 
 let neverHasSideEffects = 1, isReMaterializable = 1 in
 def tLEApcrel   : tPseudoInst<(outs tGPR:$Rd), (ins i32imm:$label, pred:$p),
-                              Size2Bytes, IIC_iALUi, []>;
+                              2, IIC_iALUi, []>;
 
+let hasSideEffects = 1 in
 def tLEApcrelJT : tPseudoInst<(outs tGPR:$Rd),
                               (ins i32imm:$label, nohash_imm:$id, pred:$p),
-                              Size2Bytes, IIC_iALUi, []>;
-
-//===----------------------------------------------------------------------===//
-// Move between coprocessor and ARM core register -- for disassembly only
-//
-
-class tMovRCopro<string opc, bit direction, dag oops, dag iops,
-                 list<dag> pattern>
-  : T1Cop<oops, iops, !strconcat(opc, "\t$cop, $opc1, $Rt, $CRn, $CRm, $opc2"),
-          pattern> {
-  let Inst{27-24} = 0b1110;
-  let Inst{20} = direction;
-  let Inst{4} = 1;
-
-  bits<4> Rt;
-  bits<4> cop;
-  bits<3> opc1;
-  bits<3> opc2;
-  bits<4> CRm;
-  bits<4> CRn;
-
-  let Inst{15-12} = Rt;
-  let Inst{11-8}  = cop;
-  let Inst{23-21} = opc1;
-  let Inst{7-5}   = opc2;
-  let Inst{3-0}   = CRm;
-  let Inst{19-16} = CRn;
-}
-
-def tMCR : tMovRCopro<"mcr", 0 /* from ARM core register to coprocessor */,
-           (outs),
-           (ins p_imm:$cop, i32imm:$opc1, GPR:$Rt, c_imm:$CRn,
-                c_imm:$CRm, i32imm:$opc2),
-           [(int_arm_mcr imm:$cop, imm:$opc1, GPR:$Rt, imm:$CRn,
-                         imm:$CRm, imm:$opc2)]>;
-def tMRC : tMovRCopro<"mrc", 1 /* from coprocessor to ARM core register */,
-           (outs GPR:$Rt),
-           (ins p_imm:$cop, i32imm:$opc1, c_imm:$CRn, c_imm:$CRm, i32imm:$opc2),
-           []>;
-
-def : Pat<(int_arm_mrc imm:$cop, imm:$opc1, imm:$CRn, imm:$CRm, imm:$opc2),
-          (tMRC imm:$cop, imm:$opc1, imm:$CRn, imm:$CRm, imm:$opc2)>,
-          Requires<[IsThumb, HasV6T2]>;
-
-class tMovRRCopro<string opc, bit direction,
-                  list<dag> pattern = [/* For disassembly only */]>
-  : T1Cop<(outs), (ins p_imm:$cop, i32imm:$opc1, GPR:$Rt, GPR:$Rt2, c_imm:$CRm),
-          !strconcat(opc, "\t$cop, $opc1, $Rt, $Rt2, $CRm"), pattern> {
-  let Inst{27-24} = 0b1100;
-  let Inst{23-21} = 0b010;
-  let Inst{20} = direction;
-
-  bits<4> Rt;
-  bits<4> Rt2;
-  bits<4> cop;
-  bits<4> opc1;
-  bits<4> CRm;
-
-  let Inst{15-12} = Rt;
-  let Inst{19-16} = Rt2;
-  let Inst{11-8}  = cop;
-  let Inst{7-4}   = opc1;
-  let Inst{3-0}   = CRm;
-}
-
-def tMCRR : tMovRRCopro<"mcrr", 0 /* from ARM core register to coprocessor */,
-                        [(int_arm_mcrr imm:$cop, imm:$opc1, GPR:$Rt, GPR:$Rt2,
-                                       imm:$CRm)]>;
-def tMRRC : tMovRRCopro<"mrrc", 1 /* from coprocessor to ARM core register */>;
-
-//===----------------------------------------------------------------------===//
-// Other Coprocessor Instructions.  For disassembly only.
-//
-def tCDP : T1Cop<(outs), (ins p_imm:$cop, i32imm:$opc1,
-                 c_imm:$CRd, c_imm:$CRn, c_imm:$CRm, i32imm:$opc2),
-                 "cdp\t$cop, $opc1, $CRd, $CRn, $CRm, $opc2",
-                 [(int_arm_cdp imm:$cop, imm:$opc1, imm:$CRd, imm:$CRn,
-                               imm:$CRm, imm:$opc2)]> {
-  let Inst{27-24} = 0b1110;
-
-  bits<4> opc1;
-  bits<4> CRn;
-  bits<4> CRd;
-  bits<4> cop;
-  bits<3> opc2;
-  bits<4> CRm;
-
-  let Inst{3-0}   = CRm;
-  let Inst{4}     = 0;
-  let Inst{7-5}   = opc2;
-  let Inst{11-8}  = cop;
-  let Inst{15-12} = CRd;
-  let Inst{19-16} = CRn;
-  let Inst{23-20} = opc1;
-}
+                              2, IIC_iALUi, []>;
 
 //===----------------------------------------------------------------------===//
 // TLS Instructions
@@ -1335,7 +1214,7 @@ def tCDP : T1Cop<(outs), (ins p_imm:$cop, i32imm:$opc1,
 // This is a pseudo inst so that we can get the encoding right,
 // complete with fixup for the aeabi_read_tp function.
 let isCall = 1, Defs = [R0, R12, LR, CPSR], Uses = [SP] in
-def tTPsoft : tPseudoInst<(outs), (ins), Size4Bytes, IIC_Br,
+def tTPsoft : tPseudoInst<(outs), (ins), 4, IIC_Br,
                           [(set R0, ARMthread_pointer)]>;
 
 //===----------------------------------------------------------------------===//
@@ -1353,19 +1232,20 @@ def tTPsoft : tPseudoInst<(outs), (ins), Size4Bytes, IIC_Br,
 // preserve all of the callee-saved resgisters, which is exactly what we want.
 // $val is a scratch register for our use.
 let Defs = [ R0,  R1,  R2,  R3,  R4,  R5,  R6,  R7, R12, CPSR ],
-    hasSideEffects = 1, isBarrier = 1, isCodeGenOnly = 1 in
+    hasSideEffects = 1, isBarrier = 1, isCodeGenOnly = 1,
+    usesCustomInserter = 1 in
 def tInt_eh_sjlj_setjmp : ThumbXI<(outs),(ins tGPR:$src, tGPR:$val),
-                                  AddrModeNone, SizeSpecial, NoItinerary, "","",
+                                  AddrModeNone, 0, NoItinerary, "","",
                           [(set R0, (ARMeh_sjlj_setjmp tGPR:$src, tGPR:$val))]>;
 
-// FIXME: Non-Darwin version(s)
+// FIXME: Non-IOS version(s)
 let isBarrier = 1, hasSideEffects = 1, isTerminator = 1, isCodeGenOnly = 1,
     Defs = [ R7, LR, SP ] in
 def tInt_eh_sjlj_longjmp : XI<(outs), (ins GPR:$src, GPR:$scratch),
-                              AddrModeNone, SizeSpecial, IndexModeNone,
+                              AddrModeNone, 0, IndexModeNone,
                               Pseudo, NoItinerary, "", "",
                               [(ARMeh_sjlj_longjmp GPR:$src, GPR:$scratch)]>,
-                             Requires<[IsThumb, IsDarwin]>;
+                             Requires<[IsThumb, IsIOS]>;
 
 //===----------------------------------------------------------------------===//
 // Non-Instruction Patterns
@@ -1403,20 +1283,14 @@ def : T1Pat<(ARMWrapperJT tjumptable:$dst, imm:$id),
 
 // Direct calls
 def : T1Pat<(ARMtcall texternalsym:$func), (tBL texternalsym:$func)>,
-      Requires<[IsThumb, IsNotDarwin]>;
-def : T1Pat<(ARMtcall texternalsym:$func), (tBLr9 texternalsym:$func)>,
-      Requires<[IsThumb, IsDarwin]>;
+      Requires<[IsThumb]>;
 
 def : Tv5Pat<(ARMcall texternalsym:$func), (tBLXi texternalsym:$func)>,
-      Requires<[IsThumb, HasV5T, IsNotDarwin]>;
-def : Tv5Pat<(ARMcall texternalsym:$func), (tBLXi_r9 texternalsym:$func)>,
-      Requires<[IsThumb, HasV5T, IsDarwin]>;
+      Requires<[IsThumb, HasV5T]>;
 
 // Indirect calls to ARM routines
 def : Tv5Pat<(ARMcall GPR:$dst), (tBLXr GPR:$dst)>,
-      Requires<[IsThumb, HasV5T, IsNotDarwin]>;
-def : Tv5Pat<(ARMcall GPR:$dst), (tBLXr_r9 GPR:$dst)>,
-      Requires<[IsThumb, HasV5T, IsDarwin]>;
+      Requires<[IsThumb, HasV5T]>;
 
 // zextload i1 -> zextload i8
 def : T1Pat<(zextloadi1 t_addrmode_rrs1:$addr),
@@ -1456,6 +1330,31 @@ def : T1Pat<(sextloadi16 t_addrmode_rrs2:$addr),
 def : T1Pat<(sextloadi16 t_addrmode_is2:$addr),
             (tASRri (tLSLri (tLDRHi t_addrmode_is2:$addr), 16), 16)>;
 
+def : T1Pat<(atomic_load_8 t_addrmode_is1:$src),
+             (tLDRBi t_addrmode_is1:$src)>;
+def : T1Pat<(atomic_load_8 t_addrmode_rrs1:$src),
+             (tLDRBr t_addrmode_rrs1:$src)>;
+def : T1Pat<(atomic_load_16 t_addrmode_is2:$src),
+             (tLDRHi t_addrmode_is2:$src)>;
+def : T1Pat<(atomic_load_16 t_addrmode_rrs2:$src),
+             (tLDRHr t_addrmode_rrs2:$src)>;
+def : T1Pat<(atomic_load_32 t_addrmode_is4:$src),
+             (tLDRi t_addrmode_is4:$src)>;
+def : T1Pat<(atomic_load_32 t_addrmode_rrs4:$src),
+             (tLDRr t_addrmode_rrs4:$src)>;
+def : T1Pat<(atomic_store_8 t_addrmode_is1:$ptr, tGPR:$val),
+             (tSTRBi tGPR:$val, t_addrmode_is1:$ptr)>;
+def : T1Pat<(atomic_store_8 t_addrmode_rrs1:$ptr, tGPR:$val),
+             (tSTRBr tGPR:$val, t_addrmode_rrs1:$ptr)>;
+def : T1Pat<(atomic_store_16 t_addrmode_is2:$ptr, tGPR:$val),
+             (tSTRHi tGPR:$val, t_addrmode_is2:$ptr)>;
+def : T1Pat<(atomic_store_16 t_addrmode_rrs2:$ptr, tGPR:$val),
+             (tSTRHr tGPR:$val, t_addrmode_rrs2:$ptr)>;
+def : T1Pat<(atomic_store_32 t_addrmode_is4:$ptr, tGPR:$val),
+             (tSTRi tGPR:$val, t_addrmode_is4:$ptr)>;
+def : T1Pat<(atomic_store_32 t_addrmode_rrs4:$ptr, tGPR:$val),
+             (tSTRr tGPR:$val, t_addrmode_rrs4:$ptr)>;
+
 // Large immediate handling.
 
 // Two piece imms.
@@ -1481,12 +1380,38 @@ def tLDRpci_pic : PseudoInst<(outs GPR:$dst), (ins i32imm:$addr, pclabel:$cp),
 let isReturn = 1, isTerminator = 1, isBarrier = 1, mayLoad = 1,
     hasExtraDefRegAllocReq = 1 in
 def tPOP_RET : tPseudoExpand<(outs), (ins pred:$p, reglist:$regs, variable_ops),
-                           Size2Bytes, IIC_iPop_Br, [],
+                           2, IIC_iPop_Br, [],
                            (tPOP pred:$p, reglist:$regs)>;
 
 // Indirect branch using "mov pc, $Rm"
 let isBranch = 1, isTerminator = 1, isBarrier = 1, isIndirectBranch = 1 in {
-  def tBRIND : tPseudoExpand<(outs), (ins GPR:$Rm),
-                  Size2Bytes, IIC_Br, [(brind GPR:$Rm)],
-                  (tMOVr PC, GPR:$Rm, (ops 14, zero_reg))>;
+  def tBRIND : tPseudoExpand<(outs), (ins GPR:$Rm, pred:$p),
+                  2, IIC_Br, [(brind GPR:$Rm)],
+                  (tMOVr PC, GPR:$Rm, pred:$p)>;
 }
+
+
+// In Thumb1, "nop" is encoded as a "mov r8, r8". Technically, the bf00
+// encoding is available on ARMv6K, but we don't differentiate that finely.
+def : InstAlias<"nop", (tMOVr R8, R8, 14, 0)>,Requires<[IsThumb, IsThumb1Only]>;
+
+
+// For round-trip assembly/disassembly, we have to handle a CPS instruction
+// without any iflags. That's not, strictly speaking, valid syntax, but it's
+// a useful extension and assembles to defined behaviour (the insn does
+// nothing).
+def : tInstAlias<"cps$imod", (tCPS imod_op:$imod, 0)>;
+def : tInstAlias<"cps$imod", (tCPS imod_op:$imod, 0)>;
+
+// "neg" is and alias for "rsb rd, rn, #0"
+def : tInstAlias<"neg${s}${p} $Rd, $Rm",
+                 (tRSB tGPR:$Rd, s_cc_out:$s, tGPR:$Rm, pred:$p)>;
+
+
+// Implied destination operand forms for shifts.
+def : tInstAlias<"lsl${s}${p} $Rdm, $imm",
+             (tLSLri tGPR:$Rdm, cc_out:$s, tGPR:$Rdm, imm0_31:$imm, pred:$p)>;
+def : tInstAlias<"lsr${s}${p} $Rdm, $imm",
+             (tLSRri tGPR:$Rdm, cc_out:$s, tGPR:$Rdm, imm_sr:$imm, pred:$p)>;
+def : tInstAlias<"asr${s}${p} $Rdm, $imm",
+             (tASRri tGPR:$Rdm, cc_out:$s, tGPR:$Rdm, imm_sr:$imm, pred:$p)>;