[mips][mips64r6] c.cond.fmt, mov[fntz], and mov[fntz].[ds] are not available on MIPS32r6/MIPS64r6

[dsanders]

Summary

c.cond.fmt has been replaced by cmp.cond.fmt. Where c.cond.fmt wrote to
dedicated condition registers, cmp.cond.fmt writes 1 or 0 to normal FGR's
(like the GPR comparisons).

mov[fntz] have been replaced by seleqz and selnez. These instructions
conditionally zero a register based on a bool in a GPR. The results can
then be or'd together to act as a select without, for example, requiring a third
register read port.

mov[fntz].[ds] have been replaced with sel.[ds]

MIPS64r6 currently generates unnecessary sign-extensions for most selects.
This is because the result of a SETCC is currently an i32. Bits 32-63 are
undefined in i32 and the behaviour of seleqz/selnez would otherwise depend
on undefined bits. Later, we will fix this by making the result of SETCC an
i64 on MIPS64 targets.

Depends on D3958

[dsanders]

Removed the comment about the removal of c.cond.fmt since this has been done.

[zoran.jovanovic]

LGTM with a nit.

lib/Target/Mips/Mips32r6InstrInfo.td

612

Some of the lines are longer than 80 characters.

[vmedic]

Same comment as Zoran, looks good with a nit(80 character limit).

[dsanders]

Refresh after resolving rebase conflicts

[dsanders]

The 80 col corrections were committed in r210778 since I forgot to do them for r210777

lib/Target/Mips/Disassembler/MipsDisassembler.cpp

MipsDisassembler(const MCSubtargetInfo &STI, MCContext &Ctx,
     MipsDisassemblerBase(STI, Ctx, bigEndian) {
       IsMicroMips = STI.getFeatureBits() & Mips::FeatureMicroMips;
     }
 
   bool isMips32r6() const {
     return STI.getFeatureBits() & Mips::FeatureMips32r6;
   }
 
+  bool isGP64() const { return STI.getFeatureBits() & Mips::FeatureGP64Bit; }
+
   /// getInstruction - See MCDisassembler.
   DecodeStatus getInstruction(MCInst &instr,
                               uint64_t &size,
                               const MemoryObject &region,
                               uint64_t address,
                               raw_ostream &vStream,
                               raw_ostream &cStream) const override;
 };
static DecodeStatus DecodeCCRRegisterClass(MCInst &Inst,
                                            uint64_t Address,
                                            const void *Decoder);
 
 static DecodeStatus DecodeFCCRegisterClass(MCInst &Inst,
                                            unsigned RegNo,
                                            uint64_t Address,
                                            const void *Decoder);
 
+static DecodeStatus DecodeFGRCCRegisterClass(MCInst &Inst, unsigned RegNo,
+                                             uint64_t Address,
+                                             const void *Decoder);
+
 static DecodeStatus DecodeHWRegsRegisterClass(MCInst &Inst,
                                               unsigned Insn,
                                               uint64_t Address,
                                               const void *Decoder);
 
 static DecodeStatus DecodeAFGR64RegisterClass(MCInst &Inst,
                                               unsigned RegNo,
                                               uint64_t Address,
MipsDisassembler::getInstruction(MCInst &instr,
   uint32_t Insn;
 
   DecodeStatus Result = readInstruction32(Region, Address, Size,
                                           Insn, isBigEndian, IsMicroMips);
   if (Result == MCDisassembler::Fail)
     return MCDisassembler::Fail;
 
   if (IsMicroMips) {
+    DEBUG(dbgs() << "Trying MicroMips32 table (32-bit opcodes):\n");
     // Calling the auto-generated decoder function.
     Result = decodeInstruction(DecoderTableMicroMips32, instr, Insn, Address,
                                this, STI);
     if (Result != MCDisassembler::Fail) {
       Size = 4;
       return Result;
     }
     return MCDisassembler::Fail;
   }
 
+  if (isMips32r6() && isGP64()) {
+    DEBUG(dbgs() << "Trying Mips32r6_64r6 (GPR64) table (32-bit opcodes):\n");
+    Result = decodeInstruction(DecoderTableMips32r6_64r6_GP6432, instr, Insn,
+                               Address, this, STI);
+    if (Result != MCDisassembler::Fail) {
+      Size = 4;
+      return Result;
+    }
+  }
+
   if (isMips32r6()) {
+    DEBUG(dbgs() << "Trying Mips32r6_64r6 table (32-bit opcodes):\n");
     Result = decodeInstruction(DecoderTableMips32r6_64r632, instr, Insn,
                                Address, this, STI);
     if (Result != MCDisassembler::Fail) {
       Size = 4;
       return Result;
     }
   }
 
+  DEBUG(dbgs() << "Trying Mips table (32-bit opcodes):\n");
   // Calling the auto-generated decoder function.
   Result = decodeInstruction(DecoderTableMips32, instr, Insn, Address,
                              this, STI);
   if (Result != MCDisassembler::Fail) {
     Size = 4;
     return Result;
   }
 
static DecodeStatus DecodeFCCRegisterClass(MCInst &Inst,
                                            const void *Decoder) {
   if (RegNo > 7)
     return MCDisassembler::Fail;
   unsigned Reg = getReg(Decoder, Mips::FCCRegClassID, RegNo);
   Inst.addOperand(MCOperand::CreateReg(Reg));
   return MCDisassembler::Success;
 }
 
+static DecodeStatus DecodeFGRCCRegisterClass(MCInst &Inst, unsigned RegNo,
+                                             uint64_t Address,
+                                             const void *Decoder) {
+  if (RegNo > 31)
+    return MCDisassembler::Fail;
+
+  unsigned Reg = getReg(Decoder, Mips::FGRCCRegClassID, RegNo);
+  Inst.addOperand(MCOperand::CreateReg(Reg));
+  return MCDisassembler::Success;
+}
+
 static DecodeStatus DecodeMem(MCInst &Inst,
                               unsigned Insn,
                               uint64_t Address,
                               const void *Decoder) {
   int Offset = SignExtend32<16>(Insn & 0xffff);
   unsigned Reg = fieldFromInstruction(Insn, 16, 5);
   unsigned Base = fieldFromInstruction(Insn, 21, 5);
 

lib/Target/Mips/Mips32r6InstrInfo.td

 // Reencoded: sdbbp
 // Reencoded: sdc2
 // Reencoded: swc2
 // Removed: bc1any2, bc1any4
 // Removed: bc2[ft]
 // Removed: bc2f, bc2t
 // Removed: bgezal
 // Removed: bltzal
-// Removed: c.cond.fmt, bc1[ft]
+// Removed: bc1[ft]
 // Removed: ldxc1
 // Removed: luxc1
 // Removed: lwxc1
-// Removed: movf, movt
-// Removed: movf.fmt, movt.fmt, movn.fmt, movz.fmt
-// Removed: movn, movz
 // Removed: prefx
 // Removed: sdxc1
 // Removed: suxc1
 // Removed: swxc1
 // Rencoded: [ls][wd]c2
 
 def brtarget21 : Operand<OtherVT> {
   let EncoderMethod = "getBranchTarget21OpValue";

 class SELNEZ_S_ENC : COP1_3R_FM<0b010111, FIELD_FMT_S>;
 class SELNEZ_D_ENC : COP1_3R_FM<0b010111, FIELD_FMT_D>;
 
 class RINT_S_ENC : COP1_2R_FM<0b011010, FIELD_FMT_S>;
 class RINT_D_ENC : COP1_2R_FM<0b011010, FIELD_FMT_D>;
 class CLASS_S_ENC : COP1_2R_FM<0b011011, FIELD_FMT_S>;
 class CLASS_D_ENC : COP1_2R_FM<0b011011, FIELD_FMT_D>;
 
-class CMP_CONDN_DESC_BASE<string CondStr, string Typestr, RegisterOperand FGROpnd> {
-  dag OutOperandList = (outs FGROpnd:$fd);
+class CMP_CONDN_DESC_BASE<string CondStr, string Typestr,
+                          RegisterOperand FGROpnd,
+                          SDPatternOperator Op = null_frag> {
+  dag OutOperandList = (outs FGRCCOpnd:$fd);
   dag InOperandList = (ins FGROpnd:$fs, FGROpnd:$ft);
   string AsmString = !strconcat("cmp.", CondStr, ".", Typestr, "\t$fd, $fs, $ft");
-  list<dag> Pattern = [];
+  list<dag> Pattern = [(set FGRCCOpnd:$fd, (Op FGROpnd:$fs, FGROpnd:$ft))];
 }
 
 //===----------------------------------------------------------------------===//
 //
 // Instruction Multiclasses
 //
 //===----------------------------------------------------------------------===//
 
 multiclass CMP_CC_M <FIELD_CMP_FORMAT Format, string Typestr,
                      RegisterOperand FGROpnd>{
   def CMP_F_#NAME : COP1_CMP_CONDN_FM<Format, FIELD_CMP_COND_F>,
                     CMP_CONDN_DESC_BASE<"f", Typestr, FGROpnd>,
                     ISA_MIPS32R6;
   def CMP_UN_#NAME : COP1_CMP_CONDN_FM<Format, FIELD_CMP_COND_UN>,
-                     CMP_CONDN_DESC_BASE<"un", Typestr, FGROpnd>,
+                     CMP_CONDN_DESC_BASE<"un", Typestr, FGROpnd, setuo>,
                      ISA_MIPS32R6;
   def CMP_EQ_#NAME : COP1_CMP_CONDN_FM<Format, FIELD_CMP_COND_EQ>,
-                     CMP_CONDN_DESC_BASE<"eq", Typestr, FGROpnd>,
+                     CMP_CONDN_DESC_BASE<"eq", Typestr, FGROpnd, setoeq>,
                      ISA_MIPS32R6;
   def CMP_UEQ_#NAME : COP1_CMP_CONDN_FM<Format, FIELD_CMP_COND_UEQ>,
-                      CMP_CONDN_DESC_BASE<"ueq", Typestr, FGROpnd>,
+                      CMP_CONDN_DESC_BASE<"ueq", Typestr, FGROpnd, setueq>,
                       ISA_MIPS32R6;
   def CMP_OLT_#NAME : COP1_CMP_CONDN_FM<Format, FIELD_CMP_COND_OLT>,
-                      CMP_CONDN_DESC_BASE<"olt", Typestr, FGROpnd>,
+                      CMP_CONDN_DESC_BASE<"olt", Typestr, FGROpnd, setolt>,
                       ISA_MIPS32R6;
   def CMP_ULT_#NAME : COP1_CMP_CONDN_FM<Format, FIELD_CMP_COND_ULT>,
-                      CMP_CONDN_DESC_BASE<"ult", Typestr, FGROpnd>,
+                      CMP_CONDN_DESC_BASE<"ult", Typestr, FGROpnd, setult>,
                       ISA_MIPS32R6;
   def CMP_OLE_#NAME : COP1_CMP_CONDN_FM<Format, FIELD_CMP_COND_OLE>,
-                      CMP_CONDN_DESC_BASE<"ole", Typestr, FGROpnd>,
+                      CMP_CONDN_DESC_BASE<"ole", Typestr, FGROpnd, setole>,
                       ISA_MIPS32R6;
   def CMP_ULE_#NAME : COP1_CMP_CONDN_FM<Format, FIELD_CMP_COND_ULE>,
-                      CMP_CONDN_DESC_BASE<"ule", Typestr, FGROpnd>,
+                      CMP_CONDN_DESC_BASE<"ule", Typestr, FGROpnd, setule>,
                       ISA_MIPS32R6;
   def CMP_SF_#NAME : COP1_CMP_CONDN_FM<Format, FIELD_CMP_COND_SF>,
                      CMP_CONDN_DESC_BASE<"sf", Typestr, FGROpnd>,
                      ISA_MIPS32R6;
   def CMP_NGLE_#NAME : COP1_CMP_CONDN_FM<Format, FIELD_CMP_COND_NGLE>,
                        CMP_CONDN_DESC_BASE<"ngle", Typestr, FGROpnd>,
                        ISA_MIPS32R6;
   def CMP_SEQ_#NAME : COP1_CMP_CONDN_FM<Format, FIELD_CMP_COND_SEQ>,
class MUL_R6_DESC_BASE<string instr_asm, RegisterOperand GPROpnd,
   list<dag> Pattern = [(set GPROpnd:$rd, (Op GPROpnd:$rs, GPROpnd:$rt))];
 }
 
 class MUH_DESC    : MUL_R6_DESC_BASE<"muh", GPR32Opnd, mulhs>;
 class MUHU_DESC   : MUL_R6_DESC_BASE<"muhu", GPR32Opnd, mulhu>;
 class MUL_R6_DESC : MUL_R6_DESC_BASE<"mul", GPR32Opnd, mul>;
 class MULU_DESC   : MUL_R6_DESC_BASE<"mulu", GPR32Opnd>;
 
-class COP1_4R_DESC_BASE<string instr_asm, RegisterOperand FGROpnd> {
+class COP1_SEL_DESC_BASE<string instr_asm, RegisterOperand FGROpnd> {
   dag OutOperandList = (outs FGROpnd:$fd);
-  dag InOperandList = (ins FGROpnd:$fd_in, FGROpnd:$fs, FGROpnd:$ft);
+  dag InOperandList = (ins FGRCCOpnd:$fd_in, FGROpnd:$fs, FGROpnd:$ft);
   string AsmString = !strconcat(instr_asm, "\t$fd, $fs, $ft");
-  list<dag> Pattern = [];
+  list<dag> Pattern = [(set FGROpnd:$fd, (select FGRCCOpnd:$fd_in,
+                                                 FGROpnd:$ft,
+                                                 FGROpnd:$fs))];
   string Constraints = "$fd_in = $fd";
 }
 
-class SEL_D_DESC : COP1_4R_DESC_BASE<"sel.d", FGR64Opnd>;
-class SEL_S_DESC : COP1_4R_DESC_BASE<"sel.s", FGR32Opnd>;
+class SEL_D_DESC : COP1_SEL_DESC_BASE<"sel.d", FGR64Opnd> {
+  // We must insert a SUBREG_TO_REG around $fd_in
+  bit usesCustomInserter = 1;
+}
+class SEL_S_DESC : COP1_SEL_DESC_BASE<"sel.s", FGR32Opnd>;
 
 class SELEQNE_Z_DESC_BASE<string instr_asm, RegisterOperand GPROpnd> {
   dag OutOperandList = (outs GPROpnd:$rd);
   dag InOperandList = (ins GPROpnd:$rs, GPROpnd:$rt);
   string AsmString = !strconcat(instr_asm, "\t$rd, $rs, $rt");
   list<dag> Pattern = [];
 }
 
 class SELEQZ_DESC : SELEQNE_Z_DESC_BASE<"seleqz", GPR32Opnd>;
 class SELNEZ_DESC : SELEQNE_Z_DESC_BASE<"selnez", GPR32Opnd>;
 
+class COP1_4R_DESC_BASE<string instr_asm, RegisterOperand FGROpnd> {
+  dag OutOperandList = (outs FGROpnd:$fd);
+  dag InOperandList = (ins FGROpnd:$fd_in, FGROpnd:$fs, FGROpnd:$ft);
+  string AsmString = !strconcat(instr_asm, "\t$fd, $fs, $ft");
+  list<dag> Pattern = [];
+  string Constraints = "$fd_in = $fd";
+}
+
 class MADDF_S_DESC  : COP1_4R_DESC_BASE<"maddf.s", FGR32Opnd>;
 class MADDF_D_DESC  : COP1_4R_DESC_BASE<"maddf.d", FGR64Opnd>;
 class MSUBF_S_DESC  : COP1_4R_DESC_BASE<"msubf.s", FGR32Opnd>;
 class MSUBF_D_DESC  : COP1_4R_DESC_BASE<"msubf.d", FGR64Opnd>;
 
 class MAX_MIN_DESC_BASE<string instr_asm, RegisterOperand FGROpnd> {
   dag OutOperandList = (outs FGROpnd:$fd);
   dag InOperandList = (ins FGROpnd:$fs, FGROpnd:$ft);

 def MSUBF_D : MSUBF_D_ENC, MSUBF_D_DESC, ISA_MIPS32R6;
 def MUH    : MUH_ENC, MUH_DESC, ISA_MIPS32R6;
 def MUHU   : MUHU_ENC, MUHU_DESC, ISA_MIPS32R6;
 def MUL_R6 : MUL_R6_ENC, MUL_R6_DESC, ISA_MIPS32R6;
 def MULU   : MULU_ENC, MULU_DESC, ISA_MIPS32R6;
 def NAL; // BAL with rd=0
 def RINT_D : RINT_D_ENC, RINT_D_DESC, ISA_MIPS32R6;
 def RINT_S : RINT_S_ENC, RINT_S_DESC, ISA_MIPS32R6;
-def SELEQZ : SELEQZ_ENC, SELEQZ_DESC, ISA_MIPS32R6;
+def SELEQZ : SELEQZ_ENC, SELEQZ_DESC, ISA_MIPS32R6, GPR_32;
 def SELEQZ_D : SELEQZ_D_ENC, SELEQZ_D_DESC, ISA_MIPS32R6;
 def SELEQZ_S : SELEQZ_S_ENC, SELEQZ_S_DESC, ISA_MIPS32R6;
-def SELNEZ : SELNEZ_ENC, SELNEZ_DESC, ISA_MIPS32R6;
+def SELNEZ : SELNEZ_ENC, SELNEZ_DESC, ISA_MIPS32R6, GPR_32;
 def SELNEZ_D : SELNEZ_D_ENC, SELNEZ_D_DESC, ISA_MIPS32R6;
 def SELNEZ_S : SELNEZ_S_ENC, SELNEZ_S_DESC, ISA_MIPS32R6;
 def SEL_D : SEL_D_ENC, SEL_D_DESC, ISA_MIPS32R6;
 def SEL_S : SEL_S_ENC, SEL_S_DESC, ISA_MIPS32R6;
+
+//===----------------------------------------------------------------------===//
+//
+// Patterns and Pseudo Instructions
+//
+//===----------------------------------------------------------------------===//
+
+// f32 comparisons supported via another comparison
+def : MipsPat<(setone f32:$lhs, f32:$rhs),
+              (NOR (CMP_UEQ_S f32:$lhs, f32:$rhs), ZERO)>, ISA_MIPS32R6;
+def : MipsPat<(seto f32:$lhs, f32:$rhs),
+              (NOR (CMP_UN_S f32:$lhs, f32:$rhs), ZERO)>, ISA_MIPS32R6;
+def : MipsPat<(setune f32:$lhs, f32:$rhs),
+              (NOR (CMP_EQ_S f32:$lhs, f32:$rhs), ZERO)>, ISA_MIPS32R6;
+def : MipsPat<(seteq f32:$lhs, f32:$rhs), (CMP_EQ_S f32:$lhs, f32:$rhs)>, ISA_MIPS32R6;
+def : MipsPat<(setgt f32:$lhs, f32:$rhs), (CMP_LE_S f32:$rhs, f32:$lhs)>, ISA_MIPS32R6;
+def : MipsPat<(setge f32:$lhs, f32:$rhs), (CMP_LT_S f32:$rhs, f32:$lhs)>, ISA_MIPS32R6;
+def : MipsPat<(setlt f32:$lhs, f32:$rhs), (CMP_OLT_S f32:$lhs, f32:$rhs)>, ISA_MIPS32R6;
+def : MipsPat<(setlt f32:$lhs, f32:$rhs), (CMP_OLE_S f32:$lhs, f32:$rhs)>, ISA_MIPS32R6;
+def : MipsPat<(setne f32:$lhs, f32:$rhs),
+              (NOR (CMP_EQ_S f32:$lhs, f32:$rhs), ZERO)>, ISA_MIPS32R6;
+
+// f64 comparisons supported via another comparison
+def : MipsPat<(setone f64:$lhs, f64:$rhs),
+              (NOR (CMP_UEQ_D f64:$lhs, f64:$rhs), ZERO)>, ISA_MIPS32R6;
+def : MipsPat<(seto f64:$lhs, f64:$rhs),
+              (NOR (CMP_UN_D f64:$lhs, f64:$rhs), ZERO)>, ISA_MIPS32R6;
+def : MipsPat<(setune f64:$lhs, f64:$rhs),
+              (NOR (CMP_EQ_D f64:$lhs, f64:$rhs), ZERO)>, ISA_MIPS32R6;
+def : MipsPat<(seteq f64:$lhs, f64:$rhs), (CMP_EQ_D f64:$lhs, f64:$rhs)>, ISA_MIPS32R6;
+def : MipsPat<(setgt f64:$lhs, f64:$rhs), (CMP_LE_D f64:$rhs, f64:$lhs)>, ISA_MIPS32R6;
+def : MipsPat<(setge f64:$lhs, f64:$rhs), (CMP_LT_D f64:$rhs, f64:$lhs)>, ISA_MIPS32R6;
+def : MipsPat<(setlt f64:$lhs, f64:$rhs), (CMP_OLT_D f64:$lhs, f64:$rhs)>, ISA_MIPS32R6;
+def : MipsPat<(setlt f64:$lhs, f64:$rhs), (CMP_OLE_D f64:$lhs, f64:$rhs)>, ISA_MIPS32R6;
+def : MipsPat<(setne f64:$lhs, f64:$rhs),
+              (NOR (CMP_EQ_D f64:$lhs, f64:$rhs), ZERO)>, ISA_MIPS32R6;
+
+// i32 selects
+def : MipsPat<(select i32:$cond, i32:$t, i32:$f),
+              (OR (SELNEZ i32:$t, i32:$cond), (SELEQZ i32:$f, i32:$cond))>,
+              ISA_MIPS32R6;
+def : MipsPat<(select (i32 (seteq i32:$cond, immz)), i32:$t, i32:$f),
+              (OR (SELNEZ i32:$t, i32:$cond), (SELEQZ i32:$f, i32:$cond))>,
+              ISA_MIPS32R6;
+def : MipsPat<(select (i32 (setne i32:$cond, immz)), i32:$t, i32:$f),
+              (OR (SELNEZ i32:$f, i32:$cond), (SELEQZ i32:$t, i32:$cond))>,
+              ISA_MIPS32R6;
+def : MipsPat<(select (i32 (seteq i32:$cond, immZExt16:$imm)), i32:$t, i32:$f),
+              (OR (SELNEZ i32:$t, (XORi i32:$cond, immZExt16:$imm)),
+                  (SELEQZ i32:$f, (XORi i32:$cond, immZExt16:$imm)))>,
+              ISA_MIPS32R6;
+def : MipsPat<(select (i32 (setne i32:$cond, immZExt16:$imm)), i32:$t, i32:$f),
+              (OR (SELNEZ i32:$f, (XORi i32:$cond, immZExt16:$imm)),
+                  (SELEQZ i32:$t, (XORi i32:$cond, immZExt16:$imm)))>,
+              ISA_MIPS32R6;
+def : MipsPat<(select (i32 (setgt i32:$cond, immSExt16Plus1:$imm)), i32:$t,
+                      i32:$f),
+              (OR (SELNEZ i32:$t, (SLTi i32:$cond, (Plus1 imm:$imm))),
+                  (SELEQZ i32:$f, (SLTi i32:$cond, (Plus1 imm:$imm))))>,
+              ISA_MIPS32R6;
+def : MipsPat<(select (i32 (setugt i32:$cond, immSExt16Plus1:$imm)),
+                      i32:$t, i32:$f),
+              (OR (SELNEZ i32:$t, (SLTiu i32:$cond, (Plus1 imm:$imm))),
+                  (SELEQZ i32:$f, (SLTiu i32:$cond, (Plus1 imm:$imm))))>,
+              ISA_MIPS32R6;
+
+def : MipsPat<(select i32:$cond, i32:$t, immz),
+              (SELNEZ i32:$t, i32:$cond)>, ISA_MIPS32R6;
+def : MipsPat<(select (i32 (setne i32:$cond, immz)), i32:$t, immz),
+              (SELNEZ i32:$t, i32:$cond)>, ISA_MIPS32R6;
+def : MipsPat<(select (i32 (seteq i32:$cond, immz)), i32:$t, immz),
+              (SELEQZ i32:$t, i32:$cond)>, ISA_MIPS32R6;
+def : MipsPat<(select i32:$cond, immz, i32:$f),
+              (SELEQZ i32:$f, i32:$cond)>, ISA_MIPS32R6;
+def : MipsPat<(select (i32 (setne i32:$cond, immz)), immz, i32:$f),
+              (SELEQZ i32:$f, i32:$cond)>, ISA_MIPS32R6;
+def : MipsPat<(select (i32 (seteq i32:$cond, immz)), immz, i32:$f),
+              (SELNEZ i32:$f, i32:$cond)>, ISA_MIPS32R6;

lib/Target/Mips/Mips64InstrInfo.td

 // shamt must fit in 6 bits.
 def immZExt6 : ImmLeaf<i32, [{return Imm == (Imm & 0x3f);}]>;
 
 // Node immediate fits as 10-bit sign extended on target immediate.
 // e.g. seqi, snei
 def immSExt10_64 : PatLeaf<(i64 imm),
                            [{ return isInt<10>(N->getSExtValue()); }]>;
 
+def immZExt16_64 : PatLeaf<(i64 imm),
+                           [{ return isInt<16>(N->getZExtValue()); }]>;
+
 //===----------------------------------------------------------------------===//
 // Instructions specific format
 //===----------------------------------------------------------------------===//
 let usesCustomInserter = 1 in {
   def ATOMIC_LOAD_ADD_I64  : Atomic2Ops<atomic_load_add_64, GPR64>;
   def ATOMIC_LOAD_SUB_I64  : Atomic2Ops<atomic_load_sub_64, GPR64>;
   def ATOMIC_LOAD_AND_I64  : Atomic2Ops<atomic_load_and_64, GPR64>;
   def ATOMIC_LOAD_OR_I64   : Atomic2Ops<atomic_load_or_64, GPR64>;

lib/Target/Mips/Mips64r6InstrInfo.td

 class DDIVU_DESC   : DIVMOD_DESC_BASE<"ddivu", GPR64Opnd, udiv>;
 class DMOD_DESC    : DIVMOD_DESC_BASE<"dmod", GPR64Opnd, srem>;
 class DMODU_DESC   : DIVMOD_DESC_BASE<"dmodu", GPR64Opnd, urem>;
 class DMUH_DESC    : MUL_R6_DESC_BASE<"dmuh", GPR64Opnd, mulhs>;
 class DMUHU_DESC   : MUL_R6_DESC_BASE<"dmuhu", GPR64Opnd, mulhu>;
 class DMUL_R6_DESC : MUL_R6_DESC_BASE<"dmul", GPR64Opnd, mul>;
 class DMULU_DESC   : MUL_R6_DESC_BASE<"dmulu", GPR64Opnd>;
 
+class SELEQZ64_DESC : SELEQNE_Z_DESC_BASE<"seleqz", GPR64Opnd>;
+class SELNEZ64_DESC : SELEQNE_Z_DESC_BASE<"selnez", GPR64Opnd>;
+
 //===----------------------------------------------------------------------===//
 //
 // Instruction Definitions
 //
 //===----------------------------------------------------------------------===//
 
 def DAHI : DAHI_ENC, DAHI_DESC, ISA_MIPS64R6;
 def DALIGN : DALIGN_ENC, DALIGN_DESC, ISA_MIPS64R6;
 def DATI : DATI_ENC, DATI_DESC, ISA_MIPS64R6;
 def DAUI : DAUI_ENC, DAUI_DESC, ISA_MIPS64R6;
 def DBITSWAP : DBITSWAP_ENC, DBITSWAP_DESC, ISA_MIPS64R6;
 def DDIV : DDIV_ENC, DDIV_DESC, ISA_MIPS64R6;
 def DDIVU : DDIVU_ENC, DDIVU_DESC, ISA_MIPS64R6;
 // def DLSA; // See MSA
 def DMOD : DMOD_ENC, DMOD_DESC, ISA_MIPS64R6;
 def DMODU : DMODU_ENC, DMODU_DESC, ISA_MIPS64R6;
 def DMUH: DMUH_ENC, DMUH_DESC, ISA_MIPS64R6;
 def DMUHU: DMUHU_ENC, DMUHU_DESC, ISA_MIPS64R6;
 def DMUL_R6: DMUL_R6_ENC, DMUL_R6_DESC, ISA_MIPS64R6;
 def DMULU: DMULU_ENC, DMULU_DESC, ISA_MIPS64R6;
 def LDPC;
+let DecoderNamespace = "Mips32r6_64r6_GP64" in {
+  def SELEQZ64 : SELEQZ_ENC, SELEQZ64_DESC, ISA_MIPS32R6, GPR_64;
+  def SELNEZ64 : SELNEZ_ENC, SELNEZ64_DESC, ISA_MIPS32R6, GPR_64;
+}
+
+//===----------------------------------------------------------------------===//
+//
+// Patterns and Pseudo Instructions
+//
+//===----------------------------------------------------------------------===//
+
+// i64 selects
+def : MipsPat<(select i64:$cond, i64:$t, i64:$f),
+              (OR64 (SELNEZ64 i64:$t, i64:$cond),
+                    (SELEQZ64 i64:$f, i64:$cond))>,
+              ISA_MIPS64R6;
+def : MipsPat<(select (i32 (seteq i64:$cond, immz)), i64:$t, i64:$f),
+              (OR64 (SELNEZ64 i64:$t, i64:$cond),
+                    (SELEQZ64 i64:$f, i64:$cond))>,
+              ISA_MIPS64R6;
+def : MipsPat<(select (i32 (setne i64:$cond, immz)), i64:$t, i64:$f),
+              (OR64 (SELNEZ64 i64:$f, i64:$cond),
+                    (SELEQZ64 i64:$t, i64:$cond))>,
+              ISA_MIPS64R6;
+def : MipsPat<(select (i32 (seteq i64:$cond, immZExt16_64:$imm)), i64:$t, i64:$f),
+              (OR64 (SELNEZ64 i64:$t, (XORi64 i64:$cond, immZExt16_64:$imm)),
+                    (SELEQZ64 i64:$f, (XORi64 i64:$cond, immZExt16_64:$imm)))>,
+              ISA_MIPS64R6;
+def : MipsPat<(select (i32 (setne i64:$cond, immZExt16_64:$imm)), i64:$t, i64:$f),
+              (OR64 (SELNEZ64 i64:$f, (XORi64 i64:$cond, immZExt16_64:$imm)),
+                    (SELEQZ64 i64:$t, (XORi64 i64:$cond, immZExt16_64:$imm)))>,
+              ISA_MIPS64R6;
+def : MipsPat<
+  (select (i32 (setgt i64:$cond, immSExt16Plus1:$imm)), i64:$t, i64:$f),
+  (OR64 (SELNEZ64 i64:$t,
+                  (SUBREG_TO_REG (i64 0), (SLTi64 i64:$cond, (Plus1 imm:$imm)),
+                                 sub_32)),
+        (SELEQZ64 i64:$f,
+                  (SUBREG_TO_REG (i64 0), (SLTi64 i64:$cond, (Plus1 imm:$imm)),
+                                 sub_32)))>,
+  ISA_MIPS64R6;
+def : MipsPat<
+  (select (i32 (setugt i64:$cond, immSExt16Plus1:$imm)), i64:$t, i64:$f),
+  (OR64 (SELNEZ64 i64:$t,
+                  (SUBREG_TO_REG (i64 0), (SLTiu64 i64:$cond, (Plus1 imm:$imm)),
+                                 sub_32)),
+        (SELEQZ64 i64:$f,
+                  (SUBREG_TO_REG (i64 0), (SLTiu64 i64:$cond, (Plus1 imm:$imm)),
+                                 sub_32)))>,
+  ISA_MIPS64R6;
+
+def : MipsPat<(select (i32 (setne i64:$cond, immz)), i64:$t, immz),
+              (SELNEZ64 i64:$t, i64:$cond)>, ISA_MIPS64R6;
+def : MipsPat<(select (i32 (seteq i64:$cond, immz)), i64:$t, immz),
+              (SELEQZ64 i64:$t, i64:$cond)>, ISA_MIPS64R6;
+def : MipsPat<(select (i32 (setne i64:$cond, immz)), immz, i64:$f),
+              (SELEQZ64 i64:$f, i64:$cond)>, ISA_MIPS64R6;
+def : MipsPat<(select (i32 (seteq i64:$cond, immz)), immz, i64:$f),
+              (SELNEZ64 i64:$f, i64:$cond)>, ISA_MIPS64R6;
+
+// i64 selects from an i32 comparison
+// One complicating factor here is that bits 32-63 of an i32 are undefined.
+// FIXME: Ideally, setcc would always produce an i64 on MIPS64 targets.
+//        This would allow us to remove the sign-extensions here.
+def : MipsPat<(select i32:$cond, i64:$t, i64:$f),
+              (OR64 (SELNEZ64 i64:$t, (SLL64_32 i32:$cond)),
+                    (SELEQZ64 i64:$f, (SLL64_32 i32:$cond)))>,
+              ISA_MIPS64R6;
+def : MipsPat<(select (i32 (seteq i32:$cond, immz)), i64:$t, i64:$f),
+              (OR64 (SELNEZ64 i64:$t, (SLL64_32 i32:$cond)),
+                    (SELEQZ64 i64:$f, (SLL64_32 i32:$cond)))>,
+              ISA_MIPS64R6;
+def : MipsPat<(select (i32 (setne i32:$cond, immz)), i64:$t, i64:$f),
+              (OR64 (SELNEZ64 i64:$f, (SLL64_32 i32:$cond)),
+                    (SELEQZ64 i64:$t, (SLL64_32 i32:$cond)))>,
+              ISA_MIPS64R6;
+def : MipsPat<(select (i32 (seteq i32:$cond, immZExt16:$imm)), i64:$t, i64:$f),
+              (OR64 (SELNEZ64 i64:$t, (SLL64_32 (XORi i32:$cond,
+                                                      immZExt16:$imm))),
+                    (SELEQZ64 i64:$f, (SLL64_32 (XORi i32:$cond,
+                                                      immZExt16:$imm))))>,
+              ISA_MIPS64R6;
+def : MipsPat<(select (i32 (setne i32:$cond, immZExt16:$imm)), i64:$t, i64:$f),
+              (OR64 (SELNEZ64 i64:$f, (SLL64_32 (XORi i32:$cond,
+                                                      immZExt16:$imm))),
+                    (SELEQZ64 i64:$t, (SLL64_32 (XORi i32:$cond,
+                                                      immZExt16:$imm))))>,
+              ISA_MIPS64R6;
+
+def : MipsPat<(select i32:$cond, i64:$t, immz),
+              (SELNEZ64 i64:$t, (SLL64_32 i32:$cond))>,
+              ISA_MIPS64R6;
+def : MipsPat<(select (i32 (setne i32:$cond, immz)), i64:$t, immz),
+              (SELNEZ64 i64:$t, (SLL64_32 i32:$cond))>,
+              ISA_MIPS64R6;
+def : MipsPat<(select (i32 (seteq i32:$cond, immz)), i64:$t, immz),
+              (SELEQZ64 i64:$t, (SLL64_32 i32:$cond))>,
+              ISA_MIPS64R6;
+def : MipsPat<(select i32:$cond, immz, i64:$f),
+              (SELEQZ64 i64:$f, (SLL64_32 i32:$cond))>,
+              ISA_MIPS64R6;
+def : MipsPat<(select (i32 (setne i32:$cond, immz)), immz, i64:$f),
+              (SELEQZ64 i64:$f, (SLL64_32 i32:$cond))>,
+              ISA_MIPS64R6;
+def : MipsPat<(select (i32 (seteq i32:$cond, immz)), immz, i64:$f),
+              (SELNEZ64 i64:$f, (SLL64_32 i32:$cond))>,
+              ISA_MIPS64R6;

lib/Target/Mips/MipsCondMov.td

multiclass MovnPats<RegisterClass CRC, RegisterClass DRC, Instruction MOVNInst,
   def : MipsPat<(select CRC:$cond, DRC:$T, DRC:$F),
                 (MOVNInst DRC:$T, CRC:$cond, DRC:$F)>;
   def : MipsPat<(select (i32 (setne CRC:$lhs, 0)),DRC:$T, DRC:$F),
                 (MOVNInst DRC:$T, CRC:$lhs, DRC:$F)>;
 }
 
 // Instantiation of instructions.
 def MOVZ_I_I : MMRel, CMov_I_I_FT<"movz", GPR32Opnd, GPR32Opnd, II_MOVZ>,
-               ADD_FM<0, 0xa>, INSN_MIPS4_32;
+               ADD_FM<0, 0xa>, INSN_MIPS4_32_NOT_32R6_64R6;
 
 let isCodeGenOnly = 1 in {
   def MOVZ_I_I64   : CMov_I_I_FT<"movz", GPR32Opnd, GPR64Opnd, II_MOVZ>,
-                     ADD_FM<0, 0xa>;
+                     ADD_FM<0, 0xa>, INSN_MIPS4_32_NOT_32R6_64R6;
   def MOVZ_I64_I   : CMov_I_I_FT<"movz", GPR64Opnd, GPR32Opnd, II_MOVZ>,
-                     ADD_FM<0, 0xa>;
+                     ADD_FM<0, 0xa>, INSN_MIPS4_32_NOT_32R6_64R6;
   def MOVZ_I64_I64 : CMov_I_I_FT<"movz", GPR64Opnd, GPR64Opnd, II_MOVZ>,
-                     ADD_FM<0, 0xa>;
+                     ADD_FM<0, 0xa>, INSN_MIPS4_32_NOT_32R6_64R6;
 }
 
 def MOVN_I_I       : MMRel, CMov_I_I_FT<"movn", GPR32Opnd, GPR32Opnd, II_MOVN>,
-                     ADD_FM<0, 0xb>, INSN_MIPS4_32;
+                     ADD_FM<0, 0xb>, INSN_MIPS4_32_NOT_32R6_64R6;
 
 let isCodeGenOnly = 1 in {
   def MOVN_I_I64   : CMov_I_I_FT<"movn", GPR32Opnd, GPR64Opnd, II_MOVN>,
-                     ADD_FM<0, 0xb>;
+                     ADD_FM<0, 0xb>, INSN_MIPS4_32_NOT_32R6_64R6;
   def MOVN_I64_I   : CMov_I_I_FT<"movn", GPR64Opnd, GPR32Opnd, II_MOVN>,
-                     ADD_FM<0, 0xb>;
+                     ADD_FM<0, 0xb>, INSN_MIPS4_32_NOT_32R6_64R6;
   def MOVN_I64_I64 : CMov_I_I_FT<"movn", GPR64Opnd, GPR64Opnd, II_MOVN>,
-                     ADD_FM<0, 0xb>;
+                     ADD_FM<0, 0xb>, INSN_MIPS4_32_NOT_32R6_64R6;
 }
 
 def MOVZ_I_S : MMRel, CMov_I_F_FT<"movz.s", GPR32Opnd, FGR32Opnd, II_MOVZ_S>,
-               CMov_I_F_FM<18, 16>, INSN_MIPS4_32;
+               CMov_I_F_FM<18, 16>, INSN_MIPS4_32_NOT_32R6_64R6;
 
 let isCodeGenOnly = 1 in
 def MOVZ_I64_S : CMov_I_F_FT<"movz.s", GPR64Opnd, FGR32Opnd, II_MOVZ_S>,
-                 CMov_I_F_FM<18, 16>, AdditionalRequires<[HasMips64]>;
+                 CMov_I_F_FM<18, 16>, INSN_MIPS4_32_NOT_32R6_64R6,
+                 AdditionalRequires<[HasMips64]>;
 
 def MOVN_I_S : MMRel, CMov_I_F_FT<"movn.s", GPR32Opnd, FGR32Opnd, II_MOVN_S>,
-               CMov_I_F_FM<19, 16>, INSN_MIPS4_32;
+               CMov_I_F_FM<19, 16>, INSN_MIPS4_32_NOT_32R6_64R6;
 
 let isCodeGenOnly = 1 in
 def MOVN_I64_S : CMov_I_F_FT<"movn.s", GPR64Opnd, FGR32Opnd, II_MOVN_S>,
-                 CMov_I_F_FM<19, 16>, AdditionalRequires<[IsGP64bit]>;
+                 CMov_I_F_FM<19, 16>, INSN_MIPS4_32_NOT_32R6_64R6,
+                 AdditionalRequires<[IsGP64bit]>;
 
 def MOVZ_I_D32 : MMRel, CMov_I_F_FT<"movz.d", GPR32Opnd, AFGR64Opnd,
                                     II_MOVZ_D>, CMov_I_F_FM<18, 17>,
-                 INSN_MIPS4_32, FGR_32;
+                 INSN_MIPS4_32_NOT_32R6_64R6, FGR_32;
 def MOVN_I_D32 : MMRel, CMov_I_F_FT<"movn.d", GPR32Opnd, AFGR64Opnd,
                                     II_MOVN_D>, CMov_I_F_FM<19, 17>,
-                 INSN_MIPS4_32, FGR_32;
+                 INSN_MIPS4_32_NOT_32R6_64R6, FGR_32;
 
 let DecoderNamespace = "Mips64" in {
   def MOVZ_I_D64 : CMov_I_F_FT<"movz.d", GPR32Opnd, FGR64Opnd, II_MOVZ_D>,
-                   CMov_I_F_FM<18, 17>, INSN_MIPS4_32, FGR_64;
+                   CMov_I_F_FM<18, 17>, INSN_MIPS4_32_NOT_32R6_64R6, FGR_64;
   def MOVN_I_D64 : CMov_I_F_FT<"movn.d", GPR32Opnd, FGR64Opnd, II_MOVN_D>,
-                   CMov_I_F_FM<19, 17>, INSN_MIPS4_32, FGR_64;
+                   CMov_I_F_FM<19, 17>, INSN_MIPS4_32_NOT_32R6_64R6, FGR_64;
   let isCodeGenOnly = 1 in {
-    def MOVZ_I64_D64 : CMov_I_F_FT<"movz.d", GPR64Opnd, FGR64Opnd,
-                                   II_MOVZ_D>, CMov_I_F_FM<18, 17>, FGR_64;
-    def MOVN_I64_D64 : CMov_I_F_FT<"movn.d", GPR64Opnd, FGR64Opnd,
-                                   II_MOVN_D>, CMov_I_F_FM<19, 17>, FGR_64;
+    def MOVZ_I64_D64 : CMov_I_F_FT<"movz.d", GPR64Opnd, FGR64Opnd, II_MOVZ_D>,
+                       CMov_I_F_FM<18, 17>, INSN_MIPS4_32_NOT_32R6_64R6, FGR_64;
+    def MOVN_I64_D64 : CMov_I_F_FT<"movn.d", GPR64Opnd, FGR64Opnd, II_MOVN_D>,
+                       CMov_I_F_FM<19, 17>, INSN_MIPS4_32_NOT_32R6_64R6, FGR_64;
   }
 }
 
 def MOVT_I : MMRel, CMov_F_I_FT<"movt", GPR32Opnd, II_MOVT, MipsCMovFP_T>,
-             CMov_F_I_FM<1>, INSN_MIPS4_32;
+             CMov_F_I_FM<1>, INSN_MIPS4_32_NOT_32R6_64R6;
 
 let isCodeGenOnly = 1 in
 def MOVT_I64 : CMov_F_I_FT<"movt", GPR64Opnd, II_MOVT, MipsCMovFP_T>,
-               CMov_F_I_FM<1>, AdditionalRequires<[IsGP64bit]>;
+               CMov_F_I_FM<1>, INSN_MIPS4_32_NOT_32R6_64R6,
+               AdditionalRequires<[IsGP64bit]>;
 
 def MOVF_I : MMRel, CMov_F_I_FT<"movf", GPR32Opnd, II_MOVF, MipsCMovFP_F>,
-             CMov_F_I_FM<0>, INSN_MIPS4_32;
+             CMov_F_I_FM<0>, INSN_MIPS4_32_NOT_32R6_64R6;
 
 let isCodeGenOnly = 1 in
 def MOVF_I64 : CMov_F_I_FT<"movf", GPR64Opnd, II_MOVF, MipsCMovFP_F>,
-               CMov_F_I_FM<0>, AdditionalRequires<[IsGP64bit]>;
+               CMov_F_I_FM<0>, INSN_MIPS4_32_NOT_32R6_64R6,
+               AdditionalRequires<[IsGP64bit]>;
 
 def MOVT_S : MMRel, CMov_F_F_FT<"movt.s", FGR32Opnd, II_MOVT_S, MipsCMovFP_T>,
-             CMov_F_F_FM<16, 1>, INSN_MIPS4_32;
+             CMov_F_F_FM<16, 1>, INSN_MIPS4_32_NOT_32R6_64R6;
 def MOVF_S : MMRel, CMov_F_F_FT<"movf.s", FGR32Opnd, II_MOVF_S, MipsCMovFP_F>,
-             CMov_F_F_FM<16, 0>, INSN_MIPS4_32;
+             CMov_F_F_FM<16, 0>, INSN_MIPS4_32_NOT_32R6_64R6;
 
 def MOVT_D32 : MMRel, CMov_F_F_FT<"movt.d", AFGR64Opnd, II_MOVT_D,
                                   MipsCMovFP_T>, CMov_F_F_FM<17, 1>,
-               INSN_MIPS4_32, FGR_32;
+               INSN_MIPS4_32_NOT_32R6_64R6, FGR_32;
 def MOVF_D32 : MMRel, CMov_F_F_FT<"movf.d", AFGR64Opnd, II_MOVF_D,
                                   MipsCMovFP_F>, CMov_F_F_FM<17, 0>,
-               INSN_MIPS4_32, FGR_32;
+               INSN_MIPS4_32_NOT_32R6_64R6, FGR_32;
 
 let DecoderNamespace = "Mips64" in {
   def MOVT_D64 : CMov_F_F_FT<"movt.d", FGR64Opnd, II_MOVT_D, MipsCMovFP_T>,
-                 CMov_F_F_FM<17, 1>, INSN_MIPS4_32, FGR_64;
+                 CMov_F_F_FM<17, 1>, INSN_MIPS4_32_NOT_32R6_64R6, FGR_64;
   def MOVF_D64 : CMov_F_F_FT<"movf.d", FGR64Opnd, II_MOVF_D, MipsCMovFP_F>,
-                 CMov_F_F_FM<17, 0>, INSN_MIPS4_32, FGR_64;
+                 CMov_F_F_FM<17, 0>, INSN_MIPS4_32_NOT_32R6_64R6, FGR_64;
 }
 
 // Instantiation of conditional move patterns.
-defm : MovzPats0<GPR32, GPR32, MOVZ_I_I, SLT, SLTu, SLTi, SLTiu>;
-defm : MovzPats1<GPR32, GPR32, MOVZ_I_I, XOR>;
-defm : MovzPats2<GPR32, GPR32, MOVZ_I_I, XORi>;
+defm : MovzPats0<GPR32, GPR32, MOVZ_I_I, SLT, SLTu, SLTi, SLTiu>,
+       INSN_MIPS4_32_NOT_32R6_64R6;
+defm : MovzPats1<GPR32, GPR32, MOVZ_I_I, XOR>, INSN_MIPS4_32_NOT_32R6_64R6;
+defm : MovzPats2<GPR32, GPR32, MOVZ_I_I, XORi>, INSN_MIPS4_32_NOT_32R6_64R6;
 
-defm : MovzPats0<GPR32, GPR64, MOVZ_I_I64, SLT, SLTu, SLTi, SLTiu>, GPR_64;
+defm : MovzPats0<GPR32, GPR64, MOVZ_I_I64, SLT, SLTu, SLTi, SLTiu>,
+       INSN_MIPS4_32_NOT_32R6_64R6, GPR_64;
 defm : MovzPats0<GPR64, GPR32, MOVZ_I_I, SLT64, SLTu64, SLTi64, SLTiu64>,
-       GPR_64;
+       INSN_MIPS4_32_NOT_32R6_64R6, GPR_64;
 defm : MovzPats0<GPR64, GPR64, MOVZ_I_I64, SLT64, SLTu64, SLTi64, SLTiu64>,
+       INSN_MIPS4_32_NOT_32R6_64R6, GPR_64;
+defm : MovzPats1<GPR32, GPR64, MOVZ_I_I64, XOR>,
+       INSN_MIPS4_32_NOT_32R6_64R6, GPR_64;
+defm : MovzPats1<GPR64, GPR32, MOVZ_I64_I, XOR64>,
+       INSN_MIPS4_32_NOT_32R6_64R6, GPR_64;
+defm : MovzPats1<GPR64, GPR64, MOVZ_I64_I64, XOR64>,
+       INSN_MIPS4_32_NOT_32R6_64R6, GPR_64;
+defm : MovzPats2<GPR32, GPR64, MOVZ_I_I64, XORi>,
+       INSN_MIPS4_32_NOT_32R6_64R6, GPR_64;
+defm : MovzPats2<GPR64, GPR32, MOVZ_I64_I, XORi64>,
+       INSN_MIPS4_32_NOT_32R6_64R6, GPR_64;
+defm : MovzPats2<GPR64, GPR64, MOVZ_I64_I64, XORi64>,
+       INSN_MIPS4_32_NOT_32R6_64R6, GPR_64;
+
+defm : MovnPats<GPR32, GPR32, MOVN_I_I, XOR>, INSN_MIPS4_32_NOT_32R6_64R6;
+
+defm : MovnPats<GPR32, GPR64, MOVN_I_I64, XOR>, INSN_MIPS4_32_NOT_32R6_64R6,
+       GPR_64;
+defm : MovnPats<GPR64, GPR32, MOVN_I64_I, XOR64>, INSN_MIPS4_32_NOT_32R6_64R6,
+       GPR_64;
+defm : MovnPats<GPR64, GPR64, MOVN_I64_I64, XOR64>, INSN_MIPS4_32_NOT_32R6_64R6,
        GPR_64;
-defm : MovzPats1<GPR32, GPR64, MOVZ_I_I64, XOR>, GPR_64;
-defm : MovzPats1<GPR64, GPR32, MOVZ_I64_I, XOR64>, GPR_64;
-defm : MovzPats1<GPR64, GPR64, MOVZ_I64_I64, XOR64>, GPR_64;
-defm : MovzPats2<GPR32, GPR64, MOVZ_I_I64, XORi>, GPR_64;
-defm : MovzPats2<GPR64, GPR32, MOVZ_I64_I, XORi64>, GPR_64;
-defm : MovzPats2<GPR64, GPR64, MOVZ_I64_I64, XORi64>, GPR_64;
-
-defm : MovnPats<GPR32, GPR32, MOVN_I_I, XOR>;
-
-defm : MovnPats<GPR32, GPR64, MOVN_I_I64, XOR>, GPR_64;
-defm : MovnPats<GPR64, GPR32, MOVN_I64_I, XOR64>, GPR_64;
-defm : MovnPats<GPR64, GPR64, MOVN_I64_I64, XOR64>, GPR_64;
-
-defm : MovzPats0<GPR32, FGR32, MOVZ_I_S, SLT, SLTu, SLTi, SLTiu>;
-defm : MovzPats1<GPR32, FGR32, MOVZ_I_S, XOR>;
-defm : MovnPats<GPR32, FGR32, MOVN_I_S, XOR>;
+
+defm : MovzPats0<GPR32, FGR32, MOVZ_I_S, SLT, SLTu, SLTi, SLTiu>,
+       INSN_MIPS4_32_NOT_32R6_64R6;
+defm : MovzPats1<GPR32, FGR32, MOVZ_I_S, XOR>, INSN_MIPS4_32_NOT_32R6_64R6;
+defm : MovnPats<GPR32, FGR32, MOVN_I_S, XOR>, INSN_MIPS4_32_NOT_32R6_64R6;
 
 defm : MovzPats0<GPR64, FGR32, MOVZ_I_S, SLT64, SLTu64, SLTi64, SLTiu64>,
+       INSN_MIPS4_32_NOT_32R6_64R6, GPR_64;
+defm : MovzPats1<GPR64, FGR32, MOVZ_I64_S, XOR64>, INSN_MIPS4_32_NOT_32R6_64R6,
+       GPR_64;
+defm : MovnPats<GPR64, FGR32, MOVN_I64_S, XOR64>, INSN_MIPS4_32_NOT_32R6_64R6,
        GPR_64;
-defm : MovzPats1<GPR64, FGR32, MOVZ_I64_S, XOR64>, GPR_64;
-defm : MovnPats<GPR64, FGR32, MOVN_I64_S, XOR64>, GPR_64;
 
-defm : MovzPats0<GPR32, AFGR64, MOVZ_I_D32, SLT, SLTu, SLTi, SLTiu>, FGR_32;
-defm : MovzPats1<GPR32, AFGR64, MOVZ_I_D32, XOR>, FGR_32;
-defm : MovnPats<GPR32, AFGR64, MOVN_I_D32, XOR>, FGR_32;
+defm : MovzPats0<GPR32, AFGR64, MOVZ_I_D32, SLT, SLTu, SLTi, SLTiu>,
+       INSN_MIPS4_32_NOT_32R6_64R6, FGR_32;
+defm : MovzPats1<GPR32, AFGR64, MOVZ_I_D32, XOR>, INSN_MIPS4_32_NOT_32R6_64R6,
+       FGR_32;
+defm : MovnPats<GPR32, AFGR64, MOVN_I_D32, XOR>, INSN_MIPS4_32_NOT_32R6_64R6,
+       FGR_32;
 
-defm : MovzPats0<GPR32, FGR64, MOVZ_I_D64, SLT, SLTu, SLTi, SLTiu>, FGR_64;
+defm : MovzPats0<GPR32, FGR64, MOVZ_I_D64, SLT, SLTu, SLTi, SLTiu>,
+       INSN_MIPS4_32_NOT_32R6_64R6, FGR_64;
 defm : MovzPats0<GPR64, FGR64, MOVZ_I_D64, SLT64, SLTu64, SLTi64, SLTiu64>,
+       INSN_MIPS4_32_NOT_32R6_64R6, FGR_64;
+defm : MovzPats1<GPR32, FGR64, MOVZ_I_D64, XOR>, INSN_MIPS4_32_NOT_32R6_64R6,
+       FGR_64;
+defm : MovzPats1<GPR64, FGR64, MOVZ_I64_D64, XOR64>,
+       INSN_MIPS4_32_NOT_32R6_64R6, FGR_64;
+defm : MovnPats<GPR32, FGR64, MOVN_I_D64, XOR>, INSN_MIPS4_32_NOT_32R6_64R6,
+       FGR_64;
+defm : MovnPats<GPR64, FGR64, MOVN_I64_D64, XOR64>, INSN_MIPS4_32_NOT_32R6_64R6,
        FGR_64;
-defm : MovzPats1<GPR32, FGR64, MOVZ_I_D64, XOR>, FGR_64;
-defm : MovzPats1<GPR64, FGR64, MOVZ_I64_D64, XOR64>, FGR_64;
-defm : MovnPats<GPR32, FGR64, MOVN_I_D64, XOR>, FGR_64;
-defm : MovnPats<GPR64, FGR64, MOVN_I64_D64, XOR64>, FGR_64;

lib/Target/Mips/MipsISelLowering.h

MachineBasicBlock *emitAtomicBinary(MachineInstr *MI, MachineBasicBlock *BB,
                     unsigned Size, unsigned BinOpcode, bool Nand = false) const;
     MachineBasicBlock *emitAtomicBinaryPartword(MachineInstr *MI,
                     MachineBasicBlock *BB, unsigned Size, unsigned BinOpcode,
                     bool Nand = false) const;
     MachineBasicBlock *emitAtomicCmpSwap(MachineInstr *MI,
                                   MachineBasicBlock *BB, unsigned Size) const;
     MachineBasicBlock *emitAtomicCmpSwapPartword(MachineInstr *MI,
                                   MachineBasicBlock *BB, unsigned Size) const;
+    MachineBasicBlock *emitSEL_D(MachineInstr *MI, MachineBasicBlock *BB) const;
   };
 
   /// Create MipsTargetLowering objects.
   const MipsTargetLowering *createMips16TargetLowering(MipsTargetMachine &TM);
   const MipsTargetLowering *createMipsSETargetLowering(MipsTargetMachine &TM);
 
   namespace Mips {
     FastISel *createFastISel(FunctionLoweringInfo &funcInfo,
                              const TargetLibraryInfo *libInfo);
   }
 }
 
 #endif // MipsISELLOWERING_H

lib/Target/Mips/MipsISelLowering.cpp

MipsTargetLowering::EmitInstrWithCustomInserter(MachineInstr *MI,
   case Mips::PseudoDSDIV:
   case Mips::PseudoDUDIV:
   case Mips::DDIV:
   case Mips::DDIVU:
   case Mips::DMOD:
   case Mips::DMODU:
     return insertDivByZeroTrap(MI, *BB, *getTargetMachine().getInstrInfo(),
                                true);
+  case Mips::SEL_D:
+    return emitSEL_D(MI, BB);
   }
 }
 
 // This function also handles Mips::ATOMIC_SWAP_I32 (when BinOpcode == 0), and
 // Mips::ATOMIC_LOAD_NAND_I32 (when Nand == true)
 MachineBasicBlock *
 MipsTargetLowering::emitAtomicBinary(MachineInstr *MI, MachineBasicBlock *BB,
                                      unsigned Size, unsigned BinOpcode,
MipsTargetLowering::emitAtomicCmpSwapPartword(MachineInstr *MI,
   BuildMI(BB, DL, TII->get(Mips::SRA), Dest)
       .addReg(SllRes).addImm(ShiftImm);
 
   MI->eraseFromParent();   // The instruction is gone now.
 
   return exitMBB;
 }
 
+MachineBasicBlock *MipsTargetLowering::emitSEL_D(MachineInstr *MI,
+                                                 MachineBasicBlock *BB) const {
+  MachineFunction *MF = BB->getParent();
+  const TargetRegisterInfo *TRI = getTargetMachine().getRegisterInfo();
+  const TargetInstrInfo *TII = getTargetMachine().getInstrInfo();
+  MachineRegisterInfo &RegInfo = MF->getRegInfo();
+  DebugLoc DL = MI->getDebugLoc();
+  MachineBasicBlock::iterator II(MI);
+
+  unsigned Fc = MI->getOperand(1).getReg();
+  const auto &FGR64RegClass = TRI->getRegClass(Mips::FGR64RegClassID);
+
+  unsigned Fc2 = RegInfo.createVirtualRegister(FGR64RegClass);
+
+  BuildMI(*BB, II, DL, TII->get(Mips::SUBREG_TO_REG), Fc2)
+      .addImm(0)
+      .addReg(Fc)
+      .addImm(Mips::sub_lo);
+
+  // We don't erase the original instruction, we just replace the condition
+  // register with the 64-bit super-register.
+  MI->getOperand(1).setReg(Fc2);
+
+  return BB;
+}
+
 //===----------------------------------------------------------------------===//
 //  Misc Lower Operation implementation
 //===----------------------------------------------------------------------===//
 SDValue MipsTargetLowering::lowerBR_JT(SDValue Op, SelectionDAG &DAG) const {
   SDValue Chain = Op.getOperand(0);
   SDValue Table = Op.getOperand(1);
   SDValue Index = Op.getOperand(2);
   SDLoc DL(Op);

 
 SDValue MipsTargetLowering::lowerBRCOND(SDValue Op, SelectionDAG &DAG) const {
   // The first operand is the chain, the second is the condition, the third is
   // the block to branch to if the condition is true.
   SDValue Chain = Op.getOperand(0);
   SDValue Dest = Op.getOperand(2);
   SDLoc DL(Op);
 
+  assert(!Subtarget->hasMips32r6() && !Subtarget->hasMips64r6());
   SDValue CondRes = createFPCmp(DAG, Op.getOperand(1));
 
   // Return if flag is not set by a floating point comparison.
   if (CondRes.getOpcode() != MipsISD::FPCmp)
     return Op;
 
   SDValue CCNode  = CondRes.getOperand(2);
   Mips::CondCode CC =
     (Mips::CondCode)cast<ConstantSDNode>(CCNode)->getZExtValue();
   unsigned Opc = invertFPCondCodeUser(CC) ? Mips::BRANCH_F : Mips::BRANCH_T;
   SDValue BrCode = DAG.getConstant(Opc, MVT::i32);
   SDValue FCC0 = DAG.getRegister(Mips::FCC0, MVT::i32);
   return DAG.getNode(MipsISD::FPBrcond, DL, Op.getValueType(), Chain, BrCode,
                      FCC0, Dest, CondRes);
 }
 
 SDValue MipsTargetLowering::
 lowerSELECT(SDValue Op, SelectionDAG &DAG) const
 {
+  assert(!Subtarget->hasMips32r6() && !Subtarget->hasMips64r6());
   SDValue Cond = createFPCmp(DAG, Op.getOperand(0));
 
   // Return if flag is not set by a floating point comparison.
   if (Cond.getOpcode() != MipsISD::FPCmp)
     return Op;
 
   return createCMovFP(DAG, Cond, Op.getOperand(1), Op.getOperand(2),
                       SDLoc(Op));
SDValue Cond = DAG.getNode(ISD::SETCC, DL,
                              Op.getOperand(0), Op.getOperand(1),
                              Op.getOperand(4));
 
   return DAG.getNode(ISD::SELECT, DL, Op.getValueType(), Cond, Op.getOperand(2),
                      Op.getOperand(3));
 }
 
 SDValue MipsTargetLowering::lowerSETCC(SDValue Op, SelectionDAG &DAG) const {
+  assert(!Subtarget->hasMips32r6() && !Subtarget->hasMips64r6());
   SDValue Cond = createFPCmp(DAG, Op);
 
   assert(Cond.getOpcode() == MipsISD::FPCmp &&
          "Floating point operand expected.");
 
   SDValue True  = DAG.getConstant(1, MVT::i32);
   SDValue False = DAG.getConstant(0, MVT::i32);
 

lib/Target/Mips/MipsInstrFPU.td

multiclass C_COND_M<string TypeStr, RegisterOperand RC, bits<5> fmt,
   def C_SEQ_#NAME : C_COND_FT<"seq", TypeStr, RC, itin>, C_COND_FM<fmt, 10>;
   def C_NGL_#NAME : C_COND_FT<"ngl", TypeStr, RC, itin>, C_COND_FM<fmt, 11>;
   def C_LT_#NAME : C_COND_FT<"lt", TypeStr, RC, itin>, C_COND_FM<fmt, 12>;
   def C_NGE_#NAME : C_COND_FT<"nge", TypeStr, RC, itin>, C_COND_FM<fmt, 13>;
   def C_LE_#NAME : C_COND_FT<"le", TypeStr, RC, itin>, C_COND_FM<fmt, 14>;
   def C_NGT_#NAME : C_COND_FT<"ngt", TypeStr, RC, itin>, C_COND_FM<fmt, 15>;
 }
 
-defm S : C_COND_M<"s", FGR32Opnd, 16, II_C_CC_S>;
-defm D32 : C_COND_M<"d", AFGR64Opnd, 17, II_C_CC_D>,
+defm S : C_COND_M<"s", FGR32Opnd, 16, II_C_CC_S>, ISA_MIPS1_NOT_32R6_64R6;
+defm D32 : C_COND_M<"d", AFGR64Opnd, 17, II_C_CC_D>, ISA_MIPS1_NOT_32R6_64R6,
            AdditionalRequires<[NotFP64bit]>;
 let DecoderNamespace = "Mips64" in
-defm D64 : C_COND_M<"d", FGR64Opnd, 17, II_C_CC_D>,
+defm D64 : C_COND_M<"d", FGR64Opnd, 17, II_C_CC_D>, ISA_MIPS1_NOT_32R6_64R6,
            AdditionalRequires<[IsFP64bit]>;
 
 //===----------------------------------------------------------------------===//
 // Floating Point Instructions
 //===----------------------------------------------------------------------===//
 def ROUND_W_S  : MMRel, ABSS_FT<"round.w.s", FGR32Opnd, FGR32Opnd, II_ROUND>,
                  ABSS_FM<0xc, 16>, ISA_MIPS2;
 def TRUNC_W_S  : MMRel, ABSS_FT<"trunc.w.s", FGR32Opnd, FGR32Opnd, II_TRUNC>,

 def MIPS_FCOND_SEQ  : PatLeaf<(i32 10)>;
 def MIPS_FCOND_NGL  : PatLeaf<(i32 11)>;
 def MIPS_FCOND_LT   : PatLeaf<(i32 12)>;
 def MIPS_FCOND_NGE  : PatLeaf<(i32 13)>;
 def MIPS_FCOND_LE   : PatLeaf<(i32 14)>;
 def MIPS_FCOND_NGT  : PatLeaf<(i32 15)>;
 
 /// Floating Point Compare
-def FCMP_S32 : MMRel, CEQS_FT<"s", FGR32, II_C_CC_S, MipsFPCmp>, CEQS_FM<16>;
+def FCMP_S32 : MMRel, CEQS_FT<"s", FGR32, II_C_CC_S, MipsFPCmp>, CEQS_FM<16>,
+               ISA_MIPS1_NOT_32R6_64R6;
 def FCMP_D32 : MMRel, CEQS_FT<"d", AFGR64, II_C_CC_D, MipsFPCmp>, CEQS_FM<17>,
-               AdditionalRequires<[NotFP64bit]>;
+               ISA_MIPS1_NOT_32R6_64R6, AdditionalRequires<[NotFP64bit]>;
 let DecoderNamespace = "Mips64" in
 def FCMP_D64 : CEQS_FT<"d", FGR64, II_C_CC_D, MipsFPCmp>, CEQS_FM<17>,
-               AdditionalRequires<[IsFP64bit]>;
+               ISA_MIPS1_NOT_32R6_64R6, AdditionalRequires<[IsFP64bit]>;
 
 //===----------------------------------------------------------------------===//
 // Floating Point Pseudo-Instructions
 //===----------------------------------------------------------------------===//
 
 // This pseudo instr gets expanded into 2 mtc1 instrs after register
 // allocation.
 class BuildPairF64Base<RegisterOperand RO> :

lib/Target/Mips/MipsInstrInfo.td

 class ISA_MIPS64R6 { list<Predicate> InsnPredicates = [HasMips64r6]; }
 
 // The portions of MIPS-III that were also added to MIPS32
 class INSN_MIPS3_32 { list<Predicate> InsnPredicates = [HasMips3_32]; }
 
 // The portions of MIPS-III that were also added to MIPS32
 class INSN_MIPS3_32R2 { list<Predicate> InsnPredicates = [HasMips3_32r2]; }
 
-// The portions of MIPS-IV that were also added to MIPS32
-class INSN_MIPS4_32 { list<Predicate> InsnPredicates = [HasMips4_32]; }
+// The portions of MIPS-IV that were also added to MIPS32 but were removed in
+// MIPS32r6 and MIPS64r6.
+class INSN_MIPS4_32_NOT_32R6_64R6 {
+  list<Predicate> InsnPredicates = [HasMips4_32, NotMips32r6, NotMips64r6];
+}
 
 // The portions of MIPS-IV that were also added to MIPS32R2
 class INSN_MIPS4_32R2 { list<Predicate> InsnPredicates = [HasMips4_32r2]; }
 
 // The portions of MIPS-V that were also added to MIPS32R2
 class INSN_MIPS5_32R2 { list<Predicate> InsnPredicates = [HasMips5_32r2]; }
 
 //===----------------------------------------------------------------------===//

lib/Target/Mips/MipsRegisterInfo.td

 // FP control registers.
 def CCR : RegisterClass<"Mips", [i32], 32, (sequence "FCR%u", 0, 31)>,
           Unallocatable;
 
 // FP condition code registers.
 def FCC : RegisterClass<"Mips", [i32], 32, (sequence "FCC%u", 0, 7)>,
           Unallocatable;
 
+// MIPS32r6/MIPS64r6 store FPU condition codes in normal FGR registers.
+// This class allows us to represent this in codegen patterns.
+def FGRCC : RegisterClass<"Mips", [i32], 32, (sequence "F%u", 0, 31)>;
+
 def MSA128B: RegisterClass<"Mips", [v16i8], 128,
                            (sequence "W%u", 0, 31)>;
 def MSA128H: RegisterClass<"Mips", [v8i16, v8f16], 128,
                            (sequence "W%u", 0, 31)>;
 def MSA128W: RegisterClass<"Mips", [v4i32, v4f32], 128,
                            (sequence "W%u", 0, 31)>;
 def MSA128D: RegisterClass<"Mips", [v2i64, v2f64], 128,
                            (sequence "W%u", 0, 31)>;

 def FGR64Opnd : RegisterOperand<FGR64> {
   let ParserMatchClass = FGR64AsmOperand;
 }
 
 def FGR32Opnd : RegisterOperand<FGR32> {
   let ParserMatchClass = FGR32AsmOperand;
 }
 
+def FGRCCOpnd : RegisterOperand<FGRCC> {
+  // The assembler doesn't use register classes so we can re-use
+  // FGR32AsmOperand.
+  let ParserMatchClass = FGR32AsmOperand;
+}
+
 def FGRH32Opnd : RegisterOperand<FGRH32> {
   let ParserMatchClass = FGRH32AsmOperand;
 }
 
 def FCCRegsOpnd : RegisterOperand<FCC> {
   let ParserMatchClass = FCCRegsAsmOperand;
 }
 

lib/Target/Mips/MipsSEISelLowering.cpp

if (Subtarget->hasMips32r6()) {
     // MIPS32r6 replaces the accumulator-based division/remainder with separate
     // three register division and remainder instructions.
     setOperationAction(ISD::SDIVREM, MVT::i32, Expand);
     setOperationAction(ISD::UDIVREM, MVT::i32, Expand);
     setOperationAction(ISD::SDIV, MVT::i32, Legal);
     setOperationAction(ISD::UDIV, MVT::i32, Legal);
     setOperationAction(ISD::SREM, MVT::i32, Legal);
     setOperationAction(ISD::UREM, MVT::i32, Legal);
+
+    // MIPS32r6 replaces conditional moves with an equivalent that removes the
+    // need for three GPR read ports.
+    setOperationAction(ISD::SETCC, MVT::i32, Legal);
+    setOperationAction(ISD::SELECT, MVT::i32, Legal);
+    setOperationAction(ISD::SELECT_CC, MVT::i32, Expand);
+
+    setOperationAction(ISD::SETCC, MVT::f32, Legal);
+    setOperationAction(ISD::SELECT, MVT::f32, Legal);
+    setOperationAction(ISD::SELECT_CC, MVT::f32, Expand);
+
+    assert(Subtarget->isFP64bit() && "FR=1 is required for MIPS32r6");
+    setOperationAction(ISD::SETCC, MVT::f64, Legal);
+    setOperationAction(ISD::SELECT, MVT::f64, Legal);
+    setOperationAction(ISD::SELECT_CC, MVT::f64, Expand);
+
+    // Floating point > and >= are supported via < and <=
+    setCondCodeAction(ISD::SETOGE, MVT::f32, Expand);
+    setCondCodeAction(ISD::SETOGT, MVT::f32, Expand);
+    setCondCodeAction(ISD::SETUGE, MVT::f32, Expand);
+    setCondCodeAction(ISD::SETUGT, MVT::f32, Expand);
+
+    setCondCodeAction(ISD::SETOGE, MVT::f64, Expand);
+    setCondCodeAction(ISD::SETOGT, MVT::f64, Expand);
+    setCondCodeAction(ISD::SETUGE, MVT::f64, Expand);
+    setCondCodeAction(ISD::SETUGT, MVT::f64, Expand);
   }
 
   if (Subtarget->hasMips64r6()) {
     // MIPS64r6 replaces the accumulator-based multiplies with a three register
     // instruction
     setOperationAction(ISD::MUL, MVT::i64, Legal);
     setOperationAction(ISD::MULHS, MVT::i64, Legal);
     setOperationAction(ISD::MULHU, MVT::i64, Legal);
 
     // MIPS32r6 replaces the accumulator-based division/remainder with separate
     // three register division and remainder instructions.
     setOperationAction(ISD::SDIVREM, MVT::i64, Expand);
     setOperationAction(ISD::UDIVREM, MVT::i64, Expand);
     setOperationAction(ISD::SDIV, MVT::i64, Legal);
     setOperationAction(ISD::UDIV, MVT::i64, Legal);
     setOperationAction(ISD::SREM, MVT::i64, Legal);
     setOperationAction(ISD::UREM, MVT::i64, Legal);
+
+    // MIPS64r6 replaces conditional moves with an equivalent that removes the
+    // need for three GPR read ports.
+    setOperationAction(ISD::SETCC, MVT::i64, Legal);
+    setOperationAction(ISD::SELECT, MVT::i64, Legal);
+    setOperationAction(ISD::SELECT_CC, MVT::i64, Expand);
   }
 
   computeRegisterProperties();
 }
 
 const MipsTargetLowering *
 llvm::createMipsSETargetLowering(MipsTargetMachine &TM) {
   return new MipsSETargetLowering(TM);

test/CodeGen/Mips/cmov.ll

-; RUN: llc -march=mips < %s | FileCheck %s -check-prefix=O32
-; RUN: llc -march=mips -regalloc=basic < %s | FileCheck %s -check-prefix=O32
-; RUN: llc -march=mips64el -mcpu=mips4 -mattr=n64 < %s | FileCheck %s -check-prefix=N64
-; RUN: llc -march=mips64el -mcpu=mips64 -mattr=n64 < %s | FileCheck %s -check-prefix=N64
+; RUN: llc -march=mips     -mcpu=mips32                 < %s | FileCheck %s -check-prefix=ALL -check-prefix=32-CMOV
+; RUN: llc -march=mips     -mcpu=mips32 -regalloc=basic < %s | FileCheck %s -check-prefix=ALL -check-prefix=32-CMOV
+; RUN: llc -march=mips     -mcpu=mips32r2               < %s | FileCheck %s -check-prefix=ALL -check-prefix=32-CMOV
+; RUN: llc -march=mips     -mcpu=mips32r6               < %s | FileCheck %s -check-prefix=ALL -check-prefix=32-CMP
+; RUN: llc -march=mips64el -mcpu=mips4                  < %s | FileCheck %s -check-prefix=ALL -check-prefix=64-CMOV
+; RUN: llc -march=mips64el -mcpu=mips64                 < %s | FileCheck %s -check-prefix=ALL -check-prefix=64-CMOV
+; RUN: llc -march=mips64el -mcpu=mips64r6               < %s | FileCheck %s -check-prefix=ALL -check-prefix=64-CMP
 
 @i1 = global [3 x i32] [i32 1, i32 2, i32 3], align 4
 @i3 = common global i32* null, align 4
 
-; O32-DAG:  lw $[[R0:[0-9]+]], %got(i3)
-; O32-DAG:  addiu $[[R1:[0-9]+]], ${{[0-9]+}}, %got(i1)
-; O32:      movn $[[R0]], $[[R1]], ${{[0-9]+}}
-; N64-DAG:  ldr $[[R0:[0-9]+]]
-; N64-DAG:  ld $[[R1:[0-9]+]], %got_disp(i1)
-; N64:      movn $[[R0]], $[[R1]], ${{[0-9]+}}
+; ALL-LABEL: cmov1:
+
+; 32-CMOV-DAG:  lw $[[R0:[0-9]+]], %got(i3)
+; 32-CMOV-DAG:  addiu $[[R1:[0-9]+]], ${{[0-9]+}}, %got(i1)
+; 32-CMOV-DAG:  movn $[[R0]], $[[R1]], $4
+; 32-CMOV-DAG:  lw $2, 0($[[R0]])
+
+; 32-CMP-DAG:   lw $[[R0:[0-9]+]], %got(i3)
+; 32-CMP-DAG:   addiu $[[R1:[0-9]+]], ${{[0-9]+}}, %got(i1)
+; 32-CMP-DAG:   seleqz $[[T0:[0-9]+]], $[[R1]], $4
+; 32-CMP-DAG:   selnez $[[T1:[0-9]+]], $[[R0]], $4
+; 32-CMP-DAG:   or $[[T2:[0-9]+]], $[[T1]], $[[T0]]
+; 32-CMP-DAG:   lw $2, 0($[[T2]])
+
+; 64-CMOV-DAG:  ldr $[[R0:[0-9]+]]
+; 64-CMOV-DAG:  ld $[[R1:[0-9]+]], %got_disp(i1)
+; 64-CMOV-DAG:  movn $[[R0]], $[[R1]], $4
+
+; 64-CMP-DAG:   ld $[[R0:[0-9]+]], %got_disp(i3)(
+; 64-CMP-DAG:   daddiu $[[R1:[0-9]+]], ${{[0-9]+}}, %got_disp(i1)
+; FIXME: This sll works around an implementation detail in the code generator
+;        (setcc's result is i32 so bits 32-63 are undefined). It's not really
+;        needed.
+; 64-CMP-DAG:   sll $[[CC:[0-9]+]], $4, 0
+; 64-CMP-DAG:   seleqz $[[T0:[0-9]+]], $[[R1]], $[[CC]]
+; 64-CMP-DAG:   selnez $[[T1:[0-9]+]], $[[R0]], $[[CC]]
+; 64-CMP-DAG:   or $[[T2:[0-9]+]], $[[T1]], $[[T0]]
+; 64-CMP-DAG:   ld $2, 0($[[T2]])
+
 define i32* @cmov1(i32 %s) nounwind readonly {
 entry:
   %tobool = icmp ne i32 %s, 0
   %tmp1 = load i32** @i3, align 4
   %cond = select i1 %tobool, i32* getelementptr inbounds ([3 x i32]* @i1, i32 0, i32 0), i32* %tmp1
   ret i32* %cond
 }
 
 @c = global i32 1, align 4
 @d = global i32 0, align 4
 
-; O32-LABEL: cmov2:
-; O32: addiu $[[R1:[0-9]+]], ${{[a-z0-9]+}}, %got(d)
-; O32: addiu $[[R0:[0-9]+]], ${{[a-z0-9]+}}, %got(c)
-; O32: movn  $[[R1]], $[[R0]], ${{[0-9]+}}
-; N64-LABEL: cmov2:
-; N64: daddiu $[[R1:[0-9]+]], ${{[0-9]+}}, %got_disp(d)
-; N64: daddiu $[[R0:[0-9]+]], ${{[0-9]+}}, %got_disp(c)
-; N64: movn  $[[R1]], $[[R0]], ${{[0-9]+}}
+; ALL-LABEL: cmov2:
+
+; 32-CMOV-DAG:  addiu $[[R1:[0-9]+]], ${{[0-9]+}}, %got(d)
+; 32-CMOV-DAG:  addiu $[[R0:[0-9]+]], ${{[0-9]+}}, %got(c)
+; 32-CMOV-DAG:  movn  $[[R1]], $[[R0]], $4
+; 32-CMOV-DAG:  lw $2, 0($[[R0]])
+
+; 32-CMP-DAG:   addiu $[[R1:[0-9]+]], ${{[0-9]+}}, %got(d)
+; 32-CMP-DAG:   addiu $[[R0:[0-9]+]], ${{[0-9]+}}, %got(c)
+; 32-CMP-DAG:   seleqz $[[T0:[0-9]+]], $[[R0]], $4
+; 32-CMP-DAG:   selnez $[[T1:[0-9]+]], $[[R1]], $4
+; 32-CMP-DAG:   or $[[T2:[0-9]+]], $[[T1]], $[[T0]]
+; 32-CMP-DAG:   lw $2, 0($[[T2]])
+
+; 64-CMOV:      daddiu $[[R1:[0-9]+]], ${{[0-9]+}}, %got_disp(d)
+; 64-CMOV:      daddiu $[[R0:[0-9]+]], ${{[0-9]+}}, %got_disp(c)
+; 64-CMOV:      movn  $[[R1]], $[[R0]], $4
+
+; 64-CMP-DAG:   daddiu $[[R1:[0-9]+]], ${{[0-9]+}}, %got_disp(d)
+; 64-CMP-DAG:   daddiu $[[R0:[0-9]+]], ${{[0-9]+}}, %got_disp(c)
+; FIXME: This sll works around an implementation detail in the code generator
+;        (setcc's result is i32 so bits 32-63 are undefined). It's not really
+;        needed.
+; 64-CMP-DAG:   sll $[[CC:[0-9]+]], $4, 0
+; 64-CMP-DAG:   seleqz $[[T0:[0-9]+]], $[[R0]], $[[CC]]
+; 64-CMP-DAG:   selnez $[[T1:[0-9]+]], $[[R1]], $[[CC]]
+; 64-CMP-DAG:   or $[[T2:[0-9]+]], $[[T1]], $[[T0]]
+; 64-CMP-DAG:   lw $2, 0($[[T2]])
+
 define i32 @cmov2(i32 %s) nounwind readonly {
 entry:
   %tobool = icmp ne i32 %s, 0
   %tmp1 = load i32* @c, align 4
   %tmp2 = load i32* @d, align 4
   %cond = select i1 %tobool, i32 %tmp1, i32 %tmp2
   ret i32 %cond
 }
 
-; O32-LABEL: cmov3:
-; O32: xori $[[R0:[0-9]+]], ${{[0-9]+}}, 234
-; O32: movz ${{[0-9]+}}, ${{[0-9]+}}, $[[R0]]
+; ALL-LABEL: cmov3:
+
+; We won't check the result register since we can't know if the move is first
+; or last. We do know it will be either one of two registers so we can at least
+; check that.
+
+; 32-CMOV:      xori $[[R0:[0-9]+]], $4, 234
+; 32-CMOV:      movz ${{[26]}}, $5, $[[R0]]
+
+; 32-CMP-DAG:   xori $[[CC:[0-9]+]], $4, 234
+; 32-CMP-DAG:   selnez $[[T0:[0-9]+]], $5, $[[CC]]
+; 32-CMP-DAG:   seleqz $[[T1:[0-9]+]], $6, $[[CC]]
+; 32-CMP-DAG:   or $2, $[[T0]], $[[T1]]
+
+; 64-CMOV:      xori $[[R0:[0-9]+]], $4, 234
+; 64-CMOV:      movz ${{[26]}}, $5, $[[R0]]
+
+; 64-CMP-DAG:   xori $[[CC:[0-9]+]], $4, 234
+; 64-CMP-DAG:   selnez $[[T0:[0-9]+]], $5, $[[CC]]
+; 64-CMP-DAG:   seleqz $[[T1:[0-9]+]], $6, $[[CC]]
+; 64-CMP-DAG:   or $2, $[[T0]], $[[T1]]
+
 define i32 @cmov3(i32 %a, i32 %b, i32 %c) nounwind readnone {
 entry:
   %cmp = icmp eq i32 %a, 234
   %cond = select i1 %cmp, i32 %b, i32 %c
   ret i32 %cond
 }
 
-; N64-LABEL: cmov4:
-; N64: xori $[[R0:[0-9]+]], ${{[0-9]+}}, 234
-; N64: movz ${{[0-9]+}}, ${{[0-9]+}}, $[[R0]]
+; ALL-LABEL: cmov4:
+
+; We won't check the result register since we can't know if the move is first
+; or last. We do know it will be one of two registers so we can at least check
+; that.
+
+; 32-CMOV-DAG: xori $[[R0:[0-9]+]], $4, 234
+; 32-CMOV-DAG: lw $[[R1:2]], 16($sp)
+; 32-CMOV-DAG: lw $[[R2:3]], 20($sp)
+; 32-CMOV-DAG: movz $[[R1]], $6, $[[R0]]
+; 32-CMOV-DAG: movz $[[R2]], $7, $[[R0]]
+
+; 32-CMP-DAG:  xori $[[R0:[0-9]+]], $4, 234
+; 32-CMP-DAG:  lw $[[R1:[0-9]+]], 16($sp)
+; 32-CMP-DAG:  lw $[[R2:[0-9]+]], 20($sp)
+; 32-CMP-DAG:  selnez $[[T0:[0-9]+]], $6, $[[R0]]
+; 32-CMP-DAG:  selnez $[[T1:[0-9]+]], $7, $[[R0]]
+; 32-CMP-DAG:  seleqz $[[T2:[0-9]+]], $[[R1]], $[[R0]]
+; 32-CMP-DAG:  seleqz $[[T3:[0-9]+]], $[[R2]], $[[R0]]
+; 32-CMP-DAG:  or $2, $[[T0]], $[[T2]]
+; 32-CMP-DAG:  or $3, $[[T1]], $[[T3]]
+
+; 64-CMOV: xori $[[R0:[0-9]+]], $4, 234
+; 64-CMOV: movz ${{[26]}}, $5, $[[R0]]
+
+; 64-CMP-DAG:  xori $[[R0:[0-9]+]], $4, 234
+; 64-CMP-DAG:  selnez $[[T0:[0-9]+]], $5, $[[R0]]
+; 64-CMP-DAG:  seleqz $[[T1:[0-9]+]], $6, $[[R0]]
+; 64-CMP-DAG:  or $2, $[[T0]], $[[T1]]
+
 define i64 @cmov4(i32 %a, i64 %b, i64 %c) nounwind readnone {
 entry:
   %cmp = icmp eq i32 %a, 234
   %cond = select i1 %cmp, i64 %b, i64 %c
   ret i64 %cond
 }
 
 ; slti and conditional move.
 ;
 ; Check that, pattern
 ;  (select (setgt a, N), t, f)
 ; turns into
 ;  (movz t, (setlt a, N + 1), f)
 ; if N + 1 fits in 16-bit.
 
-; O32-LABEL: slti0:
-; O32: slti $[[R0:[0-9]+]], ${{[0-9]+}}, 32767
-; O32: movz ${{[0-9]+}}, ${{[0-9]+}}, $[[R0]]
+; ALL-LABEL: slti0:
+
+; 32-CMOV-DAG: addiu $[[I3:[0-9]+]], $zero, 3
+; 32-CMOV-DAG: addiu $[[I5:2]], $zero, 5
+; 32-CMOV-DAG: slti $[[R0:[0-9]+]], $4, 32767
+; 32-CMOV-DAG: movz $[[I5]], $[[I3]], $[[R0]]
+
+; 32-CMP-DAG:  addiu $[[I3:[0-9]+]], $zero, 3
+; 32-CMP-DAG:  addiu $[[I5:[0-9]+]], $zero, 5
+; 32-CMP-DAG:  slti $[[R0:[0-9]+]], $4, 32767
+; FIXME: We can do better than this by using selccz to choose between +0 and +2
+; 32-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I3]], $[[R0]]
+; 32-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I5]], $[[R0]]
+; 32-CMP-DAG:  or $2, $[[T0]], $[[T1]]
+
+; 64-CMOV-DAG: addiu $[[I3:[0-9]+]], $zero, 3
+; 64-CMOV-DAG: addiu $[[I5:2]], $zero, 5
+; 64-CMOV-DAG: slti $[[R0:[0-9]+]], $4, 32767
+; 64-CMOV-DAG: movz $[[I5]], $[[I3]], $[[R0]]
+
+; 64-CMP-DAG:  addiu $[[I3:[0-9]+]], $zero, 3
+; 64-CMP-DAG:  addiu $[[I5:[0-9]+]], $zero, 5
+; 64-CMP-DAG:  slti $[[R0:[0-9]+]], $4, 32767
+; FIXME: We can do better than this by using selccz to choose between +0 and +2
+; 64-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I3]], $[[R0]]
+; 64-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I5]], $[[R0]]
+; 64-CMP-DAG:  or $2, $[[T0]], $[[T1]]
 
 define i32 @slti0(i32 %a) {
 entry:
   %cmp = icmp sgt i32 %a, 32766
   %cond = select i1 %cmp, i32 3, i32 5
   ret i32 %cond
 }
 
-; O32-LABEL: slti1:
-; O32: slt ${{[0-9]+}}
+; ALL-LABEL: slti1:
+
+; 32-CMOV-DAG: addiu $[[I7:[0-9]+]], $zero, 7
+; 32-CMOV-DAG: addiu $[[I5:2]], $zero, 5
+; 32-CMOV-DAG: addiu $[[R1:[0-9]+]], $zero, 32767
+; 32-CMOV-DAG: slt $[[R0:[0-9]+]], $[[R1]], $4
+; 32-CMOV-DAG: movn $[[I5]], $[[I7]], $[[R0]]
+
+; 32-CMP-DAG:  addiu $[[I7:[0-9]+]], $zero, 7
+; 32-CMP-DAG:  addiu $[[I5:[0-9]+]], $zero, 5
+; 32-CMP-DAG:  addiu $[[I32767:[0-9]+]], $zero, 32767
+; 32-CMP-DAG:  slt $[[R0:[0-9]+]], $[[I32767]], $4
+; FIXME: We can do better than this by using selccz to choose between -0 and -2
+; 32-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I7]], $[[R0]]
+; 32-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I5]], $[[R0]]
+; 32-CMP-DAG:  or $2, $[[T0]], $[[T1]]
+
+; 64-CMOV-DAG: addiu $[[I7:[0-9]+]], $zero, 7
+; 64-CMOV-DAG: addiu $[[I5:2]], $zero, 5
+; 64-CMOV-DAG: addiu $[[R1:[0-9]+]], $zero, 32767
+; 64-CMOV-DAG: slt $[[R0:[0-9]+]], $[[R1]], $4
+; 64-CMOV-DAG: movn $[[I5]], $[[I7]], $[[R0]]
+
+; 64-CMP-DAG:  addiu $[[I7:[0-9]+]], $zero, 7
+; 64-CMP-DAG:  addiu $[[I5:2]], $zero, 5
+; 64-CMP-DAG:  addiu $[[R1:[0-9]+]], $zero, 32767
+; 64-CMP-DAG:  slt $[[R0:[0-9]+]], $[[R1]], $4
+; FIXME: We can do better than this by using selccz to choose between -0 and -2
+; 64-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I7]], $[[R0]]
+; 64-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I5]], $[[R0]]
+; 64-CMP-DAG:  or $2, $[[T0]], $[[T1]]
 
 define i32 @slti1(i32 %a) {
 entry:
   %cmp = icmp sgt i32 %a, 32767
-  %cond = select i1 %cmp, i32 3, i32 5
+  %cond = select i1 %cmp, i32 7, i32 5
   ret i32 %cond
 }
 
-; O32-LABEL: slti2:
-; O32: slti $[[R0:[0-9]+]], ${{[0-9]+}}, -32768
-; O32: movz ${{[0-9]+}}, ${{[0-9]+}}, $[[R0]]
+; ALL-LABEL: slti2:
+
+; 32-CMOV-DAG: addiu $[[I3:[0-9]+]], $zero, 3
+; 32-CMOV-DAG: addiu $[[I5:2]], $zero, 5
+; 32-CMOV-DAG: slti $[[R0:[0-9]+]], $4, -32768
+; 32-CMOV-DAG: movz $[[I5]], $[[I3]], $[[R0]]
+
+; 32-CMP-DAG:  addiu $[[I3:[0-9]+]], $zero, 3
+; 32-CMP-DAG:  addiu $[[I5:[0-9]+]], $zero, 5
+; 32-CMP-DAG:  slti $[[R0:[0-9]+]], $4, -32768
+; FIXME: We can do better than this by using selccz to choose between +0 and +2
+; 32-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I3]], $[[R0]]
+; 32-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I5]], $[[R0]]
+; 32-CMP-DAG:  or $2, $[[T0]], $[[T1]]
+
+; 64-CMOV-DAG: addiu $[[I3:[0-9]+]], $zero, 3
+; 64-CMOV-DAG: addiu $[[I5:2]], $zero, 5
+; 64-CMOV-DAG: slti $[[R0:[0-9]+]], $4, -32768
+; 64-CMOV-DAG: movz $[[I5]], $[[I3]], $[[R0]]
+
+; 64-CMP-DAG:  addiu $[[I3:[0-9]+]], $zero, 3
+; 64-CMP-DAG:  addiu $[[I5:[0-9]+]], $zero, 5
+; 64-CMP-DAG:  slti $[[R0:[0-9]+]], $4, -32768
+; FIXME: We can do better than this by using selccz to choose between +0 and +2
+; 64-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I3]], $[[R0]]
+; 64-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I5]], $[[R0]]
+; 64-CMP-DAG:  or $2, $[[T0]], $[[T1]]
 
 define i32 @slti2(i32 %a) {
 entry:
   %cmp = icmp sgt i32 %a, -32769
   %cond = select i1 %cmp, i32 3, i32 5
   ret i32 %cond
 }
 
-; O32-LABEL: slti3:
-; O32: slt ${{[0-9]+}}
+; ALL-LABEL: slti3:
+
+; 32-CMOV-DAG: addiu $[[I3:[0-9]+]], $zero, 3
+; 32-CMOV-DAG: addiu $[[I5:2]], $zero, 5
+; 32-CMOV-DAG: lui $[[R1:[0-9]+]], 65535
+; 32-CMOV-DAG: ori $[[R1]], $[[R1]], 32766
+; 32-CMOV-DAG: slt $[[R0:[0-9]+]], $[[R1]], $4
+; 32-CMOV-DAG: movn $[[I5]], $[[I3]], $[[R0]]
+
+; 32-CMP-DAG:  addiu $[[I3:[0-9]+]], $zero, 3
+; 32-CMP-DAG:  addiu $[[I5:[0-9]+]], $zero, 5
+; 32-CMP-DAG:  lui $[[IMM:[0-9]+]], 65535
+; 32-CMP-DAG:  ori $[[IMM]], $[[IMM]], 32766
+; 32-CMP-DAG:  slt $[[R0:[0-9]+]], $[[I32767]], $4
+; FIXME: We can do better than this by using selccz to choose between -0 and -2
+; 32-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I3]], $[[R0]]
+; 32-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I5]], $[[R0]]
+; 32-CMP-DAG:  or $2, $[[T0]], $[[T1]]
+
+; 64-CMOV-DAG: addiu $[[I3:[0-9]+]], $zero, 3
+; 64-CMOV-DAG: addiu $[[I5:2]], $zero, 5
+; 64-CMOV-DAG: lui $[[R1:[0-9]+]], 65535
+; 64-CMOV-DAG: ori $[[R1]], $[[R1]], 32766
+; 64-CMOV-DAG: slt $[[R0:[0-9]+]], $[[R1]], $4
+; 64-CMOV-DAG: movn $[[I5]], $[[I3]], $[[R0]]
+
+; 64-CMP-DAG:  addiu $[[I3:[0-9]+]], $zero, 3
+; 64-CMP-DAG:  addiu $[[I5:2]], $zero, 5
+; 64-CMP-DAG:  lui $[[IMM:[0-9]+]], 65535
+; 64-CMP-DAG:  ori $[[IMM]], $[[IMM]], 32766
+; 64-CMP-DAG:  slt $[[R0:[0-9]+]], $[[IMM]], $4
+; FIXME: We can do better than this by using selccz to choose between -0 and -2
+; 64-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I3]], $[[R0]]
+; 64-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I5]], $[[R0]]
+; 64-CMP-DAG:  or $2, $[[T0]], $[[T1]]
 
 define i32 @slti3(i32 %a) {
 entry:
   %cmp = icmp sgt i32 %a, -32770
   %cond = select i1 %cmp, i32 3, i32 5
   ret i32 %cond
 }
 
 ; 64-bit patterns.
 
-; N64-LABEL: slti64_0:
-; N64: slti $[[R0:[0-9]+]], ${{[0-9]+}}, 32767
-; N64: movz ${{[0-9]+}}, ${{[0-9]+}}, $[[R0]]
+; ALL-LABEL: slti64_0:
+
+; 32-CMOV-DAG:  slt $[[CC:[0-9]+]], $zero, $4
+; 32-CMOV-DAG:  addiu $[[I32766:[0-9]+]], $zero, 32766
+; 32-CMOV-DAG:  sltu $[[R1:[0-9]+]], $[[I32766]], $5
+; 32-CMOV-DAG:  movz $[[CC:[0-9]+]], $[[R1]], $4
+; 32-CMOV-DAG:  addiu $[[I5:[0-9]+]], $zero, 5
+; 32-CMOV-DAG:  addiu $[[I4:3]], $zero, 4
+; 32-CMOV-DAG:  movn $[[I4]], $[[I5]], $[[CC]]
+; 32-CMOV-DAG:  addiu $2, $zero, 0
+
+; 32-CMP-DAG:   slt $[[CC0:[0-9]+]], $zero, $4
+; 32-CMP-DAG:   addiu $[[I32766:[0-9]+]], $zero, 32766
+; 32-CMP-DAG:   sltu $[[CC1:[0-9]+]], $[[I32766]], $5
+; 32-CMP-DAG:   seleqz $[[CC2:[0-9]+]], $[[CC0]], $4
+; 32-CMP-DAG:   selnez $[[CC3:[0-9]+]], $[[CC1]], $4
+; 32-CMP:       or $[[CC:[0-9]+]], $[[CC3]], $[[CC2]]
+; 32-CMP-DAG:   addiu $[[I5:[0-9]+]], $zero, 5
+; 32-CMP-DAG:   addiu $[[I4:[0-9]+]], $zero, 4
+; 32-CMP-DAG:   seleqz $[[T0:[0-9]+]], $[[I4]], $[[CC]]
+; 32-CMP-DAG:   selnez $[[T1:[0-9]+]], $[[I5]], $[[CC]]
+; 32-CMP-DAG:   or $3, $[[T1]], $[[T0]]
+; 32-CMP-DAG:   addiu $2, $zero, 0
+
+; 64-CMOV-DAG:  addiu $[[I5:[0-9]+]], $zero, 5
+; 64-CMOV-DAG:  addiu $[[I4:2]], $zero, 4
+; 64-CMOV-DAG:  slti $[[R0:[0-9]+]], $4, 32767
+; 64-CMOV-DAG:  movz $[[I4]], $[[I5]], $[[R0]]
+
+; 64-CMP-DAG:  addiu $[[I5:[0-9]+]], $zero, 5
+; 64-CMP-DAG:  addiu $[[I4:[0-9]+]], $zero, 4
+; 64-CMP-DAG:  slti $[[R0:[0-9]+]], $4, 32767
+; FIXME: We can do better than this by adding/subtracting the result of slti
+;        to/from one of the constants.
+; 64-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I5]], $[[R0]]
+; 64-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I4]], $[[R0]]
+; 64-CMP-DAG:  or $2, $[[T0]], $[[T1]]
 
 define i64 @slti64_0(i64 %a) {
 entry:
   %cmp = icmp sgt i64 %a, 32766
-  %conv = select i1 %cmp, i64 3, i64 4
+  %conv = select i1 %cmp, i64 5, i64 4
   ret i64 %conv
 }
 
-; N64-LABEL: slti64_1:
-; N64: slt ${{[0-9]+}}
+; ALL-LABEL: slti64_1:
+
+; 32-CMOV-DAG:  slt $[[CC:[0-9]+]], $zero, $4
+; 32-CMOV-DAG:  addiu $[[I32766:[0-9]+]], $zero, 32767
+; 32-CMOV-DAG:  sltu $[[R1:[0-9]+]], $[[I32766]], $5
+; 32-CMOV-DAG:  movz $[[CC:[0-9]+]], $[[R1]], $4
+; 32-CMOV-DAG:  addiu $[[I5:[0-9]+]], $zero, 5
+; 32-CMOV-DAG:  addiu $[[I4:3]], $zero, 4
+; 32-CMOV-DAG:  movn $[[I4]], $[[I5]], $[[CC]]
+; 32-CMOV-DAG:  addiu $2, $zero, 0
+
+; 32-CMP-DAG:   slt $[[CC0:[0-9]+]], $zero, $4
+; 32-CMP-DAG:   addiu $[[I32766:[0-9]+]], $zero, 32767
+; 32-CMP-DAG:   sltu $[[CC1:[0-9]+]], $[[I32766]], $5
+; 32-CMP-DAG:   seleqz $[[CC2:[0-9]+]], $[[CC0]], $4
+; 32-CMP-DAG:   selnez $[[CC3:[0-9]+]], $[[CC1]], $4
+; 32-CMP:       or $[[CC:[0-9]+]], $[[CC3]], $[[CC2]]
+; 32-CMP-DAG:   addiu $[[I5:[0-9]+]], $zero, 5
+; 32-CMP-DAG:   addiu $[[I4:[0-9]+]], $zero, 4
+; 32-CMP-DAG:   seleqz $[[T0:[0-9]+]], $[[I4]], $[[CC]]
+; 32-CMP-DAG:   selnez $[[T1:[0-9]+]], $[[I5]], $[[CC]]
+; 32-CMP-DAG:   or $3, $[[T1]], $[[T0]]
+; 32-CMP-DAG:   addiu $2, $zero, 0
+
+; 64-CMOV-DAG: daddiu $[[I5:[0-9]+]], $zero, 5
+; 64-CMOV-DAG: daddiu $[[I4:2]], $zero, 4
+; 64-CMOV-DAG: daddiu $[[R1:[0-9]+]], $zero, 32767
+; 64-CMOV-DAG: slt $[[R0:[0-9]+]], $[[R1]], $4
+; 64-CMOV-DAG: movn $[[I4]], $[[I5]], $[[R0]]
+
+; 64-CMP-DAG:  daddiu $[[I5:[0-9]+]], $zero, 5
+; 64-CMP-DAG:  daddiu $[[I4:2]], $zero, 4
+; 64-CMP-DAG:  daddiu $[[R1:[0-9]+]], $zero, 32767
+; 64-CMP-DAG:  slt $[[R0:[0-9]+]], $[[R1]], $4
+; FIXME: We can do better than this by using selccz to choose between -0 and -2
+; 64-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I5]], $[[R0]]
+; 64-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I4]], $[[R0]]
+; 64-CMP-DAG:  or $2, $[[T0]], $[[T1]]
 
 define i64 @slti64_1(i64 %a) {
 entry:
   %cmp = icmp sgt i64 %a, 32767
-  %conv = select i1 %cmp, i64 3, i64 4
+  %conv = select i1 %cmp, i64 5, i64 4
   ret i64 %conv
 }
 
-; N64-LABEL: slti64_2:
-; N64: slti $[[R0:[0-9]+]], ${{[0-9]+}}, -32768
-; N64: movz ${{[0-9]+}}, ${{[0-9]+}}, $[[R0]]
+; ALL-LABEL: slti64_2:
+
+; FIXME: The 32-bit versions of this test are too complicated to reasonably
+;        match at the moment. They do show some missing optimizations though
+;        such as:
+;           (movz $a, $b, (neg $c)) -> (movn $a, $b, $c)
+
+; 64-CMOV-DAG:  addiu $[[I3:[0-9]+]], $zero, 3
+; 64-CMOV-DAG:  addiu $[[I4:2]], $zero, 4
+; 64-CMOV-DAG:  slti $[[R0:[0-9]+]], $4, -32768
+; 64-CMOV-DAG:  movz $[[I4]], $[[I3]], $[[R0]]
+
+; 64-CMP-DAG:  addiu $[[I3:[0-9]+]], $zero, 3
+; 64-CMP-DAG:  addiu $[[I4:[0-9]+]], $zero, 4
+; 64-CMP-DAG:  slti $[[R0:[0-9]+]], $4, -32768
+; FIXME: We can do better than this by adding/subtracting the result of slti
+;        to/from one of the constants.
+; 64-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I3]], $[[R0]]
+; 64-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I4]], $[[R0]]
+; 64-CMP-DAG:  or $2, $[[T0]], $[[T1]]
 
 define i64 @slti64_2(i64 %a) {
 entry:
   %cmp = icmp sgt i64 %a, -32769
   %conv = select i1 %cmp, i64 3, i64 4
   ret i64 %conv
 }
 
-; N64-LABEL: slti64_3:
-; N64: slt ${{[0-9]+}}
+; ALL-LABEL: slti64_3:
+
+; FIXME: The 32-bit versions of this test are too complicated to reasonably
+;        match at the moment. They do show some missing optimizations though
+;        such as:
+;           (movz $a, $b, (neg $c)) -> (movn $a, $b, $c)
+
+; 64-CMOV-DAG: daddiu $[[I5:[0-9]+]], $zero, 5
+; 64-CMOV-DAG: daddiu $[[I4:2]], $zero, 4
+; 64-CMOV-DAG: daddiu $[[R1:[0-9]+]], ${{[0-9]+}}, 32766
+; 64-CMOV-DAG: slt $[[R0:[0-9]+]], $[[R1]], $4
+; 64-CMOV-DAG: movn $[[I4]], $[[I5]], $[[R0]]
+
+; 64-CMP-DAG:  daddiu $[[I5:[0-9]+]], $zero, 5
+; 64-CMP-DAG:  daddiu $[[I4:2]], $zero, 4
+; 64-CMP-DAG:  daddiu $[[R1:[0-9]+]], ${{[0-9]+}}, 32766
+; 64-CMP-DAG:  slt $[[R0:[0-9]+]], $[[R1]], $4
+; FIXME: We can do better than this by using selccz to choose between -0 and -2
+; 64-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I5]], $[[R0]]
+; 64-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I4]], $[[R0]]
+; 64-CMP-DAG:  or $2, $[[T0]], $[[T1]]
 
 define i64 @slti64_3(i64 %a) {
 entry:
   %cmp = icmp sgt i64 %a, -32770
-  %conv = select i1 %cmp, i64 3, i64 4
+  %conv = select i1 %cmp, i64 5, i64 4
   ret i64 %conv
 }
 
 ; sltiu instructions.
 
-; O32-LABEL: sltiu0:
-; O32: sltiu $[[R0:[0-9]+]], ${{[0-9]+}}, 32767
-; O32: movz ${{[0-9]+}}, ${{[0-9]+}}, $[[R0]]
+; ALL-LABEL: sltiu0:
+
+; 32-CMOV-DAG: addiu $[[I3:[0-9]+]], $zero, 3
+; 32-CMOV-DAG: addiu $[[I5:2]], $zero, 5
+; 32-CMOV-DAG: sltiu $[[R0:[0-9]+]], $4, 32767
+; 32-CMOV-DAG: movz $[[I5]], $[[I3]], $[[R0]]
+
+; 32-CMP-DAG:  addiu $[[I3:[0-9]+]], $zero, 3
+; 32-CMP-DAG:  addiu $[[I5:[0-9]+]], $zero, 5
+; 32-CMP-DAG:  sltiu $[[R0:[0-9]+]], $4, 32767
+; FIXME: We can do better than this by using selccz to choose between +0 and +2
+; 32-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I3]], $[[R0]]
+; 32-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I5]], $[[R0]]
+; 32-CMP-DAG:  or $2, $[[T0]], $[[T1]]
+
+; 64-CMOV-DAG: addiu $[[I3:[0-9]+]], $zero, 3
+; 64-CMOV-DAG: addiu $[[I5:2]], $zero, 5
+; 64-CMOV-DAG: sltiu $[[R0:[0-9]+]], $4, 32767
+; 64-CMOV-DAG: movz $[[I5]], $[[I3]], $[[R0]]
+
+; 64-CMP-DAG:  addiu $[[I3:[0-9]+]], $zero, 3
+; 64-CMP-DAG:  addiu $[[I5:[0-9]+]], $zero, 5
+; 64-CMP-DAG:  sltiu $[[R0:[0-9]+]], $4, 32767
+; FIXME: We can do better than this by using selccz to choose between +0 and +2
+; 64-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I3]], $[[R0]]
+; 64-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I5]], $[[R0]]
+; 64-CMP-DAG:  or $2, $[[T0]], $[[T1]]
 
 define i32 @sltiu0(i32 %a) {
 entry:
   %cmp = icmp ugt i32 %a, 32766
   %cond = select i1 %cmp, i32 3, i32 5
   ret i32 %cond
 }
 
-; O32-LABEL: sltiu1:
-; O32: sltu ${{[0-9]+}}
+; ALL-LABEL: sltiu1:
+
+; 32-CMOV-DAG: addiu $[[I7:[0-9]+]], $zero, 7
+; 32-CMOV-DAG: addiu $[[I5:2]], $zero, 5
+; 32-CMOV-DAG: addiu $[[R1:[0-9]+]], $zero, 32767
+; 32-CMOV-DAG: sltu $[[R0:[0-9]+]], $[[R1]], $4
+; 32-CMOV-DAG: movn $[[I5]], $[[I7]], $[[R0]]
+
+; 32-CMP-DAG:  addiu $[[I7:[0-9]+]], $zero, 7
+; 32-CMP-DAG:  addiu $[[I5:[0-9]+]], $zero, 5
+; 32-CMP-DAG:  addiu $[[I32767:[0-9]+]], $zero, 32767
+; 32-CMP-DAG:  sltu $[[R0:[0-9]+]], $[[I32767]], $4
+; FIXME: We can do better than this by using selccz to choose between -0 and -2
+; 32-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I7]], $[[R0]]
+; 32-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I5]], $[[R0]]
+; 32-CMP-DAG:  or $2, $[[T0]], $[[T1]]
+
+; 64-CMOV-DAG: addiu $[[I7:[0-9]+]], $zero, 7
+; 64-CMOV-DAG: addiu $[[I5:2]], $zero, 5
+; 64-CMOV-DAG: addiu $[[R1:[0-9]+]], $zero, 32767
+; 64-CMOV-DAG: sltu $[[R0:[0-9]+]], $[[R1]], $4
+; 64-CMOV-DAG: movn $[[I5]], $[[I7]], $[[R0]]
+
+; 64-CMP-DAG:  addiu $[[I7:[0-9]+]], $zero, 7
+; 64-CMP-DAG:  addiu $[[I5:2]], $zero, 5
+; 64-CMP-DAG:  addiu $[[R1:[0-9]+]], $zero, 32767
+; 64-CMP-DAG:  sltu $[[R0:[0-9]+]], $[[R1]], $4
+; FIXME: We can do better than this by using selccz to choose between -0 and -2
+; 64-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I7]], $[[R0]]
+; 64-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I5]], $[[R0]]
+; 64-CMP-DAG:  or $2, $[[T0]], $[[T1]]
 
 define i32 @sltiu1(i32 %a) {
 entry:
   %cmp = icmp ugt i32 %a, 32767
-  %cond = select i1 %cmp, i32 3, i32 5
+  %cond = select i1 %cmp, i32 7, i32 5
   ret i32 %cond
 }
 
-; O32-LABEL: sltiu2:
-; O32: sltiu $[[R0:[0-9]+]], ${{[0-9]+}}, -32768
-; O32: movz ${{[0-9]+}}, ${{[0-9]+}}, $[[R0]]
+; ALL-LABEL: sltiu2:
+
+; 32-CMOV-DAG: addiu $[[I3:[0-9]+]], $zero, 3
+; 32-CMOV-DAG: addiu $[[I5:2]], $zero, 5
+; 32-CMOV-DAG: sltiu $[[R0:[0-9]+]], $4, -32768
+; 32-CMOV-DAG: movz $[[I5]], $[[I3]], $[[R0]]
+
+; 32-CMP-DAG:  addiu $[[I3:[0-9]+]], $zero, 3
+; 32-CMP-DAG:  addiu $[[I5:[0-9]+]], $zero, 5
+; 32-CMP-DAG:  sltiu $[[R0:[0-9]+]], $4, -32768
+; FIXME: We can do better than this by using selccz to choose between +0 and +2
+; 32-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I3]], $[[R0]]
+; 32-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I5]], $[[R0]]
+; 32-CMP-DAG:  or $2, $[[T0]], $[[T1]]
+
+; 64-CMOV-DAG: addiu $[[I3:[0-9]+]], $zero, 3
+; 64-CMOV-DAG: addiu $[[I5:2]], $zero, 5
+; 64-CMOV-DAG: sltiu $[[R0:[0-9]+]], $4, -32768
+; 64-CMOV-DAG: movz $[[I5]], $[[I3]], $[[R0]]
+
+; 64-CMP-DAG:  addiu $[[I3:[0-9]+]], $zero, 3
+; 64-CMP-DAG:  addiu $[[I5:[0-9]+]], $zero, 5
+; 64-CMP-DAG:  sltiu $[[R0:[0-9]+]], $4, -32768
+; FIXME: We can do better than this by using selccz to choose between +0 and +2
+; 64-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I3]], $[[R0]]
+; 64-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I5]], $[[R0]]
+; 64-CMP-DAG:  or $2, $[[T0]], $[[T1]]
 
 define i32 @sltiu2(i32 %a) {
 entry:
   %cmp = icmp ugt i32 %a, -32769
   %cond = select i1 %cmp, i32 3, i32 5
   ret i32 %cond
 }
 
-; O32-LABEL: sltiu3:
-; O32: sltu ${{[0-9]+}}
+; ALL-LABEL: sltiu3:
+
+; 32-CMOV-DAG: addiu $[[I3:[0-9]+]], $zero, 3
+; 32-CMOV-DAG: addiu $[[I5:2]], $zero, 5
+; 32-CMOV-DAG: lui $[[R1:[0-9]+]], 65535
+; 32-CMOV-DAG: ori $[[R1]], $[[R1]], 32766
+; 32-CMOV-DAG: sltu $[[R0:[0-9]+]], $[[R1]], $4
+; 32-CMOV-DAG: movn $[[I5]], $[[I3]], $[[R0]]
+
+; 32-CMP-DAG:  addiu $[[I3:[0-9]+]], $zero, 3
+; 32-CMP-DAG:  addiu $[[I5:[0-9]+]], $zero, 5
+; 32-CMP-DAG:  lui $[[IMM:[0-9]+]], 65535
+; 32-CMP-DAG:  ori $[[IMM]], $[[IMM]], 32766
+; 32-CMP-DAG:  sltu $[[R0:[0-9]+]], $[[I32767]], $4
+; FIXME: We can do better than this by using selccz to choose between -0 and -2
+; 32-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I3]], $[[R0]]
+; 32-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I5]], $[[R0]]
+; 32-CMP-DAG:  or $2, $[[T0]], $[[T1]]
+
+; 64-CMOV-DAG: addiu $[[I3:[0-9]+]], $zero, 3
+; 64-CMOV-DAG: addiu $[[I5:2]], $zero, 5
+; 64-CMOV-DAG: lui $[[R1:[0-9]+]], 65535
+; 64-CMOV-DAG: ori $[[R1]], $[[R1]], 32766
+; 64-CMOV-DAG: sltu $[[R0:[0-9]+]], $[[R1]], $4
+; 64-CMOV-DAG: movn $[[I5]], $[[I3]], $[[R0]]
+
+; 64-CMP-DAG:  addiu $[[I3:[0-9]+]], $zero, 3
+; 64-CMP-DAG:  addiu $[[I5:2]], $zero, 5
+; 64-CMP-DAG:  lui $[[IMM:[0-9]+]], 65535
+; 64-CMP-DAG:  ori $[[IMM]], $[[IMM]], 32766
+; 64-CMP-DAG:  sltu $[[R0:[0-9]+]], $[[IMM]], $4
+; FIXME: We can do better than this by using selccz to choose between -0 and -2
+; 64-CMP-DAG:  selnez $[[T0:[0-9]+]], $[[I3]], $[[R0]]
+; 64-CMP-DAG:  seleqz $[[T1:[0-9]+]], $[[I5]], $[[R0]]
+; 64-CMP-DAG:  or $2, $[[T0]], $[[T1]]
 
 define i32 @sltiu3(i32 %a) {
 entry:
   %cmp = icmp ugt i32 %a, -32770
   %cond = select i1 %cmp, i32 3, i32 5
   ret i32 %cond
 }
 
 ; Check if
 ;  (select (setxx a, N), x, x-1) or
 ;  (select (setxx a, N), x-1, x)
 ; doesn't generate conditional moves
 ; for constant operands whose difference is |1|
 
 define i32 @slti4(i32 %a) nounwind readnone {
   %1 = icmp slt i32 %a, 7
   %2 = select i1 %1, i32 4, i32 3
   ret i32 %2
 }
 
-; O32-LABEL: slti4:
-; O32-DAG: slti [[R1:\$[0-9]+]], $4, 7
-; O32-DAG: addiu [[R2:\$[0-9]+]], [[R1]], 3
-; O32-NOT: movn
-; O32:.size slti4
+; ALL-LABEL: slti4:
+
+; 32-CMOV-DAG: slti [[R1:\$[0-9]+]], $4, 7
+; 32-CMOV-DAG: addiu $2, [[R1]], 3
+; 32-CMOV-NOT: movn
+
+; 32-CMP-DAG:  slti [[R1:\$[0-9]+]], $4, 7
+; 32-CMP-DAG:  addiu $2, [[R1]], 3
+; 32-CMP-NOT:  seleqz
+; 32-CMP-NOT:  selnez
+
+; 64-CMOV-DAG: slti [[R1:\$[0-9]+]], $4, 7
+; 64-CMOV-DAG: addiu $2, [[R1]], 3
+; 64-CMOV-NOT: movn
+
+; 64-CMP-DAG:  slti [[R1:\$[0-9]+]], $4, 7
+; 64-CMP-DAG:  addiu $2, [[R1]], 3
+; 64-CMP-NOT:  seleqz
+; 64-CMP-NOT:  selnez
 
 define i32 @slti5(i32 %a) nounwind readnone {
   %1 = icmp slt i32 %a, 7
   %2 = select i1 %1, i32 -3, i32 -4
   ret i32 %2
 }
 
-; O32-LABEL: slti5:
-; O32-DAG: slti [[R1:\$[0-9]+]], $4, 7
-; O32-DAG: addiu [[R3:\$[0-9]+]], [[R2:\$[a-z0-9]+]], -4
-; O32-NOT: movn
-; O32:.size slti5
+; ALL-LABEL: slti5:
+
+; 32-CMOV-DAG: slti [[R1:\$[0-9]+]], $4, 7
+; 32-CMOV-DAG: addiu [[R3:\$[0-9]+]], [[R2:\$[a-z0-9]+]], -4
+; 32-CMOV-NOT: movn
+
+; 32-CMP-DAG:  slti [[R1:\$[0-9]+]], $4, 7
+; 32-CMP-DAG:  addiu [[R3:\$[0-9]+]], [[R2:\$[a-z0-9]+]], -4
+; 32-CMP-NOT:  seleqz
+; 32-CMP-NOT:  selnez
+
+; 64-CMOV-DAG: slti [[R1:\$[0-9]+]], $4, 7
+; 64-CMOV-DAG: addiu [[R3:\$[0-9]+]], [[R2:\$[a-z0-9]+]], -4
+; 64-CMOV-NOT: movn
+
+; 64-CMP-DAG:  slti [[R1:\$[0-9]+]], $4, 7
+; 64-CMP-DAG:  addiu [[R3:\$[0-9]+]], [[R2:\$[a-z0-9]+]], -4
+; 64-CMP-NOT:  seleqz
+; 64-CMP-NOT:  selnez
 
 define i32 @slti6(i32 %a) nounwind readnone {
   %1 = icmp slt i32 %a, 7
   %2 = select i1 %1, i32 3, i32 4
   ret i32 %2
 }
 
-; O32-LABEL: slti6:
-; O32-DAG: slti [[R1:\$[0-9]+]], $4, 7
-; O32-DAG: xori [[R1]], [[R1]], 1
-; O32-DAG: addiu [[R2:\$[0-9]+]], [[R1]], 3
-; O32-NOT: movn
-; O32:.size slti6
+; ALL-LABEL: slti6:
+
+; 32-CMOV-DAG: slti [[R1:\$[0-9]+]], $4, 7
+; 32-CMOV-DAG: xori [[R1]], [[R1]], 1
+; 32-CMOV-DAG: addiu [[R2:\$[0-9]+]], [[R1]], 3
+; 32-CMOV-NOT: movn
+
+; 32-CMP-DAG:  slti [[R1:\$[0-9]+]], $4, 7
+; 32-CMP-DAG:  xori [[R1]], [[R1]], 1
+; 32-CMP-DAG:  addiu [[R2:\$[0-9]+]], [[R1]], 3
+; 32-CMP-NOT:  seleqz
+; 32-CMP-NOT:  selnez
+
+; 64-CMOV-DAG: slti [[R1:\$[0-9]+]], $4, 7
+; 64-CMOV-DAG: xori [[R1]], [[R1]], 1
+; 64-CMOV-DAG: addiu [[R2:\$[0-9]+]], [[R1]], 3
+; 64-CMOV-NOT: movn
+
+; 64-CMP-DAG:  slti [[R1:\$[0-9]+]], $4, 7
+; 64-CMP-DAG:  xori [[R1]], [[R1]], 1
+; 64-CMP-DAG:  addiu [[R2:\$[0-9]+]], [[R1]], 3
+; 64-CMP-NOT:  seleqz
+; 64-CMP-NOT:  selnez

test/CodeGen/Mips/fcmp.ll

+; RUN: llc < %s -march=mipsel   -mcpu=mips32   | FileCheck %s -check-prefix=ALL -check-prefix=32-C
+; RUN: llc < %s -march=mipsel   -mcpu=mips32r2 | FileCheck %s -check-prefix=ALL -check-prefix=32-C
+; RUN: llc < %s -march=mipsel   -mcpu=mips32r6 | FileCheck %s -check-prefix=ALL -check-prefix=32-CMP
+; RUN: llc < %s -march=mips64el -mcpu=mips4    | FileCheck %s -check-prefix=ALL -check-prefix=64-C
+; RUN: llc < %s -march=mips64el -mcpu=mips64   | FileCheck %s -check-prefix=ALL -check-prefix=64-C
+; RUN: llc < %s -march=mips64el -mcpu=mips64r2 | FileCheck %s -check-prefix=ALL -check-prefix=64-C
+; RUN: llc < %s -march=mips64el -mcpu=mips64r6 | FileCheck %s -check-prefix=ALL -check-prefix=64-CMP
+
+define i32 @false_f32(float %a, float %b) nounwind {
+; ALL-LABEL: false_f32:
+; ALL:           addiu $2, $zero, 0
+
+  %1 = fcmp false float %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @oeq_f32(float %a, float %b) nounwind {
+; ALL-LABEL: oeq_f32:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.eq.s $f12, $f14
+; 32-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.eq.s $f12, $f13
+; 64-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.eq.s $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.eq.s $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp oeq float %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @ogt_f32(float %a, float %b) nounwind {
+; ALL-LABEL: ogt_f32:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.ule.s $f12, $f14
+; 32-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.ule.s $f12, $f13
+; 64-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.olt.s $[[T0:f[0-9]+]], $f14, $f12
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.olt.s $[[T0:f[0-9]+]], $f13, $f12
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp ogt float %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @oge_f32(float %a, float %b) nounwind {
+; ALL-LABEL: oge_f32:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.ult.s $f12, $f14
+; 32-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.ult.s $f12, $f13
+; 64-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.ole.s $[[T0:f[0-9]+]], $f14, $f12
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.ole.s $[[T0:f[0-9]+]], $f13, $f12
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp oge float %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @olt_f32(float %a, float %b) nounwind {
+; ALL-LABEL: olt_f32:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.olt.s $f12, $f14
+; 32-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.olt.s $f12, $f13
+; 64-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.olt.s $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.olt.s $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp olt float %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @ole_f32(float %a, float %b) nounwind {
+; ALL-LABEL: ole_f32:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.ole.s $f12, $f14
+; 32-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.ole.s $f12, $f13
+; 64-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.ole.s $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.ole.s $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp ole float %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @one_f32(float %a, float %b) nounwind {
+; ALL-LABEL: one_f32:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.ueq.s $f12, $f14
+; 32-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.ueq.s $f12, $f13
+; 64-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.ueq.s $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $[[T1:[0-9]+]], $[[T0]]
+; 32-CMP-DAG:    not $2, $[[T1]]
+
+; 64-CMP-DAG:    cmp.ueq.s $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $[[T1:[0-9]+]], $[[T0]]
+; 64-CMP-DAG:    not $2, $[[T1]]
+
+  %1 = fcmp one float %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @ord_f32(float %a, float %b) nounwind {
+; ALL-LABEL: ord_f32:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.un.s $f12, $f14
+; 32-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.un.s $f12, $f13
+; 64-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.un.s $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $[[T1:[0-9]+]], $[[T0]]
+; 32-CMP-DAG:    not $2, $[[T1]]
+
+; 64-CMP-DAG:    cmp.un.s $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $[[T1:[0-9]+]], $[[T0]]
+; 64-CMP-DAG:    not $2, $[[T1]]
+
+  %1 = fcmp ord float %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @ueq_f32(float %a, float %b) nounwind {
+; ALL-LABEL: ueq_f32:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.ueq.s $f12, $f14
+; 32-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.ueq.s $f12, $f13
+; 64-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.ueq.s $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.ueq.s $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp ueq float %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @ugt_f32(float %a, float %b) nounwind {
+; ALL-LABEL: ugt_f32:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.ole.s $f12, $f14
+; 32-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.ole.s $f12, $f13
+; 64-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.ult.s $[[T0:f[0-9]+]], $f14, $f12
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.ult.s $[[T0:f[0-9]+]], $f13, $f12
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp ugt float %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @uge_f32(float %a, float %b) nounwind {
+; ALL-LABEL: uge_f32:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.olt.s $f12, $f14
+; 32-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.olt.s $f12, $f13
+; 64-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.ule.s $[[T0:f[0-9]+]], $f14, $f12
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.ule.s $[[T0:f[0-9]+]], $f13, $f12
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp uge float %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @ult_f32(float %a, float %b) nounwind {
+; ALL-LABEL: ult_f32:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.ult.s $f12, $f14
+; 32-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.ult.s $f12, $f13
+; 64-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.ult.s $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.ult.s $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp ult float %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @ule_f32(float %a, float %b) nounwind {
+; ALL-LABEL: ule_f32:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.ule.s $f12, $f14
+; 32-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.ule.s $f12, $f13
+; 64-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.ule.s $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.ule.s $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp ule float %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @une_f32(float %a, float %b) nounwind {
+; ALL-LABEL: une_f32:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.eq.s $f12, $f14
+; 32-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.eq.s $f12, $f13
+; 64-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.eq.s $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $[[T1:[0-9]+]], $[[T0]]
+; 32-CMP-DAG:    not $2, $[[T1]]
+
+; 64-CMP-DAG:    cmp.eq.s $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $[[T1:[0-9]+]], $[[T0]]
+; 64-CMP-DAG:    not $2, $[[T1]]
+
+  %1 = fcmp une float %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @uno_f32(float %a, float %b) nounwind {
+; ALL-LABEL: uno_f32:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.un.s $f12, $f14
+; 32-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.un.s $f12, $f13
+; 64-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.un.s $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.un.s $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp uno float %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @true_f32(float %a, float %b) nounwind {
+; ALL-LABEL: true_f32:
+; ALL:           addiu $2, $zero, 1
+
+  %1 = fcmp true float %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @false_f64(double %a, double %b) nounwind {
+; ALL-LABEL: false_f64:
+; ALL:           addiu $2, $zero, 0
+
+  %1 = fcmp false double %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @oeq_f64(double %a, double %b) nounwind {
+; ALL-LABEL: oeq_f64:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.eq.d $f12, $f14
+; 32-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.eq.d $f12, $f13
+; 64-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.eq.d $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.eq.d $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp oeq double %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @ogt_f64(double %a, double %b) nounwind {
+; ALL-LABEL: ogt_f64:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.ule.d $f12, $f14
+; 32-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.ule.d $f12, $f13
+; 64-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.olt.d $[[T0:f[0-9]+]], $f14, $f12
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.olt.d $[[T0:f[0-9]+]], $f13, $f12
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp ogt double %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @oge_f64(double %a, double %b) nounwind {
+; ALL-LABEL: oge_f64:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.ult.d $f12, $f14
+; 32-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.ult.d $f12, $f13
+; 64-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.ole.d $[[T0:f[0-9]+]], $f14, $f12
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.ole.d $[[T0:f[0-9]+]], $f13, $f12
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp oge double %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @olt_f64(double %a, double %b) nounwind {
+; ALL-LABEL: olt_f64:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.olt.d $f12, $f14
+; 32-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.olt.d $f12, $f13
+; 64-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.olt.d $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.olt.d $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp olt double %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @ole_f64(double %a, double %b) nounwind {
+; ALL-LABEL: ole_f64:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.ole.d $f12, $f14
+; 32-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.ole.d $f12, $f13
+; 64-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.ole.d $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.ole.d $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp ole double %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @one_f64(double %a, double %b) nounwind {
+; ALL-LABEL: one_f64:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.ueq.d $f12, $f14
+; 32-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.ueq.d $f12, $f13
+; 64-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.ueq.d $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $[[T1:[0-9]+]], $[[T0]]
+; 32-CMP-DAG:    not $2, $[[T1]]
+
+; 64-CMP-DAG:    cmp.ueq.d $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $[[T1:[0-9]+]], $[[T0]]
+; 64-CMP-DAG:    not $2, $[[T1]]
+
+  %1 = fcmp one double %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @ord_f64(double %a, double %b) nounwind {
+; ALL-LABEL: ord_f64:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.un.d $f12, $f14
+; 32-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.un.d $f12, $f13
+; 64-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.un.d $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $[[T1:[0-9]+]], $[[T0]]
+; 32-CMP-DAG:    not $2, $[[T1]]
+
+; 64-CMP-DAG:    cmp.un.d $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $[[T1:[0-9]+]], $[[T0]]
+; 64-CMP-DAG:    not $2, $[[T1]]
+
+  %1 = fcmp ord double %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @ueq_f64(double %a, double %b) nounwind {
+; ALL-LABEL: ueq_f64:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.ueq.d $f12, $f14
+; 32-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.ueq.d $f12, $f13
+; 64-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.ueq.d $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.ueq.d $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp ueq double %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @ugt_f64(double %a, double %b) nounwind {
+; ALL-LABEL: ugt_f64:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.ole.d $f12, $f14
+; 32-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.ole.d $f12, $f13
+; 64-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.ult.d $[[T0:f[0-9]+]], $f14, $f12
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.ult.d $[[T0:f[0-9]+]], $f13, $f12
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp ugt double %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @uge_f64(double %a, double %b) nounwind {
+; ALL-LABEL: uge_f64:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.olt.d $f12, $f14
+; 32-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.olt.d $f12, $f13
+; 64-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.ule.d $[[T0:f[0-9]+]], $f14, $f12
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.ule.d $[[T0:f[0-9]+]], $f13, $f12
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp uge double %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @ult_f64(double %a, double %b) nounwind {
+; ALL-LABEL: ult_f64:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.ult.d $f12, $f14
+; 32-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.ult.d $f12, $f13
+; 64-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.ult.d $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.ult.d $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp ult double %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @ule_f64(double %a, double %b) nounwind {
+; ALL-LABEL: ule_f64:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.ule.d $f12, $f14
+; 32-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.ule.d $f12, $f13
+; 64-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.ule.d $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.ule.d $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp ule double %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @une_f64(double %a, double %b) nounwind {
+; ALL-LABEL: une_f64:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.eq.d $f12, $f14
+; 32-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.eq.d $f12, $f13
+; 64-C-DAG:      movf $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.eq.d $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $[[T1:[0-9]+]], $[[T0]]
+; 32-CMP-DAG:    not $2, $[[T1]]
+
+; 64-CMP-DAG:    cmp.eq.d $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $[[T1:[0-9]+]], $[[T0]]
+; 64-CMP-DAG:    not $2, $[[T1]]
+
+  %1 = fcmp une double %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @uno_f64(double %a, double %b) nounwind {
+; ALL-LABEL: uno_f64:
+
+; 32-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 32-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 32-C-DAG:      c.un.d $f12, $f14
+; 32-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 64-C-DAG:      addiu $[[T0:2]], $zero, 0
+; 64-C-DAG:      addiu $[[T1:[0-9]+]], $zero, 1
+; 64-C-DAG:      c.un.d $f12, $f13
+; 64-C-DAG:      movt $[[T0]], $1, $fcc0
+
+; 32-CMP-DAG:    cmp.un.d $[[T0:f[0-9]+]], $f12, $f14
+; 32-CMP-DAG:    mfc1 $2, $[[T0]]
+
+; 64-CMP-DAG:    cmp.un.d $[[T0:f[0-9]+]], $f12, $f13
+; 64-CMP-DAG:    mfc1 $2, $[[T0]]
+
+  %1 = fcmp uno double %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}
+
+define i32 @true_f64(double %a, double %b) nounwind {
+; ALL-LABEL: true_f64:
+; ALL:           addiu $2, $zero, 1
+
+  %1 = fcmp true double %a, %b
+  %2 = zext i1 %1 to i32
+  ret i32 %2
+}

test/CodeGen/Mips/mips64-f128.ll

 ; RUN: llc -mtriple=mips64el-unknown-unknown -mcpu=mips4 -soft-float -O1 \
-; RUN:     -disable-mips-delay-filler < %s | FileCheck %s
+; RUN:     -disable-mips-delay-filler < %s | FileCheck %s -check-prefix=ALL -check-prefix=C_CC_FMT
 ; RUN: llc -mtriple=mips64el-unknown-unknown -mcpu=mips64 -soft-float -O1 \
-; RUN:     -disable-mips-delay-filler < %s | FileCheck %s
+; RUN:     -disable-mips-delay-filler < %s | FileCheck %s -check-prefix=ALL -check-prefix=C_CC_FMT
+; RUN: llc -mtriple=mips64el-unknown-unknown -mcpu=mips64r2 -soft-float -O1 \
+; RUN:     -disable-mips-delay-filler < %s | FileCheck %s -check-prefix=ALL -check-prefix=C_CC_FMT
+; RUN: llc -mtriple=mips64el-unknown-unknown -mcpu=mips64r6 -soft-float -O1 \
+; RUN:     -disable-mips-delay-filler < %s | FileCheck %s -check-prefix=ALL -check-prefix=CMP_CC_FMT
 
 @gld0 = external global fp128
 @gld1 = external global fp128
 @gld2 = external global fp128
 @gf1 = external global float
 @gd1 = external global double
 
-; CHECK-LABEL: addLD:
-; CHECK: ld $25, %call16(__addtf3)
+; ALL-LABEL: addLD:
+; ALL: ld $25, %call16(__addtf3)
 
 define fp128 @addLD() {
 entry:
   %0 = load fp128* @gld0, align 16
   %1 = load fp128* @gld1, align 16
   %add = fadd fp128 %0, %1
   ret fp128 %add
 }
 
-; CHECK-LABEL: subLD:
-; CHECK: ld $25, %call16(__subtf3)
+; ALL-LABEL: subLD:
+; ALL: ld $25, %call16(__subtf3)
 
 define fp128 @subLD() {
 entry:
   %0 = load fp128* @gld0, align 16
   %1 = load fp128* @gld1, align 16
   %sub = fsub fp128 %0, %1
   ret fp128 %sub
 }
 
-; CHECK-LABEL: mulLD:
-; CHECK: ld $25, %call16(__multf3)
+; ALL-LABEL: mulLD:
+; ALL: ld $25, %call16(__multf3)
 
 define fp128 @mulLD() {
 entry:
   %0 = load fp128* @gld0, align 16
   %1 = load fp128* @gld1, align 16
   %mul = fmul fp128 %0, %1
   ret fp128 %mul
 }
 
-; CHECK-LABEL: divLD:
-; CHECK: ld $25, %call16(__divtf3)
+; ALL-LABEL: divLD:
+; ALL: ld $25, %call16(__divtf3)
 
 define fp128 @divLD() {
 entry:
   %0 = load fp128* @gld0, align 16
   %1 = load fp128* @gld1, align 16
   %div = fdiv fp128 %0, %1
   ret fp128 %div
 }
 
-; CHECK-LABEL: conv_LD_char:
-; CHECK: ld $25, %call16(__floatsitf)
+; ALL-LABEL: conv_LD_char:
+; ALL: ld $25, %call16(__floatsitf)
 
 define fp128 @conv_LD_char(i8 signext %a) {
 entry:
   %conv = sitofp i8 %a to fp128
   ret fp128 %conv
 }
 
-; CHECK-LABEL: conv_LD_short:
-; CHECK: ld $25, %call16(__floatsitf)
+; ALL-LABEL: conv_LD_short:
+; ALL: ld $25, %call16(__floatsitf)
 
 define fp128 @conv_LD_short(i16 signext %a) {
 entry:
   %conv = sitofp i16 %a to fp128
   ret fp128 %conv
 }
 
-; CHECK-LABEL: conv_LD_int:
-; CHECK: ld $25, %call16(__floatsitf)
+; ALL-LABEL: conv_LD_int:
+; ALL: ld $25, %call16(__floatsitf)
 
 define fp128 @conv_LD_int(i32 %a) {
 entry:
   %conv = sitofp i32 %a to fp128
   ret fp128 %conv
 }
 
-; CHECK-LABEL: conv_LD_LL:
-; CHECK: ld $25, %call16(__floatditf)
+; ALL-LABEL: conv_LD_LL:
+; ALL: ld $25, %call16(__floatditf)
 
 define fp128 @conv_LD_LL(i64 %a) {
 entry:
   %conv = sitofp i64 %a to fp128
   ret fp128 %conv
 }
 
-; CHECK-LABEL: conv_LD_UChar:
-; CHECK: ld $25, %call16(__floatunsitf)
+; ALL-LABEL: conv_LD_UChar:
+; ALL: ld $25, %call16(__floatunsitf)
 
 define fp128 @conv_LD_UChar(i8 zeroext %a) {
 entry:
   %conv = uitofp i8 %a to fp128
   ret fp128 %conv
 }
 
-; CHECK-LABEL: conv_LD_UShort:
-; CHECK: ld $25, %call16(__floatunsitf)
+; ALL-LABEL: conv_LD_UShort:
+; ALL: ld $25, %call16(__floatunsitf)
 
 define fp128 @conv_LD_UShort(i16 zeroext %a) {
 entry:
   %conv = uitofp i16 %a to fp128
   ret fp128 %conv
 }
 
-; CHECK-LABEL: conv_LD_UInt:
-; CHECK: ld $25, %call16(__floatunsitf)
+; ALL-LABEL: conv_LD_UInt:
+; ALL: ld $25, %call16(__floatunsitf)
 
 define fp128 @conv_LD_UInt(i32 %a) {
 entry:
   %conv = uitofp i32 %a to fp128
   ret fp128 %conv
 }
 
-; CHECK-LABEL: conv_LD_ULL:
-; CHECK: ld $25, %call16(__floatunditf)
+; ALL-LABEL: conv_LD_ULL:
+; ALL: ld $25, %call16(__floatunditf)
 
 define fp128 @conv_LD_ULL(i64 %a) {
 entry:
   %conv = uitofp i64 %a to fp128
   ret fp128 %conv
 }
 
-; CHECK-LABEL: conv_char_LD:
-; CHECK: ld $25, %call16(__fixtfsi)
+; ALL-LABEL: conv_char_LD:
+; ALL: ld $25, %call16(__fixtfsi)
 
 define signext i8 @conv_char_LD(fp128 %a) {
 entry:
   %conv = fptosi fp128 %a to i8
   ret i8 %conv
 }
 
-; CHECK-LABEL: conv_short_LD:
-; CHECK: ld $25, %call16(__fixtfsi)
+; ALL-LABEL: conv_short_LD:
+; ALL: ld $25, %call16(__fixtfsi)
 
 define signext i16 @conv_short_LD(fp128 %a) {
 entry:
   %conv = fptosi fp128 %a to i16
   ret i16 %conv
 }
 
-; CHECK-LABEL: conv_int_LD:
-; CHECK: ld $25, %call16(__fixtfsi)
+; ALL-LABEL: conv_int_LD:
+; ALL: ld $25, %call16(__fixtfsi)
 
 define i32 @conv_int_LD(fp128 %a) {
 entry:
   %conv = fptosi fp128 %a to i32
   ret i32 %conv
 }
 
-; CHECK-LABEL: conv_LL_LD:
-; CHECK: ld $25, %call16(__fixtfdi)
+; ALL-LABEL: conv_LL_LD:
+; ALL: ld $25, %call16(__fixtfdi)
 
 define i64 @conv_LL_LD(fp128 %a) {
 entry:
   %conv = fptosi fp128 %a to i64
   ret i64 %conv
 }
 
-; CHECK-LABEL: conv_UChar_LD:
-; CHECK: ld $25, %call16(__fixtfsi)
+; ALL-LABEL: conv_UChar_LD:
+; ALL: ld $25, %call16(__fixtfsi)
 
 define zeroext i8 @conv_UChar_LD(fp128 %a) {
 entry:
   %conv = fptoui fp128 %a to i8
   ret i8 %conv
 }
 
-; CHECK-LABEL: conv_UShort_LD:
-; CHECK: ld $25, %call16(__fixtfsi)
+; ALL-LABEL: conv_UShort_LD:
+; ALL: ld $25, %call16(__fixtfsi)
 
 define zeroext i16 @conv_UShort_LD(fp128 %a) {
 entry:
   %conv = fptoui fp128 %a to i16
   ret i16 %conv
 }
 
-; CHECK-LABEL: conv_UInt_LD:
-; CHECK: ld $25, %call16(__fixunstfsi)
+; ALL-LABEL: conv_UInt_LD:
+; ALL: ld $25, %call16(__fixunstfsi)
 
 define i32 @conv_UInt_LD(fp128 %a) {
 entry:
   %conv = fptoui fp128 %a to i32
   ret i32 %conv
 }
 
-; CHECK-LABEL: conv_ULL_LD:
-; CHECK: ld $25, %call16(__fixunstfdi)
+; ALL-LABEL: conv_ULL_LD:
+; ALL: ld $25, %call16(__fixunstfdi)
 
 define i64 @conv_ULL_LD(fp128 %a) {
 entry:
   %conv = fptoui fp128 %a to i64
   ret i64 %conv
 }
 
-; CHECK-LABEL: conv_LD_float:
-; CHECK: ld $25, %call16(__extendsftf2)
+; ALL-LABEL: conv_LD_float:
+; ALL: ld $25, %call16(__extendsftf2)
 
 define fp128 @conv_LD_float(float %a) {
 entry:
   %conv = fpext float %a to fp128
   ret fp128 %conv
 }
 
-; CHECK-LABEL: conv_LD_double:
-; CHECK: ld $25, %call16(__extenddftf2)
+; ALL-LABEL: conv_LD_double:
+; ALL: ld $25, %call16(__extenddftf2)
 
 define fp128 @conv_LD_double(double %a) {
 entry:
   %conv = fpext double %a to fp128
   ret fp128 %conv
 }
 
-; CHECK-LABEL: conv_float_LD:
-; CHECK: ld $25, %call16(__trunctfsf2)
+; ALL-LABEL: conv_float_LD:
+; ALL: ld $25, %call16(__trunctfsf2)
 
 define float @conv_float_LD(fp128 %a) {
 entry:
   %conv = fptrunc fp128 %a to float
   ret float %conv
 }
 
-; CHECK-LABEL: conv_double_LD:
-; CHECK: ld $25, %call16(__trunctfdf2)
+; ALL-LABEL: conv_double_LD:
+; ALL: ld $25, %call16(__trunctfdf2)
 
 define double @conv_double_LD(fp128 %a) {
 entry:
   %conv = fptrunc fp128 %a to double
   ret double %conv
 }
 
-; CHECK-LABEL:             libcall1_fabsl:
-; CHECK-DAG: ld      $[[R0:[0-9]+]], 8($[[R4:[0-9]+]])
-; CHECK-DAG: daddiu  $[[R1:[0-9]+]], $zero, 1
-; CHECK-DAG: dsll    $[[R2:[0-9]+]], $[[R1]], 63
-; CHECK-DAG: daddiu  $[[R3:[0-9]+]], $[[R2]], -1
-; CHECK-DAG: and     $4, $[[R0]], $[[R3]]
-; CHECK-DAG: ld      $2, 0($[[R4]])
+; ALL-LABEL:             libcall1_fabsl:
+; ALL-DAG: ld      $[[R0:[0-9]+]], 8($[[R4:[0-9]+]])
+; ALL-DAG: daddiu  $[[R1:[0-9]+]], $zero, 1
+; ALL-DAG: dsll    $[[R2:[0-9]+]], $[[R1]], 63
+; ALL-DAG: daddiu  $[[R3:[0-9]+]], $[[R2]], -1
+; ALL-DAG: and     $4, $[[R0]], $[[R3]]
+; ALL-DAG: ld      $2, 0($[[R4]])
 
 define fp128 @libcall1_fabsl() {
 entry:
   %0 = load fp128* @gld0, align 16
   %call = tail call fp128 @fabsl(fp128 %0) nounwind readnone
   ret fp128 %call
 }
 
 declare fp128 @fabsl(fp128) #1
 
-; CHECK-LABEL: libcall1_ceill:
-; CHECK: ld $25, %call16(ceill)
+; ALL-LABEL: libcall1_ceill:
+; ALL: ld $25, %call16(ceill)
 
 define fp128 @libcall1_ceill() {
 entry:
   %0 = load fp128* @gld0, align 16
   %call = tail call fp128 @ceill(fp128 %0) nounwind readnone
   ret fp128 %call
 }
 
 declare fp128 @ceill(fp128) #1
 
-; CHECK-LABEL: libcall1_sinl:
-; CHECK: ld $25, %call16(sinl)
+; ALL-LABEL: libcall1_sinl:
+; ALL: ld $25, %call16(sinl)
 
 define fp128 @libcall1_sinl() {
 entry:
   %0 = load fp128* @gld0, align 16
   %call = tail call fp128 @sinl(fp128 %0) nounwind
   ret fp128 %call
 }
 
 declare fp128 @sinl(fp128) #2
 
-; CHECK-LABEL: libcall1_cosl:
-; CHECK: ld $25, %call16(cosl)
+; ALL-LABEL: libcall1_cosl:
+; ALL: ld $25, %call16(cosl)
 
 define fp128 @libcall1_cosl() {
 entry:
   %0 = load fp128* @gld0, align 16
   %call = tail call fp128 @cosl(fp128 %0) nounwind
   ret fp128 %call
 }
 
 declare fp128 @cosl(fp128) #2
 
-; CHECK-LABEL: libcall1_expl:
-; CHECK: ld $25, %call16(expl)
+; ALL-LABEL: libcall1_expl:
+; ALL: ld $25, %call16(expl)
 
 define fp128 @libcall1_expl() {
 entry:
   %0 = load fp128* @gld0, align 16
   %call = tail call fp128 @expl(fp128 %0) nounwind
   ret fp128 %call
 }
 
 declare fp128 @expl(fp128) #2
 
-; CHECK-LABEL: libcall1_exp2l:
-; CHECK: ld $25, %call16(exp2l)
+; ALL-LABEL: libcall1_exp2l:
+; ALL: ld $25, %call16(exp2l)
 
 define fp128 @libcall1_exp2l() {
 entry:
   %0 = load fp128* @gld0, align 16
   %call = tail call fp128 @exp2l(fp128 %0) nounwind
   ret fp128 %call
 }
 
 declare fp128 @exp2l(fp128) #2
 
-; CHECK-LABEL: libcall1_logl:
-; CHECK: ld $25, %call16(logl)
+; ALL-LABEL: libcall1_logl:
+; ALL: ld $25, %call16(logl)
 
 define fp128 @libcall1_logl() {
 entry:
   %0 = load fp128* @gld0, align 16
   %call = tail call fp128 @logl(fp128 %0) nounwind
   ret fp128 %call
 }
 
 declare fp128 @logl(fp128) #2
 
-; CHECK-LABEL: libcall1_log2l:
-; CHECK: ld $25, %call16(log2l)
+; ALL-LABEL: libcall1_log2l:
+; ALL: ld $25, %call16(log2l)
 
 define fp128 @libcall1_log2l() {
 entry:
   %0 = load fp128* @gld0, align 16
   %call = tail call fp128 @log2l(fp128 %0) nounwind
   ret fp128 %call
 }
 
 declare fp128 @log2l(fp128) #2
 
-; CHECK-LABEL: libcall1_log10l:
-; CHECK: ld $25, %call16(log10l)
+; ALL-LABEL: libcall1_log10l:
+; ALL: ld $25, %call16(log10l)
 
 define fp128 @libcall1_log10l() {
 entry:
   %0 = load fp128* @gld0, align 16
   %call = tail call fp128 @log10l(fp128 %0) nounwind
   ret fp128 %call
 }
 
 declare fp128 @log10l(fp128) #2
 
-; CHECK-LABEL: libcall1_nearbyintl:
-; CHECK: ld $25, %call16(nearbyintl)
+; ALL-LABEL: libcall1_nearbyintl:
+; ALL: ld $25, %call16(nearbyintl)
 
 define fp128 @libcall1_nearbyintl() {
 entry:
   %0 = load fp128* @gld0, align 16
   %call = tail call fp128 @nearbyintl(fp128 %0) nounwind readnone
   ret fp128 %call
 }
 
 declare fp128 @nearbyintl(fp128) #1
 
-; CHECK-LABEL: libcall1_floorl:
-; CHECK: ld $25, %call16(floorl)
+; ALL-LABEL: libcall1_floorl:
+; ALL: ld $25, %call16(floorl)
 
 define fp128 @libcall1_floorl() {
 entry:
   %0 = load fp128* @gld0, align 16
   %call = tail call fp128 @floorl(fp128 %0) nounwind readnone
   ret fp128 %call
 }
 
 declare fp128 @floorl(fp128) #1
 
-; CHECK-LABEL: libcall1_sqrtl:
-; CHECK: ld $25, %call16(sqrtl)
+; ALL-LABEL: libcall1_sqrtl:
+; ALL: ld $25, %call16(sqrtl)
 
 define fp128 @libcall1_sqrtl() {
 entry:
   %0 = load fp128* @gld0, align 16
   %call = tail call fp128 @sqrtl(fp128 %0) nounwind
   ret fp128 %call
 }
 
 declare fp128 @sqrtl(fp128) #2
 
-; CHECK-LABEL: libcall1_rintl:
-; CHECK: ld $25, %call16(rintl)
+; ALL-LABEL: libcall1_rintl:
+; ALL: ld $25, %call16(rintl)
 
 define fp128 @libcall1_rintl() {
 entry:
   %0 = load fp128* @gld0, align 16
   %call = tail call fp128 @rintl(fp128 %0) nounwind readnone
   ret fp128 %call
 }
 
 declare fp128 @rintl(fp128) #1
 
-; CHECK-LABEL: libcall_powil:
-; CHECK: ld $25, %call16(__powitf2)
+; ALL-LABEL: libcall_powil:
+; ALL: ld $25, %call16(__powitf2)
 
 define fp128 @libcall_powil(fp128 %a, i32 %b) {
 entry:
   %0 = tail call fp128 @llvm.powi.f128(fp128 %a, i32 %b)
   ret fp128 %0
 }
 
 declare fp128 @llvm.powi.f128(fp128, i32) #3
 
-; CHECK-LABEL:     libcall2_copysignl:
-; CHECK-DAG: daddiu $[[R2:[0-9]+]], $zero, 1
-; CHECK-DAG: dsll   $[[R3:[0-9]+]], $[[R2]], 63
-; CHECK-DAG: ld     $[[R0:[0-9]+]], %got_disp(gld1)
-; CHECK-DAG: ld     $[[R1:[0-9]+]], 8($[[R0]])
-; CHECK-DAG: and    $[[R4:[0-9]+]], $[[R1]], $[[R3]]
-; CHECK-DAG: ld     $[[R5:[0-9]+]], %got_disp(gld0)
-; CHECK-DAG: ld     $[[R6:[0-9]+]], 8($[[R5]])
-; CHECK-DAG: daddiu $[[R7:[0-9]+]], $[[R3]], -1
-; CHECK-DAG: and    $[[R8:[0-9]+]], $[[R6]], $[[R7]]
-; CHECK-DAG: or     $4, $[[R8]], $[[R4]]
-; CHECK-DAG: ld     $2, 0($[[R5]])
+; ALL-LABEL:     libcall2_copysignl:
+; ALL-DAG: daddiu $[[R2:[0-9]+]], $zero, 1
+; ALL-DAG: dsll   $[[R3:[0-9]+]], $[[R2]], 63
+; ALL-DAG: ld     $[[R0:[0-9]+]], %got_disp(gld1)
+; ALL-DAG: ld     $[[R1:[0-9]+]], 8($[[R0]])
+; ALL-DAG: and    $[[R4:[0-9]+]], $[[R1]], $[[R3]]
+; ALL-DAG: ld     $[[R5:[0-9]+]], %got_disp(gld0)
+; ALL-DAG: ld     $[[R6:[0-9]+]], 8($[[R5]])
+; ALL-DAG: daddiu $[[R7:[0-9]+]], $[[R3]], -1
+; ALL-DAG: and    $[[R8:[0-9]+]], $[[R6]], $[[R7]]
+; ALL-DAG: or     $4, $[[R8]], $[[R4]]
+; ALL-DAG: ld     $2, 0($[[R5]])
 
 define fp128 @libcall2_copysignl() {
 entry:
   %0 = load fp128* @gld0, align 16
   %1 = load fp128* @gld1, align 16
   %call = tail call fp128 @copysignl(fp128 %0, fp128 %1) nounwind readnone
   ret fp128 %call
 }
 
 declare fp128 @copysignl(fp128, fp128) #1
 
-; CHECK-LABEL: libcall2_powl:
-; CHECK: ld $25, %call16(powl)
+; ALL-LABEL: libcall2_powl:
+; ALL: ld $25, %call16(powl)
 
 define fp128 @libcall2_powl() {
 entry:
   %0 = load fp128* @gld0, align 16
   %1 = load fp128* @gld1, align 16
   %call = tail call fp128 @powl(fp128 %0, fp128 %1) nounwind
   ret fp128 %call
 }
 
 declare fp128 @powl(fp128, fp128) #2
 
-; CHECK-LABEL: libcall2_fmodl:
-; CHECK: ld $25, %call16(fmodl)
+; ALL-LABEL: libcall2_fmodl:
+; ALL: ld $25, %call16(fmodl)
 
 define fp128 @libcall2_fmodl() {
 entry:
   %0 = load fp128* @gld0, align 16
   %1 = load fp128* @gld1, align 16
   %call = tail call fp128 @fmodl(fp128 %0, fp128 %1) nounwind
   ret fp128 %call
 }
 
 declare fp128 @fmodl(fp128, fp128) #2
 
-; CHECK-LABEL: libcall3_fmal:
-; CHECK: ld $25, %call16(fmal)
+; ALL-LABEL: libcall3_fmal:
+; ALL: ld $25, %call16(fmal)
 
 define fp128 @libcall3_fmal() {
 entry:
   %0 = load fp128* @gld0, align 16
   %1 = load fp128* @gld2, align 16
   %2 = load fp128* @gld1, align 16
   %3 = tail call fp128 @llvm.fma.f128(fp128 %0, fp128 %2, fp128 %1)
   ret fp128 %3
 }
 
 declare fp128 @llvm.fma.f128(fp128, fp128, fp128) #4
 
-; CHECK-LABEL: cmp_lt:
-; CHECK: ld $25, %call16(__lttf2)
+; ALL-LABEL: cmp_lt:
+; ALL: ld $25, %call16(__lttf2)
 
 define i32 @cmp_lt(fp128 %a, fp128 %b) {
 entry:
   %cmp = fcmp olt fp128 %a, %b
   %conv = zext i1 %cmp to i32
   ret i32 %conv
 }
 
-; CHECK-LABEL: cmp_le:
-; CHECK: ld $25, %call16(__letf2)
+; ALL-LABEL: cmp_le:
+; ALL: ld $25, %call16(__letf2)
 
 define i32 @cmp_le(fp128 %a, fp128 %b) {
 entry:
   %cmp = fcmp ole fp128 %a, %b
   %conv = zext i1 %cmp to i32
   ret i32 %conv
 }
 
-; CHECK-LABEL: cmp_gt:
-; CHECK: ld $25, %call16(__gttf2)
+; ALL-LABEL: cmp_gt:
+; ALL: ld $25, %call16(__gttf2)
 
 define i32 @cmp_gt(fp128 %a, fp128 %b) {
 entry:
   %cmp = fcmp ogt fp128 %a, %b
   %conv = zext i1 %cmp to i32
   ret i32 %conv
 }
 
-; CHECK-LABEL: cmp_ge:
-; CHECK: ld $25, %call16(__getf2)
+; ALL-LABEL: cmp_ge:
+; ALL: ld $25, %call16(__getf2)
 
 define i32 @cmp_ge(fp128 %a, fp128 %b) {
 entry:
   %cmp = fcmp oge fp128 %a, %b
   %conv = zext i1 %cmp to i32
   ret i32 %conv
 }
 
-; CHECK-LABEL: cmp_eq:
-; CHECK: ld $25, %call16(__eqtf2)
+; ALL-LABEL: cmp_eq:
+; ALL: ld $25, %call16(__eqtf2)
 
 define i32 @cmp_eq(fp128 %a, fp128 %b) {
 entry:
   %cmp = fcmp oeq fp128 %a, %b
   %conv = zext i1 %cmp to i32
   ret i32 %conv
 }
 
-; CHECK-LABEL: cmp_ne:
-; CHECK: ld $25, %call16(__netf2)
+; ALL-LABEL: cmp_ne:
+; ALL: ld $25, %call16(__netf2)
 
 define i32 @cmp_ne(fp128 %a, fp128 %b) {
 entry:
   %cmp = fcmp une fp128 %a, %b
   %conv = zext i1 %cmp to i32
   ret i32 %conv
 }
 
-; CHECK-LABEL: load_LD_LD:
-; CHECK: ld $[[R0:[0-9]+]], %got_disp(gld1)
-; CHECK: ld $2, 0($[[R0]])
-; CHECK: ld $4, 8($[[R0]])
+; ALL-LABEL: load_LD_LD:
+; ALL: ld $[[R0:[0-9]+]], %got_disp(gld1)
+; ALL: ld $2, 0($[[R0]])
+; ALL: ld $4, 8($[[R0]])
 
 define fp128 @load_LD_LD() {
 entry:
   %0 = load fp128* @gld1, align 16
   ret fp128 %0
 }
 
-; CHECK-LABEL: load_LD_float:
-; CHECK: ld   $[[R0:[0-9]+]], %got_disp(gf1)
-; CHECK: lw   $4, 0($[[R0]])
-; CHECK: ld   $25, %call16(__extendsftf2)
-; CHECK: jalr $25
+; ALL-LABEL: load_LD_float:
+; ALL: ld   $[[R0:[0-9]+]], %got_disp(gf1)
+; ALL: lw   $4, 0($[[R0]])
+; ALL: ld   $25, %call16(__extendsftf2)
+; ALL: jalr $25
 
 define fp128 @load_LD_float() {
 entry:
   %0 = load float* @gf1, align 4
   %conv = fpext float %0 to fp128
   ret fp128 %conv
 }
 
-; CHECK-LABEL: load_LD_double:
-; CHECK: ld   $[[R0:[0-9]+]], %got_disp(gd1)
-; CHECK: ld   $4, 0($[[R0]])
-; CHECK: ld   $25, %call16(__extenddftf2)
-; CHECK: jalr $25
+; ALL-LABEL: load_LD_double:
+; ALL: ld   $[[R0:[0-9]+]], %got_disp(gd1)
+; ALL: ld   $4, 0($[[R0]])
+; ALL: ld   $25, %call16(__extenddftf2)
+; ALL: jalr $25
 
 define fp128 @load_LD_double() {
 entry:
   %0 = load double* @gd1, align 8
   %conv = fpext double %0 to fp128
   ret fp128 %conv
 }
 
-; CHECK-LABEL: store_LD_LD:
-; CHECK: ld $[[R0:[0-9]+]], %got_disp(gld1)
-; CHECK: ld $[[R1:[0-9]+]], 0($[[R0]])
-; CHECK: ld $[[R2:[0-9]+]], 8($[[R0]])
-; CHECK: ld $[[R3:[0-9]+]], %got_disp(gld0)
-; CHECK: sd $[[R2]], 8($[[R3]])
-; CHECK: sd $[[R1]], 0($[[R3]])
+; ALL-LABEL: store_LD_LD:
+; ALL: ld $[[R0:[0-9]+]], %got_disp(gld1)
+; ALL: ld $[[R1:[0-9]+]], 0($[[R0]])
+; ALL: ld $[[R2:[0-9]+]], 8($[[R0]])
+; ALL: ld $[[R3:[0-9]+]], %got_disp(gld0)
+; ALL: sd $[[R2]], 8($[[R3]])
+; ALL: sd $[[R1]], 0($[[R3]])
 
 define void @store_LD_LD() {
 entry:
   %0 = load fp128* @gld1, align 16
   store fp128 %0, fp128* @gld0, align 16
   ret void
 }
 
-; CHECK-LABEL: store_LD_float:
-; CHECK: ld   $[[R0:[0-9]+]], %got_disp(gld1)
-; CHECK: ld   $4, 0($[[R0]])
-; CHECK: ld   $5, 8($[[R0]])
-; CHECK: ld   $25, %call16(__trunctfsf2)
-; CHECK: jalr $25
-; CHECK: ld   $[[R1:[0-9]+]], %got_disp(gf1)
-; CHECK: sw   $2, 0($[[R1]])
+; ALL-LABEL: store_LD_float:
+; ALL: ld   $[[R0:[0-9]+]], %got_disp(gld1)
+; ALL: ld   $4, 0($[[R0]])
+; ALL: ld   $5, 8($[[R0]])
+; ALL: ld   $25, %call16(__trunctfsf2)
+; ALL: jalr $25
+; ALL: ld   $[[R1:[0-9]+]], %got_disp(gf1)
+; ALL: sw   $2, 0($[[R1]])
 
 define void @store_LD_float() {
 entry:
   %0 = load fp128* @gld1, align 16
   %conv = fptrunc fp128 %0 to float
   store float %conv, float* @gf1, align 4
   ret void
 }
 
-; CHECK-LABEL: store_LD_double:
-; CHECK: ld   $[[R0:[0-9]+]], %got_disp(gld1)
-; CHECK: ld   $4, 0($[[R0]])
-; CHECK: ld   $5, 8($[[R0]])
-; CHECK: ld   $25, %call16(__trunctfdf2)
-; CHECK: jalr $25
-; CHECK: ld   $[[R1:[0-9]+]], %got_disp(gd1)
-; CHECK: sd   $2, 0($[[R1]])
+; ALL-LABEL: store_LD_double:
+; ALL: ld   $[[R0:[0-9]+]], %got_disp(gld1)
+; ALL: ld   $4, 0($[[R0]])
+; ALL: ld   $5, 8($[[R0]])
+; ALL: ld   $25, %call16(__trunctfdf2)
+; ALL: jalr $25
+; ALL: ld   $[[R1:[0-9]+]], %got_disp(gd1)
+; ALL: sd   $2, 0($[[R1]])
 
 define void @store_LD_double() {
 entry:
   %0 = load fp128* @gld1, align 16
   %conv = fptrunc fp128 %0 to double
   store double %conv, double* @gd1, align 8
   ret void
 }
 
-; CHECK-LABEL: select_LD:
-; CHECK: movn $8, $6, $4
-; CHECK: movn $9, $7, $4
-; CHECK: move $2, $8
-; CHECK: move $4, $9
+; ALL-LABEL: select_LD:
+; C_CC_FMT:      movn $8, $6, $4
+; C_CC_FMT:      movn $9, $7, $4
+; C_CC_FMT:      move $2, $8
+; C_CC_FMT:      move $4, $9
+
+; FIXME: This sll works around an implementation detail in the code generator
+;        (setcc's result is i32 so bits 32-63 are undefined). It's not really
+;        needed.
+; CMP_CC_FMT-DAG: sll $[[CC:[0-9]+]], $4, 0
+; CMP_CC_FMT-DAG: selnez $[[EQ1:[0-9]+]], $8, $[[CC]]
+; CMP_CC_FMT-DAG: seleqz $[[NE1:[0-9]+]], $6, $[[CC]]
+; CMP_CC_FMT-DAG: or $2, $[[EQ1]], $[[NE1]]
+; CMP_CC_FMT-DAG: selnez $[[EQ2:[0-9]+]], $9, $[[CC]]
+; CMP_CC_FMT-DAG: seleqz $[[NE2:[0-9]+]], $7, $[[CC]]
+; CMP_CC_FMT-DAG: or $4, $[[EQ2]], $[[NE2]]
 
 define fp128 @select_LD(i32 %a, i64, fp128 %b, fp128 %c) {
 entry:
   %tobool = icmp ne i32 %a, 0
   %cond = select i1 %tobool, fp128 %b, fp128 %c
   ret fp128 %cond
 }
 
-; CHECK-LABEL: selectCC_LD:
-; CHECK: move $[[R0:[0-9]+]], $11
-; CHECK: move $[[R1:[0-9]+]], $10
-; CHECK: move $[[R2:[0-9]+]], $9
-; CHECK: move $[[R3:[0-9]+]], $8
-; CHECK: ld   $25, %call16(__gttf2)($gp)
-; CHECK: jalr $25
-; CHECK: slti $1, $2, 1
-; CHECK: movz $[[R1]], $[[R3]], $1
-; CHECK: movz $[[R0]], $[[R2]], $1
-; CHECK: move $2, $[[R1]]
-; CHECK: move $4, $[[R0]]
+; ALL-LABEL: selectCC_LD:
+; ALL:           move $[[R0:[0-9]+]], $11
+; ALL:           move $[[R1:[0-9]+]], $10
+; ALL:           move $[[R2:[0-9]+]], $9
+; ALL:           move $[[R3:[0-9]+]], $8
+; ALL:           ld   $25, %call16(__gttf2)($gp)
+; ALL:           jalr $25
+
+; C_CC_FMT:      slti $[[CC:[0-9]+]], $2, 1
+; C_CC_FMT:      movz $[[R1]], $[[R3]], $[[CC]]
+; C_CC_FMT:      movz $[[R0]], $[[R2]], $[[CC]]
+; C_CC_FMT:      move $2, $[[R1]]
+; C_CC_FMT:      move $4, $[[R0]]
+
+; CMP_CC_FMT:    slt $[[CC:[0-9]+]], $zero, $2
+; CMP_CC_FMT:    seleqz $[[EQ1:[0-9]+]], $[[R1]], $[[CC]]
+; CMP_CC_FMT:    selnez $[[NE1:[0-9]+]], $[[R3]], $[[CC]]
+; CMP_CC_FMT:    or $2, $[[NE1]], $[[EQ1]]
+; CMP_CC_FMT:    seleqz $[[EQ2:[0-9]+]], $[[R0]], $[[CC]]
+; CMP_CC_FMT:    selnez $[[NE2:[0-9]+]], $[[R2]], $[[CC]]
+; CMP_CC_FMT:    or $4, $[[NE2]], $[[EQ2]]
 
 define fp128 @selectCC_LD(fp128 %a, fp128 %b, fp128 %c, fp128 %d) {
 entry:
   %cmp = fcmp ogt fp128 %a, %b
   %cond = select i1 %cmp, fp128 %c, fp128 %d
   ret fp128 %cond
 }

test/CodeGen/Mips/select.ll

-; RUN: llc  < %s -march=mipsel | FileCheck %s -check-prefix=CHECK
+; RUN: llc < %s -march=mipsel   -mcpu=mips32   | FileCheck %s -check-prefix=ALL -check-prefix=32
+; RUN: llc < %s -march=mipsel   -mcpu=mips32r2 | FileCheck %s -check-prefix=ALL -check-prefix=32R2
+; RUN: llc < %s -march=mipsel   -mcpu=mips32r6 | FileCheck %s -check-prefix=ALL -check-prefix=32R6
+; RUN: llc < %s -march=mips64el -mcpu=mips64   | FileCheck %s -check-prefix=ALL -check-prefix=64
+; RUN: llc < %s -march=mips64el -mcpu=mips64r2 | FileCheck %s -check-prefix=ALL -check-prefix=64R2
+; RUN: llc < %s -march=mips64el -mcpu=mips64r6 | FileCheck %s -check-prefix=ALL -check-prefix=64R6
 
 @d2 = external global double
 @d3 = external global double
 
-define i32 @sel1(i32 %s, i32 %f0, i32 %f1) nounwind readnone {
+define i32 @i32_icmp_ne_i32_val(i32 %s, i32 %f0, i32 %f1) nounwind readnone {
 entry:
-; CHECK: movn
+; ALL-LABEL: i32_icmp_ne_i32_val:
+
+; 32:            movn $5, $6, $4
+; 32:            move $2, $5
+
+; 32R2:          movn $5, $6, $4
+; 32R2:          move $2, $5
+
+; 32R6-DAG:      selnez $[[T0:[0-9]+]], $5, $4
+; 32R6-DAG:      seleqz $[[T1:[0-9]+]], $6, $4
+; 32R6:          or $2, $[[T0]], $[[T1]]
+
+; 64:            movn $5, $6, $4
+; 64:            move $2, $5
+
+; 64R2:          movn $5, $6, $4
+; 64R2:          move $2, $5
+
+; 64R6-DAG:      selnez $[[T0:[0-9]+]], $5, $4
+; 64R6-DAG:      seleqz $[[T1:[0-9]+]], $6, $4
+; 64R6:          or $2, $[[T0]], $[[T1]]
+
   %tobool = icmp ne i32 %s, 0
   %cond = select i1 %tobool, i32 %f1, i32 %f0
   ret i32 %cond
 }
 
-define float @sel2(i32 %s, float %f0, float %f1) nounwind readnone {
+define i64 @i32_icmp_ne_i64_val(i32 %s, i64 %f0, i64 %f1) nounwind readnone {
+entry:
+; ALL-LABEL: i32_icmp_ne_i64_val:
+
+; 32-DAG:        lw $[[F1:[0-9]+]], 16($sp)
+; 32-DAG:        movn $6, $[[F1]], $4
+; 32-DAG:        lw $[[F1H:[0-9]+]], 20($sp)
+; 32:            movn $7, $[[F1H]], $4
+; 32:            move $2, $6
+; 32:            move $3, $7
+
+; 32R2-DAG:      lw $[[F1:[0-9]+]], 16($sp)
+; 32R2-DAG:      movn $6, $[[F1]], $4
+; 32R2-DAG:      lw $[[F1H:[0-9]+]], 20($sp)
+; 32R2:          movn $7, $[[F1H]], $4
+; 32R2:          move $2, $6
+; 32R2:          move $3, $7
+
+; 32R6-DAG:      lw $[[F1:[0-9]+]], 16($sp)
+; 32R6-DAG:      selnez $[[T0:[0-9]+]], $6, $4
+; 32R6-DAG:      seleqz $[[T1:[0-9]+]], $[[F1]], $4
+; 32R6:          or $2, $[[T0]], $[[T1]]
+; 32R6-DAG:      lw $[[F1H:[0-9]+]], 20($sp)
+; 32R6-DAG:      selnez $[[T0:[0-9]+]], $7, $4
+; 32R6-DAG:      seleqz $[[T1:[0-9]+]], $[[F1H]], $4
+; 32R6:          or $3, $[[T0]], $[[T1]]
+
+; 64:            movn $5, $6, $4
+; 64:            move $2, $5
+
+; 64R2:          movn $5, $6, $4
+; 64R2:          move $2, $5
+
+; FIXME: This sll works around an implementation detail in the code generator
+;        (setcc's result is i32 so bits 32-63 are undefined). It's not really
+;        needed.
+; 64R6-DAG:      sll $[[CC:[0-9]+]], $4, 0
+; 64R6-DAG:      selnez $[[T0:[0-9]+]], $5, $[[CC]]
+; 64R6-DAG:      seleqz $[[T1:[0-9]+]], $6, $[[CC]]
+; 64R6:          or $2, $[[T0]], $[[T1]]
+
+  %tobool = icmp ne i32 %s, 0
+  %cond = select i1 %tobool, i64 %f1, i64 %f0
+  ret i64 %cond
+}
+
+define i64 @i64_icmp_ne_i64_val(i64 %s, i64 %f0, i64 %f1) nounwind readnone {
 entry:
-; CHECK: movn.s
+; ALL-LABEL: i64_icmp_ne_i64_val:
+
+; 32-DAG:        or $[[CC:[0-9]+]], $4
+; 32-DAG:        lw $[[F1:[0-9]+]], 16($sp)
+; 32-DAG:        movn $6, $[[F1]], $[[CC]]
+; 32-DAG:        lw $[[F1H:[0-9]+]], 20($sp)
+; 32:            movn $7, $[[F1H]], $[[CC]]
+; 32:            move $2, $6
+; 32:            move $3, $7
+
+; 32R2-DAG:      or $[[CC:[0-9]+]], $4
+; 32R2-DAG:      lw $[[F1:[0-9]+]], 16($sp)
+; 32R2-DAG:      movn $6, $[[F1]], $[[CC]]
+; 32R2-DAG:      lw $[[F1H:[0-9]+]], 20($sp)
+; 32R2:          movn $7, $[[F1H]], $[[CC]]
+; 32R2:          move $2, $6
+; 32R2:          move $3, $7
+
+; 32R6-DAG:      lw $[[F1:[0-9]+]], 16($sp)
+; 32R6-DAG:      or $[[T2:[0-9]+]], $4, $5
+; 32R6-DAG:      selnez $[[T0:[0-9]+]], $6, $[[T2]]
+; 32R6-DAG:      seleqz $[[T1:[0-9]+]], $[[F1]], $[[T2]]
+; 32R6:          or $2, $[[T0]], $[[T1]]
+; 32R6-DAG:      lw $[[F1H:[0-9]+]], 20($sp)
+; 32R6-DAG:      selnez $[[T0:[0-9]+]], $7, $[[T2]]
+; 32R6-DAG:      seleqz $[[T1:[0-9]+]], $[[F1H]], $[[T2]]
+; 32R6:          or $3, $[[T0]], $[[T1]]
+
+; 64:            movn $5, $6, $4
+; 64:            move $2, $5
+
+; 64R2:          movn $5, $6, $4
+; 64R2:          move $2, $5
+
+; 64R6-DAG:      selnez $[[T0:[0-9]+]], $5, $4
+; 64R6-DAG:      seleqz $[[T1:[0-9]+]], $6, $4
+; 64R6:          or $2, $[[T0]], $[[T1]]
+
+  %tobool = icmp ne i64 %s, 0
+  %cond = select i1 %tobool, i64 %f1, i64 %f0
+  ret i64 %cond
+}
+
+define float @i32_icmp_ne_f32_val(i32 %s, float %f0, float %f1) nounwind readnone {
+entry:
+; ALL-LABEL: i32_icmp_ne_f32_val:
+
+; 32-DAG:        mtc1 $5, $[[F0:f[0-9]+]]
+; 32-DAG:        mtc1 $6, $[[F1:f0]]
+; 32:            movn.s $[[F1]], $[[F0]], $4
+
+; 32R2-DAG:      mtc1 $5, $[[F0:f[0-9]+]]
+; 32R2-DAG:      mtc1 $6, $[[F1:f0]]
+; 32R2:          movn.s $[[F1]], $[[F0]], $4
+
+; 32R6-DAG:      mtc1 $5, $[[F0:f[0-9]+]]
+; 32R6-DAG:      mtc1 $6, $[[F1:f[0-9]+]]
+; 32R6:          sltu $[[T0:[0-9]+]], $zero, $4
+; 32R6:          mtc1 $[[T0]], $[[CC:f0]]
+; 32R6:          sel.s $[[CC]], $[[F1]], $[[F0]]
+
+; 64:            movn.s $f14, $f13, $4
+; 64:            mov.s $f0, $f14
+
+; 64R2:          movn.s $f14, $f13, $4
+; 64R2:          mov.s $f0, $f14
+
+; 64R6:          sltu $[[T0:[0-9]+]], $zero, $4
+; 64R6:          mtc1 $[[T0]], $[[CC:f0]]
+; 64R6:          sel.s $[[CC]], $f14, $f13
+
   %tobool = icmp ne i32 %s, 0
   %cond = select i1 %tobool, float %f0, float %f1
   ret float %cond
 }
 
-define double @sel2_1(i32 %s, double %f0, double %f1) nounwind readnone {
+define double @i32_icmp_ne_f64_val(i32 %s, double %f0, double %f1) nounwind readnone {
 entry:
-; CHECK: movn.d
+; ALL-LABEL: i32_icmp_ne_f64_val:
+
+; 32-DAG:        mtc1 $6, $[[F0:f[1-3]*[02468]+]]
+; 32-DAG:        mtc1 $7, $[[F0H:f[1-3]*[13579]+]]
+; 32-DAG:        ldc1 $[[F1:f0]], 16($sp)
+; 32:            movn.d $[[F1]], $[[F0]], $4
+
+; 32R2-DAG:      mtc1 $6, $[[F0:f[0-9]+]]
+; 32R2-DAG:      mthc1 $7, $[[F0]]
+; 32R2-DAG:      ldc1 $[[F1:f0]], 16($sp)
+; 32R2:          movn.d $[[F1]], $[[F0]], $4
+
+; 32R6-DAG:      mtc1 $6, $[[F0:f[0-9]+]]
+; 32R6-DAG:      mthc1 $7, $[[F0]]
+; 32R6-DAG:      sltu $[[T0:[0-9]+]], $zero, $4
+; 32R6-DAG:      mtc1 $[[T0]], $[[CC:f0]]
+; 32R6-DAG:      ldc1 $[[F1:f[0-9]+]], 16($sp)
+; 32R6:          sel.d $[[CC]], $[[F1]], $[[F0]]
+
+; 64:            movn.d $f14, $f13, $4
+; 64:            mov.d $f0, $f14
+
+; 64R2:          movn.d $f14, $f13, $4
+; 64R2:          mov.d $f0, $f14
+
+; 64R6-DAG:      sltu $[[T0:[0-9]+]], $zero, $4
+; 64R6-DAG:      mtc1 $[[T0]], $[[CC:f0]]
+; 64R6:          sel.d $[[CC]], $f14, $f13
+
   %tobool = icmp ne i32 %s, 0
   %cond = select i1 %tobool, double %f0, double %f1
   ret double %cond
 }
 
-define float @sel3(float %f0, float %f1, float %f2, float %f3) nounwind readnone {
+define float @f32_fcmp_oeq_f32_val(float %f0, float %f1, float %f2, float %f3) nounwind readnone {
 entry:
-; CHECK: c.eq.s
-; CHECK: movt.s
+; ALL-LABEL: f32_fcmp_oeq_f32_val:
+
+; 32-DAG:        mtc1 $6, $[[F2:f[0-9]+]]
+; 32-DAG:        mtc1 $7, $[[F3:f[0-9]+]]
+; 32:            c.eq.s $[[F2]], $[[F3]]
+; 32:            movt.s $f14, $f12, $fcc0
+; 32:            mov.s $f0, $f14
+
+; 32R2-DAG:      mtc1 $6, $[[F2:f[0-9]+]]
+; 32R2-DAG:      mtc1 $7, $[[F3:f[0-9]+]]
+; 32R2:          c.eq.s $[[F2]], $[[F3]]
+; 32R2:          movt.s $f14, $f12, $fcc0
+; 32R2:          mov.s $f0, $f14
+
+; 32R6-DAG:      mtc1 $6, $[[F2:f[0-9]+]]
+; 32R6-DAG:      mtc1 $7, $[[F3:f[0-9]+]]
+; 32R6:          cmp.eq.s $[[CC:f0]], $[[F2]], $[[F3]]
+; 32R6:          sel.s $[[CC]], $f14, $f12
+
+; 64:            c.eq.s $f14, $f15
+; 64:            movt.s $f13, $f12, $fcc0
+; 64:            mov.s $f0, $f13
+
+; 64R2:          c.eq.s $f14, $f15
+; 64R2:          movt.s $f13, $f12, $fcc0
+; 64R2:          mov.s $f0, $f13
+
+; 64R6:          cmp.eq.s $[[CC:f0]], $f14, $f15
+; 64R6:          sel.s $[[CC]], $f13, $f12
+
   %cmp = fcmp oeq float %f2, %f3
   %cond = select i1 %cmp, float %f0, float %f1
   ret float %cond
 }
 
-define float @sel4(float %f0, float %f1, float %f2, float %f3) nounwind readnone {
+define float @f32_fcmp_olt_f32_val(float %f0, float %f1, float %f2, float %f3) nounwind readnone {
 entry:
-; CHECK: c.olt.s
-; CHECK: movt.s
+; ALL-LABEL: f32_fcmp_olt_f32_val:
+
+; 32-DAG:        mtc1 $6, $[[F2:f[0-9]+]]
+; 32-DAG:        mtc1 $7, $[[F3:f[0-9]+]]
+; 32:            c.olt.s $[[F2]], $[[F3]]
+; 32:            movt.s $f14, $f12, $fcc0
+; 32:            mov.s $f0, $f14
+
+; 32R2-DAG:      mtc1 $6, $[[F2:f[0-9]+]]
+; 32R2-DAG:      mtc1 $7, $[[F3:f[0-9]+]]
+; 32R2:          c.olt.s $[[F2]], $[[F3]]
+; 32R2:          movt.s $f14, $f12, $fcc0
+; 32R2:          mov.s $f0, $f14
+
+; 32R6-DAG:      mtc1 $6, $[[F2:f[0-9]+]]
+; 32R6-DAG:      mtc1 $7, $[[F3:f[0-9]+]]
+; 32R6:          cmp.olt.s $[[CC:f0]], $[[F2]], $[[F3]]
+; 32R6:          sel.s $[[CC]], $f14, $f12
+
+; 64:            c.olt.s $f14, $f15
+; 64:            movt.s $f13, $f12, $fcc0
+; 64:            mov.s $f0, $f13
+
+; 64R2:          c.olt.s $f14, $f15
+; 64R2:          movt.s $f13, $f12, $fcc0
+; 64R2:          mov.s $f0, $f13
+
+; 64R6:          cmp.olt.s $[[CC:f0]], $f14, $f15
+; 64R6:          sel.s $[[CC]], $f13, $f12
+
   %cmp = fcmp olt float %f2, %f3
   %cond = select i1 %cmp, float %f0, float %f1
   ret float %cond
 }
 
-define float @sel5(float %f0, float %f1, float %f2, float %f3) nounwind readnone {
+define float @f32_fcmp_ogt_f32_val(float %f0, float %f1, float %f2, float %f3) nounwind readnone {
 entry:
-; CHECK: c.ule.s
-; CHECK: movf.s
+; ALL-LABEL: f32_fcmp_ogt_f32_val:
+
+; 32-DAG:        mtc1 $6, $[[F2:f[0-9]+]]
+; 32-DAG:        mtc1 $7, $[[F3:f[0-9]+]]
+; 32:            c.ule.s $[[F2]], $[[F3]]
+; 32:            movf.s $f14, $f12, $fcc0
+; 32:            mov.s $f0, $f14
+
+; 32R2-DAG:      mtc1 $6, $[[F2:f[0-9]+]]
+; 32R2-DAG:      mtc1 $7, $[[F3:f[0-9]+]]
+; 32R2:          c.ule.s $[[F2]], $[[F3]]
+; 32R2:          movf.s $f14, $f12, $fcc0
+; 32R2:          mov.s $f0, $f14
+
+; 32R6-DAG:      mtc1 $6, $[[F2:f[0-9]+]]
+; 32R6-DAG:      mtc1 $7, $[[F3:f[0-9]+]]
+; 32R6:          cmp.olt.s $[[CC:f0]], $[[F3]], $[[F2]]
+; 32R6:          sel.s $[[CC]], $f14, $f12
+
+; 64:            c.ule.s $f14, $f15
+; 64:            movf.s $f13, $f12, $fcc0
+; 64:            mov.s $f0, $f13
+
+; 64R2:          c.ule.s $f14, $f15
+; 64R2:          movf.s $f13, $f12, $fcc0
+; 64R2:          mov.s $f0, $f13
+
+; 64R6:          cmp.olt.s $[[CC:f0]], $f15, $f14
+; 64R6:          sel.s $[[CC]], $f13, $f12
+
   %cmp = fcmp ogt float %f2, %f3
   %cond = select i1 %cmp, float %f0, float %f1
   ret float %cond
 }
 
-define double @sel5_1(double %f0, double %f1, float %f2, float %f3) nounwind readnone {
+define double @f32_fcmp_ogt_f64_val(double %f0, double %f1, float %f2, float %f3) nounwind readnone {
 entry:
-; CHECK: c.ule.s
-; CHECK: movf.d
+; ALL-LABEL: f32_fcmp_ogt_f64_val:
+
+; 32-DAG:        lwc1 $[[F2:f[0-9]+]], 16($sp)
+; 32-DAG:        lwc1 $[[F3:f[0-9]+]], 20($sp)
+; 32:            c.ule.s $[[F2]], $[[F3]]
+; 32:            movf.d $f14, $f12, $fcc0
+; 32:            mov.d $f0, $f14
+
+; 32R2-DAG:      lwc1 $[[F2:f[0-9]+]], 16($sp)
+; 32R2-DAG:      lwc1 $[[F3:f[0-9]+]], 20($sp)
+; 32R2:          c.ule.s $[[F2]], $[[F3]]
+; 32R2:          movf.d $f14, $f12, $fcc0
+; 32R2:          mov.d $f0, $f14
+
+; 32R6-DAG:      lwc1 $[[F2:f[0-9]+]], 16($sp)
+; 32R6-DAG:      lwc1 $[[F3:f[0-9]+]], 20($sp)
+; 32R6:          cmp.olt.s $[[CC:f0]], $[[F3]], $[[F2]]
+; 32R6:          sel.d $[[CC]], $f14, $f12
+
+; 64:            c.ule.s $f14, $f15
+; 64:            movf.d $f13, $f12, $fcc0
+; 64:            mov.d $f0, $f13
+
+; 64R2:          c.ule.s $f14, $f15
+; 64R2:          movf.d $f13, $f12, $fcc0
+; 64R2:          mov.d $f0, $f13
+
+; 64R6:          cmp.olt.s $[[CC:f0]], $f15, $f14
+; 64R6:          sel.d $[[CC]], $f13, $f12
+
   %cmp = fcmp ogt float %f2, %f3
   %cond = select i1 %cmp, double %f0, double %f1
   ret double %cond
 }
 
-define double @sel6(double %f0, double %f1, double %f2, double %f3) nounwind readnone {
+define double @f64_fcmp_oeq_f64_val(double %f0, double %f1, double %f2, double %f3) nounwind readnone {
 entry:
-; CHECK: c.eq.d
-; CHECK: movt.d
+; ALL-LABEL: f64_fcmp_oeq_f64_val:
+
+; 32-DAG:        ldc1 $[[F2:f[0-9]+]], 16($sp)
+; 32-DAG:        ldc1 $[[F3:f[0-9]+]], 24($sp)
+; 32:            c.eq.d $[[F2]], $[[F3]]
+; 32:            movt.d $f14, $f12, $fcc0
+; 32:            mov.d $f0, $f14
+
+; 32R2-DAG:      ldc1 $[[F2:f[0-9]+]], 16($sp)
+; 32R2-DAG:      ldc1 $[[F3:f[0-9]+]], 24($sp)
+; 32R2:          c.eq.d $[[F2]], $[[F3]]
+; 32R2:          movt.d $f14, $f12, $fcc0
+; 32R2:          mov.d $f0, $f14
+
+; 32R6-DAG:      ldc1 $[[F2:f[0-9]+]], 16($sp)
+; 32R6-DAG:      ldc1 $[[F3:f[0-9]+]], 24($sp)
+; 32R6:          cmp.eq.d $[[CC:f0]], $[[F2]], $[[F3]]
+; 32R6:          sel.d $[[CC]], $f14, $f12
+
+; 64:            c.eq.d $f14, $f15
+; 64:            movt.d $f13, $f12, $fcc0
+; 64:            mov.d $f0, $f13
+
+; 64R2:          c.eq.d $f14, $f15
+; 64R2:          movt.d $f13, $f12, $fcc0
+; 64R2:          mov.d $f0, $f13
+
+; 64R6:          cmp.eq.d $[[CC:f0]], $f14, $f15
+; 64R6:          sel.d $[[CC]], $f13, $f12
+
   %cmp = fcmp oeq double %f2, %f3
   %cond = select i1 %cmp, double %f0, double %f1
   ret double %cond
 }
 
-define double @sel7(double %f0, double %f1, double %f2, double %f3) nounwind readnone {
+define double @f64_fcmp_olt_f64_val(double %f0, double %f1, double %f2, double %f3) nounwind readnone {
 entry:
-; CHECK: c.olt.d
-; CHECK: movt.d
+; ALL-LABEL: f64_fcmp_olt_f64_val:
+
+; 32-DAG:        ldc1 $[[F2:f[0-9]+]], 16($sp)
+; 32-DAG:        ldc1 $[[F3:f[0-9]+]], 24($sp)
+; 32:            c.olt.d $[[F2]], $[[F3]]
+; 32:            movt.d $f14, $f12, $fcc0
+; 32:            mov.d $f0, $f14
+
+; 32R2-DAG:      ldc1 $[[F2:f[0-9]+]], 16($sp)
+; 32R2-DAG:      ldc1 $[[F3:f[0-9]+]], 24($sp)
+; 32R2:          c.olt.d $[[F2]], $[[F3]]
+; 32R2:          movt.d $f14, $f12, $fcc0
+; 32R2:          mov.d $f0, $f14
+
+; 32R6-DAG:      ldc1 $[[F2:f[0-9]+]], 16($sp)
+; 32R6-DAG:      ldc1 $[[F3:f[0-9]+]], 24($sp)
+; 32R6:          cmp.olt.d $[[CC:f0]], $[[F2]], $[[F3]]
+; 32R6:          sel.d $[[CC]], $f14, $f12
+
+; 64:            c.olt.d $f14, $f15
+; 64:            movt.d $f13, $f12, $fcc0
+; 64:            mov.d $f0, $f13
+
+; 64R2:          c.olt.d $f14, $f15
+; 64R2:          movt.d $f13, $f12, $fcc0
+; 64R2:          mov.d $f0, $f13
+
+; 64R6:          cmp.olt.d $[[CC:f0]], $f14, $f15
+; 64R6:          sel.d $[[CC]], $f13, $f12
+
   %cmp = fcmp olt double %f2, %f3
   %cond = select i1 %cmp, double %f0, double %f1
   ret double %cond
 }
 
-define double @sel8(double %f0, double %f1, double %f2, double %f3) nounwind readnone {
+define double @f64_fcmp_ogt_f64_val(double %f0, double %f1, double %f2, double %f3) nounwind readnone {
 entry:
-; CHECK: c.ule.d
-; CHECK: movf.d
+; ALL-LABEL: f64_fcmp_ogt_f64_val:
+
+; 32-DAG:        ldc1 $[[F2:f[0-9]+]], 16($sp)
+; 32-DAG:        ldc1 $[[F3:f[0-9]+]], 24($sp)
+; 32:            c.ule.d $[[F2]], $[[F3]]
+; 32:            movf.d $f14, $f12, $fcc0
+; 32:            mov.d $f0, $f14
+
+; 32R2-DAG:      ldc1 $[[F2:f[0-9]+]], 16($sp)
+; 32R2-DAG:      ldc1 $[[F3:f[0-9]+]], 24($sp)
+; 32R2:          c.ule.d $[[F2]], $[[F3]]
+; 32R2:          movf.d $f14, $f12, $fcc0
+; 32R2:          mov.d $f0, $f14
+
+; 32R6-DAG:      ldc1 $[[F2:f[0-9]+]], 16($sp)
+; 32R6-DAG:      ldc1 $[[F3:f[0-9]+]], 24($sp)
+; 32R6:          cmp.olt.d $[[CC:f0]], $[[F3]], $[[F2]]
+; 32R6:          sel.d $[[CC]], $f14, $f12
+
+; 64:            c.ule.d $f14, $f15
+; 64:            movf.d $f13, $f12, $fcc0
+; 64:            mov.d $f0, $f13
+
+; 64R2:          c.ule.d $f14, $f15
+; 64R2:          movf.d $f13, $f12, $fcc0
+; 64R2:          mov.d $f0, $f13
+
+; 64R6:          cmp.olt.d $[[CC:f0]], $f15, $f14
+; 64R6:          sel.d $[[CC]], $f13, $f12
+
   %cmp = fcmp ogt double %f2, %f3
   %cond = select i1 %cmp, double %f0, double %f1
   ret double %cond
 }
 
-define float @sel8_1(float %f0, float %f1, double %f2, double %f3) nounwind readnone {
+define float @f64_fcmp_ogt_f32_val(float %f0, float %f1, double %f2, double %f3) nounwind readnone {
 entry:
-; CHECK: c.ule.d
-; CHECK: movf.s
+; ALL-LABEL: f64_fcmp_ogt_f32_val:
+
+; 32-DAG:        mtc1 $6, $[[F2:f[1-3]*[02468]+]]
+; 32-DAG:        mtc1 $7, $[[F2H:f[1-3]*[13579]+]]
+; 32-DAG:        ldc1 $[[F3:f[0-9]+]], 16($sp)
+; 32:            c.ule.d $[[F2]], $[[F3]]
+; 32:            movf.s $f14, $f12, $fcc0
+; 32:            mov.s $f0, $f14
+
+; 32R2-DAG:      mtc1 $6, $[[F2:f[0-9]+]]
+; 32R2-DAG:      mthc1 $7, $[[F2]]
+; 32R2-DAG:      ldc1 $[[F3:f[0-9]+]], 16($sp)
+; 32R2:          c.ule.d $[[F2]], $[[F3]]
+; 32R2:          movf.s $f14, $f12, $fcc0
+; 32R2:          mov.s $f0, $f14
+
+; 32R6-DAG:      mtc1 $6, $[[F2:f[0-9]+]]
+; 32R6-DAG:      mthc1 $7, $[[F2]]
+; 32R6-DAG:      ldc1 $[[F3:f[0-9]+]], 16($sp)
+; 32R6:          cmp.olt.d $[[CC:f0]], $[[F3]], $[[F2]]
+; 32R6:          sel.s $[[CC]], $f14, $f12
+
+; 64:            c.ule.d $f14, $f15
+; 64:            movf.s $f13, $f12, $fcc0
+; 64:            mov.s $f0, $f13
+
+; 64R2:          c.ule.d $f14, $f15
+; 64R2:          movf.s $f13, $f12, $fcc0
+; 64R2:          mov.s $f0, $f13
+
+; 64R6:          cmp.olt.d $[[CC:f0]], $f15, $f14
+; 64R6:          sel.s $[[CC]], $f13, $f12
+
   %cmp = fcmp ogt double %f2, %f3
   %cond = select i1 %cmp, float %f0, float %f1
   ret float %cond
 }
 
-define i32 @sel9(i32 %f0, i32 %f1, float %f2, float %f3) nounwind readnone {
+define i32 @f32_fcmp_oeq_i32_val(i32 %f0, i32 %f1, float %f2, float %f3) nounwind readnone {
 entry:
-; CHECK: c.eq.s
-; CHECK: movt
+; ALL-LABEL: f32_fcmp_oeq_i32_val:
+
+; 32-DAG:        mtc1 $6, $[[F2:f[0-9]+]]
+; 32-DAG:        mtc1 $7, $[[F3:f[0-9]+]]
+; 32:            c.eq.s $[[F2]], $[[F3]]
+; 32:            movt $5, $4, $fcc0
+; 32:            move $2, $5
+
+; 32R2-DAG:      mtc1 $6, $[[F2:f[0-9]+]]
+; 32R2-DAG:      mtc1 $7, $[[F3:f[0-9]+]]
+; 32R2:          c.eq.s $[[F2]], $[[F3]]
+; 32R2:          movt $5, $4, $fcc0
+; 32R2:          move $2, $5
+
+; 32R6-DAG:      mtc1 $6, $[[F2:f[0-9]+]]
+; 32R6-DAG:      mtc1 $7, $[[F3:f[0-9]+]]
+; 32R6:          cmp.eq.s $[[CC:f[0-9]+]], $[[F2]], $[[F3]]
+; 32R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 32R6:          seleqz $[[EQ:[0-9]+]], $5, $[[CCGPR]]
+; FIXME: This move is redundant
+; 32R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 32R6:          selnez $[[NE:[0-9]+]], $4, $[[CCGPR]]
+; 32R6:          or $2, $[[NE]], $[[EQ]]
+
+; 64:            c.eq.s $f14, $f15
+; 64:            movt $5, $4, $fcc0
+; 64:            move $2, $5
+
+; 64R2:          c.eq.s $f14, $f15
+; 64R2:          movt $5, $4, $fcc0
+; 64R2:          move $2, $5
+
+; 64R6:          cmp.eq.s $[[CC:f[0-9]+]], $f14, $f15
+; 64R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 64R6:          seleqz $[[EQ:[0-9]+]], $5, $[[CCGPR]]
+; FIXME: This move is redundant
+; 64R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 64R6:          selnez $[[NE:[0-9]+]], $4, $[[CCGPR]]
+; 64R6:          or $2, $[[NE]], $[[EQ]]
+
   %cmp = fcmp oeq float %f2, %f3
   %cond = select i1 %cmp, i32 %f0, i32 %f1
   ret i32 %cond
 }
 
-define i32 @sel10(i32 %f0, i32 %f1, float %f2, float %f3) nounwind readnone {
+define i32 @f32_fcmp_olt_i32_val(i32 %f0, i32 %f1, float %f2, float %f3) nounwind readnone {
 entry:
-; CHECK: c.olt.s
-; CHECK: movt
+; ALL-LABEL: f32_fcmp_olt_i32_val:
+
+; 32-DAG:        mtc1 $6, $[[F2:f[0-9]+]]
+; 32-DAG:        mtc1 $7, $[[F3:f[0-9]+]]
+; 32:            c.olt.s $[[F2]], $[[F3]]
+; 32:            movt $5, $4, $fcc0
+; 32:            move $2, $5
+
+; 32R2-DAG:      mtc1 $6, $[[F2:f[0-9]+]]
+; 32R2-DAG:      mtc1 $7, $[[F3:f[0-9]+]]
+; 32R2:          c.olt.s $[[F2]], $[[F3]]
+; 32R2:          movt $5, $4, $fcc0
+; 32R2:          move $2, $5
+
+; 32R6-DAG:      mtc1 $6, $[[F2:f[0-9]+]]
+; 32R6-DAG:      mtc1 $7, $[[F3:f[0-9]+]]
+; 32R6:          cmp.olt.s $[[CC:f[0-9]+]], $[[F2]], $[[F3]]
+; 32R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 32R6:          seleqz $[[EQ:[0-9]+]], $5, $[[CCGPR]]
+; FIXME: This move is redundant
+; 32R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 32R6:          selnez $[[NE:[0-9]+]], $4, $[[CCGPR]]
+; 32R6:          or $2, $[[NE]], $[[EQ]]
+
+; 64:            c.olt.s $f14, $f15
+; 64:            movt $5, $4, $fcc0
+; 64:            move $2, $5
+
+; 64R2:          c.olt.s $f14, $f15
+; 64R2:          movt $5, $4, $fcc0
+; 64R2:          move $2, $5
+
+; 64R6:          cmp.olt.s $[[CC:f[0-9]+]], $f14, $f15
+; 64R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 64R6:          seleqz $[[EQ:[0-9]+]], $5, $[[CCGPR]]
+; FIXME: This move is redundant
+; 64R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 64R6:          selnez $[[NE:[0-9]+]], $4, $[[CCGPR]]
+; 64R6:          or $2, $[[NE]], $[[EQ]]
   %cmp = fcmp olt float %f2, %f3
   %cond = select i1 %cmp, i32 %f0, i32 %f1
   ret i32 %cond
 }
 
-define i32 @sel11(i32 %f0, i32 %f1, float %f2, float %f3) nounwind readnone {
+define i32 @f32_fcmp_ogt_i32_val(i32 %f0, i32 %f1, float %f2, float %f3) nounwind readnone {
 entry:
-; CHECK: c.ule.s
-; CHECK: movf
+; ALL-LABEL: f32_fcmp_ogt_i32_val:
+
+; 32-DAG:        mtc1 $6, $[[F2:f[0-9]+]]
+; 32-DAG:        mtc1 $7, $[[F3:f[0-9]+]]
+; 32:            c.ule.s $[[F2]], $[[F3]]
+; 32:            movf $5, $4, $fcc0
+; 32:            move $2, $5
+
+; 32R2-DAG:      mtc1 $6, $[[F2:f[0-9]+]]
+; 32R2-DAG:      mtc1 $7, $[[F3:f[0-9]+]]
+; 32R2:          c.ule.s $[[F2]], $[[F3]]
+; 32R2:          movf $5, $4, $fcc0
+; 32R2:          move $2, $5
+
+; 32R6-DAG:      mtc1 $6, $[[F2:f[0-9]+]]
+; 32R6-DAG:      mtc1 $7, $[[F3:f[0-9]+]]
+; 32R6:          cmp.olt.s $[[CC:f[0-9]+]], $[[F3]], $[[F2]]
+; 32R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 32R6:          seleqz $[[EQ:[0-9]+]], $5, $[[CCGPR]]
+; FIXME: This move is redundant
+; 32R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 32R6:          selnez $[[NE:[0-9]+]], $4, $[[CCGPR]]
+; 32R6:          or $2, $[[NE]], $[[EQ]]
+
+; 64:            c.ule.s $f14, $f15
+; 64:            movf $5, $4, $fcc0
+; 64:            move $2, $5
+
+; 64R2:          c.ule.s $f14, $f15
+; 64R2:          movf $5, $4, $fcc0
+; 64R2:          move $2, $5
+
+; 64R6:          cmp.olt.s $[[CC:f[0-9]+]], $f15, $f14
+; 64R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 64R6:          seleqz $[[EQ:[0-9]+]], $5, $[[CCGPR]]
+; FIXME: This move is redundant
+; 64R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 64R6:          selnez $[[NE:[0-9]+]], $4, $[[CCGPR]]
+; 64R6:          or $2, $[[NE]], $[[EQ]]
+
   %cmp = fcmp ogt float %f2, %f3
   %cond = select i1 %cmp, i32 %f0, i32 %f1
   ret i32 %cond
 }
 
-define i32 @sel12(i32 %f0, i32 %f1) nounwind readonly {
+define i32 @f64_fcmp_oeq_i32_val(i32 %f0, i32 %f1) nounwind readonly {
 entry:
-; CHECK: c.eq.d
-; CHECK: movt
+; ALL-LABEL: f64_fcmp_oeq_i32_val:
+
+; 32-DAG:        addiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(_gp_disp)
+; 32-DAG:        addu $[[GOT:[0-9]+]], $[[T0]], $25
+; 32-DAG:        lw $[[D2:[0-9]+]], %got(d2)($1)
+; 32-DAG:        ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 32-DAG:        lw $[[D3:[0-9]+]], %got(d3)($1)
+; 32-DAG:        ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 32:            c.eq.d $[[TMP]], $[[TMP1]]
+; 32:            movt $5, $4, $fcc0
+; 32:            move $2, $5
+
+; 32R2-DAG:      addiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(_gp_disp)
+; 32R2-DAG:      addu $[[GOT:[0-9]+]], $[[T0]], $25
+; 32R2-DAG:      lw $[[D2:[0-9]+]], %got(d2)($1)
+; 32R2-DAG:      ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 32R2-DAG:      lw $[[D3:[0-9]+]], %got(d3)($1)
+; 32R2-DAG:      ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 32R2:          c.eq.d $[[TMP]], $[[TMP1]]
+; 32R2:          movt $5, $4, $fcc0
+; 32R2:          move $2, $5
+
+; 32R6-DAG:      addiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(_gp_disp)
+; 32R6-DAG:      addu $[[GOT:[0-9]+]], $[[T0]], $25
+; 32R6-DAG:      lw $[[D2:[0-9]+]], %got(d2)($1)
+; 32R6-DAG:      ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 32R6-DAG:      lw $[[D3:[0-9]+]], %got(d3)($1)
+; 32R6-DAG:      ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 32R6:          cmp.eq.d $[[CC:f[0-9]+]], $[[TMP]], $[[TMP1]]
+; 32R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 32R6:          seleqz $[[EQ:[0-9]+]], $5, $[[CCGPR]]
+; FIXME: This move is redundant
+; 32R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 32R6:          selnez $[[NE:[0-9]+]], $4, $[[CCGPR]]
+; 32R6:          or $2, $[[NE]], $[[EQ]]
+
+; 64-DAG:        daddiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(%neg(%gp_rel(f64_fcmp_oeq_i32_val)))
+; 64-DAG:        daddu $[[GOT:[0-9]+]], $[[T0]], $25
+; 64-DAG:        ld $[[D2:[0-9]+]], %got_disp(d2)($1)
+; 64-DAG:        ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 64-DAG:        ld $[[D3:[0-9]+]], %got_disp(d3)($1)
+; 64-DAG:        ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 64:            c.eq.d $[[TMP]], $[[TMP1]]
+; 64:            movt $5, $4, $fcc0
+; 64:            move $2, $5
+
+; 64R2-DAG:      daddiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(%neg(%gp_rel(f64_fcmp_oeq_i32_val)))
+; 64R2-DAG:      daddu $[[GOT:[0-9]+]], $[[T0]], $25
+; 64R2-DAG:      ld $[[D2:[0-9]+]], %got_disp(d2)($1)
+; 64R2-DAG:      ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 64R2-DAG:      ld $[[D3:[0-9]+]], %got_disp(d3)($1)
+; 64R2-DAG:      ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 64R2:          c.eq.d $[[TMP]], $[[TMP1]]
+; 64R2:          movt $5, $4, $fcc0
+; 64R2:          move $2, $5
+
+; 64R6-DAG:      daddiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(%neg(%gp_rel(f64_fcmp_oeq_i32_val)))
+; 64R6-DAG:      daddu $[[GOT:[0-9]+]], $[[T0]], $25
+; 64R6-DAG:      ld $[[D2:[0-9]+]], %got_disp(d2)($1)
+; 64R6-DAG:      ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 64R6-DAG:      ld $[[D3:[0-9]+]], %got_disp(d3)($1)
+; 64R6-DAG:      ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 64R6:          cmp.eq.d $[[CC:f[0-9]+]], $[[TMP]], $[[TMP1]]
+; 64R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 64R6:          seleqz $[[EQ:[0-9]+]], $5, $[[CCGPR]]
+; FIXME: This move is redundant
+; 64R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 64R6:          selnez $[[NE:[0-9]+]], $4, $[[CCGPR]]
+; 64R6:          or $2, $[[NE]], $[[EQ]]
+
   %tmp = load double* @d2, align 8
   %tmp1 = load double* @d3, align 8
   %cmp = fcmp oeq double %tmp, %tmp1
   %cond = select i1 %cmp, i32 %f0, i32 %f1
   ret i32 %cond
 }
 
-define i32 @sel13(i32 %f0, i32 %f1) nounwind readonly {
+define i32 @f64_fcmp_olt_i32_val(i32 %f0, i32 %f1) nounwind readonly {
 entry:
-; CHECK: c.olt.d
-; CHECK: movt
+; ALL-LABEL: f64_fcmp_olt_i32_val:
+
+; 32-DAG:        addiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(_gp_disp)
+; 32-DAG:        addu $[[GOT:[0-9]+]], $[[T0]], $25
+; 32-DAG:        lw $[[D2:[0-9]+]], %got(d2)($1)
+; 32-DAG:        ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 32-DAG:        lw $[[D3:[0-9]+]], %got(d3)($1)
+; 32-DAG:        ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 32:            c.olt.d $[[TMP]], $[[TMP1]]
+; 32:            movt $5, $4, $fcc0
+; 32:            move $2, $5
+
+; 32R2-DAG:      addiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(_gp_disp)
+; 32R2-DAG:      addu $[[GOT:[0-9]+]], $[[T0]], $25
+; 32R2-DAG:      lw $[[D2:[0-9]+]], %got(d2)($1)
+; 32R2-DAG:      ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 32R2-DAG:      lw $[[D3:[0-9]+]], %got(d3)($1)
+; 32R2-DAG:      ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 32R2:          c.olt.d $[[TMP]], $[[TMP1]]
+; 32R2:          movt $5, $4, $fcc0
+; 32R2:          move $2, $5
+
+; 32R6-DAG:      addiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(_gp_disp)
+; 32R6-DAG:      addu $[[GOT:[0-9]+]], $[[T0]], $25
+; 32R6-DAG:      lw $[[D2:[0-9]+]], %got(d2)($1)
+; 32R6-DAG:      ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 32R6-DAG:      lw $[[D3:[0-9]+]], %got(d3)($1)
+; 32R6-DAG:      ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 32R6:          cmp.olt.d $[[CC:f[0-9]+]], $[[TMP]], $[[TMP1]]
+; 32R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 32R6:          seleqz $[[EQ:[0-9]+]], $5, $[[CCGPR]]
+; FIXME: This move is redundant
+; 32R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 32R6:          selnez $[[NE:[0-9]+]], $4, $[[CCGPR]]
+; 32R6:          or $2, $[[NE]], $[[EQ]]
+
+; 64-DAG:        daddiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(%neg(%gp_rel(f64_fcmp_olt_i32_val)))
+; 64-DAG:        daddu $[[GOT:[0-9]+]], $[[T0]], $25
+; 64-DAG:        ld $[[D2:[0-9]+]], %got_disp(d2)($1)
+; 64-DAG:        ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 64-DAG:        ld $[[D3:[0-9]+]], %got_disp(d3)($1)
+; 64-DAG:        ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 64:            c.olt.d $[[TMP]], $[[TMP1]]
+; 64:            movt $5, $4, $fcc0
+; 64:            move $2, $5
+
+; 64R2-DAG:      daddiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(%neg(%gp_rel(f64_fcmp_olt_i32_val)))
+; 64R2-DAG:      daddu $[[GOT:[0-9]+]], $[[T0]], $25
+; 64R2-DAG:      ld $[[D2:[0-9]+]], %got_disp(d2)($1)
+; 64R2-DAG:      ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 64R2-DAG:      ld $[[D3:[0-9]+]], %got_disp(d3)($1)
+; 64R2-DAG:      ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 64R2:          c.olt.d $[[TMP]], $[[TMP1]]
+; 64R2:          movt $5, $4, $fcc0
+; 64R2:          move $2, $5
+
+; 64R6-DAG:      daddiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(%neg(%gp_rel(f64_fcmp_olt_i32_val)))
+; 64R6-DAG:      daddu $[[GOT:[0-9]+]], $[[T0]], $25
+; 64R6-DAG:      ld $[[D2:[0-9]+]], %got_disp(d2)($1)
+; 64R6-DAG:      ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 64R6-DAG:      ld $[[D3:[0-9]+]], %got_disp(d3)($1)
+; 64R6-DAG:      ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 64R6:          cmp.olt.d $[[CC:f[0-9]+]], $[[TMP]], $[[TMP1]]
+; 64R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 64R6:          seleqz $[[EQ:[0-9]+]], $5, $[[CCGPR]]
+; FIXME: This move is redundant
+; 64R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 64R6:          selnez $[[NE:[0-9]+]], $4, $[[CCGPR]]
+; 64R6:          or $2, $[[NE]], $[[EQ]]
+
   %tmp = load double* @d2, align 8
   %tmp1 = load double* @d3, align 8
   %cmp = fcmp olt double %tmp, %tmp1
   %cond = select i1 %cmp, i32 %f0, i32 %f1
   ret i32 %cond
 }
 
-define i32 @sel14(i32 %f0, i32 %f1) nounwind readonly {
+define i32 @f64_fcmp_ogt_i32_val(i32 %f0, i32 %f1) nounwind readonly {
 entry:
-; CHECK: c.ule.d
-; CHECK: movf
+; ALL-LABEL: f64_fcmp_ogt_i32_val:
+
+; 32-DAG:        addiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(_gp_disp)
+; 32-DAG:        addu $[[GOT:[0-9]+]], $[[T0]], $25
+; 32-DAG:        lw $[[D2:[0-9]+]], %got(d2)($1)
+; 32-DAG:        ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 32-DAG:        lw $[[D3:[0-9]+]], %got(d3)($1)
+; 32-DAG:        ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 32:            c.ule.d $[[TMP]], $[[TMP1]]
+; 32:            movf $5, $4, $fcc0
+; 32:            move $2, $5
+
+; 32R2-DAG:      addiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(_gp_disp)
+; 32R2-DAG:      addu $[[GOT:[0-9]+]], $[[T0]], $25
+; 32R2-DAG:      lw $[[D2:[0-9]+]], %got(d2)($1)
+; 32R2-DAG:      ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 32R2-DAG:      lw $[[D3:[0-9]+]], %got(d3)($1)
+; 32R2-DAG:      ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 32R2:          c.ule.d $[[TMP]], $[[TMP1]]
+; 32R2:          movf $5, $4, $fcc0
+; 32R2:          move $2, $5
+
+; 32R6-DAG:      addiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(_gp_disp)
+; 32R6-DAG:      addu $[[GOT:[0-9]+]], $[[T0]], $25
+; 32R6-DAG:      lw $[[D2:[0-9]+]], %got(d2)($1)
+; 32R6-DAG:      ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 32R6-DAG:      lw $[[D3:[0-9]+]], %got(d3)($1)
+; 32R6-DAG:      ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 32R6:          cmp.olt.d $[[CC:f[0-9]+]], $[[TMP1]], $[[TMP]]
+; 32R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 32R6:          seleqz $[[EQ:[0-9]+]], $5, $[[CCGPR]]
+; FIXME: This move is redundant
+; 32R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 32R6:          selnez $[[NE:[0-9]+]], $4, $[[CCGPR]]
+; 32R6:          or $2, $[[NE]], $[[EQ]]
+
+; 64-DAG:        daddiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(%neg(%gp_rel(f64_fcmp_ogt_i32_val)))
+; 64-DAG:        daddu $[[GOT:[0-9]+]], $[[T0]], $25
+; 64-DAG:        ld $[[D2:[0-9]+]], %got_disp(d2)($1)
+; 64-DAG:        ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 64-DAG:        ld $[[D3:[0-9]+]], %got_disp(d3)($1)
+; 64-DAG:        ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 64:            c.ule.d $[[TMP]], $[[TMP1]]
+; 64:            movf $5, $4, $fcc0
+; 64:            move $2, $5
+
+; 64R2-DAG:      daddiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(%neg(%gp_rel(f64_fcmp_ogt_i32_val)))
+; 64R2-DAG:      daddu $[[GOT:[0-9]+]], $[[T0]], $25
+; 64R2-DAG:      ld $[[D2:[0-9]+]], %got_disp(d2)($1)
+; 64R2-DAG:      ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 64R2-DAG:      ld $[[D3:[0-9]+]], %got_disp(d3)($1)
+; 64R2-DAG:      ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 64R2:          c.ule.d $[[TMP]], $[[TMP1]]
+; 64R2:          movf $5, $4, $fcc0
+; 64R2:          move $2, $5
+
+; 64R6-DAG:      daddiu $[[T0:[0-9]+]], ${{[0-9]+}}, %lo(%neg(%gp_rel(f64_fcmp_ogt_i32_val)))
+; 64R6-DAG:      daddu $[[GOT:[0-9]+]], $[[T0]], $25
+; 64R6-DAG:      ld $[[D2:[0-9]+]], %got_disp(d2)($1)
+; 64R6-DAG:      ldc1 $[[TMP:f[0-9]+]], 0($[[D2]])
+; 64R6-DAG:      ld $[[D3:[0-9]+]], %got_disp(d3)($1)
+; 64R6-DAG:      ldc1 $[[TMP1:f[0-9]+]], 0($[[D3]])
+; 64R6:          cmp.olt.d $[[CC:f[0-9]+]], $[[TMP1]], $[[TMP]]
+; 64R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 64R6:          seleqz $[[EQ:[0-9]+]], $5, $[[CCGPR]]
+; FIXME: This move is redundant
+; 64R6:          mfc1 $[[CCGPR:[0-9]+]], $[[CC]]
+; 64R6:          selnez $[[NE:[0-9]+]], $4, $[[CCGPR]]
+; 64R6:          or $2, $[[NE]], $[[EQ]]
+
   %tmp = load double* @d2, align 8
   %tmp1 = load double* @d3, align 8
   %cmp = fcmp ogt double %tmp, %tmp1
   %cond = select i1 %cmp, i32 %f0, i32 %f1
   ret i32 %cond
 }

test/CodeGen/Mips/selectcc.ll

-; RUN: llc -march=mipsel < %s
-; RUN: llc -march=mipsel -pre-RA-sched=source < %s | FileCheck %s --check-prefix=SOURCE-SCHED
+; RUN: llc -march=mipsel -mcpu=mips32 < %s
+; RUN: llc -march=mipsel -mcpu=mips32 -pre-RA-sched=source < %s | FileCheck %s --check-prefix=SOURCE-SCHED
+; RUN: llc -march=mipsel -mcpu=mips32r2 < %s
+; RUN: llc -march=mipsel -mcpu=mips32r2 -pre-RA-sched=source < %s | FileCheck %s --check-prefix=SOURCE-SCHED
 
 @gf0 = external global float
 @gf1 = external global float
 @gd0 = external global double
 @gd1 = external global double
 
 define float @select_cc_f32(float %a, float %b) nounwind {
 entry:

test/CodeGen/Mips/zeroreg.ll

-; RUN: llc < %s -march=mipsel | FileCheck %s
+; RUN: llc < %s -march=mipsel -mcpu=mips32   | FileCheck %s -check-prefix=ALL -check-prefix=32-CMOV
+; RUN: llc < %s -march=mipsel -mcpu=mips32r2 | FileCheck %s -check-prefix=ALL -check-prefix=32-CMOV
+; RUN: llc < %s -march=mipsel -mcpu=mips32r6 | FileCheck %s -check-prefix=ALL -check-prefix=32R6
+; RUN: llc < %s -march=mipsel -mcpu=mips4    | FileCheck %s -check-prefix=ALL -check-prefix=64-CMOV
+; RUN: llc < %s -march=mipsel -mcpu=mips64   | FileCheck %s -check-prefix=ALL -check-prefix=64-CMOV
+; RUN: llc < %s -march=mipsel -mcpu=mips64r2 | FileCheck %s -check-prefix=ALL -check-prefix=64-CMOV
+; RUN: llc < %s -march=mipsel -mcpu=mips64r6 | FileCheck %s -check-prefix=ALL -check-prefix=64R6
 
 @g1 = external global i32
 
-define i32 @foo0(i32 %s) nounwind readonly {
+define i32 @sel_icmp_nez_i32_z0(i32 %s) nounwind readonly {
 entry:
-; CHECK:     movn ${{[0-9]+}}, $zero
+; ALL-LABEL: sel_icmp_nez_i32_z0:
+
+; 32-CMOV:       lw $2, 0(${{[0-9]+}})
+; 32-CMOV:       movn $2, $zero, $4
+
+; 32R6:          lw $[[R0:[0-9]+]], 0(${{[0-9]+}})
+; 32R6:          seleqz $2, $[[R0]], $4
+
+; 64-CMOV:       lw $2, 0(${{[0-9]+}})
+; 64-CMOV:       movn $2, $zero, $4
+
+; 64R6:          lw $[[R0:[0-9]+]], 0(${{[0-9]+}})
+; 64R6:          seleqz $2, $[[R0]], $4
+
   %tobool = icmp ne i32 %s, 0
   %0 = load i32* @g1, align 4
   %cond = select i1 %tobool, i32 0, i32 %0
   ret i32 %cond
 }
 
-define i32 @foo1(i32 %s) nounwind readonly {
+define i32 @sel_icmp_nez_i32_z1(i32 %s) nounwind readonly {
 entry:
-; CHECK:     movz ${{[0-9]+}}, $zero
+; ALL-LABEL: sel_icmp_nez_i32_z1:
+
+; 32-CMOV:       lw $2, 0(${{[0-9]+}})
+; 32-CMOV:       movz $2, $zero, $4
+
+; 32R6:          lw $[[R0:[0-9]+]], 0(${{[0-9]+}})
+; 32R6:          selnez $2, $[[R0]], $4
+
+; 64-CMOV:       lw $2, 0(${{[0-9]+}})
+; 64-CMOV:       movz $2, $zero, $4
+
+; 64R6:          lw $[[R0:[0-9]+]], 0(${{[0-9]+}})
+; 64R6:          selnez $2, $[[R0]], $4
+
   %tobool = icmp ne i32 %s, 0
   %0 = load i32* @g1, align 4
   %cond = select i1 %tobool, i32 %0, i32 0
   ret i32 %cond
 }
+
+@g2 = external global i64
+
+define i64 @sel_icmp_nez_i64_z0(i64 %s) nounwind readonly {
+entry:
+; ALL-LABEL: sel_icmp_nez_i64_z0:
+
+; 32-CMOV-DAG:   lw $[[R0:2]], 0(${{[0-9]+}})
+; 32-CMOV-DAG:   lw $[[R1:3]], 4(${{[0-9]+}})
+; 32-CMOV-DAG:   movn $[[R0]], $zero, $4
+; 32-CMOV-DAG:   movn $[[R1]], $zero, $4
+
+; 32R6-DAG:      lw $[[R0:[0-9]+]], 0(${{[0-9]+}})
+; 32R6-DAG:      lw $[[R1:[0-9]+]], 4(${{[0-9]+}})
+; 32R6-DAG:      or $[[CC:[0-9]+]], $4, $5
+; 32R6-DAG:      seleqz $2, $[[R0]], $[[CC]]
+; 32R6-DAG:      seleqz $3, $[[R1]], $[[CC]]
+
+; 64-CMOV:       ld $2, 0(${{[0-9]+}})
+; 64-CMOV:       movn $2, $zero, $4
+
+; 64R6:          ld $[[R0:[0-9]+]], 0(${{[0-9]+}})
+; 64R6:          seleqz $2, $[[R0]], $4
+
+  %tobool = icmp ne i64 %s, 0
+  %0 = load i64* @g2, align 4
+  %cond = select i1 %tobool, i64 0, i64 %0
+  ret i64 %cond
+}
+
+define i64 @sel_icmp_nez_i64_z1(i64 %s) nounwind readonly {
+entry:
+; ALL-LABEL: sel_icmp_nez_i64_z1:
+
+; 32-CMOV-DAG:   lw $[[R0:2]], 0(${{[0-9]+}})
+; 32-CMOV-DAG:   lw $[[R1:3]], 4(${{[0-9]+}})
+; 32-CMOV-DAG:   movz $[[R0]], $zero, $4
+; 32-CMOV-DAG:   movz $[[R1]], $zero, $4
+
+; 32R6-DAG:      lw $[[R0:[0-9]+]], 0(${{[0-9]+}})
+; 32R6-DAG:      lw $[[R1:[0-9]+]], 4(${{[0-9]+}})
+; 32R6-DAG:      or $[[CC:[0-9]+]], $4, $5
+; 32R6-DAG:      selnez $2, $[[R0]], $[[CC]]
+; 32R6-DAG:      selnez $3, $[[R1]], $[[CC]]
+
+; 64-CMOV:       ld $2, 0(${{[0-9]+}})
+; 64-CMOV:       movz $2, $zero, $4
+
+; 64R6:          ld $[[R0:[0-9]+]], 0(${{[0-9]+}})
+; 64R6:          selnez $2, $[[R0]], $4
+
+  %tobool = icmp ne i64 %s, 0
+  %0 = load i64* @g2, align 4
+  %cond = select i1 %tobool, i64 %0, i64 0
+  ret i64 %cond
+}

test/MC/Mips/mips32r6/invalid-mips1.s

 # Instructions that are invalid
 #
 # RUN: not llvm-mc %s -triple=mips-unknown-linux -show-encoding -mcpu=mips32r6 \
 # RUN:     2>%t1
 # RUN: FileCheck %s < %t1
 
   .set noat
         addi      $13,$9,26322        # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        c.ngl.d   $f29,$f29           # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        c.ngle.d  $f0,$f16            # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        c.sf.d    $f30,$f0            # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        c.sf.s    $f14,$f22           # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mfhi      $s3                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mfhi      $sp                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mflo      $s1                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mthi      $s1                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mtlo      $25                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mtlo      $sp                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mult      $sp,$s4             # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mult      $sp,$v0             # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         multu     $9,$s2              # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         multu     $gp,$k0             # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
 #       div has been re-encoded. See valid.s
 #       divu has been re-encoded. See valid.s

test/MC/Mips/mips32r6/invalid-mips32.s

 # Instructions that are invalid
 #
 # RUN: not llvm-mc %s -triple=mips-unknown-linux -show-encoding -mcpu=mips32r6 \
 # RUN:     2>%t1
 # RUN: FileCheck %s < %t1
 
         .set noat
         madd      $s6,$13       # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         madd      $zero,$9      # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         maddu     $s3,$gp       # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         maddu     $24,$s2       # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movf      $gp,$8,$fcc7        # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movf.d    $f6,$f11,$fcc5      # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movf.s    $f23,$f5,$fcc6      # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movn      $v1,$s1,$s0         # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movn.d    $f27,$f21,$k0       # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movn.s    $f12,$f0,$s7        # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movt      $zero,$s4,$fcc5     # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movt.d    $f0,$f2,$fcc0       # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movt.s    $f30,$f2,$fcc1      # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movz      $a1,$s6,$9          # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movz.d    $f12,$f29,$9        # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movz.s    $f25,$f7,$v1        # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         msub      $s7,$k1       # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         msubu     $15,$a1       # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled

test/MC/Mips/mips64r6/invalid-mips1.s

 # Instructions that are invalid
 #
 # RUN: not llvm-mc %s -triple=mips64-unknown-linux -show-encoding -mcpu=mips64r6 \
 # RUN:     2>%t1
 # RUN: FileCheck %s < %t1
 
   .set noat
         addi      $13,$9,26322        # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        c.ngl.d   $f29,$f29           # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        c.ngle.d  $f0,$f16            # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        c.sf.d    $f30,$f0            # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        c.sf.s    $f14,$f22           # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mfhi      $s3                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mfhi      $sp                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mflo      $s1                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mthi      $s1                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mtlo      $25                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mtlo      $sp                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mult      $sp,$s4             # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mult      $sp,$v0             # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         multu     $9,$s2              # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         multu     $gp,$k0             # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
 #       div has been re-encoded. See valid.s
 #       divu has been re-encoded. See valid.s

test/MC/Mips/mips64r6/invalid-mips64.s

 # Instructions that are invalid
 #
 # RUN: not llvm-mc %s -triple=mips64-unknown-linux -show-encoding -mcpu=mips64r6 \
 # RUN:     2>%t1
 # RUN: FileCheck %s < %t1
 
   .set noat
         addi      $13,$9,26322        # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         dmult     $s7,$9              # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         dmultu    $a1,$a2             # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         jalx      4                   # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mfhi      $s3                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mfhi      $sp                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mflo      $s1                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movf      $gp,$8,$fcc7        # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movf.d    $f6,$f11,$fcc5      # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movf.s    $f23,$f5,$fcc6      # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movn      $v1,$s1,$s0         # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movn.d    $f27,$f21,$k0       # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movn.s    $f12,$f0,$s7        # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movt      $zero,$s4,$fcc5     # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movt.d    $f0,$f2,$fcc0       # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movt.s    $f30,$f2,$fcc1      # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movz      $a1,$s6,$9          # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movz.d    $f12,$f29,$9        # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
+        movz.s    $f25,$f7,$v1        # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mthi      $s1                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mtlo      $25                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mtlo      $sp                 # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mult      $sp,$s4             # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         mult      $sp,$v0             # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         multu     $9,$s2              # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         multu     $gp,$k0             # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled
         teqi      $s5,-17504          # CHECK: :[[@LINE]]:{{[0-9]+}}: error: instruction requires a CPU feature not currently enabled