AMD Zen Scheduler Model Update

[GGanesh]

Summary

This patch is an update to the zen scheduler model.
This is the second patch in the list of incremental patches.

This patch has the following

1. Regex based Instruction itineraries for integer instructions.
2. The instructions are grouped as per the nature of the instructions (move, arithmetic, logic, Misc, Control Transfer).
3. FP instructions and their itineraries are added which includes values for SSE4A, BMI, BMI2 and SHA instructions.

[RKSimon]

lib/Target/X86/X86ScheduleZnver1.td

267

Shouldn't all the memory instructions (folding etc.) have an AGU resource dependency?

[GGanesh]

Updated for the itineraries of memory variants of the instructions.

[GGanesh]

Simon, Craig Topper! My next increment is ready. If this patch can be accepted and committed, I will rebase and submit the next patch.
Or should I submit the next patch as an incremental patch with the changes put forth in this patch? Please help!

[RKSimon]

Since its the only other AMD model, please can you add the SSE4A values?

[GGanesh]

Yes Simon! I will include the SSE4A instructions, their itineraries in the next patch. I will include tests verifying them as well.
If this patch is okay, can you please commit this patch on my behalf.

[GGanesh]

Updated the patch as per Simon's comments.
Added the FP instruction itineraries which includes SSE4A and SHA instructions.

[javed.absar]

lib/Target/X86/X86ScheduleZnver1.td

514

nitpick. Latency = 1 is default, so mention can be omitted.

787

ResourceCycles = [1] is default.

[GGanesh]

Updated as per Javed's comments!

[GGanesh]

Simon! If you are okay with the patch, can you please commit the patch on my behalf!

[andreadb]

I noticed that you forgot to include llvm-commits as a subscriber.

[craig.topper]

lib/Target/X86/X86ScheduleZnver1.td

782

Shouldn't VCVTSD2SSrm use the AGU?

[GGanesh]

Updated for review comments from Craig Topper!

[craig.topper]

LGTM

lib/Target/X86/X86ScheduleZnver1.td

 
 // Integer division issued on ALU2.
 def ZnDivider : ProcResource<1>;
 
 // 4 Cycles load-to use Latency is captured
 def : ReadAdvance<ReadAfterLd, 4>;
 
 // (a folded load is an instruction that loads and does some operation)
-// Ex: ADDPD xmm,[mem]-> This instruction has two micro-ops 
+// Ex: ADDPD xmm,[mem]-> This instruction has two micro-ops
 // Instructions with folded loads are usually micro-fused, so they only appear
 // as two micro-ops.
 //      a. load and
 //      b. addpd
 // This multiclass is for folded loads for integer units.
 multiclass ZnWriteResPair<X86FoldableSchedWrite SchedRW,
                           ProcResourceKind ExePort,
                           int Lat> {
   // Register variant takes 1-cycle on Execution Port.
   def : WriteRes<SchedRW, [ExePort]> { let Latency = Lat; }
 
-  // Memory variant also uses a cycle on ZnAGU 
+  // Memory variant also uses a cycle on ZnAGU
   // adds 4 cycles to the latency.
   def : WriteRes<SchedRW.Folded, [ZnAGU, ExePort]> {
      let Latency = !add(Lat, 4);
   }
 }
 
 // This multiclass is for folded loads for floating point units.
 multiclass ZnWriteResFpuPair<X86FoldableSchedWrite SchedRW,
                           ProcResourceKind ExePort,
                           int Lat> {
   // Register variant takes 1-cycle on Execution Port.
   def : WriteRes<SchedRW, [ExePort]> { let Latency = Lat; }
 
   // Memory variant also uses a cycle on ZnAGU
   // adds 7 cycles to the latency.
   def : WriteRes<SchedRW.Folded, [ZnAGU, ExePort]> {
      let Latency = !add(Lat, 7);
   }
 }
 
-// WriteRMW is set for instructions with Memory write 
+// WriteRMW is set for instructions with Memory write
 // operation in codegen
 def : WriteRes<WriteRMW, [ZnAGU]>;
 
 def : WriteRes<WriteStore, [ZnAGU]>;
 def : WriteRes<WriteMove,  [ZnALU]>;
 def : WriteRes<WriteLoad,  [ZnAGU]> { let Latency = 8; }
 
 def : WriteRes<WriteZero,  []>;
let Latency = 100 in {
   def : WriteRes<WritePCmpIStrMLd, []>;
   def : WriteRes<WritePCmpEStrI, []>;
   def : WriteRes<WritePCmpEStrILd, []>;
   def : WriteRes<WritePCmpEStrM, []>;
   def : WriteRes<WritePCmpEStrMLd, []>;
   def : WriteRes<WritePCmpIStrI, []>;
   def : WriteRes<WritePCmpIStrILd, []>;
   }
+
+//=== Regex based itineraries ===//
+// Notation:
+// - r: register.
+// - m = memory.
+// - i = immediate
+// - mm: 64 bit mmx register.
+// - x = 128 bit xmm register.
+// - (x)mm = mmx or xmm register.
+// - y = 256 bit ymm register.
+// - v = any vector register.
+
+//=== Integer Instructions ===//
+//-- Move instructions --//
+// MOV.
+// r16,m.
+def : InstRW<[WriteALULd], (instregex "MOV16rm")>;
+
+// MOVSX, MOVZX.
+// r,m.
+def : InstRW<[WriteLoad], (instregex "MOV(S|Z)X32rm(8|16)")>;
+
+// CMOVcc.
+// r,r.
+def : InstRW<[WriteALU],
+      (instregex "CMOV(O|NO|B|AE|E|NE|BE|A|S|NS|P|NP|L|GE|LE|G)(16|32|64)rr")>;
+// r,m.
+def : InstRW<[WriteALULd, ReadAfterLd],
+      (instregex "CMOV(O|NO|B|AE|E|NE|BE|A|S|NS|P|NP|L|GE|LE|G)(16|32|64)rm")>;
+
+// XCHG.
+// r,r.
+def ZnWriteXCHG : SchedWriteRes<[ZnALU]> {
+  let NumMicroOps = 2;
+  let ResourceCycles = [2];
+}
+
+def : InstRW<[ZnWriteXCHG], (instregex "XCHG(8|16|32|64)rr", "XCHG(16|32|64)ar")>;
+
+// r,m.
+def ZnWriteXCHGrm : SchedWriteRes<[ZnALU]> {
+  let Latency = 5;
+  let NumMicroOps = 2;
+}
+def : InstRW<[ZnWriteXCHGrm], (instregex "XCHG(8|16|32|64)rm")>;
+
+def : InstRW<[WriteMicrocoded], (instregex "XLAT")>;
+
+// POP16.
+// r.
+def ZnWritePop16r : SchedWriteRes<[ZnAGU]>{
+  let Latency = 5;
+  let NumMicroOps = 2;
+}
+def : InstRW<[ZnWritePop16r], (instregex "POP16rmm")>;
+def : InstRW<[WriteMicrocoded], (instregex "POPF(16|32)")>;
+def : InstRW<[WriteMicrocoded], (instregex "POPA(16|32)")>;
+
+
+// PUSH.
+// r. Has default values.
+// m.
+def ZnWritePUSH : SchedWriteRes<[ZnAGU]>{
+  let Latency = 4;
+}
+def : InstRW<[ZnWritePUSH], (instregex "PUSH(16|32)rmm")>;
+
+//PUSHF
+def : InstRW<[WriteMicrocoded], (instregex "PUSHF(16|32)")>;
+
+// PUSHA.
+def ZnWritePushA : SchedWriteRes<[ZnAGU]> {
+  let Latency = 8;
+}
+def : InstRW<[ZnWritePushA], (instregex "PUSHA(16|32)")>;
+
+//LAHF
+def : InstRW<[WriteMicrocoded], (instregex "LAHF")>;
+
+// SAHF.
+def ZnWriteSAHF : SchedWriteRes<[ZnALU]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+def : InstRW<[ZnWriteSAHF], (instregex "SAHF")>;
+
+// BSWAP.
+def ZnWriteBSwap : SchedWriteRes<[ZnALU]> {
+  let ResourceCycles = [4];
+}
+def : InstRW<[ZnWriteBSwap], (instregex "BSWAP")>;
+
+// MOVBE.
+// r,m.
+def ZnWriteMOVBE : SchedWriteRes<[ZnALU]> {
+  let ResourceCycles = [2];
+}
+def : InstRW<[WriteALU], (instregex "MOVBE(16|32|64)rm")>;
+
+// m16,r16.
+def : InstRW<[WriteALU], (instregex "MOVBE(16|32|64)mr")>;
+
+//-- Arithmetic instructions --//
+
+// ADD SUB.
+// m,r/i.
+def : InstRW<[WriteALU], (instregex "(ADD|SUB)(8|16|32|64)m(r|i)",
+                          "(ADD|SUB)(8|16|32|64)mi8",
+                          "(ADD|SUB)64mi32")>;
+
+// ADC SBB.
+// r,r/i.
+def : InstRW<[WriteALU], (instregex "(ADC|SBB)(8|16|32|64)r(r|i)",
+                          "(ADC|SBB)(16|32|64)ri8",
+                          "(ADC|SBB)64ri32",
+                          "(ADC|SBB)(8|16|32|64)rr_REV")>;
+
+// r,m.
+def : InstRW<[WriteALU, ReadAfterLd],
+            (instregex "(ADC|SBB)(8|16|32|64)rm")>;
+
+// m,r/i.
+def : InstRW<[WriteALU],
+             (instregex "(ADC|SBB)(8|16|32|64)m(r|i)",
+              "(ADC|SBB)(16|32|64)mi8",
+              "(ADC|SBB)64mi32")>;
+
+// INC DEC NOT NEG.
+// m.
+def : InstRW<[WriteALU],
+             (instregex "(INC|DEC|NOT|NEG)(8|16|32|64)m",
+              "(INC|DEC)64(16|32)m")>;
+
+// MUL IMUL.
+// r16.
+def ZnWriteMul16 : SchedWriteRes<[ZnALU1, ZnMultiplier]> {
+  let Latency = 3;
+}
+def : InstRW<[ZnWriteMul16], (instregex "IMUL16r", "MUL16r")>;
+
+// m16.
+def ZnWriteMul16Ld : SchedWriteRes<[ZnAGU, ZnALU1, ZnMultiplier]> {
+  let Latency = 8;
+}
+def : InstRW<[ZnWriteMul16Ld], (instregex "IMUL16m", "MUL16m")>;
+
+// r32.
+def ZnWriteMul32 : SchedWriteRes<[ZnALU1, ZnMultiplier]> {
+  let Latency = 3;
+}
+def : InstRW<[ZnWriteMul32], (instregex "IMUL32r", "MUL32r")>;
+
+// m32.
+def ZnWriteMul32Ld : SchedWriteRes<[ZnALU1, ZnMultiplier]> {
+  let Latency = 8;
+}
+def : InstRW<[ZnWriteMul32Ld], (instregex "IMUL32m", "MUL32m")>;
+
+// r64.
+def ZnWriteMul64 : SchedWriteRes<[ZnALU1, ZnMultiplier]> {
+  let Latency = 4;
+  let NumMicroOps = 2;
+}
+def : InstRW<[ZnWriteMul64], (instregex "IMUL64r", "MUL64r")>;
+
+// m64.
+def ZnWriteMul64Ld : SchedWriteRes<[ZnAGU, ZnALU1, ZnMultiplier]> {
+  let Latency = 9;
+  let NumMicroOps = 2;
+}
+def : InstRW<[ZnWriteMul64Ld], (instregex "IMUL64m", "MUL64m")>;
+
+// r16,r16.
+def ZnWriteMul16rri : SchedWriteRes<[ZnALU1, ZnMultiplier]> {
+  let Latency = 3;
+}
+def : InstRW<[ZnWriteMul16rri], (instregex "IMUL16rri", "IMUL16rri8")>;
+
+// r16,m16.
+def ZnWriteMul16rmi : SchedWriteRes<[ZnALU1, ZnMultiplier]> {
+  let Latency = 8;
 }
+def : InstRW<[ZnWriteMul16rmi], (instregex "IMUL16rmi", "IMUL16rmi8")>;
+
+// MULX.
+// r32,r32,r32.
+def ZnWriteMulX32 : SchedWriteRes<[ZnALU1, ZnMultiplier]> {
+  let Latency = 3;
+  let ResourceCycles = [1, 2];
+}
+def : InstRW<[ZnWriteMulX32], (instregex "MULX32rr")>;
+
+// r32,r32,m32.
+def ZnWriteMulX32Ld : SchedWriteRes<[ZnAGU, ZnALU1, ZnMultiplier]> {
+  let Latency = 3;
+  let ResourceCycles = [1, 2, 2];
+}
+def : InstRW<[ZnWriteMulX32Ld], (instregex "MULX32rm")>;
+
+// r64,r64,r64.
+def ZnWriteMulX64 : SchedWriteRes<[ZnALU1]> {
+  let Latency = 3;
+}
+def : InstRW<[ZnWriteMulX64], (instregex "MULX64rr")>;
+
+// r64,r64,m64.
+def ZnWriteMulX64Ld : SchedWriteRes<[ZnAGU, ZnALU1, ZnMultiplier]> {
+  let Latency = 8;
+}
+def : InstRW<[ZnWriteMulX64Ld], (instregex "MULX64rm")>;
+
+// DIV, IDIV.
+// r8.
+def ZnWriteDiv8 : SchedWriteRes<[ZnALU2, ZnDivider]> {
+  let Latency = 15;
+}
+def : InstRW<[ZnWriteDiv8], (instregex "DIV8r", "IDIV8r")>;
+
+// r16.
+def ZnWriteDiv16 : SchedWriteRes<[ZnALU2, ZnDivider]> {
+  let Latency = 17;
+  let NumMicroOps = 2;
+}
+def : InstRW<[ZnWriteDiv16], (instregex "DIV16r", "IDIV16r")>;
+
+// r32.
+def ZnWriteDiv32 : SchedWriteRes<[ZnALU2, ZnDivider]> {
+  let Latency = 25;
+  let NumMicroOps = 2;
+}
+def : InstRW<[ZnWriteDiv32], (instregex "DIV32r", "IDIV32r")>;
+
+// r64.
+def ZnWriteDiv64 : SchedWriteRes<[ZnALU2, ZnDivider]> {
+  let Latency = 41;
+  let NumMicroOps = 2;
+}
+def : InstRW<[ZnWriteDiv64], (instregex "DIV64r", "IDIV64r")>;
+
+//-- Control transfer instructions --//
+
+// J(E|R)CXZ.
+def ZnWriteJCXZ : SchedWriteRes<[ZnALU03]>;
+def : InstRW<[ZnWriteJCXZ], (instregex "JCXZ", "JECXZ_(32|64)", "JRCXZ")>;
+
+// INTO
+def : InstRW<[WriteMicrocoded], (instregex "INTO")>;
+
+// LOOP.
+def ZnWriteLOOP : SchedWriteRes<[ZnALU03]>;
+def : InstRW<[ZnWriteLOOP], (instregex "LOOP")>;
+
+// LOOP(N)E, LOOP(N)Z
+def ZnWriteLOOPE : SchedWriteRes<[ZnALU03]>;
+def : InstRW<[ZnWriteLOOPE], (instregex "LOOPE", "LOOPNE",
+                              "LOOPZ", "LOOPNZ")>;
+
+// CALL.
+// r.
+def ZnWriteCALLr : SchedWriteRes<[ZnAGU, ZnALU03]>;
+def : InstRW<[ZnWriteCALLr], (instregex "CALL(16|32)r")>;
+
+def : InstRW<[WriteMicrocoded], (instregex "CALL(16|32)m")>;
+
+// RET.
+def ZnWriteRET : SchedWriteRes<[ZnALU03]> {
+  let NumMicroOps = 2;
+}
+def : InstRW<[ZnWriteRET], (instregex "RET(L|Q|W)", "LRET(L|Q|W)",
+                            "IRET(D|Q)", "RETF")>;
+
+//-- Logic instructions --//
+
+// AND OR XOR.
+// m,r/i.
+def : InstRW<[WriteALU],
+             (instregex "(AND|OR|XOR)(8|16|32|64)m(r|i)",
+              "(AND|OR|XOR)(8|16|32|64)mi8", "(AND|OR|XOR)64mi32")>;
+
+// ANDN.
+// r,r.
+def : InstRW<[WriteALU], (instregex "ANDN(32|64)rr")>;
+// r,m.
+def : InstRW<[WriteALULd], (instregex "ANDN(32|64)rm")>;
+
+// Define ALU latency variants
+def ZnWriteALULat2 : SchedWriteRes<[ZnALU]> {
+  let Latency = 2;
+}
+def ZnWriteALULat2Ld : SchedWriteRes<[ZnAGU, ZnALU]> {
+  let Latency = 6;
+}
+
+def ZnWriteALULat3 : SchedWriteRes<[ZnALU]> {
+  let Latency = 3;
+}
+def ZnWriteALULat3Ld : SchedWriteRes<[ZnAGU, ZnALU]> {
+  let Latency = 7;
+}
+
+// BSF BSR.
+// r,r.
+def : InstRW<[ZnWriteALULat3], (instregex "BS(R|F)(16|32|64)rr")>;
+// r,m.
+def : InstRW<[ZnWriteALULat3Ld], (instregex "BS(R|F)(16|32|64)rm")>;
+
+// BT.
+// r,r/i.
+def : InstRW<[WriteShift], (instregex "BT(16|32|64)r(r|i8)")>;
+def : InstRW<[WriteShift], (instregex "BT(16|32|64)mr")>;
+def : InstRW<[WriteShiftLd], (instregex "BT(16|32|64)mi8")>;
+
+// BTR BTS BTC.
+// r,r,i.
+def ZnWriteBTRSC : SchedWriteRes<[ZnALU]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+def : InstRW<[ZnWriteBTRSC], (instregex "BT(R|S|C)(16|32|64)r(r|i8)")>;
+
+// m,r,i.
+def : InstRW<[ZnWriteBTRSC], (instregex "BT(R|S|C)(16|32|64)m(r|i8)")>;
+
+// BLSI BLSMSK BLSR.
+// r,r.
+def : InstRW<[ZnWriteALULat2], (instregex "BLS(I|MSK|R)(32|64)rr")>;
+// r,m.
+def : InstRW<[ZnWriteALULat2Ld], (instregex "BLS(I|MSK|R)(32|64)rm")>;
+
+// BEXTR.
+// r,r,r.
+def : InstRW<[WriteALU], (instregex "BEXTR(32|64)rr")>;
+// r,m,r.
+def : InstRW<[WriteALULd], (instregex "BEXTR(32|64)rm")>;
+
+// BZHI.
+// r,r,r.
+def : InstRW<[WriteALU], (instregex "BZHI(32|64)rr")>;
+// r,m,r.
+def : InstRW<[WriteALULd], (instregex "BZHI(32|64)rm")>;
+
+// CLD STD.
+def : InstRW<[WriteALU], (instregex "STD", "CLD")>;
+
+// PDEP PEXT.
+// r,r,r.
+def : InstRW<[WriteMicrocoded], (instregex "PDEP(32|64)rr", "PEXT(32|64)rr")>;
+// r,m,r.
+def : InstRW<[WriteMicrocoded], (instregex "PDEP(32|64)rm", "PEXT(32|64)rm")>;
+
+// ROR ROL.
+def : InstRW<[WriteShift], (instregex "RO(R|L)(8|16|32|64)r1")>;
+
+// RCR RCL.
+// r,1.
+def : InstRW<[WriteShift], (instregex "RC(R|L)(8|16|32|64)r1")>;
+
+// m,1.
+def : InstRW<[WriteMicrocoded], (instregex "RC(R|L)(8|16|32|64)m1")>;
+
+// i.
+def : InstRW<[WriteShift], (instregex "RC(R|L)(8|16|32|64)r(i|CL)")>;
+
+// m,i.
+def : InstRW<[WriteMicrocoded], (instregex "RC(R|L)(8|16|32|64)m(i|CL)")>;
+
+// SHR SHL SAR.
+// m,i.
+def : InstRW<[WriteShift], (instregex "S(A|H)(R|L)(8|16|32|64)m(i|1)")>;
+
+// SHRD SHLD.
+def : InstRW<[WriteShift], (instregex "SH(R|L)D(16|32|64)(r|m)ri8")>;
+
+// r,r,cl.
+def : InstRW<[WriteMicrocoded], (instregex "SHLD(16|32|64)rrCL")>;
+
+// r,r,cl.
+def : InstRW<[WriteMicrocoded], (instregex "SHRD(16|32|64)rrCL")>;
+
+// m,r,cl.
+def : InstRW<[WriteMicrocoded], (instregex "SH(R|L)D(16|32|64)mrCL")>;
+
+// SETcc.
+// r.
+def : InstRW<[WriteShift],
+             (instregex "SET(O|NO|B|AE|E|NE|BE|A|S|NS|P|NP|L|GE|LE|G)r")>;
+// m.
+def : InstRW<[WriteShift],
+             (instregex "SET(O|NO|B|AE|E|NE|BE|A|S|NS|P|NP|L|GE|LE|G)m")>;
+
+// LZCNT TZCNT.
+// r,r.
+def : InstRW<[ZnWriteALULat2], (instregex "(L|TZCNT)(16|32|64)rr")>;
+// r,m.
+def : InstRW<[ZnWriteALULat2Ld], (instregex "(L|TZCNT)(16|32|64)rm")>;
+
+//-- Misc instructions --//
+// CMPXCHG.
+def ZnWriteCMPXCHG : SchedWriteRes<[ZnALU]> {
+  let Latency = 3;
+  let NumMicroOps = 5;
+}
+def : InstRW<[ZnWriteCMPXCHG], (instregex "CMPXCHG(8|16|32|64)rm")>;
+
+// CMPXCHG8B.
+def ZnWriteCMPXCHG8B : SchedWriteRes<[ZnAGU, ZnALU, ZnAGU]> {
+  let NumMicroOps = 18;
+}
+def : InstRW<[ZnWriteCMPXCHG8B], (instregex "CMPXCHG8B")>;
+
+def : InstRW<[WriteMicrocoded], (instregex "CMPXCHG16B")>;
+
+// LEAVE
+def ZnWriteLEAVE : SchedWriteRes<[ZnALU, ZnAGU]> {
+  let NumMicroOps = 2;
+}
+def : InstRW<[ZnWriteLEAVE], (instregex "LEAVE")>;
+
+// PAUSE.
+def : InstRW<[WriteMicrocoded], (instregex "PAUSE")>;
+
+// RDTSC.
+def : InstRW<[WriteMicrocoded], (instregex "RDTSC")>;
+
+// RDPMC.
+def : InstRW<[WriteMicrocoded], (instregex "RDPMC")>;
+
+// RDRAND.
+def : InstRW<[WriteMicrocoded], (instregex "RDRAND(16|32|64)r")>;
+
+// XGETBV.
+def : InstRW<[WriteMicrocoded], (instregex "XGETBV")>;
+
+//-- String instructions --//
+// CMPS.
+def : InstRW<[WriteMicrocoded], (instregex "CMPS(B|L|Q|W)")>;
+
+// LODSB/W.
+def : InstRW<[WriteMicrocoded], (instregex "LODS(B|W)")>;
+
+// LODSD/Q.
+def : InstRW<[WriteMicrocoded], (instregex "LODS(L|Q)")>;
+
+// MOVS.
+def : InstRW<[WriteMicrocoded], (instregex "MOVS(B|L|Q|W)")>;
+
+// SCAS.
+def : InstRW<[WriteMicrocoded], (instregex "SCAS(B|W|L|Q)")>;
+
+// STOS
+def : InstRW<[WriteMicrocoded], (instregex "STOS(B|L|Q|W)")>;
+
+// XADD.
+def : InstRW<[WriteMicrocoded], (instregex "XADD(8|16|32|64)rm")>;
+
+} // SchedModel

test/CodeGen/X86/avx-schedule.ll

 ; BTVER2-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    vaddpd (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_addpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vaddpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fadd <4 x double> %a0, %a1
   %2 = load <4 x double>, <4 x double> *%a2, align 32
   %3 = fadd <4 x double> %1, %2
   ret <4 x double> %3
 }
 
 define <8 x float> @test_addps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
 ; GENERIC-LABEL: test_addps:

 ; BTVER2-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    vaddps (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_addps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vaddps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fadd <8 x float> %a0, %a1
   %2 = load <8 x float>, <8 x float> *%a2, align 32
   %3 = fadd <8 x float> %1, %2
   ret <8 x float> %3
 }
 
 define <4 x double> @test_addsubpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
 ; GENERIC-LABEL: test_addsubpd:

 ; BTVER2-NEXT:    vaddsubpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    vaddsubpd (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_addsubpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vaddsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vaddsubpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x double> @llvm.x86.avx.addsub.pd.256(<4 x double> %a0, <4 x double> %a1)
   %2 = load <4 x double>, <4 x double> *%a2, align 32
   %3 = call <4 x double> @llvm.x86.avx.addsub.pd.256(<4 x double> %1, <4 x double> %2)
   ret <4 x double> %3
 }
 declare <4 x double> @llvm.x86.avx.addsub.pd.256(<4 x double>, <4 x double>) nounwind readnone
 
 define <8 x float> @test_addsubps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {

 ; BTVER2-NEXT:    vaddsubps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    vaddsubps (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_addsubps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vaddsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vaddsubps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x float> @llvm.x86.avx.addsub.ps.256(<8 x float> %a0, <8 x float> %a1)
   %2 = load <8 x float>, <8 x float> *%a2, align 32
   %3 = call <8 x float> @llvm.x86.avx.addsub.ps.256(<8 x float> %1, <8 x float> %2)
   ret <8 x float> %3
 }
 declare <8 x float> @llvm.x86.avx.addsub.ps.256(<8 x float>, <8 x float>) nounwind readnone
 
 define <4 x double> @test_andnotpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {

 ; BTVER2-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_andnotpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vandnpd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vandnpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast <4 x double> %a0 to <4 x i64>
   %2 = bitcast <4 x double> %a1 to <4 x i64>
   %3 = xor <4 x i64> %1, <i64 -1, i64 -1, i64 -1, i64 -1>
   %4 = and <4 x i64> %3, %2
   %5 = load <4 x double>, <4 x double> *%a2, align 32
   %6 = bitcast <4 x double> %5 to <4 x i64>
   %7 = xor <4 x i64> %4, <i64 -1, i64 -1, i64 -1, i64 -1>
   %8 = and <4 x i64> %6, %7

 ; BTVER2-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_andnotps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vandnps %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vandnps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast <8 x float> %a0 to <4 x i64>
   %2 = bitcast <8 x float> %a1 to <4 x i64>
   %3 = xor <4 x i64> %1, <i64 -1, i64 -1, i64 -1, i64 -1>
   %4 = and <4 x i64> %3, %2
   %5 = load <8 x float>, <8 x float> *%a2, align 32
   %6 = bitcast <8 x float> %5 to <4 x i64>
   %7 = xor <4 x i64> %4, <i64 -1, i64 -1, i64 -1, i64 -1>
   %8 = and <4 x i64> %6, %7

 ; BTVER2-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_andpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vandpd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vandpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast <4 x double> %a0 to <4 x i64>
   %2 = bitcast <4 x double> %a1 to <4 x i64>
   %3 = and <4 x i64> %1, %2
   %4 = load <4 x double>, <4 x double> *%a2, align 32
   %5 = bitcast <4 x double> %4 to <4 x i64>
   %6 = and <4 x i64> %3, %5
   %7 = bitcast <4 x i64> %6 to <4 x double>
   %8 = fadd <4 x double> %a1, %7

 ; BTVER2-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_andps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vandps %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vandps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast <8 x float> %a0 to <4 x i64>
   %2 = bitcast <8 x float> %a1 to <4 x i64>
   %3 = and <4 x i64> %1, %2
   %4 = load <8 x float>, <8 x float> *%a2, align 32
   %5 = bitcast <8 x float> %4 to <4 x i64>
   %6 = and <4 x i64> %3, %5
   %7 = bitcast <4 x i64> %6 to <8 x float>
   %8 = fadd <8 x float> %a1, %7

 ; BTVER2-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],mem[1,2],ymm0[3] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_blendpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3] sched: [1:0.50]
 ; ZNVER1-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vblendpd {{.*#+}} ymm0 = ymm0[0],mem[1,2],ymm0[3] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x double> %a0, <4 x double> %a1, <4 x i32> <i32 0, i32 5, i32 6, i32 3>
   %2 = load <4 x double>, <4 x double> *%a2, align 32
   %3 = fadd <4 x double> %a1, %1
   %4 = shufflevector <4 x double> %3, <4 x double> %2, <4 x i32> <i32 0, i32 5, i32 6, i32 3>
   ret <4 x double> %4
 }
 
 define <8 x float> @test_blendps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {

 ; BTVER2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3,4,5,6,7] sched: [1:0.50]
 ; BTVER2-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2],ymm0[3],mem[4,5,6],ymm0[7] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_blendps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3,4,5,6,7] sched: [1:0.50]
 ; ZNVER1-NEXT:    vblendps {{.*#+}} ymm0 = ymm0[0,1],mem[2],ymm0[3],mem[4,5,6],ymm0[7] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x float> %a0, <8 x float> %a1, <8 x i32> <i32 0, i32 9, i32 10, i32 3, i32 4, i32 5, i32 6, i32 7>
   %2 = load <8 x float>, <8 x float> *%a2, align 32
   %3 = shufflevector <8 x float> %1, <8 x float> %2, <8 x i32> <i32 0, i32 1, i32 10, i32 3, i32 12, i32 13, i32 14, i32 7>
   ret <8 x float> %3
 }
 
 define <4 x double> @test_blendvpd(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, <4 x double> *%a3) {
 ; GENERIC-LABEL: test_blendvpd:

 ; BTVER2-NEXT:    vblendvpd %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vblendvpd %ymm2, (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_blendvpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vblendvpd %ymm2, %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; ZNVER1-NEXT:    vblendvpd %ymm2, (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x double> @llvm.x86.avx.blendv.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2)
   %2 = load <4 x double>, <4 x double> *%a3, align 32
   %3 = call <4 x double> @llvm.x86.avx.blendv.pd.256(<4 x double> %1, <4 x double> %2, <4 x double> %a2)
   ret <4 x double> %3
 }
 declare <4 x double> @llvm.x86.avx.blendv.pd.256(<4 x double>, <4 x double>, <4 x double>) nounwind readnone
 
 define <8 x float> @test_blendvps(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, <8 x float> *%a3) {

 ; BTVER2-NEXT:    vblendvps %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vblendvps %ymm2, (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_blendvps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vblendvps %ymm2, %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; ZNVER1-NEXT:    vblendvps %ymm2, (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x float> @llvm.x86.avx.blendv.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2)
   %2 = load <8 x float>, <8 x float> *%a3, align 32
   %3 = call <8 x float> @llvm.x86.avx.blendv.ps.256(<8 x float> %1, <8 x float> %2, <8 x float> %a2)
   ret <8 x float> %3
 }
 declare <8 x float> @llvm.x86.avx.blendv.ps.256(<8 x float>, <8 x float>, <8 x float>) nounwind readnone
 
 define <8 x float> @test_broadcastf128(<4 x float> *%a0) {

 ; BTVER2-LABEL: test_broadcastf128:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_broadcastf128:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load <4 x float>, <4 x float> *%a0, align 32
   %2 = shufflevector <4 x float> %1, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
   ret <8 x float> %2
 }
 
 define <4 x double> @test_broadcastsd_ymm(double *%a0) {
 ; GENERIC-LABEL: test_broadcastsd_ymm:
 ; GENERIC:       # BB#0:

 ; BTVER2-LABEL: test_broadcastsd_ymm:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vbroadcastsd (%rdi), %ymm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_broadcastsd_ymm:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vbroadcastsd (%rdi), %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load double, double *%a0, align 8
   %2 = insertelement <4 x double> undef, double %1, i32 0
   %3 = shufflevector <4 x double> %2, <4 x double> undef, <4 x i32> zeroinitializer
   ret <4 x double> %3
 }
 
 define <4 x float> @test_broadcastss(float *%a0) {
 ; GENERIC-LABEL: test_broadcastss:

 ; BTVER2-LABEL: test_broadcastss:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vbroadcastss (%rdi), %xmm0 # sched: [5:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_broadcastss:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vbroadcastss (%rdi), %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load float, float *%a0, align 4
   %2 = insertelement <4 x float> undef, float %1, i32 0
   %3 = shufflevector <4 x float> %2, <4 x float> undef, <4 x i32> zeroinitializer
   ret <4 x float> %3
 }
 
 define <8 x float> @test_broadcastss_ymm(float *%a0) {
 ; GENERIC-LABEL: test_broadcastss_ymm:

 ; BTVER2-LABEL: test_broadcastss_ymm:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vbroadcastss (%rdi), %ymm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_broadcastss_ymm:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vbroadcastss (%rdi), %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load float, float *%a0, align 4
   %2 = insertelement <8 x float> undef, float %1, i32 0
   %3 = shufflevector <8 x float> %2, <8 x float> undef, <8 x i32> zeroinitializer
   ret <8 x float> %3
 }
 
 define <4 x double> @test_cmppd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
 ; GENERIC-LABEL: test_cmppd:

 ; BTVER2-NEXT:    vorpd %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cmppd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcmpeqpd %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vcmpeqpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
 ; ZNVER1-NEXT:    vorpd %ymm0, %ymm1, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fcmp oeq <4 x double> %a0, %a1
   %2 = load <4 x double>, <4 x double> *%a2, align 32
   %3 = fcmp oeq <4 x double> %a0, %2
   %4 = sext <4 x i1> %1 to <4 x i64>
   %5 = sext <4 x i1> %3 to <4 x i64>
   %6 = or <4 x i64> %4, %5
   %7 = bitcast <4 x i64> %6 to <4 x double>
   ret <4 x double> %7

 ; BTVER2-NEXT:    vorps %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cmpps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcmpeqps %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vcmpeqps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
 ; ZNVER1-NEXT:    vorps %ymm0, %ymm1, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fcmp oeq <8 x float> %a0, %a1
   %2 = load <8 x float>, <8 x float> *%a2, align 32
   %3 = fcmp oeq <8 x float> %a0, %2
   %4 = sext <8 x i1> %1 to <8 x i32>
   %5 = sext <8 x i1> %3 to <8 x i32>
   %6 = or <8 x i32> %4, %5
   %7 = bitcast <8 x i32> %6 to <8 x float>
   ret <8 x float> %7

 ; BTVER2-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtdq2pd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtdq2pd (%rdi), %ymm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvtdq2pd %xmm0, %ymm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = sitofp <4 x i32> %a0 to <4 x double>
   %2 = load <4 x i32>, <4 x i32> *%a1, align 16
   %3 = sitofp <4 x i32> %2 to <4 x double>
   %4 = fadd <4 x double> %1, %3
   ret <4 x double> %4
 }
 
 define <8 x float> @test_cvtdq2ps(<8 x i32> %a0, <8 x i32> *%a1) {

 ; BTVER2-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtdq2ps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtdq2ps (%rdi), %ymm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvtdq2ps %ymm0, %ymm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = sitofp <8 x i32> %a0 to <8 x float>
   %2 = load <8 x i32>, <8 x i32> *%a1, align 16
   %3 = sitofp <8 x i32> %2 to <8 x float>
   %4 = fadd <8 x float> %1, %3
   ret <8 x float> %4
 }
 
 define <8 x i32> @test_cvtpd2dq(<4 x double> %a0, <4 x double> *%a1) {

 ; BTVER2-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtpd2dq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvttpd2dqy (%rdi), %xmm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvttpd2dq %ymm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fptosi <4 x double> %a0 to <4 x i32>
   %2 = load <4 x double>, <4 x double> *%a1, align 32
   %3 = fptosi <4 x double> %2 to <4 x i32>
   %4 = shufflevector <4 x i32> %1, <4 x i32> %3, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
   ret <8 x i32> %4
 }
 
 define <8 x float> @test_cvtpd2ps(<4 x double> %a0, <4 x double> *%a1) {

 ; BTVER2-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtpd2ps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtpd2psy (%rdi), %xmm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvtpd2ps %ymm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fptrunc <4 x double> %a0 to <4 x float>
   %2 = load <4 x double>, <4 x double> *%a1, align 32
   %3 = fptrunc <4 x double> %2 to <4 x float>
   %4 = shufflevector <4 x float> %1, <4 x float> %3, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
   ret <8 x float> %4
 }
 
 define <8 x i32> @test_cvtps2dq(<8 x float> %a0, <8 x float> *%a1) {

 ; BTVER2-NEXT:    vorps %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtps2dq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvttps2dq (%rdi), %ymm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvttps2dq %ymm0, %ymm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vorps %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fptosi <8 x float> %a0 to <8 x i32>
   %2 = load <8 x float>, <8 x float> *%a1, align 32
   %3 = fptosi <8 x float> %2 to <8 x i32>
   %4 = or <8 x i32> %1, %3
   ret <8 x i32> %4
 }
 
 define <4 x double> @test_divpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {

 ; BTVER2-NEXT:    vdivpd %ymm1, %ymm0, %ymm0 # sched: [38:38.00]
 ; BTVER2-NEXT:    vdivpd (%rdi), %ymm0, %ymm0 # sched: [43:38.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_divpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vdivpd %ymm1, %ymm0, %ymm0 # sched: [15:1.00]
 ; ZNVER1-NEXT:    vdivpd (%rdi), %ymm0, %ymm0 # sched: [22:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fdiv <4 x double> %a0, %a1
   %2 = load <4 x double>, <4 x double> *%a2, align 32
   %3 = fdiv <4 x double> %1, %2
   ret <4 x double> %3
 }
 
 define <8 x float> @test_divps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
 ; GENERIC-LABEL: test_divps:

 ; BTVER2-NEXT:    vdivps %ymm1, %ymm0, %ymm0 # sched: [38:38.00]
 ; BTVER2-NEXT:    vdivps (%rdi), %ymm0, %ymm0 # sched: [43:38.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_divps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vdivps %ymm1, %ymm0, %ymm0 # sched: [15:1.00]
 ; ZNVER1-NEXT:    vdivps (%rdi), %ymm0, %ymm0 # sched: [22:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fdiv <8 x float> %a0, %a1
   %2 = load <8 x float>, <8 x float> *%a2, align 32
   %3 = fdiv <8 x float> %1, %2
   ret <8 x float> %3
 }
 
 define <8 x float> @test_dpps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
 ; GENERIC-LABEL: test_dpps:

 ; BTVER2-NEXT:    vdpps $7, %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vdpps $7, (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_dpps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vdpps $7, %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vdpps $7, (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x float> @llvm.x86.avx.dp.ps.256(<8 x float> %a0, <8 x float> %a1, i8 7)
   %2 = load <8 x float>, <8 x float> *%a2, align 32
   %3 = call <8 x float> @llvm.x86.avx.dp.ps.256(<8 x float> %1, <8 x float> %2, i8 7)
   ret <8 x float> %3
 }
 declare <8 x float> @llvm.x86.avx.dp.ps.256(<8 x float>, <8 x float>, i8) nounwind readnone
 
 define <4 x float> @test_extractf128(<8 x float> %a0, <8 x float> %a1, <4 x float> *%a2) {

 ; BTVER2-NEXT:    vextractf128 $1, %ymm1, (%rdi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_extractf128:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vextractf128 $1, %ymm0, %xmm0 # sched: [1:0.50]
 ; ZNVER1-NEXT:    vextractf128 $1, %ymm1, (%rdi) # sched: [1:0.50]
 ; ZNVER1-NEXT:    vzeroupper
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x float> %a0, <8 x float> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
   %2 = shufflevector <8 x float> %a1, <8 x float> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
   store <4 x float> %2, <4 x float> *%a2
   ret <4 x float> %1
 }
 
 define <4 x double> @test_haddpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
 ; GENERIC-LABEL: test_haddpd:

 ; BTVER2-NEXT:    vhaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    vhaddpd (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_haddpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vhaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vhaddpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x double> @llvm.x86.avx.hadd.pd.256(<4 x double> %a0, <4 x double> %a1)
   %2 = load <4 x double>, <4 x double> *%a2, align 32
   %3 = call <4 x double> @llvm.x86.avx.hadd.pd.256(<4 x double> %1, <4 x double> %2)
   ret <4 x double> %3
 }
 declare <4 x double> @llvm.x86.avx.hadd.pd.256(<4 x double>, <4 x double>) nounwind readnone
 
 define <8 x float> @test_haddps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {

 ; BTVER2-NEXT:    vhaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    vhaddps (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_haddps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vhaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vhaddps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float> %a0, <8 x float> %a1)
   %2 = load <8 x float>, <8 x float> *%a2, align 32
   %3 = call <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float> %1, <8 x float> %2)
   ret <8 x float> %3
 }
 declare <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float>, <8 x float>) nounwind readnone
 
 define <4 x double> @test_hsubpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {

 ; BTVER2-NEXT:    vhsubpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    vhsubpd (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_hsubpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vhsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vhsubpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x double> @llvm.x86.avx.hsub.pd.256(<4 x double> %a0, <4 x double> %a1)
   %2 = load <4 x double>, <4 x double> *%a2, align 32
   %3 = call <4 x double> @llvm.x86.avx.hsub.pd.256(<4 x double> %1, <4 x double> %2)
   ret <4 x double> %3
 }
 declare <4 x double> @llvm.x86.avx.hsub.pd.256(<4 x double>, <4 x double>) nounwind readnone
 
 define <8 x float> @test_hsubps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {

 ; BTVER2-NEXT:    vhsubps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    vhsubps (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_hsubps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vhsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vhsubps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x float> @llvm.x86.avx.hsub.ps.256(<8 x float> %a0, <8 x float> %a1)
   %2 = load <8 x float>, <8 x float> *%a2, align 32
   %3 = call <8 x float> @llvm.x86.avx.hsub.ps.256(<8 x float> %1, <8 x float> %2)
   ret <8 x float> %3
 }
 declare <8 x float> @llvm.x86.avx.hsub.ps.256(<8 x float>, <8 x float>) nounwind readnone
 
 define <8 x float> @test_insertf128(<8 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {

 ; BTVER2-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_insertf128:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm1 # sched: [1:0.50]
 ; ZNVER1-NEXT:    vinsertf128 $1, (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x float> %a1, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
   %2 = shufflevector <8 x float> %a0, <8 x float> %1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
   %3 = load <4 x float>, <4 x float> *%a2, align 16
   %4 = shufflevector <4 x float> %3, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
   %5 = shufflevector <8 x float> %a0, <8 x float> %4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
   %6 = fadd <8 x float> %2, %5
   ret <8 x float> %6
 }

 ; BTVER2-LABEL: test_lddqu:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vlddqu (%rdi), %ymm0 # sched: [5:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lddqu:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vlddqu (%rdi), %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <32 x i8> @llvm.x86.avx.ldu.dq.256(i8* %a0)
   ret <32 x i8> %1
 }
 declare <32 x i8> @llvm.x86.avx.ldu.dq.256(i8*) nounwind readonly
 
 define <2 x double> @test_maskmovpd(i8* %a0, <2 x i64> %a1, <2 x double> %a2) {
 ; GENERIC-LABEL: test_maskmovpd:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    vmovapd %xmm2, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_maskmovpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmaskmovpd (%rdi), %xmm0, %xmm2
 ; ZNVER1-NEXT:    vmaskmovpd %xmm1, %xmm0, (%rdi)
 ; ZNVER1-NEXT:    vmovapd %xmm2, %xmm0 # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x double> @llvm.x86.avx.maskload.pd(i8* %a0, <2 x i64> %a1)
   call void @llvm.x86.avx.maskstore.pd(i8* %a0, <2 x i64> %a1, <2 x double> %a2)
   ret <2 x double> %1
 }
 declare <2 x double> @llvm.x86.avx.maskload.pd(i8*, <2 x i64>) nounwind readonly
 declare void @llvm.x86.avx.maskstore.pd(i8*, <2 x i64>, <2 x double>) nounwind
 
 define <4 x double> @test_maskmovpd_ymm(i8* %a0, <4 x i64> %a1, <4 x double> %a2) {

 ; BTVER2-NEXT:    vmovapd %ymm2, %ymm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_maskmovpd_ymm:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmaskmovpd (%rdi), %ymm0, %ymm2
 ; ZNVER1-NEXT:    vmaskmovpd %ymm1, %ymm0, (%rdi)
 ; ZNVER1-NEXT:    vmovapd %ymm2, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x double> @llvm.x86.avx.maskload.pd.256(i8* %a0, <4 x i64> %a1)
   call void @llvm.x86.avx.maskstore.pd.256(i8* %a0, <4 x i64> %a1, <4 x double> %a2)
   ret <4 x double> %1
 }
 declare <4 x double> @llvm.x86.avx.maskload.pd.256(i8*, <4 x i64>) nounwind readonly
 declare void @llvm.x86.avx.maskstore.pd.256(i8*, <4 x i64>, <4 x double>) nounwind
 
 define <4 x float> @test_maskmovps(i8* %a0, <4 x i32> %a1, <4 x float> %a2) {

 ; BTVER2-NEXT:    vmovaps %xmm2, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_maskmovps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmaskmovps (%rdi), %xmm0, %xmm2
 ; ZNVER1-NEXT:    vmaskmovps %xmm1, %xmm0, (%rdi)
 ; ZNVER1-NEXT:    vmovaps %xmm2, %xmm0 # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.avx.maskload.ps(i8* %a0, <4 x i32> %a1)
   call void @llvm.x86.avx.maskstore.ps(i8* %a0, <4 x i32> %a1, <4 x float> %a2)
   ret <4 x float> %1
 }
 declare <4 x float> @llvm.x86.avx.maskload.ps(i8*, <4 x i32>) nounwind readonly
 declare void @llvm.x86.avx.maskstore.ps(i8*, <4 x i32>, <4 x float>) nounwind
 
 define <8 x float> @test_maskmovps_ymm(i8* %a0, <8 x i32> %a1, <8 x float> %a2) {

 ; BTVER2-NEXT:    vmovaps %ymm2, %ymm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_maskmovps_ymm:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmaskmovps (%rdi), %ymm0, %ymm2
 ; ZNVER1-NEXT:    vmaskmovps %ymm1, %ymm0, (%rdi)
 ; ZNVER1-NEXT:    vmovaps %ymm2, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x float> @llvm.x86.avx.maskload.ps.256(i8* %a0, <8 x i32> %a1)
   call void @llvm.x86.avx.maskstore.ps.256(i8* %a0, <8 x i32> %a1, <8 x float> %a2)
   ret <8 x float> %1
 }
 declare <8 x float> @llvm.x86.avx.maskload.ps.256(i8*, <8 x i32>) nounwind readonly
 declare void @llvm.x86.avx.maskstore.ps.256(i8*, <8 x i32>, <8 x float>) nounwind
 
 define <4 x double> @test_maxpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {

 ; BTVER2-NEXT:    vmaxpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vmaxpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_maxpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmaxpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmaxpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x double> @llvm.x86.avx.max.pd.256(<4 x double> %a0, <4 x double> %a1)
   %2 = load <4 x double>, <4 x double> *%a2, align 32
   %3 = call <4 x double> @llvm.x86.avx.max.pd.256(<4 x double> %1, <4 x double> %2)
   ret <4 x double> %3
 }
 declare <4 x double> @llvm.x86.avx.max.pd.256(<4 x double>, <4 x double>) nounwind readnone
 
 define <8 x float> @test_maxps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {

 ; BTVER2-NEXT:    vmaxps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vmaxps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_maxps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmaxps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmaxps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x float> @llvm.x86.avx.max.ps.256(<8 x float> %a0, <8 x float> %a1)
   %2 = load <8 x float>, <8 x float> *%a2, align 32
   %3 = call <8 x float> @llvm.x86.avx.max.ps.256(<8 x float> %1, <8 x float> %2)
   ret <8 x float> %3
 }
 declare <8 x float> @llvm.x86.avx.max.ps.256(<8 x float>, <8 x float>) nounwind readnone
 
 define <4 x double> @test_minpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {

 ; BTVER2-NEXT:    vminpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vminpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_minpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vminpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vminpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x double> @llvm.x86.avx.min.pd.256(<4 x double> %a0, <4 x double> %a1)
   %2 = load <4 x double>, <4 x double> *%a2, align 32
   %3 = call <4 x double> @llvm.x86.avx.min.pd.256(<4 x double> %1, <4 x double> %2)
   ret <4 x double> %3
 }
 declare <4 x double> @llvm.x86.avx.min.pd.256(<4 x double>, <4 x double>) nounwind readnone
 
 define <8 x float> @test_minps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {

 ; BTVER2-NEXT:    vminps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vminps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_minps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vminps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vminps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x float> @llvm.x86.avx.min.ps.256(<8 x float> %a0, <8 x float> %a1)
   %2 = load <8 x float>, <8 x float> *%a2, align 32
   %3 = call <8 x float> @llvm.x86.avx.min.ps.256(<8 x float> %1, <8 x float> %2)
   ret <8 x float> %3
 }
 declare <8 x float> @llvm.x86.avx.min.ps.256(<8 x float>, <8 x float>) nounwind readnone
 
 define <4 x double> @test_movapd(<4 x double> *%a0, <4 x double> *%a1) {

 ; BTVER2-NEXT:    vmovapd %ymm0, (%rsi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movapd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovapd (%rdi), %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmovapd %ymm0, (%rsi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load <4 x double>, <4 x double> *%a0, align 32
   %2 = fadd <4 x double> %1, %1
   store <4 x double> %2, <4 x double> *%a1, align 32
   ret <4 x double> %2
 }
 
 define <8 x float> @test_movaps(<8 x float> *%a0, <8 x float> *%a1) {
 ; GENERIC-LABEL: test_movaps:

 ; BTVER2-NEXT:    vmovaps %ymm0, (%rsi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movaps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovaps (%rdi), %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmovaps %ymm0, (%rsi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load <8 x float>, <8 x float> *%a0, align 32
   %2 = fadd <8 x float> %1, %1
   store <8 x float> %2, <8 x float> *%a1, align 32
   ret <8 x float> %2
 }
 
 define <4 x double> @test_movddup(<4 x double> %a0, <4 x double> *%a1) {
 ; GENERIC-LABEL: test_movddup:

 ; BTVER2-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movddup:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovddup {{.*#+}} ymm1 = mem[0,0,2,2] sched: [8:0.50]
 ; ZNVER1-NEXT:    vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2] sched: [1:0.50]
 ; ZNVER1-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x double> %a0, <4 x double> undef, <4 x i32> <i32 0, i32 0, i32 2, i32 2>
   %2 = load <4 x double>, <4 x double> *%a1, align 32
   %3 = shufflevector <4 x double> %2, <4 x double> undef, <4 x i32> <i32 0, i32 0, i32 2, i32 2>
   %4 = fadd <4 x double> %1, %3
   ret <4 x double> %4
 }
 
 define i32 @test_movmskpd(<4 x double> %a0) {

 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vmovmskpd %ymm0, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movmskpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovmskpd %ymm0, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    vzeroupper
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.avx.movmsk.pd.256(<4 x double> %a0)
   ret i32 %1
 }
 declare i32 @llvm.x86.avx.movmsk.pd.256(<4 x double>) nounwind readnone
 
 define i32 @test_movmskps(<8 x float> %a0) {
 ; GENERIC-LABEL: test_movmskps:
 ; GENERIC:       # BB#0:

 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vmovmskps %ymm0, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movmskps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovmskps %ymm0, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    vzeroupper
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.avx.movmsk.ps.256(<8 x float> %a0)
   ret i32 %1
 }
 declare i32 @llvm.x86.avx.movmsk.ps.256(<8 x float>) nounwind readnone
 
 define <4 x double> @test_movntpd(<4 x double> %a0, <4 x double> *%a1) {
 ; GENERIC-LABEL: test_movntpd:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    vmovntpd %ymm0, (%rdi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movntpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmovntpd %ymm0, (%rdi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fadd <4 x double> %a0, %a0
   store <4 x double> %1, <4 x double> *%a1, align 32, !nontemporal !0
   ret <4 x double> %1
 }
 
 define <8 x float> @test_movntps(<8 x float> %a0, <8 x float> *%a1) {
 ; GENERIC-LABEL: test_movntps:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    vaddps %ymm0, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    vmovntps %ymm0, (%rdi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movntps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmovntps %ymm0, (%rdi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fadd <8 x float> %a0, %a0
   store <8 x float> %1, <8 x float> *%a1, align 32, !nontemporal !0
   ret <8 x float> %1
 }
 
 define <8 x float> @test_movshdup(<8 x float> %a0, <8 x float> *%a1) {
 ; GENERIC-LABEL: test_movshdup:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movshdup:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovshdup {{.*#+}} ymm1 = mem[1,1,3,3,5,5,7,7] sched: [8:0.50]
 ; ZNVER1-NEXT:    vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7] sched: [1:0.50]
 ; ZNVER1-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x float> %a0, <8 x float> undef, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 5, i32 5, i32 7, i32 7>
   %2 = load <8 x float>, <8 x float> *%a1, align 32
   %3 = shufflevector <8 x float> %2, <8 x float> undef, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 5, i32 5, i32 7, i32 7>
   %4 = fadd <8 x float> %1, %3
   ret <8 x float> %4
 }
 
 define <8 x float> @test_movsldup(<8 x float> %a0, <8 x float> *%a1) {

 ; BTVER2-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movsldup:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovsldup {{.*#+}} ymm1 = mem[0,0,2,2,4,4,6,6] sched: [8:0.50]
 ; ZNVER1-NEXT:    vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6] sched: [1:0.50]
 ; ZNVER1-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x float> %a0, <8 x float> undef, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6>
   %2 = load <8 x float>, <8 x float> *%a1, align 32
   %3 = shufflevector <8 x float> %2, <8 x float> undef, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6>
   %4 = fadd <8 x float> %1, %3
   ret <8 x float> %4
 }
 
 define <4 x double> @test_movupd(<4 x double> *%a0, <4 x double> *%a1) {

 ; BTVER2-NEXT:    vmovupd %ymm0, (%rsi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movupd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovupd (%rdi), %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmovupd %ymm0, (%rsi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load <4 x double>, <4 x double> *%a0, align 1
   %2 = fadd <4 x double> %1, %1
   store <4 x double> %2, <4 x double> *%a1, align 1
   ret <4 x double> %2
 }
 
 define <8 x float> @test_movups(<8 x float> *%a0, <8 x float> *%a1) {
 ; GENERIC-LABEL: test_movups:

 ; BTVER2-NEXT:    vmovups %ymm0, (%rsi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movups:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovups (%rdi), %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmovups %ymm0, (%rsi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load <8 x float>, <8 x float> *%a0, align 1
   %2 = fadd <8 x float> %1, %1
   store <8 x float> %2, <8 x float> *%a1, align 1
   ret <8 x float> %2
 }
 
 define <4 x double> @test_mulpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
 ; GENERIC-LABEL: test_mulpd:

 ; BTVER2-NEXT:    vmulpd %ymm1, %ymm0, %ymm0 # sched: [4:4.00]
 ; BTVER2-NEXT:    vmulpd (%rdi), %ymm0, %ymm0 # sched: [9:4.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_mulpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmulpd %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vmulpd (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fmul <4 x double> %a0, %a1
   %2 = load <4 x double>, <4 x double> *%a2, align 32
   %3 = fmul <4 x double> %1, %2
   ret <4 x double> %3
 }
 
 define <8 x float> @test_mulps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
 ; GENERIC-LABEL: test_mulps:

 ; BTVER2-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
 ; BTVER2-NEXT:    vmulps (%rdi), %ymm0, %ymm0 # sched: [7:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_mulps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vmulps (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fmul <8 x float> %a0, %a1
   %2 = load <8 x float>, <8 x float> *%a2, align 32
   %3 = fmul <8 x float> %1, %2
   ret <8 x float> %3
 }
 
 define <4 x double> @orpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
 ; GENERIC-LABEL: orpd:

 ; BTVER2-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: orpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vorpd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vorpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast <4 x double> %a0 to <4 x i64>
   %2 = bitcast <4 x double> %a1 to <4 x i64>
   %3 = or <4 x i64> %1, %2
   %4 = load <4 x double>, <4 x double> *%a2, align 32
   %5 = bitcast <4 x double> %4 to <4 x i64>
   %6 = or <4 x i64> %3, %5
   %7 = bitcast <4 x i64> %6 to <4 x double>
   %8 = fadd <4 x double> %a1, %7

 ; BTVER2-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_orps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vorps %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vorps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast <8 x float> %a0 to <4 x i64>
   %2 = bitcast <8 x float> %a1 to <4 x i64>
   %3 = or <4 x i64> %1, %2
   %4 = load <8 x float>, <8 x float> *%a2, align 32
   %5 = bitcast <8 x float> %4 to <4 x i64>
   %6 = or <4 x i64> %3, %5
   %7 = bitcast <4 x i64> %6 to <8 x float>
   %8 = fadd <8 x float> %a1, %7

 ; BTVER2-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_permilpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpermilpd {{.*#+}} xmm1 = mem[1,0] sched: [8:0.50]
 ; ZNVER1-NEXT:    vpermilpd {{.*#+}} xmm0 = xmm0[1,0] sched: [1:0.50]
 ; ZNVER1-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <2 x double> %a0, <2 x double> undef, <2 x i32> <i32 1, i32 0>
   %2 = load <2 x double>, <2 x double> *%a1, align 16
   %3 = shufflevector <2 x double> %2, <2 x double> undef, <2 x i32> <i32 1, i32 0>
   %4 = fadd <2 x double> %1, %3
   ret <2 x double> %4
 }
 
 define <4 x double> @test_permilpd_ymm(<4 x double> %a0, <4 x double> *%a1) {

 ; BTVER2-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_permilpd_ymm:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpermilpd {{.*#+}} ymm1 = mem[1,0,2,3] sched: [8:0.50]
 ; ZNVER1-NEXT:    vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3] sched: [1:0.50]
 ; ZNVER1-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x double> %a0, <4 x double> undef, <4 x i32> <i32 1, i32 0, i32 2, i32 3>
   %2 = load <4 x double>, <4 x double> *%a1, align 32
   %3 = shufflevector <4 x double> %2, <4 x double> undef, <4 x i32> <i32 1, i32 0, i32 2, i32 3>
   %4 = fadd <4 x double> %1, %3
   ret <4 x double> %4
 }
 
 define <4 x float> @test_permilps(<4 x float> %a0, <4 x float> *%a1) {

 ; BTVER2-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_permilps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpermilps {{.*#+}} xmm1 = mem[3,2,1,0] sched: [8:0.50]
 ; ZNVER1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0] sched: [1:0.50]
 ; ZNVER1-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x float> %a0, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   %2 = load <4 x float>, <4 x float> *%a1, align 16
   %3 = shufflevector <4 x float> %2, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   %4 = fadd <4 x float> %1, %3
   ret <4 x float> %4
 }
 
 define <8 x float> @test_permilps_ymm(<8 x float> %a0, <8 x float> *%a1) {

 ; BTVER2-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_permilps_ymm:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpermilps {{.*#+}} ymm1 = mem[3,2,1,0,7,6,5,4] sched: [8:0.50]
 ; ZNVER1-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:0.50]
 ; ZNVER1-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x float> %a0, <8 x float> undef, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
   %2 = load <8 x float>, <8 x float> *%a1, align 32
   %3 = shufflevector <8 x float> %2, <8 x float> undef, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
   %4 = fadd <8 x float> %1, %3
   ret <8 x float> %4
 }
 
 define <2 x double> @test_permilvarpd(<2 x double> %a0, <2 x i64> %a1, <2 x i64> *%a2) {

 ; BTVER2-NEXT:    vpermilpd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpermilpd (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_permilvarpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpermilpd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; ZNVER1-NEXT:    vpermilpd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x double> @llvm.x86.avx.vpermilvar.pd(<2 x double> %a0, <2 x i64> %a1)
   %2 = load <2 x i64>, <2 x i64> *%a2, align 16
   %3 = call <2 x double> @llvm.x86.avx.vpermilvar.pd(<2 x double> %1, <2 x i64> %2)
   ret <2 x double> %3
 }
 declare <2 x double> @llvm.x86.avx.vpermilvar.pd(<2 x double>, <2 x i64>) nounwind readnone
 
 define <4 x double> @test_permilvarpd_ymm(<4 x double> %a0, <4 x i64> %a1, <4 x i64> *%a2) {

 ; BTVER2-NEXT:    vpermilpd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpermilpd (%rdi), %ymm0, %ymm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_permilvarpd_ymm:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpermilpd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; ZNVER1-NEXT:    vpermilpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x double> @llvm.x86.avx.vpermilvar.pd.256(<4 x double> %a0, <4 x i64> %a1)
   %2 = load <4 x i64>, <4 x i64> *%a2, align 32
   %3 = call <4 x double> @llvm.x86.avx.vpermilvar.pd.256(<4 x double> %1, <4 x i64> %2)
   ret <4 x double> %3
 }
 declare <4 x double> @llvm.x86.avx.vpermilvar.pd.256(<4 x double>, <4 x i64>) nounwind readnone
 
 define <4 x float> @test_permilvarps(<4 x float> %a0, <4 x i32> %a1, <4 x i32> *%a2) {

 ; BTVER2-NEXT:    vpermilps %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpermilps (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_permilvarps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpermilps %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; ZNVER1-NEXT:    vpermilps (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.avx.vpermilvar.ps(<4 x float> %a0, <4 x i32> %a1)
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = call <4 x float> @llvm.x86.avx.vpermilvar.ps(<4 x float> %1, <4 x i32> %2)
   ret <4 x float> %3
 }
 declare <4 x float> @llvm.x86.avx.vpermilvar.ps(<4 x float>, <4 x i32>) nounwind readnone
 
 define <8 x float> @test_permilvarps_ymm(<8 x float> %a0, <8 x i32> %a1, <8 x i32> *%a2) {

 ; BTVER2-NEXT:    vpermilps %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpermilps (%rdi), %ymm0, %ymm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_permilvarps_ymm:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpermilps %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; ZNVER1-NEXT:    vpermilps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x float> @llvm.x86.avx.vpermilvar.ps.256(<8 x float> %a0, <8 x i32> %a1)
   %2 = load <8 x i32>, <8 x i32> *%a2, align 32
   %3 = call <8 x float> @llvm.x86.avx.vpermilvar.ps.256(<8 x float> %1, <8 x i32> %2)
   ret <8 x float> %3
 }
 declare <8 x float> @llvm.x86.avx.vpermilvar.ps.256(<8 x float>, <8 x i32>) nounwind readnone
 
 define <8 x float> @test_rcpps(<8 x float> %a0, <8 x float> *%a1) {

 ; BTVER2-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_rcpps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vrcpps (%rdi), %ymm1 # sched: [12:0.50]
 ; ZNVER1-NEXT:    vrcpps %ymm0, %ymm0 # sched: [5:0.50]
 ; ZNVER1-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x float> @llvm.x86.avx.rcp.ps.256(<8 x float> %a0)
   %2 = load <8 x float>, <8 x float> *%a1, align 32
   %3 = call <8 x float> @llvm.x86.avx.rcp.ps.256(<8 x float> %2)
   %4 = fadd <8 x float> %1, %3
   ret <8 x float> %4
 }
 declare <8 x float> @llvm.x86.avx.rcp.ps.256(<8 x float>) nounwind readnone
 

 ; BTVER2-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_roundpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vroundpd $7, (%rdi), %ymm1 # sched: [10:1.00]
 ; ZNVER1-NEXT:    vroundpd $7, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x double> @llvm.x86.avx.round.pd.256(<4 x double> %a0, i32 7)
   %2 = load <4 x double>, <4 x double> *%a1, align 32
   %3 = call <4 x double> @llvm.x86.avx.round.pd.256(<4 x double> %2, i32 7)
   %4 = fadd <4 x double> %1, %3
   ret <4 x double> %4
 }
 declare <4 x double> @llvm.x86.avx.round.pd.256(<4 x double>, i32) nounwind readnone
 

 ; BTVER2-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_roundps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vroundps $7, (%rdi), %ymm1 # sched: [10:1.00]
 ; ZNVER1-NEXT:    vroundps $7, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x float> @llvm.x86.avx.round.ps.256(<8 x float> %a0, i32 7)
   %2 = load <8 x float>, <8 x float> *%a1, align 32
   %3 = call <8 x float> @llvm.x86.avx.round.ps.256(<8 x float> %2, i32 7)
   %4 = fadd <8 x float> %1, %3
   ret <8 x float> %4
 }
 declare <8 x float> @llvm.x86.avx.round.ps.256(<8 x float>, i32) nounwind readnone
 

 ; BTVER2-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_rsqrtps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vrsqrtps (%rdi), %ymm1 # sched: [12:0.50]
 ; ZNVER1-NEXT:    vrsqrtps %ymm0, %ymm0 # sched: [5:0.50]
 ; ZNVER1-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x float> @llvm.x86.avx.rsqrt.ps.256(<8 x float> %a0)
   %2 = load <8 x float>, <8 x float> *%a1, align 32
   %3 = call <8 x float> @llvm.x86.avx.rsqrt.ps.256(<8 x float> %2)
   %4 = fadd <8 x float> %1, %3
   ret <8 x float> %4
 }
 declare <8 x float> @llvm.x86.avx.rsqrt.ps.256(<8 x float>) nounwind readnone
 

 ; BTVER2-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_shufpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[2],ymm1[3] sched: [1:0.50]
 ; ZNVER1-NEXT:    vshufpd {{.*#+}} ymm1 = ymm1[1],mem[0],ymm1[2],mem[3] sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x double> %a0, <4 x double> %a1, <4 x i32> <i32 1, i32 4, i32 2, i32 7>
   %2 = load <4 x double>, <4 x double> *%a2, align 32
   %3 = shufflevector <4 x double> %a1, <4 x double> %2, <4 x i32> <i32 1, i32 4, i32 2, i32 7>
   %4 = fadd <4 x double> %1, %3
   ret <4 x double> %4
 }
 
 define <8 x float> @test_shufps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) nounwind {

 ; BTVER2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4] sched: [1:0.50]
 ; BTVER2-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,3],mem[0,0],ymm0[4,7],mem[4,4] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_shufps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4] sched: [1:0.50]
 ; ZNVER1-NEXT:    vshufps {{.*#+}} ymm0 = ymm0[0,3],mem[0,0],ymm0[4,7],mem[4,4] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x float> %a0, <8 x float> %a1, <8 x i32> <i32 0, i32 0, i32 8, i32 8, i32 4, i32 4, i32 12, i32 12>
   %2 = load <8 x float>, <8 x float> *%a2, align 32
   %3 = shufflevector <8 x float> %1, <8 x float> %2, <8 x i32> <i32 0, i32 3, i32 8, i32 8, i32 4, i32 7, i32 12, i32 12>
   ret <8 x float> %3
 }
 
 define <4 x double> @test_sqrtpd(<4 x double> %a0, <4 x double> *%a1) {
 ; GENERIC-LABEL: test_sqrtpd:

 ; BTVER2-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_sqrtpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vsqrtpd (%rdi), %ymm1 # sched: [27:1.00]
 ; ZNVER1-NEXT:    vsqrtpd %ymm0, %ymm0 # sched: [20:1.00]
 ; ZNVER1-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x double> @llvm.x86.avx.sqrt.pd.256(<4 x double> %a0)
   %2 = load <4 x double>, <4 x double> *%a1, align 32
   %3 = call <4 x double> @llvm.x86.avx.sqrt.pd.256(<4 x double> %2)
   %4 = fadd <4 x double> %1, %3
   ret <4 x double> %4
 }
 declare <4 x double> @llvm.x86.avx.sqrt.pd.256(<4 x double>) nounwind readnone
 

 ; BTVER2-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_sqrtps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vsqrtps (%rdi), %ymm1 # sched: [27:1.00]
 ; ZNVER1-NEXT:    vsqrtps %ymm0, %ymm0 # sched: [20:1.00]
 ; ZNVER1-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x float> @llvm.x86.avx.sqrt.ps.256(<8 x float> %a0)
   %2 = load <8 x float>, <8 x float> *%a1, align 32
   %3 = call <8 x float> @llvm.x86.avx.sqrt.ps.256(<8 x float> %2)
   %4 = fadd <8 x float> %1, %3
   ret <8 x float> %4
 }
 declare <8 x float> @llvm.x86.avx.sqrt.ps.256(<8 x float>) nounwind readnone
 

 ; BTVER2-NEXT:    vsubpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    vsubpd (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_subpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vsubpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fsub <4 x double> %a0, %a1
   %2 = load <4 x double>, <4 x double> *%a2, align 32
   %3 = fsub <4 x double> %1, %2
   ret <4 x double> %3
 }
 
 define <8 x float> @test_subps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
 ; GENERIC-LABEL: test_subps:

 ; BTVER2-NEXT:    vsubps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    vsubps (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_subps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vsubps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fsub <8 x float> %a0, %a1
   %2 = load <8 x float>, <8 x float> *%a2, align 32
   %3 = fsub <8 x float> %1, %2
   ret <8 x float> %3
 }
 
 define i32 @test_testpd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) {
 ; GENERIC-LABEL: test_testpd:

 ;
 ; ZNVER1-LABEL: test_testpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    xorl %eax, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    vtestpd %xmm1, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    setb %al # sched: [1:0.25]
 ; ZNVER1-NEXT:    vtestpd (%rdi), %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    adcl $0, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.avx.vtestc.pd(<2 x double> %a0, <2 x double> %a1)
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = call i32 @llvm.x86.avx.vtestc.pd(<2 x double> %a0, <2 x double> %2)
   %4 = add i32 %1, %3
   ret i32 %4
 }
 declare i32 @llvm.x86.avx.vtestc.pd(<2 x double>, <2 x double>) nounwind readnone
 

 ; ZNVER1-LABEL: test_testpd_ymm:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    xorl %eax, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    vtestpd %ymm1, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    setb %al # sched: [1:0.25]
 ; ZNVER1-NEXT:    vtestpd (%rdi), %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    adcl $0, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    vzeroupper
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.avx.vtestc.pd.256(<4 x double> %a0, <4 x double> %a1)
   %2 = load <4 x double>, <4 x double> *%a2, align 32
   %3 = call i32 @llvm.x86.avx.vtestc.pd.256(<4 x double> %a0, <4 x double> %2)
   %4 = add i32 %1, %3
   ret i32 %4
 }
 declare i32 @llvm.x86.avx.vtestc.pd.256(<4 x double>, <4 x double>) nounwind readnone
 

 ;
 ; ZNVER1-LABEL: test_testps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    xorl %eax, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    vtestps %xmm1, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    setb %al # sched: [1:0.25]
 ; ZNVER1-NEXT:    vtestps (%rdi), %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    adcl $0, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.avx.vtestc.ps(<4 x float> %a0, <4 x float> %a1)
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = call i32 @llvm.x86.avx.vtestc.ps(<4 x float> %a0, <4 x float> %2)
   %4 = add i32 %1, %3
   ret i32 %4
 }
 declare i32 @llvm.x86.avx.vtestc.ps(<4 x float>, <4 x float>) nounwind readnone
 

 ; ZNVER1-LABEL: test_testps_ymm:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    xorl %eax, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    vtestps %ymm1, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    setb %al # sched: [1:0.25]
 ; ZNVER1-NEXT:    vtestps (%rdi), %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    adcl $0, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    vzeroupper
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.avx.vtestc.ps.256(<8 x float> %a0, <8 x float> %a1)
   %2 = load <8 x float>, <8 x float> *%a2, align 32
   %3 = call i32 @llvm.x86.avx.vtestc.ps.256(<8 x float> %a0, <8 x float> %2)
   %4 = add i32 %1, %3
   ret i32 %4
 }
 declare i32 @llvm.x86.avx.vtestc.ps.256(<8 x float>, <8 x float>) nounwind readnone
 

 ; BTVER2-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_unpckhpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:0.50]
 ; ZNVER1-NEXT:    vunpckhpd {{.*#+}} ymm1 = ymm1[1],mem[1],ymm1[3],mem[3] sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x double> %a0, <4 x double> %a1, <4 x i32> <i32 1, i32 5, i32 3, i32 7>
   %2 = load <4 x double>, <4 x double> *%a2, align 32
   %3 = shufflevector <4 x double> %a1, <4 x double> %2, <4 x i32> <i32 1, i32 5, i32 3, i32 7>
   %4 = fadd <4 x double> %1, %3
   ret <4 x double> %4
 }
 
 define <8 x float> @test_unpckhps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) nounwind {

 ; BTVER2-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:0.50]
 ; BTVER2-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_unpckhps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:0.50]
 ; ZNVER1-NEXT:    vunpckhps {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x float> %a0, <8 x float> %a1, <8 x i32> <i32 2, i32 10, i32 3, i32 11, i32 6, i32 14, i32 7, i32 15>
   %2 = load <8 x float>, <8 x float> *%a2, align 32
   %3 = shufflevector <8 x float> %1, <8 x float> %2, <8 x i32> <i32 2, i32 10, i32 3, i32 11, i32 6, i32 14, i32 7, i32 15>
   ret <8 x float> %3
 }
 
 define <4 x double> @test_unpcklpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
 ; GENERIC-LABEL: test_unpcklpd:

 ; BTVER2-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_unpcklpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:0.50]
 ; ZNVER1-NEXT:    vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x double> %a0, <4 x double> %a1, <4 x i32> <i32 0, i32 4, i32 2, i32 6>
   %2 = load <4 x double>, <4 x double> *%a2, align 32
   %3 = shufflevector <4 x double> %a1, <4 x double> %2, <4 x i32> <i32 0, i32 4, i32 2, i32 6>
   %4 = fadd <4 x double> %1, %3
   ret <4 x double> %4
 }
 
 define <8 x float> @test_unpcklps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) nounwind {

 ; BTVER2-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:0.50]
 ; BTVER2-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_unpcklps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:0.50]
 ; ZNVER1-NEXT:    vunpcklps {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x float> %a0, <8 x float> %a1, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 4, i32 12, i32 5, i32 13>
   %2 = load <8 x float>, <8 x float> *%a2, align 32
   %3 = shufflevector <8 x float> %1, <8 x float> %2, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 4, i32 12, i32 5, i32 13>
   ret <8 x float> %3
 }
 
 define <4 x double> @test_xorpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
 ; GENERIC-LABEL: test_xorpd:

 ; BTVER2-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_xorpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vxorpd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vxorpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast <4 x double> %a0 to <4 x i64>
   %2 = bitcast <4 x double> %a1 to <4 x i64>
   %3 = xor <4 x i64> %1, %2
   %4 = load <4 x double>, <4 x double> *%a2, align 32
   %5 = bitcast <4 x double> %4 to <4 x i64>
   %6 = xor <4 x i64> %3, %5
   %7 = bitcast <4 x i64> %6 to <4 x double>
   %8 = fadd <4 x double> %a1, %7

 ; BTVER2-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_xorps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vxorps %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vxorps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast <8 x float> %a0 to <4 x i64>
   %2 = bitcast <8 x float> %a1 to <4 x i64>
   %3 = xor <4 x i64> %1, %2
   %4 = load <8 x float>, <8 x float> *%a2, align 32
   %5 = bitcast <8 x float> %4 to <4 x i64>
   %6 = xor <4 x i64> %3, %5
   %7 = bitcast <4 x i64> %6 to <8 x float>
   %8 = fadd <8 x float> %a1, %7

 ; BTVER2-LABEL: test_zeroall:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vzeroall # sched: [90:?]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_zeroall:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vzeroall
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   call void @llvm.x86.avx.vzeroall()
   ret void
 }
 declare void @llvm.x86.avx.vzeroall() nounwind
 
 define void @test_zeroupper() {
 ; GENERIC-LABEL: test_zeroupper:
 ; GENERIC:       # BB#0:

 ; BTVER2-LABEL: test_zeroupper:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vzeroupper # sched: [46:?]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_zeroupper:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vzeroupper
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   call void @llvm.x86.avx.vzeroupper()
   ret void
 }
 declare void @llvm.x86.avx.vzeroupper() nounwind
 
 !0 = !{i32 1}

test/CodeGen/X86/avx2-schedule.ll

 ; HASWELL-NEXT:    vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_pabsb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpabsb (%rdi), %ymm1 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpabsb %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <32 x i8> @llvm.x86.avx2.pabs.b(<32 x i8> %a0)
   %2 = load <32 x i8>, <32 x i8> *%a1, align 32
   %3 = call <32 x i8> @llvm.x86.avx2.pabs.b(<32 x i8> %2)
   %4 = or <32 x i8> %1, %3
   ret <32 x i8> %4
 }
 declare <32 x i8> @llvm.x86.avx2.pabs.b(<32 x i8>) nounwind readnone
 

 ; HASWELL-NEXT:    vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_pabsd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpabsd (%rdi), %ymm1 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpabsd %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i32> @llvm.x86.avx2.pabs.d(<8 x i32> %a0)
   %2 = load <8 x i32>, <8 x i32> *%a1, align 32
   %3 = call <8 x i32> @llvm.x86.avx2.pabs.d(<8 x i32> %2)
   %4 = or <8 x i32> %1, %3
   ret <8 x i32> %4
 }
 declare <8 x i32> @llvm.x86.avx2.pabs.d(<8 x i32>) nounwind readnone
 

 ; HASWELL-NEXT:    vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_pabsw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpabsw (%rdi), %ymm1 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpabsw %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i16> @llvm.x86.avx2.pabs.w(<16 x i16> %a0)
   %2 = load <16 x i16>, <16 x i16> *%a1, align 32
   %3 = call <16 x i16> @llvm.x86.avx2.pabs.w(<16 x i16> %2)
   %4 = or <16 x i16> %1, %3
   ret <16 x i16> %4
 }
 declare <16 x i16> @llvm.x86.avx2.pabs.w(<16 x i16>) nounwind readnone
 

 ; HASWELL-NEXT:    vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpaddb (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_paddb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = add <32 x i8> %a0, %a1
   %2 = load <32 x i8>, <32 x i8> *%a2, align 32
   %3 = add <32 x i8> %1, %2
   ret <32 x i8> %3
 }
 
 define <8 x i32> @test_paddd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
 ; GENERIC-LABEL: test_paddd:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    vpaddd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; GENERIC-NEXT:    vpaddd (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_paddd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpaddd (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_paddd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = add <8 x i32> %a0, %a1
   %2 = load <8 x i32>, <8 x i32> *%a2, align 32
   %3 = add <8 x i32> %1, %2
   ret <8 x i32> %3
 }
 
 define <4 x i64> @test_paddq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
 ; GENERIC-LABEL: test_paddq:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; GENERIC-NEXT:    vpaddq (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_paddq:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpaddq (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_paddq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = add <4 x i64> %a0, %a1
   %2 = load <4 x i64>, <4 x i64> *%a2, align 32
   %3 = add <4 x i64> %1, %2
   ret <4 x i64> %3
 }
 
 define <16 x i16> @test_paddw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
 ; GENERIC-LABEL: test_paddw:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    vpaddw %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; GENERIC-NEXT:    vpaddw (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_paddw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpaddw (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_paddw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = add <16 x i16> %a0, %a1
   %2 = load <16 x i16>, <16 x i16> *%a2, align 32
   %3 = add <16 x i16> %1, %2
   ret <16 x i16> %3
 }
 
 define <4 x i64> @test_pand(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
 ; GENERIC-LABEL: test_pand:

 ; HASWELL-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_pand:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpand %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpand (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = and <4 x i64> %a0, %a1
   %2 = load <4 x i64>, <4 x i64> *%a2, align 32
   %3 = and <4 x i64> %1, %2
   %4 = add <4 x i64> %3, %a1
   ret <4 x i64> %4
 }
 
 define <4 x i64> @test_pandn(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {

 ; HASWELL-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_pandn:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpandn %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpandn (%rdi), %ymm0, %ymm1 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = xor <4 x i64> %a0, <i64 -1, i64 -1, i64 -1, i64 -1>
   %2 = and <4 x i64> %a1, %1
   %3 = load <4 x i64>, <4 x i64> *%a2, align 32
   %4 = xor <4 x i64> %2, <i64 -1, i64 -1, i64 -1, i64 -1>
   %5 = and <4 x i64> %3, %4
   %6 = add <4 x i64> %2, %5
   ret <4 x i64> %6
 }

 ; HASWELL-NEXT:    vpmulld %ymm1, %ymm0, %ymm0 # sched: [10:2.00]
 ; HASWELL-NEXT:    vpmulld (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_pmulld:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmulld %ymm1, %ymm0, %ymm0 # sched: [4:1.00]
 ; ZNVER1-NEXT:    vpmulld (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = mul <8 x i32> %a0, %a1
   %2 = load <8 x i32>, <8 x i32> *%a2, align 32
   %3 = mul <8 x i32> %1, %2
   ret <8 x i32> %3
 }
 
 define <16 x i16> @test_pmullw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
 ; GENERIC-LABEL: test_pmullw:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    vpmullw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
 ; GENERIC-NEXT:    vpmullw (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_pmullw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpmullw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
 ; HASWELL-NEXT:    vpmullw (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_pmullw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmullw %ymm1, %ymm0, %ymm0 # sched: [4:1.00]
 ; ZNVER1-NEXT:    vpmullw (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = mul <16 x i16> %a0, %a1
   %2 = load <16 x i16>, <16 x i16> *%a2, align 32
   %3 = mul <16 x i16> %1, %2
   ret <16 x i16> %3
 }
 
 define <4 x i64> @test_por(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
 ; GENERIC-LABEL: test_por:

 ; HASWELL-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_por:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpor (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = or <4 x i64> %a0, %a1
   %2 = load <4 x i64>, <4 x i64> *%a2, align 32
   %3 = or <4 x i64> %1, %2
   %4 = add <4 x i64> %3, %a1
   ret <4 x i64> %4
 }
 
 define <32 x i8> @test_psubb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
 ; GENERIC-LABEL: test_psubb:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    vpsubb %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; GENERIC-NEXT:    vpsubb (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_psubb:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpsubb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpsubb (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_psubb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsubb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsubb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = sub <32 x i8> %a0, %a1
   %2 = load <32 x i8>, <32 x i8> *%a2, align 32
   %3 = sub <32 x i8> %1, %2
   ret <32 x i8> %3
 }
 
 define <8 x i32> @test_psubd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
 ; GENERIC-LABEL: test_psubd:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    vpsubd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; GENERIC-NEXT:    vpsubd (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_psubd:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpsubd (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_psubd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsubd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = sub <8 x i32> %a0, %a1
   %2 = load <8 x i32>, <8 x i32> *%a2, align 32
   %3 = sub <8 x i32> %1, %2
   ret <8 x i32> %3
 }
 
 define <4 x i64> @test_psubq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
 ; GENERIC-LABEL: test_psubq:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    vpsubq %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; GENERIC-NEXT:    vpsubq (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_psubq:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpsubq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpsubq (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_psubq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsubq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsubq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = sub <4 x i64> %a0, %a1
   %2 = load <4 x i64>, <4 x i64> *%a2, align 32
   %3 = sub <4 x i64> %1, %2
   ret <4 x i64> %3
 }
 
 define <16 x i16> @test_psubw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
 ; GENERIC-LABEL: test_psubw:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    vpsubw %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; GENERIC-NEXT:    vpsubw (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_psubw:
 ; HASWELL:       # BB#0:
 ; HASWELL-NEXT:    vpsubw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    vpsubw (%rdi), %ymm0, %ymm0 # sched: [5:0.50]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_psubw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsubw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsubw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = sub <16 x i16> %a0, %a1
   %2 = load <16 x i16>, <16 x i16> *%a2, align 32
   %3 = sub <16 x i16> %1, %2
   ret <16 x i16> %3
 }
 
 define <4 x i64> @test_pxor(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
 ; GENERIC-LABEL: test_pxor:

 ; HASWELL-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_pxor:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpxor %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpxor (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = xor <4 x i64> %a0, %a1
   %2 = load <4 x i64>, <4 x i64> *%a2, align 32
   %3 = xor <4 x i64> %1, %2
   %4 = add <4 x i64> %3, %a1
   ret <4 x i64> %4
 }
 
 !0 = !{i32 1}

test/CodeGen/X86/bmi-schedule.ll

 ;
 ; ZNVER1-LABEL: test_andn_i16:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    andnl %esi, %edi, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    notl %edi # sched: [1:0.25]
 ; ZNVER1-NEXT:    andw (%rdx), %di # sched: [5:0.50]
 ; ZNVER1-NEXT:    addl %edi, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i16, i16 *%a2
   %2 = xor i16 %a0, -1
   %3 = and i16 %2, %a1
   %4 = and i16 %2, %1
   %5 = add i16 %3, %4
   ret i16 %5
 }
 

 ; BTVER2-NEXT:    addl %ecx, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_andn_i32:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    andnl (%rdx), %edi, %eax # sched: [5:0.50]
 ; ZNVER1-NEXT:    andnl %esi, %edi, %ecx # sched: [1:0.25]
 ; ZNVER1-NEXT:    addl %ecx, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i32, i32 *%a2
   %2 = xor i32 %a0, -1
   %3 = and i32 %2, %a1
   %4 = and i32 %2, %1
   %5 = add i32 %3, %4
   ret i32 %5
 }
 

 ; BTVER2-NEXT:    addq %rcx, %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_andn_i64:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    andnq (%rdx), %rdi, %rax # sched: [5:0.50]
 ; ZNVER1-NEXT:    andnq %rsi, %rdi, %rcx # sched: [1:0.25]
 ; ZNVER1-NEXT:    addq %rcx, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i64, i64 *%a2
   %2 = xor i64 %a0, -1
   %3 = and i64 %2, %a1
   %4 = and i64 %2, %1
   %5 = add i64 %3, %4
   ret i64 %5
 }
 

 ; BTVER2-NEXT:    addl %ecx, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_bextr_i32:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    bextrl %edi, (%rdx), %ecx
 ; ZNVER1-NEXT:    bextrl %edi, %esi, %eax
 ; ZNVER1-NEXT:    addl %ecx, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i32, i32 *%a2
   %2 = tail call i32 @llvm.x86.bmi.bextr.32(i32 %1, i32 %a0)
   %3 = tail call i32 @llvm.x86.bmi.bextr.32(i32 %a1, i32 %a0)
   %4 = add i32 %2, %3
   ret i32 %4
 }
 declare i32 @llvm.x86.bmi.bextr.32(i32, i32)
 

 ; BTVER2-NEXT:    addq %rcx, %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_bextr_i64:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    bextrq %rdi, (%rdx), %rcx
 ; ZNVER1-NEXT:    bextrq %rdi, %rsi, %rax
 ; ZNVER1-NEXT:    addq %rcx, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i64, i64 *%a2
   %2 = tail call i64 @llvm.x86.bmi.bextr.64(i64 %1, i64 %a0)
   %3 = tail call i64 @llvm.x86.bmi.bextr.64(i64 %a1, i64 %a0)
   %4 = add i64 %2, %3
   ret i64 %4
 }
 declare i64 @llvm.x86.bmi.bextr.64(i64, i64)
 

 ; BTVER2-NEXT:    addl %ecx, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_blsi_i32:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    blsil (%rsi), %ecx
 ; ZNVER1-NEXT:    blsil %edi, %eax
 ; ZNVER1-NEXT:    addl %ecx, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i32, i32 *%a1
   %2 = sub i32 0, %1
   %3 = sub i32 0, %a0
   %4 = and i32 %1, %2
   %5 = and i32 %a0, %3
   %6 = add i32 %4, %5
   ret i32 %6
 }

 ; BTVER2-NEXT:    addq %rcx, %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_blsi_i64:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    blsiq (%rsi), %rcx
 ; ZNVER1-NEXT:    blsiq %rdi, %rax
 ; ZNVER1-NEXT:    addq %rcx, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i64, i64 *%a1
   %2 = sub i64 0, %1
   %3 = sub i64 0, %a0
   %4 = and i64 %1, %2
   %5 = and i64 %a0, %3
   %6 = add i64 %4, %5
   ret i64 %6
 }

 ; BTVER2-NEXT:    addl %ecx, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_blsmsk_i32:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    blsmskl (%rsi), %ecx
 ; ZNVER1-NEXT:    blsmskl %edi, %eax
 ; ZNVER1-NEXT:    addl %ecx, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i32, i32 *%a1
   %2 = sub i32 %1, 1
   %3 = sub i32 %a0, 1
   %4 = xor i32 %1, %2
   %5 = xor i32 %a0, %3
   %6 = add i32 %4, %5
   ret i32 %6
 }

 ; BTVER2-NEXT:    addq %rcx, %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_blsmsk_i64:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    blsmskq (%rsi), %rcx
 ; ZNVER1-NEXT:    blsmskq %rdi, %rax
 ; ZNVER1-NEXT:    addq %rcx, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i64, i64 *%a1
   %2 = sub i64 %1, 1
   %3 = sub i64 %a0, 1
   %4 = xor i64 %1, %2
   %5 = xor i64 %a0, %3
   %6 = add i64 %4, %5
   ret i64 %6
 }

 ; BTVER2-NEXT:    addl %ecx, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_blsr_i32:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    blsrl (%rsi), %ecx
 ; ZNVER1-NEXT:    blsrl %edi, %eax
 ; ZNVER1-NEXT:    addl %ecx, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i32, i32 *%a1
   %2 = sub i32 %1, 1
   %3 = sub i32 %a0, 1
   %4 = and i32 %1, %2
   %5 = and i32 %a0, %3
   %6 = add i32 %4, %5
   ret i32 %6
 }

 ; BTVER2-NEXT:    addq %rcx, %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_blsr_i64:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    blsrq (%rsi), %rcx
 ; ZNVER1-NEXT:    blsrq %rdi, %rax
 ; ZNVER1-NEXT:    addq %rcx, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i64, i64 *%a1
   %2 = sub i64 %1, 1
   %3 = sub i64 %a0, 1
   %4 = and i64 %1, %2
   %5 = and i64 %a0, %3
   %6 = add i64 %4, %5
   ret i64 %6
 }

 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cttz_i16:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    tzcntw (%rsi), %cx
 ; ZNVER1-NEXT:    tzcntw %di, %ax
 ; ZNVER1-NEXT:    orl %ecx, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i16, i16 *%a1
   %2 = tail call i16 @llvm.cttz.i16( i16 %1, i1 false )
   %3 = tail call i16 @llvm.cttz.i16( i16 %a0, i1 false )
   %4 = or i16 %2, %3
   ret i16 %4
 }
 declare i16 @llvm.cttz.i16(i16, i1)
 

 ; BTVER2-NEXT:    orl %ecx, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cttz_i32:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    tzcntl (%rsi), %ecx
 ; ZNVER1-NEXT:    tzcntl %edi, %eax
 ; ZNVER1-NEXT:    orl %ecx, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i32, i32 *%a1
   %2 = tail call i32 @llvm.cttz.i32( i32 %1, i1 false )
   %3 = tail call i32 @llvm.cttz.i32( i32 %a0, i1 false )
   %4 = or i32 %2, %3
   ret i32 %4
 }
 declare i32 @llvm.cttz.i32(i32, i1)
 

 ; BTVER2-NEXT:    orq %rcx, %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cttz_i64:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    tzcntq (%rsi), %rcx
 ; ZNVER1-NEXT:    tzcntq %rdi, %rax
 ; ZNVER1-NEXT:    orq %rcx, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i64, i64 *%a1
   %2 = tail call i64 @llvm.cttz.i64( i64 %1, i1 false )
   %3 = tail call i64 @llvm.cttz.i64( i64 %a0, i1 false )
   %4 = or i64 %2, %3
   ret i64 %4
 }
 declare i64 @llvm.cttz.i64(i64, i1)

test/CodeGen/X86/bmi2-schedule.ll

 ; HASWELL-NEXT:    addl %ecx, %eax # sched: [1:0.25]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_bzhi_i32:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    bzhil %edi, (%rdx), %ecx
 ; ZNVER1-NEXT:    bzhil %edi, %esi, %eax
 ; ZNVER1-NEXT:    addl %ecx, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i32, i32 *%a2
   %2 = tail call i32 @llvm.x86.bmi.bzhi.32(i32 %1, i32 %a0)
   %3 = tail call i32 @llvm.x86.bmi.bzhi.32(i32 %a1, i32 %a0)
   %4 = add i32 %2, %3
   ret i32 %4
 }
 declare i32 @llvm.x86.bmi.bzhi.32(i32, i32)
 

 ; HASWELL-NEXT:    addq %rcx, %rax # sched: [1:0.25]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_bzhi_i64:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    bzhiq %rdi, (%rdx), %rcx
 ; ZNVER1-NEXT:    bzhiq %rdi, %rsi, %rax
 ; ZNVER1-NEXT:    addq %rcx, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i64, i64 *%a2
   %2 = tail call i64 @llvm.x86.bmi.bzhi.64(i64 %1, i64 %a0)
   %3 = tail call i64 @llvm.x86.bmi.bzhi.64(i64 %a1, i64 %a0)
   %4 = add i64 %2, %3
   ret i64 %4
 }
 declare i64 @llvm.x86.bmi.bzhi.64(i64, i64)
 

 ; HASWELL-NEXT:    addl %ecx, %eax # sched: [1:0.25]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_pdep_i32:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    pdepl (%rdx), %edi, %ecx
 ; ZNVER1-NEXT:    pdepl %esi, %edi, %eax
 ; ZNVER1-NEXT:    addl %ecx, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i32, i32 *%a2
   %2 = tail call i32 @llvm.x86.bmi.pdep.32(i32 %a0, i32 %1)
   %3 = tail call i32 @llvm.x86.bmi.pdep.32(i32 %a0, i32 %a1)
   %4 = add i32 %2, %3
   ret i32 %4
 }
 declare i32 @llvm.x86.bmi.pdep.32(i32, i32)
 

 ; HASWELL-NEXT:    addq %rcx, %rax # sched: [1:0.25]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_pdep_i64:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    pdepq (%rdx), %rdi, %rcx
 ; ZNVER1-NEXT:    pdepq %rsi, %rdi, %rax
 ; ZNVER1-NEXT:    addq %rcx, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i64, i64 *%a2
   %2 = tail call i64 @llvm.x86.bmi.pdep.64(i64 %a0, i64 %1)
   %3 = tail call i64 @llvm.x86.bmi.pdep.64(i64 %a0, i64 %a1)
   %4 = add i64 %2, %3
   ret i64 %4
 }
 declare i64 @llvm.x86.bmi.pdep.64(i64, i64)
 

 ; HASWELL-NEXT:    addl %ecx, %eax # sched: [1:0.25]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_pext_i32:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    pextl (%rdx), %edi, %ecx
 ; ZNVER1-NEXT:    pextl %esi, %edi, %eax
 ; ZNVER1-NEXT:    addl %ecx, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i32, i32 *%a2
   %2 = tail call i32 @llvm.x86.bmi.pext.32(i32 %a0, i32 %1)
   %3 = tail call i32 @llvm.x86.bmi.pext.32(i32 %a0, i32 %a1)
   %4 = add i32 %2, %3
   ret i32 %4
 }
 declare i32 @llvm.x86.bmi.pext.32(i32, i32)
 

 ; HASWELL-NEXT:    addq %rcx, %rax # sched: [1:0.25]
 ; HASWELL-NEXT:    retq # sched: [1:1.00]
 ;
 ; ZNVER1-LABEL: test_pext_i64:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    pextq (%rdx), %rdi, %rcx
 ; ZNVER1-NEXT:    pextq %rsi, %rdi, %rax
 ; ZNVER1-NEXT:    addq %rcx, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i64, i64 *%a2
   %2 = tail call i64 @llvm.x86.bmi.pext.64(i64 %a0, i64 %1)
   %3 = tail call i64 @llvm.x86.bmi.pext.64(i64 %a0, i64 %a1)
   %4 = add i64 %2, %3
   ret i64 %4
 }
 declare i64 @llvm.x86.bmi.pext.64(i64, i64)

test/CodeGen/X86/f16c-schedule.ll

 ; BTVER2-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_vcvtph2ps_128:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtph2ps (%rdi), %xmm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvtph2ps %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load <8 x i16>, <8 x i16> *%a1
   %2 = call <4 x float> @llvm.x86.vcvtph2ps.128(<8 x i16> %1)
   %3 = call <4 x float> @llvm.x86.vcvtph2ps.128(<8 x i16> %a0)
   %4 = fadd <4 x float> %2, %3
   ret <4 x float> %4
 }
 declare <4 x float> @llvm.x86.vcvtph2ps.128(<8 x i16>)
 

 ; BTVER2-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_vcvtph2ps_256:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtph2ps (%rdi), %ymm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvtph2ps %xmm0, %ymm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load <8 x i16>, <8 x i16> *%a1
   %2 = call <8 x float> @llvm.x86.vcvtph2ps.256(<8 x i16> %1)
   %3 = call <8 x float> @llvm.x86.vcvtph2ps.256(<8 x i16> %a0)
   %4 = fadd <8 x float> %2, %3
   ret <8 x float> %4
 }
 declare <8 x float> @llvm.x86.vcvtph2ps.256(<8 x i16>)
 

 ; BTVER2-NEXT:    vcvtps2ph $0, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vcvtps2ph $0, %xmm1, (%rdi) # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_vcvtps2ph_128:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtps2ph $0, %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vcvtps2ph $0, %xmm1, (%rdi) # sched: [12:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float> %a0, i32 0)
   %2 = call <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float> %a1, i32 0)
   %3 = shufflevector <8 x i16> %2, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   store <4 x i16> %3, <4 x i16> *%a2
   ret <8 x i16> %1
 }
 declare <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float>, i32)
 

 ; BTVER2-NEXT:    vcvtps2ph $0, %ymm1, (%rdi) # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_vcvtps2ph_256:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtps2ph $0, %ymm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vcvtps2ph $0, %ymm1, (%rdi) # sched: [12:1.00]
 ; ZNVER1-NEXT:    vzeroupper
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.vcvtps2ph.256(<8 x float> %a0, i32 0)
   %2 = call <8 x i16> @llvm.x86.vcvtps2ph.256(<8 x float> %a1, i32 0)
   store <8 x i16> %2, <8 x i16> *%a2
   ret <8 x i16> %1
 }
 declare <8 x i16> @llvm.x86.vcvtps2ph.256(<8 x float>, i32)

test/CodeGen/X86/lea32-schedule.ll

 ; BTVER2-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; BTVER2-NEXT:    leal -24(%rdi), %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_offset:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; ZNVER1-NEXT:    leal -24(%rdi), %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %2 = add nsw i32 %0, -24
   ret i32 %2
 }
 
 define i32 @test_lea_offset_big(i32) {
 ; GENERIC-LABEL: test_lea_offset_big:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>

 ; BTVER2-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; BTVER2-NEXT:    leal 1024(%rdi), %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_offset_big:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; ZNVER1-NEXT:    leal 1024(%rdi), %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %2 = add nsw i32 %0, 1024
   ret i32 %2
 }
 
 ; Function Attrs: norecurse nounwind readnone uwtable
 define i32 @test_lea_add(i32, i32) {
 ; GENERIC-LABEL: test_lea_add:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    leal (%rdi,%rsi), %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_add:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    # kill: %ESI<def> %ESI<kill> %RSI<def>
 ; ZNVER1-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; ZNVER1-NEXT:    leal (%rdi,%rsi), %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %3 = add nsw i32 %1, %0
   ret i32 %3
 }
 
 define i32 @test_lea_add_offset(i32, i32) {
 ; GENERIC-LABEL: test_lea_add_offset:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    # kill: %ESI<def> %ESI<kill> %RSI<def>

 ; BTVER2-NEXT:    leal 16(%rdi,%rsi), %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_add_offset:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    # kill: %ESI<def> %ESI<kill> %RSI<def>
 ; ZNVER1-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; ZNVER1-NEXT:    leal 16(%rdi,%rsi), %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %3 = add i32 %0, 16
   %4 = add i32 %3, %1
   ret i32 %4
 }
 
 define i32 @test_lea_add_offset_big(i32, i32) {
 ; GENERIC-LABEL: test_lea_add_offset_big:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    leal -4096(%rdi,%rsi), %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_add_offset_big:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    # kill: %ESI<def> %ESI<kill> %RSI<def>
 ; ZNVER1-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; ZNVER1-NEXT:    leal -4096(%rdi,%rsi), %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %3 = add i32 %0, -4096
   %4 = add i32 %3, %1
   ret i32 %4
 }
 
 define i32 @test_lea_mul(i32) {
 ; GENERIC-LABEL: test_lea_mul:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; BTVER2-NEXT:    leal (%rdi,%rdi,2), %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_mul:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; ZNVER1-NEXT:    leal (%rdi,%rdi,2), %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %2 = mul nsw i32 %0, 3
   ret i32 %2
 }
 
 define i32 @test_lea_mul_offset(i32) {
 ; GENERIC-LABEL: test_lea_mul_offset:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>

 ; BTVER2-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; BTVER2-NEXT:    leal -32(%rdi,%rdi,2), %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_mul_offset:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; ZNVER1-NEXT:    leal -32(%rdi,%rdi,2), %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %2 = mul nsw i32 %0, 3
   %3 = add nsw i32 %2, -32
   ret i32 %3
 }
 
 define i32 @test_lea_mul_offset_big(i32) {
 ; GENERIC-LABEL: test_lea_mul_offset_big:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; BTVER2-NEXT:    leal 10000(%rdi,%rdi,8), %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_mul_offset_big:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; ZNVER1-NEXT:    leal 10000(%rdi,%rdi,8), %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %2 = mul nsw i32 %0, 9
   %3 = add nsw i32 %2, 10000
   ret i32 %3
 }
 
 define i32 @test_lea_add_scale(i32, i32) {
 ; GENERIC-LABEL: test_lea_add_scale:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    leal (%rdi,%rsi,2), %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_add_scale:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    # kill: %ESI<def> %ESI<kill> %RSI<def>
 ; ZNVER1-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; ZNVER1-NEXT:    leal (%rdi,%rsi,2), %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %3 = shl i32 %1, 1
   %4 = add nsw i32 %3, %0
   ret i32 %4
 }
 
 define i32 @test_lea_add_scale_offset(i32, i32) {
 ; GENERIC-LABEL: test_lea_add_scale_offset:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    leal 96(%rdi,%rsi,4), %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_add_scale_offset:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    # kill: %ESI<def> %ESI<kill> %RSI<def>
 ; ZNVER1-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; ZNVER1-NEXT:    leal 96(%rdi,%rsi,4), %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %3 = shl i32 %1, 2
   %4 = add i32 %0, 96
   %5 = add i32 %4, %3
   ret i32 %5
 }
 
 define i32 @test_lea_add_scale_offset_big(i32, i32) {
 ; GENERIC-LABEL: test_lea_add_scale_offset_big:

 ; BTVER2-NEXT:    leal -1200(%rdi,%rsi,8), %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_add_scale_offset_big:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    # kill: %ESI<def> %ESI<kill> %RSI<def>
 ; ZNVER1-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; ZNVER1-NEXT:    leal -1200(%rdi,%rsi,8), %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %3 = shl i32 %1, 3
   %4 = add i32 %0, -1200
   %5 = add i32 %4, %3
   ret i32 %5
 }

test/CodeGen/X86/lea64-schedule.ll

 ; BTVER2-LABEL: test_lea_offset:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    leaq -24(%rdi), %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_offset:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    leaq -24(%rdi), %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %2 = add nsw i64 %0, -24
   ret i64 %2
 }
 
 define i64 @test_lea_offset_big(i64) {
 ; GENERIC-LABEL: test_lea_offset_big:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    leaq 1024(%rdi), %rax # sched: [1:0.50]

 ; BTVER2-LABEL: test_lea_offset_big:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    leaq 1024(%rdi), %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_offset_big:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    leaq 1024(%rdi), %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %2 = add nsw i64 %0, 1024
   ret i64 %2
 }
 
 ; Function Attrs: norecurse nounwind readnone uwtable
 define i64 @test_lea_add(i64, i64) {
 ; GENERIC-LABEL: test_lea_add:
 ; GENERIC:       # BB#0:

 ; BTVER2-LABEL: test_lea_add:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    leaq (%rdi,%rsi), %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_add:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    leaq (%rdi,%rsi), %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %3 = add nsw i64 %1, %0
   ret i64 %3
 }
 
 define i64 @test_lea_add_offset(i64, i64) {
 ; GENERIC-LABEL: test_lea_add_offset:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    leaq 16(%rdi,%rsi), %rax # sched: [1:0.50]

 ; BTVER2-LABEL: test_lea_add_offset:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    leaq 16(%rdi,%rsi), %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_add_offset:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    leaq 16(%rdi,%rsi), %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %3 = add i64 %0, 16
   %4 = add i64 %3, %1
   ret i64 %4
 }
 
 define i64 @test_lea_add_offset_big(i64, i64) {
 ; GENERIC-LABEL: test_lea_add_offset_big:
 ; GENERIC:       # BB#0:

 ; BTVER2-LABEL: test_lea_add_offset_big:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    leaq -4096(%rdi,%rsi), %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_add_offset_big:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    leaq -4096(%rdi,%rsi), %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %3 = add i64 %0, -4096
   %4 = add i64 %3, %1
   ret i64 %4
 }
 
 define i64 @test_lea_mul(i64) {
 ; GENERIC-LABEL: test_lea_mul:
 ; GENERIC:       # BB#0:

 ; BTVER2-LABEL: test_lea_mul:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    leaq (%rdi,%rdi,2), %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_mul:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    leaq (%rdi,%rdi,2), %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %2 = mul nsw i64 %0, 3
   ret i64 %2
 }
 
 define i64 @test_lea_mul_offset(i64) {
 ; GENERIC-LABEL: test_lea_mul_offset:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    leaq -32(%rdi,%rdi,2), %rax # sched: [1:0.50]

 ; BTVER2-LABEL: test_lea_mul_offset:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    leaq -32(%rdi,%rdi,2), %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_mul_offset:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    leaq -32(%rdi,%rdi,2), %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %2 = mul nsw i64 %0, 3
   %3 = add nsw i64 %2, -32
   ret i64 %3
 }
 
 define i64 @test_lea_mul_offset_big(i64) {
 ; GENERIC-LABEL: test_lea_mul_offset_big:
 ; GENERIC:       # BB#0:

 ; BTVER2-LABEL: test_lea_mul_offset_big:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    leaq 10000(%rdi,%rdi,8), %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_mul_offset_big:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    leaq 10000(%rdi,%rdi,8), %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %2 = mul nsw i64 %0, 9
   %3 = add nsw i64 %2, 10000
   ret i64 %3
 }
 
 define i64 @test_lea_add_scale(i64, i64) {
 ; GENERIC-LABEL: test_lea_add_scale:
 ; GENERIC:       # BB#0:

 ; BTVER2-LABEL: test_lea_add_scale:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    leaq (%rdi,%rsi,2), %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_add_scale:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    leaq (%rdi,%rsi,2), %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %3 = shl i64 %1, 1
   %4 = add nsw i64 %3, %0
   ret i64 %4
 }
 
 define i64 @test_lea_add_scale_offset(i64, i64) {
 ; GENERIC-LABEL: test_lea_add_scale_offset:
 ; GENERIC:       # BB#0:

 ; BTVER2-LABEL: test_lea_add_scale_offset:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    leaq 96(%rdi,%rsi,4), %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_add_scale_offset:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    leaq 96(%rdi,%rsi,4), %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %3 = shl i64 %1, 2
   %4 = add i64 %0, 96
   %5 = add i64 %4, %3
   ret i64 %5
 }
 
 define i64 @test_lea_add_scale_offset_big(i64, i64) {
 ; GENERIC-LABEL: test_lea_add_scale_offset_big:

 ; BTVER2-LABEL: test_lea_add_scale_offset_big:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    leaq -1200(%rdi,%rsi,8), %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lea_add_scale_offset_big:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    leaq -1200(%rdi,%rsi,8), %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %3 = shl i64 %1, 3
   %4 = add i64 %0, -1200
   %5 = add i64 %4, %3
   ret i64 %5
 }

test/CodeGen/X86/lzcnt-schedule.ll

 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_ctlz_i16:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    lzcntw (%rsi), %cx
 ; ZNVER1-NEXT:    lzcntw %di, %ax
 ; ZNVER1-NEXT:    orl %ecx, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i16, i16 *%a1
   %2 = tail call i16 @llvm.ctlz.i16( i16 %1, i1 false )
   %3 = tail call i16 @llvm.ctlz.i16( i16 %a0, i1 false )
   %4 = or i16 %2, %3
   ret i16 %4
 }
 declare i16 @llvm.ctlz.i16(i16, i1)
 

 ; BTVER2-NEXT:    orl %ecx, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_ctlz_i32:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    lzcntl (%rsi), %ecx
 ; ZNVER1-NEXT:    lzcntl %edi, %eax
 ; ZNVER1-NEXT:    orl %ecx, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i32, i32 *%a1
   %2 = tail call i32 @llvm.ctlz.i32( i32 %1, i1 false )
   %3 = tail call i32 @llvm.ctlz.i32( i32 %a0, i1 false )
   %4 = or i32 %2, %3
   ret i32 %4
 }
 declare i32 @llvm.ctlz.i32(i32, i1)
 

 ; BTVER2-NEXT:    orq %rcx, %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_ctlz_i64:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    lzcntq (%rsi), %rcx
 ; ZNVER1-NEXT:    lzcntq %rdi, %rax
 ; ZNVER1-NEXT:    orq %rcx, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i64, i64 *%a1
   %2 = tail call i64 @llvm.ctlz.i64( i64 %1, i1 false )
   %3 = tail call i64 @llvm.ctlz.i64( i64 %a0, i1 false )
   %4 = or i64 %2, %3
   ret i64 %4
 }
 declare i64 @llvm.ctlz.i64(i64, i1)

test/CodeGen/X86/popcnt-schedule.ll

 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_ctpop_i16:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    popcntw (%rsi), %cx # sched: [10:1.00]
 ; ZNVER1-NEXT:    popcntw %di, %ax # sched: [3:1.00]
 ; ZNVER1-NEXT:    orl %ecx, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i16, i16 *%a1
   %2 = tail call i16 @llvm.ctpop.i16( i16 %1 )
   %3 = tail call i16 @llvm.ctpop.i16( i16 %a0 )
   %4 = or i16 %2, %3
   ret i16 %4
 }
 declare i16 @llvm.ctpop.i16(i16)
 

 ; BTVER2-NEXT:    orl %ecx, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_ctpop_i32:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    popcntl (%rsi), %ecx # sched: [10:1.00]
 ; ZNVER1-NEXT:    popcntl %edi, %eax # sched: [3:1.00]
 ; ZNVER1-NEXT:    orl %ecx, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i32, i32 *%a1
   %2 = tail call i32 @llvm.ctpop.i32( i32 %1 )
   %3 = tail call i32 @llvm.ctpop.i32( i32 %a0 )
   %4 = or i32 %2, %3
   ret i32 %4
 }
 declare i32 @llvm.ctpop.i32(i32)
 

 ; BTVER2-NEXT:    orq %rcx, %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_ctpop_i64:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    popcntq (%rsi), %rcx # sched: [10:1.00]
 ; ZNVER1-NEXT:    popcntq %rdi, %rax # sched: [3:1.00]
 ; ZNVER1-NEXT:    orq %rcx, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i64, i64 *%a1
   %2 = tail call i64 @llvm.ctpop.i64( i64 %1 )
   %3 = tail call i64 @llvm.ctpop.i64( i64 %a0 )
   %4 = or i64 %2, %3
   ret i64 %4
 }
 declare i64 @llvm.ctpop.i64(i64)

test/CodeGen/X86/sse-schedule.ll

 ; BTVER2-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vaddps (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_addps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vaddps (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fadd <4 x float> %a0, %a1
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = fadd <4 x float> %1, %2
   ret <4 x float> %3
 }
 
 define float @test_addss(float %a0, float %a1, float *%a2) {
 ; GENERIC-LABEL: test_addss:

 ; BTVER2-NEXT:    vaddss %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vaddss (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_addss:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vaddss %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vaddss (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fadd float %a0, %a1
   %2 = load float, float *%a2, align 4
   %3 = fadd float %1, %2
   ret float %3
 }
 
 define <4 x float> @test_andps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {
 ; GENERIC-LABEL: test_andps:

 ; BTVER2-NEXT:    vandps %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vandps (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_andps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vandps %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vandps (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast <4 x float> %a0 to <4 x i32>
   %2 = bitcast <4 x float> %a1 to <4 x i32>
   %3 = and <4 x i32> %1, %2
   %4 = load <4 x float>, <4 x float> *%a2, align 16
   %5 = bitcast <4 x float> %4 to <4 x i32>
   %6 = and <4 x i32> %3, %5
   %7 = bitcast <4 x i32> %6 to <4 x float>
   ret <4 x float> %7

 ; BTVER2-NEXT:    vandnps %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vandnps (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_andnotps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vandnps %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vandnps (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast <4 x float> %a0 to <4 x i32>
   %2 = bitcast <4 x float> %a1 to <4 x i32>
   %3 = xor <4 x i32> %1, <i32 -1, i32 -1, i32 -1, i32 -1>
   %4 = and <4 x i32> %3, %2
   %5 = load <4 x float>, <4 x float> *%a2, align 16
   %6 = bitcast <4 x float> %5 to <4 x i32>
   %7 = xor <4 x i32> %4, <i32 -1, i32 -1, i32 -1, i32 -1>
   %8 = and <4 x i32> %6, %7

 ; BTVER2-NEXT:    vorps %xmm0, %xmm1, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cmpps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcmpeqps %xmm1, %xmm0, %xmm1 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vcmpeqps (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
 ; ZNVER1-NEXT:    vorps %xmm0, %xmm1, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fcmp oeq <4 x float> %a0, %a1
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = fcmp oeq <4 x float> %a0, %2
   %4 = or <4 x i1> %1, %3
   %5 = sext <4 x i1> %4 to <4 x i32>
   %6 = bitcast <4 x i32> %5 to <4 x float>
   ret <4 x float> %6
 }

 ; BTVER2-NEXT:    vcmpeqss %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vcmpeqss (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cmpss:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcmpeqss %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vcmpeqss (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = insertelement <4 x float> undef, float %a0, i32 0
   %2 = insertelement <4 x float> undef, float %a1, i32 0
   %3 = call <4 x float> @llvm.x86.sse.cmp.ss(<4 x float> %1, <4 x float> %2, i8 0)
   %4 = load float, float *%a2, align 4
   %5 = insertelement <4 x float> undef, float %4, i32 0
   %6 = call <4 x float> @llvm.x86.sse.cmp.ss(<4 x float> %3, <4 x float> %5, i8 0)
   %7 = extractelement <4 x float> %6, i32 0
   ret float %7

 ; ZNVER1-NEXT:    sete %cl # sched: [1:0.25]
 ; ZNVER1-NEXT:    andb %al, %cl # sched: [1:0.25]
 ; ZNVER1-NEXT:    vcomiss (%rdi), %xmm0 # sched: [10:1.00]
 ; ZNVER1-NEXT:    setnp %al # sched: [1:0.25]
 ; ZNVER1-NEXT:    sete %dl # sched: [1:0.25]
 ; ZNVER1-NEXT:    andb %al, %dl # sched: [1:0.25]
 ; ZNVER1-NEXT:    orb %cl, %dl # sched: [1:0.25]
 ; ZNVER1-NEXT:    movzbl %dl, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.sse.comieq.ss(<4 x float> %a0, <4 x float> %a1)
   %2 = load <4 x float>, <4 x float> *%a2, align 4
   %3 = call i32 @llvm.x86.sse.comieq.ss(<4 x float> %a0, <4 x float> %2)
   %4 = or i32 %1, %3
   ret i32 %4
 }
 declare i32 @llvm.x86.sse.comieq.ss(<4 x float>, <4 x float>) nounwind readnone
 

 ; BTVER2-NEXT:    vaddss %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtsi2ss:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtsi2ssl %edi, %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vcvtsi2ssl (%rsi), %xmm1, %xmm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vaddss %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = sitofp i32 %a0 to float
   %2 = load i32, i32 *%a1, align 4
   %3 = sitofp i32 %2 to float
   %4 = fadd float %1, %3
   ret float %4
 }
 
 define float @test_cvtsi2ssq(i64 %a0, i64 *%a1) {

 ; BTVER2-NEXT:    vaddss %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtsi2ssq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtsi2ssq %rdi, %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vcvtsi2ssq (%rsi), %xmm1, %xmm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vaddss %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = sitofp i64 %a0 to float
   %2 = load i64, i64 *%a1, align 8
   %3 = sitofp i64 %2 to float
   %4 = fadd float %1, %3
   ret float %4
 }
 
 define i32 @test_cvtss2si(float %a0, float *%a1) {

 ; BTVER2-NEXT:    addl %ecx, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtss2si:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtss2si (%rdi), %eax # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvtss2si %xmm0, %ecx # sched: [5:1.00]
 ; ZNVER1-NEXT:    addl %ecx, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = insertelement <4 x float> undef, float %a0, i32 0
   %2 = call i32 @llvm.x86.sse.cvtss2si(<4 x float> %1)
   %3 = load float, float *%a1, align 4
   %4 = insertelement <4 x float> undef, float %3, i32 0
   %5 = call i32 @llvm.x86.sse.cvtss2si(<4 x float> %4)
   %6 = add i32 %2, %5
   ret i32 %6
 }

 ; BTVER2-NEXT:    addq %rcx, %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtss2siq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtss2si (%rdi), %rax # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvtss2si %xmm0, %rcx # sched: [5:1.00]
 ; ZNVER1-NEXT:    addq %rcx, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = insertelement <4 x float> undef, float %a0, i32 0
   %2 = call i64 @llvm.x86.sse.cvtss2si64(<4 x float> %1)
   %3 = load float, float *%a1, align 4
   %4 = insertelement <4 x float> undef, float %3, i32 0
   %5 = call i64 @llvm.x86.sse.cvtss2si64(<4 x float> %4)
   %6 = add i64 %2, %5
   ret i64 %6
 }

 ; BTVER2-NEXT:    addl %ecx, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvttss2si:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvttss2si (%rdi), %eax # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvttss2si %xmm0, %ecx # sched: [5:1.00]
 ; ZNVER1-NEXT:    addl %ecx, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fptosi float %a0 to i32
   %2 = load float, float *%a1, align 4
   %3 = fptosi float %2 to i32
   %4 = add i32 %1, %3
   ret i32 %4
 }
 
 define i64 @test_cvttss2siq(float %a0, float *%a1) {

 ; BTVER2-NEXT:    addq %rcx, %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvttss2siq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvttss2si (%rdi), %rax # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvttss2si %xmm0, %rcx # sched: [5:1.00]
 ; ZNVER1-NEXT:    addq %rcx, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fptosi float %a0 to i64
   %2 = load float, float *%a1, align 4
   %3 = fptosi float %2 to i64
   %4 = add i64 %1, %3
   ret i64 %4
 }
 
 define <4 x float> @test_divps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {

 ; BTVER2-NEXT:    vdivps %xmm1, %xmm0, %xmm0 # sched: [19:19.00]
 ; BTVER2-NEXT:    vdivps (%rdi), %xmm0, %xmm0 # sched: [24:19.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_divps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vdivps %xmm1, %xmm0, %xmm0 # sched: [15:1.00]
 ; ZNVER1-NEXT:    vdivps (%rdi), %xmm0, %xmm0 # sched: [22:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fdiv <4 x float> %a0, %a1
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = fdiv <4 x float> %1, %2
   ret <4 x float> %3
 }
 
 define float @test_divss(float %a0, float %a1, float *%a2) {
 ; GENERIC-LABEL: test_divss:

 ; BTVER2-NEXT:    vdivss %xmm1, %xmm0, %xmm0 # sched: [19:19.00]
 ; BTVER2-NEXT:    vdivss (%rdi), %xmm0, %xmm0 # sched: [24:19.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_divss:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vdivss %xmm1, %xmm0, %xmm0 # sched: [15:1.00]
 ; ZNVER1-NEXT:    vdivss (%rdi), %xmm0, %xmm0 # sched: [22:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fdiv float %a0, %a1
   %2 = load float, float *%a2, align 4
   %3 = fdiv float %1, %2
   ret float %3
 }
 
 define void @test_ldmxcsr(i32 %a0) {
 ; GENERIC-LABEL: test_ldmxcsr:

 ; BTVER2-NEXT:    movl %edi, -{{[0-9]+}}(%rsp) # sched: [1:1.00]
 ; BTVER2-NEXT:    vldmxcsr -{{[0-9]+}}(%rsp) # sched: [5:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_ldmxcsr:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    movl %edi, -{{[0-9]+}}(%rsp) # sched: [1:0.50]
 ; ZNVER1-NEXT:    vldmxcsr -{{[0-9]+}}(%rsp) # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = alloca i32, align 4
   %2 = bitcast i32* %1 to i8*
   store i32 %a0, i32* %1
   call void @llvm.x86.sse.ldmxcsr(i8* %2)
   ret void
 }
 declare void @llvm.x86.sse.ldmxcsr(i8*) nounwind readnone
 

 ; BTVER2-NEXT:    vmaxps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vmaxps (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_maxps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmaxps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmaxps (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.sse.max.ps(<4 x float> %a0, <4 x float> %a1)
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = call <4 x float> @llvm.x86.sse.max.ps(<4 x float> %1, <4 x float> %2)
   ret <4 x float> %3
 }
 declare <4 x float> @llvm.x86.sse.max.ps(<4 x float>, <4 x float>) nounwind readnone
 
 define <4 x float> @test_maxss(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {

 ; BTVER2-NEXT:    vmaxss %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vmaxss (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_maxss:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmaxss %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmaxss (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.sse.max.ss(<4 x float> %a0, <4 x float> %a1)
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = call <4 x float> @llvm.x86.sse.max.ss(<4 x float> %1, <4 x float> %2)
   ret <4 x float> %3
 }
 declare <4 x float> @llvm.x86.sse.max.ss(<4 x float>, <4 x float>) nounwind readnone
 
 define <4 x float> @test_minps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {

 ; BTVER2-NEXT:    vminps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vminps (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_minps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vminps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vminps (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.sse.min.ps(<4 x float> %a0, <4 x float> %a1)
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = call <4 x float> @llvm.x86.sse.min.ps(<4 x float> %1, <4 x float> %2)
   ret <4 x float> %3
 }
 declare <4 x float> @llvm.x86.sse.min.ps(<4 x float>, <4 x float>) nounwind readnone
 
 define <4 x float> @test_minss(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {

 ; BTVER2-NEXT:    vminss %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vminss (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_minss:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vminss %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vminss (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.sse.min.ss(<4 x float> %a0, <4 x float> %a1)
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = call <4 x float> @llvm.x86.sse.min.ss(<4 x float> %1, <4 x float> %2)
   ret <4 x float> %3
 }
 declare <4 x float> @llvm.x86.sse.min.ss(<4 x float>, <4 x float>) nounwind readnone
 
 define void @test_movaps(<4 x float> *%a0, <4 x float> *%a1) {

 ; BTVER2-NEXT:    vmovaps %xmm0, (%rsi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movaps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovaps (%rdi), %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddps %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmovaps %xmm0, (%rsi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load <4 x float>, <4 x float> *%a0, align 16
   %2 = fadd <4 x float> %1, %1
   store <4 x float> %2, <4 x float> *%a1, align 16
   ret void
 }
 
 ; TODO (v)movhlps
 

 ; BTVER2-LABEL: test_movhlps:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1] sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movhlps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm1[1],xmm0[1] sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x float> %a0, <4 x float> %a1, <4 x i32> <i32 6, i32 7, i32 2, i32 3>
   ret <4 x float> %1
 }
 
 ; TODO (v)movhps
 
 define void @test_movhps(<4 x float> %a0, <4 x float> %a1, x86_mmx *%a2) {
 ; GENERIC-LABEL: test_movhps:

 ; BTVER2-NEXT:    vpextrq $1, %xmm0, (%rdi) # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movhps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovhpd {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vpextrq $1, %xmm0, (%rdi) # sched: [8:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast x86_mmx* %a2 to <2 x float>*
   %2 = load <2 x float>, <2 x float> *%1, align 8
   %3 = shufflevector <2 x float> %2, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %4 = shufflevector <4 x float> %a1, <4 x float> %3, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
   %5 = fadd <4 x float> %a0, %4
   %6 = shufflevector <4 x float> %5, <4 x float> undef, <2 x i32> <i32 2, i32 3>
   store <2 x float> %6, <2 x float>* %1
   ret void

 ; BTVER2-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0] sched: [1:0.50]
 ; BTVER2-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movlhps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0] sched: [1:0.50]
 ; ZNVER1-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x float> %a0, <4 x float> %a1, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
   %2 = fadd <4 x float> %a1, %1
   ret <4 x float> %2
 }
 
 define void @test_movlps(<4 x float> %a0, <4 x float> %a1, x86_mmx *%a2) {
 ; GENERIC-LABEL: test_movlps:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    vmovlps %xmm0, (%rdi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movlps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovlpd {{.*#+}} xmm1 = mem[0],xmm1[1] sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmovlps %xmm0, (%rdi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast x86_mmx* %a2 to <2 x float>*
   %2 = load <2 x float>, <2 x float> *%1, align 8
   %3 = shufflevector <2 x float> %2, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %4 = shufflevector <4 x float> %a1, <4 x float> %3, <4 x i32> <i32 4, i32 5, i32 2, i32 3>
   %5 = fadd <4 x float> %a0, %4
   %6 = shufflevector <4 x float> %5, <4 x float> undef, <2 x i32> <i32 0, i32 1>
   store <2 x float> %6, <2 x float>* %1
   ret void

 ; BTVER2-LABEL: test_movmskps:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vmovmskps %xmm0, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movmskps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovmskps %xmm0, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.sse.movmsk.ps(<4 x float> %a0)
   ret i32 %1
 }
 declare i32 @llvm.x86.sse.movmsk.ps(<4 x float>) nounwind readnone
 
 define void @test_movntps(<4 x float> %a0, <4 x float> *%a1) {
 ; GENERIC-LABEL: test_movntps:
 ; GENERIC:       # BB#0:

 ; BTVER2-LABEL: test_movntps:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vmovntps %xmm0, (%rdi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movntps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovntps %xmm0, (%rdi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   store <4 x float> %a0, <4 x float> *%a1, align 16, !nontemporal !0
   ret void
 }
 
 define void @test_movss_mem(float* %a0, float* %a1) {
 ; GENERIC-LABEL: test_movss_mem:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero sched: [6:0.50]

 ; BTVER2-NEXT:    vmovss %xmm0, (%rsi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movss_mem:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddss %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmovss %xmm0, (%rsi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load float, float* %a0, align 1
   %2 = fadd float %1, %1
   store float %2, float *%a1, align 1
   ret void
 }
 
 define <4 x float> @test_movss_reg(<4 x float> %a0, <4 x float> %a1) {
 ; GENERIC-LABEL: test_movss_reg:

 ; BTVER2-LABEL: test_movss_reg:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3] sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movss_reg:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3] sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x float> %a0, <4 x float> %a1, <4 x i32> <i32 4, i32 1, i32 2, i32 3>
   ret <4 x float> %1
 }
 
 define void @test_movups(<4 x float> *%a0, <4 x float> *%a1) {
 ; GENERIC-LABEL: test_movups:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    movups (%rdi), %xmm0 # sched: [6:0.50]

 ; BTVER2-NEXT:    vmovups %xmm0, (%rsi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movups:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovups (%rdi), %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddps %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmovups %xmm0, (%rsi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load <4 x float>, <4 x float> *%a0, align 1
   %2 = fadd <4 x float> %1, %1
   store <4 x float> %2, <4 x float> *%a1, align 1
   ret void
 }
 
 define <4 x float> @test_mulps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {
 ; GENERIC-LABEL: test_mulps:

 ; BTVER2-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vmulps (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_mulps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmulps %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vmulps (%rdi), %xmm0, %xmm0 # sched: [12:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fmul <4 x float> %a0, %a1
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = fmul <4 x float> %1, %2
   ret <4 x float> %3
 }
 
 define float @test_mulss(float %a0, float %a1, float *%a2) {
 ; GENERIC-LABEL: test_mulss:

 ; BTVER2-NEXT:    vmulss %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vmulss (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_mulss:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmulss %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vmulss (%rdi), %xmm0, %xmm0 # sched: [12:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fmul float %a0, %a1
   %2 = load float, float *%a2, align 4
   %3 = fmul float %1, %2
   ret float %3
 }
 
 define <4 x float> @test_orps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {
 ; GENERIC-LABEL: test_orps:

 ; BTVER2-NEXT:    vorps %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vorps (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_orps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vorps %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vorps (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast <4 x float> %a0 to <4 x i32>
   %2 = bitcast <4 x float> %a1 to <4 x i32>
   %3 = or <4 x i32> %1, %2
   %4 = load <4 x float>, <4 x float> *%a2, align 16
   %5 = bitcast <4 x float> %4 to <4 x i32>
   %6 = or <4 x i32> %3, %5
   %7 = bitcast <4 x i32> %6 to <4 x float>
   ret <4 x float> %7

 ; BTVER2-LABEL: test_prefetchnta:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    prefetchnta (%rdi) # sched: [5:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_prefetchnta:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    prefetchnta (%rdi) # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   call void @llvm.prefetch(i8* %a0, i32 0, i32 0, i32 1)
   ret void
 }
 declare void @llvm.prefetch(i8* nocapture, i32, i32, i32) nounwind readnone
 
 define <4 x float> @test_rcpps(<4 x float> %a0, <4 x float> *%a1) {
 ; GENERIC-LABEL: test_rcpps:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_rcpps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vrcpps (%rdi), %xmm1 # sched: [12:0.50]
 ; ZNVER1-NEXT:    vrcpps %xmm0, %xmm0 # sched: [5:0.50]
 ; ZNVER1-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.sse.rcp.ps(<4 x float> %a0)
   %2 = load <4 x float>, <4 x float> *%a1, align 16
   %3 = call <4 x float> @llvm.x86.sse.rcp.ps(<4 x float> %2)
   %4 = fadd <4 x float> %1, %3
   ret <4 x float> %4
 }
 declare <4 x float> @llvm.x86.sse.rcp.ps(<4 x float>) nounwind readnone
 

 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_rcpss:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [8:0.50]
 ; ZNVER1-NEXT:    vrcpss %xmm0, %xmm0, %xmm0 # sched: [12:0.50]
 ; ZNVER1-NEXT:    vrcpss %xmm1, %xmm1, %xmm1 # sched: [12:0.50]
 ; ZNVER1-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = insertelement <4 x float> undef, float %a0, i32 0
   %2 = call <4 x float> @llvm.x86.sse.rcp.ss(<4 x float> %1)
   %3 = load float, float *%a1, align 4
   %4 = insertelement <4 x float> undef, float %3, i32 0
   %5 = call <4 x float> @llvm.x86.sse.rcp.ss(<4 x float> %4)
   %6 = fadd <4 x float> %2, %5
   ret <4 x float> %6
 }

 ; BTVER2-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_rsqrtps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vrsqrtps (%rdi), %xmm1 # sched: [12:0.50]
 ; ZNVER1-NEXT:    vrsqrtps %xmm0, %xmm0 # sched: [5:0.50]
 ; ZNVER1-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.sse.rsqrt.ps(<4 x float> %a0)
   %2 = load <4 x float>, <4 x float> *%a1, align 16
   %3 = call <4 x float> @llvm.x86.sse.rsqrt.ps(<4 x float> %2)
   %4 = fadd <4 x float> %1, %3
   ret <4 x float> %4
 }
 declare <4 x float> @llvm.x86.sse.rsqrt.ps(<4 x float>) nounwind readnone
 

 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_rsqrtss:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [8:0.50]
 ; ZNVER1-NEXT:    vrsqrtss %xmm0, %xmm0, %xmm0 # sched: [12:0.50]
 ; ZNVER1-NEXT:    vrsqrtss %xmm1, %xmm1, %xmm1 # sched: [12:0.50]
 ; ZNVER1-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = insertelement <4 x float> undef, float %a0, i32 0
   %2 = call <4 x float> @llvm.x86.sse.rsqrt.ss(<4 x float> %1)
   %3 = load float, float *%a1, align 4
   %4 = insertelement <4 x float> undef, float %3, i32 0
   %5 = call <4 x float> @llvm.x86.sse.rsqrt.ss(<4 x float> %4)
   %6 = fadd <4 x float> %2, %5
   ret <4 x float> %6
 }

 ; BTVER2-LABEL: test_sfence:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    sfence # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_sfence:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    sfence # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   call void @llvm.x86.sse.sfence()
   ret void
 }
 declare void @llvm.x86.sse.sfence() nounwind readnone
 
 define <4 x float> @test_shufps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) nounwind {
 ; GENERIC-LABEL: test_shufps:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0] sched: [1:0.50]
 ; BTVER2-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,3],mem[0,0] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_shufps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,0],xmm1[0,0] sched: [1:0.50]
 ; ZNVER1-NEXT:    vshufps {{.*#+}} xmm0 = xmm0[0,3],mem[0,0] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x float> %a0, <4 x float> %a1, <4 x i32> <i32 0, i32 0, i32 4, i32 4>
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = shufflevector <4 x float> %1, <4 x float> %2, <4 x i32> <i32 0, i32 3, i32 4, i32 4>
   ret <4 x float> %3
 }
 
 define <4 x float> @test_sqrtps(<4 x float> %a0, <4 x float> *%a1) {
 ; GENERIC-LABEL: test_sqrtps:

 ; BTVER2-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_sqrtps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vsqrtps (%rdi), %xmm1 # sched: [27:1.00]
 ; ZNVER1-NEXT:    vsqrtps %xmm0, %xmm0 # sched: [20:1.00]
 ; ZNVER1-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.sse.sqrt.ps(<4 x float> %a0)
   %2 = load <4 x float>, <4 x float> *%a1, align 16
   %3 = call <4 x float> @llvm.x86.sse.sqrt.ps(<4 x float> %2)
   %4 = fadd <4 x float> %1, %3
   ret <4 x float> %4
 }
 declare <4 x float> @llvm.x86.sse.sqrt.ps(<4 x float>) nounwind readnone
 

 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_sqrtss:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovaps (%rdi), %xmm1 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vsqrtss %xmm0, %xmm0, %xmm0 # sched: [27:1.00]
 ; ZNVER1-NEXT:    vsqrtss %xmm1, %xmm1, %xmm1 # sched: [27:1.00]
 ; ZNVER1-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.sse.sqrt.ss(<4 x float> %a0)
   %2 = load <4 x float>, <4 x float> *%a1, align 16
   %3 = call <4 x float> @llvm.x86.sse.sqrt.ss(<4 x float> %2)
   %4 = fadd <4 x float> %1, %3
   ret <4 x float> %4
 }
 declare <4 x float> @llvm.x86.sse.sqrt.ss(<4 x float>) nounwind readnone
 

 ; BTVER2-NEXT:    vstmxcsr -{{[0-9]+}}(%rsp) # sched: [1:1.00]
 ; BTVER2-NEXT:    movl -{{[0-9]+}}(%rsp), %eax # sched: [5:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_stmxcsr:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vstmxcsr -{{[0-9]+}}(%rsp) # sched: [1:0.50]
 ; ZNVER1-NEXT:    movl -{{[0-9]+}}(%rsp), %eax # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = alloca i32, align 4
   %2 = bitcast i32* %1 to i8*
   call void @llvm.x86.sse.stmxcsr(i8* %2)
   %3 = load i32, i32* %1, align 4
   ret i32 %3
 }
 declare void @llvm.x86.sse.stmxcsr(i8*) nounwind readnone
 

 ; BTVER2-NEXT:    vsubps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vsubps (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_subps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vsubps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vsubps (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fsub <4 x float> %a0, %a1
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = fsub <4 x float> %1, %2
   ret <4 x float> %3
 }
 
 define float @test_subss(float %a0, float %a1, float *%a2) {
 ; GENERIC-LABEL: test_subss:

 ; BTVER2-NEXT:    vsubss %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vsubss (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_subss:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vsubss %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vsubss (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fsub float %a0, %a1
   %2 = load float, float *%a2, align 4
   %3 = fsub float %1, %2
   ret float %3
 }
 
 define i32 @test_ucomiss(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {
 ; GENERIC-LABEL: test_ucomiss:

 ; ZNVER1-NEXT:    sete %cl # sched: [1:0.25]
 ; ZNVER1-NEXT:    andb %al, %cl # sched: [1:0.25]
 ; ZNVER1-NEXT:    vucomiss (%rdi), %xmm0 # sched: [10:1.00]
 ; ZNVER1-NEXT:    setnp %al # sched: [1:0.25]
 ; ZNVER1-NEXT:    sete %dl # sched: [1:0.25]
 ; ZNVER1-NEXT:    andb %al, %dl # sched: [1:0.25]
 ; ZNVER1-NEXT:    orb %cl, %dl # sched: [1:0.25]
 ; ZNVER1-NEXT:    movzbl %dl, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.sse.ucomieq.ss(<4 x float> %a0, <4 x float> %a1)
   %2 = load <4 x float>, <4 x float> *%a2, align 4
   %3 = call i32 @llvm.x86.sse.ucomieq.ss(<4 x float> %a0, <4 x float> %2)
   %4 = or i32 %1, %3
   ret i32 %4
 }
 declare i32 @llvm.x86.sse.ucomieq.ss(<4 x float>, <4 x float>) nounwind readnone
 

 ; BTVER2-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3] sched: [1:0.50]
 ; BTVER2-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],mem[2],xmm0[3],mem[3] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_unpckhps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3] sched: [1:0.50]
 ; ZNVER1-NEXT:    vunpckhps {{.*#+}} xmm0 = xmm0[2],mem[2],xmm0[3],mem[3] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x float> %a0, <4 x float> %a1, <4 x i32> <i32 2, i32 6, i32 3, i32 7>
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = shufflevector <4 x float> %1, <4 x float> %2, <4 x i32> <i32 2, i32 6, i32 3, i32 7>
   ret <4 x float> %3
 }
 
 define <4 x float> @test_unpcklps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {
 ; GENERIC-LABEL: test_unpcklps:

 ; BTVER2-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] sched: [1:0.50]
 ; BTVER2-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_unpcklps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] sched: [1:0.50]
 ; ZNVER1-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x float> %a0, <4 x float> %a1, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = shufflevector <4 x float> %1, <4 x float> %2, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
   ret <4 x float> %3
 }
 
 define <4 x float> @test_xorps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {
 ; GENERIC-LABEL: test_xorps:

 ; BTVER2-NEXT:    vxorps %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vxorps (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_xorps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vxorps %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vxorps (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast <4 x float> %a0 to <4 x i32>
   %2 = bitcast <4 x float> %a1 to <4 x i32>
   %3 = xor <4 x i32> %1, %2
   %4 = load <4 x float>, <4 x float> *%a2, align 16
   %5 = bitcast <4 x float> %4 to <4 x i32>
   %6 = xor <4 x i32> %3, %5
   %7 = bitcast <4 x i32> %6 to <4 x float>
   ret <4 x float> %7
 }
 
 !0 = !{i32 1}

test/CodeGen/X86/sse2-schedule.ll

 ; BTVER2-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vaddpd (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_addpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vaddpd (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fadd <2 x double> %a0, %a1
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = fadd <2 x double> %1, %2
   ret <2 x double> %3
 }
 
 define double @test_addsd(double %a0, double %a1, double *%a2) {
 ; GENERIC-LABEL: test_addsd:

 ; BTVER2-NEXT:    vaddsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vaddsd (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_addsd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vaddsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vaddsd (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fadd double %a0, %a1
   %2 = load double, double *%a2, align 8
   %3 = fadd double %1, %2
   ret double %3
 }
 
 define <2 x double> @test_andpd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) {
 ; GENERIC-LABEL: test_andpd:

 ; BTVER2-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_andpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vandpd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vandpd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast <2 x double> %a0 to <4 x i32>
   %2 = bitcast <2 x double> %a1 to <4 x i32>
   %3 = and <4 x i32> %1, %2
   %4 = load <2 x double>, <2 x double> *%a2, align 16
   %5 = bitcast <2 x double> %4 to <4 x i32>
   %6 = and <4 x i32> %3, %5
   %7 = bitcast <4 x i32> %6 to <2 x double>
   %8 = fadd <2 x double> %a1, %7

 ; BTVER2-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_andnotpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vandnpd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vandnpd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast <2 x double> %a0 to <4 x i32>
   %2 = bitcast <2 x double> %a1 to <4 x i32>
   %3 = xor <4 x i32> %1, <i32 -1, i32 -1, i32 -1, i32 -1>
   %4 = and <4 x i32> %3, %2
   %5 = load <2 x double>, <2 x double> *%a2, align 16
   %6 = bitcast <2 x double> %5 to <4 x i32>
   %7 = xor <4 x i32> %4, <i32 -1, i32 -1, i32 -1, i32 -1>
   %8 = and <4 x i32> %6, %7

 ; BTVER2-NEXT:    vorpd %xmm0, %xmm1, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cmppd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcmpeqpd %xmm1, %xmm0, %xmm1 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vcmpeqpd (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
 ; ZNVER1-NEXT:    vorpd %xmm0, %xmm1, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fcmp oeq <2 x double> %a0, %a1
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = fcmp oeq <2 x double> %a0, %2
   %4 = or <2 x i1> %1, %3
   %5 = sext <2 x i1> %4 to <2 x i64>
   %6 = bitcast <2 x i64> %5 to <2 x double>
   ret <2 x double> %6
 }

 ; BTVER2-NEXT:    vcmpeqsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vcmpeqsd (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cmpsd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcmpeqsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vcmpeqsd (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = insertelement <2 x double> undef, double %a0, i32 0
   %2 = insertelement <2 x double> undef, double %a1, i32 0
   %3 = call <2 x double> @llvm.x86.sse2.cmp.sd(<2 x double> %1, <2 x double> %2, i8 0)
   %4 = load double, double *%a2, align 8
   %5 = insertelement <2 x double> undef, double %4, i32 0
   %6 = call <2 x double> @llvm.x86.sse2.cmp.sd(<2 x double> %3, <2 x double> %5, i8 0)
   %7 = extractelement <2 x double> %6, i32 0
   ret double %7

 ; ZNVER1-NEXT:    sete %cl # sched: [1:0.25]
 ; ZNVER1-NEXT:    andb %al, %cl # sched: [1:0.25]
 ; ZNVER1-NEXT:    vcomisd (%rdi), %xmm0 # sched: [10:1.00]
 ; ZNVER1-NEXT:    setnp %al # sched: [1:0.25]
 ; ZNVER1-NEXT:    sete %dl # sched: [1:0.25]
 ; ZNVER1-NEXT:    andb %al, %dl # sched: [1:0.25]
 ; ZNVER1-NEXT:    orb %cl, %dl # sched: [1:0.25]
 ; ZNVER1-NEXT:    movzbl %dl, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.sse2.comieq.sd(<2 x double> %a0, <2 x double> %a1)
   %2 = load <2 x double>, <2 x double> *%a2, align 8
   %3 = call i32 @llvm.x86.sse2.comieq.sd(<2 x double> %a0, <2 x double> %2)
   %4 = or i32 %1, %3
   ret i32 %4
 }
 declare i32 @llvm.x86.sse2.comieq.sd(<2 x double>, <2 x double>) nounwind readnone
 

 ; BTVER2-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtdq2pd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtdq2pd (%rdi), %xmm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvtdq2pd %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x i32> %a0, <4 x i32> undef, <2 x i32> <i32 0, i32 1>
   %2 = sitofp <2 x i32> %1 to <2 x double>
   %3 = load <4 x i32>, <4 x i32>*%a1, align 16
   %4 = shufflevector <4 x i32> %3, <4 x i32> undef, <2 x i32> <i32 0, i32 1>
   %5 = sitofp <2 x i32> %4 to <2 x double>
   %6 = fadd <2 x double> %2, %5
   ret <2 x double> %6
 }

 ; BTVER2-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtdq2ps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtdq2ps (%rdi), %xmm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvtdq2ps %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = sitofp <4 x i32> %a0 to <4 x float>
   %2 = load <4 x i32>, <4 x i32>*%a1, align 16
   %3 = sitofp <4 x i32> %2 to <4 x float>
   %4 = fadd <4 x float> %1, %3
   ret <4 x float> %4
 }
 
 define <4 x i32> @test_cvtpd2dq(<2 x double> %a0, <2 x double> *%a1) {

 ; BTVER2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtpd2dq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtpd2dqx (%rdi), %xmm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvtpd2dq %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x i32> @llvm.x86.sse2.cvtpd2dq(<2 x double> %a0)
   %2 = load <2 x double>, <2 x double> *%a1, align 16
   %3 = call <4 x i32> @llvm.x86.sse2.cvtpd2dq(<2 x double> %2)
   %4 = add <4 x i32> %1, %3
   ret <4 x i32> %4
 }
 declare <4 x i32> @llvm.x86.sse2.cvtpd2dq(<2 x double>) nounwind readnone
 

 ; BTVER2-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtpd2ps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtpd2psx (%rdi), %xmm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvtpd2ps %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.sse2.cvtpd2ps(<2 x double> %a0)
   %2 = load <2 x double>, <2 x double> *%a1, align 16
   %3 = call <4 x float> @llvm.x86.sse2.cvtpd2ps(<2 x double> %2)
   %4 = fadd <4 x float> %1, %3
   ret <4 x float> %4
 }
 declare <4 x float> @llvm.x86.sse2.cvtpd2ps(<2 x double>) nounwind readnone
 

 ; BTVER2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtps2dq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtps2dq (%rdi), %xmm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvtps2dq %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x i32> @llvm.x86.sse2.cvtps2dq(<4 x float> %a0)
   %2 = load <4 x float>, <4 x float> *%a1, align 16
   %3 = call <4 x i32> @llvm.x86.sse2.cvtps2dq(<4 x float> %2)
   %4 = add <4 x i32> %1, %3
   ret <4 x i32> %4
 }
 declare <4 x i32> @llvm.x86.sse2.cvtps2dq(<4 x float>) nounwind readnone
 

 ; BTVER2-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtps2pd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtps2pd (%rdi), %xmm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvtps2pd %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x float> %a0, <4 x float> undef, <2 x i32> <i32 0, i32 1>
   %2 = fpext <2 x float> %1 to <2 x double>
   %3 = load <4 x float>, <4 x float> *%a1, align 16
   %4 = shufflevector <4 x float> %3, <4 x float> undef, <2 x i32> <i32 0, i32 1>
   %5 = fpext <2 x float> %4 to <2 x double>
   %6 = fadd <2 x double> %2, %5
   ret <2 x double> %6
 }

 ; BTVER2-NEXT:    addl %ecx, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtsd2si:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtsd2si (%rdi), %eax # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvtsd2si %xmm0, %ecx # sched: [5:1.00]
 ; ZNVER1-NEXT:    addl %ecx, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = insertelement <2 x double> undef, double %a0, i32 0
   %2 = call i32 @llvm.x86.sse2.cvtsd2si(<2 x double> %1)
   %3 = load double, double *%a1, align 8
   %4 = insertelement <2 x double> undef, double %3, i32 0
   %5 = call i32 @llvm.x86.sse2.cvtsd2si(<2 x double> %4)
   %6 = add i32 %2, %5
   ret i32 %6
 }

 ; BTVER2-NEXT:    addq %rcx, %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtsd2siq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtsd2si (%rdi), %rax # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvtsd2si %xmm0, %rcx # sched: [5:1.00]
 ; ZNVER1-NEXT:    addq %rcx, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = insertelement <2 x double> undef, double %a0, i32 0
   %2 = call i64 @llvm.x86.sse2.cvtsd2si64(<2 x double> %1)
   %3 = load double, double *%a1, align 8
   %4 = insertelement <2 x double> undef, double %3, i32 0
   %5 = call i64 @llvm.x86.sse2.cvtsd2si64(<2 x double> %4)
   %6 = add i64 %2, %5
   ret i64 %6
 }

 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtsd2ss:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero sched: [8:0.50]
 ; ZNVER1-NEXT:    vcvtsd2ss %xmm0, %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vcvtsd2ss %xmm1, %xmm1, %xmm1 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vaddss %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fptrunc double %a0 to float
   %2 = load double, double *%a1, align 8
   %3 = fptrunc double %2 to float
   %4 = fadd float %1, %3
   ret float %4
 }
 
 define double @test_cvtsi2sd(i32 %a0, i32 *%a1) {

 ; BTVER2-NEXT:    vaddsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtsi2sd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtsi2sdl %edi, %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vcvtsi2sdl (%rsi), %xmm1, %xmm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vaddsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = sitofp i32 %a0 to double
   %2 = load i32, i32 *%a1, align 8
   %3 = sitofp i32 %2 to double
   %4 = fadd double %1, %3
   ret double %4
 }
 
 define double @test_cvtsi2sdq(i64 %a0, i64 *%a1) {

 ; BTVER2-NEXT:    vaddsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtsi2sdq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvtsi2sdq %rdi, %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vcvtsi2sdq (%rsi), %xmm1, %xmm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vaddsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = sitofp i64 %a0 to double
   %2 = load i64, i64 *%a1, align 8
   %3 = sitofp i64 %2 to double
   %4 = fadd double %1, %3
   ret double %4
 }
 
 ; TODO - cvtss2sd_m

 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvtss2sd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero sched: [8:0.50]
 ; ZNVER1-NEXT:    vcvtss2sd %xmm0, %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vcvtss2sd %xmm1, %xmm1, %xmm1 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vaddsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fpext float %a0 to double
   %2 = load float, float *%a1, align 4
   %3 = fpext float %2 to double
   %4 = fadd double %1, %3
   ret double %4
 }
 
 define <4 x i32> @test_cvttpd2dq(<2 x double> %a0, <2 x double> *%a1) {

 ; BTVER2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvttpd2dq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvttpd2dqx (%rdi), %xmm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvttpd2dq %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fptosi <2 x double> %a0 to <2 x i32>
   %2 = shufflevector <2 x i32> %1, <2 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %3 = load <2 x double>, <2 x double> *%a1, align 16
   %4 = fptosi <2 x double> %3 to <2 x i32>
   %5 = shufflevector <2 x i32> %4, <2 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %6 = add <4 x i32> %2, %5
   ret <4 x i32> %6
 }

 ; BTVER2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvttps2dq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvttps2dq (%rdi), %xmm1 # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvttps2dq %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fptosi <4 x float> %a0 to <4 x i32>
   %2 = load <4 x float>, <4 x float> *%a1, align 16
   %3 = fptosi <4 x float> %2 to <4 x i32>
   %4 = add <4 x i32> %1, %3
   ret <4 x i32> %4
 }
 
 define i32 @test_cvttsd2si(double %a0, double *%a1) {

 ; BTVER2-NEXT:    addl %ecx, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvttsd2si:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvttsd2si (%rdi), %eax # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvttsd2si %xmm0, %ecx # sched: [5:1.00]
 ; ZNVER1-NEXT:    addl %ecx, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fptosi double %a0 to i32
   %2 = load double, double *%a1, align 8
   %3 = fptosi double %2 to i32
   %4 = add i32 %1, %3
   ret i32 %4
 }
 
 define i64 @test_cvttsd2siq(double %a0, double *%a1) {

 ; BTVER2-NEXT:    addq %rcx, %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_cvttsd2siq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vcvttsd2si (%rdi), %rax # sched: [12:1.00]
 ; ZNVER1-NEXT:    vcvttsd2si %xmm0, %rcx # sched: [5:1.00]
 ; ZNVER1-NEXT:    addq %rcx, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fptosi double %a0 to i64
   %2 = load double, double *%a1, align 8
   %3 = fptosi double %2 to i64
   %4 = add i64 %1, %3
   ret i64 %4
 }
 
 define <2 x double> @test_divpd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) {

 ; BTVER2-NEXT:    vdivpd %xmm1, %xmm0, %xmm0 # sched: [19:19.00]
 ; BTVER2-NEXT:    vdivpd (%rdi), %xmm0, %xmm0 # sched: [24:19.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_divpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vdivpd %xmm1, %xmm0, %xmm0 # sched: [15:1.00]
 ; ZNVER1-NEXT:    vdivpd (%rdi), %xmm0, %xmm0 # sched: [22:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fdiv <2 x double> %a0, %a1
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = fdiv <2 x double> %1, %2
   ret <2 x double> %3
 }
 
 define double @test_divsd(double %a0, double %a1, double *%a2) {
 ; GENERIC-LABEL: test_divsd:

 ; BTVER2-NEXT:    vdivsd %xmm1, %xmm0, %xmm0 # sched: [19:19.00]
 ; BTVER2-NEXT:    vdivsd (%rdi), %xmm0, %xmm0 # sched: [24:19.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_divsd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vdivsd %xmm1, %xmm0, %xmm0 # sched: [15:1.00]
 ; ZNVER1-NEXT:    vdivsd (%rdi), %xmm0, %xmm0 # sched: [22:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fdiv double %a0, %a1
   %2 = load double, double *%a2, align 8
   %3 = fdiv double %1, %2
   ret double %3
 }
 
 define void @test_lfence() {
 ; GENERIC-LABEL: test_lfence:

 ; BTVER2-LABEL: test_lfence:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    lfence # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lfence:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    lfence # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   call void @llvm.x86.sse2.lfence()
   ret void
 }
 declare void @llvm.x86.sse2.lfence() nounwind readnone
 
 define void @test_mfence() {
 ; GENERIC-LABEL: test_mfence:
 ; GENERIC:       # BB#0:

 ; BTVER2-LABEL: test_mfence:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    mfence # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_mfence:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    mfence # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   call void @llvm.x86.sse2.mfence()
   ret void
 }
 declare void @llvm.x86.sse2.mfence() nounwind readnone
 
 define void @test_maskmovdqu(<16 x i8> %a0, <16 x i8> %a1, i8* %a2) {
 ; GENERIC-LABEL: test_maskmovdqu:
 ; GENERIC:       # BB#0:

 ; BTVER2-LABEL: test_maskmovdqu:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vmaskmovdqu %xmm1, %xmm0 # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_maskmovdqu:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmaskmovdqu %xmm1, %xmm0 # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   call void @llvm.x86.sse2.maskmov.dqu(<16 x i8> %a0, <16 x i8> %a1, i8* %a2)
   ret void
 }
 declare void @llvm.x86.sse2.maskmov.dqu(<16 x i8>, <16 x i8>, i8*) nounwind
 
 define <2 x double> @test_maxpd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) {
 ; GENERIC-LABEL: test_maxpd:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    vmaxpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vmaxpd (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_maxpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmaxpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmaxpd (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x double> @llvm.x86.sse2.max.pd(<2 x double> %a0, <2 x double> %a1)
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = call <2 x double> @llvm.x86.sse2.max.pd(<2 x double> %1, <2 x double> %2)
   ret <2 x double> %3
 }
 declare <2 x double> @llvm.x86.sse2.max.pd(<2 x double>, <2 x double>) nounwind readnone
 
 define <2 x double> @test_maxsd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) {

 ; BTVER2-NEXT:    vmaxsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vmaxsd (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_maxsd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmaxsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmaxsd (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x double> @llvm.x86.sse2.max.sd(<2 x double> %a0, <2 x double> %a1)
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = call <2 x double> @llvm.x86.sse2.max.sd(<2 x double> %1, <2 x double> %2)
   ret <2 x double> %3
 }
 declare <2 x double> @llvm.x86.sse2.max.sd(<2 x double>, <2 x double>) nounwind readnone
 
 define <2 x double> @test_minpd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) {

 ; BTVER2-NEXT:    vminpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vminpd (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_minpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vminpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vminpd (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x double> @llvm.x86.sse2.min.pd(<2 x double> %a0, <2 x double> %a1)
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = call <2 x double> @llvm.x86.sse2.min.pd(<2 x double> %1, <2 x double> %2)
   ret <2 x double> %3
 }
 declare <2 x double> @llvm.x86.sse2.min.pd(<2 x double>, <2 x double>) nounwind readnone
 
 define <2 x double> @test_minsd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) {

 ; BTVER2-NEXT:    vminsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vminsd (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_minsd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vminsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vminsd (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x double> @llvm.x86.sse2.min.sd(<2 x double> %a0, <2 x double> %a1)
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = call <2 x double> @llvm.x86.sse2.min.sd(<2 x double> %1, <2 x double> %2)
   ret <2 x double> %3
 }
 declare <2 x double> @llvm.x86.sse2.min.sd(<2 x double>, <2 x double>) nounwind readnone
 
 define void @test_movapd(<2 x double> *%a0, <2 x double> *%a1) {

 ; BTVER2-NEXT:    vmovapd %xmm0, (%rsi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movapd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovapd (%rdi), %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmovapd %xmm0, (%rsi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load <2 x double>, <2 x double> *%a0, align 16
   %2 = fadd <2 x double> %1, %1
   store <2 x double> %2, <2 x double> *%a1, align 16
   ret void
 }
 
 define void @test_movdqa(<2 x i64> *%a0, <2 x i64> *%a1) {
 ; GENERIC-LABEL: test_movdqa:

 ; BTVER2-NEXT:    vmovdqa %xmm0, (%rsi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movdqa:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovdqa (%rdi), %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpaddq %xmm0, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vmovdqa %xmm0, (%rsi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load <2 x i64>, <2 x i64> *%a0, align 16
   %2 = add <2 x i64> %1, %1
   store <2 x i64> %2, <2 x i64> *%a1, align 16
   ret void
 }
 
 define void @test_movdqu(<2 x i64> *%a0, <2 x i64> *%a1) {
 ; GENERIC-LABEL: test_movdqu:

 ; BTVER2-NEXT:    vmovdqu %xmm0, (%rsi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movdqu:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovdqu (%rdi), %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpaddq %xmm0, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vmovdqu %xmm0, (%rsi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load <2 x i64>, <2 x i64> *%a0, align 1
   %2 = add <2 x i64> %1, %1
   store <2 x i64> %2, <2 x i64> *%a1, align 1
   ret void
 }
 
 define i32 @test_movd(<4 x i32> %a0, i32 %a1, i32 *%a2) {
 ; GENERIC-LABEL: test_movd:

 ; ZNVER1-LABEL: test_movd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovd {{.*#+}} xmm2 = mem[0],zero,zero,zero sched: [8:0.50]
 ; ZNVER1-NEXT:    vmovd %edi, %xmm1 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddd %xmm1, %xmm0, %xmm1 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vmovd %xmm1, (%rsi) # sched: [1:0.50]
 ; ZNVER1-NEXT:    vpaddd %xmm2, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vmovd %xmm0, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = insertelement <4 x i32> undef, i32 %a1, i32 0
   %2 = load i32, i32 *%a2
   %3 = insertelement <4 x i32> undef, i32 %2, i32 0
   %4 = add <4 x i32> %a0, %1
   %5 = add <4 x i32> %a0, %3
   %6 = extractelement <4 x i32> %4, i32 0
   %7 = extractelement <4 x i32> %5, i32 0
   store i32 %6, i32* %a2

 ; ZNVER1-LABEL: test_movd_64:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovq {{.*#+}} xmm2 = mem[0],zero sched: [8:0.50]
 ; ZNVER1-NEXT:    vmovq %rdi, %xmm1 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddq %xmm1, %xmm0, %xmm1 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vmovq %xmm1, (%rsi) # sched: [1:0.50]
 ; ZNVER1-NEXT:    vpaddq %xmm2, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vmovq %xmm0, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = insertelement <2 x i64> undef, i64 %a1, i64 0
   %2 = load i64, i64 *%a2
   %3 = insertelement <2 x i64> undef, i64 %2, i64 0
   %4 = add <2 x i64> %a0, %1
   %5 = add <2 x i64> %a0, %3
   %6 = extractelement <2 x i64> %4, i64 0
   %7 = extractelement <2 x i64> %5, i64 0
   store i64 %6, i64* %a2

 ; BTVER2-NEXT:    vmovhpd %xmm0, (%rdi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movhpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovhpd {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmovhpd %xmm0, (%rdi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast x86_mmx* %a2 to double*
   %2 = load double, double *%1, align 8
   %3 = insertelement <2 x double> %a1, double %2, i32 1
   %4 = fadd <2 x double> %a0, %3
   %5 = extractelement <2 x double> %4, i32 1
   store double %5, double* %1
   ret void
 }

 ; BTVER2-NEXT:    vmovlpd %xmm0, (%rdi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movlpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovlpd {{.*#+}} xmm1 = mem[0],xmm1[1] sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmovlpd %xmm0, (%rdi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast x86_mmx* %a2 to double*
   %2 = load double, double *%1, align 8
   %3 = insertelement <2 x double> %a1, double %2, i32 0
   %4 = fadd <2 x double> %a0, %3
   %5 = extractelement <2 x double> %4, i32 0
   store double %5, double* %1
   ret void
 }

 ; BTVER2-LABEL: test_movmskpd:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vmovmskpd %xmm0, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movmskpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovmskpd %xmm0, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.sse2.movmsk.pd(<2 x double> %a0)
   ret i32 %1
 }
 declare i32 @llvm.x86.sse2.movmsk.pd(<2 x double>) nounwind readnone
 
 define void @test_movntdqa(<2 x i64> %a0, <2 x i64> *%a1) {
 ; GENERIC-LABEL: test_movntdqa:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    vpaddq %xmm0, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vmovntdq %xmm0, (%rdi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movntdqa:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpaddq %xmm0, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vmovntdq %xmm0, (%rdi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = add <2 x i64> %a0, %a0
   store <2 x i64> %1, <2 x i64> *%a1, align 16, !nontemporal !0
   ret void
 }
 
 define void @test_movntpd(<2 x double> %a0, <2 x double> *%a1) {
 ; GENERIC-LABEL: test_movntpd:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    vaddpd %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vmovntpd %xmm0, (%rdi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movntpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vaddpd %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmovntpd %xmm0, (%rdi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fadd <2 x double> %a0, %a0
   store <2 x double> %1, <2 x double> *%a1, align 16, !nontemporal !0
   ret void
 }
 
 define <2 x i64> @test_movq_mem(<2 x i64> %a0, i64 *%a1) {
 ; GENERIC-LABEL: test_movq_mem:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    vmovq %xmm0, (%rdi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movq_mem:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovq {{.*#+}} xmm1 = mem[0],zero sched: [8:0.50]
 ; ZNVER1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vmovq %xmm0, (%rdi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load i64, i64* %a1, align 1
   %2 = insertelement <2 x i64> zeroinitializer, i64 %1, i32 0
   %3 = add <2 x i64> %a0, %2
   %4 = extractelement <2 x i64> %3, i32 0
   store i64 %4, i64 *%a1, align 1
   ret <2 x i64> %3
 }
 

 ; BTVER2-NEXT:    vmovq {{.*#+}} xmm0 = xmm0[0],zero sched: [1:0.50]
 ; BTVER2-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movq_reg:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovq {{.*#+}} xmm0 = xmm0[0],zero sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddq %xmm0, %xmm1, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <2 x i64> %a0, <2 x i64> zeroinitializer, <2 x i32> <i32 0, i32 2>
   %2 = add <2 x i64> %a1, %1
   ret <2 x i64> %2
 }
 
 define void @test_movsd_mem(double* %a0, double* %a1) {
 ; GENERIC-LABEL: test_movsd_mem:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    vmovsd %xmm0, (%rsi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movsd_mem:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddsd %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmovsd %xmm0, (%rsi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load double, double* %a0, align 1
   %2 = fadd double %1, %1
   store double %2, double *%a1, align 1
   ret void
 }
 
 define <2 x double> @test_movsd_reg(<2 x double> %a0, <2 x double> %a1) {
 ; GENERIC-LABEL: test_movsd_reg:

 ; BTVER2-LABEL: test_movsd_reg:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0] sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movsd_reg:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0] sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <2 x double> %a0, <2 x double> %a1, <2 x i32> <i32 2, i32 0>
   ret <2 x double> %1
 }
 
 define void @test_movupd(<2 x double> *%a0, <2 x double> *%a1) {
 ; GENERIC-LABEL: test_movupd:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    movupd (%rdi), %xmm0 # sched: [6:0.50]

 ; BTVER2-NEXT:    vmovupd %xmm0, (%rsi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movupd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovupd (%rdi), %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vmovupd %xmm0, (%rsi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load <2 x double>, <2 x double> *%a0, align 1
   %2 = fadd <2 x double> %1, %1
   store <2 x double> %2, <2 x double> *%a1, align 1
   ret void
 }
 
 define <2 x double> @test_mulpd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) {
 ; GENERIC-LABEL: test_mulpd:

 ; BTVER2-NEXT:    vmulpd %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vmulpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_mulpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmulpd %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vmulpd (%rdi), %xmm0, %xmm0 # sched: [12:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fmul <2 x double> %a0, %a1
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = fmul <2 x double> %1, %2
   ret <2 x double> %3
 }
 
 define double @test_mulsd(double %a0, double %a1, double *%a2) {
 ; GENERIC-LABEL: test_mulsd:

 ; BTVER2-NEXT:    vmulsd %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vmulsd (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_mulsd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmulsd %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
 ; ZNVER1-NEXT:    vmulsd (%rdi), %xmm0, %xmm0 # sched: [12:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fmul double %a0, %a1
   %2 = load double, double *%a2, align 8
   %3 = fmul double %1, %2
   ret double %3
 }
 
 define <2 x double> @test_orpd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) {
 ; GENERIC-LABEL: test_orpd:

 ; BTVER2-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_orpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vorpd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vorpd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast <2 x double> %a0 to <4 x i32>
   %2 = bitcast <2 x double> %a1 to <4 x i32>
   %3 = or <4 x i32> %1, %2
   %4 = load <2 x double>, <2 x double> *%a2, align 16
   %5 = bitcast <2 x double> %4 to <4 x i32>
   %6 = or <4 x i32> %3, %5
   %7 = bitcast <4 x i32> %6 to <2 x double>
   %8 = fadd <2 x double> %a1, %7

 ; BTVER2-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpackssdw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_packssdw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpackssdw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a0, <4 x i32> %a1)
   %2 = bitcast <8 x i16> %1 to <4 x i32>
   %3 = load <4 x i32>, <4 x i32> *%a2, align 16
   %4 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %2, <4 x i32> %3)
   ret <8 x i16> %4
 }
 declare <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32>, <4 x i32>) nounwind readnone
 

 ; BTVER2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpacksswb (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_packsswb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpacksswb (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %a0, <8 x i16> %a1)
   %2 = bitcast <16 x i8> %1 to <8 x i16>
   %3 = load <8 x i16>, <8 x i16> *%a2, align 16
   %4 = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %2, <8 x i16> %3)
   ret <16 x i8> %4
 }
 declare <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16>, <8 x i16>) nounwind readnone
 

 ; BTVER2-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpackuswb (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_packuswb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpackuswb (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> %a0, <8 x i16> %a1)
   %2 = bitcast <16 x i8> %1 to <8 x i16>
   %3 = load <8 x i16>, <8 x i16> *%a2, align 16
   %4 = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> %2, <8 x i16> %3)
   ret <16 x i8> %4
 }
 declare <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16>, <8 x i16>) nounwind readnone
 

 ; BTVER2-NEXT:    vpaddb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpaddb (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_paddb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpaddb %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddb (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = add <16 x i8> %a0, %a1
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = add <16 x i8> %1, %2
   ret <16 x i8> %3
 }
 
 define <4 x i32> @test_paddd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
 ; GENERIC-LABEL: test_paddd:

 ; BTVER2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpaddd (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_paddd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = add <4 x i32> %a0, %a1
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = add <4 x i32> %1, %2
   ret <4 x i32> %3
 }
 
 define <2 x i64> @test_paddq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
 ; GENERIC-LABEL: test_paddq:

 ; BTVER2-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpaddq (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_paddq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddq (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = add <2 x i64> %a0, %a1
   %2 = load <2 x i64>, <2 x i64> *%a2, align 16
   %3 = add <2 x i64> %1, %2
   ret <2 x i64> %3
 }
 
 define <16 x i8> @test_paddsb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
 ; GENERIC-LABEL: test_paddsb:

 ; BTVER2-NEXT:    vpaddsb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpaddsb (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_paddsb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpaddsb %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddsb (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a0, <16 x i8> %a1)
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %1, <16 x i8> %2)
   ret <16 x i8> %3
 }
 declare <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8>, <16 x i8>) nounwind readnone
 
 define <8 x i16> @test_paddsw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vpaddsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpaddsw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_paddsw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpaddsw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddsw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %1, <8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16>, <8 x i16>) nounwind readnone
 
 define <16 x i8> @test_paddusb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {

 ; BTVER2-NEXT:    vpaddusb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpaddusb (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_paddusb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpaddusb %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddusb (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.sse2.paddus.b(<16 x i8> %a0, <16 x i8> %a1)
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = call <16 x i8> @llvm.x86.sse2.paddus.b(<16 x i8> %1, <16 x i8> %2)
   ret <16 x i8> %3
 }
 declare <16 x i8> @llvm.x86.sse2.paddus.b(<16 x i8>, <16 x i8>) nounwind readnone
 
 define <8 x i16> @test_paddusw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vpaddusw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpaddusw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_paddusw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpaddusw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddusw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.sse2.paddus.w(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.sse2.paddus.w(<8 x i16> %1, <8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.sse2.paddus.w(<8 x i16>, <8 x i16>) nounwind readnone
 
 define <8 x i16> @test_paddw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpaddw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_paddw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = add <8 x i16> %a0, %a1
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = add <8 x i16> %1, %2
   ret <8 x i16> %3
 }
 
 define <2 x i64> @test_pand(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
 ; GENERIC-LABEL: test_pand:

 ; BTVER2-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pand:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpand %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpand (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = and <2 x i64> %a0, %a1
   %2 = load <2 x i64>, <2 x i64> *%a2, align 16
   %3 = and <2 x i64> %1, %2
   %4 = add <2 x i64> %3, %a1
   ret <2 x i64> %4
 }
 
 define <2 x i64> @test_pandn(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {

 ; BTVER2-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pandn:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpandn %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpandn (%rdi), %xmm0, %xmm1 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = xor <2 x i64> %a0, <i64 -1, i64 -1>
   %2 = and <2 x i64> %a1, %1
   %3 = load <2 x i64>, <2 x i64> *%a2, align 16
   %4 = xor <2 x i64> %2, <i64 -1, i64 -1>
   %5 = and <2 x i64> %3, %4
   %6 = add <2 x i64> %2, %5
   ret <2 x i64> %6
 }

 ; BTVER2-NEXT:    vpavgb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpavgb (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pavgb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpavgb %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpavgb (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.sse2.pavg.b(<16 x i8> %a0, <16 x i8> %a1)
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = call <16 x i8> @llvm.x86.sse2.pavg.b(<16 x i8> %1, <16 x i8> %2)
   ret <16 x i8> %3
 }
 declare <16 x i8> @llvm.x86.sse2.pavg.b(<16 x i8> %arg0, <16 x i8> %arg1) nounwind readnone
 
 define <8 x i16> @test_pavgw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vpavgw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpavgw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pavgw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpavgw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpavgw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16> %1, <8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16>, <8 x i16>) nounwind readnone
 
 define <16 x i8> @test_pcmpeqb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {

 ; BTVER2-NEXT:    vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pcmpeqb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm1 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpcmpeqb (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = icmp eq <16 x i8> %a0, %a1
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = icmp eq <16 x i8> %a0, %2
   %4 = or <16 x i1> %1, %3
   %5 = sext <16 x i1> %4 to <16 x i8>
   ret <16 x i8> %5
 }
 

 ; BTVER2-NEXT:    vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pcmpeqd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm1 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpcmpeqd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = icmp eq <4 x i32> %a0, %a1
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = icmp eq <4 x i32> %a0, %2
   %4 = or <4 x i1> %1, %3
   %5 = sext <4 x i1> %4 to <4 x i32>
   ret <4 x i32> %5
 }
 

 ; BTVER2-NEXT:    vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pcmpeqw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm1 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpcmpeqw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = icmp eq <8 x i16> %a0, %a1
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = icmp eq <8 x i16> %a0, %2
   %4 = or <8 x i1> %1, %3
   %5 = sext <8 x i1> %4 to <8 x i16>
   ret <8 x i16> %5
 }
 

 ; BTVER2-NEXT:    vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pcmpgtb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpcmpgtb %xmm1, %xmm0, %xmm1 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpcmpgtb (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = icmp sgt <16 x i8> %a0, %a1
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = icmp sgt <16 x i8> %a0, %2
   %4 = or <16 x i1> %1, %3
   %5 = sext <16 x i1> %4 to <16 x i8>
   ret <16 x i8> %5
 }
 

 ; BTVER2-NEXT:    vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pcmpgtd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm1 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpcmpeqd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = icmp sgt <4 x i32> %a0, %a1
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = icmp eq <4 x i32> %a0, %2
   %4 = or <4 x i1> %1, %3
   %5 = sext <4 x i1> %4 to <4 x i32>
   ret <4 x i32> %5
 }
 

 ; BTVER2-NEXT:    vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pcmpgtw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm1 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpcmpgtw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpor %xmm0, %xmm1, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = icmp sgt <8 x i16> %a0, %a1
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = icmp sgt <8 x i16> %a0, %2
   %4 = or <8 x i1> %1, %3
   %5 = sext <8 x i1> %4 to <8 x i16>
   ret <8 x i16> %5
 }
 

 ; BTVER2-NEXT:    vpextrw $6, %xmm0, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pextrw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpextrw $6, %xmm0, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    # kill: %AX<def> %AX<kill> %EAX<kill>
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = extractelement <8 x i16> %a0, i32 6
   ret i16 %1
 }
 
 define <8 x i16> @test_pinsrw(<8 x i16> %a0, i16 %a1, i16 *%a2) {
 ; GENERIC-LABEL: test_pinsrw:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    pinsrw $1, %edi, %xmm0 # sched: [2:1.00]

 ; BTVER2-NEXT:    vpinsrw $1, %edi, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpinsrw $3, (%rsi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pinsrw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpinsrw $1, %edi, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpinsrw $3, (%rsi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = insertelement <8 x i16> %a0, i16 %a1, i32 1
   %2 = load i16, i16 *%a2
   %3 = insertelement <8 x i16> %1, i16 %2, i32 3
   ret <8 x i16> %3
 }
 
 define <4 x i32> @test_pmaddwd(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {
 ; GENERIC-LABEL: test_pmaddwd:

 ; BTVER2-NEXT:    vpmaddwd %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vpmaddwd (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmaddwd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmaddwd %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
 ; ZNVER1-NEXT:    vpmaddwd (%rdi), %xmm0, %xmm0 # sched: [11:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x i32> @llvm.x86.sse2.pmadd.wd(<8 x i16> %a0, <8 x i16> %a1)
   %2 = bitcast <4 x i32> %1 to <8 x i16>
   %3 = load <8 x i16>, <8 x i16> *%a2, align 16
   %4 = call <4 x i32> @llvm.x86.sse2.pmadd.wd(<8 x i16> %2, <8 x i16> %3)
   ret <4 x i32> %4
 }
 declare <4 x i32> @llvm.x86.sse2.pmadd.wd(<8 x i16>, <8 x i16>) nounwind readnone
 

 ; BTVER2-NEXT:    vpmaxsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpmaxsw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmaxsw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmaxsw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpmaxsw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.sse2.pmaxs.w(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.sse2.pmaxs.w(<8 x i16> %1, <8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.sse2.pmaxs.w(<8 x i16>, <8 x i16>) nounwind readnone
 
 define <16 x i8> @test_pmaxub(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {

 ; BTVER2-NEXT:    vpmaxub %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpmaxub (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmaxub:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmaxub %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpmaxub (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.sse2.pmaxu.b(<16 x i8> %a0, <16 x i8> %a1)
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = call <16 x i8> @llvm.x86.sse2.pmaxu.b(<16 x i8> %1, <16 x i8> %2)
   ret <16 x i8> %3
 }
 declare <16 x i8> @llvm.x86.sse2.pmaxu.b(<16 x i8>, <16 x i8>) nounwind readnone
 
 define <8 x i16> @test_pminsw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vpminsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpminsw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pminsw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpminsw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpminsw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.sse2.pmins.w(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.sse2.pmins.w(<8 x i16> %1, <8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.sse2.pmins.w(<8 x i16>, <8 x i16>) nounwind readnone
 
 define <16 x i8> @test_pminub(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {

 ; BTVER2-NEXT:    vpminub %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpminub (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pminub:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpminub %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpminub (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.sse2.pminu.b(<16 x i8> %a0, <16 x i8> %a1)
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = call <16 x i8> @llvm.x86.sse2.pminu.b(<16 x i8> %1, <16 x i8> %2)
   ret <16 x i8> %3
 }
 declare <16 x i8> @llvm.x86.sse2.pminu.b(<16 x i8>, <16 x i8>) nounwind readnone
 
 define i32 @test_pmovmskb(<16 x i8> %a0) {

 ; BTVER2-LABEL: test_pmovmskb:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vpmovmskb %xmm0, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmovmskb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmovmskb %xmm0, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.sse2.pmovmskb.128(<16 x i8> %a0)
   ret i32 %1
 }
 declare i32 @llvm.x86.sse2.pmovmskb.128(<16 x i8>) nounwind readnone
 
 define <8 x i16> @test_pmulhuw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {
 ; GENERIC-LABEL: test_pmulhuw:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    vpmulhuw %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vpmulhuw (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmulhuw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmulhuw %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
 ; ZNVER1-NEXT:    vpmulhuw (%rdi), %xmm0, %xmm0 # sched: [11:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.sse2.pmulhu.w(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.sse2.pmulhu.w(<8 x i16> %1, <8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.sse2.pmulhu.w(<8 x i16>, <8 x i16>) nounwind readnone
 
 define <8 x i16> @test_pmulhw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vpmulhw %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vpmulhw (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmulhw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmulhw %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
 ; ZNVER1-NEXT:    vpmulhw (%rdi), %xmm0, %xmm0 # sched: [11:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16> %1, <8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16>, <8 x i16>) nounwind readnone
 
 define <8 x i16> @test_pmullw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vpmullw %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vpmullw (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmullw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmullw %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
 ; ZNVER1-NEXT:    vpmullw (%rdi), %xmm0, %xmm0 # sched: [11:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = mul <8 x i16> %a0, %a1
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = mul <8 x i16> %1, %2
   ret <8 x i16> %3
 }
 
 define <2 x i64> @test_pmuludq(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
 ; GENERIC-LABEL: test_pmuludq:

 ; BTVER2-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vpmuludq (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmuludq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmuludq %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
 ; ZNVER1-NEXT:    vpmuludq (%rdi), %xmm0, %xmm0 # sched: [11:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x i64> @llvm.x86.sse2.pmulu.dq(<4 x i32> %a0, <4 x i32> %a1)
   %2 = bitcast <2 x i64> %1 to <4 x i32>
   %3 = load <4 x i32>, <4 x i32> *%a2, align 16
   %4 = call <2 x i64> @llvm.x86.sse2.pmulu.dq(<4 x i32> %2, <4 x i32> %3)
   ret <2 x i64> %4
 }
 declare <2 x i64> @llvm.x86.sse2.pmulu.dq(<4 x i32>, <4 x i32>) nounwind readnone
 

 ; BTVER2-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_por:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpor (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = or <2 x i64> %a0, %a1
   %2 = load <2 x i64>, <2 x i64> *%a2, align 16
   %3 = or <2 x i64> %1, %2
   %4 = add <2 x i64> %3, %a1
   ret <2 x i64> %4
 }
 
 define <2 x i64> @test_psadbw(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {

 ; BTVER2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vpsadbw (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psadbw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
 ; ZNVER1-NEXT:    vpsadbw (%rdi), %xmm0, %xmm0 # sched: [11:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x i64> @llvm.x86.sse2.psad.bw(<16 x i8> %a0, <16 x i8> %a1)
   %2 = bitcast <2 x i64> %1 to <16 x i8>
   %3 = load <16 x i8>, <16 x i8> *%a2, align 16
   %4 = call <2 x i64> @llvm.x86.sse2.psad.bw(<16 x i8> %2, <16 x i8> %3)
   ret <2 x i64> %4
 }
 declare <2 x i64> @llvm.x86.sse2.psad.bw(<16 x i8>, <16 x i8>) nounwind readnone
 

 ; BTVER2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pshufd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpshufd {{.*#+}} xmm1 = mem[3,2,1,0] sched: [8:0.50]
 ; ZNVER1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,0,3,2] sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x i32> %a0, <4 x i32> undef, <4 x i32> <i32 1, i32 0, i32 3, i32 2>
   %2 = load <4 x i32>, <4 x i32> *%a1, align 16
   %3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
   %4 = add <4 x i32> %1, %3
   ret <4 x i32> %4
 }
 
 define <8 x i16> @test_pshufhw(<8 x i16> %a0, <8 x i16> *%a1) {

 ; BTVER2-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pshufhw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpshufhw {{.*#+}} xmm1 = mem[0,1,2,3,7,6,5,4] sched: [8:0.50]
 ; ZNVER1-NEXT:    vpshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,4,7,6] sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x i16> %a0, <8 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 5, i32 4, i32 7, i32 6>
   %2 = load <8 x i16>, <8 x i16> *%a1, align 16
   %3 = shufflevector <8 x i16> %2, <8 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 7, i32 6, i32 5, i32 4>
   %4 = add <8 x i16> %1, %3
   ret <8 x i16> %4
 }
 
 define <8 x i16> @test_pshuflw(<8 x i16> %a0, <8 x i16> *%a1) {

 ; BTVER2-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pshuflw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpshuflw {{.*#+}} xmm1 = mem[3,2,1,0,4,5,6,7] sched: [8:0.50]
 ; ZNVER1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[1,0,3,2,4,5,6,7] sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x i16> %a0, <8 x i16> undef, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 4, i32 5, i32 6, i32 7>
   %2 = load <8 x i16>, <8 x i16> *%a1, align 16
   %3 = shufflevector <8 x i16> %2, <8 x i16> undef, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 4, i32 5, i32 6, i32 7>
   %4 = add <8 x i16> %1, %3
   ret <8 x i16> %4
 }
 
 define <4 x i32> @test_pslld(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {

 ; BTVER2-NEXT:    vpslld $2, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pslld:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpslld %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpslld (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpslld $2, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x i32> @llvm.x86.sse2.psll.d(<4 x i32> %a0, <4 x i32> %a1)
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = call <4 x i32> @llvm.x86.sse2.psll.d(<4 x i32> %1, <4 x i32> %2)
   %4 = call <4 x i32> @llvm.x86.sse2.pslli.d(<4 x i32> %3, i32 2)
   ret <4 x i32> %4
 }
 declare <4 x i32> @llvm.x86.sse2.psll.d(<4 x i32>, <4 x i32>) nounwind readnone
 declare <4 x i32> @llvm.x86.sse2.pslli.d(<4 x i32>, i32) nounwind readnone

 ; BTVER2-LABEL: test_pslldq:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11] sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pslldq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpslldq {{.*#+}} xmm0 = zero,zero,zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11] sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x i32> %a0, <4 x i32> zeroinitializer, <4 x i32> <i32 4, i32 0, i32 1, i32 2>
   ret <4 x i32> %1
 }
 
 define <2 x i64> @test_psllq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
 ; GENERIC-LABEL: test_psllq:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    psllq %xmm1, %xmm0 # sched: [2:1.00]

 ; BTVER2-NEXT:    vpsllq $2, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psllq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsllq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsllq (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpsllq $2, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x i64> @llvm.x86.sse2.psll.q(<2 x i64> %a0, <2 x i64> %a1)
   %2 = load <2 x i64>, <2 x i64> *%a2, align 16
   %3 = call <2 x i64> @llvm.x86.sse2.psll.q(<2 x i64> %1, <2 x i64> %2)
   %4 = call <2 x i64> @llvm.x86.sse2.pslli.q(<2 x i64> %3, i32 2)
   ret <2 x i64> %4
 }
 declare <2 x i64> @llvm.x86.sse2.psll.q(<2 x i64>, <2 x i64>) nounwind readnone
 declare <2 x i64> @llvm.x86.sse2.pslli.q(<2 x i64>, i32) nounwind readnone

 ; BTVER2-NEXT:    vpsllw $2, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psllw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsllw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsllw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpsllw $2, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.sse2.psll.w(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.sse2.psll.w(<8 x i16> %1, <8 x i16> %2)
   %4 = call <8 x i16> @llvm.x86.sse2.pslli.w(<8 x i16> %3, i32 2)
   ret <8 x i16> %4
 }
 declare <8 x i16> @llvm.x86.sse2.psll.w(<8 x i16>, <8 x i16>) nounwind readnone
 declare <8 x i16> @llvm.x86.sse2.pslli.w(<8 x i16>, i32) nounwind readnone

 ; BTVER2-NEXT:    vpsrad $2, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psrad:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsrad %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsrad (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpsrad $2, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x i32> @llvm.x86.sse2.psra.d(<4 x i32> %a0, <4 x i32> %a1)
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = call <4 x i32> @llvm.x86.sse2.psra.d(<4 x i32> %1, <4 x i32> %2)
   %4 = call <4 x i32> @llvm.x86.sse2.psrai.d(<4 x i32> %3, i32 2)
   ret <4 x i32> %4
 }
 declare <4 x i32> @llvm.x86.sse2.psra.d(<4 x i32>, <4 x i32>) nounwind readnone
 declare <4 x i32> @llvm.x86.sse2.psrai.d(<4 x i32>, i32) nounwind readnone

 ; BTVER2-NEXT:    vpsraw $2, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psraw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsraw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsraw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpsraw $2, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.sse2.psra.w(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.sse2.psra.w(<8 x i16> %1, <8 x i16> %2)
   %4 = call <8 x i16> @llvm.x86.sse2.psrai.w(<8 x i16> %3, i32 2)
   ret <8 x i16> %4
 }
 declare <8 x i16> @llvm.x86.sse2.psra.w(<8 x i16>, <8 x i16>) nounwind readnone
 declare <8 x i16> @llvm.x86.sse2.psrai.w(<8 x i16>, i32) nounwind readnone

 ; BTVER2-NEXT:    vpsrld $2, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psrld:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsrld %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsrld (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpsrld $2, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x i32> @llvm.x86.sse2.psrl.d(<4 x i32> %a0, <4 x i32> %a1)
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = call <4 x i32> @llvm.x86.sse2.psrl.d(<4 x i32> %1, <4 x i32> %2)
   %4 = call <4 x i32> @llvm.x86.sse2.psrli.d(<4 x i32> %3, i32 2)
   ret <4 x i32> %4
 }
 declare <4 x i32> @llvm.x86.sse2.psrl.d(<4 x i32>, <4 x i32>) nounwind readnone
 declare <4 x i32> @llvm.x86.sse2.psrli.d(<4 x i32>, i32) nounwind readnone

 ; BTVER2-LABEL: test_psrldq:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm0[4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psrldq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsrldq {{.*#+}} xmm0 = xmm0[4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,zero sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x i32> %a0, <4 x i32> zeroinitializer, <4 x i32> <i32 1, i32 2, i32 3, i32 4>
   ret <4 x i32> %1
 }
 
 define <2 x i64> @test_psrlq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
 ; GENERIC-LABEL: test_psrlq:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    psrlq %xmm1, %xmm0 # sched: [2:1.00]

 ; BTVER2-NEXT:    vpsrlq $2, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psrlq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsrlq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsrlq (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpsrlq $2, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x i64> @llvm.x86.sse2.psrl.q(<2 x i64> %a0, <2 x i64> %a1)
   %2 = load <2 x i64>, <2 x i64> *%a2, align 16
   %3 = call <2 x i64> @llvm.x86.sse2.psrl.q(<2 x i64> %1, <2 x i64> %2)
   %4 = call <2 x i64> @llvm.x86.sse2.psrli.q(<2 x i64> %3, i32 2)
   ret <2 x i64> %4
 }
 declare <2 x i64> @llvm.x86.sse2.psrl.q(<2 x i64>, <2 x i64>) nounwind readnone
 declare <2 x i64> @llvm.x86.sse2.psrli.q(<2 x i64>, i32) nounwind readnone

 ; BTVER2-NEXT:    vpsrlw $2, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psrlw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsrlw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsrlw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpsrlw $2, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.sse2.psrl.w(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.sse2.psrl.w(<8 x i16> %1, <8 x i16> %2)
   %4 = call <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16> %3, i32 2)
   ret <8 x i16> %4
 }
 declare <8 x i16> @llvm.x86.sse2.psrl.w(<8 x i16>, <8 x i16>) nounwind readnone
 declare <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16>, i32) nounwind readnone

 ; BTVER2-NEXT:    vpsubb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpsubb (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psubb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsubb %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsubb (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = sub <16 x i8> %a0, %a1
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = sub <16 x i8> %1, %2
   ret <16 x i8> %3
 }
 
 define <4 x i32> @test_psubd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
 ; GENERIC-LABEL: test_psubd:

 ; BTVER2-NEXT:    vpsubd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpsubd (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psubd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsubd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsubd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = sub <4 x i32> %a0, %a1
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = sub <4 x i32> %1, %2
   ret <4 x i32> %3
 }
 
 define <2 x i64> @test_psubq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
 ; GENERIC-LABEL: test_psubq:

 ; BTVER2-NEXT:    vpsubq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpsubq (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psubq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsubq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsubq (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = sub <2 x i64> %a0, %a1
   %2 = load <2 x i64>, <2 x i64> *%a2, align 16
   %3 = sub <2 x i64> %1, %2
   ret <2 x i64> %3
 }
 
 define <16 x i8> @test_psubsb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
 ; GENERIC-LABEL: test_psubsb:

 ; BTVER2-NEXT:    vpsubsb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpsubsb (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psubsb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsubsb %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsubsb (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %a0, <16 x i8> %a1)
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %1, <16 x i8> %2)
   ret <16 x i8> %3
 }
 declare <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8>, <16 x i8>) nounwind readnone
 
 define <8 x i16> @test_psubsw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vpsubsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpsubsw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psubsw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsubsw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsubsw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %1, <8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16>, <8 x i16>) nounwind readnone
 
 define <16 x i8> @test_psubusb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {

 ; BTVER2-NEXT:    vpsubusb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpsubusb (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psubusb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsubusb %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsubusb (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.sse2.psubus.b(<16 x i8> %a0, <16 x i8> %a1)
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = call <16 x i8> @llvm.x86.sse2.psubus.b(<16 x i8> %1, <16 x i8> %2)
   ret <16 x i8> %3
 }
 declare <16 x i8> @llvm.x86.sse2.psubus.b(<16 x i8>, <16 x i8>) nounwind readnone
 
 define <8 x i16> @test_psubusw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vpsubusw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpsubusw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psubusw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsubusw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsubusw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.sse2.psubus.w(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.sse2.psubus.w(<8 x i16> %1, <8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.sse2.psubus.w(<8 x i16>, <8 x i16>) nounwind readnone
 
 define <8 x i16> @test_psubw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vpsubw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpsubw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psubw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsubw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsubw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = sub <8 x i16> %a0, %a1
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = sub <8 x i16> %1, %2
   ret <8 x i16> %3
 }
 
 define <16 x i8> @test_punpckhbw(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
 ; GENERIC-LABEL: test_punpckhbw:

 ; BTVER2-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15] sched: [1:0.50]
 ; BTVER2-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],mem[8],xmm0[9],mem[9],xmm0[10],mem[10],xmm0[11],mem[11],xmm0[12],mem[12],xmm0[13],mem[13],xmm0[14],mem[14],xmm0[15],mem[15] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_punpckhbw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15] sched: [1:0.25]
 ; ZNVER1-NEXT:    vpunpckhbw {{.*#+}} xmm0 = xmm0[8],mem[8],xmm0[9],mem[9],xmm0[10],mem[10],xmm0[11],mem[11],xmm0[12],mem[12],xmm0[13],mem[13],xmm0[14],mem[14],xmm0[15],mem[15] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <16 x i8> %a0, <16 x i8> %a1, <16 x i32> <i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = shufflevector <16 x i8> %1, <16 x i8> %2, <16 x i32> <i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
   ret <16 x i8> %3
 }
 
 define <4 x i32> @test_punpckhdq(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
 ; GENERIC-LABEL: test_punpckhdq:

 ; BTVER2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_punpckhdq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpunpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3] sched: [1:0.25]
 ; ZNVER1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm1[2],mem[2],xmm1[3],mem[3] sched: [8:0.50]
 ; ZNVER1-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x i32> %a0, <4 x i32> %a1, <4 x i32> <i32 2, i32 6, i32 3, i32 7>
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = shufflevector <4 x i32> %a1, <4 x i32> %2, <4 x i32> <i32 2, i32 6, i32 3, i32 7>
   %4 = add <4 x i32> %1, %3
   ret <4 x i32> %4
 }
 
 define <2 x i64> @test_punpckhqdq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {

 ; BTVER2-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_punpckhqdq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpunpckhqdq {{.*#+}} xmm0 = xmm0[1],xmm1[1] sched: [1:0.25]
 ; ZNVER1-NEXT:    vpunpckhqdq {{.*#+}} xmm1 = xmm1[1],mem[1] sched: [8:0.50]
 ; ZNVER1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <2 x i64> %a0, <2 x i64> %a1, <2 x i32> <i32 1, i32 3>
   %2 = load <2 x i64>, <2 x i64> *%a2, align 16
   %3 = shufflevector <2 x i64> %a1, <2 x i64> %2, <2x i32> <i32 1, i32 3>
   %4 = add <2 x i64> %1, %3
   ret <2 x i64> %4
 }
 
 define <8 x i16> @test_punpckhwd(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7] sched: [1:0.50]
 ; BTVER2-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_punpckhwd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7] sched: [1:0.25]
 ; ZNVER1-NEXT:    vpunpckhwd {{.*#+}} xmm0 = xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x i16> %a0, <8 x i16> %a1, <8 x i32> <i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = shufflevector <8 x i16> %1, <8 x i16> %2, <8 x i32> <i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
   ret <8 x i16> %3
 }
 
 define <16 x i8> @test_punpcklbw(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
 ; GENERIC-LABEL: test_punpcklbw:

 ; BTVER2-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7] sched: [1:0.50]
 ; BTVER2-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3],xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_punpcklbw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7] sched: [1:0.25]
 ; ZNVER1-NEXT:    vpunpcklbw {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3],xmm0[4],mem[4],xmm0[5],mem[5],xmm0[6],mem[6],xmm0[7],mem[7] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <16 x i8> %a0, <16 x i8> %a1, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23>
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = shufflevector <16 x i8> %1, <16 x i8> %2, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23>
   ret <16 x i8> %3
 }
 
 define <4 x i32> @test_punpckldq(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
 ; GENERIC-LABEL: test_punpckldq:

 ; BTVER2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_punpckldq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] sched: [1:0.25]
 ; ZNVER1-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1] sched: [8:0.50]
 ; ZNVER1-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x i32> %a0, <4 x i32> %a1, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = shufflevector <4 x i32> %a1, <4 x i32> %2, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
   %4 = add <4 x i32> %1, %3
   ret <4 x i32> %4
 }
 
 define <2 x i64> @test_punpcklqdq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {

 ; BTVER2-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_punpcklqdq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] sched: [1:0.25]
 ; ZNVER1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],mem[0] sched: [8:0.50]
 ; ZNVER1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <2 x i64> %a0, <2 x i64> %a1, <2 x i32> <i32 0, i32 2>
   %2 = load <2 x i64>, <2 x i64> *%a2, align 16
   %3 = shufflevector <2 x i64> %a1, <2 x i64> %2, <2x i32> <i32 0, i32 2>
   %4 = add <2 x i64> %1, %3
   ret <2 x i64> %4
 }
 
 define <8 x i16> @test_punpcklwd(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] sched: [1:0.50]
 ; BTVER2-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_punpcklwd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] sched: [1:0.25]
 ; ZNVER1-NEXT:    vpunpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x i16> %a0, <8 x i16> %a1, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = shufflevector <8 x i16> %1, <8 x i16> %2, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
   ret <8 x i16> %3
 }
 
 define <2 x i64> @test_pxor(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
 ; GENERIC-LABEL: test_pxor:

 ; BTVER2-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pxor:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpxor %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpxor (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = xor <2 x i64> %a0, %a1
   %2 = load <2 x i64>, <2 x i64> *%a2, align 16
   %3 = xor <2 x i64> %1, %2
   %4 = add <2 x i64> %3, %a1
   ret <2 x i64> %4
 }
 
 define <2 x double> @test_shufpd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) {

 ; BTVER2-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_shufpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vshufpd {{.*#+}} xmm0 = xmm0[1],xmm1[0] sched: [1:0.50]
 ; ZNVER1-NEXT:    vshufpd {{.*#+}} xmm1 = xmm1[1],mem[0] sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <2 x double> %a0, <2 x double> %a1, <2 x i32> <i32 1, i32 2>
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = shufflevector <2 x double> %a1, <2 x double> %2, <2 x i32> <i32 1, i32 2>
   %4 = fadd <2 x double> %1, %3
   ret <2 x double> %4
 }
 
 define <2 x double> @test_sqrtpd(<2 x double> %a0, <2 x double> *%a1) {

 ; BTVER2-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_sqrtpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vsqrtpd (%rdi), %xmm1 # sched: [27:1.00]
 ; ZNVER1-NEXT:    vsqrtpd %xmm0, %xmm0 # sched: [20:1.00]
 ; ZNVER1-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x double> @llvm.x86.sse2.sqrt.pd(<2 x double> %a0)
   %2 = load <2 x double>, <2 x double> *%a1, align 16
   %3 = call <2 x double> @llvm.x86.sse2.sqrt.pd(<2 x double> %2)
   %4 = fadd <2 x double> %1, %3
   ret <2 x double> %4
 }
 declare <2 x double> @llvm.x86.sse2.sqrt.pd(<2 x double>) nounwind readnone
 

 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_sqrtsd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovapd (%rdi), %xmm1 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vsqrtsd %xmm0, %xmm0, %xmm0 # sched: [27:1.00]
 ; ZNVER1-NEXT:    vsqrtsd %xmm1, %xmm1, %xmm1 # sched: [27:1.00]
 ; ZNVER1-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x double> @llvm.x86.sse2.sqrt.sd(<2 x double> %a0)
   %2 = load <2 x double>, <2 x double> *%a1, align 16
   %3 = call <2 x double> @llvm.x86.sse2.sqrt.sd(<2 x double> %2)
   %4 = fadd <2 x double> %1, %3
   ret <2 x double> %4
 }
 declare <2 x double> @llvm.x86.sse2.sqrt.sd(<2 x double>) nounwind readnone
 

 ; BTVER2-NEXT:    vsubpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vsubpd (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_subpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vsubpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vsubpd (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fsub <2 x double> %a0, %a1
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = fsub <2 x double> %1, %2
   ret <2 x double> %3
 }
 
 define double @test_subsd(double %a0, double %a1, double *%a2) {
 ; GENERIC-LABEL: test_subsd:

 ; BTVER2-NEXT:    vsubsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vsubsd (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_subsd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vsubsd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vsubsd (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = fsub double %a0, %a1
   %2 = load double, double *%a2, align 8
   %3 = fsub double %1, %2
   ret double %3
 }
 
 define i32 @test_ucomisd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) {
 ; GENERIC-LABEL: test_ucomisd:

 ; ZNVER1-NEXT:    sete %cl # sched: [1:0.25]
 ; ZNVER1-NEXT:    andb %al, %cl # sched: [1:0.25]
 ; ZNVER1-NEXT:    vucomisd (%rdi), %xmm0 # sched: [10:1.00]
 ; ZNVER1-NEXT:    setnp %al # sched: [1:0.25]
 ; ZNVER1-NEXT:    sete %dl # sched: [1:0.25]
 ; ZNVER1-NEXT:    andb %al, %dl # sched: [1:0.25]
 ; ZNVER1-NEXT:    orb %cl, %dl # sched: [1:0.25]
 ; ZNVER1-NEXT:    movzbl %dl, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.sse2.ucomieq.sd(<2 x double> %a0, <2 x double> %a1)
   %2 = load <2 x double>, <2 x double> *%a2, align 8
   %3 = call i32 @llvm.x86.sse2.ucomieq.sd(<2 x double> %a0, <2 x double> %2)
   %4 = or i32 %1, %3
   ret i32 %4
 }
 declare i32 @llvm.x86.sse2.ucomieq.sd(<2 x double>, <2 x double>) nounwind readnone
 

 ; BTVER2-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_unpckhpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vunpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1] sched: [1:0.50]
 ; ZNVER1-NEXT:    vunpckhpd {{.*#+}} xmm1 = xmm1[1],mem[1] sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <2 x double> %a0, <2 x double> %a1, <2 x i32> <i32 1, i32 3>
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = shufflevector <2 x double> %a1, <2 x double> %2, <2 x i32> <i32 1, i32 3>
   %4 = fadd <2 x double> %1, %3
   ret <2 x double> %4
 }
 
 define <2 x double> @test_unpcklpd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) {

 ; BTVER2-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_unpcklpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0] sched: [1:0.50]
 ; ZNVER1-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm0[0],mem[0] sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <2 x double> %a0, <2 x double> %a1, <2 x i32> <i32 0, i32 2>
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = shufflevector <2 x double> %1, <2 x double> %2, <2 x i32> <i32 0, i32 2>
   %4 = fadd <2 x double> %1, %3
   ret <2 x double> %4
 }
 
 define <2 x double> @test_xorpd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) {

 ; BTVER2-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_xorpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vxorpd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vxorpd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = bitcast <2 x double> %a0 to <4 x i32>
   %2 = bitcast <2 x double> %a1 to <4 x i32>
   %3 = xor <4 x i32> %1, %2
   %4 = load <2 x double>, <2 x double> *%a2, align 16
   %5 = bitcast <2 x double> %4 to <4 x i32>
   %6 = xor <4 x i32> %3, %5
   %7 = bitcast <4 x i32> %6 to <2 x double>
   %8 = fadd <2 x double> %a1, %7
   ret <2 x double> %8
 }
 
 !0 = !{i32 1}

test/CodeGen/X86/sse3-schedule.ll

 ; BTVER2-NEXT:    vaddsubpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vaddsubpd (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_addsubpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vaddsubpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vaddsubpd (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x double> @llvm.x86.sse3.addsub.pd(<2 x double> %a0, <2 x double> %a1)
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = call <2 x double> @llvm.x86.sse3.addsub.pd(<2 x double> %1, <2 x double> %2)
   ret <2 x double> %3
 }
 declare <2 x double> @llvm.x86.sse3.addsub.pd(<2 x double>, <2 x double>) nounwind readnone
 
 define <4 x float> @test_addsubps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {

 ; BTVER2-NEXT:    vaddsubps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vaddsubps (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_addsubps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vaddsubps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vaddsubps (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.sse3.addsub.ps(<4 x float> %a0, <4 x float> %a1)
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = call <4 x float> @llvm.x86.sse3.addsub.ps(<4 x float> %1, <4 x float> %2)
   ret <4 x float> %3
 }
 declare <4 x float> @llvm.x86.sse3.addsub.ps(<4 x float>, <4 x float>) nounwind readnone
 
 define <2 x double> @test_haddpd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) {

 ; BTVER2-NEXT:    vhaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vhaddpd (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_haddpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vhaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vhaddpd (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x double> @llvm.x86.sse3.hadd.pd(<2 x double> %a0, <2 x double> %a1)
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = call <2 x double> @llvm.x86.sse3.hadd.pd(<2 x double> %1, <2 x double> %2)
   ret <2 x double> %3
 }
 declare <2 x double> @llvm.x86.sse3.hadd.pd(<2 x double>, <2 x double>) nounwind readnone
 
 define <4 x float> @test_haddps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {

 ; BTVER2-NEXT:    vhaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vhaddps (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_haddps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vhaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vhaddps (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.sse3.hadd.ps(<4 x float> %a0, <4 x float> %a1)
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = call <4 x float> @llvm.x86.sse3.hadd.ps(<4 x float> %1, <4 x float> %2)
   ret <4 x float> %3
 }
 declare <4 x float> @llvm.x86.sse3.hadd.ps(<4 x float>, <4 x float>) nounwind readnone
 
 define <2 x double> @test_hsubpd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) {

 ; BTVER2-NEXT:    vhsubpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vhsubpd (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_hsubpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vhsubpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vhsubpd (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x double> @llvm.x86.sse3.hsub.pd(<2 x double> %a0, <2 x double> %a1)
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = call <2 x double> @llvm.x86.sse3.hsub.pd(<2 x double> %1, <2 x double> %2)
   ret <2 x double> %3
 }
 declare <2 x double> @llvm.x86.sse3.hsub.pd(<2 x double>, <2 x double>) nounwind readnone
 
 define <4 x float> @test_hsubps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {

 ; BTVER2-NEXT:    vhsubps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vhsubps (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_hsubps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vhsubps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vhsubps (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.sse3.hsub.ps(<4 x float> %a0, <4 x float> %a1)
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = call <4 x float> @llvm.x86.sse3.hsub.ps(<4 x float> %1, <4 x float> %2)
   ret <4 x float> %3
 }
 declare <4 x float> @llvm.x86.sse3.hsub.ps(<4 x float>, <4 x float>) nounwind readnone
 
 define <16 x i8> @test_lddqu(i8* %a0) {

 ; BTVER2-LABEL: test_lddqu:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vlddqu (%rdi), %xmm0 # sched: [5:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_lddqu:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vlddqu (%rdi), %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.sse3.ldu.dq(i8* %a0)
   ret <16 x i8> %1
 }
 declare <16 x i8> @llvm.x86.sse3.ldu.dq(i8*) nounwind readonly
 
 define void @test_monitor(i8* %a0, i32 %a1, i32 %a2) {
 ; GENERIC-LABEL: test_monitor:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    monitor # sched: [100:0.17]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_monitor:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    leaq (%rdi), %rax # sched: [1:0.25]
 ; ZNVER1-NEXT:    movl %esi, %ecx # sched: [1:0.25]
 ; ZNVER1-NEXT:    monitor # sched: [100:?]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   tail call void @llvm.x86.sse3.monitor(i8* %a0, i32 %a1, i32 %a2)
   ret void
 }
 declare void @llvm.x86.sse3.monitor(i8*, i32, i32)
 
 define <2 x double> @test_movddup(<2 x double> %a0, <2 x double> *%a1) {
 ; GENERIC-LABEL: test_movddup:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movddup:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovddup {{.*#+}} xmm1 = mem[0,0] sched: [8:0.50]
 ; ZNVER1-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0] sched: [1:0.50]
 ; ZNVER1-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <2 x double> %a0, <2 x double> undef, <2 x i32> zeroinitializer
   %2 = load <2 x double>, <2 x double> *%a1, align 16
   %3 = shufflevector <2 x double> %2, <2 x double> undef, <2 x i32> zeroinitializer
   %4 = fadd <2 x double> %1, %3
   ret <2 x double> %4
 }
 
 define <4 x float> @test_movshdup(<4 x float> %a0, <4 x float> *%a1) {

 ; BTVER2-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movshdup:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovshdup {{.*#+}} xmm1 = mem[1,1,3,3] sched: [8:0.50]
 ; ZNVER1-NEXT:    vmovshdup {{.*#+}} xmm0 = xmm0[1,1,3,3] sched: [1:0.50]
 ; ZNVER1-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x float> %a0, <4 x float> undef, <4 x i32> <i32 1, i32 1, i32 3, i32 3>
   %2 = load <4 x float>, <4 x float> *%a1, align 16
   %3 = shufflevector <4 x float> %2, <4 x float> undef, <4 x i32> <i32 1, i32 1, i32 3, i32 3>
   %4 = fadd <4 x float> %1, %3
   ret <4 x float> %4
 }
 
 define <4 x float> @test_movsldup(<4 x float> %a0, <4 x float> *%a1) {

 ; BTVER2-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movsldup:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovsldup {{.*#+}} xmm1 = mem[0,0,2,2] sched: [8:0.50]
 ; ZNVER1-NEXT:    vmovsldup {{.*#+}} xmm0 = xmm0[0,0,2,2] sched: [1:0.50]
 ; ZNVER1-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x float> %a0, <4 x float> undef, <4 x i32> <i32 0, i32 0, i32 2, i32 2>
   %2 = load <4 x float>, <4 x float> *%a1, align 16
   %3 = shufflevector <4 x float> %2, <4 x float> undef, <4 x i32> <i32 0, i32 0, i32 2, i32 2>
   %4 = fadd <4 x float> %1, %3
   ret <4 x float> %4
 }
 
 define void @test_mwait(i32 %a0, i32 %a1) {

 ; BTVER2-NEXT:    mwait # sched: [100:0.17]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_mwait:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    movl %edi, %ecx # sched: [1:0.25]
 ; ZNVER1-NEXT:    movl %esi, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    mwait # sched: [100:?]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   tail call void @llvm.x86.sse3.mwait(i32 %a0, i32 %a1)
   ret void
 }
 declare void @llvm.x86.sse3.mwait(i32, i32)

test/CodeGen/X86/sse41-schedule.ll

 ; BTVER2-NEXT:    vblendpd {{.*#+}} xmm0 = xmm0[0],mem[1] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_blendpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vblendpd {{.*#+}} xmm0 = xmm0[0],xmm1[1] sched: [1:0.50]
 ; ZNVER1-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vblendpd {{.*#+}} xmm0 = xmm0[0],mem[1] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <2 x double> %a0, <2 x double> %a1, <2 x i32> <i32 0, i32 3>
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = fadd <2 x double> %a1, %1
   %4 = shufflevector <2 x double> %3, <2 x double> %2, <2 x i32> <i32 0, i32 3>
   ret <2 x double> %4
 }
 
 define <4 x float> @test_blendps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {

 ; BTVER2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2],xmm0[3] sched: [1:0.50]
 ; BTVER2-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],mem[1],xmm0[2,3] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_blendps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1,2],xmm0[3] sched: [1:0.50]
 ; ZNVER1-NEXT:    vblendps {{.*#+}} xmm0 = xmm0[0],mem[1],xmm0[2,3] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x float> %a0, <4 x float> %a1, <4 x i32> <i32 0, i32 5, i32 6, i32 3>
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = shufflevector <4 x float> %1, <4 x float> %2, <4 x i32> <i32 0, i32 5, i32 2, i32 3>
   ret <4 x float> %3
 }
 
 define <2 x double> @test_blendvpd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2, <2 x double> *%a3) {
 ; GENERIC-LABEL: test_blendvpd:

 ; BTVER2-NEXT:    vblendvpd %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vblendvpd %xmm2, (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_blendvpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vblendvpd %xmm2, %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; ZNVER1-NEXT:    vblendvpd %xmm2, (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x double> @llvm.x86.sse41.blendvpd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2)
   %2 = load <2 x double>, <2 x double> *%a3, align 16
   %3 = call <2 x double> @llvm.x86.sse41.blendvpd(<2 x double> %1, <2 x double> %2, <2 x double> %a2)
   ret <2 x double> %3
 }
 declare <2 x double> @llvm.x86.sse41.blendvpd(<2 x double>, <2 x double>, <2 x double>) nounwind readnone
 
 define <4 x float> @test_blendvps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2, <4 x float> *%a3) {

 ; BTVER2-NEXT:    vblendvps %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vblendvps %xmm2, (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_blendvps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vblendvps %xmm2, %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; ZNVER1-NEXT:    vblendvps %xmm2, (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.sse41.blendvps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
   %2 = load <4 x float>, <4 x float> *%a3
   %3 = call <4 x float> @llvm.x86.sse41.blendvps(<4 x float> %1, <4 x float> %2, <4 x float> %a2)
   ret <4 x float> %3
 }
 declare <4 x float> @llvm.x86.sse41.blendvps(<4 x float>, <4 x float>, <4 x float>) nounwind readnone
 
 define <2 x double> @test_dppd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) {

 ; BTVER2-NEXT:    vdppd $7, %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vdppd $7, (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_dppd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vdppd $7, %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vdppd $7, (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x double> @llvm.x86.sse41.dppd(<2 x double> %a0, <2 x double> %a1, i8 7)
   %2 = load <2 x double>, <2 x double> *%a2, align 16
   %3 = call <2 x double> @llvm.x86.sse41.dppd(<2 x double> %1, <2 x double> %2, i8 7)
   ret <2 x double> %3
 }
 declare <2 x double> @llvm.x86.sse41.dppd(<2 x double>, <2 x double>, i8) nounwind readnone
 
 define <4 x float> @test_dpps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {

 ; BTVER2-NEXT:    vdpps $7, %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    vdpps $7, (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_dpps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vdpps $7, %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vdpps $7, (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.sse41.dpps(<4 x float> %a0, <4 x float> %a1, i8 7)
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = call <4 x float> @llvm.x86.sse41.dpps(<4 x float> %1, <4 x float> %2, i8 7)
   ret <4 x float> %3
 }
 declare <4 x float> @llvm.x86.sse41.dpps(<4 x float>, <4 x float>, i8) nounwind readnone
 
 define <4 x float> @test_insertps(<4 x float> %a0, <4 x float> %a1, float *%a2) {

 ; BTVER2-NEXT:    vinsertps {{.*#+}} xmm0 = zero,xmm1[0],xmm0[2,3] sched: [1:0.50]
 ; BTVER2-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_insertps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vinsertps {{.*#+}} xmm0 = zero,xmm1[0],xmm0[2,3] sched: [1:0.50]
 ; ZNVER1-NEXT:    vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.sse41.insertps(<4 x float> %a0, <4 x float> %a1, i8 17)
   %2 = load float, float *%a2
   %3 = insertelement <4 x float> %1, float %2, i32 3
   ret <4 x float> %3
 }
 declare <4 x float> @llvm.x86.sse41.insertps(<4 x float>, <4 x float>, i8) nounwind readnone
 
 define <2 x i64> @test_movntdqa(i8* %a0) {

 ; BTVER2-LABEL: test_movntdqa:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    vmovntdqa (%rdi), %xmm0 # sched: [5:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movntdqa:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmovntdqa (%rdi), %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x i64> @llvm.x86.sse41.movntdqa(i8* %a0)
   ret <2 x i64> %1
 }
 declare <2 x i64> @llvm.x86.sse41.movntdqa(i8*) nounwind readnone
 
 define <8 x i16> @test_mpsadbw(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {
 ; GENERIC-LABEL: test_mpsadbw:
 ; GENERIC:       # BB#0:

 ; BTVER2-NEXT:    vmpsadbw $7, %xmm1, %xmm0, %xmm0 # sched: [3:2.00]
 ; BTVER2-NEXT:    vmpsadbw $7, (%rdi), %xmm0, %xmm0 # sched: [8:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_mpsadbw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vmpsadbw $7, %xmm1, %xmm0, %xmm0 # sched: [100:?]
 ; ZNVER1-NEXT:    vmpsadbw $7, (%rdi), %xmm0, %xmm0 # sched: [100:?]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.sse41.mpsadbw(<16 x i8> %a0, <16 x i8> %a1, i8 7)
   %2 = bitcast <8 x i16> %1 to <16 x i8>
   %3 = load <16 x i8>, <16 x i8> *%a2, align 16
   %4 = call <8 x i16> @llvm.x86.sse41.mpsadbw(<16 x i8> %2, <16 x i8> %3, i8 7)
   ret <8 x i16> %4
 }
 declare <8 x i16> @llvm.x86.sse41.mpsadbw(<16 x i8>, <16 x i8>, i8) nounwind readnone
 

 ; BTVER2-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpackusdw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_packusdw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpackusdw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a0, <4 x i32> %a1)
   %2 = bitcast <8 x i16> %1 to <4 x i32>
   %3 = load <4 x i32>, <4 x i32> *%a2, align 16
   %4 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %2, <4 x i32> %3)
   ret <8 x i16> %4
 }
 declare <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32>, <4 x i32>) nounwind readnone
 

 ; BTVER2-NEXT:    vpblendvb %xmm2, %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vpblendvb %xmm2, (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pblendvb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpblendvb %xmm2, %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
 ; ZNVER1-NEXT:    vpblendvb %xmm2, (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> %a2)
   %2 = load <16 x i8>, <16 x i8> *%a3, align 16
   %3 = call <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8> %1, <16 x i8> %2, <16 x i8> %a2)
   ret <16 x i8> %3
 }
 declare <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8>, <16 x i8>, <16 x i8>) nounwind readnone
 
 define <8 x i16> @test_pblendw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7] sched: [1:0.50]
 ; BTVER2-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],mem[2,3],xmm0[4,5,6],mem[7] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pblendw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5],xmm0[6],xmm1[7] sched: [1:0.50]
 ; ZNVER1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],mem[2,3],xmm0[4,5,6],mem[7] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x i16> %a0, <8 x i16> %a1, <8 x i32> <i32 0, i32 9, i32 2, i32 11, i32 4, i32 13, i32 6, i32 15>
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = shufflevector <8 x i16> %1, <8 x i16> %2, <8 x i32> <i32 0, i32 1, i32 10, i32 11, i32 4, i32 5, i32 6, i32 15>
   ret <8 x i16> %3
 }
 
 define <2 x i64> @test_pcmpeqq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
 ; GENERIC-LABEL: test_pcmpeqq:

 ; BTVER2-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpcmpeqq (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pcmpeqq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpcmpeqq (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = icmp eq <2 x i64> %a0, %a1
   %2 = sext <2 x i1> %1 to <2 x i64>
   %3 = load <2 x i64>, <2 x i64>*%a2, align 16
   %4 = icmp eq <2 x i64> %2, %3
   %5 = sext <2 x i1> %4 to <2 x i64>
   ret <2 x i64> %5
 }
 

 ; BTVER2-NEXT:    vpextrb $3, %xmm0, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    vpextrb $1, %xmm0, (%rdi) # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pextrb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpextrb $3, %xmm0, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpextrb $1, %xmm0, (%rdi) # sched: [8:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = extractelement <16 x i8> %a0, i32 3
   %2 = extractelement <16 x i8> %a0, i32 1
   store i8 %2, i8 *%a1
   %3 = zext i8 %1 to i32
   ret i32 %3
 }
 
 define i32 @test_pextrd(<4 x i32> %a0, i32 *%a1) {

 ; BTVER2-NEXT:    vpextrd $3, %xmm0, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    vpextrd $1, %xmm0, (%rdi) # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pextrd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpextrd $3, %xmm0, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpextrd $1, %xmm0, (%rdi) # sched: [8:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = extractelement <4 x i32> %a0, i32 3
   %2 = extractelement <4 x i32> %a0, i32 1
   store i32 %2, i32 *%a1
   ret i32 %1
 }
 
 define i64 @test_pextrq(<2 x i64> %a0, <2 x i64> %a1, i64 *%a2) {
 ; GENERIC-LABEL: test_pextrq:

 ; BTVER2-NEXT:    vpextrq $1, %xmm0, %rax # sched: [1:0.50]
 ; BTVER2-NEXT:    vpextrq $1, %xmm0, (%rdi) # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pextrq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpextrq $1, %xmm0, %rax # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpextrq $1, %xmm0, (%rdi) # sched: [8:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = extractelement <2 x i64> %a0, i32 1
   %2 = extractelement <2 x i64> %a0, i32 1
   store i64 %2, i64 *%a2
   ret i64 %1
 }
 
 define i32 @test_pextrw(<8 x i16> %a0, i16 *%a1) {
 ; GENERIC-LABEL: test_pextrw:

 ; BTVER2-NEXT:    vpextrw $3, %xmm0, %eax # sched: [1:0.50]
 ; BTVER2-NEXT:    vpextrw $1, %xmm0, (%rdi) # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pextrw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpextrw $3, %xmm0, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpextrw $1, %xmm0, (%rdi) # sched: [8:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = extractelement <8 x i16> %a0, i32 3
   %2 = extractelement <8 x i16> %a0, i32 1
   store i16 %2, i16 *%a1
   %3 = zext i16 %1 to i32
   ret i32 %3
 }
 
 define <8 x i16> @test_phminposuw(<8 x i16> *%a0) {

 ; BTVER2-NEXT:    vphminposuw (%rdi), %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    vphminposuw %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_phminposuw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vphminposuw (%rdi), %xmm0 # sched: [11:1.00]
 ; ZNVER1-NEXT:    vphminposuw %xmm0, %xmm0 # sched: [4:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = load <8 x i16>, <8 x i16> *%a0, align 16
   %2 = call <8 x i16> @llvm.x86.sse41.phminposuw(<8 x i16> %1)
   %3 = call <8 x i16> @llvm.x86.sse41.phminposuw(<8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.sse41.phminposuw(<8 x i16>) nounwind readnone
 
 define <16 x i8> @test_pinsrb(<16 x i8> %a0, i8 %a1, i8 *%a2) {

 ; BTVER2-NEXT:    vpinsrb $1, %edi, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpinsrb $3, (%rsi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pinsrb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpinsrb $1, %edi, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpinsrb $3, (%rsi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = insertelement <16 x i8> %a0, i8 %a1, i32 1
   %2 = load i8, i8 *%a2
   %3 = insertelement <16 x i8> %1, i8 %2, i32 3
   ret <16 x i8> %3
 }
 
 define <4 x i32> @test_pinsrd(<4 x i32> %a0, i32 %a1, i32 *%a2) {
 ; GENERIC-LABEL: test_pinsrd:

 ; BTVER2-NEXT:    vpinsrd $1, %edi, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpinsrd $3, (%rsi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pinsrd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpinsrd $1, %edi, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpinsrd $3, (%rsi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = insertelement <4 x i32> %a0, i32 %a1, i32 1
   %2 = load i32, i32 *%a2
   %3 = insertelement <4 x i32> %1, i32 %2, i32 3
   ret <4 x i32> %3
 }
 
 define <2 x i64> @test_pinsrq(<2 x i64> %a0, <2 x i64> %a1, i64 %a2, i64 *%a3) {
 ; GENERIC-LABEL: test_pinsrq:

 ; BTVER2-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pinsrq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpinsrq $1, (%rsi), %xmm1, %xmm1 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpinsrq $1, %rdi, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = insertelement <2 x i64> %a0, i64 %a2, i32 1
   %2 = load i64, i64 *%a3
   %3 = insertelement <2 x i64> %a1, i64 %2, i32 1
   %4 = add <2 x i64> %1, %3
   ret <2 x i64> %4
 }
 
 define <16 x i8> @test_pmaxsb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {

 ; BTVER2-NEXT:    vpmaxsb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpmaxsb (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmaxsb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmaxsb %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpmaxsb (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.sse41.pmaxsb(<16 x i8> %a0, <16 x i8> %a1)
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = call <16 x i8> @llvm.x86.sse41.pmaxsb(<16 x i8> %1, <16 x i8> %2)
   ret <16 x i8> %3
 }
 declare <16 x i8> @llvm.x86.sse41.pmaxsb(<16 x i8>, <16 x i8>) nounwind readnone
 
 define <4 x i32> @test_pmaxsd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {

 ; BTVER2-NEXT:    vpmaxsd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpmaxsd (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmaxsd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmaxsd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpmaxsd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x i32> @llvm.x86.sse41.pmaxsd(<4 x i32> %a0, <4 x i32> %a1)
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = call <4 x i32> @llvm.x86.sse41.pmaxsd(<4 x i32> %1, <4 x i32> %2)
   ret <4 x i32> %3
 }
 declare <4 x i32> @llvm.x86.sse41.pmaxsd(<4 x i32>, <4 x i32>) nounwind readnone
 
 define <4 x i32> @test_pmaxud(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {

 ; BTVER2-NEXT:    vpmaxud %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpmaxud (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmaxud:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmaxud %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpmaxud (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x i32> @llvm.x86.sse41.pmaxud(<4 x i32> %a0, <4 x i32> %a1)
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = call <4 x i32> @llvm.x86.sse41.pmaxud(<4 x i32> %1, <4 x i32> %2)
   ret <4 x i32> %3
 }
 declare <4 x i32> @llvm.x86.sse41.pmaxud(<4 x i32>, <4 x i32>) nounwind readnone
 
 define <8 x i16> @test_pmaxuw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vpmaxuw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpmaxuw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmaxuw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmaxuw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpmaxuw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.sse41.pmaxuw(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.sse41.pmaxuw(<8 x i16> %1, <8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.sse41.pmaxuw(<8 x i16>, <8 x i16>) nounwind readnone
 
 define <16 x i8> @test_pminsb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {

 ; BTVER2-NEXT:    vpminsb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpminsb (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pminsb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpminsb %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpminsb (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.sse41.pminsb(<16 x i8> %a0, <16 x i8> %a1)
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = call <16 x i8> @llvm.x86.sse41.pminsb(<16 x i8> %1, <16 x i8> %2)
   ret <16 x i8> %3
 }
 declare <16 x i8> @llvm.x86.sse41.pminsb(<16 x i8>, <16 x i8>) nounwind readnone
 
 define <4 x i32> @test_pminsd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {

 ; BTVER2-NEXT:    vpminsd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpminsd (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pminsd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpminsd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpminsd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x i32> @llvm.x86.sse41.pminsd(<4 x i32> %a0, <4 x i32> %a1)
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = call <4 x i32> @llvm.x86.sse41.pminsd(<4 x i32> %1, <4 x i32> %2)
   ret <4 x i32> %3
 }
 declare <4 x i32> @llvm.x86.sse41.pminsd(<4 x i32>, <4 x i32>) nounwind readnone
 
 define <4 x i32> @test_pminud(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {

 ; BTVER2-NEXT:    vpminud %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpminud (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pminud:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpminud %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpminud (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x i32> @llvm.x86.sse41.pminud(<4 x i32> %a0, <4 x i32> %a1)
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = call <4 x i32> @llvm.x86.sse41.pminud(<4 x i32> %1, <4 x i32> %2)
   ret <4 x i32> %3
 }
 declare <4 x i32> @llvm.x86.sse41.pminud(<4 x i32>, <4 x i32>) nounwind readnone
 
 define <8 x i16> @test_pminuw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vpminuw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpminuw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pminuw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpminuw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpminuw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.sse41.pminuw(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.sse41.pminuw(<8 x i16> %1, <8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.sse41.pminuw(<8 x i16>, <8 x i16>) nounwind readnone
 
 define <8 x i16> @test_pmovsxbw(<16 x i8> %a0, <8 x i8> *%a1) {

 ; BTVER2-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmovsxbw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmovsxbw (%rdi), %xmm1 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpmovsxbw %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
   %2 = sext <8 x i8> %1 to <8 x i16>
   %3 = load <8 x i8>, <8 x i8>* %a1, align 1
   %4 = sext <8 x i8> %3 to <8 x i16>
   %5 = add <8 x i16> %2, %4
   ret <8 x i16> %5
 }
 

 ; BTVER2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmovsxbd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmovsxbd (%rdi), %xmm1 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpmovsxbd %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %2 = sext <4 x i8> %1 to <4 x i32>
   %3 = load <4 x i8>, <4 x i8>* %a1, align 1
   %4 = sext <4 x i8> %3 to <4 x i32>
   %5 = add <4 x i32> %2, %4
   ret <4 x i32> %5
 }
 

 ; BTVER2-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmovsxbq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmovsxbq (%rdi), %xmm1 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpmovsxbq %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <2 x i32> <i32 0, i32 1>
   %2 = sext <2 x i8> %1 to <2 x i64>
   %3 = load <2 x i8>, <2 x i8>* %a1, align 1
   %4 = sext <2 x i8> %3 to <2 x i64>
   %5 = add <2 x i64> %2, %4
   ret <2 x i64> %5
 }
 

 ; BTVER2-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmovsxdq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmovsxdq (%rdi), %xmm1 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpmovsxdq %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x i32> %a0, <4 x i32> undef, <2 x i32> <i32 0, i32 1>
   %2 = sext <2 x i32> %1 to <2 x i64>
   %3 = load <2 x i32>, <2 x i32>* %a1, align 1
   %4 = sext <2 x i32> %3 to <2 x i64>
   %5 = add <2 x i64> %2, %4
   ret <2 x i64> %5
 }
 

 ; BTVER2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmovsxwd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmovsxwd (%rdi), %xmm1 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpmovsxwd %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x i16> %a0, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %2 = sext <4 x i16> %1 to <4 x i32>
   %3 = load <4 x i16>, <4 x i16>* %a1, align 1
   %4 = sext <4 x i16> %3 to <4 x i32>
   %5 = add <4 x i32> %2, %4
   ret <4 x i32> %5
 }
 

 ; BTVER2-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmovsxwq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmovsxwq (%rdi), %xmm1 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpmovsxwq %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x i16> %a0, <8 x i16> undef, <2 x i32> <i32 0, i32 1>
   %2 = sext <2 x i16> %1 to <2 x i64>
   %3 = load <2 x i16>, <2 x i16>* %a1, align 1
   %4 = sext <2 x i16> %3 to <2 x i64>
   %5 = add <2 x i64> %2, %4
   ret <2 x i64> %5
 }
 

 ; BTVER2-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmovzxbw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmovzxbw {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero sched: [8:0.50]
 ; ZNVER1-NEXT:    vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
   %2 = zext <8 x i8> %1 to <8 x i16>
   %3 = load <8 x i8>, <8 x i8>* %a1, align 1
   %4 = zext <8 x i8> %3 to <8 x i16>
   %5 = add <8 x i16> %2, %4
   ret <8 x i16> %5
 }
 

 ; BTVER2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmovzxbd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmovzxbd {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [8:0.50]
 ; ZNVER1-NEXT:    vpmovzxbd {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %2 = zext <4 x i8> %1 to <4 x i32>
   %3 = load <4 x i8>, <4 x i8>* %a1, align 1
   %4 = zext <4 x i8> %3 to <4 x i32>
   %5 = add <4 x i32> %2, %4
   ret <4 x i32> %5
 }
 

 ; BTVER2-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmovzxbq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmovzxbq {{.*#+}} xmm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero sched: [8:0.50]
 ; ZNVER1-NEXT:    vpmovzxbq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <2 x i32> <i32 0, i32 1>
   %2 = zext <2 x i8> %1 to <2 x i64>
   %3 = load <2 x i8>, <2 x i8>* %a1, align 1
   %4 = zext <2 x i8> %3 to <2 x i64>
   %5 = add <2 x i64> %2, %4
   ret <2 x i64> %5
 }
 

 ; BTVER2-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmovzxdq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmovzxdq {{.*#+}} xmm1 = mem[0],zero,mem[1],zero sched: [8:0.50]
 ; ZNVER1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <4 x i32> %a0, <4 x i32> undef, <2 x i32> <i32 0, i32 1>
   %2 = zext <2 x i32> %1 to <2 x i64>
   %3 = load <2 x i32>, <2 x i32>* %a1, align 1
   %4 = zext <2 x i32> %3 to <2 x i64>
   %5 = add <2 x i64> %2, %4
   ret <2 x i64> %5
 }
 

 ; BTVER2-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmovzxwd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmovzxwd {{.*#+}} xmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [8:0.50]
 ; ZNVER1-NEXT:    vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x i16> %a0, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %2 = zext <4 x i16> %1 to <4 x i32>
   %3 = load <4 x i16>, <4 x i16>* %a1, align 1
   %4 = zext <4 x i16> %3 to <4 x i32>
   %5 = add <4 x i32> %2, %4
   ret <4 x i32> %5
 }
 

 ; BTVER2-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmovzxwq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmovzxwq {{.*#+}} xmm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero sched: [8:0.50]
 ; ZNVER1-NEXT:    vpmovzxwq {{.*#+}} xmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero sched: [1:0.25]
 ; ZNVER1-NEXT:    vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x i16> %a0, <8 x i16> undef, <2 x i32> <i32 0, i32 1>
   %2 = zext <2 x i16> %1 to <2 x i64>
   %3 = load <2 x i16>, <2 x i16>* %a1, align 1
   %4 = zext <2 x i16> %3 to <2 x i64>
   %5 = add <2 x i64> %2, %4
   ret <2 x i64> %5
 }
 

 ; BTVER2-NEXT:    vpmuldq %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vpmuldq (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmuldq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmuldq %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
 ; ZNVER1-NEXT:    vpmuldq (%rdi), %xmm0, %xmm0 # sched: [11:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x i64> @llvm.x86.sse41.pmuldq(<4 x i32> %a0, <4 x i32> %a1)
   %2 = bitcast <2 x i64> %1 to <4 x i32>
   %3 = load <4 x i32>, <4 x i32> *%a2, align 16
   %4 = call <2 x i64> @llvm.x86.sse41.pmuldq(<4 x i32> %2, <4 x i32> %3)
   ret <2 x i64> %4
 }
 declare <2 x i64> @llvm.x86.sse41.pmuldq(<4 x i32>, <4 x i32>) nounwind readnone
 

 ; BTVER2-NEXT:    vpmulld %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vpmulld (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmulld:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmulld %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
 ; ZNVER1-NEXT:    vpmulld (%rdi), %xmm0, %xmm0 # sched: [11:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = mul <4 x i32> %a0, %a1
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = mul <4 x i32> %1, %2
   ret <4 x i32> %3
 }
 
 define i32 @test_ptest(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
 ; GENERIC-LABEL: test_ptest:

 ; ZNVER1-LABEL: test_ptest:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vptest %xmm1, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    setb %al # sched: [1:0.25]
 ; ZNVER1-NEXT:    vptest (%rdi), %xmm0 # sched: [8:0.50]
 ; ZNVER1-NEXT:    setb %cl # sched: [1:0.25]
 ; ZNVER1-NEXT:    andb %al, %cl # sched: [1:0.25]
 ; ZNVER1-NEXT:    movzbl %cl, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.sse41.ptestc(<2 x i64> %a0, <2 x i64> %a1)
   %2 = load <2 x i64>, <2 x i64> *%a2, align 16
   %3 = call i32 @llvm.x86.sse41.ptestc(<2 x i64> %a0, <2 x i64> %2)
   %4 = and i32 %1, %3
   ret i32 %4
 }
 declare i32 @llvm.x86.sse41.ptestc(<2 x i64>, <2 x i64>) nounwind readnone
 

 ; BTVER2-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_roundpd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vroundpd $7, (%rdi), %xmm1 # sched: [10:1.00]
 ; ZNVER1-NEXT:    vroundpd $7, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x double> @llvm.x86.sse41.round.pd(<2 x double> %a0, i32 7)
   %2 = load <2 x double>, <2 x double> *%a1, align 16
   %3 = call <2 x double> @llvm.x86.sse41.round.pd(<2 x double> %2, i32 7)
   %4 = fadd <2 x double> %1, %3
   ret <2 x double> %4
 }
 declare <2 x double> @llvm.x86.sse41.round.pd(<2 x double>, i32) nounwind readnone
 

 ; BTVER2-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_roundps:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vroundps $7, (%rdi), %xmm1 # sched: [10:1.00]
 ; ZNVER1-NEXT:    vroundps $7, %xmm0, %xmm0 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.sse41.round.ps(<4 x float> %a0, i32 7)
   %2 = load <4 x float>, <4 x float> *%a1, align 16
   %3 = call <4 x float> @llvm.x86.sse41.round.ps(<4 x float> %2, i32 7)
   %4 = fadd <4 x float> %1, %3
   ret <4 x float> %4
 }
 declare <4 x float> @llvm.x86.sse41.round.ps(<4 x float>, i32) nounwind readnone
 

 ; BTVER2-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_roundsd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vroundsd $7, %xmm1, %xmm0, %xmm1 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vroundsd $7, (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
 ; ZNVER1-NEXT:    vaddpd %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <2 x double> @llvm.x86.sse41.round.sd(<2 x double> %a0, <2 x double> %a1, i32 7)
   %2 = load <2 x double>, <2 x double>* %a2, align 16
   %3 = call <2 x double> @llvm.x86.sse41.round.sd(<2 x double> %a0, <2 x double> %2, i32 7)
   %4 = fadd <2 x double> %1, %3
   ret <2 x double> %4
 }
 declare <2 x double> @llvm.x86.sse41.round.sd(<2 x double>, <2 x double>, i32) nounwind readnone
 

 ; BTVER2-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_roundss:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vroundss $7, %xmm1, %xmm0, %xmm1 # sched: [3:1.00]
 ; ZNVER1-NEXT:    vroundss $7, (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
 ; ZNVER1-NEXT:    vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x float> @llvm.x86.sse41.round.ss(<4 x float> %a0, <4 x float> %a1, i32 7)
   %2 = load <4 x float>, <4 x float> *%a2, align 16
   %3 = call <4 x float> @llvm.x86.sse41.round.ss(<4 x float> %a0, <4 x float> %2, i32 7)
   %4 = fadd <4 x float> %1, %3
   ret <4 x float> %4
 }
 declare <4 x float> @llvm.x86.sse41.round.ss(<4 x float>, <4 x float>, i32) nounwind readnone

test/CodeGen/X86/sse42-schedule.ll

 ; BTVER2-NEXT:    movl %edi, %eax # sched: [1:0.17]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: crc32_32_8:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    crc32b %sil, %edi # sched: [3:1.00]
 ; ZNVER1-NEXT:    crc32b (%rdx), %edi # sched: [10:1.00]
 ; ZNVER1-NEXT:    movl %edi, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.sse42.crc32.32.8(i32 %a0, i8 %a1)
   %2 = load i8, i8 *%a2
   %3 = call i32 @llvm.x86.sse42.crc32.32.8(i32 %1, i8 %2)
   ret i32 %3
 }
 declare i32 @llvm.x86.sse42.crc32.32.8(i32, i8) nounwind
 
 define i32 @crc32_32_16(i32 %a0, i16 %a1, i16 *%a2) {

 ; BTVER2-NEXT:    movl %edi, %eax # sched: [1:0.17]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: crc32_32_16:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    crc32w %si, %edi # sched: [3:1.00]
 ; ZNVER1-NEXT:    crc32w (%rdx), %edi # sched: [10:1.00]
 ; ZNVER1-NEXT:    movl %edi, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.sse42.crc32.32.16(i32 %a0, i16 %a1)
   %2 = load i16, i16 *%a2
   %3 = call i32 @llvm.x86.sse42.crc32.32.16(i32 %1, i16 %2)
   ret i32 %3
 }
 declare i32 @llvm.x86.sse42.crc32.32.16(i32, i16) nounwind
 
 define i32 @crc32_32_32(i32 %a0, i32 %a1, i32 *%a2) {

 ; BTVER2-NEXT:    movl %edi, %eax # sched: [1:0.17]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: crc32_32_32:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    crc32l %esi, %edi # sched: [3:1.00]
 ; ZNVER1-NEXT:    crc32l (%rdx), %edi # sched: [10:1.00]
 ; ZNVER1-NEXT:    movl %edi, %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.sse42.crc32.32.32(i32 %a0, i32 %a1)
   %2 = load i32, i32 *%a2
   %3 = call i32 @llvm.x86.sse42.crc32.32.32(i32 %1, i32 %2)
   ret i32 %3
 }
 declare i32 @llvm.x86.sse42.crc32.32.32(i32, i32) nounwind
 
 define i64 @crc32_64_8(i64 %a0, i8 %a1, i8 *%a2) nounwind {

 ; BTVER2-NEXT:    movq %rdi, %rax # sched: [1:0.17]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: crc32_64_8:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    crc32b %sil, %edi # sched: [3:1.00]
 ; ZNVER1-NEXT:    crc32b (%rdx), %edi # sched: [10:1.00]
 ; ZNVER1-NEXT:    movq %rdi, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i64 @llvm.x86.sse42.crc32.64.8(i64 %a0, i8 %a1)
   %2 = load i8, i8 *%a2
   %3 = call i64 @llvm.x86.sse42.crc32.64.8(i64 %1, i8 %2)
   ret i64 %3
 }
 declare i64 @llvm.x86.sse42.crc32.64.8(i64, i8) nounwind
 
 define i64 @crc32_64_64(i64 %a0, i64 %a1, i64 *%a2) {

 ; BTVER2-NEXT:    movq %rdi, %rax # sched: [1:0.17]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: crc32_64_64:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    crc32q %rsi, %rdi # sched: [3:1.00]
 ; ZNVER1-NEXT:    crc32q (%rdx), %rdi # sched: [10:1.00]
 ; ZNVER1-NEXT:    movq %rdi, %rax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i64 @llvm.x86.sse42.crc32.64.64(i64 %a0, i64 %a1)
   %2 = load i64, i64 *%a2
   %3 = call i64 @llvm.x86.sse42.crc32.64.64(i64 %1, i64 %2)
   ret i64 %3
 }
 declare i64 @llvm.x86.sse42.crc32.64.64(i64, i64) nounwind
 
 define i32 @test_pcmpestri(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {

 ; ZNVER1-NEXT:    movl $7, %edx # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpcmpestri $7, %xmm1, %xmm0 # sched: [100:?]
 ; ZNVER1-NEXT:    movl $7, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    movl $7, %edx # sched: [1:0.25]
 ; ZNVER1-NEXT:    movl %ecx, %esi # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpcmpestri $7, (%rdi), %xmm0 # sched: [100:?]
 ; ZNVER1-NEXT:    # kill: %ECX<def> %ECX<kill> %RCX<def>
 ; ZNVER1-NEXT:    leal (%rcx,%rsi), %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.sse42.pcmpestri128(<16 x i8> %a0, i32 7, <16 x i8> %a1, i32 7, i8 7)
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = call i32 @llvm.x86.sse42.pcmpestri128(<16 x i8> %a0, i32 7, <16 x i8> %2, i32 7, i8 7)
   %4 = add i32 %1, %3
   ret i32 %4
 }
 declare i32 @llvm.x86.sse42.pcmpestri128(<16 x i8>, i32, <16 x i8>, i32, i8) nounwind readnone
 

 ; ZNVER1-LABEL: test_pcmpestrm:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    movl $7, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    movl $7, %edx # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpcmpestrm $7, %xmm1, %xmm0 # sched: [100:?]
 ; ZNVER1-NEXT:    movl $7, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    movl $7, %edx # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpcmpestrm $7, (%rdi), %xmm0 # sched: [100:?]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.sse42.pcmpestrm128(<16 x i8> %a0, i32 7, <16 x i8> %a1, i32 7, i8 7)
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = call <16 x i8> @llvm.x86.sse42.pcmpestrm128(<16 x i8> %1, i32 7, <16 x i8> %2, i32 7, i8 7)
   ret <16 x i8> %3
 }
 declare <16 x i8> @llvm.x86.sse42.pcmpestrm128(<16 x i8>, i32, <16 x i8>, i32, i8) nounwind readnone
 
 define i32 @test_pcmpistri(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {

 ;
 ; ZNVER1-LABEL: test_pcmpistri:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpcmpistri $7, %xmm1, %xmm0 # sched: [100:?]
 ; ZNVER1-NEXT:    movl %ecx, %eax # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpcmpistri $7, (%rdi), %xmm0 # sched: [100:?]
 ; ZNVER1-NEXT:    # kill: %ECX<def> %ECX<kill> %RCX<def>
 ; ZNVER1-NEXT:    leal (%rcx,%rax), %eax # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call i32 @llvm.x86.sse42.pcmpistri128(<16 x i8> %a0, <16 x i8> %a1, i8 7)
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = call i32 @llvm.x86.sse42.pcmpistri128(<16 x i8> %a0, <16 x i8> %2, i8 7)
   %4 = add i32 %1, %3
   ret i32 %4
 }
 declare i32 @llvm.x86.sse42.pcmpistri128(<16 x i8>, <16 x i8>, i8) nounwind readnone
 

 ; BTVER2-NEXT:    vpcmpistrm $7, %xmm1, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    vpcmpistrm $7, (%rdi), %xmm0 # sched: [12:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pcmpistrm:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpcmpistrm $7, %xmm1, %xmm0 # sched: [100:?]
 ; ZNVER1-NEXT:    vpcmpistrm $7, (%rdi), %xmm0 # sched: [100:?]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.sse42.pcmpistrm128(<16 x i8> %a0, <16 x i8> %a1, i8 7)
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = call <16 x i8> @llvm.x86.sse42.pcmpistrm128(<16 x i8> %1, <16 x i8> %2, i8 7)
   ret <16 x i8> %3
 }
 declare <16 x i8> @llvm.x86.sse42.pcmpistrm128(<16 x i8>, <16 x i8>, i8) nounwind readnone
 
 define <2 x i64> @test_pcmpgtq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {

 ; BTVER2-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpcmpgtq (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pcmpgtq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpcmpgtq (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = icmp sgt <2 x i64> %a0, %a1
   %2 = sext <2 x i1> %1 to <2 x i64>
   %3 = load <2 x i64>, <2 x i64>*%a2, align 16
   %4 = icmp sgt <2 x i64> %2, %3
   %5 = sext <2 x i1> %4 to <2 x i64>
   ret <2 x i64> %5
 }

test/CodeGen/X86/sse4a-schedule.ll

 ; BTVER2-LABEL: test_extrq:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    extrq %xmm1, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_extrq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    extrq %xmm1, %xmm0
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = tail call <2 x i64> @llvm.x86.sse4a.extrq(<2 x i64> %a0, <16 x i8> %a1)
   ret <2 x i64> %1
 }
 declare <2 x i64> @llvm.x86.sse4a.extrq(<2 x i64>, <16 x i8>)
 
 define <2 x i64> @test_extrqi(<2 x i64> %a0) {
 ; GENERIC-LABEL: test_extrqi:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    extrq $2, $3, %xmm0
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; BTVER2-LABEL: test_extrqi:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    extrq $2, $3, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_extrqi:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    extrq $2, $3, %xmm0
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = tail call <2 x i64> @llvm.x86.sse4a.extrqi(<2 x i64> %a0, i8 3, i8 2)
   ret <2 x i64> %1
 }
 declare <2 x i64> @llvm.x86.sse4a.extrqi(<2 x i64>, i8, i8)
 
 define <2 x i64> @test_insertq(<2 x i64> %a0, <2 x i64> %a1) {
 ; GENERIC-LABEL: test_insertq:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    insertq %xmm1, %xmm0
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; BTVER2-LABEL: test_insertq:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    insertq %xmm1, %xmm0 # sched: [2:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_insertq:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    insertq %xmm1, %xmm0
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = tail call <2 x i64> @llvm.x86.sse4a.insertq(<2 x i64> %a0, <2 x i64> %a1)
   ret <2 x i64> %1
 }
 declare <2 x i64> @llvm.x86.sse4a.insertq(<2 x i64>, <2 x i64>)
 
 define <2 x i64> @test_insertqi(<2 x i64> %a0, <2 x i64> %a1) {
 ; GENERIC-LABEL: test_insertqi:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    insertq $6, $5, %xmm1, %xmm0
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; BTVER2-LABEL: test_insertqi:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    insertq $6, $5, %xmm1, %xmm0 # sched: [2:2.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_insertqi:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    insertq $6, $5, %xmm1, %xmm0
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = tail call <2 x i64> @llvm.x86.sse4a.insertqi(<2 x i64> %a0, <2 x i64> %a1, i8 5, i8 6)
   ret <2 x i64> %1
 }
 declare <2 x i64> @llvm.x86.sse4a.insertqi(<2 x i64>, <2 x i64>, i8, i8)
 
 define void @test_movntsd(i8* %p, <2 x double> %a) {
 ; GENERIC-LABEL: test_movntsd:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    movntsd %xmm0, (%rdi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; BTVER2-LABEL: test_movntsd:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    movntsd %xmm0, (%rdi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movntsd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    movntsd %xmm0, (%rdi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   tail call void @llvm.x86.sse4a.movnt.sd(i8* %p, <2 x double> %a)
   ret void
 }
 declare void @llvm.x86.sse4a.movnt.sd(i8*, <2 x double>)
 
 define void @test_movntss(i8* %p, <4 x float> %a) {
 ; GENERIC-LABEL: test_movntss:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    movntss %xmm0, (%rdi) # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; BTVER2-LABEL: test_movntss:
 ; BTVER2:       # BB#0:
 ; BTVER2-NEXT:    movntss %xmm0, (%rdi) # sched: [1:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_movntss:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    movntss %xmm0, (%rdi) # sched: [1:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   tail call void @llvm.x86.sse4a.movnt.ss(i8* %p, <4 x float> %a)
   ret void
 }
 declare void @llvm.x86.sse4a.movnt.ss(i8*, <4 x float>)
 

test/CodeGen/X86/ssse3-schedule.ll

 ; BTVER2-NEXT:    vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pabsb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpabsb (%rdi), %xmm1 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpabsb %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.ssse3.pabs.b.128(<16 x i8> %a0)
   %2 = load <16 x i8>, <16 x i8> *%a1, align 16
   %3 = call <16 x i8> @llvm.x86.ssse3.pabs.b.128(<16 x i8> %2)
   %4 = or <16 x i8> %1, %3
   ret <16 x i8> %4
 }
 declare <16 x i8> @llvm.x86.ssse3.pabs.b.128(<16 x i8>) nounwind readnone
 

 ; BTVER2-NEXT:    vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pabsd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpabsd (%rdi), %xmm1 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpabsd %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x i32> @llvm.x86.ssse3.pabs.d.128(<4 x i32> %a0)
   %2 = load <4 x i32>, <4 x i32> *%a1, align 16
   %3 = call <4 x i32> @llvm.x86.ssse3.pabs.d.128(<4 x i32> %2)
   %4 = or <4 x i32> %1, %3
   ret <4 x i32> %4
 }
 declare <4 x i32> @llvm.x86.ssse3.pabs.d.128(<4 x i32>) nounwind readnone
 

 ; BTVER2-NEXT:    vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pabsw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpabsw (%rdi), %xmm1 # sched: [8:0.50]
 ; ZNVER1-NEXT:    vpabsw %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.ssse3.pabs.w.128(<8 x i16> %a0)
   %2 = load <8 x i16>, <8 x i16> *%a1, align 16
   %3 = call <8 x i16> @llvm.x86.ssse3.pabs.w.128(<8 x i16> %2)
   %4 = or <8 x i16> %1, %3
   ret <8 x i16> %4
 }
 declare <8 x i16> @llvm.x86.ssse3.pabs.w.128(<8 x i16>) nounwind readnone
 

 ; BTVER2-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[6,7,8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5] sched: [1:0.50]
 ; BTVER2-NEXT:    vpalignr {{.*#+}} xmm0 = mem[14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13] sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_palignr:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpalignr {{.*#+}} xmm0 = xmm0[6,7,8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4,5] sched: [1:0.25]
 ; ZNVER1-NEXT:    vpalignr {{.*#+}} xmm0 = mem[14,15],xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13] sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = shufflevector <8 x i16> %a0, <8 x i16> %a1, <8 x i32> <i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10>
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = shufflevector <8 x i16> %2, <8 x i16> %1, <8 x i32> <i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14>
   ret <8 x i16> %3
 }
 
 define <4 x i32> @test_phaddd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
 ; GENERIC-LABEL: test_phaddd:

 ; BTVER2-NEXT:    vphaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vphaddd (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_phaddd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vphaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vphaddd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x i32> @llvm.x86.ssse3.phadd.d.128(<4 x i32> %a0, <4 x i32> %a1)
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = call <4 x i32> @llvm.x86.ssse3.phadd.d.128(<4 x i32> %1, <4 x i32> %2)
   ret <4 x i32> %3
 }
 declare <4 x i32> @llvm.x86.ssse3.phadd.d.128(<4 x i32>, <4 x i32>) nounwind readnone
 
 define <8 x i16> @test_phaddsw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vphaddsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vphaddsw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_phaddsw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vphaddsw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vphaddsw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.ssse3.phadd.sw.128(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.ssse3.phadd.sw.128(<8 x i16> %1, <8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.ssse3.phadd.sw.128(<8 x i16>, <8 x i16>) nounwind readnone
 
 define <8 x i16> @test_phaddw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vphaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vphaddw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_phaddw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vphaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vphaddw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.ssse3.phadd.w.128(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.ssse3.phadd.w.128(<8 x i16> %1, <8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.ssse3.phadd.w.128(<8 x i16>, <8 x i16>) nounwind readnone
 
 define <4 x i32> @test_phsubd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {

 ; BTVER2-NEXT:    vphsubd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vphsubd (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_phsubd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vphsubd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vphsubd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x i32> @llvm.x86.ssse3.phsub.d.128(<4 x i32> %a0, <4 x i32> %a1)
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = call <4 x i32> @llvm.x86.ssse3.phsub.d.128(<4 x i32> %1, <4 x i32> %2)
   ret <4 x i32> %3
 }
 declare <4 x i32> @llvm.x86.ssse3.phsub.d.128(<4 x i32>, <4 x i32>) nounwind readnone
 
 define <8 x i16> @test_phsubsw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vphsubsw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vphsubsw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_phsubsw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vphsubsw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vphsubsw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.ssse3.phsub.sw.128(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.ssse3.phsub.sw.128(<8 x i16> %1, <8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.ssse3.phsub.sw.128(<8 x i16>, <8 x i16>) nounwind readnone
 
 define <8 x i16> @test_phsubw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vphsubw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vphsubw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_phsubw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vphsubw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vphsubw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.ssse3.phsub.w.128(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.ssse3.phsub.w.128(<8 x i16> %1, <8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.ssse3.phsub.w.128(<8 x i16>, <8 x i16>) nounwind readnone
 
 define <8 x i16> @test_pmaddubsw(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {

 ; BTVER2-NEXT:    vpmaddubsw %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vpmaddubsw (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmaddubsw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmaddubsw %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
 ; ZNVER1-NEXT:    vpmaddubsw (%rdi), %xmm0, %xmm0 # sched: [11:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.ssse3.pmadd.ub.sw.128(<16 x i8> %a0, <16 x i8> %a1)
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = bitcast <8 x i16> %1 to <16 x i8>
   %4 = call <8 x i16> @llvm.x86.ssse3.pmadd.ub.sw.128(<16 x i8> %3, <16 x i8> %2)
   ret <8 x i16> %4
 }
 declare <8 x i16> @llvm.x86.ssse3.pmadd.ub.sw.128(<16 x i8>, <16 x i8>) nounwind readnone
 

 ; BTVER2-NEXT:    vpmulhrsw %xmm1, %xmm0, %xmm0 # sched: [2:1.00]
 ; BTVER2-NEXT:    vpmulhrsw (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pmulhrsw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpmulhrsw %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
 ; ZNVER1-NEXT:    vpmulhrsw (%rdi), %xmm0, %xmm0 # sched: [11:1.00]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.ssse3.pmul.hr.sw.128(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.ssse3.pmul.hr.sw.128(<8 x i16> %1, <8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.ssse3.pmul.hr.sw.128(<8 x i16>, <8 x i16>) nounwind readnone
 
 define <16 x i8> @test_pshufb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {

 ; BTVER2-NEXT:    vpshufb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpshufb (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_pshufb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpshufb %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpshufb (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %a0, <16 x i8> %a1)
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %1, <16 x i8> %2)
   ret <16 x i8> %3
 }
 declare <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8>, <16 x i8>) nounwind readnone
 
 define <16 x i8> @test_psignb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> *%a2) {

 ; BTVER2-NEXT:    vpsignb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpsignb (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psignb:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsignb %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsignb (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <16 x i8> @llvm.x86.ssse3.psign.b.128(<16 x i8> %a0, <16 x i8> %a1)
   %2 = load <16 x i8>, <16 x i8> *%a2, align 16
   %3 = call <16 x i8> @llvm.x86.ssse3.psign.b.128(<16 x i8> %1, <16 x i8> %2)
   ret <16 x i8> %3
 }
 declare <16 x i8> @llvm.x86.ssse3.psign.b.128(<16 x i8>, <16 x i8>) nounwind readnone
 
 define <4 x i32> @test_psignd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {

 ; BTVER2-NEXT:    vpsignd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpsignd (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psignd:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsignd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsignd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <4 x i32> @llvm.x86.ssse3.psign.d.128(<4 x i32> %a0, <4 x i32> %a1)
   %2 = load <4 x i32>, <4 x i32> *%a2, align 16
   %3 = call <4 x i32> @llvm.x86.ssse3.psign.d.128(<4 x i32> %1, <4 x i32> %2)
   ret <4 x i32> %3
 }
 declare <4 x i32> @llvm.x86.ssse3.psign.d.128(<4 x i32>, <4 x i32>) nounwind readnone
 
 define <8 x i16> @test_psignw(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {

 ; BTVER2-NEXT:    vpsignw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
 ; BTVER2-NEXT:    vpsignw (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
 ; BTVER2-NEXT:    retq # sched: [4:1.00]
 ;
 ; ZNVER1-LABEL: test_psignw:
 ; ZNVER1:       # BB#0:
 ; ZNVER1-NEXT:    vpsignw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
 ; ZNVER1-NEXT:    vpsignw (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT:    retq # sched: [5:0.50]
+; ZNVER1-NEXT:    retq # sched: [1:0.50]
   %1 = call <8 x i16> @llvm.x86.ssse3.psign.w.128(<8 x i16> %a0, <8 x i16> %a1)
   %2 = load <8 x i16>, <8 x i16> *%a2, align 16
   %3 = call <8 x i16> @llvm.x86.ssse3.psign.w.128(<8 x i16> %1, <8 x i16> %2)
   ret <8 x i16> %3
 }
 declare <8 x i16> @llvm.x86.ssse3.psign.w.128(<8 x i16>, <8 x i16>) nounwind readnone