[X86] Improved sched models for X86 CMPXCHG* instructions

[avt77]

Summary

This patch improves sched models for CMPXCHG* X86 instrs removing unnecessary redefinitions of instr infos. This patch is based on results of D48222.

[RKSimon]

lib/Target/X86/X86SchedBroadwell.td

122 ↗

(On Diff #158267)

What about the WriteCMPXCHGLd cases?

[avt77]

lib/Target/X86/X86SchedBroadwell.td

122 ↗

(On Diff #158267)

I decided (as with BT*rr instrs) it'd be better to have separate patches for 'rr' and 'Ld' versions.The problem is:

automaticaly generated WriteCMPXCHGLd does not fit to real numbers that's why it's better to use X86WriteRes for both 'rr' and 'Ld' versions instead of X86SchedWritePair. Should I do it in this patch?

[avt77]

The WriteCMPXCHGLd was implemented.

[craig.topper]

lib/Target/X86/X86InstrInfo.td

2083 ↗

(On Diff #158552)

If I remember right from arithmetic instructions, this doesn't make the latency additive the way it should be. This should probably be just WriteCMPXCHGRMW implemented as a WriteSequence like we do for WriteALURMW

[avt77]

lib/Target/X86/X86InstrInfo.td

2083 ↗

(On Diff #158552)

The given implementation is very similar to suggested WriteSequence (in fact it's the same) and keeps all current sched values w/o changes. Could we stay with the current implementation or it's better to re-implement it?

[avt77]

The code was re-based. We have 2 changed tests for SandyBridge (Generic). I did not find the proper numbers in Intel SDMs or on Agner site that's why I don't know what's beter the new results or the old ones. The old sched model does not use SBPort4 for memory ops and that looks strange for me. Could anyone help me with these tests?

[craig.topper]

lib/Target/X86/X86SchedSandyBridge.td

953 ↗

(On Diff #160325)

My best guess is 6 uops and 8 latency like the other Intel CPUs.

Resource cycles
Port015 - 1 cycle (equivalent to 0156 on other Intel CPUs)
Port05 - 2 cycles (equivalent to the Port06 on other Intel CPUs)
Port4 - 1 cycle
Port23 - 2 cycle

[avt77]

I fixed sched parameters accordingly to craig.topper suggestions.

[craig.topper]

lib/Target/X86/X86SchedSandyBridge.td

956 ↗

(On Diff #160554)

I didn't notice the regular expression here had changed. The information I gave was for WriteCMPXCHGLd

[avt77]

I fixed the issue with WriteCMPXCHGLd - tnx to craig.topper.

[RKSimon]

lib/Target/X86/X86InstrInfo.td

2083 ↗

(On Diff #161201)

Wasn't the plan to add a WriteCMPXCHGRMW WriteSequence class?

lib/Target/X86/X86SchedSandyBridge.td

956 ↗

(On Diff #160554)

Why has the regex changed?

[avt77]

lib/Target/X86/X86InstrInfo.td

2083 ↗

(On Diff #161201)

It was not the plan because it's the same change. And even more: I tried to switch to WriteCMPXCHGRMW and it did not work here. Should I try it again?

lib/Target/X86/X86SchedSandyBridge.td

956 ↗

(On Diff #160554)

Because we don't need this SBWriteResGroup81 any more. Maybe I would change the name but I kept it.

[craig.topper]

lib/Target/X86/X86SchedSandyBridge.td

119 ↗

(On Diff #161201)

The port4 and one of the port23 shouldn't be here. They come from WriteRMW that WriteCMPXCHGLd is paired with.

[avt77]

lib/Target/X86/X86SchedSandyBridge.td

119 ↗

(On Diff #161201)

Oh, yes, you're right. Tnx.

[avt77]

lib/Target/X86/X86SchedSandyBridge.td

119 ↗

(On Diff #161201)

The problem here is the following: we can't use the paired version because it uses the default values [1,1] but we need 2 cycles for port23. It seems we have to redefine these intrs for SB or to remove the pairing with WriteRMW. I'd prefer to remove the pairing if you don't mind.

[craig.topper]

lib/Target/X86/X86SchedSandyBridge.td

119 ↗

(On Diff #161201)

I'm fine with removing the pairing. But change the name to WriteCMPXCHGRMW

[avt77]

Now we have WriteCMPXCHGRMW accordingly to craig.topper requirement.

[craig.topper]

lib/Target/X86/X86SchedSandyBridge.td

119 ↗

(On Diff #162740)

Shouldn't this be 6 uops not 7?

[avt77]

lib/Target/X86/X86SchedSandyBridge.td

119 ↗

(On Diff #162740)

Yes, you suggested 6 in your first comments I'll fix it asap.

[avt77]

The issue with SB uops was fixed.

[craig.topper]

LGTM

llvm/trunk/lib/Target/X86/X86InstrInfo.td

def XADD32rm  : I<0xC1, MRMSrcMem, (outs GR32:$dst),
                  "xadd{l}\t{$val, $ptr|$ptr, $val}", []>, TB,
                  OpSize32;
 def XADD64rm  : RI<0xC1, MRMSrcMem, (outs GR64:$dst),
                    (ins GR64:$val, i64mem:$ptr),
                    "xadd{q}\t{$val, $ptr|$ptr, $val}", []>, TB;
 
 }
 
-let SchedRW = [WriteALU], hasSideEffects = 0 in {
+let SchedRW = [WriteCMPXCHG], hasSideEffects = 0 in {
 let Defs = [AL, EFLAGS], Uses = [AL] in
 def CMPXCHG8rr : I<0xB0, MRMDestReg, (outs GR8:$dst), (ins GR8:$src),
                    "cmpxchg{b}\t{$src, $dst|$dst, $src}", []>, TB,
                    NotMemoryFoldable;
 let Defs = [AX, EFLAGS], Uses = [AX] in
 def CMPXCHG16rr : I<0xB1, MRMDestReg, (outs GR16:$dst), (ins GR16:$src),
                     "cmpxchg{w}\t{$src, $dst|$dst, $src}", []>, TB, OpSize16,
                     NotMemoryFoldable;
 let Defs = [EAX, EFLAGS], Uses = [EAX] in
 def CMPXCHG32rr  : I<0xB1, MRMDestReg, (outs GR32:$dst), (ins GR32:$src),
                      "cmpxchg{l}\t{$src, $dst|$dst, $src}", []>, TB, OpSize32,
                      NotMemoryFoldable;
 let Defs = [RAX, EFLAGS], Uses = [RAX] in
 def CMPXCHG64rr  : RI<0xB1, MRMDestReg, (outs GR64:$dst), (ins GR64:$src),
                       "cmpxchg{q}\t{$src, $dst|$dst, $src}", []>, TB,
                       NotMemoryFoldable;
 } // SchedRW, hasSideEffects
 
-let SchedRW = [WriteALULd, WriteRMW], mayLoad = 1, mayStore = 1,
+let SchedRW = [WriteCMPXCHGRMW], mayLoad = 1, mayStore = 1,
     hasSideEffects = 0 in {
 let Defs = [AL, EFLAGS], Uses = [AL] in
 def CMPXCHG8rm   : I<0xB0, MRMDestMem, (outs), (ins i8mem:$dst, GR8:$src),
                      "cmpxchg{b}\t{$src, $dst|$dst, $src}", []>, TB,
                      NotMemoryFoldable;
 let Defs = [AX, EFLAGS], Uses = [AX] in
 def CMPXCHG16rm  : I<0xB1, MRMDestMem, (outs), (ins i16mem:$dst, GR16:$src),
                      "cmpxchg{w}\t{$src, $dst|$dst, $src}", []>, TB, OpSize16,

llvm/trunk/lib/Target/X86/X86SchedBroadwell.td

 defm : BWWriteResPair<WriteDiv16,  [BWPort0, BWDivider], 25, [1, 10]>;
 defm : BWWriteResPair<WriteDiv32,  [BWPort0, BWDivider], 25, [1, 10]>;
 defm : BWWriteResPair<WriteDiv64,  [BWPort0, BWDivider], 25, [1, 10]>;
 defm : BWWriteResPair<WriteIDiv8,  [BWPort0, BWDivider], 25, [1, 10]>;
 defm : BWWriteResPair<WriteIDiv16, [BWPort0, BWDivider], 25, [1, 10]>;
 defm : BWWriteResPair<WriteIDiv32, [BWPort0, BWDivider], 25, [1, 10]>;
 defm : BWWriteResPair<WriteIDiv64, [BWPort0, BWDivider], 25, [1, 10]>;
 
+defm : X86WriteRes<WriteCMPXCHG,[BWPort06, BWPort0156], 5, [2, 3], 5>;
+defm : X86WriteRes<WriteCMPXCHGRMW,[BWPort23, BWPort06, BWPort0156, BWPort237, BWPort4], 8, [1, 2, 1, 1, 1], 6>;
 defm : X86WriteRes<WriteBSWAP32,   [BWPort15], 1, [1], 1>;
 defm : X86WriteRes<WriteBSWAP64,   [BWPort06, BWPort15], 2, [1, 1], 2>;
 defm : X86WriteRes<WriteXCHG,      [BWPort0156], 2, [3], 3>;
 
 defm : BWWriteResPair<WriteCRC32, [BWPort1],   3>;
 def : WriteRes<WriteIMulH, []> { let Latency = 3; } // Integer multiplication, high part.
 
 def : WriteRes<WriteLEA, [BWPort15]>; // LEA instructions can't fold loads.

 
 def BWWriteResGroup55 : SchedWriteRes<[BWPort06,BWPort0156]> {
   let Latency = 5;
   let NumMicroOps = 5;
   let ResourceCycles = [1,4];
 }
 def: InstRW<[BWWriteResGroup55], (instrs XSETBV)>;
 
-def BWWriteResGroup56 : SchedWriteRes<[BWPort06,BWPort0156]> {
-  let Latency = 5;
-  let NumMicroOps = 5;
-  let ResourceCycles = [2,3];
-}
-def: InstRW<[BWWriteResGroup56], (instregex "CMPXCHG(8|16|32|64)rr")>;
-
 def BWWriteResGroup57 : SchedWriteRes<[BWPort4,BWPort237,BWPort0156]> {
   let Latency = 5;
   let NumMicroOps = 6;
   let ResourceCycles = [1,1,4];
 }
 def: InstRW<[BWWriteResGroup57], (instregex "PUSHF(16|64)")>;
 
 def BWWriteResGroup58 : SchedWriteRes<[BWPort23]> {

 def: InstRW<[BWWriteResGroup99], (instregex "XCHG(8|16|32|64)rm")>;
 
 def BWWriteResGroup100 : SchedWriteRes<[BWPort4,BWPort23,BWPort237,BWPort06,BWPort0156]> {
   let Latency = 8;
   let NumMicroOps = 6;
   let ResourceCycles = [1,1,1,2,1];
 }
 def : SchedAlias<WriteADCRMW, BWWriteResGroup100>;
-def: InstRW<[BWWriteResGroup100], (instregex "CMPXCHG(8|16|32|64)rm",
-                                             "ROL(8|16|32|64)mCL",
+def: InstRW<[BWWriteResGroup100], (instregex "ROL(8|16|32|64)mCL",
                                              "SAR(8|16|32|64)mCL",
                                              "SHL(8|16|32|64)mCL",
                                              "SHR(8|16|32|64)mCL")>;
 
 def BWWriteResGroup101 : SchedWriteRes<[BWPort1,BWPort23]> {
   let Latency = 9;
   let NumMicroOps = 2;
   let ResourceCycles = [1,1];

llvm/trunk/lib/Target/X86/X86SchedHaswell.td

 // Arithmetic.
 defm : HWWriteResPair<WriteALU,    [HWPort0156], 1>;
 defm : HWWriteResPair<WriteADC,    [HWPort06, HWPort0156], 2, [1,1], 2>;
 defm : HWWriteResPair<WriteIMul,   [HWPort1],   3>;
 defm : HWWriteResPair<WriteIMul64, [HWPort1],   3>;
 
 defm : X86WriteRes<WriteBSWAP32,   [HWPort15], 1, [1], 1>;
 defm : X86WriteRes<WriteBSWAP64,   [HWPort06, HWPort15], 2, [1,1], 2>;
+defm : X86WriteRes<WriteCMPXCHG,[HWPort06, HWPort0156], 5, [2,3], 5>;
+defm : X86WriteRes<WriteCMPXCHGRMW,[HWPort23,HWPort06,HWPort0156,HWPort237,HWPort4], 9, [1,2,1,1,1], 6>;
 defm : X86WriteRes<WriteXCHG, [HWPort0156], 2, [3], 3>;
 
 def  : WriteRes<WriteIMulH, []> { let Latency = 3; }
 
 // Integer shifts and rotates.
 defm : HWWriteResPair<WriteShift,  [HWPort06],  1>;
 
 // SHLD/SHRD.

 }
 def: InstRW<[HWWriteResGroup68], (instregex "XCHG(8|16|32|64)rm")>;
 
 def HWWriteResGroup69 : SchedWriteRes<[HWPort4,HWPort23,HWPort237,HWPort06,HWPort0156]> {
   let Latency = 9;
   let NumMicroOps = 6;
   let ResourceCycles = [1,1,1,2,1];
 }
-def: InstRW<[HWWriteResGroup69], (instregex "CMPXCHG(8|16|32|64)rm",
-                                            "ROL(8|16|32|64)mCL",
+def: InstRW<[HWWriteResGroup69], (instregex "ROL(8|16|32|64)mCL",
                                             "SAR(8|16|32|64)mCL",
                                             "SHL(8|16|32|64)mCL",
                                             "SHR(8|16|32|64)mCL")>;
 def: SchedAlias<WriteADCRMW, HWWriteResGroup69>;
 
 def HWWriteResGroup70 : SchedWriteRes<[HWPort0,HWPort1]> {
   let Latency = 4;
   let NumMicroOps = 2;

 
 def HWWriteResGroup100 : SchedWriteRes<[HWPort06,HWPort0156]> {
   let Latency = 5;
   let NumMicroOps = 5;
   let ResourceCycles = [1,4];
 }
 def: InstRW<[HWWriteResGroup100], (instrs XSETBV)>;
 
-def HWWriteResGroup101 : SchedWriteRes<[HWPort06,HWPort0156]> {
-  let Latency = 5;
-  let NumMicroOps = 5;
-  let ResourceCycles = [2,3];
-}
-def: InstRW<[HWWriteResGroup101], (instregex "CMPXCHG(8|16|32|64)rr")>;
-
 def HWWriteResGroup102 : SchedWriteRes<[HWPort1,HWPort5]> {
   let Latency = 6;
   let NumMicroOps = 2;
   let ResourceCycles = [1,1];
 }
 def: InstRW<[HWWriteResGroup102], (instrs VCVTDQ2PDYrr,
                                           VCVTPD2PSYrr,
                                           VCVTPD2DQYrr,

llvm/trunk/lib/Target/X86/X86SchedSandyBridge.td

 // Arithmetic.
 defm : SBWriteResPair<WriteALU,    [SBPort015], 1>;
 defm : SBWriteResPair<WriteADC,    [SBPort05,SBPort015], 2, [1,1], 2>;
 defm : SBWriteResPair<WriteIMul,   [SBPort1],   3>;
 defm : SBWriteResPair<WriteIMul64, [SBPort1],   3>;
 defm : X86WriteRes<WriteXCHG,      [SBPort015], 2, [3], 3>;
 
 defm : X86WriteRes<WriteBSWAP32,   [SBPort1], 1, [1], 1>;
-defm : X86WriteRes<WriteBSWAP64,   [SBPort1,SBPort05], 2, [1,1], 2>;
+defm : X86WriteRes<WriteBSWAP64,   [SBPort1, SBPort05], 2, [1,1], 2>;
+defm : X86WriteRes<WriteCMPXCHG,   [SBPort05, SBPort015], 5, [1,3], 4>;
+defm : X86WriteRes<WriteCMPXCHGRMW,[SBPort015, SBPort5, SBPort23, SBPort4], 8, [1, 2, 2, 1], 6>;
 
 defm : SBWriteResPair<WriteDiv8,   [SBPort0, SBDivider], 25, [1, 10]>;
 defm : SBWriteResPair<WriteDiv16,  [SBPort0, SBDivider], 25, [1, 10]>;
 defm : SBWriteResPair<WriteDiv32,  [SBPort0, SBDivider], 25, [1, 10]>;
 defm : SBWriteResPair<WriteDiv64,  [SBPort0, SBDivider], 25, [1, 10]>;
 defm : SBWriteResPair<WriteIDiv8,  [SBPort0, SBDivider], 25, [1, 10]>;
 defm : SBWriteResPair<WriteIDiv16, [SBPort0, SBDivider], 25, [1, 10]>;
 defm : SBWriteResPair<WriteIDiv32, [SBPort0, SBDivider], 25, [1, 10]>;

 
 def SBWriteResGroup41 : SchedWriteRes<[SBPort5,SBPort015]> {
   let Latency = 5;
   let NumMicroOps = 4;
   let ResourceCycles = [1,3];
 }
 def: InstRW<[SBWriteResGroup41], (instrs FNINIT)>;
 
-def SBWriteResGroup42 : SchedWriteRes<[SBPort05,SBPort015]> {
-  let Latency = 5;
-  let NumMicroOps = 4;
-  let ResourceCycles = [1,3];
-}
-def: InstRW<[SBWriteResGroup42], (instregex "CMPXCHG(8|16|32|64)rr")>;
-
 def SBWriteResGroup43 : SchedWriteRes<[SBPort4,SBPort23,SBPort05]> {
   let Latency = 3;
   let NumMicroOps = 4;
   let ResourceCycles = [1,1,2];
 }
 def: InstRW<[SBWriteResGroup43], (instregex "SET(A|BE)m")>;
 
 def SBWriteResGroup45 : SchedWriteRes<[SBPort0,SBPort4,SBPort23,SBPort15]> {

 
 def SBWriteResGroup77 : SchedWriteRes<[SBPort0,SBPort1,SBPort23]> {
   let Latency = 8;
   let NumMicroOps = 3;
   let ResourceCycles = [1,1,1];
 }
 def: InstRW<[SBWriteResGroup77], (instregex "(V?)(U?)COMI(SD|SS)rm")>;
 
-def SBWriteResGroup81 : SchedWriteRes<[SBPort23,SBPort015]> {
-  let Latency = 8;
-  let NumMicroOps = 4;
-  let ResourceCycles = [1,3];
+def SBWriteResGroup81 : SchedWriteRes<[SBPort4, SBPort23, SBPort015]> {
+  let Latency = 6;
+  let NumMicroOps = 3;
+  let ResourceCycles = [1, 2, 1];
 }
-def: InstRW<[SBWriteResGroup81], (instregex "CMPXCHG(8|16|32|64)rm")>;
+def: InstRW<[SBWriteResGroup81], (instregex "CMPXCHG(8|16)B")>;
 
 def SBWriteResGroup83 : SchedWriteRes<[SBPort23,SBPort015]> {
   let Latency = 8;
   let NumMicroOps = 5;
   let ResourceCycles = [2,3];
 }
 def: InstRW<[SBWriteResGroup83], (instrs CMPSB,
                                          CMPSL,

llvm/trunk/lib/Target/X86/X86SchedSkylakeClient.td

 // Arithmetic.
 defm : SKLWriteResPair<WriteALU,    [SKLPort0156], 1>; // Simple integer ALU op.
 defm : SKLWriteResPair<WriteADC,    [SKLPort06],   1>; // Integer ALU + flags op.
 defm : SKLWriteResPair<WriteIMul,   [SKLPort1],    3>; // Integer multiplication.
 defm : SKLWriteResPair<WriteIMul64, [SKLPort1],    3>; // Integer 64-bit multiplication.
 
 defm : X86WriteRes<WriteBSWAP32,    [SKLPort15], 1, [1], 1>;
 defm : X86WriteRes<WriteBSWAP64,    [SKLPort06, SKLPort15], 2, [1,1], 2>;
+defm : X86WriteRes<WriteCMPXCHG,[SKLPort06, SKLPort0156], 5, [2,3], 5>;
+defm : X86WriteRes<WriteCMPXCHGRMW,[SKLPort23,SKLPort06,SKLPort0156,SKLPort237,SKLPort4], 8, [1,2,1,1,1], 6>;
 defm : X86WriteRes<WriteXCHG, [SKLPort0156], 2, [3], 3>;
 
 defm : SKLWriteResPair<WriteDiv8,   [SKLPort0, SKLDivider], 25, [1,10], 1, 4>;
 defm : SKLWriteResPair<WriteDiv16,  [SKLPort0, SKLDivider], 25, [1,10], 1, 4>;
 defm : SKLWriteResPair<WriteDiv32,  [SKLPort0, SKLDivider], 25, [1,10], 1, 4>;
 defm : SKLWriteResPair<WriteDiv64,  [SKLPort0, SKLDivider], 25, [1,10], 1, 4>;
 defm : SKLWriteResPair<WriteIDiv8,  [SKLPort0, SKLDivider], 25, [1,10], 1, 4>;
 defm : SKLWriteResPair<WriteIDiv16, [SKLPort0, SKLDivider], 25, [1,10], 1, 4>;

 
 def SKLWriteResGroup63 : SchedWriteRes<[SKLPort06,SKLPort0156]> {
   let Latency = 5;
   let NumMicroOps = 5;
   let ResourceCycles = [1,4];
 }
 def: InstRW<[SKLWriteResGroup63], (instrs XSETBV)>;
 
-def SKLWriteResGroup64 : SchedWriteRes<[SKLPort06,SKLPort0156]> {
-  let Latency = 5;
-  let NumMicroOps = 5;
-  let ResourceCycles = [2,3];
-}
-def: InstRW<[SKLWriteResGroup64], (instregex "CMPXCHG(8|16|32|64)rr")>;
-
 def SKLWriteResGroup65 : SchedWriteRes<[SKLPort4,SKLPort237,SKLPort0156]> {
   let Latency = 5;
   let NumMicroOps = 6;
   let ResourceCycles = [1,1,4];
 }
 def: InstRW<[SKLWriteResGroup65], (instregex "PUSHF(16|64)")>;
 
 def SKLWriteResGroup67 : SchedWriteRes<[SKLPort23]> {
def: InstRW<[SKLWriteResGroup117], (instregex "ROL(8|16|32|64)mCL",
                                               "SHR(8|16|32|64)mCL")>;
 
 def SKLWriteResGroup119 : SchedWriteRes<[SKLPort4,SKLPort23,SKLPort237,SKLPort06,SKLPort0156]> {
   let Latency = 8;
   let NumMicroOps = 6;
   let ResourceCycles = [1,1,1,2,1];
 }
 def: SchedAlias<WriteADCRMW, SKLWriteResGroup119>;
-def: InstRW<[SKLWriteResGroup119], (instregex "CMPXCHG(8|16|32|64)rm")>;
 
 def SKLWriteResGroup120 : SchedWriteRes<[SKLPort0,SKLPort23]> {
   let Latency = 9;
   let NumMicroOps = 2;
   let ResourceCycles = [1,1];
 }
 def: InstRW<[SKLWriteResGroup120], (instrs MMX_CVTPI2PSirm)>;
 

llvm/trunk/lib/Target/X86/X86SchedSkylakeServer.td

 // Arithmetic.
 defm : SKXWriteResPair<WriteALU,    [SKXPort0156], 1>; // Simple integer ALU op.
 defm : SKXWriteResPair<WriteADC,    [SKXPort06],   1>; // Integer ALU + flags op.
 defm : SKXWriteResPair<WriteIMul,   [SKXPort1],    3>; // Integer multiplication.
 defm : SKXWriteResPair<WriteIMul64, [SKXPort1],    3>; // Integer 64-bit multiplication.
 
 defm : X86WriteRes<WriteBSWAP32, [SKXPort15], 1, [1], 1>;
 defm : X86WriteRes<WriteBSWAP64, [SKXPort06, SKXPort15], 2, [1,1], 2>;
+defm : X86WriteRes<WriteCMPXCHG,[SKXPort06, SKXPort0156], 5, [2,3], 5>;
+defm : X86WriteRes<WriteCMPXCHGRMW,[SKXPort23,SKXPort06,SKXPort0156,SKXPort237,SKXPort4], 8, [1,2,1,1,1], 6>;
 defm : X86WriteRes<WriteXCHG,       [SKXPort0156], 2, [3], 3>;
 
 defm : SKXWriteResPair<WriteDiv8,   [SKXPort0, SKXDivider], 25, [1,10], 1, 4>;
 defm : SKXWriteResPair<WriteDiv16,  [SKXPort0, SKXDivider], 25, [1,10], 1, 4>;
 defm : SKXWriteResPair<WriteDiv32,  [SKXPort0, SKXDivider], 25, [1,10], 1, 4>;
 defm : SKXWriteResPair<WriteDiv64,  [SKXPort0, SKXDivider], 25, [1,10], 1, 4>;
 defm : SKXWriteResPair<WriteIDiv8,  [SKXPort0, SKXDivider], 25, [1,10], 1, 4>;
 defm : SKXWriteResPair<WriteIDiv16, [SKXPort0, SKXDivider], 25, [1,10], 1, 4>;

 
 def SKXWriteResGroup67 : SchedWriteRes<[SKXPort06,SKXPort0156]> {
   let Latency = 5;
   let NumMicroOps = 5;
   let ResourceCycles = [1,4];
 }
 def: InstRW<[SKXWriteResGroup67], (instrs XSETBV)>;
 
-def SKXWriteResGroup68 : SchedWriteRes<[SKXPort06,SKXPort0156]> {
-  let Latency = 5;
-  let NumMicroOps = 5;
-  let ResourceCycles = [2,3];
-}
-def: InstRW<[SKXWriteResGroup68], (instregex "CMPXCHG(8|16|32|64)rr")>;
-
 def SKXWriteResGroup69 : SchedWriteRes<[SKXPort4,SKXPort237,SKXPort0156]> {
   let Latency = 5;
   let NumMicroOps = 6;
   let ResourceCycles = [1,1,4];
 }
 def: InstRW<[SKXWriteResGroup69], (instregex "PUSHF(16|64)")>;
 
 def SKXWriteResGroup71 : SchedWriteRes<[SKXPort23]> {
def: InstRW<[SKXWriteResGroup128], (instregex "ROL(8|16|32|64)mCL",
                                               "SHR(8|16|32|64)mCL")>;
 
 def SKXWriteResGroup130 : SchedWriteRes<[SKXPort4,SKXPort23,SKXPort237,SKXPort06,SKXPort0156]> {
   let Latency = 8;
   let NumMicroOps = 6;
   let ResourceCycles = [1,1,1,2,1];
 }
 def: SchedAlias<WriteADCRMW, SKXWriteResGroup130>;
-def: InstRW<[SKXWriteResGroup130], (instregex "CMPXCHG(8|16|32|64)rm")>;
 
 def SKXWriteResGroup131 : SchedWriteRes<[SKXPort0,SKXPort4,SKXPort5,SKXPort237,SKXPort0156]> {
   let Latency = 8;
   let NumMicroOps = 8;
   let ResourceCycles = [1,2,1,2,2];
 }
 def: InstRW<[SKXWriteResGroup131], (instrs VPSCATTERQDZ128mr,
                                            VPSCATTERQDZ256mr,

llvm/trunk/lib/Target/X86/X86Schedule.td

 def  WriteADCRMW : WriteSequence<[WriteADCLd, WriteStore]>;
 defm WriteIMul   : X86SchedWritePair; // Integer multiplication.
 defm WriteIMul64 : X86SchedWritePair; // Integer 64-bit multiplication.
 def  WriteIMulH  : SchedWrite;        // Integer multiplication, high part.
 def  WriteLEA    : SchedWrite;        // LEA instructions can't fold loads.
 
 def  WriteBSWAP32 : SchedWrite; // Byte Order (Endianness) 32-bit Swap.
 def  WriteBSWAP64 : SchedWrite; // Byte Order (Endianness) 64-bit Swap.
-def  WriteXCHG   : SchedWrite;        // Compare+Exchange - TODO RMW support.
+defm WriteCMPXCHG : X86SchedWritePair; // Compare and set, compare and swap.
+def  WriteCMPXCHGRMW : SchedWrite;     // Compare and set, compare and swap.
+def  WriteXCHG    : SchedWrite;        // Compare+Exchange - TODO RMW support.
 
 // Integer division.
 defm WriteDiv8   : X86SchedWritePair;
 defm WriteDiv16  : X86SchedWritePair;
 defm WriteDiv32  : X86SchedWritePair;
 defm WriteDiv64  : X86SchedWritePair;
 defm WriteIDiv8  : X86SchedWritePair;
 defm WriteIDiv16 : X86SchedWritePair;

llvm/trunk/lib/Target/X86/X86ScheduleAtom.td

 defm : AtomWriteResPair<WriteALU,    [AtomPort01], [AtomPort0]>;
 defm : AtomWriteResPair<WriteADC,    [AtomPort01], [AtomPort0]>;
 defm : AtomWriteResPair<WriteIMul,   [AtomPort01], [AtomPort01],  7,  7,  [7],  [7]>;
 defm : AtomWriteResPair<WriteIMul64, [AtomPort01], [AtomPort01], 12, 12, [12], [12]>;
 
 defm : X86WriteRes<WriteXCHG,        [AtomPort01], 2, [2], 1>;
 defm : X86WriteRes<WriteBSWAP32,     [AtomPort0], 1, [1], 1>;
 defm : X86WriteRes<WriteBSWAP64,     [AtomPort0], 1, [1], 1>;
+defm : AtomWriteResPair<WriteCMPXCHG, [AtomPort01], [AtomPort01], 15, 15, [15]>;
+defm : X86WriteRes<WriteCMPXCHGRMW,   [AtomPort01, AtomPort0], 1, [1, 1], 1>;
 
 defm : AtomWriteResPair<WriteDiv8,   [AtomPort01], [AtomPort01], 50, 68, [50], [68]>;
 defm : AtomWriteResPair<WriteDiv16,  [AtomPort01], [AtomPort01], 50, 50, [50], [50]>;
 defm : AtomWriteResPair<WriteDiv32,  [AtomPort01], [AtomPort01], 50, 50, [50], [50]>;
 defm : AtomWriteResPair<WriteDiv64,  [AtomPort01], [AtomPort01],130,130,[130],[130]>;
 defm : AtomWriteResPair<WriteIDiv8,  [AtomPort01], [AtomPort01], 62, 62, [62], [62]>;
 defm : AtomWriteResPair<WriteIDiv16, [AtomPort01], [AtomPort01], 62, 62, [62], [62]>;
 defm : AtomWriteResPair<WriteIDiv32, [AtomPort01], [AtomPort01], 62, 62, [62], [62]>;

 def : InstRW<[AtomWrite01_13], (instrs AAA, AAS)>;
 
 def AtomWrite01_14 : SchedWriteRes<[AtomPort01]> {
   let Latency = 14;
   let ResourceCycles = [14];
 }
 def : InstRW<[AtomWrite01_14], (instrs CMPXCHG16rm, CMPXCHG32rm, CMPXCHG64rm)>;
 
-def AtomWrite01_15 : SchedWriteRes<[AtomPort01]> {
-  let Latency = 15;
-  let ResourceCycles = [15];
-}
-def : InstRW<[AtomWrite01_15], (instrs CMPXCHG16rr, CMPXCHG32rr, CMPXCHG64rr)>;
-
 def AtomWrite01_17 : SchedWriteRes<[AtomPort01]> {
   let Latency = 17;
   let ResourceCycles = [17];
 }
 def : InstRW<[AtomWrite01_17], (instrs LOOPNE, PAUSE)>;
 
 def AtomWrite01_18 : SchedWriteRes<[AtomPort01]> {
   let Latency = 18;

llvm/trunk/lib/Target/X86/X86ScheduleBtVer2.td

 defm : JWriteResIntPair<WriteALU,    [JALU01], 1>;
 defm : JWriteResIntPair<WriteADC,    [JALU01], 1, [2]>;
 defm : JWriteResIntPair<WriteIMul,   [JALU1, JMul], 3, [1, 1], 2>; // i8/i16/i32 multiplication
 defm : JWriteResIntPair<WriteIMul64, [JALU1, JMul], 6, [1, 4], 2>; // i64 multiplication
 defm : X86WriteRes<WriteIMulH,       [JALU1], 6, [4], 1>;
 
 defm : X86WriteRes<WriteBSWAP32, [JALU01], 1, [1], 1>;
 defm : X86WriteRes<WriteBSWAP64, [JALU01], 1, [1], 1>;
+defm : X86WriteRes<WriteCMPXCHG,[JALU01], 1, [1], 1>;
+defm : X86WriteRes<WriteCMPXCHGRMW,[JALU01, JSAGU, JLAGU], 4, [1, 1, 1], 2>;
 defm : X86WriteRes<WriteXCHG,        [JALU01], 1, [1], 1>;
 
 defm : JWriteResIntPair<WriteDiv8,   [JALU1, JDiv], 12, [1, 12], 1>;
 defm : JWriteResIntPair<WriteDiv16,  [JALU1, JDiv], 17, [1, 17], 2>;
 defm : JWriteResIntPair<WriteDiv32,  [JALU1, JDiv], 25, [1, 25], 2>;
 defm : JWriteResIntPair<WriteDiv64,  [JALU1, JDiv], 41, [1, 41], 2>;
 defm : JWriteResIntPair<WriteIDiv8,  [JALU1, JDiv], 12, [1, 12], 1>;
 defm : JWriteResIntPair<WriteIDiv16, [JALU1, JDiv], 17, [1, 17], 2>;

llvm/trunk/lib/Target/X86/X86ScheduleSLM.td

 
 defm : SLMWriteResPair<WriteALU,    [SLM_IEC_RSV01], 1>;
 defm : SLMWriteResPair<WriteADC,    [SLM_IEC_RSV01], 1>;
 defm : SLMWriteResPair<WriteIMul,   [SLM_IEC_RSV1],  3>;
 defm : SLMWriteResPair<WriteIMul64, [SLM_IEC_RSV1],  3>;
 
 defm : X86WriteRes<WriteBSWAP32, [SLM_IEC_RSV01], 1, [1], 1>;
 defm : X86WriteRes<WriteBSWAP64, [SLM_IEC_RSV01], 1, [1], 1>;
-defm : X86WriteRes<WriteXCHG,       [SLM_IEC_RSV01], 1, [1], 1>;
+defm : X86WriteRes<WriteCMPXCHG, [SLM_IEC_RSV01], 1, [1], 1>;
+defm : X86WriteRes<WriteCMPXCHGRMW, [SLM_IEC_RSV01, SLM_MEC_RSV], 4, [1, 2], 2>;
+defm : X86WriteRes<WriteXCHG,      [SLM_IEC_RSV01], 1, [1], 1>;
 
 defm : SLMWriteResPair<WriteShift,  [SLM_IEC_RSV0],  1>;
 
 defm : X86WriteRes<WriteSHDrri, [SLM_IEC_RSV0],  1, [1], 1>;
 defm : X86WriteRes<WriteSHDrrcl,[SLM_IEC_RSV0],  1, [1], 1>;
 defm : X86WriteRes<WriteSHDmri, [SLM_MEC_RSV, SLM_IEC_RSV0], 4, [2, 1], 2>;
 defm : X86WriteRes<WriteSHDmrcl,[SLM_MEC_RSV, SLM_IEC_RSV0], 4, [2, 1], 2>;
 

llvm/trunk/lib/Target/X86/X86ScheduleZnver1.td

 def : WriteRes<WriteLEA, [ZnALU]>;
 defm : ZnWriteResPair<WriteALU,   [ZnALU], 1>;
 defm : ZnWriteResPair<WriteADC,   [ZnALU], 1>;
 defm : ZnWriteResPair<WriteIMul,   [ZnALU1, ZnMultiplier], 4>;
 defm : ZnWriteResPair<WriteIMul64, [ZnALU1, ZnMultiplier], 4, [1,1], 2>;
 
 defm : X86WriteRes<WriteBSWAP32, [ZnALU], 1, [4], 1>;
 defm : X86WriteRes<WriteBSWAP64, [ZnALU], 1, [4], 1>;
+defm : X86WriteRes<WriteCMPXCHG, [ZnALU], 1, [1], 1>;
+defm : X86WriteRes<WriteCMPXCHGRMW,[ZnALU,ZnAGU], 8, [1,1], 5>;
 defm : X86WriteRes<WriteXCHG, [ZnALU], 1, [2], 2>;
 
 defm : ZnWriteResPair<WriteShift, [ZnALU], 1>;
 
 defm : X86WriteRes<WriteSHDrri, [ZnALU], 1, [1], 1>;
 defm : X86WriteResUnsupported<WriteSHDrrcl>;
 defm : X86WriteResUnsupported<WriteSHDmri>;
 defm : X86WriteResUnsupported<WriteSHDmrcl>;

 
 // r,r,cl.
 def : InstRW<[WriteMicrocoded], (instregex "SH(R|L)D(16|32|64)rrCL")>;
 
 // m,r,cl.
 def : InstRW<[WriteMicrocoded], (instregex "SH(R|L)D(16|32|64)mrCL")>;
 
 //-- Misc instructions --//
-// CMPXCHG.
-def ZnWriteCMPXCHG : SchedWriteRes<[ZnAGU, ZnALU]> {
-  let Latency = 8;
-  let NumMicroOps = 5;
-}
-def : InstRW<[ZnWriteCMPXCHG], (instregex "CMPXCHG(8|16|32|64)rm")>;
-
 // CMPXCHG8B.
 def ZnWriteCMPXCHG8B : SchedWriteRes<[ZnAGU, ZnALU]> {
   let NumMicroOps = 18;
 }
 def : InstRW<[ZnWriteCMPXCHG8B], (instrs CMPXCHG8B)>;
 
 def : InstRW<[WriteMicrocoded], (instrs CMPXCHG16B)>;
 

llvm/trunk/test/CodeGen/X86/schedule-x86_64.ll

; ZNVER1-NEXT:    retq # sched: [1:0.50]
   ret void
 }
 
 define void @test_cmpxchg_8(i8 %a0, i8 %a1, i8 *%a2) optsize {
 ; GENERIC-LABEL: test_cmpxchg_8:
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    #APP
 ; GENERIC-NEXT:    cmpxchgb %dil, %sil # sched: [5:1.33]
-; GENERIC-NEXT:    cmpxchgb %dil, (%rdx) # sched: [8:1.00]
+; GENERIC-NEXT:    cmpxchgb %dil, (%rdx) # sched: [8:2.00]
 ; GENERIC-NEXT:    #NO_APP
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_cmpxchg_8:
 ; ATOM:       # %bb.0:
 ; ATOM-NEXT:    #APP
 ; ATOM-NEXT:    cmpxchgb %dil, %sil # sched: [9:4.50]
 ; ATOM-NEXT:    cmpxchgb %dil, (%rdx) # sched: [6:3.00]
 ; ATOM-NEXT:    #NO_APP
 ; ATOM-NEXT:    retq # sched: [79:39.50]
 ;
 ; SLM-LABEL: test_cmpxchg_8:
 ; SLM:       # %bb.0:
 ; SLM-NEXT:    #APP
 ; SLM-NEXT:    cmpxchgb %dil, %sil # sched: [1:0.50]
 ; SLM-NEXT:    cmpxchgb %dil, (%rdx) # sched: [4:2.00]
 ; SLM-NEXT:    #NO_APP
 ; SLM-NEXT:    retq # sched: [4:1.00]
 ;
 ; SANDY-LABEL: test_cmpxchg_8:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    #APP
 ; SANDY-NEXT:    cmpxchgb %dil, %sil # sched: [5:1.33]
-; SANDY-NEXT:    cmpxchgb %dil, (%rdx) # sched: [8:1.00]
+; SANDY-NEXT:    cmpxchgb %dil, (%rdx) # sched: [8:2.00]
 ; SANDY-NEXT:    #NO_APP
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_cmpxchg_8:
 ; HASWELL:       # %bb.0:
 ; HASWELL-NEXT:    #APP
 ; HASWELL-NEXT:    cmpxchgb %dil, %sil # sched: [5:1.25]
 ; HASWELL-NEXT:    cmpxchgb %dil, (%rdx) # sched: [9:1.00]
; ZNVER1-NEXT:    retq # sched: [1:0.50]
   tail call void asm "cmpxchgb $0, $1 \0a\09 cmpxchgb $0, $2", "r,r,*m"(i8 %a0, i8 %a1, i8 *%a2) nounwind
   ret void
 }
 define void @test_cmpxchg_16(i16 %a0, i16 %a1, i16 *%a2) optsize {
 ; GENERIC-LABEL: test_cmpxchg_16:
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    #APP
 ; GENERIC-NEXT:    cmpxchgw %di, %si # sched: [5:1.33]
-; GENERIC-NEXT:    cmpxchgw %di, (%rdx) # sched: [8:1.00]
+; GENERIC-NEXT:    cmpxchgw %di, (%rdx) # sched: [8:2.00]
 ; GENERIC-NEXT:    #NO_APP
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_cmpxchg_16:
 ; ATOM:       # %bb.0:
 ; ATOM-NEXT:    #APP
 ; ATOM-NEXT:    cmpxchgw %di, %si # sched: [15:7.50]
 ; ATOM-NEXT:    cmpxchgw %di, (%rdx) # sched: [14:7.00]
 ; ATOM-NEXT:    #NO_APP
 ; ATOM-NEXT:    retq # sched: [79:39.50]
 ;
 ; SLM-LABEL: test_cmpxchg_16:
 ; SLM:       # %bb.0:
 ; SLM-NEXT:    #APP
 ; SLM-NEXT:    cmpxchgw %di, %si # sched: [1:0.50]
 ; SLM-NEXT:    cmpxchgw %di, (%rdx) # sched: [4:2.00]
 ; SLM-NEXT:    #NO_APP
 ; SLM-NEXT:    retq # sched: [4:1.00]
 ;
 ; SANDY-LABEL: test_cmpxchg_16:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    #APP
 ; SANDY-NEXT:    cmpxchgw %di, %si # sched: [5:1.33]
-; SANDY-NEXT:    cmpxchgw %di, (%rdx) # sched: [8:1.00]
+; SANDY-NEXT:    cmpxchgw %di, (%rdx) # sched: [8:2.00]
 ; SANDY-NEXT:    #NO_APP
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_cmpxchg_16:
 ; HASWELL:       # %bb.0:
 ; HASWELL-NEXT:    #APP
 ; HASWELL-NEXT:    cmpxchgw %di, %si # sched: [5:1.25]
 ; HASWELL-NEXT:    cmpxchgw %di, (%rdx) # sched: [9:1.00]
; ZNVER1-NEXT:    retq # sched: [1:0.50]
   tail call void asm "cmpxchgw $0, $1 \0a\09 cmpxchgw $0, $2", "r,r,*m"(i16 %a0, i16 %a1, i16 *%a2) nounwind
   ret void
 }
 define void @test_cmpxchg_32(i32 %a0, i32 %a1, i32 *%a2) optsize {
 ; GENERIC-LABEL: test_cmpxchg_32:
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    #APP
 ; GENERIC-NEXT:    cmpxchgl %edi, %esi # sched: [5:1.33]
-; GENERIC-NEXT:    cmpxchgl %edi, (%rdx) # sched: [8:1.00]
+; GENERIC-NEXT:    cmpxchgl %edi, (%rdx) # sched: [8:2.00]
 ; GENERIC-NEXT:    #NO_APP
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_cmpxchg_32:
 ; ATOM:       # %bb.0:
 ; ATOM-NEXT:    #APP
 ; ATOM-NEXT:    cmpxchgl %edi, %esi # sched: [15:7.50]
 ; ATOM-NEXT:    cmpxchgl %edi, (%rdx) # sched: [14:7.00]
 ; ATOM-NEXT:    #NO_APP
 ; ATOM-NEXT:    retq # sched: [79:39.50]
 ;
 ; SLM-LABEL: test_cmpxchg_32:
 ; SLM:       # %bb.0:
 ; SLM-NEXT:    #APP
 ; SLM-NEXT:    cmpxchgl %edi, %esi # sched: [1:0.50]
 ; SLM-NEXT:    cmpxchgl %edi, (%rdx) # sched: [4:2.00]
 ; SLM-NEXT:    #NO_APP
 ; SLM-NEXT:    retq # sched: [4:1.00]
 ;
 ; SANDY-LABEL: test_cmpxchg_32:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    #APP
 ; SANDY-NEXT:    cmpxchgl %edi, %esi # sched: [5:1.33]
-; SANDY-NEXT:    cmpxchgl %edi, (%rdx) # sched: [8:1.00]
+; SANDY-NEXT:    cmpxchgl %edi, (%rdx) # sched: [8:2.00]
 ; SANDY-NEXT:    #NO_APP
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_cmpxchg_32:
 ; HASWELL:       # %bb.0:
 ; HASWELL-NEXT:    #APP
 ; HASWELL-NEXT:    cmpxchgl %edi, %esi # sched: [5:1.25]
 ; HASWELL-NEXT:    cmpxchgl %edi, (%rdx) # sched: [9:1.00]
; ZNVER1-NEXT:    retq # sched: [1:0.50]
   tail call void asm "cmpxchgl $0, $1 \0a\09 cmpxchgl $0, $2", "r,r,*m"(i32 %a0, i32 %a1, i32 *%a2) nounwind
   ret void
 }
 define void @test_cmpxchg_64(i64 %a0, i64 %a1, i64 *%a2) optsize {
 ; GENERIC-LABEL: test_cmpxchg_64:
 ; GENERIC:       # %bb.0:
 ; GENERIC-NEXT:    #APP
 ; GENERIC-NEXT:    cmpxchgq %rdi, %rsi # sched: [5:1.33]
-; GENERIC-NEXT:    cmpxchgq %rdi, (%rdx) # sched: [8:1.00]
+; GENERIC-NEXT:    cmpxchgq %rdi, (%rdx) # sched: [8:2.00]
 ; GENERIC-NEXT:    #NO_APP
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_cmpxchg_64:
 ; ATOM:       # %bb.0:
 ; ATOM-NEXT:    #APP
 ; ATOM-NEXT:    cmpxchgq %rdi, %rsi # sched: [15:7.50]
 ; ATOM-NEXT:    cmpxchgq %rdi, (%rdx) # sched: [14:7.00]
 ; ATOM-NEXT:    #NO_APP
 ; ATOM-NEXT:    retq # sched: [79:39.50]
 ;
 ; SLM-LABEL: test_cmpxchg_64:
 ; SLM:       # %bb.0:
 ; SLM-NEXT:    #APP
 ; SLM-NEXT:    cmpxchgq %rdi, %rsi # sched: [1:0.50]
 ; SLM-NEXT:    cmpxchgq %rdi, (%rdx) # sched: [4:2.00]
 ; SLM-NEXT:    #NO_APP
 ; SLM-NEXT:    retq # sched: [4:1.00]
 ;
 ; SANDY-LABEL: test_cmpxchg_64:
 ; SANDY:       # %bb.0:
 ; SANDY-NEXT:    #APP
 ; SANDY-NEXT:    cmpxchgq %rdi, %rsi # sched: [5:1.33]
-; SANDY-NEXT:    cmpxchgq %rdi, (%rdx) # sched: [8:1.00]
+; SANDY-NEXT:    cmpxchgq %rdi, (%rdx) # sched: [8:2.00]
 ; SANDY-NEXT:    #NO_APP
 ; SANDY-NEXT:    retq # sched: [1:1.00]
 ;
 ; HASWELL-LABEL: test_cmpxchg_64:
 ; HASWELL:       # %bb.0:
 ; HASWELL-NEXT:    #APP
 ; HASWELL-NEXT:    cmpxchgq %rdi, %rsi # sched: [5:1.25]
 ; HASWELL-NEXT:    cmpxchgq %rdi, (%rdx) # sched: [9:1.00]

llvm/trunk/test/tools/llvm-mca/X86/Generic/resources-x86_64.s

 # CHECK-NEXT:  1      1     0.33                        cmpq  %rsi, %rdi
 # CHECK-NEXT:  2      6     0.50    *                   cmpq  %rsi, (%rax)
 # CHECK-NEXT:  2      6     0.50    *                   cmpq  (%rax), %rdi
 # CHECK-NEXT:  5      8     1.00                  U     cmpsb %es:(%rdi), (%rsi)
 # CHECK-NEXT:  5      8     1.00                  U     cmpsw %es:(%rdi), (%rsi)
 # CHECK-NEXT:  5      8     1.00                  U     cmpsl %es:(%rdi), (%rsi)
 # CHECK-NEXT:  5      8     1.00                  U     cmpsq %es:(%rdi), (%rsi)
 # CHECK-NEXT:  4      5     1.33                        cmpxchgb  %cl, %bl
-# CHECK-NEXT:  4      8     1.00    *      *            cmpxchgb  %cl, (%rbx)
+# CHECK-NEXT:  6      8     2.00    *      *            cmpxchgb  %cl, (%rbx)
 # CHECK-NEXT:  4      5     1.33                        cmpxchgw  %cx, %bx
-# CHECK-NEXT:  4      8     1.00    *      *            cmpxchgw  %cx, (%rbx)
+# CHECK-NEXT:  6      8     2.00    *      *            cmpxchgw  %cx, (%rbx)
 # CHECK-NEXT:  4      5     1.33                        cmpxchgl  %ecx, %ebx
-# CHECK-NEXT:  4      8     1.00    *      *            cmpxchgl  %ecx, (%rbx)
+# CHECK-NEXT:  6      8     2.00    *      *            cmpxchgl  %ecx, (%rbx)
 # CHECK-NEXT:  4      5     1.33                        cmpxchgq  %rcx, %rbx
-# CHECK-NEXT:  4      8     1.00    *      *            cmpxchgq  %rcx, (%rbx)
+# CHECK-NEXT:  6      8     2.00    *      *            cmpxchgq  %rcx, (%rbx)
 # CHECK-NEXT:  1      100   0.33                  U     cpuid
 # CHECK-NEXT:  1      1     0.33                        decb  %dil
 # CHECK-NEXT:  3      7     1.00    *      *            decb  (%rax)
 # CHECK-NEXT:  1      1     0.33                        decw  %di
 # CHECK-NEXT:  3      7     1.00    *      *            decw  (%rax)
 # CHECK-NEXT:  1      1     0.33                        decl  %edi
 # CHECK-NEXT:  3      7     1.00    *      *            decl  (%rax)
 # CHECK-NEXT:  1      1     0.33                        decq  %rdi

 # CHECK-NEXT: [3]   - SBPort1
 # CHECK-NEXT: [4]   - SBPort4
 # CHECK-NEXT: [5]   - SBPort5
 # CHECK-NEXT: [6.0] - SBPort23
 # CHECK-NEXT: [6.1] - SBPort23
 
 # CHECK:      Resource pressure per iteration:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6.0]  [6.1]
-# CHECK-NEXT: 160.00  -     438.50 224.00 242.00 430.50 290.00 290.00
+# CHECK-NEXT: 160.00  -     435.83 221.33 246.00 435.83 292.00 292.00
 
 # CHECK:      Resource pressure by instruction:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6.0]  [6.1]  Instructions:
 # CHECK-NEXT:  -      -     0.83   0.33    -     0.83    -      -     adcb  $7, %al
 # CHECK-NEXT:  -      -     0.83   0.33    -     0.83    -      -     adcb  $7, %dil
 # CHECK-NEXT:  -      -     1.00   1.00   1.00   1.00   1.00   1.00   adcb  $7, (%rax)
 # CHECK-NEXT:  -      -     0.83   0.33    -     0.83    -      -     adcb  %sil, %dil
 # CHECK-NEXT:  -      -     1.33   0.33   1.00   1.33   1.00   1.00   adcb  %sil, (%rax)

 # CHECK-NEXT:  -      -     0.33   0.33    -     0.33    -      -     cmpq  %rsi, %rdi
 # CHECK-NEXT:  -      -     0.33   0.33    -     0.33   0.50   0.50   cmpq  %rsi, (%rax)
 # CHECK-NEXT:  -      -     0.33   0.33    -     0.33   0.50   0.50   cmpq  (%rax), %rdi
 # CHECK-NEXT:  -      -     1.00   1.00    -     1.00   1.00   1.00   cmpsb %es:(%rdi), (%rsi)
 # CHECK-NEXT:  -      -     1.00   1.00    -     1.00   1.00   1.00   cmpsw %es:(%rdi), (%rsi)
 # CHECK-NEXT:  -      -     1.00   1.00    -     1.00   1.00   1.00   cmpsl %es:(%rdi), (%rsi)
 # CHECK-NEXT:  -      -     1.00   1.00    -     1.00   1.00   1.00   cmpsq %es:(%rdi), (%rsi)
 # CHECK-NEXT:  -      -     1.50   1.00    -     1.50    -      -     cmpxchgb  %cl, %bl
-# CHECK-NEXT:  -      -     1.00   1.00    -     1.00   0.50   0.50   cmpxchgb  %cl, (%rbx)
+# CHECK-NEXT:  -      -     0.33   0.33   1.00   2.33   1.00   1.00   cmpxchgb  %cl, (%rbx)
 # CHECK-NEXT:  -      -     1.50   1.00    -     1.50    -      -     cmpxchgw  %cx, %bx
-# CHECK-NEXT:  -      -     1.00   1.00    -     1.00   0.50   0.50   cmpxchgw  %cx, (%rbx)
+# CHECK-NEXT:  -      -     0.33   0.33   1.00   2.33   1.00   1.00   cmpxchgw  %cx, (%rbx)
 # CHECK-NEXT:  -      -     1.50   1.00    -     1.50    -      -     cmpxchgl  %ecx, %ebx
-# CHECK-NEXT:  -      -     1.00   1.00    -     1.00   0.50   0.50   cmpxchgl  %ecx, (%rbx)
+# CHECK-NEXT:  -      -     0.33   0.33   1.00   2.33   1.00   1.00   cmpxchgl  %ecx, (%rbx)
 # CHECK-NEXT:  -      -     1.50   1.00    -     1.50    -      -     cmpxchgq  %rcx, %rbx
-# CHECK-NEXT:  -      -     1.00   1.00    -     1.00   0.50   0.50   cmpxchgq  %rcx, (%rbx)
+# CHECK-NEXT:  -      -     0.33   0.33   1.00   2.33   1.00   1.00   cmpxchgq  %rcx, (%rbx)
 # CHECK-NEXT:  -      -     0.33   0.33    -     0.33    -      -     cpuid
 # CHECK-NEXT:  -      -     0.33   0.33    -     0.33    -      -     decb  %dil
 # CHECK-NEXT:  -      -     0.33   0.33   1.00   0.33   1.00   1.00   decb  (%rax)
 # CHECK-NEXT:  -      -     0.33   0.33    -     0.33    -      -     decw  %di
 # CHECK-NEXT:  -      -     0.33   0.33   1.00   0.33   1.00   1.00   decw  (%rax)
 # CHECK-NEXT:  -      -     0.33   0.33    -     0.33    -      -     decl  %edi
 # CHECK-NEXT:  -      -     0.33   0.33   1.00   0.33   1.00   1.00   decl  (%rax)
 # CHECK-NEXT:  -      -     0.33   0.33    -     0.33    -      -     decq  %rdi

llvm/trunk/test/tools/llvm-mca/X86/SandyBridge/resources-x86_64.s

 xorq $665536, %rdi
 xorq $665536, (%rax)
 xorq $7, %rdi
 xorq $7, (%rax)
 xorq %rsi, %rdi
 xorq %rsi, (%rax)
 xorq (%rax), %rdi
 
+# CCHECK-NEXT: 160.00  -     439.83 221.33 250.00 431.83 294.00 294.00
+
 # CHECK:      Instruction Info:
 # CHECK-NEXT: [1]: #uOps
 # CHECK-NEXT: [2]: Latency
 # CHECK-NEXT: [3]: RThroughput
 # CHECK-NEXT: [4]: MayLoad
 # CHECK-NEXT: [5]: MayStore
 # CHECK-NEXT: [6]: HasSideEffects (U)
 

 # CHECK-NEXT:  1      1     0.33                        cmpq  %rsi, %rdi
 # CHECK-NEXT:  2      6     0.50    *                   cmpq  %rsi, (%rax)
 # CHECK-NEXT:  2      6     0.50    *                   cmpq  (%rax), %rdi
 # CHECK-NEXT:  5      8     1.00                  U     cmpsb %es:(%rdi), (%rsi)
 # CHECK-NEXT:  5      8     1.00                  U     cmpsw %es:(%rdi), (%rsi)
 # CHECK-NEXT:  5      8     1.00                  U     cmpsl %es:(%rdi), (%rsi)
 # CHECK-NEXT:  5      8     1.00                  U     cmpsq %es:(%rdi), (%rsi)
 # CHECK-NEXT:  4      5     1.33                        cmpxchgb  %cl, %bl
-# CHECK-NEXT:  4      8     1.00    *      *            cmpxchgb  %cl, (%rbx)
+# CHECK-NEXT:  6      8     2.00    *      *            cmpxchgb  %cl, (%rbx)
 # CHECK-NEXT:  4      5     1.33                        cmpxchgw  %cx, %bx
-# CHECK-NEXT:  4      8     1.00    *      *            cmpxchgw  %cx, (%rbx)
+# CHECK-NEXT:  6      8     2.00    *      *            cmpxchgw  %cx, (%rbx)
 # CHECK-NEXT:  4      5     1.33                        cmpxchgl  %ecx, %ebx
-# CHECK-NEXT:  4      8     1.00    *      *            cmpxchgl  %ecx, (%rbx)
+# CHECK-NEXT:  6      8     2.00    *      *            cmpxchgl  %ecx, (%rbx)
 # CHECK-NEXT:  4      5     1.33                        cmpxchgq  %rcx, %rbx
-# CHECK-NEXT:  4      8     1.00    *      *            cmpxchgq  %rcx, (%rbx)
+# CHECK-NEXT:  6      8     2.00    *      *            cmpxchgq  %rcx, (%rbx)
 # CHECK-NEXT:  1      100   0.33                  U     cpuid
 # CHECK-NEXT:  1      1     0.33                        decb  %dil
 # CHECK-NEXT:  3      7     1.00    *      *            decb  (%rax)
 # CHECK-NEXT:  1      1     0.33                        decw  %di
 # CHECK-NEXT:  3      7     1.00    *      *            decw  (%rax)
 # CHECK-NEXT:  1      1     0.33                        decl  %edi
 # CHECK-NEXT:  3      7     1.00    *      *            decl  (%rax)
 # CHECK-NEXT:  1      1     0.33                        decq  %rdi

 # CHECK-NEXT: [3]   - SBPort1
 # CHECK-NEXT: [4]   - SBPort4
 # CHECK-NEXT: [5]   - SBPort5
 # CHECK-NEXT: [6.0] - SBPort23
 # CHECK-NEXT: [6.1] - SBPort23
 
 # CHECK:      Resource pressure per iteration:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6.0]  [6.1]
-# CHECK-NEXT: 160.00  -     438.50 224.00 242.00 430.50 290.00 290.00
+# CHECK-NEXT: 160.00  -     435.83 221.33 246.00 435.83 292.00 292.00
 
 # CHECK:      Resource pressure by instruction:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6.0]  [6.1]  Instructions:
 # CHECK-NEXT:  -      -     0.83   0.33    -     0.83    -      -     adcb  $7, %al
 # CHECK-NEXT:  -      -     0.83   0.33    -     0.83    -      -     adcb  $7, %dil
 # CHECK-NEXT:  -      -     1.00   1.00   1.00   1.00   1.00   1.00   adcb  $7, (%rax)
 # CHECK-NEXT:  -      -     0.83   0.33    -     0.83    -      -     adcb  %sil, %dil
 # CHECK-NEXT:  -      -     1.33   0.33   1.00   1.33   1.00   1.00   adcb  %sil, (%rax)

 # CHECK-NEXT:  -      -     0.33   0.33    -     0.33    -      -     cmpq  %rsi, %rdi
 # CHECK-NEXT:  -      -     0.33   0.33    -     0.33   0.50   0.50   cmpq  %rsi, (%rax)
 # CHECK-NEXT:  -      -     0.33   0.33    -     0.33   0.50   0.50   cmpq  (%rax), %rdi
 # CHECK-NEXT:  -      -     1.00   1.00    -     1.00   1.00   1.00   cmpsb %es:(%rdi), (%rsi)
 # CHECK-NEXT:  -      -     1.00   1.00    -     1.00   1.00   1.00   cmpsw %es:(%rdi), (%rsi)
 # CHECK-NEXT:  -      -     1.00   1.00    -     1.00   1.00   1.00   cmpsl %es:(%rdi), (%rsi)
 # CHECK-NEXT:  -      -     1.00   1.00    -     1.00   1.00   1.00   cmpsq %es:(%rdi), (%rsi)
 # CHECK-NEXT:  -      -     1.50   1.00    -     1.50    -      -     cmpxchgb  %cl, %bl
-# CHECK-NEXT:  -      -     1.00   1.00    -     1.00   0.50   0.50   cmpxchgb  %cl, (%rbx)
+# CHECK-NEXT:  -      -     0.33   0.33   1.00   2.33   1.00   1.00   cmpxchgb  %cl, (%rbx)
 # CHECK-NEXT:  -      -     1.50   1.00    -     1.50    -      -     cmpxchgw  %cx, %bx
-# CHECK-NEXT:  -      -     1.00   1.00    -     1.00   0.50   0.50   cmpxchgw  %cx, (%rbx)
+# CHECK-NEXT:  -      -     0.33   0.33   1.00   2.33   1.00   1.00   cmpxchgw  %cx, (%rbx)
 # CHECK-NEXT:  -      -     1.50   1.00    -     1.50    -      -     cmpxchgl  %ecx, %ebx
-# CHECK-NEXT:  -      -     1.00   1.00    -     1.00   0.50   0.50   cmpxchgl  %ecx, (%rbx)
+# CHECK-NEXT:  -      -     0.33   0.33   1.00   2.33   1.00   1.00   cmpxchgl  %ecx, (%rbx)
 # CHECK-NEXT:  -      -     1.50   1.00    -     1.50    -      -     cmpxchgq  %rcx, %rbx
-# CHECK-NEXT:  -      -     1.00   1.00    -     1.00   0.50   0.50   cmpxchgq  %rcx, (%rbx)
+# CHECK-NEXT:  -      -     0.33   0.33   1.00   2.33   1.00   1.00   cmpxchgq  %rcx, (%rbx)
 # CHECK-NEXT:  -      -     0.33   0.33    -     0.33    -      -     cpuid
 # CHECK-NEXT:  -      -     0.33   0.33    -     0.33    -      -     decb  %dil
 # CHECK-NEXT:  -      -     0.33   0.33   1.00   0.33   1.00   1.00   decb  (%rax)
 # CHECK-NEXT:  -      -     0.33   0.33    -     0.33    -      -     decw  %di
 # CHECK-NEXT:  -      -     0.33   0.33   1.00   0.33   1.00   1.00   decw  (%rax)
 # CHECK-NEXT:  -      -     0.33   0.33    -     0.33    -      -     decl  %edi
 # CHECK-NEXT:  -      -     0.33   0.33   1.00   0.33   1.00   1.00   decl  (%rax)
 # CHECK-NEXT:  -      -     0.33   0.33    -     0.33    -      -     decq  %rdi