From 974a7014d8b00607f09f3f42f869e7ceea46bf61 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Sat, 12 May 2018 18:07:07 +0000
Subject: [PATCH] [X86] Add WriteFCMOV scheduler class for x87 CMOVs

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@332173 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/X86/X86InstrFPStack.td       | 2 +-
 lib/Target/X86/X86SchedBroadwell.td     | 2 ++
 lib/Target/X86/X86SchedHaswell.td       | 1 +
 lib/Target/X86/X86SchedSandyBridge.td   | 8 +-------
 lib/Target/X86/X86SchedSkylakeClient.td | 4 ++--
 lib/Target/X86/X86SchedSkylakeServer.td | 4 ++--
 lib/Target/X86/X86Schedule.td           | 1 +
 lib/Target/X86/X86ScheduleAtom.td       | 4 ++--
 lib/Target/X86/X86ScheduleBtVer2.td     | 1 +
 lib/Target/X86/X86ScheduleSLM.td        | 1 +
 lib/Target/X86/X86ScheduleZnver1.td     | 3 +--
 11 files changed, 15 insertions(+), 16 deletions(-)

diff --git a/lib/Target/X86/X86InstrFPStack.td b/lib/Target/X86/X86InstrFPStack.td
index aedd445b714..0cc6272d89a 100644
--- a/lib/Target/X86/X86InstrFPStack.td
+++ b/lib/Target/X86/X86InstrFPStack.td
@@ -384,7 +384,7 @@ multiclass FPCMov<PatLeaf cc> {
 }
 
 let Defs = [FPSW] in {
-let SchedRW = [WriteFAdd] in {
+let SchedRW = [WriteFCMOV] in {
 let Uses = [EFLAGS], Constraints = "$src1 = $dst" in {
 defm CMOVB  : FPCMov<X86_COND_B>;
 defm CMOVBE : FPCMov<X86_COND_BE>;
diff --git a/lib/Target/X86/X86SchedBroadwell.td b/lib/Target/X86/X86SchedBroadwell.td
index 268f60b8237..7efc3a28481 100755
--- a/lib/Target/X86/X86SchedBroadwell.td
+++ b/lib/Target/X86/X86SchedBroadwell.td
@@ -125,6 +125,8 @@ def : WriteRes<WriteIMulH, []> { let Latency = 3; } // Integer multiplication, h
 def : WriteRes<WriteLEA, [BWPort15]>; // LEA instructions can't fold loads.
 
 defm : BWWriteResPair<WriteCMOV,  [BWPort06], 1>; // Conditional move.
+defm : X86WriteRes<WriteFCMOV, [BWPort1], 3, [1], 1>; // x87 conditional move.
+
 def  : WriteRes<WriteSETCC, [BWPort06]>; // Setcc.
 def  : WriteRes<WriteSETCCStore, [BWPort06,BWPort4,BWPort237]> {
   let Latency = 2;
diff --git a/lib/Target/X86/X86SchedHaswell.td b/lib/Target/X86/X86SchedHaswell.td
index 7b84757baa3..0a64f2a37fe 100644
--- a/lib/Target/X86/X86SchedHaswell.td
+++ b/lib/Target/X86/X86SchedHaswell.td
@@ -121,6 +121,7 @@ defm : HWWriteResPair<WriteJump,   [HWPort06],  1>;
 defm : HWWriteResPair<WriteCRC32,  [HWPort1],   3>;
 
 defm : HWWriteResPair<WriteCMOV,  [HWPort06,HWPort0156], 2, [1,1], 2>; // Conditional move.
+defm : X86WriteRes<WriteFCMOV, [HWPort1], 3, [1], 1>; // x87 conditional move.
 def  : WriteRes<WriteSETCC, [HWPort06]>; // Setcc.
 def  : WriteRes<WriteSETCCStore, [HWPort06,HWPort4,HWPort237]> {
   let Latency = 2;
diff --git a/lib/Target/X86/X86SchedSandyBridge.td b/lib/Target/X86/X86SchedSandyBridge.td
index 74a5824a036..be380ffa2ff 100644
--- a/lib/Target/X86/X86SchedSandyBridge.td
+++ b/lib/Target/X86/X86SchedSandyBridge.td
@@ -121,6 +121,7 @@ defm : SBWriteResPair<WriteJump,  [SBPort5],   1>;
 defm : SBWriteResPair<WriteCRC32, [SBPort1],   3, [1], 1, 5>;
 
 defm : SBWriteResPair<WriteCMOV,  [SBPort05,SBPort015], 2, [1,1], 2>; // Conditional move.
+defm : X86WriteRes<WriteFCMOV, [SBPort5,SBPort05], 3, [2,1], 3>; // x87 conditional move.
 def  : WriteRes<WriteSETCC, [SBPort05]>; // Setcc.
 def  : WriteRes<WriteSETCCStore, [SBPort05,SBPort4,SBPort23]> {
   let Latency = 2;
@@ -640,13 +641,6 @@ def SBWriteResGroup25_1 : SchedWriteRes<[SBPort23,SBPort015]> {
 }
 def: InstRW<[SBWriteResGroup25_1], (instrs LEAVE, LEAVE64)>;
 
-def SBWriteResGroup25_2 : SchedWriteRes<[SBPort5,SBPort05]> {
-  let Latency = 3;
-  let NumMicroOps = 3;
-  let ResourceCycles = [2,1];
-}
-def: InstRW<[SBWriteResGroup25_2], (instregex "CMOV(N?)(B|BE|E|P)_F")>;
-
 def SBWriteResGroup26 : SchedWriteRes<[SBPort05,SBPort015]> {
   let Latency = 3;
   let NumMicroOps = 3;
diff --git a/lib/Target/X86/X86SchedSkylakeClient.td b/lib/Target/X86/X86SchedSkylakeClient.td
index e8da8b807aa..e5019938dbd 100644
--- a/lib/Target/X86/X86SchedSkylakeClient.td
+++ b/lib/Target/X86/X86SchedSkylakeClient.td
@@ -124,6 +124,7 @@ def : WriteRes<WriteIMulH, []> { let Latency = 3; } // Integer multiplication, h
 def : WriteRes<WriteLEA, [SKLPort15]>; // LEA instructions can't fold loads.
 
 defm : SKLWriteResPair<WriteCMOV,  [SKLPort06], 1>; // Conditional move.
+defm : X86WriteRes<WriteFCMOV, [SKLPort1], 3, [1], 1>; // x87 conditional move.
 def  : WriteRes<WriteSETCC, [SKLPort06]>; // Setcc.
 def  : WriteRes<WriteSETCCStore, [SKLPort06,SKLPort4,SKLPort237]> {
   let Latency = 2;
@@ -704,8 +705,7 @@ def SKLWriteResGroup29 : SchedWriteRes<[SKLPort1]> {
   let NumMicroOps = 1;
   let ResourceCycles = [1];
 }
-def: InstRW<[SKLWriteResGroup29], (instregex "CMOV(N?)(B|BE|E|P)_F",
-                                             "PDEP(32|64)rr",
+def: InstRW<[SKLWriteResGroup29], (instregex "PDEP(32|64)rr",
                                              "PEXT(32|64)rr",
                                              "SHLD(16|32|64)rri8",
                                              "SHRD(16|32|64)rri8")>;
diff --git a/lib/Target/X86/X86SchedSkylakeServer.td b/lib/Target/X86/X86SchedSkylakeServer.td
index ce3c5104900..4bd95ea494d 100755
--- a/lib/Target/X86/X86SchedSkylakeServer.td
+++ b/lib/Target/X86/X86SchedSkylakeServer.td
@@ -124,6 +124,7 @@ def : WriteRes<WriteIMulH, []> { let Latency = 3; } // Integer multiplication, h
 def : WriteRes<WriteLEA, [SKXPort15]>; // LEA instructions can't fold loads.
 
 defm : SKXWriteResPair<WriteCMOV,  [SKXPort06], 1>; // Conditional move.
+defm : X86WriteRes<WriteFCMOV, [SKXPort1], 3, [1], 1>; // x87 conditional move.
 def  : WriteRes<WriteSETCC, [SKXPort06]>; // Setcc.
 def  : WriteRes<WriteSETCCStore, [SKXPort06,SKXPort4,SKXPort237]> {
   let Latency = 2;
@@ -754,8 +755,7 @@ def SKXWriteResGroup31 : SchedWriteRes<[SKXPort1]> {
   let NumMicroOps = 1;
   let ResourceCycles = [1];
 }
-def: InstRW<[SKXWriteResGroup31], (instregex "CMOV(N?)(B|BE|E|P)_F",
-                                             "PDEP(32|64)rr",
+def: InstRW<[SKXWriteResGroup31], (instregex "PDEP(32|64)rr",
                                              "PEXT(32|64)rr",
                                              "SHLD(16|32|64)rri8",
                                              "SHRD(16|32|64)rri8")>;
diff --git a/lib/Target/X86/X86Schedule.td b/lib/Target/X86/X86Schedule.td
index f23b37ddce7..a0a9a911788 100644
--- a/lib/Target/X86/X86Schedule.td
+++ b/lib/Target/X86/X86Schedule.td
@@ -117,6 +117,7 @@ defm WritePOPCNT : X86SchedWritePair; // Bit population count.
 defm WriteLZCNT : X86SchedWritePair; // Leading zero count.
 defm WriteTZCNT : X86SchedWritePair; // Trailing zero count.
 defm WriteCMOV : X86SchedWritePair; // Conditional move.
+def  WriteFCMOV : SchedWrite; // X87 conditional move.
 def  WriteSETCC : SchedWrite; // Set register based on condition code.
 def  WriteSETCCStore : SchedWrite;
 
diff --git a/lib/Target/X86/X86ScheduleAtom.td b/lib/Target/X86/X86ScheduleAtom.td
index e81bb3605bf..b69a6281fd4 100644
--- a/lib/Target/X86/X86ScheduleAtom.td
+++ b/lib/Target/X86/X86ScheduleAtom.td
@@ -92,6 +92,7 @@ defm : AtomWriteResPair<WriteIDiv64, [AtomPort01], [AtomPort01],130,130,[130],[1
 defm : AtomWriteResPair<WriteCRC32, [AtomPort01], [AtomPort01]>; // NOTE: Doesn't exist on Atom.
 
 defm : AtomWriteResPair<WriteCMOV,  [AtomPort01], [AtomPort0]>;
+defm : X86WriteRes<WriteFCMOV, [AtomPort01], 9, [9], 1>; // x87 conditional move.
 
 def  : WriteRes<WriteSETCC, [AtomPort01]>;
 def  : WriteRes<WriteSETCCStore, [AtomPort01]> {
@@ -593,8 +594,7 @@ def : InstRW<[AtomWrite01_9], (instrs BT16mr, BT32mr, BT64mr,
                                       SHLD64mri8, SHRD64mri8,
                                       SHLD64rri8, SHRD64rri8,
                                       CMPXCHG8rr)>;
-def : InstRW<[AtomWrite01_9], (instregex "CMOV(B|BE|E|P|NB|NBE|NE|NP)_F",
-                                         "(U)?COM_FI", "TST_F",
+def : InstRW<[AtomWrite01_9], (instregex "(U)?COM_FI", "TST_F",
                                          "(U)?COMIS(D|S)rr",
                                          "CVT(T)?SS2SI64rr(_Int)?")>;
 
diff --git a/lib/Target/X86/X86ScheduleBtVer2.td b/lib/Target/X86/X86ScheduleBtVer2.td
index 01e41b9621f..112b0ebf4b2 100644
--- a/lib/Target/X86/X86ScheduleBtVer2.td
+++ b/lib/Target/X86/X86ScheduleBtVer2.td
@@ -172,6 +172,7 @@ defm : JWriteResIntPair<WriteIDiv64, [JALU1, JDiv], 41, [1, 41], 2>;
 defm : JWriteResIntPair<WriteCRC32,  [JALU01], 3, [4], 3>;
 
 defm : JWriteResIntPair<WriteCMOV,  [JALU01], 1>; // Conditional move.
+defm : X86WriteRes<WriteFCMOV, [JFPU0, JFPA], 3, [1,1], 1>; // x87 conditional move.
 def  : WriteRes<WriteSETCC, [JALU01]>; // Setcc.
 def  : WriteRes<WriteSETCCStore, [JALU01,JSAGU]>;
 
diff --git a/lib/Target/X86/X86ScheduleSLM.td b/lib/Target/X86/X86ScheduleSLM.td
index 396eb9fd814..ab39b241864 100644
--- a/lib/Target/X86/X86ScheduleSLM.td
+++ b/lib/Target/X86/X86ScheduleSLM.td
@@ -100,6 +100,7 @@ defm : SLMWriteResPair<WriteJump,   [SLM_IEC_RSV1],  1>;
 defm : SLMWriteResPair<WriteCRC32,  [SLM_IEC_RSV1],  3>;
 
 defm : SLMWriteResPair<WriteCMOV,  [SLM_IEC_RSV01], 2, [2]>;
+defm : X86WriteRes<WriteFCMOV, [SLM_FPC_RSV1], 3, [1], 1>; // x87 conditional move.
 def  : WriteRes<WriteSETCC, [SLM_IEC_RSV01]>;
 def  : WriteRes<WriteSETCCStore, [SLM_IEC_RSV01, SLM_MEC_RSV]> {
   // FIXME Latency and NumMicrOps?
diff --git a/lib/Target/X86/X86ScheduleZnver1.td b/lib/Target/X86/X86ScheduleZnver1.td
index 9af8373f549..0e68fabb017 100644
--- a/lib/Target/X86/X86ScheduleZnver1.td
+++ b/lib/Target/X86/X86ScheduleZnver1.td
@@ -367,6 +367,7 @@ def ZnWriteMicrocoded : SchedWriteRes<[]> {
 }
 
 def : SchedAlias<WriteMicrocoded, ZnWriteMicrocoded>;
+def : SchedAlias<WriteFCMOV, ZnWriteMicrocoded>;
 def : SchedAlias<WriteSystem, ZnWriteMicrocoded>;
 def : SchedAlias<WriteMPSAD, ZnWriteMicrocoded>;
 def : SchedAlias<WriteMPSADY, ZnWriteMicrocoded>;
@@ -802,8 +803,6 @@ def : InstRW<[ZnWriteFPU3], (instregex "LD_F1")>;
 // FLDPI FLDL2E etc.
 def : InstRW<[ZnWriteFPU3], (instregex "FLDPI", "FLDL2(T|E)", "FLDL(G|N)2")>;
 
-def : InstRW<[WriteMicrocoded], (instregex "CMOV(B|BE|E|P|NB|NBE|NE|NP)_F")>;
-
 // FNSTSW.
 // AX.
 def : InstRW<[WriteMicrocoded], (instrs FNSTSW16r)>;
-- 
2.11.0