[X86] Create schedule classes for BT(C|R|S)mi and BT(C|R|S)mr instructions

11 files changed, 71 insertions, 49 deletions

diff --git a/llvm/lib/Target/X86/X86InstrInfo.td b/llvm/lib/Target/X86/X86InstrInfo.td
index 673d95b0367..160401cdf7a 100644
--- a/llvm/lib/Target/X86/X86InstrInfo.td
+++ b/llvm/lib/Target/X86/X86InstrInfo.td
@@ -1830,7 +1830,7 @@ def BTC64rr : RI<0xBB, MRMDestReg, (outs GR64:$dst), (ins GR64:$src1, GR64:$src2
                  NotMemoryFoldable;
 } // SchedRW
 
-let mayLoad = 1, mayStore = 1, SchedRW = [WriteALULd, WriteRMW] in {
+let mayLoad = 1, mayStore = 1, SchedRW = [WriteBitTestSetRegRMW] in {
 def BTC16mr : I<0xBB, MRMDestMem, (outs), (ins i16mem:$src1, GR16:$src2),
                 "btc{w}\t{$src2, $src1|$src1, $src2}", []>,
                 OpSize16, TB, NotMemoryFoldable;
@@ -1851,7 +1851,7 @@ def BTC64ri8 : RIi8<0xBA, MRM7r, (outs GR64:$dst), (ins GR64:$src1, i64i8imm:$sr
                     "btc{q}\t{$src2, $src1|$src1, $src2}", []>, TB;
 } // SchedRW
 
-let mayLoad = 1, mayStore = 1, SchedRW = [WriteALULd, WriteRMW] in {
+let mayLoad = 1, mayStore = 1, SchedRW = [WriteBitTestSetImmRMW] in {
 def BTC16mi8 : Ii8<0xBA, MRM7m, (outs), (ins i16mem:$src1, i16i8imm:$src2),
                     "btc{w}\t{$src2, $src1|$src1, $src2}", []>, OpSize16, TB;
 def BTC32mi8 : Ii8<0xBA, MRM7m, (outs), (ins i32mem:$src1, i32i8imm:$src2),
@@ -1873,7 +1873,7 @@ def BTR64rr : RI<0xB3, MRMDestReg, (outs GR64:$dst), (ins GR64:$src1, GR64:$src2
                  NotMemoryFoldable;
 } // SchedRW
 
-let mayLoad = 1, mayStore = 1, SchedRW = [WriteALULd, WriteRMW] in {
+let mayLoad = 1, mayStore = 1, SchedRW = [WriteBitTestSetRegRMW] in {
 def BTR16mr : I<0xB3, MRMDestMem, (outs), (ins i16mem:$src1, GR16:$src2),
                 "btr{w}\t{$src2, $src1|$src1, $src2}", []>,
                 OpSize16, TB, NotMemoryFoldable;
@@ -1896,7 +1896,7 @@ def BTR64ri8 : RIi8<0xBA, MRM6r, (outs GR64:$dst), (ins GR64:$src1, i64i8imm:$sr
                     "btr{q}\t{$src2, $src1|$src1, $src2}", []>, TB;
 } // SchedRW
 
-let mayLoad = 1, mayStore = 1, SchedRW = [WriteALULd, WriteRMW] in {
+let mayLoad = 1, mayStore = 1, SchedRW = [WriteBitTestSetImmRMW] in {
 def BTR16mi8 : Ii8<0xBA, MRM6m, (outs), (ins i16mem:$src1, i16i8imm:$src2),
                     "btr{w}\t{$src2, $src1|$src1, $src2}", []>,
                     OpSize16, TB;
@@ -1920,7 +1920,7 @@ def BTS64rr : RI<0xAB, MRMDestReg, (outs GR64:$dst), (ins GR64:$src1, GR64:$src2
                NotMemoryFoldable;
 } // SchedRW
 
-let mayLoad = 1, mayStore = 1, SchedRW = [WriteALULd, WriteRMW] in {
+let mayLoad = 1, mayStore = 1, SchedRW = [WriteBitTestSetRegRMW] in {
 def BTS16mr : I<0xAB, MRMDestMem, (outs), (ins i16mem:$src1, GR16:$src2),
               "bts{w}\t{$src2, $src1|$src1, $src2}", []>,
               OpSize16, TB, NotMemoryFoldable;
@@ -1941,7 +1941,7 @@ def BTS64ri8 : RIi8<0xBA, MRM5r, (outs GR64:$dst), (ins GR64:$src1, i64i8imm:$sr
                     "bts{q}\t{$src2, $src1|$src1, $src2}", []>, TB;
 } // SchedRW
 
-let mayLoad = 1, mayStore = 1, SchedRW = [WriteALULd, WriteRMW] in {
+let mayLoad = 1, mayStore = 1, SchedRW = [WriteBitTestSetImmRMW] in {
 def BTS16mi8 : Ii8<0xBA, MRM5m, (outs), (ins i16mem:$src1, i16i8imm:$src2),
                     "bts{w}\t{$src2, $src1|$src1, $src2}", []>, OpSize16, TB;
 def BTS32mi8 : Ii8<0xBA, MRM5m, (outs), (ins i32mem:$src1, i32i8imm:$src2),
diff --git a/llvm/lib/Target/X86/X86SchedBroadwell.td b/llvm/lib/Target/X86/X86SchedBroadwell.td
index 94711c99afd..c24afeae641 100644
--- a/llvm/lib/Target/X86/X86SchedBroadwell.td
+++ b/llvm/lib/Target/X86/X86SchedBroadwell.td
@@ -162,11 +162,13 @@ def  : WriteRes<WriteSETCCStore, [BWPort06,BWPort4,BWPort237]> {
   let NumMicroOps = 3;
 }
 
-defm : X86WriteRes<WriteLAHFSAHF,     [BWPort06], 1, [1], 1>;
-defm : X86WriteRes<WriteBitTest,      [BWPort06], 1, [1], 1>; // Bit Test instrs
-defm : X86WriteRes<WriteBitTestImmLd, [BWPort06,BWPort23], 6, [1,1], 2>;
-defm : X86WriteRes<WriteBitTestRegLd, [BWPort0156,BWPort23], 6, [1,1], 2>;
-defm : X86WriteRes<WriteBitTestSet,   [BWPort06], 1, [1], 1>; // Bit Test + Set instrs
+defm : X86WriteRes<WriteLAHFSAHF,        [BWPort06], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTest,         [BWPort06], 1, [1], 1>; // Bit Test instrs
+defm : X86WriteRes<WriteBitTestImmLd,    [BWPort06,BWPort23], 6, [1,1], 2>;
+defm : X86WriteRes<WriteBitTestRegLd,    [BWPort0156,BWPort23], 6, [1,1], 2>;
+defm : X86WriteRes<WriteBitTestSet,      [BWPort06], 1, [1], 1>; // Bit Test + Set instrs
+defm : X86WriteRes<WriteBitTestSetImmLd, [BWPort06,BWPort23], 6, [1,1], 2>;
+defm : X86WriteRes<WriteBitTestSetRegLd, [BWPort0156,BWPort23], 5, [1,1], 2>;
 
 // Bit counts.
 defm : BWWriteResPair<WriteBSF, [BWPort1], 3>;
diff --git a/llvm/lib/Target/X86/X86SchedHaswell.td b/llvm/lib/Target/X86/X86SchedHaswell.td
index 1714acbcb84..325ed7f0aa9 100644
--- a/llvm/lib/Target/X86/X86SchedHaswell.td
+++ b/llvm/lib/Target/X86/X86SchedHaswell.td
@@ -166,11 +166,13 @@ def  : WriteRes<WriteSETCCStore, [HWPort06,HWPort4,HWPort237]> {
   let NumMicroOps = 3;
 }
 
-defm : X86WriteRes<WriteLAHFSAHF,     [HWPort06], 1, [1], 1>;
-defm : X86WriteRes<WriteBitTest,      [HWPort06], 1, [1], 1>;
-defm : X86WriteRes<WriteBitTestImmLd, [HWPort06,HWPort23], 6, [1,1], 2>;
-defm : X86WriteRes<WriteBitTestRegLd, [], 1, [], 10>;
-defm : X86WriteRes<WriteBitTestSet,   [HWPort06], 1, [1], 1>;
+defm : X86WriteRes<WriteLAHFSAHF,        [HWPort06], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTest,         [HWPort06], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTestImmLd,    [HWPort06,HWPort23], 6, [1,1], 2>;
+defm : X86WriteRes<WriteBitTestRegLd,    [], 1, [], 10>;
+defm : X86WriteRes<WriteBitTestSet,      [HWPort06], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTestSetImmLd, [HWPort06,HWPort23], 6, [1,1], 2>;
+defm : X86WriteRes<WriteBitTestSetRegLd, [], 1, [], 11>;
 
 // This is for simple LEAs with one or two input operands.
 // The complex ones can only execute on port 1, and they require two cycles on
diff --git a/llvm/lib/Target/X86/X86SchedSandyBridge.td b/llvm/lib/Target/X86/X86SchedSandyBridge.td
index 8e09c019e05..3286b8fafd0 100644
--- a/llvm/lib/Target/X86/X86SchedSandyBridge.td
+++ b/llvm/lib/Target/X86/X86SchedSandyBridge.td
@@ -161,11 +161,13 @@ def  : WriteRes<WriteSETCCStore, [SBPort05,SBPort4,SBPort23]> {
   let NumMicroOps = 3;
 }
 
-defm : X86WriteRes<WriteLAHFSAHF,     [SBPort05], 1, [1], 1>;
-defm : X86WriteRes<WriteBitTest,      [SBPort05], 1, [1], 1>;
-defm : X86WriteRes<WriteBitTestImmLd, [SBPort05,SBPort23], 6, [1,1], 2>;
-defm : X86WriteRes<WriteBitTestRegLd, [SBPort05,SBPort23], 6, [1,1], 2>;
-defm : X86WriteRes<WriteBitTestSet,   [SBPort05], 1, [1], 1>;
+defm : X86WriteRes<WriteLAHFSAHF,        [SBPort05], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTest,         [SBPort05], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTestImmLd,    [SBPort05,SBPort23], 6, [1,1], 2>;
+defm : X86WriteRes<WriteBitTestRegLd,    [SBPort05,SBPort23], 6, [1,1], 2>;
+defm : X86WriteRes<WriteBitTestSet,      [SBPort05], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTestSetImmLd, [SBPort05,SBPort23], 6, [1,1], 2>;
+defm : X86WriteRes<WriteBitTestSetRegLd, [SBPort05,SBPort23], 6, [1,1], 2>;
 
 // This is for simple LEAs with one or two input operands.
 // The complex ones can only execute on port 1, and they require two cycles on
diff --git a/llvm/lib/Target/X86/X86SchedSkylakeClient.td b/llvm/lib/Target/X86/X86SchedSkylakeClient.td
index 59f29e4b764..b525948cc9b 100644
--- a/llvm/lib/Target/X86/X86SchedSkylakeClient.td
+++ b/llvm/lib/Target/X86/X86SchedSkylakeClient.td
@@ -159,11 +159,13 @@ def  : WriteRes<WriteSETCCStore, [SKLPort06,SKLPort4,SKLPort237]> {
   let NumMicroOps = 3;
 }
 
-defm : X86WriteRes<WriteLAHFSAHF,     [SKLPort06], 1, [1], 1>;
-defm : X86WriteRes<WriteBitTest,      [SKLPort06], 1, [1], 1>;
-defm : X86WriteRes<WriteBitTestImmLd, [SKLPort06,SKLPort23], 6, [1,1], 2>;
-defm : X86WriteRes<WriteBitTestRegLd, [SKLPort0156,SKLPort23], 6, [1,1], 2>;
-defm : X86WriteRes<WriteBitTestSet,   [SKLPort06], 1, [1], 1>;
+defm : X86WriteRes<WriteLAHFSAHF,        [SKLPort06], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTest,         [SKLPort06], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTestImmLd,    [SKLPort06,SKLPort23], 6, [1,1], 2>;
+defm : X86WriteRes<WriteBitTestRegLd,    [SKLPort0156,SKLPort23], 6, [1,1], 2>;
+defm : X86WriteRes<WriteBitTestSet,      [SKLPort06], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTestSetImmLd, [SKLPort06,SKLPort23], 6, [1,1], 2>;
+defm : X86WriteRes<WriteBitTestSetRegLd, [SKLPort0156,SKLPort23], 5, [1,1], 2>;
 
 // Bit counts.
 defm : SKLWriteResPair<WriteBSF, [SKLPort1], 3>;
diff --git a/llvm/lib/Target/X86/X86SchedSkylakeServer.td b/llvm/lib/Target/X86/X86SchedSkylakeServer.td
index 8a3a9592ef8..77de215f201 100644
--- a/llvm/lib/Target/X86/X86SchedSkylakeServer.td
+++ b/llvm/lib/Target/X86/X86SchedSkylakeServer.td
@@ -159,11 +159,13 @@ def  : WriteRes<WriteSETCCStore, [SKXPort06,SKXPort4,SKXPort237]> {
   let Latency = 2;
   let NumMicroOps = 3;
 }
-defm : X86WriteRes<WriteLAHFSAHF,     [SKXPort06], 1, [1], 1>;
-defm : X86WriteRes<WriteBitTest,      [SKXPort06], 1, [1], 1>;
-defm : X86WriteRes<WriteBitTestImmLd, [SKXPort06,SKXPort23], 6, [1,1], 2>;
-defm : X86WriteRes<WriteBitTestRegLd, [SKXPort0156,SKXPort23], 6, [1,1], 2>;
-defm : X86WriteRes<WriteBitTestSet,   [SKXPort06], 1, [1], 1>;
+defm : X86WriteRes<WriteLAHFSAHF,        [SKXPort06], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTest,         [SKXPort06], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTestImmLd,    [SKXPort06,SKXPort23], 6, [1,1], 2>;
+defm : X86WriteRes<WriteBitTestRegLd,    [SKXPort0156,SKXPort23], 6, [1,1], 2>;
+defm : X86WriteRes<WriteBitTestSet,      [SKXPort06], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTestSetImmLd, [SKXPort06,SKXPort23], 6, [1,1], 2>;
+defm : X86WriteRes<WriteBitTestSetRegLd, [SKXPort0156,SKXPort23], 5, [1,1], 2>;
 
 // Integer shifts and rotates.
 defm : SKXWriteResPair<WriteShift,    [SKXPort06],  1>;
diff --git a/llvm/lib/Target/X86/X86Schedule.td b/llvm/lib/Target/X86/X86Schedule.td
index cf6ffc31b37..d226d448f14 100644
--- a/llvm/lib/Target/X86/X86Schedule.td
+++ b/llvm/lib/Target/X86/X86Schedule.td
@@ -160,7 +160,11 @@ def  WriteBitTest      : SchedWrite; // Bit Test
 def  WriteBitTestImmLd : SchedWrite;
 def  WriteBitTestRegLd : SchedWrite;
 
-def  WriteBitTestSet  : SchedWrite; // Bit Test + Set - TODO add memory folding support
+def  WriteBitTestSet       : SchedWrite; // Bit Test + Set
+def  WriteBitTestSetImmLd  : SchedWrite;
+def  WriteBitTestSetRegLd  : SchedWrite;
+def  WriteBitTestSetImmRMW : WriteSequence<[WriteBitTestSetImmLd, WriteRMW]>;
+def  WriteBitTestSetRegRMW : WriteSequence<[WriteBitTestSetRegLd, WriteRMW]>;
 
 // Integer shifts and rotates.
 defm WriteShift    : X86SchedWritePair;
diff --git a/llvm/lib/Target/X86/X86ScheduleAtom.td b/llvm/lib/Target/X86/X86ScheduleAtom.td
index 1034101a76f..dd9bd92ef16 100644
--- a/llvm/lib/Target/X86/X86ScheduleAtom.td
+++ b/llvm/lib/Target/X86/X86ScheduleAtom.td
@@ -121,10 +121,12 @@ def  : WriteRes<WriteLAHFSAHF, [AtomPort01]> {
   let Latency = 2;
   let ResourceCycles = [2];
 }
-defm : X86WriteRes<WriteBitTest,      [AtomPort1],  1, [1], 1>;
-defm : X86WriteRes<WriteBitTestImmLd, [AtomPort0],  1, [1], 1>;
-defm : X86WriteRes<WriteBitTestRegLd, [AtomPort01], 9, [9], 1>;
-defm : X86WriteRes<WriteBitTestSet,   [AtomPort1],  1, [1], 1>;
+defm : X86WriteRes<WriteBitTest,         [AtomPort1],  1, [1], 1>;
+defm : X86WriteRes<WriteBitTestImmLd,    [AtomPort0],  1, [1], 1>;
+defm : X86WriteRes<WriteBitTestRegLd,    [AtomPort01], 9, [9], 1>;
+defm : X86WriteRes<WriteBitTestSet,      [AtomPort1],  1, [1], 1>;
+defm : X86WriteRes<WriteBitTestSetImmLd, [AtomPort1],  1, [1], 1>;
+defm : X86WriteRes<WriteBitTestSetRegLd, [AtomPort1],  1, [1], 1>;
 
 // This is for simple LEAs with one or two input operands.
 def : WriteRes<WriteLEA, [AtomPort1]>;
diff --git a/llvm/lib/Target/X86/X86ScheduleBtVer2.td b/llvm/lib/Target/X86/X86ScheduleBtVer2.td
index 7871a5ebf72..8ac12c029f1 100644
--- a/llvm/lib/Target/X86/X86ScheduleBtVer2.td
+++ b/llvm/lib/Target/X86/X86ScheduleBtVer2.td
@@ -204,10 +204,12 @@ def  : WriteRes<WriteSETCC, [JALU01]>; // Setcc.
 def  : WriteRes<WriteSETCCStore, [JALU01,JSAGU]>;
 def  : WriteRes<WriteLAHFSAHF, [JALU01]>;
 
-defm : X86WriteRes<WriteBitTest,      [JALU01], 1, [1], 1>;
-defm : X86WriteRes<WriteBitTestImmLd, [JALU01, JLAGU], 4, [1, 1], 1>;
-defm : X86WriteRes<WriteBitTestRegLd, [JALU01, JLAGU], 4, [1, 1], 5>;
-defm : X86WriteRes<WriteBitTestSet,   [JALU01], 1, [1], 2>;
+defm : X86WriteRes<WriteBitTest,         [JALU01], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTestImmLd,    [JALU01,JLAGU], 4, [1,1], 1>;
+defm : X86WriteRes<WriteBitTestRegLd,    [JALU01,JLAGU], 4, [1,1], 5>;
+defm : X86WriteRes<WriteBitTestSet,      [JALU01], 1, [1], 2>;
+defm : X86WriteRes<WriteBitTestSetImmLd, [JALU01,JLAGU], 3, [1,1], 1>;
+defm : X86WriteRes<WriteBitTestSetRegLd, [JALU01,JLAGU], 3, [1,1], 1>;
 
 // This is for simple LEAs with one or two input operands.
 def : WriteRes<WriteLEA, [JALU01]>;
diff --git a/llvm/lib/Target/X86/X86ScheduleSLM.td b/llvm/lib/Target/X86/X86ScheduleSLM.td
index 372cc863ce1..b1a10c4aac7 100644
--- a/llvm/lib/Target/X86/X86ScheduleSLM.td
+++ b/llvm/lib/Target/X86/X86ScheduleSLM.td
@@ -134,11 +134,13 @@ def  : WriteRes<WriteSETCCStore, [SLM_IEC_RSV01, SLM_MEC_RSV]> {
   // FIXME Latency and NumMicrOps?
   let ResourceCycles = [2,1];
 }
-defm : X86WriteRes<WriteLAHFSAHF,     [SLM_IEC_RSV01], 1, [1], 1>;
-defm : X86WriteRes<WriteBitTest,      [SLM_IEC_RSV01], 1, [1], 1>;
-defm : X86WriteRes<WriteBitTestImmLd, [SLM_IEC_RSV01, SLM_MEC_RSV], 4, [1,1], 1>;
-defm : X86WriteRes<WriteBitTestRegLd, [SLM_IEC_RSV01, SLM_MEC_RSV], 4, [1,1], 1>;
-defm : X86WriteRes<WriteBitTestSet,   [SLM_IEC_RSV01], 1, [1], 1>;
+defm : X86WriteRes<WriteLAHFSAHF,        [SLM_IEC_RSV01], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTest,         [SLM_IEC_RSV01], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTestImmLd,    [SLM_IEC_RSV01, SLM_MEC_RSV], 4, [1,1], 1>;
+defm : X86WriteRes<WriteBitTestRegLd,    [SLM_IEC_RSV01, SLM_MEC_RSV], 4, [1,1], 1>;
+defm : X86WriteRes<WriteBitTestSet,      [SLM_IEC_RSV01], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTestSetImmLd, [SLM_IEC_RSV01, SLM_MEC_RSV], 3, [1,1], 1>;
+defm : X86WriteRes<WriteBitTestSetRegLd, [SLM_IEC_RSV01, SLM_MEC_RSV], 3, [1,1], 1>;
 
 // This is for simple LEAs with one or two input operands.
 // The complex ones can only execute on port 1, and they require two cycles on
diff --git a/llvm/lib/Target/X86/X86ScheduleZnver1.td b/llvm/lib/Target/X86/X86ScheduleZnver1.td
index f97d99e5b17..d8bf7e77ce7 100644
--- a/llvm/lib/Target/X86/X86ScheduleZnver1.td
+++ b/llvm/lib/Target/X86/X86ScheduleZnver1.td
@@ -214,10 +214,12 @@ def  : WriteRes<WriteSETCC,  [ZnALU]>;
 def  : WriteRes<WriteSETCCStore,  [ZnALU, ZnAGU]>;
 defm : X86WriteRes<WriteLAHFSAHF, [ZnALU], 2, [1], 2>;
 
-defm : X86WriteRes<WriteBitTest,      [ZnALU], 1, [1], 1>;
-defm : X86WriteRes<WriteBitTestImmLd, [ZnALU,ZnAGU], 5, [1,1], 2>;
-defm : X86WriteRes<WriteBitTestRegLd, [ZnALU,ZnAGU], 5, [1,1], 2>;
-defm : X86WriteRes<WriteBitTestSet,   [ZnALU], 2, [1], 2>;
+defm : X86WriteRes<WriteBitTest,         [ZnALU], 1, [1], 1>;
+defm : X86WriteRes<WriteBitTestImmLd,    [ZnALU,ZnAGU], 5, [1,1], 2>;
+defm : X86WriteRes<WriteBitTestRegLd,    [ZnALU,ZnAGU], 5, [1,1], 2>;
+defm : X86WriteRes<WriteBitTestSet,      [ZnALU], 2, [1], 2>;
+defm : X86WriteRes<WriteBitTestSetImmLd, [ZnALU,ZnAGU], 5, [1,1], 2>;
+defm : X86WriteRes<WriteBitTestSetRegLd, [ZnALU,ZnAGU], 5, [1,1], 2>;
 
 // Bit counts.
 defm : ZnWriteResPair<WriteBSF, [ZnALU], 3>;