From: Andrew V. Tischenko <andrew.v.tischenko@gmail.com>
Date: Fri, 20 Jul 2018 09:39:14 +0000 (+0000)
Subject: Improved sched model for X86 BSWAP* instrs.
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=f638e5271d8eb22b8a386fa176bb0ed29bad8d31;p=llvm

Improved sched model for X86 BSWAP* instrs.
Differential Revision: https://reviews.llvm.org/D49477


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@337537 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/Target/X86/X86InstrInfo.td b/lib/Target/X86/X86InstrInfo.td
index 5e30e00a1bd..7509b312c10 100644
--- a/lib/Target/X86/X86InstrInfo.td
+++ b/lib/Target/X86/X86InstrInfo.td
@@ -1352,7 +1352,7 @@ def PUSHA16  : I<0x60, RawFrm, (outs), (ins), "pushaw", []>,
                OpSize16, Requires<[Not64BitMode]>;
 }
 
-let Constraints = "$src = $dst", SchedRW = [WriteALU] in {
+let Constraints = "$src = $dst", SchedRW = [WriteBSWAP32] in {
 // This instruction is a consequence of BSWAP32r observing operand size. The
 // encoding is valid, but the behavior is undefined.
 let isCodeGenOnly = 1, ForceDisassemble = 1, hasSideEffects = 0 in
@@ -1363,6 +1363,7 @@ def BSWAP32r : I<0xC8, AddRegFrm, (outs GR32:$dst), (ins GR32:$src),
                  "bswap{l}\t$dst",
                  [(set GR32:$dst, (bswap GR32:$src))]>, OpSize32, TB;
 
+let SchedRW = [WriteBSWAP64] in
 def BSWAP64r : RI<0xC8, AddRegFrm, (outs GR64:$dst), (ins GR64:$src),
                   "bswap{q}\t$dst",
                   [(set GR64:$dst, (bswap GR64:$src))]>, TB;
diff --git a/lib/Target/X86/X86SchedBroadwell.td b/lib/Target/X86/X86SchedBroadwell.td
index 2590d9d08ac..c7713fea70f 100755
--- a/lib/Target/X86/X86SchedBroadwell.td
+++ b/lib/Target/X86/X86SchedBroadwell.td
@@ -110,7 +110,6 @@ defm : BWWriteResPair<WriteALU,    [BWPort0156], 1>; // Simple integer ALU op.
 defm : BWWriteResPair<WriteADC,    [BWPort06], 1>; // Integer ALU + flags op.
 defm : BWWriteResPair<WriteIMul,   [BWPort1], 3>; // Integer multiplication.
 defm : BWWriteResPair<WriteIMul64, [BWPort1], 3>; // Integer 64-bit multiplication.
-
 defm : BWWriteResPair<WriteDiv8,   [BWPort0, BWDivider], 25, [1, 10]>;
 defm : BWWriteResPair<WriteDiv16,  [BWPort0, BWDivider], 25, [1, 10]>;
 defm : BWWriteResPair<WriteDiv32,  [BWPort0, BWDivider], 25, [1, 10]>;
@@ -120,6 +119,9 @@ 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 : BWWriteResPair<WriteBSWAP32,[BWPort15], 1>; //
+defm : BWWriteResPair<WriteBSWAP64,[BWPort06, BWPort15], 2, [1, 1], 2>; //
+
 defm : BWWriteResPair<WriteCRC32, [BWPort1],   3>;
 def : WriteRes<WriteIMulH, []> { let Latency = 3; } // Integer multiplication, high part.
 
@@ -699,20 +701,6 @@ def BWWriteResGroup18 : SchedWriteRes<[BWPort237,BWPort0156]> {
 }
 def: InstRW<[BWWriteResGroup18], (instrs SFENCE)>;
 
-def BWWriteResGroup19 : SchedWriteRes<[BWPort06,BWPort15]> {
-  let Latency = 2;
-  let NumMicroOps = 2;
-  let ResourceCycles = [1,1];
-}
-def: InstRW<[BWWriteResGroup19], (instrs BSWAP64r)>;
-
-def BWWriteResGroup19_1 : SchedWriteRes<[BWPort15]> {
-  let Latency = 1;
-  let NumMicroOps = 1;
-  let ResourceCycles = [1];
-}
-def: InstRW<[BWWriteResGroup19_1], (instrs BSWAP32r)>;
-
 def BWWriteResGroup20 : SchedWriteRes<[BWPort06,BWPort0156]> {
   let Latency = 2;
   let NumMicroOps = 2;
diff --git a/lib/Target/X86/X86SchedHaswell.td b/lib/Target/X86/X86SchedHaswell.td
index 77d89c85df5..189dd418383 100644
--- a/lib/Target/X86/X86SchedHaswell.td
+++ b/lib/Target/X86/X86SchedHaswell.td
@@ -122,6 +122,10 @@ 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 : HWWriteResPair<WriteBSWAP32,[HWPort15],   1>;
+defm : HWWriteResPair<WriteBSWAP64,[HWPort06, HWPort15], 2, [1,1], 2>;
+
 def  : WriteRes<WriteIMulH, []> { let Latency = 3; }
 defm : HWWriteResPair<WriteShift,  [HWPort06],  1>;
 defm : HWWriteResPair<WriteShiftDouble,  [HWPort06],  1>;
@@ -1149,20 +1153,6 @@ def HWWriteResGroup33 : SchedWriteRes<[HWPort01,HWPort015]> {
 }
 def: InstRW<[HWWriteResGroup33], (instregex "MMX_MOVDQ2Qrr")>;
 
-def HWWriteResGroup34 : SchedWriteRes<[HWPort06,HWPort15]> {
-  let Latency = 2;
-  let NumMicroOps = 2;
-  let ResourceCycles = [1,1];
-}
-def: InstRW<[HWWriteResGroup34], (instrs BSWAP64r)>;
-
-def HWWriteResGroup34_1 : SchedWriteRes<[HWPort15]> {
-  let Latency = 1;
-  let NumMicroOps = 1;
-  let ResourceCycles = [1];
-}
-def: InstRW<[HWWriteResGroup34_1], (instrs BSWAP32r)>;
-
 def HWWriteResGroup35 : SchedWriteRes<[HWPort06,HWPort0156]> {
   let Latency = 2;
   let NumMicroOps = 2;
diff --git a/lib/Target/X86/X86SchedSandyBridge.td b/lib/Target/X86/X86SchedSandyBridge.td
index e6f94f703eb..3b543c680ef 100644
--- a/lib/Target/X86/X86SchedSandyBridge.td
+++ b/lib/Target/X86/X86SchedSandyBridge.td
@@ -111,6 +111,9 @@ defm : SBWriteResPair<WriteADC,    [SBPort05,SBPort015], 2, [1,1], 2>;
 defm : SBWriteResPair<WriteIMul,   [SBPort1],   3>;
 defm : SBWriteResPair<WriteIMul64, [SBPort1],   3>;
 
+defm : SBWriteResPair<WriteBSWAP32,[SBPort1], 1>;
+defm : SBWriteResPair<WriteBSWAP64,[SBPort1,SBPort05], 2, [1,1], 2>;
+
 defm : SBWriteResPair<WriteDiv8,   [SBPort0, SBDivider], 25, [1, 10]>;
 defm : SBWriteResPair<WriteDiv16,  [SBPort0, SBDivider], 25, [1, 10]>;
 defm : SBWriteResPair<WriteDiv32,  [SBPort0, SBDivider], 25, [1, 10]>;
@@ -619,20 +622,6 @@ def SBWriteResGroup15 : SchedWriteRes<[SBPort0,SBPort015]> {
 def: InstRW<[SBWriteResGroup15], (instrs CWD,
                                          FNSTSW16r)>;
 
-def SBWriteResGroup16 : SchedWriteRes<[SBPort1,SBPort05]> {
-  let Latency = 2;
-  let NumMicroOps = 2;
-  let ResourceCycles = [1,1];
-}
-def: InstRW<[SBWriteResGroup16], (instrs BSWAP64r)>;
-
-def SBWriteResGroup16_1 : SchedWriteRes<[SBPort1]> {
-  let Latency = 1;
-  let NumMicroOps = 1;
-  let ResourceCycles = [1];
-}
-def: InstRW<[SBWriteResGroup16_1], (instrs BSWAP32r)>;
-
 def SBWriteResGroup18 : SchedWriteRes<[SBPort5,SBPort015]> {
   let Latency = 2;
   let NumMicroOps = 2;
diff --git a/lib/Target/X86/X86SchedSkylakeClient.td b/lib/Target/X86/X86SchedSkylakeClient.td
index a59d2c459ee..1417799d76b 100644
--- a/lib/Target/X86/X86SchedSkylakeClient.td
+++ b/lib/Target/X86/X86SchedSkylakeClient.td
@@ -110,6 +110,9 @@ 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 : SKLWriteResPair<WriteBSWAP32,[SKLPort15],   1>; //
+defm : SKLWriteResPair<WriteBSWAP64,[SKLPort06, SKLPort15], 2, [1,1], 2>; //
+
 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>;
@@ -698,20 +701,6 @@ def SKLWriteResGroup21 : SchedWriteRes<[SKLPort237,SKLPort0156]> {
 }
 def: InstRW<[SKLWriteResGroup21], (instrs SFENCE)>;
 
-def SKLWriteResGroup22 : SchedWriteRes<[SKLPort06,SKLPort15]> {
-  let Latency = 2;
-  let NumMicroOps = 2;
-  let ResourceCycles = [1,1];
-}
-def: InstRW<[SKLWriteResGroup22], (instrs BSWAP64r)>;
-
-def SKLWriteResGroup22_1 : SchedWriteRes<[SKLPort15]> {
-  let Latency = 1;
-  let NumMicroOps = 1;
-  let ResourceCycles = [1];
-}
-def: InstRW<[SKLWriteResGroup22_1], (instrs BSWAP32r)>;
-
 def SKLWriteResGroup23 : SchedWriteRes<[SKLPort06,SKLPort0156]> {
   let Latency = 2;
   let NumMicroOps = 2;
diff --git a/lib/Target/X86/X86SchedSkylakeServer.td b/lib/Target/X86/X86SchedSkylakeServer.td
index b7327c223ff..7095ec081bd 100755
--- a/lib/Target/X86/X86SchedSkylakeServer.td
+++ b/lib/Target/X86/X86SchedSkylakeServer.td
@@ -110,6 +110,9 @@ 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 : SKXWriteResPair<WriteBSWAP32,[SKXPort15],   1>; //
+defm : SKXWriteResPair<WriteBSWAP64,[SKXPort06, SKXPort15], 2, [1,1], 2>; //
+
 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>;
@@ -722,20 +725,6 @@ def SKXWriteResGroup21 : SchedWriteRes<[SKXPort237,SKXPort0156]> {
 }
 def: InstRW<[SKXWriteResGroup21], (instrs SFENCE)>;
 
-def SKXWriteResGroup22 : SchedWriteRes<[SKXPort06,SKXPort15]> {
-  let Latency = 2;
-  let NumMicroOps = 2;
-  let ResourceCycles = [1,1];
-}
-def: InstRW<[SKXWriteResGroup22], (instrs BSWAP64r)>;
-
-def SKXWriteResGroup22_1 : SchedWriteRes<[SKXPort15]> {
-  let Latency = 1;
-  let NumMicroOps = 1;
-  let ResourceCycles = [1];
-}
-def: InstRW<[SKXWriteResGroup22_1], (instrs BSWAP32r)>;
-
 def SKXWriteResGroup23 : SchedWriteRes<[SKXPort06,SKXPort0156]> {
   let Latency = 2;
   let NumMicroOps = 2;
diff --git a/lib/Target/X86/X86Schedule.td b/lib/Target/X86/X86Schedule.td
index 270acb3c398..d0167753ccd 100644
--- a/lib/Target/X86/X86Schedule.td
+++ b/lib/Target/X86/X86Schedule.td
@@ -118,6 +118,9 @@ defm WriteIMul64 : X86SchedWritePair; // Integer 64-bit multiplication.
 def  WriteIMulH  : SchedWrite;        // Integer multiplication, high part.
 def  WriteLEA    : SchedWrite;        // LEA instructions can't fold loads.
 
+defm WriteBSWAP32: X86SchedWritePair; // Byte Order (Endiannes) Swap
+defm WriteBSWAP64: X86SchedWritePair; // Byte Order (Endiannes) Swap
+
 // Integer division.
 defm WriteDiv8   : X86SchedWritePair;
 defm WriteDiv16  : X86SchedWritePair;
diff --git a/lib/Target/X86/X86ScheduleAtom.td b/lib/Target/X86/X86ScheduleAtom.td
index 6ebc9f74270..d1e902e6c43 100644
--- a/lib/Target/X86/X86ScheduleAtom.td
+++ b/lib/Target/X86/X86ScheduleAtom.td
@@ -81,6 +81,9 @@ defm : AtomWriteResPair<WriteADC,    [AtomPort01], [AtomPort0]>;
 defm : AtomWriteResPair<WriteIMul,   [AtomPort01], [AtomPort01],  7,  7,  [7],  [7]>;
 defm : AtomWriteResPair<WriteIMul64, [AtomPort01], [AtomPort01], 12, 12, [12], [12]>;
 
+defm : AtomWriteResPair<WriteBSWAP32,    [AtomPort0], [AtomPort0]>;
+defm : AtomWriteResPair<WriteBSWAP64,    [AtomPort0], [AtomPort0]>;
+
 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]>;
@@ -489,7 +492,6 @@ def AtomWrite0_1 : SchedWriteRes<[AtomPort0]> {
   let ResourceCycles = [1];
 }
 def : InstRW<[AtomWrite0_1], (instrs FXAM, LD_Frr,
-                                     BSWAP32r, BSWAP64r,
                                      MOVSX64rr32)>;
 def : SchedAlias<WriteALURMW, AtomWrite0_1>;
 def : SchedAlias<WriteADCRMW, AtomWrite0_1>;
diff --git a/lib/Target/X86/X86ScheduleBtVer2.td b/lib/Target/X86/X86ScheduleBtVer2.td
index d539a7f668d..d78c343ebd5 100644
--- a/lib/Target/X86/X86ScheduleBtVer2.td
+++ b/lib/Target/X86/X86ScheduleBtVer2.td
@@ -168,6 +168,9 @@ defm : JWriteResIntPair<WriteIMul,   [JALU1, JMul], 3, [1, 1], 2>; // i8/i16/i32
 defm : JWriteResIntPair<WriteIMul64, [JALU1, JMul], 6, [1, 4], 2>; // i64 multiplication
 defm : X86WriteRes<WriteIMulH,       [JALU1], 6, [4], 1>;
 
+defm : JWriteResIntPair<WriteBSWAP32,[JALU01], 1>;
+defm : JWriteResIntPair<WriteBSWAP64,[JALU01], 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>;
diff --git a/lib/Target/X86/X86ScheduleSLM.td b/lib/Target/X86/X86ScheduleSLM.td
index b5425afd657..c938a4a8939 100644
--- a/lib/Target/X86/X86ScheduleSLM.td
+++ b/lib/Target/X86/X86ScheduleSLM.td
@@ -97,6 +97,10 @@ 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 : SLMWriteResPair<WriteBSWAP32,[SLM_IEC_RSV01], 1>;
+defm : SLMWriteResPair<WriteBSWAP64,[SLM_IEC_RSV01], 1>;
+
 defm : SLMWriteResPair<WriteShift,  [SLM_IEC_RSV0],  1>;
 defm : SLMWriteResPair<WriteShiftDouble,  [SLM_IEC_RSV0],  1>;
 defm : SLMWriteResPair<WriteJump,   [SLM_IEC_RSV1],  1>;
diff --git a/lib/Target/X86/X86ScheduleZnver1.td b/lib/Target/X86/X86ScheduleZnver1.td
index 6e4835feab1..3e343bc7546 100644
--- a/lib/Target/X86/X86ScheduleZnver1.td
+++ b/lib/Target/X86/X86ScheduleZnver1.td
@@ -179,6 +179,10 @@ 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 : ZnWriteResPair<WriteBSWAP32,[ZnALU], 1, [4]>;
+defm : ZnWriteResPair<WriteBSWAP64,[ZnALU], 1, [4]>;
+
 defm : ZnWriteResPair<WriteShift, [ZnALU], 1>;
 defm : ZnWriteResPair<WriteShiftDouble, [ZnALU], 1>;
 defm : ZnWriteResPair<WriteJump,  [ZnALU], 1>;
@@ -537,12 +541,6 @@ def : InstRW<[ZnWritePushA], (instregex "PUSHA(16|32)")>;
 //LAHF
 def : InstRW<[WriteMicrocoded], (instrs LAHF)>;
 
-// BSWAP.
-def ZnWriteBSwap : SchedWriteRes<[ZnALU]> {
-  let ResourceCycles = [4];
-}
-def : InstRW<[ZnWriteBSwap], (instregex "BSWAP")>;
-
 // MOVBE.
 // r,m.
 def ZnWriteMOVBE : SchedWriteRes<[ZnAGU, ZnALU]> {