From bf4c8c0ff294b86ab6769820f20e570b274e78f8 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Fri, 4 May 2018 14:54:33 +0000
Subject: [PATCH] [X86] Add WriteVecMOVMSKY scheduler class

llvm-svn: 331525
---
 llvm/lib/Target/X86/X86InstrSSE.td           |  9 +++++----
 llvm/lib/Target/X86/X86SchedBroadwell.td     |  7 ++++---
 llvm/lib/Target/X86/X86SchedHaswell.td       |  7 ++++---
 llvm/lib/Target/X86/X86SchedSandyBridge.td   |  7 ++++---
 llvm/lib/Target/X86/X86SchedSkylakeClient.td |  7 ++++---
 llvm/lib/Target/X86/X86SchedSkylakeServer.td |  7 ++++---
 llvm/lib/Target/X86/X86Schedule.td           |  7 ++++---
 llvm/lib/Target/X86/X86ScheduleAtom.td       |  7 ++++---
 llvm/lib/Target/X86/X86ScheduleBtVer2.td     |  7 ++++---
 llvm/lib/Target/X86/X86ScheduleSLM.td        |  7 ++++---
 llvm/lib/Target/X86/X86ScheduleZnver1.td     | 16 +++++++---------
 11 files changed, 48 insertions(+), 40 deletions(-)

diff --git a/llvm/lib/Target/X86/X86InstrSSE.td b/llvm/lib/Target/X86/X86InstrSSE.td
index 3857dcd58ba4..10b8cac81f84 100644
--- a/llvm/lib/Target/X86/X86InstrSSE.td
+++ b/llvm/lib/Target/X86/X86InstrSSE.td
@@ -3855,25 +3855,26 @@ defm PINSRW : sse2_pinsrw, PD;
 // SSE2 - Packed Mask Creation
 //===---------------------------------------------------------------------===//
 
-let ExeDomain = SSEPackedInt, SchedRW = [WriteVecMOVMSK] in {
+let ExeDomain = SSEPackedInt in {
 
 def VPMOVMSKBrr  : VPDI<0xD7, MRMSrcReg, (outs GR32orGR64:$dst),
            (ins VR128:$src),
            "pmovmskb\t{$src, $dst|$dst, $src}",
            [(set GR32orGR64:$dst, (X86movmsk (v16i8 VR128:$src)))]>,
-           VEX, VEX_WIG;
+           Sched<[WriteVecMOVMSK]>, VEX, VEX_WIG;
 
 let Predicates = [HasAVX2] in {
 def VPMOVMSKBYrr  : VPDI<0xD7, MRMSrcReg, (outs GR32orGR64:$dst),
            (ins VR256:$src),
            "pmovmskb\t{$src, $dst|$dst, $src}",
            [(set GR32orGR64:$dst, (X86movmsk (v32i8 VR256:$src)))]>,
-           VEX, VEX_L, VEX_WIG;
+           Sched<[WriteVecMOVMSKY]>, VEX, VEX_L, VEX_WIG;
 }
 
 def PMOVMSKBrr : PDI<0xD7, MRMSrcReg, (outs GR32orGR64:$dst), (ins VR128:$src),
            "pmovmskb\t{$src, $dst|$dst, $src}",
-           [(set GR32orGR64:$dst, (X86movmsk (v16i8 VR128:$src)))]>;
+           [(set GR32orGR64:$dst, (X86movmsk (v16i8 VR128:$src)))]>,
+           Sched<[WriteVecMOVMSK]>;
 
 } // ExeDomain = SSEPackedInt
 
diff --git a/llvm/lib/Target/X86/X86SchedBroadwell.td b/llvm/lib/Target/X86/X86SchedBroadwell.td
index 593fb6e819b3..5aac5957b228 100755
--- a/llvm/lib/Target/X86/X86SchedBroadwell.td
+++ b/llvm/lib/Target/X86/X86SchedBroadwell.td
@@ -315,9 +315,10 @@ def : WriteRes<WritePCmpEStrILd, [BWPort0, BWPort5, BWPort23, BWPort0156]> {
 }
 
 // MOVMSK Instructions.
-def : WriteRes<WriteFMOVMSK, [BWPort0]> { let Latency = 3; }
-def : WriteRes<WriteVecMOVMSK, [BWPort0]> { let Latency = 3; }
-def : WriteRes<WriteMMXMOVMSK, [BWPort0]> { let Latency = 1; }
+def : WriteRes<WriteFMOVMSK,    [BWPort0]> { let Latency = 3; }
+def : WriteRes<WriteVecMOVMSK,  [BWPort0]> { let Latency = 3; }
+def : WriteRes<WriteVecMOVMSKY, [BWPort0]> { let Latency = 3; }
+def : WriteRes<WriteMMXMOVMSK,  [BWPort0]> { let Latency = 1; }
 
 // AES instructions.
 def : WriteRes<WriteAESDecEnc, [BWPort5]> { // Decryption, encryption.
diff --git a/llvm/lib/Target/X86/X86SchedHaswell.td b/llvm/lib/Target/X86/X86SchedHaswell.td
index f4651de16083..79a8f9ee1e95 100644
--- a/llvm/lib/Target/X86/X86SchedHaswell.td
+++ b/llvm/lib/Target/X86/X86SchedHaswell.td
@@ -308,9 +308,10 @@ def : WriteRes<WritePCmpEStrILd, [HWPort0, HWPort5, HWPort23, HWPort0156]> {
 }
 
 // MOVMSK Instructions.
-def : WriteRes<WriteFMOVMSK, [HWPort0]> { let Latency = 3; }
-def : WriteRes<WriteVecMOVMSK, [HWPort0]> { let Latency = 3; }
-def : WriteRes<WriteMMXMOVMSK, [HWPort0]> { let Latency = 1; }
+def : WriteRes<WriteFMOVMSK,    [HWPort0]> { let Latency = 3; }
+def : WriteRes<WriteVecMOVMSK,  [HWPort0]> { let Latency = 3; }
+def : WriteRes<WriteVecMOVMSKY, [HWPort0]> { let Latency = 3; }
+def : WriteRes<WriteMMXMOVMSK,  [HWPort0]> { let Latency = 1; }
 
 // AES Instructions.
 def : WriteRes<WriteAESDecEnc, [HWPort5]> {
diff --git a/llvm/lib/Target/X86/X86SchedSandyBridge.td b/llvm/lib/Target/X86/X86SchedSandyBridge.td
index c5ef8835d5e1..fe8de4488728 100644
--- a/llvm/lib/Target/X86/X86SchedSandyBridge.td
+++ b/llvm/lib/Target/X86/X86SchedSandyBridge.td
@@ -287,9 +287,10 @@ def : WriteRes<WritePCmpEStrILd, [SBPort015, SBPort23]> {
 }
 
 // MOVMSK Instructions.
-def : WriteRes<WriteFMOVMSK, [SBPort0]> { let Latency = 2; }
-def : WriteRes<WriteVecMOVMSK, [SBPort0]> { let Latency = 2; }
-def : WriteRes<WriteMMXMOVMSK, [SBPort0]> { let Latency = 1; }
+def : WriteRes<WriteFMOVMSK,    [SBPort0]> { let Latency = 2; }
+def : WriteRes<WriteVecMOVMSK,  [SBPort0]> { let Latency = 2; }
+def : WriteRes<WriteVecMOVMSKY, [SBPort0]> { let Latency = 2; }
+def : WriteRes<WriteMMXMOVMSK,  [SBPort0]> { let Latency = 1; }
 
 // AES Instructions.
 def : WriteRes<WriteAESDecEnc, [SBPort5,SBPort015]> {
diff --git a/llvm/lib/Target/X86/X86SchedSkylakeClient.td b/llvm/lib/Target/X86/X86SchedSkylakeClient.td
index 7aa93195d4a8..36675a060e31 100644
--- a/llvm/lib/Target/X86/X86SchedSkylakeClient.td
+++ b/llvm/lib/Target/X86/X86SchedSkylakeClient.td
@@ -310,9 +310,10 @@ def : WriteRes<WritePCmpEStrILd, [SKLPort0, SKLPort5, SKLPort23, SKLPort0156]> {
 }
 
 // MOVMSK Instructions.
-def : WriteRes<WriteFMOVMSK, [SKLPort0]> { let Latency = 2; }
-def : WriteRes<WriteVecMOVMSK, [SKLPort0]> { let Latency = 2; }
-def : WriteRes<WriteMMXMOVMSK, [SKLPort0]> { let Latency = 2; }
+def : WriteRes<WriteFMOVMSK,    [SKLPort0]> { let Latency = 2; }
+def : WriteRes<WriteVecMOVMSK,  [SKLPort0]> { let Latency = 2; }
+def : WriteRes<WriteVecMOVMSKY, [SKLPort0]> { let Latency = 2; }
+def : WriteRes<WriteMMXMOVMSK,  [SKLPort0]> { let Latency = 2; }
 
 // AES instructions.
 def : WriteRes<WriteAESDecEnc, [SKLPort0]> { // Decryption, encryption.
diff --git a/llvm/lib/Target/X86/X86SchedSkylakeServer.td b/llvm/lib/Target/X86/X86SchedSkylakeServer.td
index 52629e53707b..7ab6d4f09525 100755
--- a/llvm/lib/Target/X86/X86SchedSkylakeServer.td
+++ b/llvm/lib/Target/X86/X86SchedSkylakeServer.td
@@ -311,9 +311,10 @@ def : WriteRes<WritePCmpEStrILd, [SKXPort0, SKXPort5, SKXPort23, SKXPort0156]> {
 }
 
 // MOVMSK Instructions.
-def : WriteRes<WriteFMOVMSK, [SKXPort0]> { let Latency = 2; }
-def : WriteRes<WriteVecMOVMSK, [SKXPort0]> { let Latency = 2; }
-def : WriteRes<WriteMMXMOVMSK, [SKXPort0]> { let Latency = 2; }
+def : WriteRes<WriteFMOVMSK,    [SKXPort0]> { let Latency = 2; }
+def : WriteRes<WriteVecMOVMSK,  [SKXPort0]> { let Latency = 2; }
+def : WriteRes<WriteVecMOVMSKY, [SKXPort0]> { let Latency = 2; }
+def : WriteRes<WriteMMXMOVMSK,  [SKXPort0]> { let Latency = 2; }
 
 // AES instructions.
 def : WriteRes<WriteAESDecEnc, [SKXPort0]> { // Decryption, encryption.
diff --git a/llvm/lib/Target/X86/X86Schedule.td b/llvm/lib/Target/X86/X86Schedule.td
index 8c2578b5d3fd..c3bea9c048c5 100644
--- a/llvm/lib/Target/X86/X86Schedule.td
+++ b/llvm/lib/Target/X86/X86Schedule.td
@@ -182,9 +182,10 @@ def  WriteVecExtract : SchedWrite; // Extract vector element to gpr.
 def  WriteVecExtractSt : SchedWrite; // Extract vector element and store.
 
 // MOVMSK operations.
-def WriteFMOVMSK : SchedWrite;
-def WriteVecMOVMSK : SchedWrite;
-def WriteMMXMOVMSK : SchedWrite;
+def WriteFMOVMSK    : SchedWrite;
+def WriteVecMOVMSK  : SchedWrite;
+def WriteVecMOVMSKY : SchedWrite;
+def WriteMMXMOVMSK  : SchedWrite;
 
 // Conversion between integer and float.
 defm WriteCvtF2I : X86SchedWritePair; // Float -> Integer.
diff --git a/llvm/lib/Target/X86/X86ScheduleAtom.td b/llvm/lib/Target/X86/X86ScheduleAtom.td
index 42ba131cce02..6c175d065192 100644
--- a/llvm/lib/Target/X86/X86ScheduleAtom.td
+++ b/llvm/lib/Target/X86/X86ScheduleAtom.td
@@ -308,9 +308,10 @@ defm : AtomWriteResPair<WritePCmpEStrM, [AtomPort01], [AtomPort01]>; // NOTE: Do
 // MOVMSK Instructions.
 ////////////////////////////////////////////////////////////////////////////////
 
-def  : WriteRes<WriteFMOVMSK,   [AtomPort0]> { let Latency = 3; let ResourceCycles = [3]; }
-def  : WriteRes<WriteVecMOVMSK, [AtomPort0]> { let Latency = 3; let ResourceCycles = [3]; }
-def  : WriteRes<WriteMMXMOVMSK, [AtomPort0]> { let Latency = 3; let ResourceCycles = [3]; }
+def : WriteRes<WriteFMOVMSK,    [AtomPort0]> { let Latency = 3; let ResourceCycles = [3]; }
+def : WriteRes<WriteVecMOVMSK,  [AtomPort0]> { let Latency = 3; let ResourceCycles = [3]; }
+def : WriteRes<WriteVecMOVMSKY, [AtomPort0]> { let Latency = 3; let ResourceCycles = [3]; }
+def : WriteRes<WriteMMXMOVMSK,  [AtomPort0]> { let Latency = 3; let ResourceCycles = [3]; }
 
 ////////////////////////////////////////////////////////////////////////////////
 // AES Instructions.
diff --git a/llvm/lib/Target/X86/X86ScheduleBtVer2.td b/llvm/lib/Target/X86/X86ScheduleBtVer2.td
index d6be3199500d..2f11ead4d8ed 100644
--- a/llvm/lib/Target/X86/X86ScheduleBtVer2.td
+++ b/llvm/lib/Target/X86/X86ScheduleBtVer2.td
@@ -460,9 +460,10 @@ defm : JWriteResFpuPair<WritePCmpEStrM, [JFPU1, JSAGU, JLAGU, JVALU, JVALU1, JFP
 // MOVMSK Instructions.
 ////////////////////////////////////////////////////////////////////////////////
 
-def  : WriteRes<WriteFMOVMSK, [JFPU0, JFPA, JALU0]> { let Latency = 3; }
-def  : WriteRes<WriteVecMOVMSK, [JFPU0, JFPA, JALU0]> { let Latency = 3; }
-def  : WriteRes<WriteMMXMOVMSK, [JFPU0, JFPA, JALU0]> { let Latency = 3; }
+def : WriteRes<WriteFMOVMSK,    [JFPU0, JFPA, JALU0]> { let Latency = 3; }
+def : WriteRes<WriteVecMOVMSK,  [JFPU0, JFPA, JALU0]> { let Latency = 3; }
+def : WriteRes<WriteVecMOVMSKY, [JFPU0, JFPA, JALU0]> { let Latency = 3; }
+def : WriteRes<WriteMMXMOVMSK,  [JFPU0, JFPA, JALU0]> { let Latency = 3; }
 
 ////////////////////////////////////////////////////////////////////////////////
 // AES Instructions.
diff --git a/llvm/lib/Target/X86/X86ScheduleSLM.td b/llvm/lib/Target/X86/X86ScheduleSLM.td
index fa3a63a3eb81..300427df6f87 100644
--- a/llvm/lib/Target/X86/X86ScheduleSLM.td
+++ b/llvm/lib/Target/X86/X86ScheduleSLM.td
@@ -255,9 +255,10 @@ def : WriteRes<WritePCmpEStrILd, [SLM_FPC_RSV0, SLM_MEC_RSV]> {
 }
 
 // MOVMSK Instructions.
-def : WriteRes<WriteFMOVMSK, [SLM_FPC_RSV1]> { let Latency = 4; }
-def : WriteRes<WriteVecMOVMSK, [SLM_FPC_RSV1]> { let Latency = 4; }
-def : WriteRes<WriteMMXMOVMSK, [SLM_FPC_RSV1]> { let Latency = 4; }
+def : WriteRes<WriteFMOVMSK,    [SLM_FPC_RSV1]> { let Latency = 4; }
+def : WriteRes<WriteVecMOVMSK,  [SLM_FPC_RSV1]> { let Latency = 4; }
+def : WriteRes<WriteVecMOVMSKY, [SLM_FPC_RSV1]> { let Latency = 4; }
+def : WriteRes<WriteMMXMOVMSK,  [SLM_FPC_RSV1]> { let Latency = 4; }
 
 // AES Instructions.
 def : WriteRes<WriteAESDecEnc, [SLM_FPC_RSV0]> {
diff --git a/llvm/lib/Target/X86/X86ScheduleZnver1.td b/llvm/lib/Target/X86/X86ScheduleZnver1.td
index 8e0cf431be26..fd863a1c1a73 100644
--- a/llvm/lib/Target/X86/X86ScheduleZnver1.td
+++ b/llvm/lib/Target/X86/X86ScheduleZnver1.td
@@ -279,8 +279,14 @@ def : WriteRes<WriteVecExtractSt, [ZnAGU, ZnFPU12, ZnFPU2]> {
 
 // MOVMSK Instructions.
 def : WriteRes<WriteFMOVMSK, [ZnFPU2]>;
-def : WriteRes<WriteVecMOVMSK, [ZnFPU2]>;
 def : WriteRes<WriteMMXMOVMSK, [ZnFPU2]>;
+def : WriteRes<WriteVecMOVMSK, [ZnFPU2]>;
+
+def : WriteRes<WriteVecMOVMSKY, [ZnFPU2]> {
+  let NumMicroOps = 2;
+  let Latency = 2;
+  let ResourceCycles = [2];
+}
 
 // AES Instructions.
 defm : ZnWriteResFpuPair<WriteAESDecEnc, [ZnFPU01], 4>;
@@ -995,14 +1001,6 @@ def : InstRW<[WriteMicrocoded],
 // m, v,v.
 def : InstRW<[WriteMicrocoded], (instregex "VPMASKMOV(D|Q)(Y?)mr")>;
 
-// PMOVMSKBY.
-def ZnWritePMOVMSKBY : SchedWriteRes<[ZnFPU2]> {
-  let NumMicroOps = 2;
-  let Latency = 2;
-  let ResourceCycles = [2];
-}
-def : InstRW<[ZnWritePMOVMSKBY], (instregex "(V|MMX_)?PMOVMSKBYrr")>;
-
 // VPBROADCAST B/W.
 // x, m8/16.
 def ZnWriteVPBROADCAST128Ld : SchedWriteRes<[ZnAGU, ZnFPU12]> {