ARM sched model: Add divsion, loads, branches, vfp cvt

Add some generic SchedWrites and assign resources for Swift and Cortex A9.

Reapply of r183257. (Removed empty InstRW for division on swift)

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@183319 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Target/ARM/ARMInstrVFP.td

@@ -870,7 +870,8 @@ let Constraints = "$a = $dst" in {
 class AVConv1XInsS_Encode<bits<5> op1, bits<2> op2, bits<4> op3, bits<4> op4,
                           bit op5, dag oops, dag iops, InstrItinClass itin,
                           string opc, string asm, list<dag> pattern>
-  : AVConv1XI<op1, op2, op3, op4, op5, oops, iops, itin, opc, asm, pattern> {
+  : AVConv1XI<op1, op2, op3, op4, op5, oops, iops, itin, opc, asm, pattern>,
+  Sched<[WriteCvtFP]> {
   bits<5> dst;
   // if dp_operation then UInt(D:Vd) else UInt(Vd:D);
   let Inst{22} = dst{0};
@@ -881,7 +882,8 @@ class AVConv1XInsS_Encode<bits<5> op1, bits<2> op2, bits<4> op3, bits<4> op4,
 class AVConv1XInsD_Encode<bits<5> op1, bits<2> op2, bits<4> op3, bits<4> op4,
                           bit op5, dag oops, dag iops, InstrItinClass itin,
                           string opc, string asm, list<dag> pattern>
-  : AVConv1XI<op1, op2, op3, op4, op5, oops, iops, itin, opc, asm, pattern> {
+  : AVConv1XI<op1, op2, op3, op4, op5, oops, iops, itin, opc, asm, pattern>,
+    Sched<[WriteCvtFP]> {
   bits<5> dst;
   // if dp_operation then UInt(D:Vd) else UInt(Vd:D);
   let Inst{22} = dst{4};

lib/Target/ARM/ARMSchedule.td

@@ -69,6 +69,21 @@ def WriteCMP : SchedWrite;
 def WriteCMPsi : SchedWrite;
 def WriteCMPsr : SchedWrite;
 
+// Division.
+def WriteDiv : SchedWrite;
+
+// Loads.
+def WriteLd : SchedWrite;
+def WritePreLd : SchedWrite;
+
+// Branches.
+def WriteBr : SchedWrite;
+def WriteBrL : SchedWrite;
+def WriteBrTbl : SchedWrite;
+
+// Fixpoint conversions.
+def WriteCvtFP : SchedWrite;
+
 // Define TII for use in SchedVariant Predicates.
 def : PredicateProlog<[{
   const ARMBaseInstrInfo *TII =

lib/Target/ARM/ARMScheduleA9.td

@@ -2493,4 +2493,11 @@ def : SchedAlias<ReadALUsr, A9ReadALU>;
 def : SchedAlias<WriteCMP, A9WriteALU>;
 def : SchedAlias<WriteCMPsi, A9WriteALU>;
 def : SchedAlias<WriteCMPsr, A9WriteALU>;
+def : WriteRes<WriteDiv, []> { let Latency = 0; }
+
+def : WriteRes<WriteBr, [A9UnitB]>;
+def : WriteRes<WriteBrL, [A9UnitB]>;
+def : WriteRes<WriteBrTbl, [A9UnitB]>;
+def : WriteRes<WritePreLd, []>;
+def : SchedAlias<WriteCvtFP, A9WriteF>;
 } // SchedModel = CortexA9Model

lib/Target/ARM/ARMScheduleSwift.td

@@ -1096,9 +1096,27 @@ let SchedModel = SwiftModel in {
   def SwiftUnitDiv : ProcResource<1>;
 
   // Generic resource requirements.
+  def SwiftWriteP0OneCycle : SchedWriteRes<[SwiftUnitP0]>;
+  def SwiftWriteP0TwoCycle : SchedWriteRes<[SwiftUnitP0]> { let Latency = 2; }
+  def SwiftWriteP0FourCycle : SchedWriteRes<[SwiftUnitP0]> { let Latency = 4; }
+  def SwiftWriteP0SixCycle : SchedWriteRes<[SwiftUnitP0]> { let Latency = 6; }
+  def SwiftWriteP0P1FourCycle : SchedWriteRes<[SwiftUnitP0, SwiftUnitP1]> {
+    let Latency = 4;
+  }
+  def SwiftWriteP0P1SixCycle : SchedWriteRes<[SwiftUnitP0, SwiftUnitP1]> {
+    let Latency = 6;
+  }
+  def SwiftWriteP01OneCycle : SchedWriteRes<[SwiftUnitP01]>;
+  def SwiftWriteP1TwoCycle : SchedWriteRes<[SwiftUnitP1]> { let Latency = 2; }
+  def SwiftWriteP1FourCycle : SchedWriteRes<[SwiftUnitP1]> { let Latency = 4; }
+  def SwiftWriteP1SixCycle : SchedWriteRes<[SwiftUnitP1]> { let Latency = 6; }
+  def SwiftWriteP1EightCycle : SchedWriteRes<[SwiftUnitP1]> { let Latency = 8; }
+  def SwiftWriteP1TwelveCyc : SchedWriteRes<[SwiftUnitP1]> { let Latency = 12; }
+  def SwiftWriteP01OneCycle2x : WriteSequence<[SwiftWriteP01OneCycle], 2>;
+  def SwiftWriteP01OneCycle3x : WriteSequence<[SwiftWriteP01OneCycle], 3>;
   def SwiftWriteP01TwoCycle : SchedWriteRes<[SwiftUnitP01]> { let Latency = 2; }
-  def SwiftWriteP01ThreeCycleTwoUops :
-    SchedWriteRes<[SwiftUnitP01, SwiftUnitP01]> {
+  def SwiftWriteP01ThreeCycleTwoUops : SchedWriteRes<[SwiftUnitP01,
+                                                      SwiftUnitP01]> {
     let Latency = 3;
     let NumMicroOps = 2;
   }
@@ -1107,7 +1125,23 @@ let SchedModel = SwiftModel in {
     let NumMicroOps = 3;
     let ResourceCycles = [3];
   }
-
+  // Plain load without writeback.
+  def SwiftWriteP2ThreeCycle : SchedWriteRes<[SwiftUnitP2]> {
+    let Latency = 3;
+  }
+  def SwiftWriteP2FourCycle : SchedWriteRes<[SwiftUnitP2]> {
+    let Latency = 4;
+  }
+  // A store does not write to a register.
+  def SwiftWriteP2 : SchedWriteRes<[SwiftUnitP2]> {
+    let Latency = 0;
+  }
+  foreach Num = 1-4 in {
+    def SwiftWrite#Num#xP2 : WriteSequence<[SwiftWriteP2], Num>;
+  }
+  def SwiftWriteP01OneCycle2x_load : WriteSequence<[SwiftWriteP01OneCycle,
+                                                    SwiftWriteP01OneCycle,
+                                                    SwiftWriteP2ThreeCycle]>;
   // 4.2.4 Arithmetic and Logical.
   // ALU operation register shifted by immediate variant.
   def SwiftWriteALUsi : SchedWriteVariant<[
@@ -1137,8 +1171,32 @@ let SchedModel = SwiftModel in {
   def : ReadAdvance<ReadALU, 0>;
   def : SchedAlias<ReadALUsr, SwiftReadAdvanceALUsr>;
 
+
+  def SwiftChooseShiftKindP01OneOrTwoCycle : SchedWriteVariant<[
+    SchedVar<IsFastImmShiftSwiftPred, [SwiftWriteP01OneCycle]>,
+    SchedVar<NoSchedPred,             [SwiftWriteP01TwoCycle]>
+  ]>;
+
   // 4.2.5 Integer comparison
   def : WriteRes<WriteCMP, [SwiftUnitP01]>;
-  def : WriteRes<WriteCMPsi, [SwiftUnitP01]>;
-  def : WriteRes<WriteCMPsr, [SwiftUnitP01]>;
+  def : SchedAlias<WriteCMPsi, SwiftChooseShiftKindP01OneOrTwoCycle>;
+  def : SchedAlias<WriteCMPsr, SwiftWriteP01TwoCycle>;
+  // 4.2.18 Integer Divide
+  def : WriteRes<WriteDiv, [SwiftUnitDiv]>; // Workaround.
+  def : InstRW < [WriteDiv],
+        (instregex "SDIV", "UDIV", "t2SDIV", "t2UDIV")>;
+  // 4.2.26 Branch
+  def : WriteRes<WriteBr, [SwiftUnitP1]> { let Latency = 0; }
+  def : WriteRes<WriteBrL, [SwiftUnitP1]> { let Latency = 2; }
+  def : WriteRes<WriteBrTbl, [SwiftUnitP1, SwiftUnitP2]> { let Latency = 0; }
+
+  // 4.2.36 Advanced SIMD and VFP, Convert
+  def : InstRW<[SwiftWriteP1FourCycle], (instregex "VCVT", "V(S|U)IT", "VTO(S|U)")>;
+  // Fixpoint conversions.
+  def : WriteRes<WriteCvtFP, [SwiftUnitP1]> { let Latency = 4; }
+  // Preload.
+  def : WriteRes<WritePreLd, [SwiftUnitP2]> { let Latency = 0;
+    let ResourceCycles = [0];
+  }
+
 }