[CodeGen] Use ProcResGroup information in SchedBoundary

When the ProcResGroup has BufferSize=0,

1. if there is a subunit in the list of write resources for the
   scheduling class, do not attempt to schedule the ProcResGroup.
2. if there is not a subunit in the list of write resources for the
   scheduling class, choose a subunit to use instead of the ProcResGroup.
3. having both the ProcResGroup and any of its subunits in the resources
   implied by a InstRW is not supported.

Used to model parallel uses from a pool of resources.

Differential Revision: https://reviews.llvm.org/D98976

Author: dpenry

llvm/include/llvm/CodeGen/MachineScheduler.h

@@ -74,6 +74,7 @@
 #ifndef LLVM_CODEGEN_MACHINESCHEDULER_H
 #define LLVM_CODEGEN_MACHINESCHEDULER_H
 
+#include "llvm/ADT/APInt.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/BitVector.h"
 #include "llvm/ADT/STLExtras.h"
@@ -674,6 +675,9 @@ class SchedBoundary {
   // it.
   SmallVector<unsigned, 16> ReservedCyclesIndex;
 
+  // For each PIdx, stores the resource group IDs of its subunits
+  SmallVector<APInt, 16> ResourceGroupSubUnitMasks;
+
 #ifndef NDEBUG
   // Remember the greatest possible stall as an upper bound on the number of
   // times we should retry the pending queue because of a hazard.
@@ -751,9 +755,15 @@ class SchedBoundary {
   unsigned getNextResourceCycleByInstance(unsigned InstanceIndex,
                                           unsigned Cycles);
 
-  std::pair<unsigned, unsigned> getNextResourceCycle(unsigned PIdx,
+  std::pair<unsigned, unsigned> getNextResourceCycle(const MCSchedClassDesc *SC,
+                                                     unsigned PIdx,
                                                      unsigned Cycles);
 
+  bool isUnbufferedGroup(unsigned PIdx) const {
+    return SchedModel->getProcResource(PIdx)->SubUnitsIdxBegin &&
+           !SchedModel->getProcResource(PIdx)->BufferSize;
+  }
+
   bool checkHazard(SUnit *SU);
 
   unsigned findMaxLatency(ArrayRef<SUnit*> ReadySUs);
@@ -775,7 +785,8 @@ class SchedBoundary {
 
   void incExecutedResources(unsigned PIdx, unsigned Count);
 
-  unsigned countResource(unsigned PIdx, unsigned Cycles, unsigned ReadyCycle);
+  unsigned countResource(const MCSchedClassDesc *SC, unsigned PIdx,
+                         unsigned Cycles, unsigned ReadyCycle);
 
   void bumpNode(SUnit *SU);
 

llvm/lib/CodeGen/MachineScheduler.cpp

@@ -2004,6 +2004,7 @@ void SchedBoundary::reset() {
   IsResourceLimited = false;
   ReservedCycles.clear();
   ReservedCyclesIndex.clear();
+  ResourceGroupSubUnitMasks.clear();
 #ifndef NDEBUG
   // Track the maximum number of stall cycles that could arise either from the
   // latency of a DAG edge or the number of cycles that a processor resource is
@@ -2045,11 +2046,18 @@ init(ScheduleDAGMI *dag, const TargetSchedModel *smodel, SchedRemainder *rem) {
     unsigned ResourceCount = SchedModel->getNumProcResourceKinds();
     ReservedCyclesIndex.resize(ResourceCount);
     ExecutedResCounts.resize(ResourceCount);
+    ResourceGroupSubUnitMasks.resize(ResourceCount, APInt(ResourceCount, 0));
     unsigned NumUnits = 0;
 
     for (unsigned i = 0; i < ResourceCount; ++i) {
       ReservedCyclesIndex[i] = NumUnits;
       NumUnits += SchedModel->getProcResource(i)->NumUnits;
+      if (isUnbufferedGroup(i)) {
+        auto SubUnits = SchedModel->getProcResource(i)->SubUnitsIdxBegin;
+        for (unsigned U = 0, UE = SchedModel->getProcResource(i)->NumUnits;
+             U != UE; ++U)
+          ResourceGroupSubUnitMasks[i].setBit(SubUnits[U]);
+      }
     }
 
     ReservedCycles.resize(NumUnits, InvalidCycle);
@@ -2091,14 +2099,45 @@ unsigned SchedBoundary::getNextResourceCycleByInstance(unsigned InstanceIdx,
 /// scheduled.  Returns the next cycle and the index of the processor resource
 /// instance in the reserved cycles vector.
 std::pair<unsigned, unsigned>
-SchedBoundary::getNextResourceCycle(unsigned PIdx, unsigned Cycles) {
+SchedBoundary::getNextResourceCycle(const MCSchedClassDesc *SC, unsigned PIdx,
+                                    unsigned Cycles) {
+
   unsigned MinNextUnreserved = InvalidCycle;
   unsigned InstanceIdx = 0;
   unsigned StartIndex = ReservedCyclesIndex[PIdx];
   unsigned NumberOfInstances = SchedModel->getProcResource(PIdx)->NumUnits;
   assert(NumberOfInstances > 0 &&
          "Cannot have zero instances of a ProcResource");
 
+  if (isUnbufferedGroup(PIdx)) {
+    // If any subunits are used by the instruction, report that the resource
+    // group is available at 0, effectively removing the group record from
+    // hazarding and basing the hazarding decisions on the subunit records.
+    // Otherwise, choose the first available instance from among the subunits.
+    // Specifications which assign cycles to both the subunits and the group or
+    // which use an unbuffered group with buffered subunits will appear to
+    // schedule strangely. In the first case, the additional cycles for the
+    // group will be ignored.  In the second, the group will be ignored
+    // entirely.
+    for (const MCWriteProcResEntry &PE :
+         make_range(SchedModel->getWriteProcResBegin(SC),
+                    SchedModel->getWriteProcResEnd(SC)))
+      if (ResourceGroupSubUnitMasks[PIdx][PE.ProcResourceIdx])
+        return std::make_pair(0u, StartIndex);
+
+    auto SubUnits = SchedModel->getProcResource(PIdx)->SubUnitsIdxBegin;
+    for (unsigned I = 0, End = NumberOfInstances; I < End; ++I) {
+      unsigned NextUnreserved, NextInstanceIdx;
+      std::tie(NextUnreserved, NextInstanceIdx) =
+          getNextResourceCycle(SC, SubUnits[I], Cycles);
+      if (MinNextUnreserved > NextUnreserved) {
+        InstanceIdx = NextInstanceIdx;
+        MinNextUnreserved = NextUnreserved;
+      }
+    }
+    return std::make_pair(MinNextUnreserved, InstanceIdx);
+  }
+
   for (unsigned I = StartIndex, End = StartIndex + NumberOfInstances; I < End;
        ++I) {
     unsigned NextUnreserved = getNextResourceCycleByInstance(I, Cycles);
@@ -2152,7 +2191,7 @@ bool SchedBoundary::checkHazard(SUnit *SU) {
       unsigned ResIdx = PE.ProcResourceIdx;
       unsigned Cycles = PE.Cycles;
       unsigned NRCycle, InstanceIdx;
-      std::tie(NRCycle, InstanceIdx) = getNextResourceCycle(ResIdx, Cycles);
+      std::tie(NRCycle, InstanceIdx) = getNextResourceCycle(SC, ResIdx, Cycles);
       if (NRCycle > CurrCycle) {
 #ifndef NDEBUG
         MaxObservedStall = std::max(Cycles, MaxObservedStall);
@@ -2302,8 +2341,8 @@ void SchedBoundary::incExecutedResources(unsigned PIdx, unsigned Count) {
 ///
 /// \return the next cycle at which the instruction may execute without
 /// oversubscribing resources.
-unsigned SchedBoundary::
-countResource(unsigned PIdx, unsigned Cycles, unsigned NextCycle) {
+unsigned SchedBoundary::countResource(const MCSchedClassDesc *SC, unsigned PIdx,
+                                      unsigned Cycles, unsigned NextCycle) {
   unsigned Factor = SchedModel->getResourceFactor(PIdx);
   unsigned Count = Factor * Cycles;
   LLVM_DEBUG(dbgs() << "  " << SchedModel->getResourceName(PIdx) << " +"
@@ -2325,7 +2364,7 @@ countResource(unsigned PIdx, unsigned Cycles, unsigned NextCycle) {
   }
   // For reserved resources, record the highest cycle using the resource.
   unsigned NextAvailable, InstanceIdx;
-  std::tie(NextAvailable, InstanceIdx) = getNextResourceCycle(PIdx, Cycles);
+  std::tie(NextAvailable, InstanceIdx) = getNextResourceCycle(SC, PIdx, Cycles);
   if (NextAvailable > CurrCycle) {
     LLVM_DEBUG(dbgs() << "  Resource conflict: "
                       << SchedModel->getResourceName(PIdx)
@@ -2405,7 +2444,7 @@ void SchedBoundary::bumpNode(SUnit *SU) {
            PI = SchedModel->getWriteProcResBegin(SC),
            PE = SchedModel->getWriteProcResEnd(SC); PI != PE; ++PI) {
       unsigned RCycle =
-        countResource(PI->ProcResourceIdx, PI->Cycles, NextCycle);
+        countResource(SC, PI->ProcResourceIdx, PI->Cycles, NextCycle);
       if (RCycle > NextCycle)
         NextCycle = RCycle;
     }
@@ -2420,7 +2459,8 @@ void SchedBoundary::bumpNode(SUnit *SU) {
         unsigned PIdx = PI->ProcResourceIdx;
         if (SchedModel->getProcResource(PIdx)->BufferSize == 0) {
           unsigned ReservedUntil, InstanceIdx;
-          std::tie(ReservedUntil, InstanceIdx) = getNextResourceCycle(PIdx, 0);
+          std::tie(ReservedUntil, InstanceIdx) =
+              getNextResourceCycle(SC, PIdx, 0);
           if (isTop()) {
             ReservedCycles[InstanceIdx] =
                 std::max(ReservedUntil, NextCycle + PI->Cycles);

llvm/test/CodeGen/ARM/cortex-m7-wideops.mir

@@ -22,9 +22,9 @@ body:             |
     ; CHECK-LABEL: name: test_groups
     ; CHECK: liveins: $d0, $r0, $r1, $r2, $r3, $r4
     ; CHECK: renamable $d0 = VADDD killed renamable $d0, renamable $d0, 14 /* CC::al */, $noreg
+    ; CHECK: renamable $r3 = t2ADDrr killed renamable $r3, renamable $r3, 14 /* CC::al */, $noreg, $noreg
     ; CHECK: renamable $s2 = VLDRS killed renamable $r0, 0, 14 /* CC::al */, $noreg
     ; CHECK: VSTRS killed renamable $s2, killed renamable $r1, 0, 14 /* CC::al */, $noreg
-    ; CHECK: renamable $r3 = t2ADDrr killed renamable $r3, renamable $r3, 14 /* CC::al */, $noreg, $noreg
     ; CHECK: t2STRi12 killed renamable $r3, killed renamable $r2, 0, 14 /* CC::al */, $noreg
     ; CHECK: renamable $r4 = t2ADDrr killed renamable $r4, renamable $r4, 14 /* CC::al */, $noreg, $noreg
     ; CHECK: tBX_RET 14 /* CC::al */, $noreg, implicit killed $d0