RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-25 20:23:11 +01:00
Code Issues Projects Releases Wiki Activity
c21336d6b8
llvm-mirror/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp
Dan Gohman 603247a4ff Fix the new scheduler assertion checks to work when 
the scheduler has inserted no-ops. This fixes
the 2006-07-03-schedulers.ll regression on ppc32.

llvm-svn: 49747
2008-04-15 22:40:14 +00:00

1876 lines
63 KiB
C++
Raw Blame History

//===----- ScheduleDAGList.cpp - Reg pressure reduction list scheduler ----===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This implements bottom-up and top-down register pressure reduction list
// schedulers, using standard algorithms. The basic approach uses a priority
// queue of available nodes to schedule. One at a time, nodes are taken from
// the priority queue (thus in priority order), checked for legality to
// schedule, and emitted if legal.
//
//===----------------------------------------------------------------------===//
#define DEBUG_TYPE "pre-RA-sched"
#include "llvm/CodeGen/ScheduleDAG.h"
#include "llvm/CodeGen/SchedulerRegistry.h"
#include "llvm/Target/TargetRegisterInfo.h"
#include "llvm/Target/TargetData.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Target/TargetInstrInfo.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/Compiler.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallSet.h"
#include "llvm/ADT/Statistic.h"
#include <climits>
#include <queue>
#include "llvm/Support/CommandLine.h"
using namespace llvm;
STATISTIC(NumBacktracks, "Number of times scheduler backtracked");
STATISTIC(NumUnfolds, "Number of nodes unfolded");
STATISTIC(NumDups, "Number of duplicated nodes");
STATISTIC(NumCCCopies, "Number of cross class copies");
static RegisterScheduler
burrListDAGScheduler("list-burr",
" Bottom-up register reduction list scheduling",
createBURRListDAGScheduler);
static RegisterScheduler
tdrListrDAGScheduler("list-tdrr",
" Top-down register reduction list scheduling",
createTDRRListDAGScheduler);
namespace {
//===----------------------------------------------------------------------===//
/// ScheduleDAGRRList - The actual register reduction list scheduler
/// implementation. This supports both top-down and bottom-up scheduling.
///
class VISIBILITY_HIDDEN ScheduleDAGRRList : public ScheduleDAG {
private:
/// isBottomUp - This is true if the scheduling problem is bottom-up, false if
/// it is top-down.
bool isBottomUp;
/// AvailableQueue - The priority queue to use for the available SUnits.
SchedulingPriorityQueue *AvailableQueue;
/// LiveRegs / LiveRegDefs - A set of physical registers and their definition
/// that are "live". These nodes must be scheduled before any other nodes that
/// modifies the registers can be scheduled.
SmallSet<unsigned, 4> LiveRegs;
std::vector<SUnit*> LiveRegDefs;
std::vector<unsigned> LiveRegCycles;
public:
ScheduleDAGRRList(SelectionDAG &dag, MachineBasicBlock *bb,
const TargetMachine &tm, bool isbottomup,
SchedulingPriorityQueue *availqueue)
: ScheduleDAG(dag, bb, tm), isBottomUp(isbottomup),
AvailableQueue(availqueue) {
}
~ScheduleDAGRRList() {
delete AvailableQueue;
}
void Schedule();
/// IsReachable - Checks if SU is reachable from TargetSU.
bool IsReachable(SUnit *SU, SUnit *TargetSU);
/// willCreateCycle - Returns true if adding an edge from SU to TargetSU will
/// create a cycle.
bool WillCreateCycle(SUnit *SU, SUnit *TargetSU);
/// AddPred - This adds the specified node X as a predecessor of
/// the current node Y if not already.
/// This returns true if this is a new predecessor.
/// Updates the topological ordering if required.
bool AddPred(SUnit *Y, SUnit *X, bool isCtrl, bool isSpecial,
unsigned PhyReg = 0, int Cost = 1);
/// RemovePred - This removes the specified node N from the predecessors of
/// the current node M. Updates the topological ordering if required.
bool RemovePred(SUnit *M, SUnit *N, bool isCtrl, bool isSpecial);
private:
void ReleasePred(SUnit*, bool, unsigned);
void ReleaseSucc(SUnit*, bool isChain, unsigned);
void CapturePred(SUnit*, SUnit*, bool);
void ScheduleNodeBottomUp(SUnit*, unsigned);
void ScheduleNodeTopDown(SUnit*, unsigned);
void UnscheduleNodeBottomUp(SUnit*);
void BacktrackBottomUp(SUnit*, unsigned, unsigned&);
SUnit *CopyAndMoveSuccessors(SUnit*);
void InsertCCCopiesAndMoveSuccs(SUnit*, unsigned,
const TargetRegisterClass*,
const TargetRegisterClass*,
SmallVector<SUnit*, 2>&);
bool DelayForLiveRegsBottomUp(SUnit*, SmallVector<unsigned, 4>&);
void ListScheduleTopDown();
void ListScheduleBottomUp();
void CommuteNodesToReducePressure();
/// CreateNewSUnit - Creates a new SUnit and returns a pointer to it.
/// Updates the topological ordering if required.
SUnit *CreateNewSUnit(SDNode *N) {
SUnit *NewNode = NewSUnit(N);
// Update the topological ordering.
if (NewNode->NodeNum >= Node2Index.size())
InitDAGTopologicalSorting();
return NewNode;
}
/// CreateClone - Creates a new SUnit from an existing one.
/// Updates the topological ordering if required.
SUnit *CreateClone(SUnit *N) {
SUnit *NewNode = Clone(N);
// Update the topological ordering.
if (NewNode->NodeNum >= Node2Index.size())
InitDAGTopologicalSorting();
return NewNode;
}
/// Functions for preserving the topological ordering
/// even after dynamic insertions of new edges.
/// This allows a very fast implementation of IsReachable.
/**
The idea of the algorithm is taken from
"Online algorithms for managing the topological order of
a directed acyclic graph" by David J. Pearce and Paul H.J. Kelly
This is the MNR algorithm, which was first introduced by
A. Marchetti-Spaccamela, U. Nanni and H. Rohnert in
"Maintaining a topological order under edge insertions".
Short description of the algorithm:
Topological ordering, ord, of a DAG maps each node to a topological
index so that for all edges X->Y it is the case that ord(X) < ord(Y).
This means that if there is a path from the node X to the node Z,
then ord(X) < ord(Z).
This property can be used to check for reachability of nodes:
if Z is reachable from X, then an insertion of the edge Z->X would
create a cycle.
The algorithm first computes a topological ordering for the DAG by initializing
the Index2Node and Node2Index arrays and then tries to keep the ordering
up-to-date after edge insertions by reordering the DAG.
On insertion of the edge X->Y, the algorithm first marks by calling DFS the
nodes reachable from Y, and then shifts them using Shift to lie immediately
after X in Index2Node.
*/
/// InitDAGTopologicalSorting - create the initial topological
/// ordering from the DAG to be scheduled.
void InitDAGTopologicalSorting();
/// DFS - make a DFS traversal and mark all nodes affected by the
/// edge insertion. These nodes will later get new topological indexes
/// by means of the Shift method.
void DFS(SUnit *SU, int UpperBound, bool& HasLoop);
/// Shift - reassign topological indexes for the nodes in the DAG
/// to preserve the topological ordering.
void Shift(BitVector& Visited, int LowerBound, int UpperBound);
/// Allocate - assign the topological index to the node n.
void Allocate(int n, int index);
/// Index2Node - Maps topological index to the node number.
std::vector<int> Index2Node;
/// Node2Index - Maps the node number to its topological index.
std::vector<int> Node2Index;
/// Visited - a set of nodes visited during a DFS traversal.
BitVector Visited;
};
} // end anonymous namespace
/// Schedule - Schedule the DAG using list scheduling.
void ScheduleDAGRRList::Schedule() {
DOUT << "********** List Scheduling **********\n";
LiveRegDefs.resize(TRI->getNumRegs(), NULL);
LiveRegCycles.resize(TRI->getNumRegs(), 0);
// Build scheduling units.
BuildSchedUnits();
DEBUG(for (unsigned su = 0, e = SUnits.size(); su != e; ++su)
SUnits[su].dumpAll(&DAG));
CalculateDepths();
CalculateHeights();
InitDAGTopologicalSorting();
AvailableQueue->initNodes(SUnitMap, SUnits);
// Execute the actual scheduling loop Top-Down or Bottom-Up as appropriate.
if (isBottomUp)
ListScheduleBottomUp();
else
ListScheduleTopDown();
AvailableQueue->releaseState();
CommuteNodesToReducePressure();
DOUT << "*** Final schedule ***\n";
DEBUG(dumpSchedule());
DOUT << "\n";
// Emit in scheduled order
EmitSchedule();
}
/// CommuteNodesToReducePressure - If a node is two-address and commutable, and
/// it is not the last use of its first operand, add it to the CommuteSet if
/// possible. It will be commuted when it is translated to a MI.
void ScheduleDAGRRList::CommuteNodesToReducePressure() {
SmallPtrSet<SUnit*, 4> OperandSeen;
for (unsigned i = Sequence.size(); i != 0; ) {
--i;
SUnit *SU = Sequence[i];
if (!SU || !SU->Node) continue;
if (SU->isCommutable) {
unsigned Opc = SU->Node->getTargetOpcode();
const TargetInstrDesc &TID = TII->get(Opc);
unsigned NumRes = TID.getNumDefs();
unsigned NumOps = TID.getNumOperands() - NumRes;
for (unsigned j = 0; j != NumOps; ++j) {
if (TID.getOperandConstraint(j+NumRes, TOI::TIED_TO) == -1)
continue;
SDNode *OpN = SU->Node->getOperand(j).Val;
SUnit *OpSU = isPassiveNode(OpN) ? NULL : SUnitMap[OpN][SU->InstanceNo];
if (OpSU && OperandSeen.count(OpSU) == 1) {
// Ok, so SU is not the last use of OpSU, but SU is two-address so
// it will clobber OpSU. Try to commute SU if no other source operands
// are live below.
bool DoCommute = true;
for (unsigned k = 0; k < NumOps; ++k) {
if (k != j) {
OpN = SU->Node->getOperand(k).Val;
OpSU = isPassiveNode(OpN) ? NULL : SUnitMap[OpN][SU->InstanceNo];
if (OpSU && OperandSeen.count(OpSU) == 1) {
DoCommute = false;
break;
}
}
}
if (DoCommute)
CommuteSet.insert(SU->Node);
}
// Only look at the first use&def node for now.
break;
}
}
for (SUnit::pred_iterator I = SU->Preds.begin(), E = SU->Preds.end();
I != E; ++I) {
if (!I->isCtrl)
OperandSeen.insert(I->Dep);
}
}
}
//===----------------------------------------------------------------------===//
// Bottom-Up Scheduling
//===----------------------------------------------------------------------===//
/// ReleasePred - Decrement the NumSuccsLeft count of a predecessor. Add it to
/// the AvailableQueue if the count reaches zero. Also update its cycle bound.
void ScheduleDAGRRList::ReleasePred(SUnit *PredSU, bool isChain,
unsigned CurCycle) {
// FIXME: the distance between two nodes is not always == the predecessor's
// latency. For example, the reader can very well read the register written
// by the predecessor later than the issue cycle. It also depends on the
// interrupt model (drain vs. freeze).
PredSU->CycleBound = std::max(PredSU->CycleBound, CurCycle + PredSU->Latency);
--PredSU->NumSuccsLeft;
#ifndef NDEBUG
if (PredSU->NumSuccsLeft < 0) {
cerr << "*** List scheduling failed! ***\n";
PredSU->dump(&DAG);
cerr << " has been released too many times!\n";
assert(0);
}
#endif
if (PredSU->NumSuccsLeft == 0) {
PredSU->isAvailable = true;
AvailableQueue->push(PredSU);
}
}
/// ScheduleNodeBottomUp - Add the node to the schedule. Decrement the pending
/// count of its predecessors. If a predecessor pending count is zero, add it to
/// the Available queue.
void ScheduleDAGRRList::ScheduleNodeBottomUp(SUnit *SU, unsigned CurCycle) {
DOUT << "*** Scheduling [" << CurCycle << "]: ";
DEBUG(SU->dump(&DAG));
SU->Cycle = CurCycle;
AvailableQueue->ScheduledNode(SU);
// Bottom up: release predecessors
for (SUnit::pred_iterator I = SU->Preds.begin(), E = SU->Preds.end();
I != E; ++I) {
ReleasePred(I->Dep, I->isCtrl, CurCycle);
if (I->Cost < 0) {
// This is a physical register dependency and it's impossible or
// expensive to copy the register. Make sure nothing that can
// clobber the register is scheduled between the predecessor and
// this node.
if (LiveRegs.insert(I->Reg)) {
LiveRegDefs[I->Reg] = I->Dep;
LiveRegCycles[I->Reg] = CurCycle;
}
}
}
// Release all the implicit physical register defs that are live.
for (SUnit::succ_iterator I = SU->Succs.begin(), E = SU->Succs.end();
I != E; ++I) {
if (I->Cost < 0) {
if (LiveRegCycles[I->Reg] == I->Dep->Cycle) {
LiveRegs.erase(I->Reg);
assert(LiveRegDefs[I->Reg] == SU &&
"Physical register dependency violated?");
LiveRegDefs[I->Reg] = NULL;
LiveRegCycles[I->Reg] = 0;
}
}
}
SU->isScheduled = true;
}
/// CapturePred - This does the opposite of ReleasePred. Since SU is being
/// unscheduled, incrcease the succ left count of its predecessors. Remove
/// them from AvailableQueue if necessary.
void ScheduleDAGRRList::CapturePred(SUnit *PredSU, SUnit *SU, bool isChain) {
PredSU->CycleBound = 0;
for (SUnit::succ_iterator I = PredSU->Succs.begin(), E = PredSU->Succs.end();
I != E; ++I) {
if (I->Dep == SU)
continue;
PredSU->CycleBound = std::max(PredSU->CycleBound,
I->Dep->Cycle + PredSU->Latency);
}
if (PredSU->isAvailable) {
PredSU->isAvailable = false;
if (!PredSU->isPending)
AvailableQueue->remove(PredSU);
}
++PredSU->NumSuccsLeft;
}
/// UnscheduleNodeBottomUp - Remove the node from the schedule, update its and
/// its predecessor states to reflect the change.
void ScheduleDAGRRList::UnscheduleNodeBottomUp(SUnit *SU) {
DOUT << "*** Unscheduling [" << SU->Cycle << "]: ";
DEBUG(SU->dump(&DAG));
AvailableQueue->UnscheduledNode(SU);
for (SUnit::pred_iterator I = SU->Preds.begin(), E = SU->Preds.end();
I != E; ++I) {
CapturePred(I->Dep, SU, I->isCtrl);
if (I->Cost < 0 && SU->Cycle == LiveRegCycles[I->Reg]) {
LiveRegs.erase(I->Reg);
assert(LiveRegDefs[I->Reg] == I->Dep &&
"Physical register dependency violated?");
LiveRegDefs[I->Reg] = NULL;
LiveRegCycles[I->Reg] = 0;
}
}
for (SUnit::succ_iterator I = SU->Succs.begin(), E = SU->Succs.end();
I != E; ++I) {
if (I->Cost < 0) {
if (LiveRegs.insert(I->Reg)) {
assert(!LiveRegDefs[I->Reg] &&
"Physical register dependency violated?");
LiveRegDefs[I->Reg] = SU;
}
if (I->Dep->Cycle < LiveRegCycles[I->Reg])
LiveRegCycles[I->Reg] = I->Dep->Cycle;
}
}
SU->Cycle = 0;
SU->isScheduled = false;
SU->isAvailable = true;
AvailableQueue->push(SU);
}
/// IsReachable - Checks if SU is reachable from TargetSU.
bool ScheduleDAGRRList::IsReachable(SUnit *SU, SUnit *TargetSU) {
// If insertion of the edge SU->TargetSU would create a cycle
// then there is a path from TargetSU to SU.
int UpperBound, LowerBound;
LowerBound = Node2Index[TargetSU->NodeNum];
UpperBound = Node2Index[SU->NodeNum];
bool HasLoop = false;
// Is Ord(TargetSU) < Ord(SU) ?
if (LowerBound < UpperBound) {
Visited.reset();
// There may be a path from TargetSU to SU. Check for it.
DFS(TargetSU, UpperBound, HasLoop);
}
return HasLoop;
}
/// Allocate - assign the topological index to the node n.
inline void ScheduleDAGRRList::Allocate(int n, int index) {
Node2Index[n] = index;
Index2Node[index] = n;
}
/// InitDAGTopologicalSorting - create the initial topological
/// ordering from the DAG to be scheduled.
void ScheduleDAGRRList::InitDAGTopologicalSorting() {
unsigned DAGSize = SUnits.size();
std::vector<unsigned> InDegree(DAGSize);
std::vector<SUnit*> WorkList;
WorkList.reserve(DAGSize);
std::vector<SUnit*> TopOrder;
TopOrder.reserve(DAGSize);
// Initialize the data structures.
for (unsigned i = 0, e = DAGSize; i != e; ++i) {
SUnit *SU = &SUnits[i];
int NodeNum = SU->NodeNum;
unsigned Degree = SU->Succs.size();
InDegree[NodeNum] = Degree;
// Is it a node without dependencies?
if (Degree == 0) {
assert(SU->Succs.empty() && "SUnit should have no successors");
// Collect leaf nodes.
WorkList.push_back(SU);
}
}
while (!WorkList.empty()) {
SUnit *SU = WorkList.back();
WorkList.pop_back();
TopOrder.push_back(SU);
for (SUnit::const_pred_iterator I = SU->Preds.begin(), E = SU->Preds.end();
I != E; ++I) {
SUnit *SU = I->Dep;
if (!--InDegree[SU->NodeNum])
// If all dependencies of the node are processed already,
// then the node can be computed now.
WorkList.push_back(SU);
}
}
// Second pass, assign the actual topological order as node ids.
int Id = 0;
Index2Node.clear();
Node2Index.clear();
Index2Node.resize(DAGSize);
Node2Index.resize(DAGSize);
Visited.resize(DAGSize);
for (std::vector<SUnit*>::reverse_iterator TI = TopOrder.rbegin(),
TE = TopOrder.rend();TI != TE; ++TI) {
Allocate((*TI)->NodeNum, Id);
Id++;
}
#ifndef NDEBUG
// Check correctness of the ordering
for (unsigned i = 0, e = DAGSize; i != e; ++i) {
SUnit *SU = &SUnits[i];
for (SUnit::const_pred_iterator I = SU->Preds.begin(), E = SU->Preds.end();
I != E; ++I) {
assert(Node2Index[SU->NodeNum] > Node2Index[I->Dep->NodeNum] &&
"Wrong topological sorting");
}
}
#endif
}
/// AddPred - adds an edge from SUnit X to SUnit Y.
/// Updates the topological ordering if required.
bool ScheduleDAGRRList::AddPred(SUnit *Y, SUnit *X, bool isCtrl, bool isSpecial,
unsigned PhyReg, int Cost) {
int UpperBound, LowerBound;
LowerBound = Node2Index[Y->NodeNum];
UpperBound = Node2Index[X->NodeNum];
bool HasLoop = false;
// Is Ord(X) < Ord(Y) ?
if (LowerBound < UpperBound) {
// Update the topological order.
Visited.reset();
DFS(Y, UpperBound, HasLoop);
assert(!HasLoop && "Inserted edge creates a loop!");
// Recompute topological indexes.
Shift(Visited, LowerBound, UpperBound);
}
// Now really insert the edge.
return Y->addPred(X, isCtrl, isSpecial, PhyReg, Cost);
}
/// RemovePred - This removes the specified node N from the predecessors of
/// the current node M. Updates the topological ordering if required.
bool ScheduleDAGRRList::RemovePred(SUnit *M, SUnit *N,
bool isCtrl, bool isSpecial) {
// InitDAGTopologicalSorting();
return M->removePred(N, isCtrl, isSpecial);
}
/// DFS - Make a DFS traversal to mark all nodes reachable from SU and mark
/// all nodes affected by the edge insertion. These nodes will later get new
/// topological indexes by means of the Shift method.
void ScheduleDAGRRList::DFS(SUnit *SU, int UpperBound, bool& HasLoop) {
std::vector<SUnit*> WorkList;
WorkList.reserve(SUnits.size());
WorkList.push_back(SU);
while (!WorkList.empty()) {
SU = WorkList.back();
WorkList.pop_back();
Visited.set(SU->NodeNum);
for (int I = SU->Succs.size()-1; I >= 0; --I) {
int s = SU->Succs[I].Dep->NodeNum;
if (Node2Index[s] == UpperBound) {
HasLoop = true;
return;
}
// Visit successors if not already and in affected region.
if (!Visited.test(s) && Node2Index[s] < UpperBound) {
WorkList.push_back(SU->Succs[I].Dep);
}
}
}
}
/// Shift - Renumber the nodes so that the topological ordering is
/// preserved.
void ScheduleDAGRRList::Shift(BitVector& Visited, int LowerBound,
int UpperBound) {
std::vector<int> L;
int shift = 0;
int i;
for (i = LowerBound; i <= UpperBound; ++i) {
// w is node at topological index i.
int w = Index2Node[i];
if (Visited.test(w)) {
// Unmark.
Visited.reset(w);
L.push_back(w);
shift = shift + 1;
} else {
Allocate(w, i - shift);
}
}
for (unsigned j = 0; j < L.size(); ++j) {
Allocate(L[j], i - shift);
i = i + 1;
}
}
/// WillCreateCycle - Returns true if adding an edge from SU to TargetSU will
/// create a cycle.
bool ScheduleDAGRRList::WillCreateCycle(SUnit *SU, SUnit *TargetSU) {
if (IsReachable(TargetSU, SU))
return true;
for (SUnit::pred_iterator I = SU->Preds.begin(), E = SU->Preds.end();
I != E; ++I)
if (I->Cost < 0 && IsReachable(TargetSU, I->Dep))
return true;
return false;
}
/// BacktrackBottomUp - Backtrack scheduling to a previous cycle specified in
/// BTCycle in order to schedule a specific node. Returns the last unscheduled
/// SUnit. Also returns if a successor is unscheduled in the process.
void ScheduleDAGRRList::BacktrackBottomUp(SUnit *SU, unsigned BtCycle,
unsigned &CurCycle) {
SUnit *OldSU = NULL;
while (CurCycle > BtCycle) {
OldSU = Sequence.back();
Sequence.pop_back();
if (SU->isSucc(OldSU))
// Don't try to remove SU from AvailableQueue.
SU->isAvailable = false;
UnscheduleNodeBottomUp(OldSU);
--CurCycle;
}
if (SU->isSucc(OldSU)) {
assert(false && "Something is wrong!");
abort();
}
++NumBacktracks;
}
/// CopyAndMoveSuccessors - Clone the specified node and move its scheduled
/// successors to the newly created node.
SUnit *ScheduleDAGRRList::CopyAndMoveSuccessors(SUnit *SU) {
if (SU->FlaggedNodes.size())
return NULL;
SDNode *N = SU->Node;
if (!N)
return NULL;
SUnit *NewSU;
bool TryUnfold = false;
for (unsigned i = 0, e = N->getNumValues(); i != e; ++i) {
MVT::ValueType VT = N->getValueType(i);
if (VT == MVT::Flag)
return NULL;
else if (VT == MVT::Other)
TryUnfold = true;
}
for (unsigned i = 0, e = N->getNumOperands(); i != e; ++i) {
const SDOperand &Op = N->getOperand(i);
MVT::ValueType VT = Op.Val->getValueType(Op.ResNo);
if (VT == MVT::Flag)
return NULL;
}
if (TryUnfold) {
SmallVector<SDNode*, 4> NewNodes;
if (!TII->unfoldMemoryOperand(DAG, N, NewNodes))
return NULL;
DOUT << "Unfolding SU # " << SU->NodeNum << "\n";
assert(NewNodes.size() == 2 && "Expected a load folding node!");
N = NewNodes[1];
SDNode *LoadNode = NewNodes[0];
unsigned NumVals = N->getNumValues();
unsigned OldNumVals = SU->Node->getNumValues();
for (unsigned i = 0; i != NumVals; ++i)
DAG.ReplaceAllUsesOfValueWith(SDOperand(SU->Node, i), SDOperand(N, i));
DAG.ReplaceAllUsesOfValueWith(SDOperand(SU->Node, OldNumVals-1),
SDOperand(LoadNode, 1));
SUnit *NewSU = CreateNewSUnit(N);
SUnitMap[N].push_back(NewSU);
const TargetInstrDesc &TID = TII->get(N->getTargetOpcode());
for (unsigned i = 0; i != TID.getNumOperands(); ++i) {
if (TID.getOperandConstraint(i, TOI::TIED_TO) != -1) {
NewSU->isTwoAddress = true;
break;
}
}
if (TID.isCommutable())
NewSU->isCommutable = true;
// FIXME: Calculate height / depth and propagate the changes?
NewSU->Depth = SU->Depth;
NewSU->Height = SU->Height;
ComputeLatency(NewSU);
// LoadNode may already exist. This can happen when there is another
// load from the same location and producing the same type of value
// but it has different alignment or volatileness.
bool isNewLoad = true;
SUnit *LoadSU;
DenseMap<SDNode*, std::vector<SUnit*> >::iterator SMI =
SUnitMap.find(LoadNode);
if (SMI != SUnitMap.end()) {
LoadSU = SMI->second.front();
isNewLoad = false;
} else {
LoadSU = CreateNewSUnit(LoadNode);
SUnitMap[LoadNode].push_back(LoadSU);
LoadSU->Depth = SU->Depth;
LoadSU->Height = SU->Height;
ComputeLatency(LoadSU);
}
SUnit *ChainPred = NULL;
SmallVector<SDep, 4> ChainSuccs;
SmallVector<SDep, 4> LoadPreds;
SmallVector<SDep, 4> NodePreds;
SmallVector<SDep, 4> NodeSuccs;
for (SUnit::pred_iterator I = SU->Preds.begin(), E = SU->Preds.end();
I != E; ++I) {
if (I->isCtrl)
ChainPred = I->Dep;
else if (I->Dep->Node && I->Dep->Node->isOperandOf(LoadNode))
LoadPreds.push_back(SDep(I->Dep, I->Reg, I->Cost, false, false));
else
NodePreds.push_back(SDep(I->Dep, I->Reg, I->Cost, false, false));
}
for (SUnit::succ_iterator I = SU->Succs.begin(), E = SU->Succs.end();
I != E; ++I) {
if (I->isCtrl)
ChainSuccs.push_back(SDep(I->Dep, I->Reg, I->Cost,
I->isCtrl, I->isSpecial));
else
NodeSuccs.push_back(SDep(I->Dep, I->Reg, I->Cost,
I->isCtrl, I->isSpecial));
}
if (ChainPred) {
RemovePred(SU, ChainPred, true, false);
if (isNewLoad)
AddPred(LoadSU, ChainPred, true, false);
}
for (unsigned i = 0, e = LoadPreds.size(); i != e; ++i) {
SDep *Pred = &LoadPreds[i];
RemovePred(SU, Pred->Dep, Pred->isCtrl, Pred->isSpecial);
if (isNewLoad) {
AddPred(LoadSU, Pred->Dep, Pred->isCtrl, Pred->isSpecial,
Pred->Reg, Pred->Cost);
}
}
for (unsigned i = 0, e = NodePreds.size(); i != e; ++i) {
SDep *Pred = &NodePreds[i];
RemovePred(SU, Pred->Dep, Pred->isCtrl, Pred->isSpecial);
AddPred(NewSU, Pred->Dep, Pred->isCtrl, Pred->isSpecial,
Pred->Reg, Pred->Cost);
}
for (unsigned i = 0, e = NodeSuccs.size(); i != e; ++i) {
SDep *Succ = &NodeSuccs[i];
RemovePred(Succ->Dep, SU, Succ->isCtrl, Succ->isSpecial);
AddPred(Succ->Dep, NewSU, Succ->isCtrl, Succ->isSpecial,
Succ->Reg, Succ->Cost);
}
for (unsigned i = 0, e = ChainSuccs.size(); i != e; ++i) {
SDep *Succ = &ChainSuccs[i];
RemovePred(Succ->Dep, SU, Succ->isCtrl, Succ->isSpecial);
if (isNewLoad) {
AddPred(Succ->Dep, LoadSU, Succ->isCtrl, Succ->isSpecial,
Succ->Reg, Succ->Cost);
}
}
if (isNewLoad) {
AddPred(NewSU, LoadSU, false, false);
}
if (isNewLoad)
AvailableQueue->addNode(LoadSU);
AvailableQueue->addNode(NewSU);
++NumUnfolds;
if (NewSU->NumSuccsLeft == 0) {
NewSU->isAvailable = true;
return NewSU;
}
SU = NewSU;
}
DOUT << "Duplicating SU # " << SU->NodeNum << "\n";
NewSU = CreateClone(SU);
// New SUnit has the exact same predecessors.
for (SUnit::pred_iterator I = SU->Preds.begin(), E = SU->Preds.end();
I != E; ++I)
if (!I->isSpecial) {
AddPred(NewSU, I->Dep, I->isCtrl, false, I->Reg, I->Cost);
NewSU->Depth = std::max(NewSU->Depth, I->Dep->Depth+1);
}
// Only copy scheduled successors. Cut them from old node's successor
// list and move them over.
SmallVector<std::pair<SUnit*, bool>, 4> DelDeps;
for (SUnit::succ_iterator I = SU->Succs.begin(), E = SU->Succs.end();
I != E; ++I) {
if (I->isSpecial)
continue;
if (I->Dep->isScheduled) {
NewSU->Height = std::max(NewSU->Height, I->Dep->Height+1);
AddPred(I->Dep, NewSU, I->isCtrl, false, I->Reg, I->Cost);
DelDeps.push_back(std::make_pair(I->Dep, I->isCtrl));
}
}
for (unsigned i = 0, e = DelDeps.size(); i != e; ++i) {
SUnit *Succ = DelDeps[i].first;
bool isCtrl = DelDeps[i].second;
RemovePred(Succ, SU, isCtrl, false);
}
AvailableQueue->updateNode(SU);
AvailableQueue->addNode(NewSU);
++NumDups;
return NewSU;
}
/// InsertCCCopiesAndMoveSuccs - Insert expensive cross register class copies
/// and move all scheduled successors of the given SUnit to the last copy.
void ScheduleDAGRRList::InsertCCCopiesAndMoveSuccs(SUnit *SU, unsigned Reg,
const TargetRegisterClass *DestRC,
const TargetRegisterClass *SrcRC,
SmallVector<SUnit*, 2> &Copies) {
SUnit *CopyFromSU = CreateNewSUnit(NULL);
CopyFromSU->CopySrcRC = SrcRC;
CopyFromSU->CopyDstRC = DestRC;
CopyFromSU->Depth = SU->Depth;
CopyFromSU->Height = SU->Height;
SUnit *CopyToSU = CreateNewSUnit(NULL);
CopyToSU->CopySrcRC = DestRC;
CopyToSU->CopyDstRC = SrcRC;
// Only copy scheduled successors. Cut them from old node's successor
// list and move them over.
SmallVector<std::pair<SUnit*, bool>, 4> DelDeps;
for (SUnit::succ_iterator I = SU->Succs.begin(), E = SU->Succs.end();
I != E; ++I) {
if (I->isSpecial)
continue;
if (I->Dep->isScheduled) {
CopyToSU->Height = std::max(CopyToSU->Height, I->Dep->Height+1);
AddPred(I->Dep, CopyToSU, I->isCtrl, false, I->Reg, I->Cost);
DelDeps.push_back(std::make_pair(I->Dep, I->isCtrl));
}
}
for (unsigned i = 0, e = DelDeps.size(); i != e; ++i) {
SUnit *Succ = DelDeps[i].first;
bool isCtrl = DelDeps[i].second;
RemovePred(Succ, SU, isCtrl, false);
}
AddPred(CopyFromSU, SU, false, false, Reg, -1);
AddPred(CopyToSU, CopyFromSU, false, false, Reg, 1);
AvailableQueue->updateNode(SU);
AvailableQueue->addNode(CopyFromSU);
AvailableQueue->addNode(CopyToSU);
Copies.push_back(CopyFromSU);
Copies.push_back(CopyToSU);
++NumCCCopies;
}
/// getPhysicalRegisterVT - Returns the ValueType of the physical register
/// definition of the specified node.
/// FIXME: Move to SelectionDAG?
static MVT::ValueType getPhysicalRegisterVT(SDNode *N, unsigned Reg,
const TargetInstrInfo *TII) {
const TargetInstrDesc &TID = TII->get(N->getTargetOpcode());
assert(TID.ImplicitDefs && "Physical reg def must be in implicit def list!");
unsigned NumRes = TID.getNumDefs();
for (const unsigned *ImpDef = TID.getImplicitDefs(); *ImpDef; ++ImpDef) {
if (Reg == *ImpDef)
break;
++NumRes;
}
return N->getValueType(NumRes);
}
/// DelayForLiveRegsBottomUp - Returns true if it is necessary to delay
/// scheduling of the given node to satisfy live physical register dependencies.
/// If the specific node is the last one that's available to schedule, do
/// whatever is necessary (i.e. backtracking or cloning) to make it possible.
bool ScheduleDAGRRList::DelayForLiveRegsBottomUp(SUnit *SU,
SmallVector<unsigned, 4> &LRegs){
if (LiveRegs.empty())
return false;
SmallSet<unsigned, 4> RegAdded;
// If this node would clobber any "live" register, then it's not ready.
for (SUnit::pred_iterator I = SU->Preds.begin(), E = SU->Preds.end();
I != E; ++I) {
if (I->Cost < 0) {
unsigned Reg = I->Reg;
if (LiveRegs.count(Reg) && LiveRegDefs[Reg] != I->Dep) {
if (RegAdded.insert(Reg))
LRegs.push_back(Reg);
}
for (const unsigned *Alias = TRI->getAliasSet(Reg);
*Alias; ++Alias)
if (LiveRegs.count(*Alias) && LiveRegDefs[*Alias] != I->Dep) {
if (RegAdded.insert(*Alias))
LRegs.push_back(*Alias);
}
}
}
for (unsigned i = 0, e = SU->FlaggedNodes.size()+1; i != e; ++i) {
SDNode *Node = (i == 0) ? SU->Node : SU->FlaggedNodes[i-1];
if (!Node || !Node->isTargetOpcode())
continue;
const TargetInstrDesc &TID = TII->get(Node->getTargetOpcode());
if (!TID.ImplicitDefs)
continue;
for (const unsigned *Reg = TID.ImplicitDefs; *Reg; ++Reg) {
if (LiveRegs.count(*Reg) && LiveRegDefs[*Reg] != SU) {
if (RegAdded.insert(*Reg))
LRegs.push_back(*Reg);
}
for (const unsigned *Alias = TRI->getAliasSet(*Reg);
*Alias; ++Alias)
if (LiveRegs.count(*Alias) && LiveRegDefs[*Alias] != SU) {
if (RegAdded.insert(*Alias))
LRegs.push_back(*Alias);
}
}
}
return !LRegs.empty();
}
/// ListScheduleBottomUp - The main loop of list scheduling for bottom-up
/// schedulers.
void ScheduleDAGRRList::ListScheduleBottomUp() {
unsigned CurCycle = 0;
// Add root to Available queue.
if (!SUnits.empty()) {
SUnit *RootSU = SUnitMap[DAG.getRoot().Val].front();
assert(RootSU->Succs.empty() && "Graph root shouldn't have successors!");
RootSU->isAvailable = true;
AvailableQueue->push(RootSU);
}
// While Available queue is not empty, grab the node with the highest
// priority. If it is not ready put it back. Schedule the node.
SmallVector<SUnit*, 4> NotReady;
while (!AvailableQueue->empty()) {
bool Delayed = false;
DenseMap<SUnit*, SmallVector<unsigned, 4> > LRegsMap;
SUnit *CurSU = AvailableQueue->pop();
while (CurSU) {
if (CurSU->CycleBound <= CurCycle) {
SmallVector<unsigned, 4> LRegs;
if (!DelayForLiveRegsBottomUp(CurSU, LRegs))
break;
Delayed = true;
LRegsMap.insert(std::make_pair(CurSU, LRegs));
}
CurSU->isPending = true; // This SU is not in AvailableQueue right now.
NotReady.push_back(CurSU);
CurSU = AvailableQueue->pop();
}
// All candidates are delayed due to live physical reg dependencies.
// Try backtracking, code duplication, or inserting cross class copies
// to resolve it.
if (Delayed && !CurSU) {
for (unsigned i = 0, e = NotReady.size(); i != e; ++i) {
SUnit *TrySU = NotReady[i];
SmallVector<unsigned, 4> &LRegs = LRegsMap[TrySU];
// Try unscheduling up to the point where it's safe to schedule
// this node.
unsigned LiveCycle = CurCycle;
for (unsigned j = 0, ee = LRegs.size(); j != ee; ++j) {
unsigned Reg = LRegs[j];
unsigned LCycle = LiveRegCycles[Reg];
LiveCycle = std::min(LiveCycle, LCycle);
}
SUnit *OldSU = Sequence[LiveCycle];
if (!WillCreateCycle(TrySU, OldSU)) {
BacktrackBottomUp(TrySU, LiveCycle, CurCycle);
// Force the current node to be scheduled before the node that
// requires the physical reg dep.
if (OldSU->isAvailable) {
OldSU->isAvailable = false;
AvailableQueue->remove(OldSU);
}
AddPred(TrySU, OldSU, true, true);
// If one or more successors has been unscheduled, then the current
// node is no longer avaialable. Schedule a successor that's now
// available instead.
if (!TrySU->isAvailable)
CurSU = AvailableQueue->pop();
else {
CurSU = TrySU;
TrySU->isPending = false;
NotReady.erase(NotReady.begin()+i);
}
break;
}
}
if (!CurSU) {
// Can't backtrack. Try duplicating the nodes that produces these
// "expensive to copy" values to break the dependency. In case even
// that doesn't work, insert cross class copies.
SUnit *TrySU = NotReady[0];
SmallVector<unsigned, 4> &LRegs = LRegsMap[TrySU];
assert(LRegs.size() == 1 && "Can't handle this yet!");
unsigned Reg = LRegs[0];
SUnit *LRDef = LiveRegDefs[Reg];
SUnit *NewDef = CopyAndMoveSuccessors(LRDef);
if (!NewDef) {
// Issue expensive cross register class copies.
MVT::ValueType VT = getPhysicalRegisterVT(LRDef->Node, Reg, TII);
const TargetRegisterClass *RC =
TRI->getPhysicalRegisterRegClass(Reg, VT);
const TargetRegisterClass *DestRC = TRI->getCrossCopyRegClass(RC);
if (!DestRC) {
assert(false && "Don't know how to copy this physical register!");
abort();
}
SmallVector<SUnit*, 2> Copies;
InsertCCCopiesAndMoveSuccs(LRDef, Reg, DestRC, RC, Copies);
DOUT << "Adding an edge from SU # " << TrySU->NodeNum
<< " to SU #" << Copies.front()->NodeNum << "\n";
AddPred(TrySU, Copies.front(), true, true);
NewDef = Copies.back();
}
DOUT << "Adding an edge from SU # " << NewDef->NodeNum
<< " to SU #" << TrySU->NodeNum << "\n";
LiveRegDefs[Reg] = NewDef;
AddPred(NewDef, TrySU, true, true);
TrySU->isAvailable = false;
CurSU = NewDef;
}
if (!CurSU) {
assert(false && "Unable to resolve live physical register dependencies!");
abort();
}
}
// Add the nodes that aren't ready back onto the available list.
for (unsigned i = 0, e = NotReady.size(); i != e; ++i) {
NotReady[i]->isPending = false;
// May no longer be available due to backtracking.
if (NotReady[i]->isAvailable)
AvailableQueue->push(NotReady[i]);
}
NotReady.clear();
if (!CurSU)
Sequence.push_back(0);
else {
ScheduleNodeBottomUp(CurSU, CurCycle);
Sequence.push_back(CurSU);
}
++CurCycle;
}
// Reverse the order if it is bottom up.
std::reverse(Sequence.begin(), Sequence.end());
#ifndef NDEBUG
// Verify that all SUnits were scheduled.
bool AnyNotSched = false;
unsigned DeadNodes = 0;
unsigned Noops = 0;
for (unsigned i = 0, e = SUnits.size(); i != e; ++i) {
if (!SUnits[i].isScheduled) {
if (SUnits[i].NumPreds == 0 && SUnits[i].NumSuccs == 0) {
++DeadNodes;
continue;
}
if (!AnyNotSched)
cerr << "*** List scheduling failed! ***\n";
SUnits[i].dump(&DAG);
cerr << "has not been scheduled!\n";
AnyNotSched = true;
}
if (SUnits[i].NumSuccsLeft != 0) {
if (!AnyNotSched)
cerr << "*** List scheduling failed! ***\n";
SUnits[i].dump(&DAG);
cerr << "has successors left!\n";
AnyNotSched = true;
}
}
for (unsigned i = 0, e = Sequence.size(); i != e; ++i)
if (!Sequence[i])
++Noops;
assert(!AnyNotSched);
assert(Sequence.size() + DeadNodes - Noops == SUnits.size() &&
"The number of nodes scheduled doesn't match the expected number!");
#endif
}
//===----------------------------------------------------------------------===//
// Top-Down Scheduling
//===----------------------------------------------------------------------===//
/// ReleaseSucc - Decrement the NumPredsLeft count of a successor. Add it to
/// the AvailableQueue if the count reaches zero. Also update its cycle bound.
void ScheduleDAGRRList::ReleaseSucc(SUnit *SuccSU, bool isChain,
unsigned CurCycle) {
// FIXME: the distance between two nodes is not always == the predecessor's
// latency. For example, the reader can very well read the register written
// by the predecessor later than the issue cycle. It also depends on the
// interrupt model (drain vs. freeze).
SuccSU->CycleBound = std::max(SuccSU->CycleBound, CurCycle + SuccSU->Latency);
--SuccSU->NumPredsLeft;
#ifndef NDEBUG
if (SuccSU->NumPredsLeft < 0) {
cerr << "*** List scheduling failed! ***\n";
SuccSU->dump(&DAG);
cerr << " has been released too many times!\n";
assert(0);
}
#endif
if (SuccSU->NumPredsLeft == 0) {
SuccSU->isAvailable = true;
AvailableQueue->push(SuccSU);
}
}
/// ScheduleNodeTopDown - Add the node to the schedule. Decrement the pending
/// count of its successors. If a successor pending count is zero, add it to
/// the Available queue.
void ScheduleDAGRRList::ScheduleNodeTopDown(SUnit *SU, unsigned CurCycle) {
DOUT << "*** Scheduling [" << CurCycle << "]: ";
DEBUG(SU->dump(&DAG));
SU->Cycle = CurCycle;
AvailableQueue->ScheduledNode(SU);
// Top down: release successors
for (SUnit::succ_iterator I = SU->Succs.begin(), E = SU->Succs.end();
I != E; ++I)
ReleaseSucc(I->Dep, I->isCtrl, CurCycle);
SU->isScheduled = true;
}
/// ListScheduleTopDown - The main loop of list scheduling for top-down
/// schedulers.
void ScheduleDAGRRList::ListScheduleTopDown() {
unsigned CurCycle = 0;
// All leaves to Available queue.
for (unsigned i = 0, e = SUnits.size(); i != e; ++i) {
// It is available if it has no predecessors.
if (SUnits[i].Preds.empty()) {
AvailableQueue->push(&SUnits[i]);
SUnits[i].isAvailable = true;
}
}
// While Available queue is not empty, grab the node with the highest
// priority. If it is not ready put it back. Schedule the node.
std::vector<SUnit*> NotReady;
while (!AvailableQueue->empty()) {
SUnit *CurSU = AvailableQueue->pop();
while (CurSU && CurSU->CycleBound > CurCycle) {
NotReady.push_back(CurSU);
CurSU = AvailableQueue->pop();
}
// Add the nodes that aren't ready back onto the available list.
AvailableQueue->push_all(NotReady);
NotReady.clear();
if (!CurSU)
Sequence.push_back(0);
else {
ScheduleNodeTopDown(CurSU, CurCycle);
Sequence.push_back(CurSU);
}
++CurCycle;
}
#ifndef NDEBUG
// Verify that all SUnits were scheduled.
bool AnyNotSched = false;
unsigned DeadNodes = 0;
unsigned Noops = 0;
for (unsigned i = 0, e = SUnits.size(); i != e; ++i) {
if (!SUnits[i].isScheduled) {
if (SUnits[i].NumPreds == 0 && SUnits[i].NumSuccs == 0) {
++DeadNodes;
continue;
}
if (!AnyNotSched)
cerr << "*** List scheduling failed! ***\n";
SUnits[i].dump(&DAG);
cerr << "has not been scheduled!\n";
AnyNotSched = true;
}
if (SUnits[i].NumPredsLeft != 0) {
if (!AnyNotSched)
cerr << "*** List scheduling failed! ***\n";
SUnits[i].dump(&DAG);
cerr << "has predecessors left!\n";
AnyNotSched = true;
}
}
for (unsigned i = 0, e = Sequence.size(); i != e; ++i)
if (!Sequence[i])
++Noops;
assert(!AnyNotSched);
assert(Sequence.size() + DeadNodes - Noops == SUnits.size() &&
"The number of nodes scheduled doesn't match the expected number!");
#endif
}
//===----------------------------------------------------------------------===//
// RegReductionPriorityQueue Implementation
//===----------------------------------------------------------------------===//
//
// This is a SchedulingPriorityQueue that schedules using Sethi Ullman numbers
// to reduce register pressure.
//
namespace {
template<class SF>
class RegReductionPriorityQueue;
/// Sorting functions for the Available queue.
struct bu_ls_rr_sort : public std::binary_function<SUnit*, SUnit*, bool> {
RegReductionPriorityQueue<bu_ls_rr_sort> *SPQ;
bu_ls_rr_sort(RegReductionPriorityQueue<bu_ls_rr_sort> *spq) : SPQ(spq) {}
bu_ls_rr_sort(const bu_ls_rr_sort &RHS) : SPQ(RHS.SPQ) {}
bool operator()(const SUnit* left, const SUnit* right) const;
};
struct td_ls_rr_sort : public std::binary_function<SUnit*, SUnit*, bool> {
RegReductionPriorityQueue<td_ls_rr_sort> *SPQ;
td_ls_rr_sort(RegReductionPriorityQueue<td_ls_rr_sort> *spq) : SPQ(spq) {}
td_ls_rr_sort(const td_ls_rr_sort &RHS) : SPQ(RHS.SPQ) {}
bool operator()(const SUnit* left, const SUnit* right) const;
};
} // end anonymous namespace
static inline bool isCopyFromLiveIn(const SUnit *SU) {
SDNode *N = SU->Node;
return N && N->getOpcode() == ISD::CopyFromReg &&
N->getOperand(N->getNumOperands()-1).getValueType() != MVT::Flag;
}
namespace {
template<class SF>
class VISIBILITY_HIDDEN RegReductionPriorityQueue
: public SchedulingPriorityQueue {
std::priority_queue<SUnit*, std::vector<SUnit*>, SF> Queue;
public:
RegReductionPriorityQueue() :
Queue(SF(this)) {}
virtual void initNodes(DenseMap<SDNode*, std::vector<SUnit*> > &sumap,
std::vector<SUnit> &sunits) {}
virtual void addNode(const SUnit *SU) {}
virtual void updateNode(const SUnit *SU) {}
virtual void releaseState() {}
virtual unsigned getNodePriority(const SUnit *SU) const {
return 0;
}
unsigned size() const { return Queue.size(); }
bool empty() const { return Queue.empty(); }
void push(SUnit *U) {
Queue.push(U);
}
void push_all(const std::vector<SUnit *> &Nodes) {
for (unsigned i = 0, e = Nodes.size(); i != e; ++i)
Queue.push(Nodes[i]);
}
SUnit *pop() {
if (empty()) return NULL;
SUnit *V = Queue.top();
Queue.pop();
return V;
}
/// remove - This is a really inefficient way to remove a node from a
/// priority queue. We should roll our own heap to make this better or
/// something.
void remove(SUnit *SU) {
std::vector<SUnit*> Temp;
assert(!Queue.empty() && "Not in queue!");
while (Queue.top() != SU) {
Temp.push_back(Queue.top());
Queue.pop();
assert(!Queue.empty() && "Not in queue!");
}
// Remove the node from the PQ.
Queue.pop();
// Add all the other nodes back.
for (unsigned i = 0, e = Temp.size(); i != e; ++i)
Queue.push(Temp[i]);
}
};
template<class SF>
class VISIBILITY_HIDDEN BURegReductionPriorityQueue
: public RegReductionPriorityQueue<SF> {
// SUnitMap SDNode to SUnit mapping (n -> n).
DenseMap<SDNode*, std::vector<SUnit*> > *SUnitMap;
// SUnits - The SUnits for the current graph.
const std::vector<SUnit> *SUnits;
// SethiUllmanNumbers - The SethiUllman number for each node.
std::vector<unsigned> SethiUllmanNumbers;
const TargetInstrInfo *TII;
const TargetRegisterInfo *TRI;
ScheduleDAGRRList *scheduleDAG;
public:
explicit BURegReductionPriorityQueue(const TargetInstrInfo *tii,
const TargetRegisterInfo *tri)
: TII(tii), TRI(tri), scheduleDAG(NULL) {}
void initNodes(DenseMap<SDNode*, std::vector<SUnit*> > &sumap,
std::vector<SUnit> &sunits) {
SUnitMap = &sumap;
SUnits = &sunits;
// Add pseudo dependency edges for two-address nodes.
AddPseudoTwoAddrDeps();
// Calculate node priorities.
CalculateSethiUllmanNumbers();
}
void addNode(const SUnit *SU) {
SethiUllmanNumbers.resize(SUnits->size(), 0);
CalcNodeSethiUllmanNumber(SU);
}
void updateNode(const SUnit *SU) {
SethiUllmanNumbers[SU->NodeNum] = 0;
CalcNodeSethiUllmanNumber(SU);
}
void releaseState() {
SUnits = 0;
SethiUllmanNumbers.clear();
}
unsigned getNodePriority(const SUnit *SU) const {
assert(SU->NodeNum < SethiUllmanNumbers.size());
unsigned Opc = SU->Node ? SU->Node->getOpcode() : 0;
if (Opc == ISD::CopyFromReg && !isCopyFromLiveIn(SU))
// CopyFromReg should be close to its def because it restricts
// allocation choices. But if it is a livein then perhaps we want it
// closer to its uses so it can be coalesced.
return 0xffff;
else if (Opc == ISD::TokenFactor || Opc == ISD::CopyToReg)
// CopyToReg should be close to its uses to facilitate coalescing and
// avoid spilling.
return 0;
else if (Opc == TargetInstrInfo::EXTRACT_SUBREG ||
Opc == TargetInstrInfo::INSERT_SUBREG)
// EXTRACT_SUBREG / INSERT_SUBREG should be close to its use to
// facilitate coalescing.
return 0;
else if (SU->NumSuccs == 0)
// If SU does not have a use, i.e. it doesn't produce a value that would
// be consumed (e.g. store), then it terminates a chain of computation.
// Give it a large SethiUllman number so it will be scheduled right
// before its predecessors that it doesn't lengthen their live ranges.
return 0xffff;
else if (SU->NumPreds == 0)
// If SU does not have a def, schedule it close to its uses because it
// does not lengthen any live ranges.
return 0;
else
return SethiUllmanNumbers[SU->NodeNum];
}
void setScheduleDAG(ScheduleDAGRRList *scheduleDag) {
scheduleDAG = scheduleDag;
}
private:
bool canClobber(const SUnit *SU, const SUnit *Op);
void AddPseudoTwoAddrDeps();
void CalculateSethiUllmanNumbers();
unsigned CalcNodeSethiUllmanNumber(const SUnit *SU);
};
template<class SF>
class VISIBILITY_HIDDEN TDRegReductionPriorityQueue
: public RegReductionPriorityQueue<SF> {
// SUnitMap SDNode to SUnit mapping (n -> n).
DenseMap<SDNode*, std::vector<SUnit*> > *SUnitMap;
// SUnits - The SUnits for the current graph.
const std::vector<SUnit> *SUnits;
// SethiUllmanNumbers - The SethiUllman number for each node.
std::vector<unsigned> SethiUllmanNumbers;
public:
TDRegReductionPriorityQueue() {}
void initNodes(DenseMap<SDNode*, std::vector<SUnit*> > &sumap,
std::vector<SUnit> &sunits) {
SUnitMap = &sumap;
SUnits = &sunits;
// Calculate node priorities.
CalculateSethiUllmanNumbers();
}
void addNode(const SUnit *SU) {
SethiUllmanNumbers.resize(SUnits->size(), 0);
CalcNodeSethiUllmanNumber(SU);
}
void updateNode(const SUnit *SU) {
SethiUllmanNumbers[SU->NodeNum] = 0;
CalcNodeSethiUllmanNumber(SU);
}
void releaseState() {
SUnits = 0;
SethiUllmanNumbers.clear();
}
unsigned getNodePriority(const SUnit *SU) const {
assert(SU->NodeNum < SethiUllmanNumbers.size());
return SethiUllmanNumbers[SU->NodeNum];
}
private:
void CalculateSethiUllmanNumbers();
unsigned CalcNodeSethiUllmanNumber(const SUnit *SU);
};
}
/// closestSucc - Returns the scheduled cycle of the successor which is
/// closet to the current cycle.
static unsigned closestSucc(const SUnit *SU) {
unsigned MaxCycle = 0;
for (SUnit::const_succ_iterator I = SU->Succs.begin(), E = SU->Succs.end();
I != E; ++I) {
unsigned Cycle = I->Dep->Cycle;
// If there are bunch of CopyToRegs stacked up, they should be considered
// to be at the same position.
if (I->Dep->Node && I->Dep->Node->getOpcode() == ISD::CopyToReg)
Cycle = closestSucc(I->Dep)+1;
if (Cycle > MaxCycle)
MaxCycle = Cycle;
}
return MaxCycle;
}
/// calcMaxScratches - Returns an cost estimate of the worse case requirement
/// for scratch registers. Live-in operands and live-out results don't count
/// since they are "fixed".
static unsigned calcMaxScratches(const SUnit *SU) {
unsigned Scratches = 0;
for (SUnit::const_pred_iterator I = SU->Preds.begin(), E = SU->Preds.end();
I != E; ++I) {
if (I->isCtrl) continue; // ignore chain preds
if (!I->Dep->Node || I->Dep->Node->getOpcode() != ISD::CopyFromReg)
Scratches++;
}
for (SUnit::const_succ_iterator I = SU->Succs.begin(), E = SU->Succs.end();
I != E; ++I) {
if (I->isCtrl) continue; // ignore chain succs
if (!I->Dep->Node || I->Dep->Node->getOpcode() != ISD::CopyToReg)
Scratches += 10;
}
return Scratches;
}
// Bottom up
bool bu_ls_rr_sort::operator()(const SUnit *left, const SUnit *right) const {
// There used to be a special tie breaker here that looked for
// two-address instructions and preferred the instruction with a
// def&use operand. The special case triggered diagnostics when
// _GLIBCXX_DEBUG was enabled because it broke the strict weak
// ordering that priority_queue requires. It didn't help much anyway
// because AddPseudoTwoAddrDeps already covers many of the cases
// where it would have applied. In addition, it's counter-intuitive
// that a tie breaker would be the first thing attempted. There's a
// "real" tie breaker below that is the operation of last resort.
// The fact that the "special tie breaker" would trigger when there
// wasn't otherwise a tie is what broke the strict weak ordering
// constraint.
unsigned LPriority = SPQ->getNodePriority(left);
unsigned RPriority = SPQ->getNodePriority(right);
if (LPriority != RPriority)
return LPriority > RPriority;
// Try schedule def + use closer when Sethi-Ullman numbers are the same.
// e.g.
// t1 = op t2, c1
// t3 = op t4, c2
//
// and the following instructions are both ready.
// t2 = op c3
// t4 = op c4
//
// Then schedule t2 = op first.
// i.e.
// t4 = op c4
// t2 = op c3
// t1 = op t2, c1
// t3 = op t4, c2
//
// This creates more short live intervals.
unsigned LDist = closestSucc(left);
unsigned RDist = closestSucc(right);
if (LDist != RDist)
return LDist < RDist;
// Intuitively, it's good to push down instructions whose results are
// liveout so their long live ranges won't conflict with other values
// which are needed inside the BB. Further prioritize liveout instructions
// by the number of operands which are calculated within the BB.
unsigned LScratch = calcMaxScratches(left);
unsigned RScratch = calcMaxScratches(right);
if (LScratch != RScratch)
return LScratch > RScratch;
if (left->Height != right->Height)
return left->Height > right->Height;
if (left->Depth != right->Depth)
return left->Depth < right->Depth;
if (left->CycleBound != right->CycleBound)
return left->CycleBound > right->CycleBound;
// FIXME: No strict ordering.
return false;
}
template<class SF> bool
BURegReductionPriorityQueue<SF>::canClobber(const SUnit *SU, const SUnit *Op) {
if (SU->isTwoAddress) {
unsigned Opc = SU->Node->getTargetOpcode();
const TargetInstrDesc &TID = TII->get(Opc);
unsigned NumRes = TID.getNumDefs();
unsigned NumOps = TID.getNumOperands() - NumRes;
for (unsigned i = 0; i != NumOps; ++i) {
if (TID.getOperandConstraint(i+NumRes, TOI::TIED_TO) != -1) {
SDNode *DU = SU->Node->getOperand(i).Val;
if ((*SUnitMap).find(DU) != (*SUnitMap).end() &&
Op == (*SUnitMap)[DU][SU->InstanceNo])
return true;
}
}
}
return false;
}
/// hasCopyToRegUse - Return true if SU has a value successor that is a
/// CopyToReg node.
static bool hasCopyToRegUse(SUnit *SU) {
for (SUnit::const_succ_iterator I = SU->Succs.begin(), E = SU->Succs.end();
I != E; ++I) {
if (I->isCtrl) continue;
SUnit *SuccSU = I->Dep;
if (SuccSU->Node && SuccSU->Node->getOpcode() == ISD::CopyToReg)
return true;
}
return false;
}
/// canClobberPhysRegDefs - True if SU would clobber one of SuccSU's
/// physical register def.
static bool canClobberPhysRegDefs(SUnit *SuccSU, SUnit *SU,
const TargetInstrInfo *TII,
const TargetRegisterInfo *TRI) {
SDNode *N = SuccSU->Node;
unsigned NumDefs = TII->get(N->getTargetOpcode()).getNumDefs();
const unsigned *ImpDefs = TII->get(N->getTargetOpcode()).getImplicitDefs();
if (!ImpDefs)
return false;
const unsigned *SUImpDefs =
TII->get(SU->Node->getTargetOpcode()).getImplicitDefs();
if (!SUImpDefs)
return false;
for (unsigned i = NumDefs, e = N->getNumValues(); i != e; ++i) {
MVT::ValueType VT = N->getValueType(i);
if (VT == MVT::Flag || VT == MVT::Other)
continue;
unsigned Reg = ImpDefs[i - NumDefs];
for (;*SUImpDefs; ++SUImpDefs) {
unsigned SUReg = *SUImpDefs;
if (TRI->regsOverlap(Reg, SUReg))
return true;
}
}
return false;
}
/// AddPseudoTwoAddrDeps - If two nodes share an operand and one of them uses
/// it as a def&use operand. Add a pseudo control edge from it to the other
/// node (if it won't create a cycle) so the two-address one will be scheduled
/// first (lower in the schedule). If both nodes are two-address, favor the
/// one that has a CopyToReg use (more likely to be a loop induction update).
/// If both are two-address, but one is commutable while the other is not
/// commutable, favor the one that's not commutable.
template<class SF>
void BURegReductionPriorityQueue<SF>::AddPseudoTwoAddrDeps() {
for (unsigned i = 0, e = SUnits->size(); i != e; ++i) {
SUnit *SU = (SUnit *)&((*SUnits)[i]);
if (!SU->isTwoAddress)
continue;
SDNode *Node = SU->Node;
if (!Node || !Node->isTargetOpcode() || SU->FlaggedNodes.size() > 0)
continue;
unsigned Opc = Node->getTargetOpcode();
const TargetInstrDesc &TID = TII->get(Opc);
unsigned NumRes = TID.getNumDefs();
unsigned NumOps = TID.getNumOperands() - NumRes;
for (unsigned j = 0; j != NumOps; ++j) {
if (TID.getOperandConstraint(j+NumRes, TOI::TIED_TO) != -1) {
SDNode *DU = SU->Node->getOperand(j).Val;
if ((*SUnitMap).find(DU) == (*SUnitMap).end())
continue;
SUnit *DUSU = (*SUnitMap)[DU][SU->InstanceNo];
if (!DUSU) continue;
for (SUnit::succ_iterator I = DUSU->Succs.begin(),E = DUSU->Succs.end();
I != E; ++I) {
if (I->isCtrl) continue;
SUnit *SuccSU = I->Dep;
if (SuccSU == SU)
continue;
// Be conservative. Ignore if nodes aren't at roughly the same
// depth and height.
if (SuccSU->Height < SU->Height && (SU->Height - SuccSU->Height) > 1)
continue;
if (!SuccSU->Node || !SuccSU->Node->isTargetOpcode())
continue;
// Don't constrain nodes with physical register defs if the
// predecessor can clobber them.
if (SuccSU->hasPhysRegDefs) {
if (canClobberPhysRegDefs(SuccSU, SU, TII, TRI))
continue;
}
// Don't constraint extract_subreg / insert_subreg these may be
// coalesced away. We don't them close to their uses.
unsigned SuccOpc = SuccSU->Node->getTargetOpcode();
if (SuccOpc == TargetInstrInfo::EXTRACT_SUBREG ||
SuccOpc == TargetInstrInfo::INSERT_SUBREG)
continue;
if ((!canClobber(SuccSU, DUSU) ||
(hasCopyToRegUse(SU) && !hasCopyToRegUse(SuccSU)) ||
(!SU->isCommutable && SuccSU->isCommutable)) &&
!scheduleDAG->IsReachable(SuccSU, SU)) {
DOUT << "Adding an edge from SU # " << SU->NodeNum
<< " to SU #" << SuccSU->NodeNum << "\n";
scheduleDAG->AddPred(SU, SuccSU, true, true);
}
}
}
}
}
}
/// CalcNodeSethiUllmanNumber - Priority is the Sethi Ullman number.
/// Smaller number is the higher priority.
template<class SF>
unsigned BURegReductionPriorityQueue<SF>::
CalcNodeSethiUllmanNumber(const SUnit *SU) {
unsigned &SethiUllmanNumber = SethiUllmanNumbers[SU->NodeNum];
if (SethiUllmanNumber != 0)
return SethiUllmanNumber;
unsigned Extra = 0;
for (SUnit::const_pred_iterator I = SU->Preds.begin(), E = SU->Preds.end();
I != E; ++I) {
if (I->isCtrl) continue; // ignore chain preds
SUnit *PredSU = I->Dep;
unsigned PredSethiUllman = CalcNodeSethiUllmanNumber(PredSU);
if (PredSethiUllman > SethiUllmanNumber) {
SethiUllmanNumber = PredSethiUllman;
Extra = 0;
} else if (PredSethiUllman == SethiUllmanNumber && !I->isCtrl)
++Extra;
}
SethiUllmanNumber += Extra;
if (SethiUllmanNumber == 0)
SethiUllmanNumber = 1;
return SethiUllmanNumber;
}
/// CalculateSethiUllmanNumbers - Calculate Sethi-Ullman numbers of all
/// scheduling units.
template<class SF>
void BURegReductionPriorityQueue<SF>::CalculateSethiUllmanNumbers() {
SethiUllmanNumbers.assign(SUnits->size(), 0);
for (unsigned i = 0, e = SUnits->size(); i != e; ++i)
CalcNodeSethiUllmanNumber(&(*SUnits)[i]);
}
/// LimitedSumOfUnscheduledPredsOfSuccs - Compute the sum of the unscheduled
/// predecessors of the successors of the SUnit SU. Stop when the provided
/// limit is exceeded.
static unsigned LimitedSumOfUnscheduledPredsOfSuccs(const SUnit *SU,
unsigned Limit) {
unsigned Sum = 0;
for (SUnit::const_succ_iterator I = SU->Succs.begin(), E = SU->Succs.end();
I != E; ++I) {
SUnit *SuccSU = I->Dep;
for (SUnit::const_pred_iterator II = SuccSU->Preds.begin(),
EE = SuccSU->Preds.end(); II != EE; ++II) {
SUnit *PredSU = II->Dep;
if (!PredSU->isScheduled)
if (++Sum > Limit)
return Sum;
}
}
return Sum;
}
// Top down
bool td_ls_rr_sort::operator()(const SUnit *left, const SUnit *right) const {
unsigned LPriority = SPQ->getNodePriority(left);
unsigned RPriority = SPQ->getNodePriority(right);
bool LIsTarget = left->Node && left->Node->isTargetOpcode();
bool RIsTarget = right->Node && right->Node->isTargetOpcode();
bool LIsFloater = LIsTarget && left->NumPreds == 0;
bool RIsFloater = RIsTarget && right->NumPreds == 0;
unsigned LBonus = (LimitedSumOfUnscheduledPredsOfSuccs(left,1) == 1) ? 2 : 0;
unsigned RBonus = (LimitedSumOfUnscheduledPredsOfSuccs(right,1) == 1) ? 2 : 0;
if (left->NumSuccs == 0 && right->NumSuccs != 0)
return false;
else if (left->NumSuccs != 0 && right->NumSuccs == 0)
return true;
if (LIsFloater)
LBonus -= 2;
if (RIsFloater)
RBonus -= 2;
if (left->NumSuccs == 1)
LBonus += 2;
if (right->NumSuccs == 1)
RBonus += 2;
if (LPriority+LBonus != RPriority+RBonus)
return LPriority+LBonus < RPriority+RBonus;
if (left->Depth != right->Depth)
return left->Depth < right->Depth;
if (left->NumSuccsLeft != right->NumSuccsLeft)
return left->NumSuccsLeft > right->NumSuccsLeft;
if (left->CycleBound != right->CycleBound)
return left->CycleBound > right->CycleBound;
// FIXME: No strict ordering.
return false;
}
/// CalcNodeSethiUllmanNumber - Priority is the Sethi Ullman number.
/// Smaller number is the higher priority.
template<class SF>
unsigned TDRegReductionPriorityQueue<SF>::
CalcNodeSethiUllmanNumber(const SUnit *SU) {
unsigned &SethiUllmanNumber = SethiUllmanNumbers[SU->NodeNum];
if (SethiUllmanNumber != 0)
return SethiUllmanNumber;
unsigned Opc = SU->Node ? SU->Node->getOpcode() : 0;
if (Opc == ISD::TokenFactor || Opc == ISD::CopyToReg)
SethiUllmanNumber = 0xffff;
else if (SU->NumSuccsLeft == 0)
// If SU does not have a use, i.e. it doesn't produce a value that would
// be consumed (e.g. store), then it terminates a chain of computation.
// Give it a small SethiUllman number so it will be scheduled right before
// its predecessors that it doesn't lengthen their live ranges.
SethiUllmanNumber = 0;
else if (SU->NumPredsLeft == 0 &&
(Opc != ISD::CopyFromReg || isCopyFromLiveIn(SU)))
SethiUllmanNumber = 0xffff;
else {
int Extra = 0;
for (SUnit::const_pred_iterator I = SU->Preds.begin(), E = SU->Preds.end();
I != E; ++I) {
if (I->isCtrl) continue; // ignore chain preds
SUnit *PredSU = I->Dep;
unsigned PredSethiUllman = CalcNodeSethiUllmanNumber(PredSU);
if (PredSethiUllman > SethiUllmanNumber) {
SethiUllmanNumber = PredSethiUllman;
Extra = 0;
} else if (PredSethiUllman == SethiUllmanNumber && !I->isCtrl)
++Extra;
}
SethiUllmanNumber += Extra;
}
return SethiUllmanNumber;
}
/// CalculateSethiUllmanNumbers - Calculate Sethi-Ullman numbers of all
/// scheduling units.
template<class SF>
void TDRegReductionPriorityQueue<SF>::CalculateSethiUllmanNumbers() {
SethiUllmanNumbers.assign(SUnits->size(), 0);
for (unsigned i = 0, e = SUnits->size(); i != e; ++i)
CalcNodeSethiUllmanNumber(&(*SUnits)[i]);
}
//===----------------------------------------------------------------------===//
// Public Constructor Functions
//===----------------------------------------------------------------------===//
llvm::ScheduleDAG* llvm::createBURRListDAGScheduler(SelectionDAGISel *IS,
SelectionDAG *DAG,
MachineBasicBlock *BB) {
const TargetInstrInfo *TII = DAG->getTarget().getInstrInfo();
const TargetRegisterInfo *TRI = DAG->getTarget().getRegisterInfo();
BURegReductionPriorityQueue<bu_ls_rr_sort> *priorityQueue =
new BURegReductionPriorityQueue<bu_ls_rr_sort>(TII, TRI);
ScheduleDAGRRList * scheduleDAG =
new ScheduleDAGRRList(*DAG, BB, DAG->getTarget(), true, priorityQueue);
priorityQueue->setScheduleDAG(scheduleDAG);
return scheduleDAG;
}
llvm::ScheduleDAG* llvm::createTDRRListDAGScheduler(SelectionDAGISel *IS,
SelectionDAG *DAG,
MachineBasicBlock *BB) {
return new ScheduleDAGRRList(*DAG, BB, DAG->getTarget(), false,
new TDRegReductionPriorityQueue<td_ls_rr_sort>());
}
Reference in New Issue View Git Blame Copy Permalink