llvm-6502/lib/CodeGen/InstrSched/SchedGraph.cpp

1047 lines
35 KiB
C++
Raw Normal View History

// $Id$
//***************************************************************************
// File:
// SchedGraph.cpp
//
// Purpose:
// Scheduling graph based on SSA graph plus extra dependence edges
// capturing dependences due to machine resources (machine registers,
// CC registers, and any others).
//
// History:
// 7/20/01 - Vikram Adve - Created
//**************************************************************************/
#include "SchedGraph.h"
#include "llvm/InstrTypes.h"
#include "llvm/Instruction.h"
#include "llvm/BasicBlock.h"
#include "llvm/Method.h"
#include "llvm/CodeGen/MachineInstr.h"
#include "llvm/CodeGen/InstrSelection.h"
#include "llvm/Target/MachineInstrInfo.h"
#include "llvm/Target/MachineRegInfo.h"
#include "llvm/iOther.h"
#include "Support/StringExtras.h"
#include "Support/STLExtras.h"
#include <algorithm>
#include <vector>
#include <iostream>
#include <ext/hash_map>
using std::vector;
using std::pair;
using std::hash_map;
using std::cerr;
//*********************** Internal Data Structures *************************/
// The following two types need to be classes, not typedefs, so we can use
// opaque declarations in SchedGraph.h
//
struct RefVec: public vector< pair<SchedGraphNode*, int> > {
typedef vector< pair<SchedGraphNode*, int> >:: iterator iterator;
typedef vector< pair<SchedGraphNode*, int> >::const_iterator const_iterator;
};
struct RegToRefVecMap: public hash_map<int, RefVec> {
typedef hash_map<int, RefVec>:: iterator iterator;
typedef hash_map<int, RefVec>::const_iterator const_iterator;
};
struct ValueToDefVecMap: public hash_map<const Instruction*, RefVec> {
typedef hash_map<const Instruction*, RefVec>:: iterator iterator;
typedef hash_map<const Instruction*, RefVec>::const_iterator const_iterator;
};
//
// class SchedGraphEdge
//
/*ctor*/
SchedGraphEdge::SchedGraphEdge(SchedGraphNode* _src,
SchedGraphNode* _sink,
SchedGraphEdgeDepType _depType,
unsigned int _depOrderType,
int _minDelay)
: src(_src),
sink(_sink),
depType(_depType),
depOrderType(_depOrderType),
minDelay((_minDelay >= 0)? _minDelay : _src->getLatency()),
val(NULL)
{
assert(src != sink && "Self-loop in scheduling graph!");
src->addOutEdge(this);
sink->addInEdge(this);
}
/*ctor*/
SchedGraphEdge::SchedGraphEdge(SchedGraphNode* _src,
SchedGraphNode* _sink,
const Value* _val,
unsigned int _depOrderType,
int _minDelay)
: src(_src),
sink(_sink),
depType(ValueDep),
depOrderType(_depOrderType),
minDelay((_minDelay >= 0)? _minDelay : _src->getLatency()),
val(_val)
{
assert(src != sink && "Self-loop in scheduling graph!");
src->addOutEdge(this);
sink->addInEdge(this);
}
/*ctor*/
SchedGraphEdge::SchedGraphEdge(SchedGraphNode* _src,
SchedGraphNode* _sink,
unsigned int _regNum,
unsigned int _depOrderType,
int _minDelay)
: src(_src),
sink(_sink),
depType(MachineRegister),
depOrderType(_depOrderType),
minDelay((_minDelay >= 0)? _minDelay : _src->getLatency()),
machineRegNum(_regNum)
{
assert(src != sink && "Self-loop in scheduling graph!");
src->addOutEdge(this);
sink->addInEdge(this);
}
/*ctor*/
SchedGraphEdge::SchedGraphEdge(SchedGraphNode* _src,
SchedGraphNode* _sink,
ResourceId _resourceId,
int _minDelay)
: src(_src),
sink(_sink),
depType(MachineResource),
depOrderType(NonDataDep),
minDelay((_minDelay >= 0)? _minDelay : _src->getLatency()),
resourceId(_resourceId)
{
assert(src != sink && "Self-loop in scheduling graph!");
src->addOutEdge(this);
sink->addInEdge(this);
}
/*dtor*/
SchedGraphEdge::~SchedGraphEdge()
{
}
void SchedGraphEdge::dump(int indent=0) const {
cerr << std::string(indent*2, ' ') << *this;
}
//
// class SchedGraphNode
//
/*ctor*/
SchedGraphNode::SchedGraphNode(unsigned int _nodeId,
const BasicBlock* _bb,
const MachineInstr* _minstr,
int indexInBB,
const TargetMachine& target)
: nodeId(_nodeId),
bb(_bb),
minstr(_minstr),
origIndexInBB(indexInBB),
latency(0)
{
if (minstr)
{
MachineOpCode mopCode = minstr->getOpCode();
latency = target.getInstrInfo().hasResultInterlock(mopCode)
? target.getInstrInfo().minLatency(mopCode)
: target.getInstrInfo().maxLatency(mopCode);
}
}
/*dtor*/
SchedGraphNode::~SchedGraphNode()
{
}
void SchedGraphNode::dump(int indent=0) const {
cerr << std::string(indent*2, ' ') << *this;
}
inline void
SchedGraphNode::addInEdge(SchedGraphEdge* edge)
{
inEdges.push_back(edge);
}
inline void
SchedGraphNode::addOutEdge(SchedGraphEdge* edge)
{
outEdges.push_back(edge);
}
inline void
SchedGraphNode::removeInEdge(const SchedGraphEdge* edge)
{
assert(edge->getSink() == this);
for (iterator I = beginInEdges(); I != endInEdges(); ++I)
if ((*I) == edge)
{
inEdges.erase(I);
break;
}
}
inline void
SchedGraphNode::removeOutEdge(const SchedGraphEdge* edge)
{
assert(edge->getSrc() == this);
for (iterator I = beginOutEdges(); I != endOutEdges(); ++I)
if ((*I) == edge)
{
outEdges.erase(I);
break;
}
}
//
// class SchedGraph
//
/*ctor*/
SchedGraph::SchedGraph(const BasicBlock* bb,
const TargetMachine& target)
{
bbVec.push_back(bb);
buildGraph(target);
}
/*dtor*/
SchedGraph::~SchedGraph()
{
for (const_iterator I = begin(); I != end(); ++I)
{
SchedGraphNode *node = I->second;
// for each node, delete its out-edges
std::for_each(node->beginOutEdges(), node->endOutEdges(),
deleter<SchedGraphEdge>);
// then delete the node itself.
delete node;
}
}
void
SchedGraph::dump() const
{
cerr << " Sched Graph for Basic Blocks: ";
for (unsigned i=0, N=bbVec.size(); i < N; i++)
{
cerr << (bbVec[i]->hasName()? bbVec[i]->getName() : "block")
<< " (" << bbVec[i] << ")"
<< ((i == N-1)? "" : ", ");
}
cerr << "\n\n Actual Root nodes : ";
for (unsigned i=0, N=graphRoot->outEdges.size(); i < N; i++)
cerr << graphRoot->outEdges[i]->getSink()->getNodeId()
<< ((i == N-1)? "" : ", ");
cerr << "\n Graph Nodes:\n";
for (const_iterator I=begin(); I != end(); ++I)
cerr << "\n" << *I->second;
cerr << "\n";
}
void
SchedGraph::eraseIncomingEdges(SchedGraphNode* node, bool addDummyEdges)
{
// Delete and disconnect all in-edges for the node
for (SchedGraphNode::iterator I = node->beginInEdges();
I != node->endInEdges(); ++I)
{
SchedGraphNode* srcNode = (*I)->getSrc();
srcNode->removeOutEdge(*I);
delete *I;
if (addDummyEdges &&
srcNode != getRoot() &&
srcNode->beginOutEdges() == srcNode->endOutEdges())
{ // srcNode has no more out edges, so add an edge to dummy EXIT node
assert(node != getLeaf() && "Adding edge that was just removed?");
(void) new SchedGraphEdge(srcNode, getLeaf(),
SchedGraphEdge::CtrlDep, SchedGraphEdge::NonDataDep, 0);
}
}
node->inEdges.clear();
}
void
SchedGraph::eraseOutgoingEdges(SchedGraphNode* node, bool addDummyEdges)
{
// Delete and disconnect all out-edges for the node
for (SchedGraphNode::iterator I = node->beginOutEdges();
I != node->endOutEdges(); ++I)
{
SchedGraphNode* sinkNode = (*I)->getSink();
sinkNode->removeInEdge(*I);
delete *I;
if (addDummyEdges &&
sinkNode != getLeaf() &&
sinkNode->beginInEdges() == sinkNode->endInEdges())
{ //sinkNode has no more in edges, so add an edge from dummy ENTRY node
assert(node != getRoot() && "Adding edge that was just removed?");
(void) new SchedGraphEdge(getRoot(), sinkNode,
SchedGraphEdge::CtrlDep, SchedGraphEdge::NonDataDep, 0);
}
}
node->outEdges.clear();
}
void
SchedGraph::eraseIncidentEdges(SchedGraphNode* node, bool addDummyEdges)
{
this->eraseIncomingEdges(node, addDummyEdges);
this->eraseOutgoingEdges(node, addDummyEdges);
}
void
SchedGraph::addDummyEdges()
{
assert(graphRoot->outEdges.size() == 0);
for (const_iterator I=begin(); I != end(); ++I)
{
SchedGraphNode* node = (*I).second;
assert(node != graphRoot && node != graphLeaf);
if (node->beginInEdges() == node->endInEdges())
(void) new SchedGraphEdge(graphRoot, node, SchedGraphEdge::CtrlDep,
SchedGraphEdge::NonDataDep, 0);
if (node->beginOutEdges() == node->endOutEdges())
(void) new SchedGraphEdge(node, graphLeaf, SchedGraphEdge::CtrlDep,
SchedGraphEdge::NonDataDep, 0);
}
}
void
SchedGraph::addCDEdges(const TerminatorInst* term,
const TargetMachine& target)
{
const MachineInstrInfo& mii = target.getInstrInfo();
MachineCodeForVMInstr& termMvec = term->getMachineInstrVec();
// Find the first branch instr in the sequence of machine instrs for term
//
unsigned first = 0;
while (! mii.isBranch(termMvec[first]->getOpCode()))
++first;
assert(first < termMvec.size() &&
"No branch instructions for BR? Ok, but weird! Delete assertion.");
if (first == termMvec.size())
return;
SchedGraphNode* firstBrNode = this->getGraphNodeForInstr(termMvec[first]);
// Add CD edges from each instruction in the sequence to the
// *last preceding* branch instr. in the sequence
// Use a latency of 0 because we only need to prevent out-of-order issue.
//
for (int i = (int) termMvec.size()-1; i > (int) first; i--)
{
SchedGraphNode* toNode = this->getGraphNodeForInstr(termMvec[i]);
assert(toNode && "No node for instr generated for branch?");
for (int j = i-1; j >= 0; j--)
if (mii.isBranch(termMvec[j]->getOpCode()))
{
SchedGraphNode* brNode = this->getGraphNodeForInstr(termMvec[j]);
assert(brNode && "No node for instr generated for branch?");
(void) new SchedGraphEdge(brNode, toNode, SchedGraphEdge::CtrlDep,
SchedGraphEdge::NonDataDep, 0);
break; // only one incoming edge is enough
}
}
// Add CD edges from each instruction preceding the first branch
// to the first branch. Use a latency of 0 as above.
//
for (int i = first-1; i >= 0; i--)
{
SchedGraphNode* fromNode = this->getGraphNodeForInstr(termMvec[i]);
assert(fromNode && "No node for instr generated for branch?");
(void) new SchedGraphEdge(fromNode, firstBrNode, SchedGraphEdge::CtrlDep,
SchedGraphEdge::NonDataDep, 0);
}
// Now add CD edges to the first branch instruction in the sequence from
// all preceding instructions in the basic block. Use 0 latency again.
//
const BasicBlock* bb = firstBrNode->getBB();
const MachineCodeForBasicBlock& mvec = bb->getMachineInstrVec();
for (unsigned i=0, N=mvec.size(); i < N; i++)
{
if (mvec[i] == termMvec[first]) // reached the first branch
break;
SchedGraphNode* fromNode = this->getGraphNodeForInstr(mvec[i]);
if (fromNode == NULL)
continue; // dummy instruction, e.g., PHI
(void) new SchedGraphEdge(fromNode, firstBrNode,
SchedGraphEdge::CtrlDep,
SchedGraphEdge::NonDataDep, 0);
// If we find any other machine instructions (other than due to
// the terminator) that also have delay slots, add an outgoing edge
// from the instruction to the instructions in the delay slots.
//
unsigned d = mii.getNumDelaySlots(mvec[i]->getOpCode());
assert(i+d < N && "Insufficient delay slots for instruction?");
for (unsigned j=1; j <= d; j++)
{
SchedGraphNode* toNode = this->getGraphNodeForInstr(mvec[i+j]);
assert(toNode && "No node for machine instr in delay slot?");
(void) new SchedGraphEdge(fromNode, toNode,
SchedGraphEdge::CtrlDep,
SchedGraphEdge::NonDataDep, 0);
}
}
}
static const int SG_LOAD_REF = 0;
static const int SG_STORE_REF = 1;
static const int SG_CALL_REF = 2;
static const unsigned int SG_DepOrderArray[][3] = {
{ SchedGraphEdge::NonDataDep,
SchedGraphEdge::AntiDep,
SchedGraphEdge::AntiDep },
{ SchedGraphEdge::TrueDep,
SchedGraphEdge::OutputDep,
SchedGraphEdge::TrueDep | SchedGraphEdge::OutputDep },
{ SchedGraphEdge::TrueDep,
SchedGraphEdge::AntiDep | SchedGraphEdge::OutputDep,
SchedGraphEdge::TrueDep | SchedGraphEdge::AntiDep
| SchedGraphEdge::OutputDep }
};
// Add a dependence edge between every pair of machine load/store/call
// instructions, where at least one is a store or a call.
// Use latency 1 just to ensure that memory operations are ordered;
// latency does not otherwise matter (true dependences enforce that).
//
void
SchedGraph::addMemEdges(const vector<SchedGraphNode*>& memNodeVec,
const TargetMachine& target)
{
const MachineInstrInfo& mii = target.getInstrInfo();
// Instructions in memNodeVec are in execution order within the basic block,
// so simply look at all pairs <memNodeVec[i], memNodeVec[j: j > i]>.
//
for (unsigned im=0, NM=memNodeVec.size(); im < NM; im++)
{
MachineOpCode fromOpCode = memNodeVec[im]->getOpCode();
int fromType = mii.isCall(fromOpCode)? SG_CALL_REF
: mii.isLoad(fromOpCode)? SG_LOAD_REF
: SG_STORE_REF;
for (unsigned jm=im+1; jm < NM; jm++)
{
MachineOpCode toOpCode = memNodeVec[jm]->getOpCode();
int toType = mii.isCall(toOpCode)? SG_CALL_REF
: mii.isLoad(toOpCode)? SG_LOAD_REF
: SG_STORE_REF;
if (fromType != SG_LOAD_REF || toType != SG_LOAD_REF)
(void) new SchedGraphEdge(memNodeVec[im], memNodeVec[jm],
SchedGraphEdge::MemoryDep,
SG_DepOrderArray[fromType][toType], 1);
}
}
}
// Add edges from/to CC reg instrs to/from call instrs.
// Essentially this prevents anything that sets or uses a CC reg from being
// reordered w.r.t. a call.
// Use a latency of 0 because we only need to prevent out-of-order issue,
// like with control dependences.
//
void
SchedGraph::addCallCCEdges(const vector<SchedGraphNode*>& memNodeVec,
MachineCodeForBasicBlock& bbMvec,
const TargetMachine& target)
{
const MachineInstrInfo& mii = target.getInstrInfo();
vector<SchedGraphNode*> callNodeVec;
// Find the call instruction nodes and put them in a vector.
for (unsigned im=0, NM=memNodeVec.size(); im < NM; im++)
if (mii.isCall(memNodeVec[im]->getOpCode()))
callNodeVec.push_back(memNodeVec[im]);
// Now walk the entire basic block, looking for CC instructions *and*
// call instructions, and keep track of the order of the instructions.
// Use the call node vec to quickly find earlier and later call nodes
// relative to the current CC instruction.
//
int lastCallNodeIdx = -1;
for (unsigned i=0, N=bbMvec.size(); i < N; i++)
if (mii.isCall(bbMvec[i]->getOpCode()))
{
++lastCallNodeIdx;
for ( ; lastCallNodeIdx < (int)callNodeVec.size(); ++lastCallNodeIdx)
if (callNodeVec[lastCallNodeIdx]->getMachineInstr() == bbMvec[i])
break;
assert(lastCallNodeIdx < (int)callNodeVec.size() && "Missed Call?");
}
else if (mii.isCCInstr(bbMvec[i]->getOpCode()))
{ // Add incoming/outgoing edges from/to preceding/later calls
SchedGraphNode* ccNode = this->getGraphNodeForInstr(bbMvec[i]);
int j=0;
for ( ; j <= lastCallNodeIdx; j++)
(void) new SchedGraphEdge(callNodeVec[j], ccNode,
MachineCCRegsRID, 0);
for ( ; j < (int) callNodeVec.size(); j++)
(void) new SchedGraphEdge(ccNode, callNodeVec[j],
MachineCCRegsRID, 0);
}
}
void
SchedGraph::addMachineRegEdges(RegToRefVecMap& regToRefVecMap,
const TargetMachine& target)
{
assert(bbVec.size() == 1 && "Only handling a single basic block here");
// This assumes that such hardwired registers are never allocated
// to any LLVM value (since register allocation happens later), i.e.,
// any uses or defs of this register have been made explicit!
// Also assumes that two registers with different numbers are
// not aliased!
//
for (RegToRefVecMap::iterator I = regToRefVecMap.begin();
I != regToRefVecMap.end(); ++I)
{
int regNum = (*I).first;
RefVec& regRefVec = (*I).second;
// regRefVec is ordered by control flow order in the basic block
for (unsigned i=0; i < regRefVec.size(); ++i)
{
SchedGraphNode* node = regRefVec[i].first;
unsigned int opNum = regRefVec[i].second;
bool isDef = node->getMachineInstr()->operandIsDefined(opNum);
for (unsigned p=0; p < i; ++p)
{
SchedGraphNode* prevNode = regRefVec[p].first;
if (prevNode != node)
{
unsigned int prevOpNum = regRefVec[p].second;
bool prevIsDef =
prevNode->getMachineInstr()->operandIsDefined(prevOpNum);
if (isDef)
new SchedGraphEdge(prevNode, node, regNum,
(prevIsDef)? SchedGraphEdge::OutputDep
: SchedGraphEdge::AntiDep);
else if (prevIsDef)
new SchedGraphEdge(prevNode, node, regNum,
SchedGraphEdge::TrueDep);
}
}
}
}
}
void
SchedGraph::addEdgesForValue(SchedGraphNode* refNode,
const RefVec& defVec,
const Value* defValue,
bool refNodeIsDef,
const TargetMachine& target)
{
// Add true or output dep edges from all def nodes before refNode in BB.
// Add anti or output dep edges to all def nodes after refNode.
for (RefVec::const_iterator I=defVec.begin(), E=defVec.end(); I != E; ++I)
{
if ((*I).first == refNode)
continue; // Dont add any self-loops
if ((*I).first->getOrigIndexInBB() < refNode->getOrigIndexInBB())
// (*).first is before refNode
(void) new SchedGraphEdge((*I).first, refNode, defValue,
(refNodeIsDef)? SchedGraphEdge::OutputDep
: SchedGraphEdge::TrueDep);
else
// (*).first is after refNode
(void) new SchedGraphEdge(refNode, (*I).first, defValue,
(refNodeIsDef)? SchedGraphEdge::OutputDep
: SchedGraphEdge::AntiDep);
}
}
void
SchedGraph::addEdgesForInstruction(const MachineInstr& minstr,
const ValueToDefVecMap& valueToDefVecMap,
const TargetMachine& target)
{
SchedGraphNode* node = this->getGraphNodeForInstr(&minstr);
if (node == NULL)
return;
// Add edges for all operands of the machine instruction.
//
for (unsigned i=0, numOps=minstr.getNumOperands(); i < numOps; i++)
{
const MachineOperand& mop = minstr.getOperand(i);
switch(mop.getOperandType())
{
case MachineOperand::MO_VirtualRegister:
case MachineOperand::MO_CCRegister:
if (const Instruction* srcI =
dyn_cast_or_null<Instruction>(mop.getVRegValue()))
{
ValueToDefVecMap::const_iterator I = valueToDefVecMap.find(srcI);
if (I != valueToDefVecMap.end())
addEdgesForValue(node, (*I).second, mop.getVRegValue(),
minstr.operandIsDefined(i), target);
}
break;
case MachineOperand::MO_MachineRegister:
break;
case MachineOperand::MO_SignExtendedImmed:
case MachineOperand::MO_UnextendedImmed:
case MachineOperand::MO_PCRelativeDisp:
break; // nothing to do for immediate fields
default:
assert(0 && "Unknown machine operand type in SchedGraph builder");
break;
}
}
// Add edges for values implicitly used by the machine instruction.
// Examples include function arguments to a Call instructions or the return
// value of a Ret instruction.
//
for (unsigned i=0, N=minstr.getNumImplicitRefs(); i < N; ++i)
if (! minstr.implicitRefIsDefined(i))
if (const Instruction* srcI =
dyn_cast_or_null<Instruction>(minstr.getImplicitRef(i)))
{
ValueToDefVecMap::const_iterator I = valueToDefVecMap.find(srcI);
if (I != valueToDefVecMap.end())
addEdgesForValue(node, (*I).second, minstr.getImplicitRef(i),
minstr.implicitRefIsDefined(i), target);
}
}
#undef NEED_SEPARATE_NONSSA_EDGES_CODE
#ifdef NEED_SEPARATE_NONSSA_EDGES_CODE
void
SchedGraph::addNonSSAEdgesForValue(const Instruction* instr,
const TargetMachine& target)
{
if (isa<PHINode>(instr))
return;
MachineCodeForVMInstr& mvec = instr->getMachineInstrVec();
const MachineInstrInfo& mii = target.getInstrInfo();
RefVec refVec;
for (unsigned i=0, N=mvec.size(); i < N; i++)
for (int o=0, N = mii.getNumOperands(mvec[i]->getOpCode()); o < N; o++)
{
const MachineOperand& mop = mvec[i]->getOperand(o);
if ((mop.getOperandType() == MachineOperand::MO_VirtualRegister ||
mop.getOperandType() == MachineOperand::MO_CCRegister)
&& mop.getVRegValue() == (Value*) instr)
{
// this operand is a definition or use of value `instr'
SchedGraphNode* node = this->getGraphNodeForInstr(mvec[i]);
assert(node && "No node for machine instruction in this BB?");
refVec.push_back(std::make_pair(node, o));
}
}
// refVec is ordered by control flow order of the machine instructions
for (unsigned i=0; i < refVec.size(); ++i)
{
SchedGraphNode* node = refVec[i].first;
unsigned int opNum = refVec[i].second;
bool isDef = node->getMachineInstr()->operandIsDefined(opNum);
if (isDef)
// add output and/or anti deps to this definition
for (unsigned p=0; p < i; ++p)
{
SchedGraphNode* prevNode = refVec[p].first;
if (prevNode != node)
{
bool prevIsDef = prevNode->getMachineInstr()->
operandIsDefined(refVec[p].second);
new SchedGraphEdge(prevNode, node, SchedGraphEdge::ValueDep,
(prevIsDef)? SchedGraphEdge::OutputDep
: SchedGraphEdge::AntiDep);
}
}
}
}
#endif //NEED_SEPARATE_NONSSA_EDGES_CODE
void
SchedGraph::findDefUseInfoAtInstr(const TargetMachine& target,
SchedGraphNode* node,
vector<SchedGraphNode*>& memNodeVec,
RegToRefVecMap& regToRefVecMap,
ValueToDefVecMap& valueToDefVecMap)
{
const MachineInstrInfo& mii = target.getInstrInfo();
MachineOpCode opCode = node->getOpCode();
if (mii.isLoad(opCode) || mii.isStore(opCode) || mii.isCall(opCode))
memNodeVec.push_back(node);
// Collect the register references and value defs. for explicit operands
//
const MachineInstr& minstr = * node->getMachineInstr();
for (int i=0, numOps = (int) minstr.getNumOperands(); i < numOps; i++)
{
const MachineOperand& mop = minstr.getOperand(i);
// if this references a register other than the hardwired
// "zero" register, record the reference.
if (mop.getOperandType() == MachineOperand::MO_MachineRegister)
{
int regNum = mop.getMachineRegNum();
if (regNum != target.getRegInfo().getZeroRegNum())
regToRefVecMap[mop.getMachineRegNum()].push_back(
std::make_pair(node, i));
continue; // nothing more to do
}
// ignore all other non-def operands
if (! minstr.operandIsDefined(i))
continue;
// We must be defining a value.
assert((mop.getOperandType() == MachineOperand::MO_VirtualRegister ||
mop.getOperandType() == MachineOperand::MO_CCRegister)
&& "Do not expect any other kind of operand to be defined!");
const Instruction* defInstr = cast<Instruction>(mop.getVRegValue());
valueToDefVecMap[defInstr].push_back(std::make_pair(node, i));
}
//
// Collect value defs. for implicit operands. The interface to extract
// them assumes they must be virtual registers!
//
for (int i=0, N = (int) minstr.getNumImplicitRefs(); i < N; ++i)
if (minstr.implicitRefIsDefined(i))
if (const Instruction* defInstr =
dyn_cast_or_null<Instruction>(minstr.getImplicitRef(i)))
{
valueToDefVecMap[defInstr].push_back(std::make_pair(node, -i));
}
}
void
SchedGraph::buildNodesforBB(const TargetMachine& target,
const BasicBlock* bb,
vector<SchedGraphNode*>& memNodeVec,
RegToRefVecMap& regToRefVecMap,
ValueToDefVecMap& valueToDefVecMap)
{
const MachineInstrInfo& mii = target.getInstrInfo();
// Build graph nodes for each VM instruction and gather def/use info.
// Do both those together in a single pass over all machine instructions.
const MachineCodeForBasicBlock& mvec = bb->getMachineInstrVec();
for (unsigned i=0; i < mvec.size(); i++)
if (! mii.isDummyPhiInstr(mvec[i]->getOpCode()))
{
SchedGraphNode* node = new SchedGraphNode(getNumNodes(), bb,
mvec[i], i, target);
this->noteGraphNodeForInstr(mvec[i], node);
// Remember all register references and value defs
findDefUseInfoAtInstr(target, node,
memNodeVec, regToRefVecMap,valueToDefVecMap);
}
#undef REALLY_NEED_TO_SEARCH_SUCCESSOR_PHIS
#ifdef REALLY_NEED_TO_SEARCH_SUCCESSOR_PHIS
// This is a BIG UGLY HACK. IT NEEDS TO BE ELIMINATED.
// Look for copy instructions inserted in this BB due to Phi instructions
// in the successor BBs.
// There MUST be exactly one copy per Phi in successor nodes.
//
for (BasicBlock::succ_const_iterator SI=bb->succ_begin(), SE=bb->succ_end();
SI != SE; ++SI)
for (BasicBlock::const_iterator PI=(*SI)->begin(), PE=(*SI)->end();
PI != PE; ++PI)
{
if ((*PI)->getOpcode() != Instruction::PHINode)
break; // No more Phis in this successor
// Find the incoming value from block bb to block (*SI)
int bbIndex = cast<PHINode>(*PI)->getBasicBlockIndex(bb);
assert(bbIndex >= 0 && "But I know bb is a predecessor of (*SI)?");
Value* inVal = cast<PHINode>(*PI)->getIncomingValue(bbIndex);
assert(inVal != NULL && "There must be an in-value on every edge");
// Find the machine instruction that makes a copy of inval to (*PI).
// This must be in the current basic block (bb).
const MachineCodeForVMInstr& mvec = (*PI)->getMachineInstrVec();
const MachineInstr* theCopy = NULL;
for (unsigned i=0; i < mvec.size() && theCopy == NULL; i++)
if (! mii.isDummyPhiInstr(mvec[i]->getOpCode()))
// not a Phi: assume this is a copy and examine its operands
for (int o=0, N=(int) mvec[i]->getNumOperands(); o < N; o++)
{
const MachineOperand& mop = mvec[i]->getOperand(o);
if (mvec[i]->operandIsDefined(o))
assert(mop.getVRegValue() == (*PI) && "dest shd be my Phi");
else if (mop.getVRegValue() == inVal)
{ // found the copy!
theCopy = mvec[i];
break;
}
}
// Found the dang instruction. Now create a node and do the rest...
if (theCopy != NULL)
{
SchedGraphNode* node = new SchedGraphNode(getNumNodes(), bb,
theCopy, origIndexInBB++, target);
this->noteGraphNodeForInstr(theCopy, node);
findDefUseInfoAtInstr(target, node,
memNodeVec, regToRefVecMap,valueToDefVecMap);
}
}
#endif //REALLY_NEED_TO_SEARCH_SUCCESSOR_PHIS
}
void
SchedGraph::buildGraph(const TargetMachine& target)
{
const BasicBlock* bb = bbVec[0];
assert(bbVec.size() == 1 && "Only handling a single basic block here");
// Use this data structure to note all machine operands that compute
// ordinary LLVM values. These must be computed defs (i.e., instructions).
// Note that there may be multiple machine instructions that define
// each Value.
ValueToDefVecMap valueToDefVecMap;
// Use this data structure to note all memory instructions.
// We use this to add memory dependence edges without a second full walk.
//
// vector<const Instruction*> memVec;
vector<SchedGraphNode*> memNodeVec;
// Use this data structure to note any uses or definitions of
// machine registers so we can add edges for those later without
// extra passes over the nodes.
// The vector holds an ordered list of references to the machine reg,
// ordered according to control-flow order. This only works for a
// single basic block, hence the assertion. Each reference is identified
// by the pair: <node, operand-number>.
//
RegToRefVecMap regToRefVecMap;
// Make a dummy root node. We'll add edges to the real roots later.
graphRoot = new SchedGraphNode(0, NULL, NULL, -1, target);
graphLeaf = new SchedGraphNode(1, NULL, NULL, -1, target);
//----------------------------------------------------------------
// First add nodes for all the machine instructions in the basic block
// because this greatly simplifies identifying which edges to add.
// Do this one VM instruction at a time since the SchedGraphNode needs that.
// Also, remember the load/store instructions to add memory deps later.
//----------------------------------------------------------------
buildNodesforBB(target, bb, memNodeVec, regToRefVecMap, valueToDefVecMap);
//----------------------------------------------------------------
// Now add edges for the following (all are incoming edges except (4)):
// (1) operands of the machine instruction, including hidden operands
// (2) machine register dependences
// (3) memory load/store dependences
// (3) other resource dependences for the machine instruction, if any
// (4) output dependences when multiple machine instructions define the
// same value; all must have been generated from a single VM instrn
// (5) control dependences to branch instructions generated for the
// terminator instruction of the BB. Because of delay slots and
// 2-way conditional branches, multiple CD edges are needed
// (see addCDEdges for details).
// Also, note any uses or defs of machine registers.
//
//----------------------------------------------------------------
MachineCodeForBasicBlock& bbMvec = bb->getMachineInstrVec();
// First, add edges to the terminator instruction of the basic block.
this->addCDEdges(bb->getTerminator(), target);
// Then add memory dep edges: store->load, load->store, and store->store.
// Call instructions are treated as both load and store.
this->addMemEdges(memNodeVec, target);
// Then add edges between call instructions and CC set/use instructions
this->addCallCCEdges(memNodeVec, bbMvec, target);
// Then add incoming def-use (SSA) edges for each machine instruction.
for (unsigned i=0, N=bbMvec.size(); i < N; i++)
addEdgesForInstruction(*bbMvec[i], valueToDefVecMap, target);
#ifdef NEED_SEPARATE_NONSSA_EDGES_CODE
// Then add non-SSA edges for all VM instructions in the block.
// We assume that all machine instructions that define a value are
// generated from the VM instruction corresponding to that value.
// TODO: This could probably be done much more efficiently.
for (BasicBlock::const_iterator II = bb->begin(); II != bb->end(); ++II)
this->addNonSSAEdgesForValue(*II, target);
#endif //NEED_SEPARATE_NONSSA_EDGES_CODE
// Then add edges for dependences on machine registers
this->addMachineRegEdges(regToRefVecMap, target);
// Finally, add edges from the dummy root and to dummy leaf
this->addDummyEdges();
}
//
// class SchedGraphSet
//
/*ctor*/
SchedGraphSet::SchedGraphSet(const Method* _method,
const TargetMachine& target) :
method(_method)
{
buildGraphsForMethod(method, target);
}
/*dtor*/
SchedGraphSet::~SchedGraphSet()
{
// delete all the graphs
for (const_iterator I = begin(); I != end(); ++I)
delete I->second;
}
void
SchedGraphSet::dump() const
{
cerr << "======== Sched graphs for method `" << method->getName()
<< "' ========\n\n";
for (const_iterator I=begin(); I != end(); ++I)
I->second->dump();
cerr << "\n====== End graphs for method `" << method->getName()
<< "' ========\n\n";
}
void
SchedGraphSet::buildGraphsForMethod(const Method *method,
const TargetMachine& target)
{
for (Method::const_iterator BI = method->begin(); BI != method->end(); ++BI)
{
SchedGraph* graph = new SchedGraph(*BI, target);
this->noteGraphForBlock(*BI, graph);
}
}
std::ostream &operator<<(std::ostream &os, const SchedGraphEdge& edge)
{
os << "edge [" << edge.src->getNodeId() << "] -> ["
<< edge.sink->getNodeId() << "] : ";
switch(edge.depType) {
case SchedGraphEdge::CtrlDep: os<< "Control Dep"; break;
case SchedGraphEdge::ValueDep: os<< "Reg Value " << edge.val; break;
case SchedGraphEdge::MemoryDep: os<< "Memory Dep"; break;
case SchedGraphEdge::MachineRegister: os<< "Reg " <<edge.machineRegNum;break;
case SchedGraphEdge::MachineResource: os<<"Resource "<<edge.resourceId;break;
default: assert(0); break;
}
os << " : delay = " << edge.minDelay << "\n";
return os;
}
std::ostream &operator<<(std::ostream &os, const SchedGraphNode& node)
{
os << std::string(8, ' ')
<< "Node " << node.nodeId << " : "
<< "latency = " << node.latency << "\n" << std::string(12, ' ');
if (node.getMachineInstr() == NULL)
os << "(Dummy node)\n";
else
{
os << *node.getMachineInstr() << "\n" << std::string(12, ' ');
os << node.inEdges.size() << " Incoming Edges:\n";
for (unsigned i=0, N=node.inEdges.size(); i < N; i++)
os << std::string(16, ' ') << *node.inEdges[i];
os << std::string(12, ' ') << node.outEdges.size()
<< " Outgoing Edges:\n";
for (unsigned i=0, N=node.outEdges.size(); i < N; i++)
os << std::string(16, ' ') << *node.outEdges[i];
}
return os;
}