6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2024-06-30 23:30:28 +00:00
Code Issues Projects Releases Wiki Activity

Replace PowerPCPEI.cpp with target independant PrologEpilogInserter

Browse Source 
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@15746 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Nate Begeman 2004-08-14 22:16:36 +00:00
parent ffde1de597
commit ca068e861b
6 changed files with 64 additions and 353 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
lib/Target/PowerPC/PPC.h
View File

@ -22,10 +22,9 @@ namespace llvm {
class FunctionPass;
class TargetMachine;
FunctionPass *createPowerPCPEI();
FunctionPass *createPPCBranchSelectionPass();
FunctionPass *createPPCAsmPrinter(std::ostream &OS,TargetMachine &TM);
FunctionPass *createPPC32ISelSimple(TargetMachine &TM);
FunctionPass *createPPC32AsmPrinter(std::ostream &OS,TargetMachine &TM);
FunctionPass *createPPC64ISelSimple(TargetMachine &TM);
FunctionPass *createPPC64AsmPrinter(std::ostream &OS,TargetMachine &TM);

45
lib/Target/PowerPC/PPCFrameInfo.h Normal file
View File

@ -0,0 +1,45 @@
//===-- PowerPCFrameInfo.h - Define TargetFrameInfo for PowerPC -*- C++ -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file was developed by the LLVM research group and is distributed under
// the University of Illinois Open Source License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
//
//----------------------------------------------------------------------------
#ifndef POWERPC_FRAMEINFO_H
#define POWERPC_FRAMEINFO_H
#include "PowerPC.h"
#include "llvm/Target/TargetFrameInfo.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Target/MRegisterInfo.h"
#include <map>
namespace llvm {
class PowerPCFrameInfo: public TargetFrameInfo {
const TargetMachine &TM;
std::pair<unsigned, int> LR[1];
public:
PowerPCFrameInfo(const TargetMachine &inTM)
: TargetFrameInfo(TargetFrameInfo::StackGrowsDown, 16, 0), TM(inTM) {
LR[0].first = PPC::LR;
LR[0].second = 8;
}
std::pair<unsigned, int> *
getCalleeSaveSpillSlots(unsigned &NumEntries) const {
NumEntries = 1;
return static_cast<std::pair<unsigned, int> *>(LR);
}
};
} // End llvm namespace
#endif

17
lib/Target/PowerPC/PPCTargetMachine.cpp
View File

@ -12,6 +12,7 @@
#include "PowerPC.h"
#include "PowerPCTargetMachine.h"
#include "PowerPCFrameInfo.h"
#include "PPC32TargetMachine.h"
#include "PPC64TargetMachine.h"
#include "PPC32JITInfo.h"
@ -48,7 +49,7 @@ namespace {
PowerPCTargetMachine::PowerPCTargetMachine(const std::string &name,
IntrinsicLowering *IL,
const TargetData &TD,
const TargetFrameInfo &TFI,
const PowerPCFrameInfo &TFI,
const PowerPCJITInfo &TJI,
bool is64b)
: TargetMachine(name, IL, TD), InstrInfo(is64b), FrameInfo(TFI), JITInfo(TJI)
@ -96,17 +97,15 @@ bool PowerPCTargetMachine::addPassesToEmitAssembly(PassManager &PM,
if (PrintMachineCode)
PM.add(createMachineFunctionPrinterPass(&std::cerr));
// PowerPC-specific prolog/epilog code inserter to put the fills/spills in the
// right spots.
PM.add(createPowerPCPEI());
PM.add(createPrologEpilogCodeInserter());
// Must run branch selection immediately preceding the printer
// Must run branch selection immediately preceding the asm printer
PM.add(createPPCBranchSelectionPass());
if (AIX)
PM.add(createPPC64AsmPrinter(Out, *this));
else
PM.add(createPPC32AsmPrinter(Out, *this));
PM.add(createPPCAsmPrinter(Out, *this));
PM.add(createMachineCodeDeleter());
return false;
@ -148,16 +147,14 @@ PPC32TargetMachine::PPC32TargetMachine(const Module &M,
IntrinsicLowering *IL)
: PowerPCTargetMachine(PPC32, IL,
TargetData(PPC32,false,4,4,4,4,4,4,2,1,4),
TargetFrameInfo(TargetFrameInfo::StackGrowsDown,16,0),
PPC32JITInfo(*this), false) {}
PowerPCFrameInfo(*this), PPC32JITInfo(*this), false) {}
/// PPC64TargetMachine ctor - Create a LP64 architecture model
///
PPC64TargetMachine::PPC64TargetMachine(const Module &M, IntrinsicLowering *IL)
: PowerPCTargetMachine(PPC64, IL,
TargetData(PPC64,false,8,4,4,4,4,4,2,1,4),
TargetFrameInfo(TargetFrameInfo::StackGrowsDown,16,0),
PPC64JITInfo(*this), true) {}
PowerPCFrameInfo(*this), PPC64JITInfo(*this), true) {}
unsigned PPC32TargetMachine::getModuleMatchQuality(const Module &M) {
if (M.getEndianness() == Module::BigEndian &&

326
lib/Target/PowerPC/PowerPCPEI.cpp
View File

@ -1,326 +0,0 @@
//===-- PrologEpilogInserter.cpp - Insert Prolog/Epilog code in function --===//
//
// The LLVM Compiler Infrastructure
//
// This file was developed by the LLVM research group and is distributed under
// the University of Illinois Open Source License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This pass is responsible for finalizing the functions frame layout, saving
// callee saved registers, and for emitting prolog & epilog code for the
// function.
//
// This pass must be run after register allocation. After this pass is
// executed, it is illegal to construct MO_FrameIndex operands.
//
//===----------------------------------------------------------------------===//
//
// FIXME: The contents of this file should be merged with the target generic
// CodeGen/PrologEpilogInserter.cpp
//
//===----------------------------------------------------------------------===//
#include "PowerPC.h"
#include "llvm/CodeGen/Passes.h"
#include "llvm/CodeGen/MachineFunctionPass.h"
#include "llvm/CodeGen/MachineInstr.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Target/MRegisterInfo.h"
#include "llvm/Target/TargetFrameInfo.h"
#include "llvm/Target/TargetInstrInfo.h"
#include "Support/Debug.h"
using namespace llvm;
namespace {
struct PPCPEI : public MachineFunctionPass {
const char *getPassName() const {
return "PowerPC Frame Finalization & Prolog/Epilog Insertion";
}
/// runOnMachineFunction - Insert prolog/epilog code and replace abstract
/// frame indexes with appropriate references.
///
bool runOnMachineFunction(MachineFunction &Fn) {
RegsToSave.clear();
StackSlots.clear();
// Scan the function for modified caller saved registers and insert spill
// code for any caller saved registers that are modified. Also calculate
// the MaxCallFrameSize and HasCalls variables for the function's frame
// information and eliminates call frame pseudo instructions.
calculateCallerSavedRegisters(Fn);
// Calculate actual frame offsets for all of the abstract stack objects...
calculateFrameObjectOffsets(Fn);
// Add prolog and epilog code to the function.
insertPrologEpilogCode(Fn);
// Add register spills and fills before prolog and after epilog so that in
// the event of a very large fixed size alloca, we don't have to do
// anything weird.
saveCallerSavedRegisters(Fn);
// Replace all MO_FrameIndex operands with physical register references
// and actual offsets.
//
replaceFrameIndices(Fn);
return true;
}
private:
std::vector<unsigned> RegsToSave;
std::vector<int> StackSlots;
void calculateCallerSavedRegisters(MachineFunction &Fn);
void saveCallerSavedRegisters(MachineFunction &Fn);
void calculateFrameObjectOffsets(MachineFunction &Fn);
void replaceFrameIndices(MachineFunction &Fn);
void insertPrologEpilogCode(MachineFunction &Fn);
};
}
/// createPowerPCPEI - This function returns a pass that inserts
/// prolog and epilog code, and eliminates abstract frame references.
///
FunctionPass *llvm::createPowerPCPEI() { return new PPCPEI(); }
/// calculateCallerSavedRegisters - Scan the function for modified caller saved
/// registers. Also calculate the MaxCallFrameSize and HasCalls variables for
/// the function's frame information and eliminates call frame pseudo
/// instructions.
///
void PPCPEI::calculateCallerSavedRegisters(MachineFunction &Fn) {
const MRegisterInfo *RegInfo = Fn.getTarget().getRegisterInfo();
const TargetFrameInfo &FrameInfo = *Fn.getTarget().getFrameInfo();
// Get the callee saved register list...
const unsigned *CSRegs = RegInfo->getCalleeSaveRegs();
// Get the function call frame set-up and tear-down instruction opcode
int FrameSetupOpcode = RegInfo->getCallFrameSetupOpcode();
int FrameDestroyOpcode = RegInfo->getCallFrameDestroyOpcode();
// Early exit for targets which have no callee saved registers and no call
// frame setup/destroy pseudo instructions.
if ((CSRegs == 0 || CSRegs[0] == 0) &&
FrameSetupOpcode == -1 && FrameDestroyOpcode == -1)
return;
// This bitset contains an entry for each physical register for the target...
std::vector<bool> ModifiedRegs(RegInfo->getNumRegs());
unsigned MaxCallFrameSize = 0;
bool HasCalls = false;
for (MachineFunction::iterator BB = Fn.begin(), E = Fn.end(); BB != E; ++BB)
for (MachineBasicBlock::iterator I = BB->begin(); I != BB->end(); )
if (I->getOpcode() == FrameSetupOpcode ||
I->getOpcode() == FrameDestroyOpcode) {
assert(I->getNumOperands() == 1 && "Call Frame Setup/Destroy Pseudo"
" instructions should have a single immediate argument!");
unsigned Size = I->getOperand(0).getImmedValue();
if (Size > MaxCallFrameSize) MaxCallFrameSize = Size;
HasCalls = true;
RegInfo->eliminateCallFramePseudoInstr(Fn, *BB, I++);
} else {
for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i) {
MachineOperand &MO = I->getOperand(i);
if (MO.isRegister() && MO.isDef()) {
assert(MRegisterInfo::isPhysicalRegister(MO.getReg()) &&
"Register allocation must be performed!");
ModifiedRegs[MO.getReg()] = true; // Register is modified
}
}
++I;
}
MachineFrameInfo *FFI = Fn.getFrameInfo();
FFI->setHasCalls(HasCalls);
FFI->setMaxCallFrameSize(MaxCallFrameSize);
// Now figure out which *callee saved* registers are modified by the current
// function, thus needing to be saved and restored in the prolog/epilog.
//
for (unsigned i = 0; CSRegs[i]; ++i) {
unsigned Reg = CSRegs[i];
if (ModifiedRegs[Reg]) {
RegsToSave.push_back(Reg); // If modified register...
} else {
for (const unsigned *AliasSet = RegInfo->getAliasSet(Reg);
*AliasSet; ++AliasSet) { // Check alias registers too...
if (ModifiedRegs[*AliasSet]) {
RegsToSave.push_back(Reg);
break;
}
}
}
}
// FIXME: should we sort the regs to save so that we always get the regs in
// the correct order?
// Now that we know which registers need to be saved and restored, allocate
// stack slots for them.
int Offset = 0;
for (unsigned i = 0, e = RegsToSave.size(); i != e; ++i) {
unsigned RegSize = RegInfo->getRegClass(RegsToSave[i])->getSize();
int FrameIdx;
if (RegsToSave[i] == PPC::LR) {
FrameIdx = FFI->CreateFixedObject(RegSize, 8); // LR lives at +8
} else {
Offset -= RegSize;
FrameIdx = FFI->CreateFixedObject(RegSize, Offset);
}
StackSlots.push_back(FrameIdx);
}
}
/// saveCallerSavedRegisters - Insert spill code for any caller saved registers
/// that are modified in the function.
///
void PPCPEI::saveCallerSavedRegisters(MachineFunction &Fn) {
// Early exit if no caller saved registers are modified!
if (RegsToSave.empty())
return;
const MRegisterInfo *RegInfo = Fn.getTarget().getRegisterInfo();
// Now that we have a stack slot for each register to be saved, insert spill
// code into the entry block...
MachineBasicBlock *MBB = Fn.begin();
MachineBasicBlock::iterator I = MBB->begin();
for (unsigned i = 0, e = RegsToSave.size(); i != e; ++i) {
const TargetRegisterClass *RC = RegInfo->getRegClass(RegsToSave[i]);
// Insert the spill to the stack frame...
RegInfo->storeRegToStackSlot(*MBB, I, RegsToSave[i], StackSlots[i], RC);
}
// Add code to restore the callee-save registers in each exiting block.
const TargetInstrInfo &TII = *Fn.getTarget().getInstrInfo();
for (MachineFunction::iterator FI = Fn.begin(), E = Fn.end(); FI != E; ++FI) {
// If last instruction is a return instruction, add an epilogue
if (!FI->empty() && TII.isReturn(FI->back().getOpcode())) {
MBB = FI;
I = MBB->end(); --I;
for (unsigned i = 0, e = RegsToSave.size(); i != e; ++i) {
const TargetRegisterClass *RC = RegInfo->getRegClass(RegsToSave[i]);
RegInfo->loadRegFromStackSlot(*MBB, I, RegsToSave[i],StackSlots[i], RC);
--I; // Insert in reverse order
}
}
}
}
/// calculateFrameObjectOffsets - Calculate actual frame offsets for all of the
/// abstract stack objects...
///
void PPCPEI::calculateFrameObjectOffsets(MachineFunction &Fn) {
const TargetFrameInfo &TFI = *Fn.getTarget().getFrameInfo();
bool StackGrowsDown =
TFI.getStackGrowthDirection() == TargetFrameInfo::StackGrowsDown;
// Loop over all of the stack objects, assigning sequential addresses...
MachineFrameInfo *FFI = Fn.getFrameInfo();
unsigned StackAlignment = TFI.getStackAlignment();
// Start at the beginning of the local area.
// The Offset is the distance from the stack top in the direction
// of stack growth -- so it's always positive.
int Offset = TFI.getOffsetOfLocalArea();
if (StackGrowsDown)
Offset = -Offset;
assert(Offset >= 0
&& "Local area offset should be in direction of stack growth");
// If there are fixed sized objects that are preallocated in the local area,
// non-fixed objects can't be allocated right at the start of local area.
// We currently don't support filling in holes in between fixed sized objects,
// so we adjust 'Offset' to point to the end of last fixed sized
// preallocated object.
for (int i = FFI->getObjectIndexBegin(); i != 0; ++i) {
int FixedOff;
if (StackGrowsDown) {
// The maximum distance from the stack pointer is at lower address of
// the object -- which is given by offset. For down growing stack
// the offset is negative, so we negate the offset to get the distance.
FixedOff = -FFI->getObjectOffset(i);
} else {
// The maximum distance from the start pointer is at the upper
// address of the object.
FixedOff = FFI->getObjectOffset(i) + FFI->getObjectSize(i);
}
if (FixedOff > Offset) Offset = FixedOff;
}
for (unsigned i = 0, e = FFI->getObjectIndexEnd(); i != e; ++i) {
// If stack grows down, we need to add size of find the lowest
// address of the object.
if (StackGrowsDown)
Offset += FFI->getObjectSize(i);
unsigned Align = FFI->getObjectAlignment(i);
assert(Align <= StackAlignment && "Cannot align stack object to higher "
"alignment boundary than the stack itself!");
Offset = (Offset+Align-1)/Align*Align; // Adjust to Alignment boundary...
if (StackGrowsDown) {
FFI->setObjectOffset(i, -Offset); // Set the computed offset
} else {
FFI->setObjectOffset(i, Offset);
Offset += FFI->getObjectSize(i);
}
}
// Set the final value of the stack pointer...
FFI->setStackSize(Offset);
}
/// insertPrologEpilogCode - Scan the function for modified caller saved
/// registers, insert spill code for these caller saved registers, then add
/// prolog and epilog code to the function.
///
void PPCPEI::insertPrologEpilogCode(MachineFunction &Fn) {
// Add prologue to the function...
Fn.getTarget().getRegisterInfo()->emitPrologue(Fn);
// Add epilogue to restore the callee-save registers in each exiting block
const TargetInstrInfo &TII = *Fn.getTarget().getInstrInfo();
for (MachineFunction::iterator I = Fn.begin(), E = Fn.end(); I != E; ++I) {
// If last instruction is a return instruction, add an epilogue
if (!I->empty() && TII.isReturn(I->back().getOpcode()))
Fn.getTarget().getRegisterInfo()->emitEpilogue(Fn, *I);
}
}
/// replaceFrameIndices - Replace all MO_FrameIndex operands with physical
/// register references and actual offsets.
///
void PPCPEI::replaceFrameIndices(MachineFunction &Fn) {
if (!Fn.getFrameInfo()->hasStackObjects()) return; // Nothing to do?
const TargetMachine &TM = Fn.getTarget();
assert(TM.getRegisterInfo() && "TM::getRegisterInfo() must be implemented!");
const MRegisterInfo &MRI = *TM.getRegisterInfo();
for (MachineFunction::iterator BB = Fn.begin(), E = Fn.end(); BB != E; ++BB)
for (MachineBasicBlock::iterator I = BB->begin(); I != BB->end(); ++I)
for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i)
if (I->getOperand(i).isFrameIndex()) {
// If this instruction has a FrameIndex operand, we need to use that
// target machine register info object to eliminate it.
MRI.eliminateFrameIndex(Fn, I);
break;
}
}

9
lib/Target/PowerPC/PowerPCTargetMachine.h
View File

@ -14,11 +14,12 @@
#ifndef POWERPC_TARGETMACHINE_H
#define POWERPC_TARGETMACHINE_H
#include "PowerPCFrameInfo.h"
#include "PowerPCInstrInfo.h"
#include "PowerPCJITInfo.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Target/TargetFrameInfo.h"
#include "llvm/PassManager.h"
#include "PowerPCInstrInfo.h"
#include "PowerPCJITInfo.h"
#include <set>
namespace llvm {
@ -28,12 +29,12 @@ class IntrinsicLowering;
class PowerPCTargetMachine : public TargetMachine {
PowerPCInstrInfo InstrInfo;
TargetFrameInfo FrameInfo;
PowerPCFrameInfo FrameInfo;
PowerPCJITInfo JITInfo;
protected:
PowerPCTargetMachine(const std::string &name, IntrinsicLowering *IL,
const TargetData &TD, const TargetFrameInfo &TFI,
const TargetData &TD, const PowerPCFrameInfo &TFI,
const PowerPCJITInfo &TJI, bool is64b);
public:
virtual const PowerPCInstrInfo *getInstrInfo() const { return &InstrInfo; }

17
lib/Target/PowerPC/README.txt
View File

@ -1,6 +1,10 @@
TODO:
* implement cast fp to bool
* implement signed right shift by reg
* implement algebraic shift right long by reg
* implement scheduling info
* implement powerpc-64 for darwin
* implement powerpc-64 for aix
* fix rlwimi generation to be use-and-def
* fix ulong to double:
floatdidf assumes signed longs. so if the high but of a ulong
just happens to be set, you get the wrong sign. The fix for this
@ -17,7 +21,6 @@ TODO:
shift right ulong a, 1 (we could use emitShift)
call floatdidf
fadd f1, f1, f1 (fp left shift by 1)
* PowerPCPEI.cpp needs to be replaced by shiny new target hook
* setCondInst needs to know branchless versions of seteq/setne/etc
* cast elimination pass (uint -> sbyte -> short, kill the byte -> short)
* should hint to the branch select pass that it doesn't need to print the
@ -26,13 +29,6 @@ TODO:
b .LBBl42__2E_expand_function_8_42 ; NewDefault
b .LBBl42__2E_expand_function_8_42 ; NewDefault
Current hacks:
* lazy insert of GlobalBaseReg definition at front of first MBB
A prime candidate for sabre's future "slightly above ISel" passes.
* cast code is huge, unwieldy. Should probably be broken up into
smaller pieces.
* visitLoadInst is getting awfully cluttered as well.
Currently failing tests:
* SingleSource
`- Regression
@ -46,8 +42,7 @@ Currently failing tests:
* MultiSource
|- Applications
| `- burg: miscompilation
| `- siod: llc bus error
| `- hbd: miscompilation
| `- d (make_dparser): miscompilation
`- Benchmarks
`- MallocBench/make: miscompilation
`- MallocBench/gs: miscompilation