llvm-6502/lib/Target/PowerPC/PPCISelPattern.cpp
2005-08-03 17:29:52 +00:00

2514 lines
94 KiB
C++

//===-- PPC32ISelPattern.cpp - A pattern matching inst selector for PPC32 -===//
//
// The LLVM Compiler Infrastructure
//
// This file was developed by Nate Begeman and is distributed under
// the University of Illinois Open Source License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file defines a pattern matching instruction selector for 32 bit PowerPC.
// Magic number generation for integer divide from the PowerPC Compiler Writer's
// Guide, section 3.2.3.5
//
//===----------------------------------------------------------------------===//
#include "PowerPC.h"
#include "PowerPCInstrBuilder.h"
#include "PowerPCInstrInfo.h"
#include "PPC32TargetMachine.h"
#include "llvm/Constants.h"
#include "llvm/Function.h"
#include "llvm/CodeGen/MachineConstantPool.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/SelectionDAG.h"
#include "llvm/CodeGen/SelectionDAGISel.h"
#include "llvm/CodeGen/SSARegMap.h"
#include "llvm/Target/TargetData.h"
#include "llvm/Target/TargetLowering.h"
#include "llvm/Target/TargetOptions.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/ADT/Statistic.h"
#include <set>
#include <algorithm>
using namespace llvm;
// IsRunOfOnes - Returns true iff Val consists of one contiguous run of 1s with
// any number of 0s on either side. The 1s are allowed to wrap from LSB to
// MSB, so 0x000FFF0, 0x0000FFFF, and 0xFF0000FF are all runs. 0x0F0F0000 is
// not, since all 1s are not contiguous.
static bool IsRunOfOnes(unsigned Val, unsigned &MB, unsigned &ME) {
if (isShiftedMask_32(Val)) {
// look for the first non-zero bit
MB = CountLeadingZeros_32(Val);
// look for the first zero bit after the run of ones
ME = CountLeadingZeros_32((Val - 1) ^ Val);
return true;
} else if (isShiftedMask_32(Val = ~Val)) { // invert mask
// effectively look for the first zero bit
ME = CountLeadingZeros_32(Val) - 1;
// effectively look for the first one bit after the run of zeros
MB = CountLeadingZeros_32((Val - 1) ^ Val) + 1;
return true;
}
// no run present
return false;
}
//===----------------------------------------------------------------------===//
// PPC32TargetLowering - PPC32 Implementation of the TargetLowering interface
namespace {
class PPC32TargetLowering : public TargetLowering {
int VarArgsFrameIndex; // FrameIndex for start of varargs area.
int ReturnAddrIndex; // FrameIndex for return slot.
public:
PPC32TargetLowering(TargetMachine &TM) : TargetLowering(TM) {
// Fold away setcc operations if possible.
setSetCCIsExpensive();
// Set up the register classes.
addRegisterClass(MVT::i32, PPC32::GPRCRegisterClass);
addRegisterClass(MVT::f32, PPC32::FPRCRegisterClass);
addRegisterClass(MVT::f64, PPC32::FPRCRegisterClass);
// PowerPC has no intrinsics for these particular operations
setOperationAction(ISD::MEMMOVE, MVT::Other, Expand);
setOperationAction(ISD::MEMSET, MVT::Other, Expand);
setOperationAction(ISD::MEMCPY, MVT::Other, Expand);
// PowerPC has an i16 but no i8 (or i1) SEXTLOAD
setOperationAction(ISD::SEXTLOAD, MVT::i1, Expand);
setOperationAction(ISD::SEXTLOAD, MVT::i8, Expand);
// PowerPC has no SREM/UREM instructions
setOperationAction(ISD::SREM, MVT::i32, Expand);
setOperationAction(ISD::UREM, MVT::i32, Expand);
// We don't support sin/cos/sqrt/fmod
setOperationAction(ISD::FSIN , MVT::f64, Expand);
setOperationAction(ISD::FCOS , MVT::f64, Expand);
setOperationAction(ISD::SREM , MVT::f64, Expand);
setOperationAction(ISD::FSIN , MVT::f32, Expand);
setOperationAction(ISD::FCOS , MVT::f32, Expand);
setOperationAction(ISD::SREM , MVT::f32, Expand);
// If we're enabling GP optimizations, use hardware square root
if (!GPOPT) {
setOperationAction(ISD::FSQRT, MVT::f64, Expand);
setOperationAction(ISD::FSQRT, MVT::f32, Expand);
}
//PowerPC does not have CTPOP or CTTZ
setOperationAction(ISD::CTPOP, MVT::i32 , Expand);
setOperationAction(ISD::CTTZ , MVT::i32 , Expand);
setSetCCResultContents(ZeroOrOneSetCCResult);
addLegalFPImmediate(+0.0); // Necessary for FSEL
addLegalFPImmediate(-0.0); //
computeRegisterProperties();
}
/// LowerArguments - This hook must be implemented to indicate how we should
/// lower the arguments for the specified function, into the specified DAG.
virtual std::vector<SDOperand>
LowerArguments(Function &F, SelectionDAG &DAG);
/// LowerCallTo - This hook lowers an abstract call to a function into an
/// actual call.
virtual std::pair<SDOperand, SDOperand>
LowerCallTo(SDOperand Chain, const Type *RetTy, bool isVarArg, unsigned CC,
bool isTailCall, SDOperand Callee, ArgListTy &Args,
SelectionDAG &DAG);
virtual SDOperand LowerVAStart(SDOperand Chain, SDOperand VAListP,
Value *VAListV, SelectionDAG &DAG);
virtual std::pair<SDOperand,SDOperand>
LowerVAArg(SDOperand Chain, SDOperand VAListP, Value *VAListV,
const Type *ArgTy, SelectionDAG &DAG);
virtual std::pair<SDOperand, SDOperand>
LowerFrameReturnAddress(bool isFrameAddr, SDOperand Chain, unsigned Depth,
SelectionDAG &DAG);
};
}
std::vector<SDOperand>
PPC32TargetLowering::LowerArguments(Function &F, SelectionDAG &DAG) {
//
// add beautiful description of PPC stack frame format, or at least some docs
//
MachineFunction &MF = DAG.getMachineFunction();
MachineFrameInfo *MFI = MF.getFrameInfo();
MachineBasicBlock& BB = MF.front();
std::vector<SDOperand> ArgValues;
// Due to the rather complicated nature of the PowerPC ABI, rather than a
// fixed size array of physical args, for the sake of simplicity let the STL
// handle tracking them for us.
std::vector<unsigned> argVR, argPR, argOp;
unsigned ArgOffset = 24;
unsigned GPR_remaining = 8;
unsigned FPR_remaining = 13;
unsigned GPR_idx = 0, FPR_idx = 0;
static const unsigned GPR[] = {
PPC::R3, PPC::R4, PPC::R5, PPC::R6,
PPC::R7, PPC::R8, PPC::R9, PPC::R10,
};
static const unsigned FPR[] = {
PPC::F1, PPC::F2, PPC::F3, PPC::F4, PPC::F5, PPC::F6, PPC::F7,
PPC::F8, PPC::F9, PPC::F10, PPC::F11, PPC::F12, PPC::F13
};
// Add DAG nodes to load the arguments... On entry to a function on PPC,
// the arguments start at offset 24, although they are likely to be passed
// in registers.
for (Function::arg_iterator I = F.arg_begin(), E = F.arg_end(); I != E; ++I) {
SDOperand newroot, argt;
unsigned ObjSize;
bool needsLoad = false;
bool ArgLive = !I->use_empty();
MVT::ValueType ObjectVT = getValueType(I->getType());
switch (ObjectVT) {
default: assert(0 && "Unhandled argument type!");
case MVT::i1:
case MVT::i8:
case MVT::i16:
case MVT::i32:
ObjSize = 4;
if (!ArgLive) break;
if (GPR_remaining > 0) {
MF.addLiveIn(GPR[GPR_idx]);
argt = newroot = DAG.getCopyFromReg(GPR[GPR_idx], MVT::i32,
DAG.getRoot());
if (ObjectVT != MVT::i32)
argt = DAG.getNode(ISD::TRUNCATE, ObjectVT, newroot);
} else {
needsLoad = true;
}
break;
case MVT::i64: ObjSize = 8;
if (!ArgLive) break;
if (GPR_remaining > 0) {
SDOperand argHi, argLo;
MF.addLiveIn(GPR[GPR_idx]);
argHi = DAG.getCopyFromReg(GPR[GPR_idx], MVT::i32, DAG.getRoot());
// If we have two or more remaining argument registers, then both halves
// of the i64 can be sourced from there. Otherwise, the lower half will
// have to come off the stack. This can happen when an i64 is preceded
// by 28 bytes of arguments.
if (GPR_remaining > 1) {
MF.addLiveIn(GPR[GPR_idx+1]);
argLo = DAG.getCopyFromReg(GPR[GPR_idx+1], MVT::i32, argHi);
} else {
int FI = MFI->CreateFixedObject(4, ArgOffset+4);
SDOperand FIN = DAG.getFrameIndex(FI, MVT::i32);
argLo = DAG.getLoad(MVT::i32, DAG.getEntryNode(), FIN,
DAG.getSrcValue(NULL));
}
// Build the outgoing arg thingy
argt = DAG.getNode(ISD::BUILD_PAIR, MVT::i64, argLo, argHi);
newroot = argLo;
} else {
needsLoad = true;
}
break;
case MVT::f32:
case MVT::f64:
ObjSize = (ObjectVT == MVT::f64) ? 8 : 4;
if (!ArgLive) break;
if (FPR_remaining > 0) {
MF.addLiveIn(FPR[FPR_idx]);
argt = newroot = DAG.getCopyFromReg(FPR[FPR_idx], ObjectVT,
DAG.getRoot());
--FPR_remaining;
++FPR_idx;
} else {
needsLoad = true;
}
break;
}
// We need to load the argument to a virtual register if we determined above
// that we ran out of physical registers of the appropriate type
if (needsLoad) {
unsigned SubregOffset = 0;
if (ObjectVT == MVT::i8 || ObjectVT == MVT::i1) SubregOffset = 3;
if (ObjectVT == MVT::i16) SubregOffset = 2;
int FI = MFI->CreateFixedObject(ObjSize, ArgOffset);
SDOperand FIN = DAG.getFrameIndex(FI, MVT::i32);
FIN = DAG.getNode(ISD::ADD, MVT::i32, FIN,
DAG.getConstant(SubregOffset, MVT::i32));
argt = newroot = DAG.getLoad(ObjectVT, DAG.getEntryNode(), FIN,
DAG.getSrcValue(NULL));
}
// Every 4 bytes of argument space consumes one of the GPRs available for
// argument passing.
if (GPR_remaining > 0) {
unsigned delta = (GPR_remaining > 1 && ObjSize == 8) ? 2 : 1;
GPR_remaining -= delta;
GPR_idx += delta;
}
ArgOffset += ObjSize;
if (newroot.Val)
DAG.setRoot(newroot.getValue(1));
ArgValues.push_back(argt);
}
// If the function takes variable number of arguments, make a frame index for
// the start of the first vararg value... for expansion of llvm.va_start.
if (F.isVarArg()) {
VarArgsFrameIndex = MFI->CreateFixedObject(4, ArgOffset);
SDOperand FIN = DAG.getFrameIndex(VarArgsFrameIndex, MVT::i32);
// If this function is vararg, store any remaining integer argument regs
// to their spots on the stack so that they may be loaded by deferencing the
// result of va_next.
std::vector<SDOperand> MemOps;
for (; GPR_remaining > 0; --GPR_remaining, ++GPR_idx) {
MF.addLiveIn(GPR[GPR_idx]);
SDOperand Val = DAG.getCopyFromReg(GPR[GPR_idx], MVT::i32, DAG.getRoot());
SDOperand Store = DAG.getNode(ISD::STORE, MVT::Other, Val.getValue(1),
Val, FIN, DAG.getSrcValue(NULL));
MemOps.push_back(Store);
// Increment the address by four for the next argument to store
SDOperand PtrOff = DAG.getConstant(4, getPointerTy());
FIN = DAG.getNode(ISD::ADD, MVT::i32, FIN, PtrOff);
}
DAG.setRoot(DAG.getNode(ISD::TokenFactor, MVT::Other, MemOps));
}
// Finally, inform the code generator which regs we return values in.
switch (getValueType(F.getReturnType())) {
default: assert(0 && "Unknown type!");
case MVT::isVoid: break;
case MVT::i1:
case MVT::i8:
case MVT::i16:
case MVT::i32:
MF.addLiveOut(PPC::R3);
break;
case MVT::i64:
MF.addLiveOut(PPC::R3);
MF.addLiveOut(PPC::R4);
break;
case MVT::f32:
case MVT::f64:
MF.addLiveOut(PPC::F1);
break;
}
return ArgValues;
}
std::pair<SDOperand, SDOperand>
PPC32TargetLowering::LowerCallTo(SDOperand Chain,
const Type *RetTy, bool isVarArg,
unsigned CallingConv, bool isTailCall,
SDOperand Callee, ArgListTy &Args,
SelectionDAG &DAG) {
// args_to_use will accumulate outgoing args for the ISD::CALL case in
// SelectExpr to use to put the arguments in the appropriate registers.
std::vector<SDOperand> args_to_use;
// Count how many bytes are to be pushed on the stack, including the linkage
// area, and parameter passing area.
unsigned NumBytes = 24;
if (Args.empty()) {
Chain = DAG.getNode(ISD::CALLSEQ_START, MVT::Other, Chain,
DAG.getConstant(NumBytes, getPointerTy()));
} else {
for (unsigned i = 0, e = Args.size(); i != e; ++i)
switch (getValueType(Args[i].second)) {
default: assert(0 && "Unknown value type!");
case MVT::i1:
case MVT::i8:
case MVT::i16:
case MVT::i32:
case MVT::f32:
NumBytes += 4;
break;
case MVT::i64:
case MVT::f64:
NumBytes += 8;
break;
}
// Just to be safe, we'll always reserve the full 24 bytes of linkage area
// plus 32 bytes of argument space in case any called code gets funky on us.
// (Required by ABI to support var arg)
if (NumBytes < 56) NumBytes = 56;
// Adjust the stack pointer for the new arguments...
// These operations are automatically eliminated by the prolog/epilog pass
Chain = DAG.getNode(ISD::CALLSEQ_START, MVT::Other, Chain,
DAG.getConstant(NumBytes, getPointerTy()));
// Set up a copy of the stack pointer for use loading and storing any
// arguments that may not fit in the registers available for argument
// passing.
SDOperand StackPtr = DAG.getCopyFromReg(PPC::R1, MVT::i32,
DAG.getEntryNode());
// Figure out which arguments are going to go in registers, and which in
// memory. Also, if this is a vararg function, floating point operations
// must be stored to our stack, and loaded into integer regs as well, if
// any integer regs are available for argument passing.
unsigned ArgOffset = 24;
unsigned GPR_remaining = 8;
unsigned FPR_remaining = 13;
std::vector<SDOperand> MemOps;
for (unsigned i = 0, e = Args.size(); i != e; ++i) {
// PtrOff will be used to store the current argument to the stack if a
// register cannot be found for it.
SDOperand PtrOff = DAG.getConstant(ArgOffset, getPointerTy());
PtrOff = DAG.getNode(ISD::ADD, MVT::i32, StackPtr, PtrOff);
MVT::ValueType ArgVT = getValueType(Args[i].second);
switch (ArgVT) {
default: assert(0 && "Unexpected ValueType for argument!");
case MVT::i1:
case MVT::i8:
case MVT::i16:
// Promote the integer to 32 bits. If the input type is signed use a
// sign extend, otherwise use a zero extend.
if (Args[i].second->isSigned())
Args[i].first =DAG.getNode(ISD::SIGN_EXTEND, MVT::i32, Args[i].first);
else
Args[i].first =DAG.getNode(ISD::ZERO_EXTEND, MVT::i32, Args[i].first);
// FALL THROUGH
case MVT::i32:
if (GPR_remaining > 0) {
args_to_use.push_back(Args[i].first);
--GPR_remaining;
} else {
MemOps.push_back(DAG.getNode(ISD::STORE, MVT::Other, Chain,
Args[i].first, PtrOff,
DAG.getSrcValue(NULL)));
}
ArgOffset += 4;
break;
case MVT::i64:
// If we have one free GPR left, we can place the upper half of the i64
// in it, and store the other half to the stack. If we have two or more
// free GPRs, then we can pass both halves of the i64 in registers.
if (GPR_remaining > 0) {
SDOperand Hi = DAG.getNode(ISD::EXTRACT_ELEMENT, MVT::i32,
Args[i].first, DAG.getConstant(1, MVT::i32));
SDOperand Lo = DAG.getNode(ISD::EXTRACT_ELEMENT, MVT::i32,
Args[i].first, DAG.getConstant(0, MVT::i32));
args_to_use.push_back(Hi);
--GPR_remaining;
if (GPR_remaining > 0) {
args_to_use.push_back(Lo);
--GPR_remaining;
} else {
SDOperand ConstFour = DAG.getConstant(4, getPointerTy());
PtrOff = DAG.getNode(ISD::ADD, MVT::i32, PtrOff, ConstFour);
MemOps.push_back(DAG.getNode(ISD::STORE, MVT::Other, Chain,
Lo, PtrOff, DAG.getSrcValue(NULL)));
}
} else {
MemOps.push_back(DAG.getNode(ISD::STORE, MVT::Other, Chain,
Args[i].first, PtrOff,
DAG.getSrcValue(NULL)));
}
ArgOffset += 8;
break;
case MVT::f32:
case MVT::f64:
if (FPR_remaining > 0) {
args_to_use.push_back(Args[i].first);
--FPR_remaining;
if (isVarArg) {
SDOperand Store = DAG.getNode(ISD::STORE, MVT::Other, Chain,
Args[i].first, PtrOff,
DAG.getSrcValue(NULL));
MemOps.push_back(Store);
// Float varargs are always shadowed in available integer registers
if (GPR_remaining > 0) {
SDOperand Load = DAG.getLoad(MVT::i32, Store, PtrOff,
DAG.getSrcValue(NULL));
MemOps.push_back(Load);
args_to_use.push_back(Load);
--GPR_remaining;
}
if (GPR_remaining > 0 && MVT::f64 == ArgVT) {
SDOperand ConstFour = DAG.getConstant(4, getPointerTy());
PtrOff = DAG.getNode(ISD::ADD, MVT::i32, PtrOff, ConstFour);
SDOperand Load = DAG.getLoad(MVT::i32, Store, PtrOff,
DAG.getSrcValue(NULL));
MemOps.push_back(Load);
args_to_use.push_back(Load);
--GPR_remaining;
}
} else {
// If we have any FPRs remaining, we may also have GPRs remaining.
// Args passed in FPRs consume either 1 (f32) or 2 (f64) available
// GPRs.
if (GPR_remaining > 0) {
args_to_use.push_back(DAG.getNode(ISD::UNDEF, MVT::i32));
--GPR_remaining;
}
if (GPR_remaining > 0 && MVT::f64 == ArgVT) {
args_to_use.push_back(DAG.getNode(ISD::UNDEF, MVT::i32));
--GPR_remaining;
}
}
} else {
MemOps.push_back(DAG.getNode(ISD::STORE, MVT::Other, Chain,
Args[i].first, PtrOff,
DAG.getSrcValue(NULL)));
}
ArgOffset += (ArgVT == MVT::f32) ? 4 : 8;
break;
}
}
if (!MemOps.empty())
Chain = DAG.getNode(ISD::TokenFactor, MVT::Other, MemOps);
}
std::vector<MVT::ValueType> RetVals;
MVT::ValueType RetTyVT = getValueType(RetTy);
if (RetTyVT != MVT::isVoid)
RetVals.push_back(RetTyVT);
RetVals.push_back(MVT::Other);
SDOperand TheCall = SDOperand(DAG.getCall(RetVals,
Chain, Callee, args_to_use), 0);
Chain = TheCall.getValue(RetTyVT != MVT::isVoid);
Chain = DAG.getNode(ISD::CALLSEQ_END, MVT::Other, Chain,
DAG.getConstant(NumBytes, getPointerTy()));
return std::make_pair(TheCall, Chain);
}
SDOperand PPC32TargetLowering::LowerVAStart(SDOperand Chain, SDOperand VAListP,
Value *VAListV, SelectionDAG &DAG) {
// vastart just stores the address of the VarArgsFrameIndex slot into the
// memory location argument.
SDOperand FR = DAG.getFrameIndex(VarArgsFrameIndex, MVT::i32);
return DAG.getNode(ISD::STORE, MVT::Other, Chain, FR, VAListP,
DAG.getSrcValue(VAListV));
}
std::pair<SDOperand,SDOperand>
PPC32TargetLowering::LowerVAArg(SDOperand Chain,
SDOperand VAListP, Value *VAListV,
const Type *ArgTy, SelectionDAG &DAG) {
MVT::ValueType ArgVT = getValueType(ArgTy);
SDOperand VAList =
DAG.getLoad(MVT::i32, Chain, VAListP, DAG.getSrcValue(VAListV));
SDOperand Result = DAG.getLoad(ArgVT, Chain, VAList, DAG.getSrcValue(NULL));
unsigned Amt;
if (ArgVT == MVT::i32 || ArgVT == MVT::f32)
Amt = 4;
else {
assert((ArgVT == MVT::i64 || ArgVT == MVT::f64) &&
"Other types should have been promoted for varargs!");
Amt = 8;
}
VAList = DAG.getNode(ISD::ADD, VAList.getValueType(), VAList,
DAG.getConstant(Amt, VAList.getValueType()));
Chain = DAG.getNode(ISD::STORE, MVT::Other, Chain,
VAList, VAListP, DAG.getSrcValue(VAListV));
return std::make_pair(Result, Chain);
}
std::pair<SDOperand, SDOperand> PPC32TargetLowering::
LowerFrameReturnAddress(bool isFrameAddress, SDOperand Chain, unsigned Depth,
SelectionDAG &DAG) {
assert(0 && "LowerFrameReturnAddress unimplemented");
abort();
}
namespace {
Statistic<>Recorded("ppc-codegen", "Number of recording ops emitted");
Statistic<>FusedFP("ppc-codegen", "Number of fused fp operations");
Statistic<>FrameOff("ppc-codegen", "Number of frame idx offsets collapsed");
//===--------------------------------------------------------------------===//
/// ISel - PPC32 specific code to select PPC32 machine instructions for
/// SelectionDAG operations.
//===--------------------------------------------------------------------===//
class ISel : public SelectionDAGISel {
PPC32TargetLowering PPC32Lowering;
SelectionDAG *ISelDAG; // Hack to support us having a dag->dag transform
// for sdiv and udiv until it is put into the future
// dag combiner.
/// ExprMap - As shared expressions are codegen'd, we keep track of which
/// vreg the value is produced in, so we only emit one copy of each compiled
/// tree.
std::map<SDOperand, unsigned> ExprMap;
unsigned GlobalBaseReg;
bool GlobalBaseInitialized;
bool RecordSuccess;
public:
ISel(TargetMachine &TM) : SelectionDAGISel(PPC32Lowering), PPC32Lowering(TM),
ISelDAG(0) {}
/// runOnFunction - Override this function in order to reset our per-function
/// variables.
virtual bool runOnFunction(Function &Fn) {
// Make sure we re-emit a set of the global base reg if necessary
GlobalBaseInitialized = false;
return SelectionDAGISel::runOnFunction(Fn);
}
/// InstructionSelectBasicBlock - This callback is invoked by
/// SelectionDAGISel when it has created a SelectionDAG for us to codegen.
virtual void InstructionSelectBasicBlock(SelectionDAG &DAG) {
DEBUG(BB->dump());
// Codegen the basic block.
ISelDAG = &DAG;
Select(DAG.getRoot());
// Clear state used for selection.
ExprMap.clear();
ISelDAG = 0;
}
// dag -> dag expanders for integer divide by constant
SDOperand BuildSDIVSequence(SDOperand N);
SDOperand BuildUDIVSequence(SDOperand N);
unsigned getGlobalBaseReg();
unsigned getConstDouble(double floatVal, unsigned Result);
void MoveCRtoGPR(unsigned CCReg, bool Inv, unsigned Idx, unsigned Result);
bool SelectBitfieldInsert(SDOperand OR, unsigned Result);
unsigned FoldIfWideZeroExtend(SDOperand N);
unsigned SelectCC(SDOperand CC, unsigned &Opc, bool &Inv, unsigned &Idx);
unsigned SelectCCExpr(SDOperand N, unsigned& Opc, bool &Inv, unsigned &Idx);
unsigned SelectExpr(SDOperand N, bool Recording=false);
void Select(SDOperand N);
unsigned SelectAddr(SDOperand N, unsigned& Reg, int& offset);
void SelectBranchCC(SDOperand N);
virtual const char *getPassName() const {
return "PowerPC Pattern Instruction Selection";
}
};
/// getImmediateForOpcode - This method returns a value indicating whether
/// the ConstantSDNode N can be used as an immediate to Opcode. The return
/// values are either 0, 1 or 2. 0 indicates that either N is not a
/// ConstantSDNode, or is not suitable for use by that opcode.
/// Return value codes for turning into an enum someday:
/// 1: constant may be used in normal immediate form.
/// 2: constant may be used in shifted immediate form.
/// 3: log base 2 of the constant may be used.
/// 4: constant is suitable for integer division conversion
/// 5: constant is a bitfield mask
///
static unsigned getImmediateForOpcode(SDOperand N, unsigned Opcode,
unsigned& Imm, bool U = false) {
if (N.getOpcode() != ISD::Constant) return 0;
int v = (int)cast<ConstantSDNode>(N)->getSignExtended();
switch(Opcode) {
default: return 0;
case ISD::ADD:
if (isInt16(v)) { Imm = v & 0xFFFF; return 1; }
if ((v & 0x0000FFFF) == 0) { Imm = v >> 16; return 2; }
break;
case ISD::AND: {
unsigned MB, ME;
if (IsRunOfOnes(v, MB, ME)) { Imm = MB << 16 | ME & 0xFFFF; return 5; }
if (isUInt16(v)) { Imm = v & 0xFFFF; return 1; }
if ((v & 0x0000FFFF) == 0) { Imm = v >> 16; return 2; }
break;
}
case ISD::XOR:
case ISD::OR:
if (isUInt16(v)) { Imm = v & 0xFFFF; return 1; }
if ((v & 0x0000FFFF) == 0) { Imm = v >> 16; return 2; }
break;
case ISD::MUL:
if (isInt16(v)) { Imm = v & 0xFFFF; return 1; }
break;
case ISD::SUB:
// handle subtract-from separately from subtract, since subi is really addi
if (U && isInt16(v)) { Imm = v & 0xFFFF; return 1; }
if (!U && isInt16(-v)) { Imm = (-v) & 0xFFFF; return 1; }
break;
case ISD::SETCC:
if (U && isUInt16(v)) { Imm = v & 0xFFFF; return 1; }
if (!U && isInt16(v)) { Imm = v & 0xFFFF; return 1; }
break;
case ISD::SDIV:
if (isPowerOf2_32(v)) { Imm = Log2_32(v); return 3; }
if (isPowerOf2_32(-v)) { Imm = Log2_32(-v); return 3; }
if (v <= -2 || v >= 2) { return 4; }
break;
case ISD::UDIV:
if (v > 1) { return 4; }
break;
}
return 0;
}
/// NodeHasRecordingVariant - If SelectExpr can always produce code for
/// NodeOpcode that also sets CR0 as a side effect, return true. Otherwise,
/// return false.
static bool NodeHasRecordingVariant(unsigned NodeOpcode) {
switch(NodeOpcode) {
default: return false;
case ISD::AND:
case ISD::OR:
return true;
}
}
/// getBCCForSetCC - Returns the PowerPC condition branch mnemonic corresponding
/// to Condition. If the Condition is unordered or unsigned, the bool argument
/// U is set to true, otherwise it is set to false.
static unsigned getBCCForSetCC(unsigned Condition, bool& U) {
U = false;
switch (Condition) {
default: assert(0 && "Unknown condition!"); abort();
case ISD::SETEQ: return PPC::BEQ;
case ISD::SETNE: return PPC::BNE;
case ISD::SETULT: U = true;
case ISD::SETLT: return PPC::BLT;
case ISD::SETULE: U = true;
case ISD::SETLE: return PPC::BLE;
case ISD::SETUGT: U = true;
case ISD::SETGT: return PPC::BGT;
case ISD::SETUGE: U = true;
case ISD::SETGE: return PPC::BGE;
}
return 0;
}
/// getCROpForOp - Return the condition register opcode (or inverted opcode)
/// associated with the SelectionDAG opcode.
static unsigned getCROpForSetCC(unsigned Opcode, bool Inv1, bool Inv2) {
switch (Opcode) {
default: assert(0 && "Unknown opcode!"); abort();
case ISD::AND:
if (Inv1 && Inv2) return PPC::CRNOR; // De Morgan's Law
if (!Inv1 && !Inv2) return PPC::CRAND;
if (Inv1 ^ Inv2) return PPC::CRANDC;
case ISD::OR:
if (Inv1 && Inv2) return PPC::CRNAND; // De Morgan's Law
if (!Inv1 && !Inv2) return PPC::CROR;
if (Inv1 ^ Inv2) return PPC::CRORC;
}
return 0;
}
/// getCRIdxForSetCC - Return the index of the condition register field
/// associated with the SetCC condition, and whether or not the field is
/// treated as inverted. That is, lt = 0; ge = 0 inverted.
static unsigned getCRIdxForSetCC(unsigned Condition, bool& Inv) {
switch (Condition) {
default: assert(0 && "Unknown condition!"); abort();
case ISD::SETULT:
case ISD::SETLT: Inv = false; return 0;
case ISD::SETUGE:
case ISD::SETGE: Inv = true; return 0;
case ISD::SETUGT:
case ISD::SETGT: Inv = false; return 1;
case ISD::SETULE:
case ISD::SETLE: Inv = true; return 1;
case ISD::SETEQ: Inv = false; return 2;
case ISD::SETNE: Inv = true; return 2;
}
return 0;
}
/// IndexedOpForOp - Return the indexed variant for each of the PowerPC load
/// and store immediate instructions.
static unsigned IndexedOpForOp(unsigned Opcode) {
switch(Opcode) {
default: assert(0 && "Unknown opcode!"); abort();
case PPC::LBZ: return PPC::LBZX; case PPC::STB: return PPC::STBX;
case PPC::LHZ: return PPC::LHZX; case PPC::STH: return PPC::STHX;
case PPC::LHA: return PPC::LHAX; case PPC::STW: return PPC::STWX;
case PPC::LWZ: return PPC::LWZX; case PPC::STFS: return PPC::STFSX;
case PPC::LFS: return PPC::LFSX; case PPC::STFD: return PPC::STFDX;
case PPC::LFD: return PPC::LFDX;
}
return 0;
}
// Structure used to return the necessary information to codegen an SDIV as
// a multiply.
struct ms {
int m; // magic number
int s; // shift amount
};
struct mu {
unsigned int m; // magic number
int a; // add indicator
int s; // shift amount
};
/// magic - calculate the magic numbers required to codegen an integer sdiv as
/// a sequence of multiply and shifts. Requires that the divisor not be 0, 1,
/// or -1.
static struct ms magic(int d) {
int p;
unsigned int ad, anc, delta, q1, r1, q2, r2, t;
const unsigned int two31 = 0x80000000U;
struct ms mag;
ad = abs(d);
t = two31 + ((unsigned int)d >> 31);
anc = t - 1 - t%ad; // absolute value of nc
p = 31; // initialize p
q1 = two31/anc; // initialize q1 = 2p/abs(nc)
r1 = two31 - q1*anc; // initialize r1 = rem(2p,abs(nc))
q2 = two31/ad; // initialize q2 = 2p/abs(d)
r2 = two31 - q2*ad; // initialize r2 = rem(2p,abs(d))
do {
p = p + 1;
q1 = 2*q1; // update q1 = 2p/abs(nc)
r1 = 2*r1; // update r1 = rem(2p/abs(nc))
if (r1 >= anc) { // must be unsigned comparison
q1 = q1 + 1;
r1 = r1 - anc;
}
q2 = 2*q2; // update q2 = 2p/abs(d)
r2 = 2*r2; // update r2 = rem(2p/abs(d))
if (r2 >= ad) { // must be unsigned comparison
q2 = q2 + 1;
r2 = r2 - ad;
}
delta = ad - r2;
} while (q1 < delta || (q1 == delta && r1 == 0));
mag.m = q2 + 1;
if (d < 0) mag.m = -mag.m; // resulting magic number
mag.s = p - 32; // resulting shift
return mag;
}
/// magicu - calculate the magic numbers required to codegen an integer udiv as
/// a sequence of multiply, add and shifts. Requires that the divisor not be 0.
static struct mu magicu(unsigned d)
{
int p;
unsigned int nc, delta, q1, r1, q2, r2;
struct mu magu;
magu.a = 0; // initialize "add" indicator
nc = - 1 - (-d)%d;
p = 31; // initialize p
q1 = 0x80000000/nc; // initialize q1 = 2p/nc
r1 = 0x80000000 - q1*nc; // initialize r1 = rem(2p,nc)
q2 = 0x7FFFFFFF/d; // initialize q2 = (2p-1)/d
r2 = 0x7FFFFFFF - q2*d; // initialize r2 = rem((2p-1),d)
do {
p = p + 1;
if (r1 >= nc - r1 ) {
q1 = 2*q1 + 1; // update q1
r1 = 2*r1 - nc; // update r1
}
else {
q1 = 2*q1; // update q1
r1 = 2*r1; // update r1
}
if (r2 + 1 >= d - r2) {
if (q2 >= 0x7FFFFFFF) magu.a = 1;
q2 = 2*q2 + 1; // update q2
r2 = 2*r2 + 1 - d; // update r2
}
else {
if (q2 >= 0x80000000) magu.a = 1;
q2 = 2*q2; // update q2
r2 = 2*r2 + 1; // update r2
}
delta = d - 1 - r2;
} while (p < 64 && (q1 < delta || (q1 == delta && r1 == 0)));
magu.m = q2 + 1; // resulting magic number
magu.s = p - 32; // resulting shift
return magu;
}
}
/// BuildSDIVSequence - Given an ISD::SDIV node expressing a divide by constant,
/// return a DAG expression to select that will generate the same value by
/// multiplying by a magic number. See:
/// <http://the.wall.riscom.net/books/proc/ppc/cwg/code2.html>
SDOperand ISel::BuildSDIVSequence(SDOperand N) {
int d = (int)cast<ConstantSDNode>(N.getOperand(1))->getSignExtended();
ms magics = magic(d);
// Multiply the numerator (operand 0) by the magic value
SDOperand Q = ISelDAG->getNode(ISD::MULHS, MVT::i32, N.getOperand(0),
ISelDAG->getConstant(magics.m, MVT::i32));
// If d > 0 and m < 0, add the numerator
if (d > 0 && magics.m < 0)
Q = ISelDAG->getNode(ISD::ADD, MVT::i32, Q, N.getOperand(0));
// If d < 0 and m > 0, subtract the numerator.
if (d < 0 && magics.m > 0)
Q = ISelDAG->getNode(ISD::SUB, MVT::i32, Q, N.getOperand(0));
// Shift right algebraic if shift value is nonzero
if (magics.s > 0)
Q = ISelDAG->getNode(ISD::SRA, MVT::i32, Q,
ISelDAG->getConstant(magics.s, MVT::i32));
// Extract the sign bit and add it to the quotient
SDOperand T =
ISelDAG->getNode(ISD::SRL, MVT::i32, Q, ISelDAG->getConstant(31, MVT::i32));
return ISelDAG->getNode(ISD::ADD, MVT::i32, Q, T);
}
/// BuildUDIVSequence - Given an ISD::UDIV node expressing a divide by constant,
/// return a DAG expression to select that will generate the same value by
/// multiplying by a magic number. See:
/// <http://the.wall.riscom.net/books/proc/ppc/cwg/code2.html>
SDOperand ISel::BuildUDIVSequence(SDOperand N) {
unsigned d =
(unsigned)cast<ConstantSDNode>(N.getOperand(1))->getSignExtended();
mu magics = magicu(d);
// Multiply the numerator (operand 0) by the magic value
SDOperand Q = ISelDAG->getNode(ISD::MULHU, MVT::i32, N.getOperand(0),
ISelDAG->getConstant(magics.m, MVT::i32));
if (magics.a == 0) {
Q = ISelDAG->getNode(ISD::SRL, MVT::i32, Q,
ISelDAG->getConstant(magics.s, MVT::i32));
} else {
SDOperand NPQ = ISelDAG->getNode(ISD::SUB, MVT::i32, N.getOperand(0), Q);
NPQ = ISelDAG->getNode(ISD::SRL, MVT::i32, NPQ,
ISelDAG->getConstant(1, MVT::i32));
NPQ = ISelDAG->getNode(ISD::ADD, MVT::i32, NPQ, Q);
Q = ISelDAG->getNode(ISD::SRL, MVT::i32, NPQ,
ISelDAG->getConstant(magics.s-1, MVT::i32));
}
return Q;
}
/// getGlobalBaseReg - Output the instructions required to put the
/// base address to use for accessing globals into a register.
///
unsigned ISel::getGlobalBaseReg() {
if (!GlobalBaseInitialized) {
// Insert the set of GlobalBaseReg into the first MBB of the function
MachineBasicBlock &FirstMBB = BB->getParent()->front();
MachineBasicBlock::iterator MBBI = FirstMBB.begin();
GlobalBaseReg = MakeReg(MVT::i32);
BuildMI(FirstMBB, MBBI, PPC::MovePCtoLR, 0, PPC::LR);
BuildMI(FirstMBB, MBBI, PPC::MFLR, 1, GlobalBaseReg).addReg(PPC::LR);
GlobalBaseInitialized = true;
}
return GlobalBaseReg;
}
/// getConstDouble - Loads a floating point value into a register, via the
/// Constant Pool. Optionally takes a register in which to load the value.
unsigned ISel::getConstDouble(double doubleVal, unsigned Result=0) {
unsigned Tmp1 = MakeReg(MVT::i32);
if (0 == Result) Result = MakeReg(MVT::f64);
MachineConstantPool *CP = BB->getParent()->getConstantPool();
ConstantFP *CFP = ConstantFP::get(Type::DoubleTy, doubleVal);
unsigned CPI = CP->getConstantPoolIndex(CFP);
if (PICEnabled)
BuildMI(BB, PPC::ADDIS, 2, Tmp1).addReg(getGlobalBaseReg())
.addConstantPoolIndex(CPI);
else
BuildMI(BB, PPC::LIS, 1, Tmp1).addConstantPoolIndex(CPI);
BuildMI(BB, PPC::LFD, 2, Result).addConstantPoolIndex(CPI).addReg(Tmp1);
return Result;
}
/// MoveCRtoGPR - Move CCReg[Idx] to the least significant bit of Result. If
/// Inv is true, then invert the result.
void ISel::MoveCRtoGPR(unsigned CCReg, bool Inv, unsigned Idx, unsigned Result){
unsigned IntCR = MakeReg(MVT::i32);
BuildMI(BB, PPC::MCRF, 1, PPC::CR7).addReg(CCReg);
BuildMI(BB, GPOPT ? PPC::MFOCRF : PPC::MFCR, 1, IntCR).addReg(PPC::CR7);
if (Inv) {
unsigned Tmp1 = MakeReg(MVT::i32);
BuildMI(BB, PPC::RLWINM, 4, Tmp1).addReg(IntCR).addImm(32-(3-Idx))
.addImm(31).addImm(31);
BuildMI(BB, PPC::XORI, 2, Result).addReg(Tmp1).addImm(1);
} else {
BuildMI(BB, PPC::RLWINM, 4, Result).addReg(IntCR).addImm(32-(3-Idx))
.addImm(31).addImm(31);
}
}
/// SelectBitfieldInsert - turn an or of two masked values into
/// the rotate left word immediate then mask insert (rlwimi) instruction.
/// Returns true on success, false if the caller still needs to select OR.
///
/// Patterns matched:
/// 1. or shl, and 5. or and, and
/// 2. or and, shl 6. or shl, shr
/// 3. or shr, and 7. or shr, shl
/// 4. or and, shr
bool ISel::SelectBitfieldInsert(SDOperand OR, unsigned Result) {
bool IsRotate = false;
unsigned TgtMask = 0xFFFFFFFF, InsMask = 0xFFFFFFFF, Amount = 0;
SDOperand Op0 = OR.getOperand(0);
SDOperand Op1 = OR.getOperand(1);
unsigned Op0Opc = Op0.getOpcode();
unsigned Op1Opc = Op1.getOpcode();
// Verify that we have the correct opcodes
if (ISD::SHL != Op0Opc && ISD::SRL != Op0Opc && ISD::AND != Op0Opc)
return false;
if (ISD::SHL != Op1Opc && ISD::SRL != Op1Opc && ISD::AND != Op1Opc)
return false;
// Generate Mask value for Target
if (ConstantSDNode *CN =
dyn_cast<ConstantSDNode>(Op0.getOperand(1).Val)) {
switch(Op0Opc) {
case ISD::SHL: TgtMask <<= (unsigned)CN->getValue(); break;
case ISD::SRL: TgtMask >>= (unsigned)CN->getValue(); break;
case ISD::AND: TgtMask &= (unsigned)CN->getValue(); break;
}
} else {
return false;
}
// Generate Mask value for Insert
if (ConstantSDNode *CN =
dyn_cast<ConstantSDNode>(Op1.getOperand(1).Val)) {
switch(Op1Opc) {
case ISD::SHL:
Amount = CN->getValue();
InsMask <<= Amount;
if (Op0Opc == ISD::SRL) IsRotate = true;
break;
case ISD::SRL:
Amount = CN->getValue();
InsMask >>= Amount;
Amount = 32-Amount;
if (Op0Opc == ISD::SHL) IsRotate = true;
break;
case ISD::AND:
InsMask &= (unsigned)CN->getValue();
break;
}
} else {
return false;
}
unsigned Tmp3 = 0;
// If both of the inputs are ANDs and one of them has a logical shift by
// constant as its input, make that the inserted value so that we can combine
// the shift into the rotate part of the rlwimi instruction
if (Op0Opc == ISD::AND && Op1Opc == ISD::AND) {
if (Op1.getOperand(0).getOpcode() == ISD::SHL ||
Op1.getOperand(0).getOpcode() == ISD::SRL) {
if (ConstantSDNode *CN =
dyn_cast<ConstantSDNode>(Op1.getOperand(0).getOperand(1).Val)) {
Amount = Op1.getOperand(0).getOpcode() == ISD::SHL ?
CN->getValue() : 32 - CN->getValue();
Tmp3 = SelectExpr(Op1.getOperand(0).getOperand(0));
}
} else if (Op0.getOperand(0).getOpcode() == ISD::SHL ||
Op0.getOperand(0).getOpcode() == ISD::SRL) {
if (ConstantSDNode *CN =
dyn_cast<ConstantSDNode>(Op0.getOperand(0).getOperand(1).Val)) {
std::swap(Op0, Op1);
std::swap(TgtMask, InsMask);
Amount = Op1.getOperand(0).getOpcode() == ISD::SHL ?
CN->getValue() : 32 - CN->getValue();
Tmp3 = SelectExpr(Op1.getOperand(0).getOperand(0));
}
}
}
// Verify that the Target mask and Insert mask together form a full word mask
// and that the Insert mask is a run of set bits (which implies both are runs
// of set bits). Given that, Select the arguments and generate the rlwimi
// instruction.
unsigned MB, ME;
if (((TgtMask & InsMask) == 0) && IsRunOfOnes(InsMask, MB, ME)) {
unsigned Tmp1, Tmp2;
bool fullMask = (TgtMask ^ InsMask) == 0xFFFFFFFF;
// Check for rotlwi / rotrwi here, a special case of bitfield insert
// where both bitfield halves are sourced from the same value.
if (IsRotate && fullMask &&
OR.getOperand(0).getOperand(0) == OR.getOperand(1).getOperand(0)) {
Tmp1 = SelectExpr(OR.getOperand(0).getOperand(0));
BuildMI(BB, PPC::RLWINM, 4, Result).addReg(Tmp1).addImm(Amount)
.addImm(0).addImm(31);
return true;
}
if (Op0Opc == ISD::AND && fullMask)
Tmp1 = SelectExpr(Op0.getOperand(0));
else
Tmp1 = SelectExpr(Op0);
Tmp2 = Tmp3 ? Tmp3 : SelectExpr(Op1.getOperand(0));
BuildMI(BB, PPC::RLWIMI, 5, Result).addReg(Tmp1).addReg(Tmp2)
.addImm(Amount).addImm(MB).addImm(ME);
return true;
}
return false;
}
/// FoldIfWideZeroExtend - 32 bit PowerPC implicit masks shift amounts to the
/// low six bits. If the shift amount is an ISD::AND node with a mask that is
/// wider than the implicit mask, then we can get rid of the AND and let the
/// shift do the mask.
unsigned ISel::FoldIfWideZeroExtend(SDOperand N) {
unsigned C;
if (N.getOpcode() == ISD::AND &&
5 == getImmediateForOpcode(N.getOperand(1), ISD::AND, C) && // isMask
31 == (C & 0xFFFF) && // ME
26 >= (C >> 16)) // MB
return SelectExpr(N.getOperand(0));
else
return SelectExpr(N);
}
unsigned ISel::SelectCC(SDOperand CC, unsigned& Opc, bool &Inv, unsigned& Idx) {
unsigned Result, Tmp1, Tmp2;
bool AlreadySelected = false;
static const unsigned CompareOpcodes[] =
{ PPC::FCMPU, PPC::FCMPU, PPC::CMPW, PPC::CMPLW };
// Allocate a condition register for this expression
Result = RegMap->createVirtualRegister(PPC32::CRRCRegisterClass);
// If the first operand to the select is a SETCC node, then we can fold it
// into the branch that selects which value to return.
if (SetCCSDNode* SetCC = dyn_cast<SetCCSDNode>(CC.Val)) {
bool U;
Opc = getBCCForSetCC(SetCC->getCondition(), U);
Idx = getCRIdxForSetCC(SetCC->getCondition(), Inv);
// Pass the optional argument U to getImmediateForOpcode for SETCC,
// so that it knows whether the SETCC immediate range is signed or not.
if (1 == getImmediateForOpcode(SetCC->getOperand(1), ISD::SETCC,
Tmp2, U)) {
// For comparisons against zero, we can implicity set CR0 if a recording
// variant (e.g. 'or.' instead of 'or') of the instruction that defines
// operand zero of the SetCC node is available.
if (0 == Tmp2 &&
NodeHasRecordingVariant(SetCC->getOperand(0).getOpcode()) &&
SetCC->getOperand(0).Val->hasOneUse()) {
RecordSuccess = false;
Tmp1 = SelectExpr(SetCC->getOperand(0), true);
if (RecordSuccess) {
++Recorded;
BuildMI(BB, PPC::MCRF, 1, Result).addReg(PPC::CR0);
return Result;
}
AlreadySelected = true;
}
// If we could not implicitly set CR0, then emit a compare immediate
// instead.
if (!AlreadySelected) Tmp1 = SelectExpr(SetCC->getOperand(0));
if (U)
BuildMI(BB, PPC::CMPLWI, 2, Result).addReg(Tmp1).addImm(Tmp2);
else
BuildMI(BB, PPC::CMPWI, 2, Result).addReg(Tmp1).addSImm(Tmp2);
} else {
bool IsInteger = MVT::isInteger(SetCC->getOperand(0).getValueType());
unsigned CompareOpc = CompareOpcodes[2 * IsInteger + U];
Tmp1 = SelectExpr(SetCC->getOperand(0));
Tmp2 = SelectExpr(SetCC->getOperand(1));
BuildMI(BB, CompareOpc, 2, Result).addReg(Tmp1).addReg(Tmp2);
}
} else {
// If this isn't a SetCC, then select the value and compare it against zero,
// treating it as if it were a boolean.
Opc = PPC::BNE;
Idx = getCRIdxForSetCC(ISD::SETNE, Inv);
Tmp1 = SelectExpr(CC);
BuildMI(BB, PPC::CMPLWI, 2, Result).addReg(Tmp1).addImm(0);
}
return Result;
}
unsigned ISel::SelectCCExpr(SDOperand N, unsigned& Opc, bool &Inv,
unsigned &Idx) {
bool Inv0, Inv1;
unsigned Idx0, Idx1, CROpc, Opc1, Tmp1, Tmp2;
// Allocate a condition register for this expression
unsigned Result = RegMap->createVirtualRegister(PPC32::CRRCRegisterClass);
// Check for the operations we support:
switch(N.getOpcode()) {
default:
Opc = PPC::BNE;
Idx = getCRIdxForSetCC(ISD::SETNE, Inv);
Tmp1 = SelectExpr(N);
BuildMI(BB, PPC::CMPLWI, 2, Result).addReg(Tmp1).addImm(0);
break;
case ISD::OR:
case ISD::AND:
Tmp1 = SelectCCExpr(N.getOperand(0), Opc, Inv0, Idx0);
Tmp2 = SelectCCExpr(N.getOperand(1), Opc1, Inv1, Idx1);
CROpc = getCROpForSetCC(N.getOpcode(), Inv0, Inv1);
if (Inv0 && !Inv1) {
std::swap(Tmp1, Tmp2);
std::swap(Idx0, Idx1);
Opc = Opc1;
}
if (Inv0 && Inv1) Opc = PPC32InstrInfo::invertPPCBranchOpcode(Opc);
BuildMI(BB, CROpc, 5, Result).addImm(Idx0).addReg(Tmp1).addImm(Idx0)
.addReg(Tmp2).addImm(Idx1);
Inv = false;
Idx = Idx0;
break;
case ISD::SETCC:
Tmp1 = SelectCC(N, Opc, Inv, Idx);
Result = Tmp1;
break;
}
return Result;
}
/// Check to see if the load is a constant offset from a base register
unsigned ISel::SelectAddr(SDOperand N, unsigned& Reg, int& offset)
{
unsigned imm = 0, opcode = N.getOpcode();
if (N.getOpcode() == ISD::ADD) {
bool isFrame = N.getOperand(0).getOpcode() == ISD::FrameIndex;
if (1 == getImmediateForOpcode(N.getOperand(1), opcode, imm)) {
offset = imm;
if (isFrame) {
++FrameOff;
Reg = cast<FrameIndexSDNode>(N.getOperand(0))->getIndex();
return 1;
} else {
Reg = SelectExpr(N.getOperand(0));
return 0;
}
} else {
Reg = SelectExpr(N.getOperand(0));
offset = SelectExpr(N.getOperand(1));
return 2;
}
}
Reg = SelectExpr(N);
offset = 0;
return 0;
}
void ISel::SelectBranchCC(SDOperand N)
{
MachineBasicBlock *Dest =
cast<BasicBlockSDNode>(N.getOperand(2))->getBasicBlock();
bool Inv;
unsigned Opc, CCReg, Idx;
Select(N.getOperand(0)); //chain
CCReg = SelectCC(N.getOperand(1), Opc, Inv, Idx);
// Iterate to the next basic block
ilist<MachineBasicBlock>::iterator It = BB;
++It;
// If this is a two way branch, then grab the fallthrough basic block argument
// and build a PowerPC branch pseudo-op, suitable for long branch conversion
// if necessary by the branch selection pass. Otherwise, emit a standard
// conditional branch.
if (N.getOpcode() == ISD::BRCONDTWOWAY) {
MachineBasicBlock *Fallthrough =
cast<BasicBlockSDNode>(N.getOperand(3))->getBasicBlock();
if (Dest != It) {
BuildMI(BB, PPC::COND_BRANCH, 4).addReg(CCReg).addImm(Opc)
.addMBB(Dest).addMBB(Fallthrough);
if (Fallthrough != It)
BuildMI(BB, PPC::B, 1).addMBB(Fallthrough);
} else {
if (Fallthrough != It) {
Opc = PPC32InstrInfo::invertPPCBranchOpcode(Opc);
BuildMI(BB, PPC::COND_BRANCH, 4).addReg(CCReg).addImm(Opc)
.addMBB(Fallthrough).addMBB(Dest);
}
}
} else {
// If the fallthrough path is off the end of the function, which would be
// undefined behavior, set it to be the same as the current block because
// we have nothing better to set it to, and leaving it alone will cause the
// PowerPC Branch Selection pass to crash.
if (It == BB->getParent()->end()) It = Dest;
BuildMI(BB, PPC::COND_BRANCH, 4).addReg(CCReg).addImm(Opc)
.addMBB(Dest).addMBB(It);
}
return;
}
unsigned ISel::SelectExpr(SDOperand N, bool Recording) {
unsigned Result;
unsigned Tmp1, Tmp2, Tmp3;
unsigned Opc = 0;
unsigned opcode = N.getOpcode();
SDNode *Node = N.Val;
MVT::ValueType DestType = N.getValueType();
if (Node->getOpcode() == ISD::CopyFromReg &&
(MRegisterInfo::isVirtualRegister(cast<RegSDNode>(Node)->getReg()) ||
cast<RegSDNode>(Node)->getReg() == PPC::R1))
// Just use the specified register as our input.
return cast<RegSDNode>(Node)->getReg();
unsigned &Reg = ExprMap[N];
if (Reg) return Reg;
switch (N.getOpcode()) {
default:
Reg = Result = (N.getValueType() != MVT::Other) ?
MakeReg(N.getValueType()) : 1;
break;
case ISD::TAILCALL:
case ISD::CALL:
// If this is a call instruction, make sure to prepare ALL of the result
// values as well as the chain.
if (Node->getNumValues() == 1)
Reg = Result = 1; // Void call, just a chain.
else {
Result = MakeReg(Node->getValueType(0));
ExprMap[N.getValue(0)] = Result;
for (unsigned i = 1, e = N.Val->getNumValues()-1; i != e; ++i)
ExprMap[N.getValue(i)] = MakeReg(Node->getValueType(i));
ExprMap[SDOperand(Node, Node->getNumValues()-1)] = 1;
}
break;
case ISD::ADD_PARTS:
case ISD::SUB_PARTS:
case ISD::SHL_PARTS:
case ISD::SRL_PARTS:
case ISD::SRA_PARTS:
Result = MakeReg(Node->getValueType(0));
ExprMap[N.getValue(0)] = Result;
for (unsigned i = 1, e = N.Val->getNumValues(); i != e; ++i)
ExprMap[N.getValue(i)] = MakeReg(Node->getValueType(i));
break;
}
switch (opcode) {
default:
Node->dump();
assert(0 && "Node not handled!\n");
case ISD::UNDEF:
BuildMI(BB, PPC::IMPLICIT_DEF, 0, Result);
return Result;
case ISD::DYNAMIC_STACKALLOC:
// Generate both result values. FIXME: Need a better commment here?
if (Result != 1)
ExprMap[N.getValue(1)] = 1;
else
Result = ExprMap[N.getValue(0)] = MakeReg(N.getValue(0).getValueType());
// FIXME: We are currently ignoring the requested alignment for handling
// greater than the stack alignment. This will need to be revisited at some
// point. Align = N.getOperand(2);
if (!isa<ConstantSDNode>(N.getOperand(2)) ||
cast<ConstantSDNode>(N.getOperand(2))->getValue() != 0) {
std::cerr << "Cannot allocate stack object with greater alignment than"
<< " the stack alignment yet!";
abort();
}
Select(N.getOperand(0));
Tmp1 = SelectExpr(N.getOperand(1));
// Subtract size from stack pointer, thereby allocating some space.
BuildMI(BB, PPC::SUBF, 2, PPC::R1).addReg(Tmp1).addReg(PPC::R1);
// Put a pointer to the space into the result register by copying the SP
BuildMI(BB, PPC::OR, 2, Result).addReg(PPC::R1).addReg(PPC::R1);
return Result;
case ISD::ConstantPool:
Tmp1 = cast<ConstantPoolSDNode>(N)->getIndex();
Tmp2 = MakeReg(MVT::i32);
if (PICEnabled)
BuildMI(BB, PPC::ADDIS, 2, Tmp2).addReg(getGlobalBaseReg())
.addConstantPoolIndex(Tmp1);
else
BuildMI(BB, PPC::LIS, 1, Tmp2).addConstantPoolIndex(Tmp1);
BuildMI(BB, PPC::LA, 2, Result).addReg(Tmp2).addConstantPoolIndex(Tmp1);
return Result;
case ISD::FrameIndex:
Tmp1 = cast<FrameIndexSDNode>(N)->getIndex();
addFrameReference(BuildMI(BB, PPC::ADDI, 2, Result), (int)Tmp1, 0, false);
return Result;
case ISD::GlobalAddress: {
GlobalValue *GV = cast<GlobalAddressSDNode>(N)->getGlobal();
Tmp1 = MakeReg(MVT::i32);
if (PICEnabled)
BuildMI(BB, PPC::ADDIS, 2, Tmp1).addReg(getGlobalBaseReg())
.addGlobalAddress(GV);
else
BuildMI(BB, PPC::LIS, 1, Tmp1).addGlobalAddress(GV);
if (GV->hasWeakLinkage() || GV->isExternal()) {
BuildMI(BB, PPC::LWZ, 2, Result).addGlobalAddress(GV).addReg(Tmp1);
} else {
BuildMI(BB, PPC::LA, 2, Result).addReg(Tmp1).addGlobalAddress(GV);
}
return Result;
}
case ISD::LOAD:
case ISD::EXTLOAD:
case ISD::ZEXTLOAD:
case ISD::SEXTLOAD: {
MVT::ValueType TypeBeingLoaded = (ISD::LOAD == opcode) ?
Node->getValueType(0) : cast<VTSDNode>(Node->getOperand(3))->getVT();
bool sext = (ISD::SEXTLOAD == opcode);
// Make sure we generate both values.
if (Result != 1)
ExprMap[N.getValue(1)] = 1; // Generate the token
else
Result = ExprMap[N.getValue(0)] = MakeReg(N.getValue(0).getValueType());
SDOperand Chain = N.getOperand(0);
SDOperand Address = N.getOperand(1);
Select(Chain);
switch (TypeBeingLoaded) {
default: Node->dump(); assert(0 && "Cannot load this type!");
case MVT::i1: Opc = PPC::LBZ; break;
case MVT::i8: Opc = PPC::LBZ; break;
case MVT::i16: Opc = sext ? PPC::LHA : PPC::LHZ; break;
case MVT::i32: Opc = PPC::LWZ; break;
case MVT::f32: Opc = PPC::LFS; break;
case MVT::f64: Opc = PPC::LFD; break;
}
if (ConstantPoolSDNode *CP = dyn_cast<ConstantPoolSDNode>(Address)) {
Tmp1 = MakeReg(MVT::i32);
int CPI = CP->getIndex();
if (PICEnabled)
BuildMI(BB, PPC::ADDIS, 2, Tmp1).addReg(getGlobalBaseReg())
.addConstantPoolIndex(CPI);
else
BuildMI(BB, PPC::LIS, 1, Tmp1).addConstantPoolIndex(CPI);
BuildMI(BB, Opc, 2, Result).addConstantPoolIndex(CPI).addReg(Tmp1);
} else if (Address.getOpcode() == ISD::FrameIndex) {
Tmp1 = cast<FrameIndexSDNode>(Address)->getIndex();
addFrameReference(BuildMI(BB, Opc, 2, Result), (int)Tmp1);
} else if(GlobalAddressSDNode *GN = dyn_cast<GlobalAddressSDNode>(Address)){
GlobalValue *GV = GN->getGlobal();
Tmp1 = MakeReg(MVT::i32);
if (PICEnabled)
BuildMI(BB, PPC::ADDIS, 2, Tmp1).addReg(getGlobalBaseReg())
.addGlobalAddress(GV);
else
BuildMI(BB, PPC::LIS, 1, Tmp1).addGlobalAddress(GV);
if (GV->hasWeakLinkage() || GV->isExternal()) {
Tmp2 = MakeReg(MVT::i32);
BuildMI(BB, PPC::LWZ, 2, Tmp2).addGlobalAddress(GV).addReg(Tmp1);
BuildMI(BB, Opc, 2, Result).addSImm(0).addReg(Tmp2);
} else {
BuildMI(BB, Opc, 2, Result).addGlobalAddress(GV).addReg(Tmp1);
}
} else {
int offset;
switch(SelectAddr(Address, Tmp1, offset)) {
default: assert(0 && "Unhandled return value from SelectAddr");
case 0: // imm offset, no frame, no index
BuildMI(BB, Opc, 2, Result).addSImm(offset).addReg(Tmp1);
break;
case 1: // imm offset + frame index
addFrameReference(BuildMI(BB, Opc, 2, Result), (int)Tmp1, offset);
break;
case 2: // base+index addressing
Opc = IndexedOpForOp(Opc);
BuildMI(BB, Opc, 2, Result).addReg(Tmp1).addReg(offset);
break;
}
}
return Result;
}
case ISD::TAILCALL:
case ISD::CALL: {
unsigned GPR_idx = 0, FPR_idx = 0;
static const unsigned GPR[] = {
PPC::R3, PPC::R4, PPC::R5, PPC::R6,
PPC::R7, PPC::R8, PPC::R9, PPC::R10,
};
static const unsigned FPR[] = {
PPC::F1, PPC::F2, PPC::F3, PPC::F4, PPC::F5, PPC::F6, PPC::F7,
PPC::F8, PPC::F9, PPC::F10, PPC::F11, PPC::F12, PPC::F13
};
// Lower the chain for this call.
Select(N.getOperand(0));
ExprMap[N.getValue(Node->getNumValues()-1)] = 1;
MachineInstr *CallMI;
// Emit the correct call instruction based on the type of symbol called.
if (GlobalAddressSDNode *GASD =
dyn_cast<GlobalAddressSDNode>(N.getOperand(1))) {
CallMI = BuildMI(PPC::CALLpcrel, 1).addGlobalAddress(GASD->getGlobal(),
true);
} else if (ExternalSymbolSDNode *ESSDN =
dyn_cast<ExternalSymbolSDNode>(N.getOperand(1))) {
CallMI = BuildMI(PPC::CALLpcrel, 1).addExternalSymbol(ESSDN->getSymbol(),
true);
} else {
Tmp1 = SelectExpr(N.getOperand(1));
BuildMI(BB, PPC::OR, 2, PPC::R12).addReg(Tmp1).addReg(Tmp1);
BuildMI(BB, PPC::MTCTR, 1).addReg(PPC::R12);
CallMI = BuildMI(PPC::CALLindirect, 3).addImm(20).addImm(0)
.addReg(PPC::R12);
}
// Load the register args to virtual regs
std::vector<unsigned> ArgVR;
for(int i = 2, e = Node->getNumOperands(); i < e; ++i)
ArgVR.push_back(SelectExpr(N.getOperand(i)));
// Copy the virtual registers into the appropriate argument register
for(int i = 0, e = ArgVR.size(); i < e; ++i) {
switch(N.getOperand(i+2).getValueType()) {
default: Node->dump(); assert(0 && "Unknown value type for call");
case MVT::i1:
case MVT::i8:
case MVT::i16:
case MVT::i32:
assert(GPR_idx < 8 && "Too many int args");
if (N.getOperand(i+2).getOpcode() != ISD::UNDEF) {
BuildMI(BB, PPC::OR,2,GPR[GPR_idx]).addReg(ArgVR[i]).addReg(ArgVR[i]);
CallMI->addRegOperand(GPR[GPR_idx], MachineOperand::Use);
}
++GPR_idx;
break;
case MVT::f64:
case MVT::f32:
assert(FPR_idx < 13 && "Too many fp args");
BuildMI(BB, PPC::FMR, 1, FPR[FPR_idx]).addReg(ArgVR[i]);
CallMI->addRegOperand(FPR[FPR_idx], MachineOperand::Use);
++FPR_idx;
break;
}
}
// Put the call instruction in the correct place in the MachineBasicBlock
BB->push_back(CallMI);
switch (Node->getValueType(0)) {
default: assert(0 && "Unknown value type for call result!");
case MVT::Other: return 1;
case MVT::i1:
case MVT::i8:
case MVT::i16:
case MVT::i32:
if (Node->getValueType(1) == MVT::i32) {
BuildMI(BB, PPC::OR, 2, Result+1).addReg(PPC::R3).addReg(PPC::R3);
BuildMI(BB, PPC::OR, 2, Result).addReg(PPC::R4).addReg(PPC::R4);
} else {
BuildMI(BB, PPC::OR, 2, Result).addReg(PPC::R3).addReg(PPC::R3);
}
break;
case MVT::f32:
case MVT::f64:
BuildMI(BB, PPC::FMR, 1, Result).addReg(PPC::F1);
break;
}
return Result+N.ResNo;
}
case ISD::SIGN_EXTEND:
case ISD::SIGN_EXTEND_INREG:
Tmp1 = SelectExpr(N.getOperand(0));
switch(cast<VTSDNode>(Node->getOperand(1))->getVT()) {
default: Node->dump(); assert(0 && "Unhandled SIGN_EXTEND type"); break;
case MVT::i16:
BuildMI(BB, PPC::EXTSH, 1, Result).addReg(Tmp1);
break;
case MVT::i8:
BuildMI(BB, PPC::EXTSB, 1, Result).addReg(Tmp1);
break;
case MVT::i1:
BuildMI(BB, PPC::SUBFIC, 2, Result).addReg(Tmp1).addSImm(0);
break;
}
return Result;
case ISD::CopyFromReg:
DestType = N.getValue(0).getValueType();
if (Result == 1)
Result = ExprMap[N.getValue(0)] = MakeReg(DestType);
Tmp1 = dyn_cast<RegSDNode>(Node)->getReg();
if (MVT::isInteger(DestType))
BuildMI(BB, PPC::OR, 2, Result).addReg(Tmp1).addReg(Tmp1);
else
BuildMI(BB, PPC::FMR, 1, Result).addReg(Tmp1);
return Result;
case ISD::SHL:
Tmp1 = SelectExpr(N.getOperand(0));
if (ConstantSDNode *CN = dyn_cast<ConstantSDNode>(N.getOperand(1))) {
Tmp2 = CN->getValue() & 0x1F;
BuildMI(BB, PPC::RLWINM, 4, Result).addReg(Tmp1).addImm(Tmp2).addImm(0)
.addImm(31-Tmp2);
} else {
Tmp2 = FoldIfWideZeroExtend(N.getOperand(1));
BuildMI(BB, PPC::SLW, 2, Result).addReg(Tmp1).addReg(Tmp2);
}
return Result;
case ISD::SRL:
Tmp1 = SelectExpr(N.getOperand(0));
if (ConstantSDNode *CN = dyn_cast<ConstantSDNode>(N.getOperand(1))) {
Tmp2 = CN->getValue() & 0x1F;
BuildMI(BB, PPC::RLWINM, 4, Result).addReg(Tmp1).addImm(32-Tmp2)
.addImm(Tmp2).addImm(31);
} else {
Tmp2 = FoldIfWideZeroExtend(N.getOperand(1));
BuildMI(BB, PPC::SRW, 2, Result).addReg(Tmp1).addReg(Tmp2);
}
return Result;
case ISD::SRA:
Tmp1 = SelectExpr(N.getOperand(0));
if (ConstantSDNode *CN = dyn_cast<ConstantSDNode>(N.getOperand(1))) {
Tmp2 = CN->getValue() & 0x1F;
BuildMI(BB, PPC::SRAWI, 2, Result).addReg(Tmp1).addImm(Tmp2);
} else {
Tmp2 = FoldIfWideZeroExtend(N.getOperand(1));
BuildMI(BB, PPC::SRAW, 2, Result).addReg(Tmp1).addReg(Tmp2);
}
return Result;
case ISD::CTLZ:
Tmp1 = SelectExpr(N.getOperand(0));
BuildMI(BB, PPC::CNTLZW, 1, Result).addReg(Tmp1);
return Result;
case ISD::ADD:
if (!MVT::isInteger(DestType)) {
if (!NoExcessFPPrecision && N.getOperand(0).getOpcode() == ISD::MUL &&
N.getOperand(0).Val->hasOneUse()) {
++FusedFP; // Statistic
Tmp1 = SelectExpr(N.getOperand(0).getOperand(0));
Tmp2 = SelectExpr(N.getOperand(0).getOperand(1));
Tmp3 = SelectExpr(N.getOperand(1));
Opc = DestType == MVT::f64 ? PPC::FMADD : PPC::FMADDS;
BuildMI(BB, Opc, 3, Result).addReg(Tmp1).addReg(Tmp2).addReg(Tmp3);
return Result;
}
if (!NoExcessFPPrecision && N.getOperand(1).getOpcode() == ISD::MUL &&
N.getOperand(1).Val->hasOneUse()) {
++FusedFP; // Statistic
Tmp1 = SelectExpr(N.getOperand(1).getOperand(0));
Tmp2 = SelectExpr(N.getOperand(1).getOperand(1));
Tmp3 = SelectExpr(N.getOperand(0));
Opc = DestType == MVT::f64 ? PPC::FMADD : PPC::FMADDS;
BuildMI(BB, Opc, 3, Result).addReg(Tmp1).addReg(Tmp2).addReg(Tmp3);
return Result;
}
Opc = DestType == MVT::f64 ? PPC::FADD : PPC::FADDS;
Tmp1 = SelectExpr(N.getOperand(0));
Tmp2 = SelectExpr(N.getOperand(1));
BuildMI(BB, Opc, 2, Result).addReg(Tmp1).addReg(Tmp2);
return Result;
}
Tmp1 = SelectExpr(N.getOperand(0));
switch(getImmediateForOpcode(N.getOperand(1), opcode, Tmp2)) {
default: assert(0 && "unhandled result code");
case 0: // No immediate
Tmp2 = SelectExpr(N.getOperand(1));
BuildMI(BB, PPC::ADD, 2, Result).addReg(Tmp1).addReg(Tmp2);
break;
case 1: // Low immediate
BuildMI(BB, PPC::ADDI, 2, Result).addReg(Tmp1).addSImm(Tmp2);
break;
case 2: // Shifted immediate
BuildMI(BB, PPC::ADDIS, 2, Result).addReg(Tmp1).addSImm(Tmp2);
break;
}
return Result;
case ISD::AND:
switch(getImmediateForOpcode(N.getOperand(1), opcode, Tmp2)) {
default: assert(0 && "unhandled result code");
case 0: // No immediate
// Check for andc: and, (xor a, -1), b
if (N.getOperand(0).getOpcode() == ISD::XOR &&
N.getOperand(0).getOperand(1).getOpcode() == ISD::Constant &&
cast<ConstantSDNode>(N.getOperand(0).getOperand(1))->isAllOnesValue()) {
Tmp1 = SelectExpr(N.getOperand(0).getOperand(0));
Tmp2 = SelectExpr(N.getOperand(1));
BuildMI(BB, PPC::ANDC, 2, Result).addReg(Tmp2).addReg(Tmp1);
return Result;
}
// It wasn't and-with-complement, emit a regular and
Tmp1 = SelectExpr(N.getOperand(0));
Tmp2 = SelectExpr(N.getOperand(1));
Opc = Recording ? PPC::ANDo : PPC::AND;
BuildMI(BB, Opc, 2, Result).addReg(Tmp1).addReg(Tmp2);
break;
case 1: // Low immediate
Tmp1 = SelectExpr(N.getOperand(0));
BuildMI(BB, PPC::ANDIo, 2, Result).addReg(Tmp1).addImm(Tmp2);
break;
case 2: // Shifted immediate
Tmp1 = SelectExpr(N.getOperand(0));
BuildMI(BB, PPC::ANDISo, 2, Result).addReg(Tmp1).addImm(Tmp2);
break;
case 5: // Bitfield mask
Opc = Recording ? PPC::RLWINMo : PPC::RLWINM;
Tmp3 = Tmp2 >> 16; // MB
Tmp2 &= 0xFFFF; // ME
// FIXME: Catch SHL-AND in addition to SRL-AND in this block.
if (N.getOperand(0).getOpcode() == ISD::SRL)
if (ConstantSDNode *SA =
dyn_cast<ConstantSDNode>(N.getOperand(0).getOperand(1))) {
// We can fold the RLWINM and the SRL together if the mask is
// clearing the top bits which are rotated around.
unsigned RotAmt = 32-(SA->getValue() & 31);
if (Tmp2 <= RotAmt) {
Tmp1 = SelectExpr(N.getOperand(0).getOperand(0));
BuildMI(BB, Opc, 4, Result).addReg(Tmp1).addImm(RotAmt)
.addImm(Tmp3).addImm(Tmp2);
break;
}
}
Tmp1 = SelectExpr(N.getOperand(0));
BuildMI(BB, Opc, 4, Result).addReg(Tmp1).addImm(0)
.addImm(Tmp3).addImm(Tmp2);
break;
}
RecordSuccess = true;
return Result;
case ISD::OR:
if (SelectBitfieldInsert(N, Result))
return Result;
Tmp1 = SelectExpr(N.getOperand(0));
switch(getImmediateForOpcode(N.getOperand(1), opcode, Tmp2)) {
default: assert(0 && "unhandled result code");
case 0: // No immediate
Tmp2 = SelectExpr(N.getOperand(1));
Opc = Recording ? PPC::ORo : PPC::OR;
RecordSuccess = true;
BuildMI(BB, Opc, 2, Result).addReg(Tmp1).addReg(Tmp2);
break;
case 1: // Low immediate
BuildMI(BB, PPC::ORI, 2, Result).addReg(Tmp1).addImm(Tmp2);
break;
case 2: // Shifted immediate
BuildMI(BB, PPC::ORIS, 2, Result).addReg(Tmp1).addImm(Tmp2);
break;
}
return Result;
case ISD::XOR: {
// Check for EQV: xor, (xor a, -1), b
if (N.getOperand(0).getOpcode() == ISD::XOR &&
N.getOperand(0).getOperand(1).getOpcode() == ISD::Constant &&
cast<ConstantSDNode>(N.getOperand(0).getOperand(1))->isAllOnesValue()) {
Tmp1 = SelectExpr(N.getOperand(0).getOperand(0));
Tmp2 = SelectExpr(N.getOperand(1));
BuildMI(BB, PPC::EQV, 2, Result).addReg(Tmp1).addReg(Tmp2);
return Result;
}
// Check for NOT, NOR, EQV, and NAND: xor (copy, or, xor, and), -1
if (N.getOperand(1).getOpcode() == ISD::Constant &&
cast<ConstantSDNode>(N.getOperand(1))->isAllOnesValue()) {
switch(N.getOperand(0).getOpcode()) {
case ISD::OR:
Tmp1 = SelectExpr(N.getOperand(0).getOperand(0));
Tmp2 = SelectExpr(N.getOperand(0).getOperand(1));
BuildMI(BB, PPC::NOR, 2, Result).addReg(Tmp1).addReg(Tmp2);
break;
case ISD::AND:
Tmp1 = SelectExpr(N.getOperand(0).getOperand(0));
Tmp2 = SelectExpr(N.getOperand(0).getOperand(1));
BuildMI(BB, PPC::NAND, 2, Result).addReg(Tmp1).addReg(Tmp2);
break;
case ISD::XOR:
Tmp1 = SelectExpr(N.getOperand(0).getOperand(0));
Tmp2 = SelectExpr(N.getOperand(0).getOperand(1));
BuildMI(BB, PPC::EQV, 2, Result).addReg(Tmp1).addReg(Tmp2);
break;
default:
Tmp1 = SelectExpr(N.getOperand(0));
BuildMI(BB, PPC::NOR, 2, Result).addReg(Tmp1).addReg(Tmp1);
break;
}
return Result;
}
Tmp1 = SelectExpr(N.getOperand(0));
switch(getImmediateForOpcode(N.getOperand(1), opcode, Tmp2)) {
default: assert(0 && "unhandled result code");
case 0: // No immediate
Tmp2 = SelectExpr(N.getOperand(1));
BuildMI(BB, PPC::XOR, 2, Result).addReg(Tmp1).addReg(Tmp2);
break;
case 1: // Low immediate
BuildMI(BB, PPC::XORI, 2, Result).addReg(Tmp1).addImm(Tmp2);
break;
case 2: // Shifted immediate
BuildMI(BB, PPC::XORIS, 2, Result).addReg(Tmp1).addImm(Tmp2);
break;
}
return Result;
}
case ISD::SUB:
if (!MVT::isInteger(DestType)) {
if (!NoExcessFPPrecision && N.getOperand(0).getOpcode() == ISD::MUL &&
N.getOperand(0).Val->hasOneUse()) {
++FusedFP; // Statistic
Tmp1 = SelectExpr(N.getOperand(0).getOperand(0));
Tmp2 = SelectExpr(N.getOperand(0).getOperand(1));
Tmp3 = SelectExpr(N.getOperand(1));
Opc = DestType == MVT::f64 ? PPC::FMSUB : PPC::FMSUBS;
BuildMI(BB, Opc, 3, Result).addReg(Tmp1).addReg(Tmp2).addReg(Tmp3);
return Result;
}
if (!NoExcessFPPrecision && N.getOperand(1).getOpcode() == ISD::MUL &&
N.getOperand(1).Val->hasOneUse()) {
++FusedFP; // Statistic
Tmp1 = SelectExpr(N.getOperand(1).getOperand(0));
Tmp2 = SelectExpr(N.getOperand(1).getOperand(1));
Tmp3 = SelectExpr(N.getOperand(0));
Opc = DestType == MVT::f64 ? PPC::FNMSUB : PPC::FNMSUBS;
BuildMI(BB, Opc, 3, Result).addReg(Tmp1).addReg(Tmp2).addReg(Tmp3);
return Result;
}
Opc = DestType == MVT::f64 ? PPC::FSUB : PPC::FSUBS;
Tmp1 = SelectExpr(N.getOperand(0));
Tmp2 = SelectExpr(N.getOperand(1));
BuildMI(BB, Opc, 2, Result).addReg(Tmp1).addReg(Tmp2);
return Result;
}
if (1 == getImmediateForOpcode(N.getOperand(0), opcode, Tmp1, true)) {
Tmp2 = SelectExpr(N.getOperand(1));
BuildMI(BB, PPC::SUBFIC, 2, Result).addReg(Tmp2).addSImm(Tmp1);
} else if (1 == getImmediateForOpcode(N.getOperand(1), opcode, Tmp2)) {
Tmp1 = SelectExpr(N.getOperand(0));
BuildMI(BB, PPC::ADDI, 2, Result).addReg(Tmp1).addSImm(Tmp2);
} else {
Tmp1 = SelectExpr(N.getOperand(0));
Tmp2 = SelectExpr(N.getOperand(1));
BuildMI(BB, PPC::SUBF, 2, Result).addReg(Tmp2).addReg(Tmp1);
}
return Result;
case ISD::MUL:
Tmp1 = SelectExpr(N.getOperand(0));
if (1 == getImmediateForOpcode(N.getOperand(1), opcode, Tmp2))
BuildMI(BB, PPC::MULLI, 2, Result).addReg(Tmp1).addSImm(Tmp2);
else {
Tmp2 = SelectExpr(N.getOperand(1));
switch (DestType) {
default: assert(0 && "Unknown type to ISD::MUL"); break;
case MVT::i32: Opc = PPC::MULLW; break;
case MVT::f32: Opc = PPC::FMULS; break;
case MVT::f64: Opc = PPC::FMUL; break;
}
BuildMI(BB, Opc, 2, Result).addReg(Tmp1).addReg(Tmp2);
}
return Result;
case ISD::MULHS:
case ISD::MULHU:
Tmp1 = SelectExpr(N.getOperand(0));
Tmp2 = SelectExpr(N.getOperand(1));
Opc = (ISD::MULHU == opcode) ? PPC::MULHWU : PPC::MULHW;
BuildMI(BB, Opc, 2, Result).addReg(Tmp1).addReg(Tmp2);
return Result;
case ISD::SDIV:
case ISD::UDIV:
switch (getImmediateForOpcode(N.getOperand(1), opcode, Tmp3)) {
default: break;
// If this is an sdiv by a power of two, we can use an srawi/addze pair.
case 3:
Tmp1 = MakeReg(MVT::i32);
Tmp2 = SelectExpr(N.getOperand(0));
if ((int)Tmp3 < 0) {
unsigned Tmp4 = MakeReg(MVT::i32);
BuildMI(BB, PPC::SRAWI, 2, Tmp1).addReg(Tmp2).addImm(-Tmp3);
BuildMI(BB, PPC::ADDZE, 1, Tmp4).addReg(Tmp1);
BuildMI(BB, PPC::NEG, 1, Result).addReg(Tmp4);
} else {
BuildMI(BB, PPC::SRAWI, 2, Tmp1).addReg(Tmp2).addImm(Tmp3);
BuildMI(BB, PPC::ADDZE, 1, Result).addReg(Tmp1);
}
return Result;
// If this is a divide by constant, we can emit code using some magic
// constants to implement it as a multiply instead.
case 4:
ExprMap.erase(N);
if (opcode == ISD::SDIV)
return SelectExpr(BuildSDIVSequence(N));
else
return SelectExpr(BuildUDIVSequence(N));
}
Tmp1 = SelectExpr(N.getOperand(0));
Tmp2 = SelectExpr(N.getOperand(1));
switch (DestType) {
default: assert(0 && "Unknown type to ISD::SDIV"); break;
case MVT::i32: Opc = (ISD::UDIV == opcode) ? PPC::DIVWU : PPC::DIVW; break;
case MVT::f32: Opc = PPC::FDIVS; break;
case MVT::f64: Opc = PPC::FDIV; break;
}
BuildMI(BB, Opc, 2, Result).addReg(Tmp1).addReg(Tmp2);
return Result;
case ISD::ADD_PARTS:
case ISD::SUB_PARTS: {
assert(N.getNumOperands() == 4 && N.getValueType() == MVT::i32 &&
"Not an i64 add/sub!");
// Emit all of the operands.
std::vector<unsigned> InVals;
for (unsigned i = 0, e = N.getNumOperands(); i != e; ++i)
InVals.push_back(SelectExpr(N.getOperand(i)));
if (N.getOpcode() == ISD::ADD_PARTS) {
BuildMI(BB, PPC::ADDC, 2, Result).addReg(InVals[0]).addReg(InVals[2]);
BuildMI(BB, PPC::ADDE, 2, Result+1).addReg(InVals[1]).addReg(InVals[3]);
} else {
BuildMI(BB, PPC::SUBFC, 2, Result).addReg(InVals[2]).addReg(InVals[0]);
BuildMI(BB, PPC::SUBFE, 2, Result+1).addReg(InVals[3]).addReg(InVals[1]);
}
return Result+N.ResNo;
}
case ISD::SHL_PARTS:
case ISD::SRA_PARTS:
case ISD::SRL_PARTS: {
assert(N.getNumOperands() == 3 && N.getValueType() == MVT::i32 &&
"Not an i64 shift!");
unsigned ShiftOpLo = SelectExpr(N.getOperand(0));
unsigned ShiftOpHi = SelectExpr(N.getOperand(1));
unsigned SHReg = FoldIfWideZeroExtend(N.getOperand(2));
Tmp1 = MakeReg(MVT::i32);
Tmp2 = MakeReg(MVT::i32);
Tmp3 = MakeReg(MVT::i32);
unsigned Tmp4 = MakeReg(MVT::i32);
unsigned Tmp5 = MakeReg(MVT::i32);
unsigned Tmp6 = MakeReg(MVT::i32);
BuildMI(BB, PPC::SUBFIC, 2, Tmp1).addReg(SHReg).addSImm(32);
if (ISD::SHL_PARTS == opcode) {
BuildMI(BB, PPC::SLW, 2, Tmp2).addReg(ShiftOpHi).addReg(SHReg);
BuildMI(BB, PPC::SRW, 2, Tmp3).addReg(ShiftOpLo).addReg(Tmp1);
BuildMI(BB, PPC::OR, 2, Tmp4).addReg(Tmp2).addReg(Tmp3);
BuildMI(BB, PPC::ADDI, 2, Tmp5).addReg(SHReg).addSImm(-32);
BuildMI(BB, PPC::SLW, 2, Tmp6).addReg(ShiftOpLo).addReg(Tmp5);
BuildMI(BB, PPC::OR, 2, Result+1).addReg(Tmp4).addReg(Tmp6);
BuildMI(BB, PPC::SLW, 2, Result).addReg(ShiftOpLo).addReg(SHReg);
} else if (ISD::SRL_PARTS == opcode) {
BuildMI(BB, PPC::SRW, 2, Tmp2).addReg(ShiftOpLo).addReg(SHReg);
BuildMI(BB, PPC::SLW, 2, Tmp3).addReg(ShiftOpHi).addReg(Tmp1);
BuildMI(BB, PPC::OR, 2, Tmp4).addReg(Tmp2).addReg(Tmp3);
BuildMI(BB, PPC::ADDI, 2, Tmp5).addReg(SHReg).addSImm(-32);
BuildMI(BB, PPC::SRW, 2, Tmp6).addReg(ShiftOpHi).addReg(Tmp5);
BuildMI(BB, PPC::OR, 2, Result).addReg(Tmp4).addReg(Tmp6);
BuildMI(BB, PPC::SRW, 2, Result+1).addReg(ShiftOpHi).addReg(SHReg);
} else {
MachineBasicBlock *TmpMBB = new MachineBasicBlock(BB->getBasicBlock());
MachineBasicBlock *PhiMBB = new MachineBasicBlock(BB->getBasicBlock());
MachineBasicBlock *OldMBB = BB;
MachineFunction *F = BB->getParent();
ilist<MachineBasicBlock>::iterator It = BB; ++It;
F->getBasicBlockList().insert(It, TmpMBB);
F->getBasicBlockList().insert(It, PhiMBB);
BB->addSuccessor(TmpMBB);
BB->addSuccessor(PhiMBB);
BuildMI(BB, PPC::SRW, 2, Tmp2).addReg(ShiftOpLo).addReg(SHReg);
BuildMI(BB, PPC::SLW, 2, Tmp3).addReg(ShiftOpHi).addReg(Tmp1);
BuildMI(BB, PPC::OR, 2, Tmp4).addReg(Tmp2).addReg(Tmp3);
BuildMI(BB, PPC::ADDICo, 2, Tmp5).addReg(SHReg).addSImm(-32);
BuildMI(BB, PPC::SRAW, 2, Tmp6).addReg(ShiftOpHi).addReg(Tmp5);
BuildMI(BB, PPC::SRAW, 2, Result+1).addReg(ShiftOpHi).addReg(SHReg);
BuildMI(BB, PPC::BLE, 2).addReg(PPC::CR0).addMBB(PhiMBB);
// Select correct least significant half if the shift amount > 32
BB = TmpMBB;
unsigned Tmp7 = MakeReg(MVT::i32);
BuildMI(BB, PPC::OR, 2, Tmp7).addReg(Tmp6).addReg(Tmp6);
TmpMBB->addSuccessor(PhiMBB);
BB = PhiMBB;
BuildMI(BB, PPC::PHI, 4, Result).addReg(Tmp4).addMBB(OldMBB)
.addReg(Tmp7).addMBB(TmpMBB);
}
return Result+N.ResNo;
}
case ISD::FP_TO_UINT:
case ISD::FP_TO_SINT: {
bool U = (ISD::FP_TO_UINT == opcode);
Tmp1 = SelectExpr(N.getOperand(0));
if (!U) {
Tmp2 = MakeReg(MVT::f64);
BuildMI(BB, PPC::FCTIWZ, 1, Tmp2).addReg(Tmp1);
int FrameIdx = BB->getParent()->getFrameInfo()->CreateStackObject(8, 8);
addFrameReference(BuildMI(BB, PPC::STFD, 3).addReg(Tmp2), FrameIdx);
addFrameReference(BuildMI(BB, PPC::LWZ, 2, Result), FrameIdx, 4);
return Result;
} else {
unsigned Zero = getConstDouble(0.0);
unsigned MaxInt = getConstDouble((1LL << 32) - 1);
unsigned Border = getConstDouble(1LL << 31);
unsigned UseZero = MakeReg(MVT::f64);
unsigned UseMaxInt = MakeReg(MVT::f64);
unsigned UseChoice = MakeReg(MVT::f64);
unsigned TmpReg = MakeReg(MVT::f64);
unsigned TmpReg2 = MakeReg(MVT::f64);
unsigned ConvReg = MakeReg(MVT::f64);
unsigned IntTmp = MakeReg(MVT::i32);
unsigned XorReg = MakeReg(MVT::i32);
MachineFunction *F = BB->getParent();
int FrameIdx = F->getFrameInfo()->CreateStackObject(8, 8);
// Update machine-CFG edges
MachineBasicBlock *XorMBB = new MachineBasicBlock(BB->getBasicBlock());
MachineBasicBlock *PhiMBB = new MachineBasicBlock(BB->getBasicBlock());
MachineBasicBlock *OldMBB = BB;
ilist<MachineBasicBlock>::iterator It = BB; ++It;
F->getBasicBlockList().insert(It, XorMBB);
F->getBasicBlockList().insert(It, PhiMBB);
BB->addSuccessor(XorMBB);
BB->addSuccessor(PhiMBB);
// Convert from floating point to unsigned 32-bit value
// Use 0 if incoming value is < 0.0
BuildMI(BB, PPC::FSEL, 3, UseZero).addReg(Tmp1).addReg(Tmp1).addReg(Zero);
// Use 2**32 - 1 if incoming value is >= 2**32
BuildMI(BB, PPC::FSUB, 2, UseMaxInt).addReg(MaxInt).addReg(Tmp1);
BuildMI(BB, PPC::FSEL, 3, UseChoice).addReg(UseMaxInt).addReg(UseZero)
.addReg(MaxInt);
// Subtract 2**31
BuildMI(BB, PPC::FSUB, 2, TmpReg).addReg(UseChoice).addReg(Border);
// Use difference if >= 2**31
BuildMI(BB, PPC::FCMPU, 2, PPC::CR0).addReg(UseChoice).addReg(Border);
BuildMI(BB, PPC::FSEL, 3, TmpReg2).addReg(TmpReg).addReg(TmpReg)
.addReg(UseChoice);
// Convert to integer
BuildMI(BB, PPC::FCTIWZ, 1, ConvReg).addReg(TmpReg2);
addFrameReference(BuildMI(BB, PPC::STFD, 3).addReg(ConvReg), FrameIdx);
addFrameReference(BuildMI(BB, PPC::LWZ, 2, IntTmp), FrameIdx, 4);
BuildMI(BB, PPC::BLT, 2).addReg(PPC::CR0).addMBB(PhiMBB);
BuildMI(BB, PPC::B, 1).addMBB(XorMBB);
// XorMBB:
// add 2**31 if input was >= 2**31
BB = XorMBB;
BuildMI(BB, PPC::XORIS, 2, XorReg).addReg(IntTmp).addImm(0x8000);
XorMBB->addSuccessor(PhiMBB);
// PhiMBB:
// DestReg = phi [ IntTmp, OldMBB ], [ XorReg, XorMBB ]
BB = PhiMBB;
BuildMI(BB, PPC::PHI, 4, Result).addReg(IntTmp).addMBB(OldMBB)
.addReg(XorReg).addMBB(XorMBB);
return Result;
}
assert(0 && "Should never get here");
return 0;
}
case ISD::SETCC:
if (SetCCSDNode *SetCC = dyn_cast<SetCCSDNode>(Node)) {
if (ConstantSDNode *CN =
dyn_cast<ConstantSDNode>(SetCC->getOperand(1).Val)) {
// We can codegen setcc op, imm very efficiently compared to a brcond.
// Check for those cases here.
// setcc op, 0
if (CN->getValue() == 0) {
Tmp1 = SelectExpr(SetCC->getOperand(0));
switch (SetCC->getCondition()) {
default: SetCC->dump(); assert(0 && "Unhandled SetCC condition"); abort();
case ISD::SETEQ:
Tmp2 = MakeReg(MVT::i32);
BuildMI(BB, PPC::CNTLZW, 1, Tmp2).addReg(Tmp1);
BuildMI(BB, PPC::RLWINM, 4, Result).addReg(Tmp2).addImm(27)
.addImm(5).addImm(31);
break;
case ISD::SETNE:
Tmp2 = MakeReg(MVT::i32);
BuildMI(BB, PPC::ADDIC, 2, Tmp2).addReg(Tmp1).addSImm(-1);
BuildMI(BB, PPC::SUBFE, 2, Result).addReg(Tmp2).addReg(Tmp1);
break;
case ISD::SETLT:
BuildMI(BB, PPC::RLWINM, 4, Result).addReg(Tmp1).addImm(1)
.addImm(31).addImm(31);
break;
case ISD::SETGT:
Tmp2 = MakeReg(MVT::i32);
Tmp3 = MakeReg(MVT::i32);
BuildMI(BB, PPC::NEG, 2, Tmp2).addReg(Tmp1);
BuildMI(BB, PPC::ANDC, 2, Tmp3).addReg(Tmp2).addReg(Tmp1);
BuildMI(BB, PPC::RLWINM, 4, Result).addReg(Tmp3).addImm(1)
.addImm(31).addImm(31);
break;
}
return Result;
}
// setcc op, -1
if (CN->isAllOnesValue()) {
Tmp1 = SelectExpr(SetCC->getOperand(0));
switch (SetCC->getCondition()) {
default: assert(0 && "Unhandled SetCC condition"); abort();
case ISD::SETEQ:
Tmp2 = MakeReg(MVT::i32);
Tmp3 = MakeReg(MVT::i32);
BuildMI(BB, PPC::ADDIC, 2, Tmp2).addReg(Tmp1).addSImm(1);
BuildMI(BB, PPC::LI, 1, Tmp3).addSImm(0);
BuildMI(BB, PPC::ADDZE, 1, Result).addReg(Tmp3);
break;
case ISD::SETNE:
Tmp2 = MakeReg(MVT::i32);
Tmp3 = MakeReg(MVT::i32);
BuildMI(BB, PPC::NOR, 2, Tmp2).addReg(Tmp1).addReg(Tmp1);
BuildMI(BB, PPC::ADDIC, 2, Tmp3).addReg(Tmp2).addSImm(-1);
BuildMI(BB, PPC::SUBFE, 2, Result).addReg(Tmp3).addReg(Tmp2);
break;
case ISD::SETLT:
Tmp2 = MakeReg(MVT::i32);
Tmp3 = MakeReg(MVT::i32);
BuildMI(BB, PPC::ADDI, 2, Tmp2).addReg(Tmp1).addSImm(1);
BuildMI(BB, PPC::AND, 2, Tmp3).addReg(Tmp2).addReg(Tmp1);
BuildMI(BB, PPC::RLWINM, 4, Result).addReg(Tmp3).addImm(1)
.addImm(31).addImm(31);
break;
case ISD::SETGT:
Tmp2 = MakeReg(MVT::i32);
BuildMI(BB, PPC::RLWINM, 4, Tmp2).addReg(Tmp1).addImm(1)
.addImm(31).addImm(31);
BuildMI(BB, PPC::XORI, 2, Result).addReg(Tmp2).addImm(1);
break;
}
return Result;
}
}
bool Inv;
unsigned CCReg = SelectCC(N, Opc, Inv, Tmp2);
MoveCRtoGPR(CCReg, Inv, Tmp2, Result);
return Result;
}
assert(0 && "Is this legal?");
return 0;
case ISD::SELECT: {
SetCCSDNode* SetCC = dyn_cast<SetCCSDNode>(N.getOperand(0).Val);
if (SetCC && N.getOperand(0).getOpcode() == ISD::SETCC &&
!MVT::isInteger(SetCC->getOperand(0).getValueType()) &&
!MVT::isInteger(N.getOperand(1).getValueType()) &&
!MVT::isInteger(N.getOperand(2).getValueType()) &&
SetCC->getCondition() != ISD::SETEQ &&
SetCC->getCondition() != ISD::SETNE) {
MVT::ValueType VT = SetCC->getOperand(0).getValueType();
unsigned TV = SelectExpr(N.getOperand(1)); // Use if TRUE
unsigned FV = SelectExpr(N.getOperand(2)); // Use if FALSE
ConstantFPSDNode *CN = dyn_cast<ConstantFPSDNode>(SetCC->getOperand(1));
if (CN && (CN->isExactlyValue(-0.0) || CN->isExactlyValue(0.0))) {
switch(SetCC->getCondition()) {
default: assert(0 && "Invalid FSEL condition"); abort();
case ISD::SETULT:
case ISD::SETLT:
std::swap(TV, FV); // fsel is natively setge, swap operands for setlt
case ISD::SETUGE:
case ISD::SETGE:
Tmp1 = SelectExpr(SetCC->getOperand(0)); // Val to compare against
BuildMI(BB, PPC::FSEL, 3, Result).addReg(Tmp1).addReg(TV).addReg(FV);
return Result;
case ISD::SETUGT:
case ISD::SETGT:
std::swap(TV, FV); // fsel is natively setge, swap operands for setlt
case ISD::SETULE:
case ISD::SETLE: {
if (SetCC->getOperand(0).getOpcode() == ISD::FNEG) {
Tmp2 = SelectExpr(SetCC->getOperand(0).getOperand(0));
} else {
Tmp2 = MakeReg(VT);
Tmp1 = SelectExpr(SetCC->getOperand(0)); // Val to compare against
BuildMI(BB, PPC::FNEG, 1, Tmp2).addReg(Tmp1);
}
BuildMI(BB, PPC::FSEL, 3, Result).addReg(Tmp2).addReg(TV).addReg(FV);
return Result;
}
}
} else {
Opc = (MVT::f64 == VT) ? PPC::FSUB : PPC::FSUBS;
Tmp1 = SelectExpr(SetCC->getOperand(0)); // Val to compare against
Tmp2 = SelectExpr(SetCC->getOperand(1));
Tmp3 = MakeReg(VT);
switch(SetCC->getCondition()) {
default: assert(0 && "Invalid FSEL condition"); abort();
case ISD::SETULT:
case ISD::SETLT:
BuildMI(BB, Opc, 2, Tmp3).addReg(Tmp1).addReg(Tmp2);
BuildMI(BB, PPC::FSEL, 3, Result).addReg(Tmp3).addReg(FV).addReg(TV);
return Result;
case ISD::SETUGE:
case ISD::SETGE:
BuildMI(BB, Opc, 2, Tmp3).addReg(Tmp1).addReg(Tmp2);
BuildMI(BB, PPC::FSEL, 3, Result).addReg(Tmp3).addReg(TV).addReg(FV);
return Result;
case ISD::SETUGT:
case ISD::SETGT:
BuildMI(BB, Opc, 2, Tmp3).addReg(Tmp2).addReg(Tmp1);
BuildMI(BB, PPC::FSEL, 3, Result).addReg(Tmp3).addReg(FV).addReg(TV);
return Result;
case ISD::SETULE:
case ISD::SETLE:
BuildMI(BB, Opc, 2, Tmp3).addReg(Tmp2).addReg(Tmp1);
BuildMI(BB, PPC::FSEL, 3, Result).addReg(Tmp3).addReg(TV).addReg(FV);
return Result;
}
}
assert(0 && "Should never get here");
return 0;
}
bool Inv;
unsigned TrueValue = SelectExpr(N.getOperand(1)); //Use if TRUE
unsigned FalseValue = SelectExpr(N.getOperand(2)); //Use if FALSE
unsigned CCReg = SelectCC(N.getOperand(0), Opc, Inv, Tmp3);
// Create an iterator with which to insert the MBB for copying the false
// value and the MBB to hold the PHI instruction for this SetCC.
MachineBasicBlock *thisMBB = BB;
const BasicBlock *LLVM_BB = BB->getBasicBlock();
ilist<MachineBasicBlock>::iterator It = BB;
++It;
// thisMBB:
// ...
// TrueVal = ...
// cmpTY ccX, r1, r2
// bCC copy1MBB
// fallthrough --> copy0MBB
MachineBasicBlock *copy0MBB = new MachineBasicBlock(LLVM_BB);
MachineBasicBlock *sinkMBB = new MachineBasicBlock(LLVM_BB);
BuildMI(BB, Opc, 2).addReg(CCReg).addMBB(sinkMBB);
MachineFunction *F = BB->getParent();
F->getBasicBlockList().insert(It, copy0MBB);
F->getBasicBlockList().insert(It, sinkMBB);
// Update machine-CFG edges
BB->addSuccessor(copy0MBB);
BB->addSuccessor(sinkMBB);
// copy0MBB:
// %FalseValue = ...
// # fallthrough to sinkMBB
BB = copy0MBB;
// Update machine-CFG edges
BB->addSuccessor(sinkMBB);
// sinkMBB:
// %Result = phi [ %FalseValue, copy0MBB ], [ %TrueValue, thisMBB ]
// ...
BB = sinkMBB;
BuildMI(BB, PPC::PHI, 4, Result).addReg(FalseValue)
.addMBB(copy0MBB).addReg(TrueValue).addMBB(thisMBB);
return Result;
}
case ISD::Constant:
switch (N.getValueType()) {
default: assert(0 && "Cannot use constants of this type!");
case MVT::i1:
BuildMI(BB, PPC::LI, 1, Result)
.addSImm(!cast<ConstantSDNode>(N)->isNullValue());
break;
case MVT::i32:
{
int v = (int)cast<ConstantSDNode>(N)->getSignExtended();
if (v < 32768 && v >= -32768) {
BuildMI(BB, PPC::LI, 1, Result).addSImm(v);
} else {
Tmp1 = MakeReg(MVT::i32);
BuildMI(BB, PPC::LIS, 1, Tmp1).addSImm(v >> 16);
BuildMI(BB, PPC::ORI, 2, Result).addReg(Tmp1).addImm(v & 0xFFFF);
}
}
}
return Result;
case ISD::ConstantFP: {
ConstantFPSDNode *CN = cast<ConstantFPSDNode>(N);
Result = getConstDouble(CN->getValue(), Result);
return Result;
}
case ISD::FNEG:
if (!NoExcessFPPrecision &&
ISD::ADD == N.getOperand(0).getOpcode() &&
N.getOperand(0).Val->hasOneUse() &&
ISD::MUL == N.getOperand(0).getOperand(0).getOpcode() &&
N.getOperand(0).getOperand(0).Val->hasOneUse()) {
++FusedFP; // Statistic
Tmp1 = SelectExpr(N.getOperand(0).getOperand(0).getOperand(0));
Tmp2 = SelectExpr(N.getOperand(0).getOperand(0).getOperand(1));
Tmp3 = SelectExpr(N.getOperand(0).getOperand(1));
Opc = DestType == MVT::f64 ? PPC::FNMADD : PPC::FNMADDS;
BuildMI(BB, Opc, 3, Result).addReg(Tmp1).addReg(Tmp2).addReg(Tmp3);
} else if (!NoExcessFPPrecision &&
ISD::ADD == N.getOperand(0).getOpcode() &&
N.getOperand(0).Val->hasOneUse() &&
ISD::MUL == N.getOperand(0).getOperand(1).getOpcode() &&
N.getOperand(0).getOperand(1).Val->hasOneUse()) {
++FusedFP; // Statistic
Tmp1 = SelectExpr(N.getOperand(0).getOperand(1).getOperand(0));
Tmp2 = SelectExpr(N.getOperand(0).getOperand(1).getOperand(1));
Tmp3 = SelectExpr(N.getOperand(0).getOperand(0));
Opc = DestType == MVT::f64 ? PPC::FNMADD : PPC::FNMADDS;
BuildMI(BB, Opc, 3, Result).addReg(Tmp1).addReg(Tmp2).addReg(Tmp3);
} else if (ISD::FABS == N.getOperand(0).getOpcode()) {
Tmp1 = SelectExpr(N.getOperand(0).getOperand(0));
BuildMI(BB, PPC::FNABS, 1, Result).addReg(Tmp1);
} else {
Tmp1 = SelectExpr(N.getOperand(0));
BuildMI(BB, PPC::FNEG, 1, Result).addReg(Tmp1);
}
return Result;
case ISD::FABS:
Tmp1 = SelectExpr(N.getOperand(0));
BuildMI(BB, PPC::FABS, 1, Result).addReg(Tmp1);
return Result;
case ISD::FSQRT:
Tmp1 = SelectExpr(N.getOperand(0));
Opc = DestType == MVT::f64 ? PPC::FSQRT : PPC::FSQRTS;
BuildMI(BB, Opc, 1, Result).addReg(Tmp1);
return Result;
case ISD::FP_ROUND:
assert (DestType == MVT::f32 &&
N.getOperand(0).getValueType() == MVT::f64 &&
"only f64 to f32 conversion supported here");
Tmp1 = SelectExpr(N.getOperand(0));
BuildMI(BB, PPC::FRSP, 1, Result).addReg(Tmp1);
return Result;
case ISD::FP_EXTEND:
assert (DestType == MVT::f64 &&
N.getOperand(0).getValueType() == MVT::f32 &&
"only f32 to f64 conversion supported here");
Tmp1 = SelectExpr(N.getOperand(0));
BuildMI(BB, PPC::FMR, 1, Result).addReg(Tmp1);
return Result;
case ISD::UINT_TO_FP:
case ISD::SINT_TO_FP: {
assert (N.getOperand(0).getValueType() == MVT::i32
&& "int to float must operate on i32");
bool IsUnsigned = (ISD::UINT_TO_FP == opcode);
Tmp1 = SelectExpr(N.getOperand(0)); // Get the operand register
Tmp2 = MakeReg(MVT::f64); // temp reg to load the integer value into
Tmp3 = MakeReg(MVT::i32); // temp reg to hold the conversion constant
int FrameIdx = BB->getParent()->getFrameInfo()->CreateStackObject(8, 8);
MachineConstantPool *CP = BB->getParent()->getConstantPool();
if (IsUnsigned) {
unsigned ConstF = getConstDouble(0x1.000000p52);
// Store the hi & low halves of the fp value, currently in int regs
BuildMI(BB, PPC::LIS, 1, Tmp3).addSImm(0x4330);
addFrameReference(BuildMI(BB, PPC::STW, 3).addReg(Tmp3), FrameIdx);
addFrameReference(BuildMI(BB, PPC::STW, 3).addReg(Tmp1), FrameIdx, 4);
addFrameReference(BuildMI(BB, PPC::LFD, 2, Tmp2), FrameIdx);
// Generate the return value with a subtract
BuildMI(BB, PPC::FSUB, 2, Result).addReg(Tmp2).addReg(ConstF);
} else {
unsigned ConstF = getConstDouble(0x1.000008p52);
unsigned TmpL = MakeReg(MVT::i32);
// Store the hi & low halves of the fp value, currently in int regs
BuildMI(BB, PPC::LIS, 1, Tmp3).addSImm(0x4330);
addFrameReference(BuildMI(BB, PPC::STW, 3).addReg(Tmp3), FrameIdx);
BuildMI(BB, PPC::XORIS, 2, TmpL).addReg(Tmp1).addImm(0x8000);
addFrameReference(BuildMI(BB, PPC::STW, 3).addReg(TmpL), FrameIdx, 4);
addFrameReference(BuildMI(BB, PPC::LFD, 2, Tmp2), FrameIdx);
// Generate the return value with a subtract
BuildMI(BB, PPC::FSUB, 2, Result).addReg(Tmp2).addReg(ConstF);
}
return Result;
}
}
return 0;
}
void ISel::Select(SDOperand N) {
unsigned Tmp1, Tmp2, Tmp3, Opc;
unsigned opcode = N.getOpcode();
if (!ExprMap.insert(std::make_pair(N, 1)).second)
return; // Already selected.
SDNode *Node = N.Val;
switch (Node->getOpcode()) {
default:
Node->dump(); std::cerr << "\n";
assert(0 && "Node not handled yet!");
case ISD::EntryToken: return; // Noop
case ISD::TokenFactor:
for (unsigned i = 0, e = Node->getNumOperands(); i != e; ++i)
Select(Node->getOperand(i));
return;
case ISD::CALLSEQ_START:
case ISD::CALLSEQ_END:
Select(N.getOperand(0));
Tmp1 = cast<ConstantSDNode>(N.getOperand(1))->getValue();
Opc = N.getOpcode() == ISD::CALLSEQ_START ? PPC::ADJCALLSTACKDOWN :
PPC::ADJCALLSTACKUP;
BuildMI(BB, Opc, 1).addImm(Tmp1);
return;
case ISD::BR: {
MachineBasicBlock *Dest =
cast<BasicBlockSDNode>(N.getOperand(1))->getBasicBlock();
Select(N.getOperand(0));
BuildMI(BB, PPC::B, 1).addMBB(Dest);
return;
}
case ISD::BRCOND:
case ISD::BRCONDTWOWAY:
SelectBranchCC(N);
return;
case ISD::CopyToReg:
Select(N.getOperand(0));
Tmp1 = SelectExpr(N.getOperand(1));
Tmp2 = cast<RegSDNode>(N)->getReg();
if (Tmp1 != Tmp2) {
if (N.getOperand(1).getValueType() == MVT::f64 ||
N.getOperand(1).getValueType() == MVT::f32)
BuildMI(BB, PPC::FMR, 1, Tmp2).addReg(Tmp1);
else
BuildMI(BB, PPC::OR, 2, Tmp2).addReg(Tmp1).addReg(Tmp1);
}
return;
case ISD::ImplicitDef:
Select(N.getOperand(0));
BuildMI(BB, PPC::IMPLICIT_DEF, 0, cast<RegSDNode>(N)->getReg());
return;
case ISD::RET:
switch (N.getNumOperands()) {
default:
assert(0 && "Unknown return instruction!");
case 3:
assert(N.getOperand(1).getValueType() == MVT::i32 &&
N.getOperand(2).getValueType() == MVT::i32 &&
"Unknown two-register value!");
Select(N.getOperand(0));
Tmp1 = SelectExpr(N.getOperand(1));
Tmp2 = SelectExpr(N.getOperand(2));
BuildMI(BB, PPC::OR, 2, PPC::R3).addReg(Tmp2).addReg(Tmp2);
BuildMI(BB, PPC::OR, 2, PPC::R4).addReg(Tmp1).addReg(Tmp1);
break;
case 2:
Select(N.getOperand(0));
Tmp1 = SelectExpr(N.getOperand(1));
switch (N.getOperand(1).getValueType()) {
default:
assert(0 && "Unknown return type!");
case MVT::f64:
case MVT::f32:
BuildMI(BB, PPC::FMR, 1, PPC::F1).addReg(Tmp1);
break;
case MVT::i32:
BuildMI(BB, PPC::OR, 2, PPC::R3).addReg(Tmp1).addReg(Tmp1);
break;
}
case 1:
Select(N.getOperand(0));
break;
}
BuildMI(BB, PPC::BLR, 0); // Just emit a 'ret' instruction
return;
case ISD::TRUNCSTORE:
case ISD::STORE: {
SDOperand Chain = N.getOperand(0);
SDOperand Value = N.getOperand(1);
SDOperand Address = N.getOperand(2);
Select(Chain);
Tmp1 = SelectExpr(Value); //value
if (opcode == ISD::STORE) {
switch(Value.getValueType()) {
default: assert(0 && "unknown Type in store");
case MVT::i32: Opc = PPC::STW; break;
case MVT::f64: Opc = PPC::STFD; break;
case MVT::f32: Opc = PPC::STFS; break;
}
} else { //ISD::TRUNCSTORE
switch(cast<VTSDNode>(Node->getOperand(4))->getVT()) {
default: assert(0 && "unknown Type in store");
case MVT::i1:
case MVT::i8: Opc = PPC::STB; break;
case MVT::i16: Opc = PPC::STH; break;
}
}
if(Address.getOpcode() == ISD::FrameIndex) {
Tmp2 = cast<FrameIndexSDNode>(Address)->getIndex();
addFrameReference(BuildMI(BB, Opc, 3).addReg(Tmp1), (int)Tmp2);
} else if(GlobalAddressSDNode *GN = dyn_cast<GlobalAddressSDNode>(Address)){
GlobalValue *GV = GN->getGlobal();
Tmp2 = MakeReg(MVT::i32);
if (PICEnabled)
BuildMI(BB, PPC::ADDIS, 2, Tmp2).addReg(getGlobalBaseReg())
.addGlobalAddress(GV);
else
BuildMI(BB, PPC::LIS, 1, Tmp2).addGlobalAddress(GV);
if (GV->hasWeakLinkage() || GV->isExternal()) {
Tmp3 = MakeReg(MVT::i32);
BuildMI(BB, PPC::LWZ, 2, Tmp3).addGlobalAddress(GV).addReg(Tmp2);
BuildMI(BB, Opc, 3).addReg(Tmp1).addSImm(0).addReg(Tmp3);
} else {
BuildMI(BB, Opc, 3).addReg(Tmp1).addGlobalAddress(GV).addReg(Tmp2);
}
} else {
int offset;
switch(SelectAddr(Address, Tmp2, offset)) {
default: assert(0 && "Unhandled return value from SelectAddr");
case 0: // imm offset, no frame, no index
BuildMI(BB, Opc, 3).addReg(Tmp1).addSImm(offset).addReg(Tmp2);
break;
case 1: // imm offset + frame index
addFrameReference(BuildMI(BB, Opc, 3).addReg(Tmp1), (int)Tmp2, offset);
break;
case 2: // base+index addressing
Opc = IndexedOpForOp(Opc);
BuildMI(BB, Opc, 3).addReg(Tmp1).addReg(Tmp2).addReg(offset);
break;
}
}
return;
}
case ISD::EXTLOAD:
case ISD::SEXTLOAD:
case ISD::ZEXTLOAD:
case ISD::LOAD:
case ISD::CopyFromReg:
case ISD::TAILCALL:
case ISD::CALL:
case ISD::DYNAMIC_STACKALLOC:
ExprMap.erase(N);
SelectExpr(N);
return;
}
assert(0 && "Should not be reached!");
}
/// createPPC32PatternInstructionSelector - This pass converts an LLVM function
/// into a machine code representation using pattern matching and a machine
/// description file.
///
FunctionPass *llvm::createPPC32ISelPattern(TargetMachine &TM) {
return new ISel(TM);
}