llvm.org GIT mirror llvm / release_16 lib / Transforms / Utils / InlineFunction.cpp

Tree @release_16 (Download .tar.gz)

InlineFunction.cpp @release_16raw · history · blame

//===- InlineFunction.cpp - Code to perform function inlining -------------===//
//                     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 file implements inlining of a function into a call site, resolving
// parameters and the return value as appropriate.
// FIXME: This pass should transform alloca instructions in the called function
// into alloca/dealloca pairs!  Or perhaps it should refuse to inline them!

#include "llvm/Transforms/Utils/Cloning.h"
#include "llvm/Constants.h"
#include "llvm/DerivedTypes.h"
#include "llvm/Module.h"
#include "llvm/Instructions.h"
#include "llvm/Intrinsics.h"
#include "llvm/Support/CallSite.h"
using namespace llvm;

bool llvm::InlineFunction(CallInst *CI) { return InlineFunction(CallSite(CI)); }
bool llvm::InlineFunction(InvokeInst *II) {return InlineFunction(CallSite(II));}

// InlineFunction - This function inlines the called function into the basic
// block of the caller.  This returns false if it is not possible to inline this
// call.  The program is still in a well defined state if this occurs though.
// Note that this only does one level of inlining.  For example, if the
// instruction 'call B' is inlined, and 'B' calls 'C', then the call to 'C' now
// exists in the instruction stream.  Similiarly this will inline a recursive
// function by one level.
bool llvm::InlineFunction(CallSite CS) {
  Instruction *TheCall = CS.getInstruction();
  assert(TheCall->getParent() && TheCall->getParent()->getParent() &&
         "Instruction not in function!");

  const Function *CalledFunc = CS.getCalledFunction();
  if (CalledFunc == 0 ||          // Can't inline external function or indirect
      CalledFunc->isExternal() || // call, or call to a vararg function!
      CalledFunc->getFunctionType()->isVarArg()) return false;

  // If the call to the callee is a non-tail call, we must clear the 'tail'
  // flags on any calls that we inline.
  bool MustClearTailCallFlags =
    isa<CallInst>(TheCall) && !cast<CallInst>(TheCall)->isTailCall();

  BasicBlock *OrigBB = TheCall->getParent();
  Function *Caller = OrigBB->getParent();

  // Get an iterator to the last basic block in the function, which will have
  // the new function inlined after it.
  Function::iterator LastBlock = &Caller->back();

  // Make sure to capture all of the return instructions from the cloned
  // function.
  std::vector<ReturnInst*> Returns;
  { // Scope to destroy ValueMap after cloning.
    // Calculate the vector of arguments to pass into the function cloner...
    std::map<const Value*, Value*> ValueMap;
    assert(std::distance(CalledFunc->arg_begin(), CalledFunc->arg_end()) ==
           std::distance(CS.arg_begin(), CS.arg_end()) &&
           "No varargs calls can be inlined!");

    CallSite::arg_iterator AI = CS.arg_begin();
    for (Function::const_arg_iterator I = CalledFunc->arg_begin(),
           E = CalledFunc->arg_end(); I != E; ++I, ++AI)
      ValueMap[I] = *AI;

    // Clone the entire body of the callee into the caller.
    CloneFunctionInto(Caller, CalledFunc, ValueMap, Returns, ".i");

  // Remember the first block that is newly cloned over.
  Function::iterator FirstNewBlock = LastBlock; ++FirstNewBlock;

  // If there are any alloca instructions in the block that used to be the entry
  // block for the callee, move them to the entry block of the caller.  First
  // calculate which instruction they should be inserted before.  We insert the
  // instructions at the end of the current alloca list.
  if (isa<AllocaInst>(FirstNewBlock->begin())) {
    BasicBlock::iterator InsertPoint = Caller->begin()->begin();
    for (BasicBlock::iterator I = FirstNewBlock->begin(),
           E = FirstNewBlock->end(); I != E; )
      if (AllocaInst *AI = dyn_cast<AllocaInst>(I++))
        if (isa<Constant>(AI->getArraySize())) {
          // Scan for the block of allocas that we can move over.
          while (isa<AllocaInst>(I) &&

          // Transfer all of the allocas over in a block.  Using splice means
          // that they instructions aren't removed from the symbol table, then
          // reinserted.
                                               AI, I);

  // If we are inlining tail call instruction through an invoke or
  if (MustClearTailCallFlags) {
    for (Function::iterator BB = FirstNewBlock, E = Caller->end();
         BB != E; ++BB)
      for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I)
        if (CallInst *CI = dyn_cast<CallInst>(I))

  // If we are inlining for an invoke instruction, we must make sure to rewrite
  // any inlined 'unwind' instructions into branches to the invoke exception
  // destination, and call instructions into invoke instructions.
  if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall)) {
    BasicBlock *InvokeDest = II->getUnwindDest();
    std::vector<Value*> InvokeDestPHIValues;

    // If there are PHI nodes in the exceptional destination block, we need to
    // keep track of which values came into them from this invoke, then remove
    // the entry for this block.
    for (BasicBlock::iterator I = InvokeDest->begin(); isa<PHINode>(I); ++I) {
      PHINode *PN = cast<PHINode>(I);
      // Save the value to use for this edge...

    for (Function::iterator BB = FirstNewBlock, E = Caller->end();
         BB != E; ++BB) {
      for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ) {
        // We only need to check for function calls: inlined invoke instructions
        // require no special handling...
        if (CallInst *CI = dyn_cast<CallInst>(I)) {
          // Convert this function call into an invoke instruction... if it's
          // not an intrinsic function call (which are known to not unwind).
          if (CI->getCalledFunction() &&
              CI->getCalledFunction()->getIntrinsicID()) {
          } else {
            // First, split the basic block...
            BasicBlock *Split = BB->splitBasicBlock(CI, CI->getName()+".noexc");

            // Next, create the new invoke instruction, inserting it at the end
            // of the old basic block.
            InvokeInst *II =
              new InvokeInst(CI->getCalledValue(), Split, InvokeDest,
                            std::vector<Value*>(CI->op_begin()+1, CI->op_end()),
                             CI->getName(), BB->getTerminator());

            // Make sure that anything using the call now uses the invoke!

            // Delete the unconditional branch inserted by splitBasicBlock
            Split->getInstList().pop_front();  // Delete the original call

            // Update any PHI nodes in the exceptional block to indicate that
            // there is now a new entry in them.
            unsigned i = 0;
            for (BasicBlock::iterator I = InvokeDest->begin();
                 isa<PHINode>(I); ++I, ++i) {
              PHINode *PN = cast<PHINode>(I);
              PN->addIncoming(InvokeDestPHIValues[i], BB);

            // This basic block is now complete, start scanning the next one.
        } else {

      if (UnwindInst *UI = dyn_cast<UnwindInst>(BB->getTerminator())) {
        // An UnwindInst requires special handling when it gets inlined into an
        // invoke site.  Once this happens, we know that the unwind would cause
        // a control transfer to the invoke exception destination, so we can
        // transform it into a direct branch to the exception destination.
        new BranchInst(InvokeDest, UI);

        // Delete the unwind instruction!

        // Update any PHI nodes in the exceptional block to indicate that
        // there is now a new entry in them.
        unsigned i = 0;
        for (BasicBlock::iterator I = InvokeDest->begin();
             isa<PHINode>(I); ++I, ++i) {
          PHINode *PN = cast<PHINode>(I);
          PN->addIncoming(InvokeDestPHIValues[i], BB);

    // Now that everything is happy, we have one final detail.  The PHI nodes in
    // the exception destination block still have entries due to the original
    // invoke instruction.  Eliminate these entries (which might even delete the
    // PHI node) now.

  // If we cloned in _exactly one_ basic block, and if that block ends in a
  // return instruction, we splice the body of the inlined callee directly into
  // the calling basic block.
  if (Returns.size() == 1 && std::distance(FirstNewBlock, Caller->end()) == 1) {
    // Move all of the instructions right before the call.
    OrigBB->getInstList().splice(TheCall, FirstNewBlock->getInstList(),
                                 FirstNewBlock->begin(), FirstNewBlock->end());
    // Remove the cloned basic block.

    // If the call site was an invoke instruction, add a branch to the normal
    // destination.
    if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall))
      new BranchInst(II->getNormalDest(), TheCall);

    // If the return instruction returned a value, replace uses of the call with
    // uses of the returned value.
    if (!TheCall->use_empty())

    // Since we are now done with the Call/Invoke, we can delete it.

    // Since we are now done with the return instruction, delete it also.

    // We are now done with the inlining.
    return true;

  // Otherwise, we have the normal case, of more than one block to inline or
  // multiple return sites.

  // We want to clone the entire callee function into the hole between the
  // "starter" and "ender" blocks.  How we accomplish this depends on whether
  // this is an invoke instruction or a call instruction.
  BasicBlock *AfterCallBB;
  if (InvokeInst *II = dyn_cast<InvokeInst>(TheCall)) {

    // Add an unconditional branch to make this look like the CallInst case...
    BranchInst *NewBr = new BranchInst(II->getNormalDest(), TheCall);

    // Split the basic block.  This guarantees that no PHI nodes will have to be
    // updated due to new incoming edges, and make the invoke case more
    // symmetric to the call case.
    AfterCallBB = OrigBB->splitBasicBlock(NewBr,

  } else {  // It's a call
    // If this is a call instruction, we need to split the basic block that
    // the call lives in.
    AfterCallBB = OrigBB->splitBasicBlock(TheCall,

  // Change the branch that used to go to AfterCallBB to branch to the first
  // basic block of the inlined function.
  TerminatorInst *Br = OrigBB->getTerminator();
  assert(Br && Br->getOpcode() == Instruction::Br &&
         "splitBasicBlock broken!");
  Br->setOperand(0, FirstNewBlock);

  // Now that the function is correct, make it a little bit nicer.  In
  // particular, move the basic blocks inserted from the end of the function
  // into the space made by splitting the source basic block.
  Caller->getBasicBlockList().splice(AfterCallBB, Caller->getBasicBlockList(),
                                     FirstNewBlock, Caller->end());

  // Handle all of the return instructions that we just cloned in, and eliminate
  // any users of the original call/invoke instruction.
  if (Returns.size() > 1) {
    // The PHI node should go at the front of the new basic block to merge all
    // possible incoming values.
    PHINode *PHI = 0;
    if (!TheCall->use_empty()) {
      PHI = new PHINode(CalledFunc->getReturnType(),
                        TheCall->getName(), AfterCallBB->begin());

      // Anything that used the result of the function call should now use the
      // PHI node as their operand.

    // Loop over all of the return instructions, turning them into unconditional
    // branches to the merge point now, and adding entries to the PHI node as
    // appropriate.
    for (unsigned i = 0, e = Returns.size(); i != e; ++i) {
      ReturnInst *RI = Returns[i];

      if (PHI) {
        assert(RI->getReturnValue() && "Ret should have value!");
        assert(RI->getReturnValue()->getType() == PHI->getType() &&
               "Ret value not consistent in function!");
        PHI->addIncoming(RI->getReturnValue(), RI->getParent());

      // Add a branch to the merge point where the PHI node lives if it exists.
      new BranchInst(AfterCallBB, RI);

      // Delete the return instruction now

  } else if (!Returns.empty()) {
    // Otherwise, if there is exactly one return value, just replace anything
    // using the return value of the call with the computed value.
    if (!TheCall->use_empty())

    // Splice the code from the return block into the block that it will return
    // to, which contains the code that was after the call.
    BasicBlock *ReturnBB = Returns[0]->getParent();

    // Update PHI nodes that use the ReturnBB to use the AfterCallBB.

    // Delete the return instruction now and empty ReturnBB now.
  } else if (!TheCall->use_empty()) {
    // No returns, but something is using the return value of the call.  Just
    // nuke the result.

  // Since we are now done with the Call/Invoke, we can delete it.

  // We should always be able to fold the entry block of the function into the
  // single predecessor of the block...
  assert(cast<BranchInst>(Br)->isUnconditional() && "splitBasicBlock broken!");
  BasicBlock *CalleeEntry = cast<BranchInst>(Br)->getSuccessor(0);

  // Splice the code entry block into calling block, right before the
  // unconditional branch.
  OrigBB->getInstList().splice(Br, CalleeEntry->getInstList());
  CalleeEntry->replaceAllUsesWith(OrigBB);  // Update PHI nodes

  // Remove the unconditional branch.

  // Now we can remove the CalleeEntry block, which is now empty.
  return true;