aboutsummaryrefslogtreecommitdiff
path: root/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp
diff options
context:
space:
mode:
authorDan Gohman <gohman@apple.com>2010-04-19 18:41:46 +0000
committerDan Gohman <gohman@apple.com>2010-04-19 18:41:46 +0000
commit46007b3712290c09d895f4bd245ac852f412556c (patch)
treeb61979519a215038b77ca4412d5281e911dfffc9 /lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp
parent3072b8aa971b6707189e9aab344475931ebfae5c (diff)
Move isInTailCallPosition out of SelectionDAGBuilder, as it isn't
SelectionDAG-specific. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101801 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp')
-rw-r--r--lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp85
1 files changed, 85 insertions, 0 deletions
diff --git a/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp b/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp
index ccccc2ad02..10b0ee4036 100644
--- a/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp
+++ b/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp
@@ -448,3 +448,88 @@ ISD::CondCode llvm::getICmpCondCode(ICmpInst::Predicate Pred) {
return ISD::SETNE;
}
}
+
+/// Test if the given instruction is in a position to be optimized
+/// with a tail-call. This roughly means that it's in a block with
+/// a return and there's nothing that needs to be scheduled
+/// between it and the return.
+///
+/// This function only tests target-independent requirements.
+bool llvm::isInTailCallPosition(ImmutableCallSite CS, Attributes CalleeRetAttr,
+ const TargetLowering &TLI) {
+ const Instruction *I = CS.getInstruction();
+ const BasicBlock *ExitBB = I->getParent();
+ const TerminatorInst *Term = ExitBB->getTerminator();
+ const ReturnInst *Ret = dyn_cast<ReturnInst>(Term);
+ const Function *F = ExitBB->getParent();
+
+ // The block must end in a return statement or unreachable.
+ //
+ // FIXME: Decline tailcall if it's not guaranteed and if the block ends in
+ // an unreachable, for now. The way tailcall optimization is currently
+ // implemented means it will add an epilogue followed by a jump. That is
+ // not profitable. Also, if the callee is a special function (e.g.
+ // longjmp on x86), it can end up causing miscompilation that has not
+ // been fully understood.
+ if (!Ret &&
+ (!GuaranteedTailCallOpt || !isa<UnreachableInst>(Term))) return false;
+
+ // If I will have a chain, make sure no other instruction that will have a
+ // chain interposes between I and the return.
+ if (I->mayHaveSideEffects() || I->mayReadFromMemory() ||
+ !I->isSafeToSpeculativelyExecute())
+ for (BasicBlock::const_iterator BBI = prior(prior(ExitBB->end())); ;
+ --BBI) {
+ if (&*BBI == I)
+ break;
+ // Debug info intrinsics do not get in the way of tail call optimization.
+ if (isa<DbgInfoIntrinsic>(BBI))
+ continue;
+ if (BBI->mayHaveSideEffects() || BBI->mayReadFromMemory() ||
+ !BBI->isSafeToSpeculativelyExecute())
+ return false;
+ }
+
+ // If the block ends with a void return or unreachable, it doesn't matter
+ // what the call's return type is.
+ if (!Ret || Ret->getNumOperands() == 0) return true;
+
+ // If the return value is undef, it doesn't matter what the call's
+ // return type is.
+ if (isa<UndefValue>(Ret->getOperand(0))) return true;
+
+ // Conservatively require the attributes of the call to match those of
+ // the return. Ignore noalias because it doesn't affect the call sequence.
+ unsigned CallerRetAttr = F->getAttributes().getRetAttributes();
+ if ((CalleeRetAttr ^ CallerRetAttr) & ~Attribute::NoAlias)
+ return false;
+
+ // It's not safe to eliminate the sign / zero extension of the return value.
+ if ((CallerRetAttr & Attribute::ZExt) || (CallerRetAttr & Attribute::SExt))
+ return false;
+
+ // Otherwise, make sure the unmodified return value of I is the return value.
+ for (const Instruction *U = dyn_cast<Instruction>(Ret->getOperand(0)); ;
+ U = dyn_cast<Instruction>(U->getOperand(0))) {
+ if (!U)
+ return false;
+ if (!U->hasOneUse())
+ return false;
+ if (U == I)
+ break;
+ // Check for a truly no-op truncate.
+ if (isa<TruncInst>(U) &&
+ TLI.isTruncateFree(U->getOperand(0)->getType(), U->getType()))
+ continue;
+ // Check for a truly no-op bitcast.
+ if (isa<BitCastInst>(U) &&
+ (U->getOperand(0)->getType() == U->getType() ||
+ (U->getOperand(0)->getType()->isPointerTy() &&
+ U->getType()->isPointerTy())))
+ continue;
+ // Otherwise it's not a true no-op.
+ return false;
+ }
+
+ return true;
+}