Reserve a stack slot if the function adjusts the stack but doesn't

simplify the call frame pseudo instructions. In that situation, the
calculations for estimating the stack size will be way off, leading to
not having an emergency spill slot when we need one. It should be possible
to be more precise about tracking the adjustment values, but not really
necessary for correctness. Upcoming cleanups for PEI in general will
render that moot.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@110258 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Jim Grosbach 2010-08-04 22:10:15 +00:00
parent 7536f72a97
commit 936ed5424c

View File

@ -851,13 +851,18 @@ ARMBaseRegisterInfo::processFunctionBeforeCalleeSavedScan(MachineFunction &MF,
// slot of the previous FP. Also, if we have variable sized objects in the
// function, stack slot references will often be negative, and some of
// our instructions are positive-offset only, so conservatively consider
// that case to want a spill slot (or register) as well.
// that case to want a spill slot (or register) as well. Similarly, if
// the function adjusts the stack pointer during execution and the
// adjustments aren't already part of our stack size estimate, our offset
// calculations may be off, so be conservative.
// FIXME: We could add logic to be more precise about negative offsets
// and which instructions will need a scratch register for them. Is it
// worth the effort and added fragility?
bool BigStack =
(RS && (estimateStackSize(MF) + (hasFP(MF) ? 4:0) >=
estimateRSStackSizeLimit(MF))) || MFI->hasVarSizedObjects();
estimateRSStackSizeLimit(MF))
|| MFI->hasVarSizedObjects()
|| (MFI->adjustsStack() && !canSimplifyCallFramePseudos(MF)));
bool ExtraCSSpill = false;
if (BigStack || !CanEliminateFrame || cannotEliminateFrame(MF)) {