[llvm] r334317 - [InstCombine] Skip dbg.value(s) when looking at stack{save, restore}.

[Vedant Kumar via llvm-commits]

> On Jun 8, 2018, at 4:27 PM, Davide Italiano <davide at freebsd.org> wrote:
> 
> On Fri, Jun 8, 2018 at 4:02 PM, Vedant Kumar <vsk at apple.com> wrote:
>> Hi Davide,
>> 
>> A question inline --
>> 
>>> On Jun 8, 2018, at 1:42 PM, Davide Italiano via llvm-commits <llvm-commits at lists.llvm.org> wrote:
>>> 
>>> Author: davide
>>> Date: Fri Jun  8 13:42:36 2018
>>> New Revision: 334317
>>> 
>>> URL: http://llvm.org/viewvc/llvm-project?rev=334317&view=rev
>>> Log:
>>> [InstCombine] Skip dbg.value(s) when looking at stack{save,restore}.
>>> 
>>> Fixes PR37713.
>>> 
>>> Added:
>>>   llvm/trunk/test/Transforms/InstCombine/stacksave-debuginfo.ll
>>> Modified:
>>>   llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
>>> 
>>> Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp?rev=334317&r1=334316&r2=334317&view=diff
>>> ==============================================================================
>>> --- llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp (original)
>>> +++ llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp Fri Jun  8 13:42:36 2018
>>> @@ -3521,8 +3521,15 @@ Instruction *InstCombiner::visitCallInst
>>>    // happen when variable allocas are DCE'd.
>>>    if (IntrinsicInst *SS = dyn_cast<IntrinsicInst>(II->getArgOperand(0))) {
>>>      if (SS->getIntrinsicID() == Intrinsic::stacksave) {
>>> -        if (&*++SS->getIterator() == II)
>>> +        // Skip over debug info instructions.
>>> +        // FIXME: This should be an utility in Instruction.h
>>> +        auto It = SS->getIterator();
>>> +        It++;
>>> +        while (isa<DbgInfoIntrinsic>(*It))
>> 
>> ^ Can you dereference the sentinel node in this while loop's condition?
>> 
> 
> I don't think so, unless the basic block is malformed. Your iterator
> points to a `stacksave`.
> At that point you are guaranteed to have at least another instruction,
> i.e. the terminator of the block (otherwise the block is malformed),
> which doesn't satisfy the `isa<DbgInfoInstrinsic>`) condition.
> Does this make sense to you?

Got it, thanks!

> 
> Thanks,
> 
> --
> Davide