[PATCH] D31062: PR32288: Describe a bool parameter's DWARF location with a simple register

[Adrian Prantl via llvm-commits]

> On Mar 19, 2017, at 1:37 PM, David Blaikie <dblaikie at gmail.com> wrote:
> 
> I /think/ this is legitimately unreachable, or probably should be. (but if you could help with how to test it, if it is testable, I'd appreciate that)
> 
> The function is for converting a dbg.declare to a dbg.value. A dbg.declare should never describe a fragment, should it?

The code in SROA.cpp:4049 will create dbg.declares with fragments in it. But I'm not sure if it runs before or after mem2reg.

-- adrian

> It's for describing the whole that's stored in a static alloca, if I understand correctly?


> 
> On Fri, Mar 17, 2017 at 11:38 AM Adrian Prantl <aprantl at apple.com <mailto:aprantl at apple.com>> wrote:
>> On Mar 17, 2017, at 11:34 AM, David Blaikie <dblaikie at gmail.com <mailto:dblaikie at gmail.com>> wrote:
>> 
>> I suppose the question is: is it worth/necessary to provide a test case? We could include a fixme & not do anything here? Or I can work up an IR test case for IR Clang (& probably Swift) don't currently generate, so behavior that won't be used.
> 
> IMO if the IR passes the Verifier, we should support it. And we didn't test other frontends, such as Julia, for example. If it is just a matter of making a small modification to IR generated by clang, I feel more comfortable with supporting it.
> 
> -- adrian
> 
>> 
>> On Fri, Mar 17, 2017 at 11:15 AM Adrian Prantl <aprantl at apple.com <mailto:aprantl at apple.com>> wrote:
>>> On Mar 17, 2017, at 9:12 AM, David Blaikie <dblaikie at gmail.com <mailto:dblaikie at gmail.com>> wrote:
>>> 
>>> Hmm, one wrinkle... 
>>> 
>>> the "if (auto Fragment ...)" case doesn't run in any existing test cases. (I put an "assert(false)" there & ran check-llvm).
>>> 
>>> Any ideas if that's still reachable? Perhaps this only happens for direct bool parameters, never fragments.
>> 
>> I just finished building the swift stdlib in 64 and 32 bit (which is typically a good source of these) and also didn't hit the assertion. We'd need to hand-craft a testcase for opt -mem2reg.
>> 
>> -- adrian
>> 
>>> 
>>> On Fri, Mar 17, 2017 at 8:28 AM Adrian Prantl via Phabricator <reviews at reviews.llvm.org <mailto:reviews at reviews.llvm.org>> wrote:
>>> aprantl accepted this revision.
>>> aprantl added a comment.
>>> This revision is now accepted and ready to land.
>>> 
>>> Took me a while to find the functional change amidst the refactoring :-)
>>> As discussed in the PR, this LGTM, thanks!
>>> One inline comment about a comment.
>>> 
>>> 
>>> 
>>> ================
>>> Comment at: lib/Transforms/Utils/Local.cpp:1118
>>>      // describing will always be smaller than the variable size, because
>>>      // VariableSize == Size of Alloca described by DDI. Since SI stores
>>>      // to the alloca described by DDI, if it's first operand is an extend,
>>> ----------------
>>> Does this comment still match our understanding?
>>> Particularly the `The fragment we're describing will always be smaller than the variable size` bit.
>>> 
>>> 
>>> https://reviews.llvm.org/D31062 <https://reviews.llvm.org/D31062>
>>> 
>>> 
>>> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170319/05600ad2/attachment.html>