r246991 - When building the alloca for a local variable, set its name

[David Blaikie via cfe-commits]

On Tue, Sep 8, 2015 at 10:52 AM, Chandler Carruth <chandlerc at gmail.com>
wrote:

> None of my performance concerns were relevant to this change FWIW.
>

Good to know - if we wanted to go down this path I figure we could just
provide overloads - StringRef and Twine, the StringRef one could always use
the name, even in -Asserts and the Twine one would not.

But yeah, your comments below seem worth considering before we figure out
about that path.


> I think the reason that this got "fixed" was because people had a tendancy
> to *rely* on the name downstream when we made it always present. =/
> Personally, I like having *no* names in a no-asserts build because it
> ensures that absolutely no one is relying on them -- we have debug
> information for that.
>
> I wonder if the right way to help the debugging scenario (which is very
> real and painful) is to expose a flag for this, or to expose a way to take
> an IR file with debug info and synthesize nice names for as much as we can
> using the debug info?
>
> On Tue, Sep 8, 2015 at 10:47 AM David Blaikie via cfe-commits <
> cfe-commits at lists.llvm.org> wrote:
>
>> On Tue, Sep 8, 2015 at 10:26 AM, John McCall <rjmccall at apple.com> wrote:
>>
>>> On Sep 8, 2015, at 8:25 AM, David Blaikie <dblaikie at gmail.com> wrote:
>>> On Tue, Sep 8, 2015 at 2:18 AM, John McCall via cfe-commits <
>>> cfe-commits at lists.llvm.org> wrote:
>>>
>>>> Author: rjmccall
>>>> Date: Tue Sep  8 04:18:30 2015
>>>> New Revision: 246991
>>>>
>>>> URL: http://llvm.org/viewvc/llvm-project?rev=246991&view=rev
>>>> Log:
>>>> When building the alloca for a local variable, set its name
>>>> separately from building the instruction so that it's
>>>> preserved even in -Asserts builds.
>>>>
>>>
>>> Why do we want to preserve this name in particular in -Asserts builds?
>>>
>>>
>>> It’s a fairly small runtime cost — no twine appending, only present on a
>>> tiny fraction of instructions, unlikely to collide within a single function
>>> — that makes reading the IR dramatically easier.  That’s still useful to be
>>> able to do with a production compiler, both for us and for sophisticated
>>> users.
>>>
>>> IIRC, Daniel did the perf work on IR names way back when; he might have
>>> other thoughts.
>>>
>>
>> *nod* I'd be curious to see/hear/better understand the tradeoffs. I know
>> I've talked to Chandler (CC'd) about it before when he's expressed a desire
>> to want to more fully remove names from non-asserts builds (specifically:
>> optimizations still create named values even in -Asserts builds, I think?
>> or maybe there was some other - oh, right, as you were saying, the Twine
>> building cost, even when the name isn't used some names are complex to
>> create, etc)
>>
>>
>>>
>>> John.
>>>
>>
>> _______________________________________________
>> cfe-commits mailing list
>> cfe-commits at lists.llvm.org
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20150908/e77093f7/attachment.html>