[cfe-commits] [PATCH] CFG automatic object destructors

[Marcin Świderski]

I did the test (twice for each version) and the build time did not differ
visibly.

W dniu 28 września 2010 09:32 użytkownik Ted Kremenek
<kremenek at apple.com>napisał:

> On Sep 27, 2010, at 10:40 PM, Marcin Świderski <marcin.sfider at gmail.com>
> wrote:
>
> > Besides specific comments (below), I'm concerned about the impact on
> CFG-construction time when scope building is *not* enabled.  This is
> currently the common case for uses of the CFG in Sema.  I think there
> shouldn't be any noticeable penalty since most of the logic is guarded with
> a flag, but it is worth verifying.
> >
> > Is there any performance benchmark prepared for such cases?
>
> I'm mainly concerned about the impact on compile time, which should be
> nonexistent since dtors shouldn't be included in the cfg by default.
>
> One good benchmark: build all of llvm/clang with a Release-Asserts build of
> clang that (1) includes your patch and (2) doesn't include your patch.  If
> the time for (1) is significantly larger than (2), that is room for concern.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20100930/ebf854b4/attachment.html>