[cfe-commits] r130453 - in /cfe/trunk: lib/CodeGen/CGExprAgg.cpp lib/CodeGen/CGExprCXX.cpp test/CodeGenCXX/value-init.cpp

[Argyrios Kyrtzidis]

I should have made it clear but the patch doesn't affect zero initialization of global variables, they are zero-initialized at declaration time with LLVM's zeroinitializer.

On Apr 28, 2011, at 4:42 PM, Chandler Carruth wrote:

> On Thu, Apr 28, 2011 at 4:36 PM, Chris Lattner <clattner at apple.com> wrote:
> 
> On Apr 28, 2011, at 3:57 PM, Argyrios Kyrtzidis wrote:
> 
> > Author: akirtzidis
> > Date: Thu Apr 28 17:57:55 2011
> > New Revision: 130453
> >
> > URL: http://llvm.org/viewvc/llvm-project?rev=130453&view=rev
> > Log:
> > Cut down unnecessary zero'ing when value-initializing arrays of C++ objects.
> >
> > -C++ objects with user-declared constructor don't need zero'ing.
> 
> Are you sure that this is true?  Apparently the C++ standard requires that global variables (at least) are zero initialized and that constructors can take advantage of this.  This is apparently well defined:
> 
> struct foo {
>  int x;
> 
>  foo() {
>   ++x;
>  }
> } f;
> 
> I'm pretty sure that this doesn't apply to temporaries etc, but it would be worth checking if you haven't already.
> 
> I checked. Variables of static and thread storage must be zero initialized first as you describe. This is governed by [basic.start.init]p2 in the C++11 ISFD.
> 
> The initialization chapter provides no such pre-initialization for the other storage classes.
>  
> 
> -Chris
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20110428/6586c567/attachment.html>