Re: calloc = malloc + memset

[Marc Glisse <marc dot glisse at inria dot fr>]

On Sat, 1 Mar 2014, Paolo Carlini wrote:

> Hi
>
> > On 28/feb/2014, at 23:48, Marc Glisse <marc.glisse@inria.fr> wrote:
> >
> > Hello,
> >
> > this is a stage 1 patch, and I'll ping it then, but if you have comments now...
> >
> > Passes bootstrap+testsuite on x86_64-linux-gnu.
> >
> > 2014-02-28  Marc Glisse  <marc.glisse@inria.fr>
> >
> >    PR tree-optimization/57742
> > gcc/
> >    * tree-ssa-forwprop.c (simplify_malloc_memset): New function.
> >    (simplify_builtin_call): Call it.
> > gcc/testsuite/
> >    * g++.dg/tree-ssa/calloc.C: New testcase.
> >    * gcc.dg/tree-ssa/calloc.c: Likewise.
> >
> > --
> > Marc Glisse
> > Index: gcc/testsuite/g++.dg/tree-ssa/calloc.C
> > ===================================================================
> > --- gcc/testsuite/g++.dg/tree-ssa/calloc.C    (revision 0)
> > +++ gcc/testsuite/g++.dg/tree-ssa/calloc.C    (working copy)
> > @@ -0,0 +1,35 @@
> > +/* { dg-do compile } */
> > +/* { dg-options "-std=gnu++11 -O3 -fdump-tree-optimized" } */
> > +
> > +#include <new>
> > +#include <vector>
> > +#include <cstdlib>
> > +
> > +void g(void*);
> > +inline void* operator new(std::size_t sz) _GLIBCXX_THROW (std::bad_alloc)
>
> Unless *really* necessary I would recommend not including the large 
> <vector> (that also couples quite seriously the front-end testsuite to 
> the library testsuite, we already discussed those topics in the past). 
> Using the internal macros seems also unnecessary.

I think it might be the first time I include large headers in a compiler 
testcase (note that there are already 16 other testcases including 
<vector> in g++.dg). In this case, it seems to be what I want to test 
though. I already have some elementary tests in gcc.dg. This testcase is 
the original motivation for working on this. It requires a combination of 
quite a few optimizations (inlining, recognizing that a loop is a memset, 
aliasing, this optimization (the complicated version with a PHI node)), 
and I want to test that we won't for instance shuffle the passes in a way 
that breaks it. Also, if the library changes vector enough that this 
doesn't optimize anymore, I want to know about it, either the library 
change was wrong or the middle-end needs to improve some optimization 
before the next release.


I wanted to keep the implementation of new as close to the one in 
libsupc++ as possible (mimic LTO), so I copy-pasted (and slightly edited, 
I may propose a patch to libsupc++ later). I agree that I should remove 
the exception specification (since I am compiling in C++11 to have access 
to get_new_handler) and replace _GLIBCXX_THROW_OR_ABORT with just throw, 
and I just did it locally, thanks.

--
Marc Glisse