[petsc-dev] [petsc-maint #69143] petsc-dev performance issues

[Satish Balay]

On Tue, 5 Apr 2011, Barry Smith wrote:

> 
> On Apr 5, 2011, at 8:56 AM, Satish Balay wrote:
> 
> > On Tue, 5 Apr 2011, Dmitry Karpeev wrote:
> > 
> >> Okay, but what's going on with petsc-3.1 on that machine?  It doesn't fail.
> >> Is that because dlsym(0,string) works?
> >> Or was dynamic loading disabled in it?
> > 
> > --with-dynamic=0 is usually the default buid.
> > 
> > petsc-3.1 never did dlopen()/dlsym() for this case - but petsc-dev
> > does do dlopen()/dlsym() even with --with-dynamic=0 for basic ex2
> > example code - hece it crashes.
> 
>    Yes, but the crash is only due to incomplete checking of functionality in configure before using it, not due to a bad model of how things should be one.
> 

Whats the difference with dlopen(executable) vs dlopen(libpetsc.so)
then? And what is the point of --with-dynamic=1 now?

If you really wish to make this the default - remove this ambiguty and
go back to a single flag [HAVE_DYNAMIC + USE_DYNAMIC] name it whatver
- and then set --with-dynamic=1 as default. And let the user disable
it if he wishes.

So lisandro should be happy that this is available by default.

Satish