This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [patch] libgomp: removing nodlopen flag for portability
- From: "Andrew Pinski" <pinskia at gmail dot com>
- To: "Jakub Jelinek" <jakub at redhat dot com>
- Cc: "Sebastian Steiger" <steigers at phys dot ethz dot ch>, gcc-patches at gcc dot gnu dot org
- Date: Wed, 2 May 2007 07:01:08 -0700
- Subject: Re: [patch] libgomp: removing nodlopen flag for portability
- References: <463890C3.6030801@phys.ethz.ch> <20070502133650.GD355@devserv.devel.redhat.com>
On 5/2/07, Jakub Jelinek <jakub@redhat.com> wrote:
This is a bad idea, global-dynamic model is much slower than initial-exec.
How much slower? And do we care that much for the library itself?
You can LD_PRELOAD=libgomp.so.1 python ...
You know there are such things as plugins and plugins can be linked in
with other shared libraries? This is exactly what Sebastian is doing.
People who use his plugins are not going to do
"LD_PRELOAD=libgomp.so.1" because it is a pain in the ass. Right now
we basically are saying we don't support plugins compiled with openmp
turned on which is wrong, we want people to write more code rather
than less.
-- Pinski