Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Sat, 12 Jun 2010 21:39:59 +0000
Subject: Re: gcc 4.5.0, auto-parallelization, OpenMP


Thanks for the page.

What about updating Openwall to use a modern yet stable compiler (such as gcc-4.3.5, which is an older and stable branch but still under maintenance upstream)?  It's great that the software is very stable.  However, the -fstack-protector implementation for the gcc3 branch is "not perfect, and can cause problems", and gcc-4 may be preferred for compatibility reasons.  Anyhow I would like to update the system compiler to use a more recent compiler, and will do so on my own system.  The updated gcc idea may not catch on but...

In my opinion, it would be great to update the Openwall toolchain to create hardened binaries based upon the ideas outlined on:

Your thoughts are welcome.

On Tue, 4 May 2010 19:14:10 +0400, Solar Designer wrote:

Here's a wiki page I've created on building and using the brand new gcc
4.5.0 on Owl:

Compared to gcc 3.4.5, which is currently included with Owl, more recent
versions of gcc add auto-vectorization (such as use of MMX/SSE*
instruction sets on x86 and x86-64, if permitted), auto-parallelization,
and support for OpenMP.  (Of course, this is a simplified and far from
exhaustive list of improvements.)

The wiki page demonstrates how to make a build of gcc capable of
auto-parallelization, briefly shows its limitations, and finally shows
how OpenMP directives may be used to get around the limitations.

This may be useful to those of you who might use (or consider to use)
Owl for software development and/or scientific computing.

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.