GMP on Core 2 Duo
tg-this-will-bounce-but-I-am-subscribed-to-the-list-honest at swox.com
Wed Nov 29 12:13:05 CET 2006
Ernest Prabhakar <prabhaka at apple.com> writes:
As the Open Source Product Manager for Mac OS X, I wanted to thank
you for the work you've done on GMP. I realize we aren't your
primary target, but a lot of scientists and researchers (including at
Apple) benefit greatly from your highly-tuned library.
Isn't it fascinating how large enterprises can benefit from a
completely non-funded hobby project?
Because of that, I wanted to see if there's anything I can do ensure
Mac OS X properly supports GMP. I understand there have been
compiler issues in the past that caused it fail "make check", but I
hope those have already been addressed. If not, please let me know.
There have been recurrent problems, I reported a handful in the
beginning of the Mac OS X era, but decided I had more fun things to do
at my leisure time.
One of the latest XCode releases cannot compile GMP on PPC. I have no
idea of the status of your compiler in X86 wrt compiling GMP.
In addition, we are very excited at the prospect of moving our whole
product line over to 64-bit processors, and we'd love to take
advantage of the enormous performance improvements you've documented
. I notice that Jason Martin has patches to build a 64-bit
version on Mac OS X and Linux, which has been picked up by Fink.
I don't think any patches are needed to build a 64-bit version on
I realize it is often tricky to integrate platform-specific changes
into a multi-platform codebase, so I was wondering if there was any
way we could help you turn these patches into something that would
"play nicely" with your existing codebase, so that GMP would work
properly out of the box on Core 2 Duo systems from Apple and other
You need to understand how most GNU projects work. Either the
maintainer(s) does something because (s)he think it is interesting, or
(s)he does it because he is paid to do it. Accepting patches from
outsiders requires a lot of work, but is usually not very interesting.
In the case of GMP, quality generic patches tend to be integrated
reasonably timely, while vendor specific patches tend to get much less
More information about the gmp-devel