On Apr 25, 6:27 am, "William Stein" <[EMAIL PROTECTED]> wrote:
> On 4/24/07, mabshoff <[EMAIL PROTECTED]> wrote:
>
>
>
> > Hello,
>
> > I did some googling and
>
> > a) The problem with "threadlist_ix -1" also happens with sage 2.4.2
> > with the current cygwin
> > b) The cygwin mailing list has lots of reports about "threadlist_ix
> > -1" with lots of different executables (emacs, pari, tar ...), but so
> > far no solution to the problem. It has been suggested that the problem
> > is caused by certain buggy drivers like certain webcams, but so far no
> > conformation.
>
> Thanks. Is this an issue only with genus2reduction and mwrank? Or
> does it happen with other SAGE components? genus2reduction is
> a standalone program that almost nobody uses. mwrank is more important
> though, but at least it's all open source and I know the author very well.
>
mwrank seems to be most effected, but I have seen the same issue for
pari mentioned on its mailing list.
> > I want to install a current snapshot of cygwin later on, but I am not
> > very confident that the problem will go away. If anybody else has an
> > idea please let me know.
>
> > I am also suprised that running the tests takes 7000s on a 1.5GHZ
> > Pentium M Laptop with 0.5G Ram. Can anybody confirm this or is it time
> > that I reinstall Windows? Maybe all the problems will then magically
> > go away :)
>
> Cygwin is fairly slow (and if you have any virus scan software it will be
> much much slower yet). Especially startup times and starting new
> interfaces takes a long time -- for some reasons Cygwin forking from
> Python is very slow, e.g., 3 seconds to start an interface.
That's pretty bad.
> Once SAGE starts an interfaces the interface itself works
> just fine -- it's the startup time that is slow. For the SAGE doctests
> that is a killer timewise, because each file requires starting SAGE
> (and sometimes interfaces) from scratch.
>
> For a while we didn't support cygwin and only distributed SAGE using
> colinux and/or vmware. But colinux isn't really that good for various
> reasons (though performance wasn't bad), and vmware can be painful
> as well -- the download for sage in vmware is huge (500MB), and
> setting up appropriate networking between vmware and windows
> to run the notebook very often doesn't just work. Overall Cygwin seems
> to be the best tradeoff. If I had way way more resources I would
> investigate other options.
>
I think that it might be worth investigating to compile sage and its
libraries with MSVC and run all the external executables with cygwin
for the time being. pyrex seems to support msvc 2003 and up, so there
is a chance I guess.
> Providing excellent support for Windows is of course a high priority
> because MS Windows is by far the most popular operating system.
> But it's challenging because SAGE is a collection of dozens
> open source math software programs, and
> most of those programs are Windows unfriendly (their developers
> mostly use Linux). Fortunately Python, which is the core of SAGE, is
> pretty Windows friendly.
>
Yep. The main issue I see on the horizon is cygwin's 32 bit limit. A
lot of systems today (even laptops) get sold with 2G+ Ram and the way
it looks cygwin will not support 64 bit binaries anytime soon now.
Obviouly that doesn't hurt the casual user, but I routinely compute
GBases that need 8GB+ (not on Windows, though). I always thought that
most serious computer algebra people would use Linux/Unix or nowadays
even MacOSX, but there is a frightening number of people out there who
only like and use Windows or are forced to use it. Virtualisation will
help in the future, but as you stated the technical issues of
networking are somewhat of an issue, while the performance penalty
will be lessened by better hardware virtualisation in the future.
> -- William
Cheers,
Michael
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to sage-devel@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at http://groups.google.com/group/sage-devel
URLs: http://sage.scipy.org/sage/ and http://modular.math.washington.edu/sage/
-~----------~----~----~----~------~----~------~--~---
