Sage 5.6 on ARM

[Julien Puydt]

Hi,

here is where we stand:

- build: ok, even though libm4rie's conversion.c is a pain ;

- ptestlong: three failing tests, related to libc's gamma function (as
usual... I should find the time to fix that...) ;

- bdist: available in the bdist/ directory of my home dir (~jpuydt) on
sage.math.washington.edu


Snark on #sagemath

PS: in case someone has access to another ARM box than ubuntu-based,
here is which files give problems ; please report if you don't see errors :
sage -t --long -force_lib devel/sage/sage/functions/other.py
sage -t --long -force_lib devel/sage/sage/symbolic/expression.pyx
sage -t --long -force_lib devel/sage/sage/rings/arith.py

[Dima Pasechnik]

I wonder why exactly these files also produce numerical noise on Cygwin,
as I reported:
http://boxen.math.washington.edu/home/dima/tmp/cygwin_sage-5.7.beta0.test.log
E.g.
File
"/usr/local/src/sage/sage-5.7.beta0/devel/sage/sage/rings/arith.py",
line 3162:
sage: binomial(0.5r, 5)
Expected:
0.02734375
Got:
0.02734375000000001

How one does get numerical noise like this on an Intel, is beyond me.

Dima

[Julien Puydt]

Le 28/01/2013 11:55, Dima Pasechnik a �crit :

Eh, I was hoping someone would tell me "I have an ARM and I don't see
the problem" so I could have a look at the libc used and see if that
helped fixing the issue.

And here you come, with another place where similar problems arise. And
I notice that the value you "Got:" is the same I see on ARM :-P

Snark on #sagemath

[Dima Pasechnik]

On Monday, 28 January 2013 19:05:31 UTC+8, Snark wrote:

Le 28/01/2013 11:55, Dima Pasechnik a �crit :

This is a eglibc/glibc problem, we reported on 

https://bugs.launchpad.net/ubuntu/+source/eglibc/+bug/713985

What I cannot   recall is why this problem does not show up on (32-bit) Linux x86.

Note that the code in question (on bug/713985) gives almost the same output as on ARM

on such a platform, too.

But somehow Sage manages to work around this.

Dima

Snark on #sagemath

[Julien Puydt]

Le 29/01/2013 09:01, Dima Pasechnik a �crit :

> This is a eglibc/glibc problem, we reported on
> https://bugs.launchpad.net/ubuntu/+source/eglibc/+bug/713985
>
> What I cannot recall is why this problem does not show up on (32-bit)
> Linux x86.
> Note that the code in question (on bug/713985) gives almost the same
> output as on ARM
> on such a platform, too.
> But somehow Sage manages to work around this.

Ah, I remember that bug report, but I didn't remember that it could also
hit non-ARM processors!

Snark on #sagemath

[mmarco]

Is it available for public download?

[Julien Puydt]

Le 29/01/2013 18:10, mmarco a �crit :

> Is it available for public download?

According to the mail I started the thread with, it is :

- bdist: available in the bdist/ directory of my home dir (~jpuydt) on
sage.math.washington.edu

:-P

Snark on #sagemath

[Dima Pasechnik]

On 2013-01-29, Julien Puydt <julien...@laposte.net> wrote:

> Le 29/01/2013 09:01, Dima Pasechnik a écrit :
>> This is a eglibc/glibc problem, we reported on
>> https://bugs.launchpad.net/ubuntu/+source/eglibc/+bug/713985
>>
>> What I cannot recall is why this problem does not show up on (32-bit)
>> Linux x86.
>> Note that the code in question (on bug/713985) gives almost the same
>> output as on ARM
>> on such a platform, too.
>> But somehow Sage manages to work around this.
>
> Ah, I remember that bug report, but I didn't remember that it could also
> hit non-ARM processors!

this must be some dark Ginac/Pynac magic happening here...

>
> Snark on #sagemath
>

[mmarco]

I cannot access to your directory.

[tom d]

I can confirm that this same numerical noise (binomial, gamma functions) happens on the ARM6 on the Raspberry Pi, running the 'raspian' distro.  (Which isn't Ubuntu, though it is a Debian derivative.)

[Julien Puydt]

Le 30/01/2013 12:49, mmarco a �crit :

> I cannot access to your directory.

Hmmm... I don't understand why:

- /home is drwxr-xr-x

- /home/jpuydt is drwxr-xr-x

- /home/jpuydt/bdist is drwxr-xr-x

- /home/jpuydt/bdist/sage-5.6-armv7l-Linux.tar.xz is -rw-r--r--

so you should have the right to enter, list and read all the way down...

Snark on #sagemath

[mmarco]

Maybe it should be inside public_html?

[Jean-Pierre Flori]

Get similar problems on Cygwin as well.
See
http://boxen.math.washington.edu/home/jpflori/ptest-5.7.beta1-cygwin.log

Dima

[Jean-Pierre Flori]

Not sure it is related, but please note that lib m does not implement long double (for several or all functions, don't know) on Cygwin.

[Dima Pasechnik]

On 2013-01-30, mmarco <mma...@unizar.es> wrote:
> Maybe it should be inside public_html?

try this:
http://boxen.math.washington.edu/home/jpuydt/bdist/