Sundials and pthread issues

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
9 messages Options
Reply | Threaded
Open this post in threaded view
|

Sundials and pthread issues

autodidact
I got the development version  of octave from the mercurial repo and tried
compiling it but I am getting these warnings:

checking for the pthreads library -lpthreads... no
configure: WARNING: SUNDIALS IDA library does not include the IDADENSE
linear solver, ode15i and ode15s will be disabled
configure: WARNING: SUNDIALS IDA library not configured with IDAKLU, ode15i
and ode15s will not support the sparse Jacobian feature

I already installed the sundials-devel rpm package and I have glibc which
contains pthread. Configure's output does show that it finds pthread.h so
I'm not sure why it can't find pthread. The output of running configure is
on  pastebin <https://pastebin.com/3bjVn055>  . I've highlight the important
bits to make skimming the code easier.



--
Sent from: http://octave.1599824.n4.nabble.com/Octave-General-f1599825.html


Reply | Threaded
Open this post in threaded view
|

Re: Sundials and pthread issues

Mike Miller-4
On Sun, Jul 01, 2018 at 10:30:20 -0700, autodidact wrote:

> I got the development version  of octave from the mercurial repo and tried
> compiling it but I am getting these warnings:
>
> checking for the pthreads library -lpthreads... no
> configure: WARNING: SUNDIALS IDA library does not include the IDADENSE
> linear solver, ode15i and ode15s will be disabled
> configure: WARNING: SUNDIALS IDA library not configured with IDAKLU, ode15i
> and ode15s will not support the sparse Jacobian feature
>
> I already installed the sundials-devel rpm package and I have glibc which
> contains pthread. Configure's output does show that it finds pthread.h so
> I'm not sure why it can't find pthread. The output of running configure is
> on  pastebin <https://pastebin.com/3bjVn055>  . I've highlight the important
> bits to make skimming the code easier.
First, there is no problem with pthreads. On GNU/Linux systems there is
no libpthreads, the feature is enabled with the -pthread compiler
option. The configure script is correctly reporting that there is no
libptheads library on your system.

Octave is not yet compatible with SUNDIALS version 3, so the warning
messages are explaining that and correctly disabling all SUNDIALS
features. If you want to help ensure that Octave 5 will be compatible
with SUNDIALS 3, you can participate in the bug report

  https://savannah.gnu.org/bugs/?52475

--
mike



signature.asc (849 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: Sundials and pthread issues

autodidact
Atleast I know now the only issue is sundials 3. Will I still be able to use
the rest of octave's features as normal if I go ahead and install? Without
sundials, does it mean that octave cannot do any ODE's, and when is the
expected release of octave 5? Last question, how can I participate in the
bug report for sundials compatibility when sundials is disabled?



--
Sent from: http://octave.1599824.n4.nabble.com/Octave-General-f1599825.html


Reply | Threaded
Open this post in threaded view
|

Re: Sundials and pthread issues

Carlo de Falco-2


> On 2 Jul 2018, at 10:43, autodidact <[hidden email]> wrote:
>
> Without
> sundials, does it mean that octave cannot do any ODE's,

No. Sundials is used as a backend for the implementation of the matlab compatible DAE solvers ode15i and ode15s only.
Without sundials those two functions won't work but you can still use daspk to solve the same class of problems (but your code will not be matlab compatible).
If your ODEs are not very stiff you can also use other solvers like ode45 and ode23 which are included in Octave 4.4.

> and when is the expected release of octave 5?

The standard answer to this question is "whenever it's ready, sooner if you help",
but Octave is trying to move to a more frequent release cycle so maybe
it will not take very long.

> Last question, how can I participate in the
> bug report for sundials compatibility when sundials is disabled?

There is a patch attached to that report, if you apply the patch before compiling
you should be able to link Octave to sundials 3 and therefore re-enable sundials.

The code in the patch is already fairly complete (thanks to Bill Greene),
what is not working correctly yet is the automatic discovery of sundials 3
with autotools tests.

If you are familiar with autotools, that is the area where help is more needed.

c.




Reply | Threaded
Open this post in threaded view
|

Re: Sundials and pthread issues

Carlo de Falco-2
In reply to this post by autodidact


> On 2 Jul 2018, at 10:43, autodidact <[hidden email]> wrote:
>
> Without
> sundials, does it mean that octave cannot do any ODE's,

No. Sundials is used as a backend for the implementation of the matlab compatible DAE solvers ode15i and ode15s only.
Without sundials those two functions won't work but you can still use daspk to solve the same class of problems (but your code will not be matlab compatible).
If your ODEs are not very stiff you can also use other solvers like ode45 and ode23 which are included in Octave 4.4.

> and when is the expected release of octave 5?

The standard answer to this question is "whenever it's ready, sooner if you help",
but Octave is trying to move to a more frequent release cycle so maybe
it will not take very long.

> Last question, how can I participate in the
> bug report for sundials compatibility when sundials is disabled?

There is a patch attached to that report, if you apply the patch before compiling
you should be able to link Octave to sundials 3 and therefore re-enable sundials.

The code in the patch is already fairly complete (thanks to Bill Greene),
what is not working correctly yet is the automatic discovery of sundials 3
with autotools tests.

If you are familiar with autotools, that is the area where help is more needed.

c.




Reply | Threaded
Open this post in threaded view
|

Re: Sundials and pthread issues

Carlo de Falco-2
In reply to this post by autodidact


> On 2 Jul 2018, at 10:43, autodidact <[hidden email]> wrote:
>
> Without
> sundials, does it mean that octave cannot do any ODE's,

No. Sundials is used as a backend for the implementation of the matlab compatible DAE solvers ode15i and ode15s only.
Without sundials those two functions won't work but you can still use daspk to solve the same class of problems (but your code will not be matlab compatible).
If your ODEs are not very stiff you can also use other solvers like ode45 and ode23 which are included in Octave 4.4.

> and when is the expected release of octave 5?

The standard answer to this question is "whenever it's ready, sooner if you help",
but Octave is trying to move to a more frequent release cycle so maybe
it will not take very long.

> Last question, how can I participate in the
> bug report for sundials compatibility when sundials is disabled?

There is a patch attached to that report, if you apply the patch before compiling
you should be able to link Octave to sundials 3 and therefore re-enable sundials.

The code in the patch is already fairly complete (thanks to Bill Greene),
what is not working correctly yet is the automatic discovery of sundials 3
with autotools tests.

If you are familiar with autotools, that is the area where help is more needed.

c.




Reply | Threaded
Open this post in threaded view
|

Re: Sundials and pthread issues

autodidact
I read the bug report and feel comforable with testing the bug fixes. First,
want to confirm what the latest bug fix I should be testing is? Is it file
#44110 added by bill greene?



--
Sent from: http://octave.1599824.n4.nabble.com/Octave-General-f1599825.html


Reply | Threaded
Open this post in threaded view
|

Re: Sundials and pthread issues

autodidact
In reply to this post by Carlo de Falco-2
I want to add on what I tried so far. After picking out the patch that seemed
most recent based on the comments, I downloaded open_TK6CSjvd.txt and
replaced __ode15__.cc with it. I also renamed the downloaded file to the
same name as the file it replaced. I then ran ./bootstrap and the first two
lines showed:

libtoolize:   error: 'build-aux/config.sub' is newer: use '--force' to
overwrite
libtoolize:   error: 'build-aux/install-sh' is newer: use '--force' to
overwrite

Then I ran /home/user/octave/src/configure  --prefix=/usr/local/octave/dev
--with-blas=openblas64 FFLAGS=-fdefault-integer-8 | less

Result:

No change, sundials NVECTOR serial library and IDA library still not found.
Where do I go from here?



--
Sent from: http://octave.1599824.n4.nabble.com/Octave-General-f1599825.html


Reply | Threaded
Open this post in threaded view
|

Re: Sundials and pthread issues

Carlo de Falco-2
In reply to this post by autodidact


> On 2 Jul 2018, at 17:58, autodidact <[hidden email]> wrote:
>
> I read the bug report and feel comforable with testing the bug fixes. First,
> want to confirm what the latest bug fix I should be testing is? Is it file
> #44110 added by bill greene?
>

please move this discussion to the tracker.
c.