Re: Helping and developing
Alistair Boyle
<alessand...@gmail.com> wrote:
> Hi
> I read you presentation and I found it very interesting.
> Your code do the same things i was trying to implement (testing large matrix
> on a shared memory machine) and use almost the same software stack (like
> pardiso and SuperLU).
Great! That's what open-source is supposed to be about.
> We want to test our fat node performance using some input file from the
> following matrix collection
> http://www.cise.ufl.edu/research/sparse/matrices/
> Unfortunately our OS (CentOS 5.5 x86-64) has autoconf version 2.59
> while Autoconf version 2.65 or higher is required.
> Can you help me in compiling your code?
That shouldn't be a problem. Where are you getting the code from: is
it the git repository or are you downloading the .tar.gz file?
The tar file is on the github page, under "downloads".
https://github.com/boyle/meagre-crowd
If you are using the git repository then you need to regenerate all
the stuff that isn't stored in the repository with ./autogen. On the
other hand, the .tar.gz file *should* have all the files you need, so
you can just untar it and run ./configure, make, make install,
irrespective of Autoconf versions. This should be true but it might
not work so give it a try and report back.
I'm using this version of Autogen because its what's on my local
computer. You could try changing the line "AC_PREREQ([2.65])" in
configure.ac to 2.59 though I suspect that some configure macros may
be broken by this.
> I'm also interested in helping you
> to developing some other features like described in your TODO file.
That is great. At the moment, meagre-crowd is still at a pretty early
stage. I've mostly been working on getting some basic functionality:
adding solvers and checking that they can all get a (correct) answer.
In the coming months I'll need to move on to tuning the solvers,
keeping them up-to-date with the latest releases, and adding more
functionality. That will certainly be made lighter work if there are
more hands. I have quite a few ideas that haven't made it to the TODO
file yet, either.
If you have suggestions or patches they'd be quite welcome.
> Thanks a lot
> AT
Alistair
Alistair Boyle
<alessand...@gmail.com> wrote:
>> That shouldn't be a problem. Where are you getting the code from: is
>> it the git repository or are you downloading the .tar.gz file?
>
>
>> The tar file is on the github page, under "downloads".
>>
>> https://github.com/boyle/meagre-crowd
>>
>> If you are using the git repository then you need to regenerate all
>> the stuff that isn't stored in the repository with ./autogen. On the
>> other hand, the .tar.gz file *should* have all the files you need, so
>> you can just untar it and run ./configure, make, make install,
>> irrespective of Autoconf versions. This should be true but it might
>> not work so give it a try and report back.
>
> (thi is what i'm downloading: boyle-meagre-crowd-0.4.2-0-g099a402.tar.gz)
Ah. I think I see the confusion.
https://github.com/downloads/boyle/meagre-crowd/meagre-crowd-0.4.2.tar.gz
downloading this file and the configure script is there. If you
clicked on "Download .tar.gz" with "branch: master" on the same line,
you get this
https://github.com/boyle/meagre-crowd/tarball/master
which is a tar of the current git repository. This looks like what
you've got. You want the first link (meagre-crowd-0.4.2.tar.gz) which
has the full distribution.
>> I'm using this version of Autogen because its what's on my local
>> computer. You could try changing the line "AC_PREREQ([2.65])" in
>> configure.ac to 2.59 though I suspect that some configure macros may
>> be broken by this.
>>
>
> Your suspect was right.
>
>
>>
>> > I'm also interested in helping you
>> > to developing some other features like described in your TODO file.
>>
>> That is great. At the moment, meagre-crowd is still at a pretty early
>> stage. I've mostly been working on getting some basic functionality:
>> adding solvers and checking that they can all get a (correct) answer.
>> In the coming months I'll need to move on to tuning the solvers,
>> keeping them up-to-date with the latest releases, and adding more
>> functionality. That will certainly be made lighter work if there are
>> more hands. I have quite a few ideas that haven't made it to the TODO
>> file yet, either.
>>
>> If you have suggestions or patches they'd be quite welcome.
>
>
> I only read the presentation by now. I hope to see the code in action soon.
Alistair Boyle
<alessand...@gmail.com> wrote:
> Thanks a lot!
> I'm running ./configure now
> that's the result:
> checking for library containing load_sparse_matrix... no
> configure: error: missing BeBOP sparse matrix I/O library
>
> I already have installed the BeBOP benchmark and set the LD_LIBRARY_PATH.
> I also set the BeBOP sparse matrix I/O library directory in the LDFLAGS
> variable for the ./configure command, but no success. How I can specify the
> path to ./configure so that BeBOP became visible?
> Thanks
> AT
Now you're at the tricky part. The dependencies all have their
particular funkiness for what needs to be done to get them installed.
I 've setup the script get_dependencies, to compile and install the
stuff I need. This probably won't work for you, but its a template. In
particular I've converted all the static libraries to dynamic
libraries (.a -> .so) and fixed some compile problems for almost every
package.
If you use the get_dependencies script as a template you'll at least
have a bit of a head start. You can also see what header files need to
be installed from the packages and what dynamic library libXYZ.so
meagre-crowd is expecting to find in the library directory.
This part is a real pain. The packaging of most of this software is
quite challenging...
By the way, please include the mailing list in your replies
(meagre...@googlegroups.com) so others can learn from what you're
going through.
Alistair