[petsc-dev] patches for ICC/Cholesky preconditioner on GPUs

[Jed Brown]

On Sun, Mar 24, 2013 at 3:17 PM, Paul Mullowney <paulm at txcorp.com> wrote:

> Jed,
>
> I think I've resolved all of your comments. I have some questions.
>
> 1) Tests for the nightly test harness
>
> (a) Is there a naming convention for the tests?
> I have written two examples named aijcusparse_ex_ilu.c and
> aijcusparse_ex_icc.c. Is this ok?
>

They source code files should be numbered unless we change policy. If we
were to change, we would still need some systematic naming so that
.gitignore would not be interrupted, perhaps 'ex071-aijcusp-icc.c'. Note
that some projects use sequential numbering to indicate "level" and
priority of tests (lower numbers run first). For now, just use numbers.


> (b) What directory should they live in?
> I assume these go in ksp/ksp/examples/tutorials
>

If they are well-documented, thus suitable as _tutorials_, they can go in
the tutorials directory. If they are _tests_, they should be in
examples/tests/.


>
> (c) How do I ensure that they tests are built and run every night with
> txpetscgpu package?
>

Use TESTEXAMPLES_TXPETSCGPU in conf/test and in the examples makefile.
You'll see lots of other packages.


> (d) Where do I put binary matrix files that these tests rely upon?
>

Put them somewhere and we'll add them here:

ftp://ftp.mcs.anl.gov/pub/petsc/matrices/

(We could consider nicer ways to maintain these.)


> 2) Doing
> > git status
> shows
>     Your branch is ahead of 'origin/pm/aijcusparse-icc' by 2 commits.
>
> I assume that when I push to the fork, these changes will be uploaded
> remotely?
>

Yes, you can use 'git push --dry-run' to see what would happen if you push.
After pushing to your fork, open a pull request.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20130324/c612c007/attachment.html>