Modified solver routines cr.c, bcgs.c and bcgsl.c

[Matthew Knepley]

On Mon, Sep 22, 2008 at 11:01 AM, Barry Smith <bsmith at mcs.anl.gov> wrote:
>
>   Stephen,
>
>     Thank you. We'll update PETSc with some of the changes.

I am doing it. However, there is a strange bug in the cr.c you sent.
If you look at
line 111, you are using ai instead of -ai. I went back through the
log, and Barry
changed this, but also removed the line

  ai = -ai;

beforehand. This is missing in your version, even though you have not made the
sign change. Not sure how this could have run.

  Matt

>    Barry
>
> On Sep 22, 2008, at 9:54 AM, Stephen R Ball wrote:
>
>>
>> Hi
>>
>> I thought it best to send this to petsc-dev rather than petsc-users.
>>
>> One of our PETSc users has noticed that they obtain better performance
>> when they use the modified versions of solver routines cr.c, bcgs.c and
>> bcgsl.c attached.
>> In places they are using routines VecDotBegin()/VecDotEnd() instead of
>> VecDot() and VecNormBegin()/VecNormEnd() instead of VecNorm().
>>
>> I am sending them to you so that you may test them with a view to you
>> possibly implementing them within PETSc if you find they improve
>> performance.
>>
>> Regards
>>
>> Stephen R. Ball
>> Advanced Technologies
>> HPC
>> DRAS
>> Rm: G17
>> Bldg: E1.1
>> AWE(A)
>> Aldermaston
>> Reading
>> Berkshire
>> ENGLAND
>> RG7 4PR
>> Tel: +44 (0)118 982 4528
>> e-mail: stephen.r.ball at awe.co.uk
>>
>>
>> <<cr.c>>  <<bcgs.c>>  <<bcgsl.c>>
>> <cr.c><bcgs.c><bcgsl.c>
>
>



-- 
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which
their experiments lead.
-- Norbert Wiener