[petsc-users] compare: performance of petsc_3.3p2 and petsc_dev-r24197

[Jed Brown]

Also,

*      ##########################################################*
*      #                                                        #*
*      #                          WARNING!!!                    #*
*      #                                                        #*
*      #   This code was compiled with a debugging option,      #*
*      #   To get timing results run ./configure                #*
*      #   using --with-debugging=no, the performance will      #*
*      #   be generally two or three times faster.              #*
*      #                                                        #*
*      ##########################################################*


On Wed, Aug 22, 2012 at 6:07 AM, Matthew Knepley <knepley at gmail.com> wrote:

> On Wed, Aug 22, 2012 at 5:52 AM, Feng Li <fengshen186 at gmail.com> wrote:
>
>> Thank you for your help!
>> The log_summary result is in the attachment.
>> Please kindly check the attachment!
>>
>
> This is not a PETSc change, its a PFLOTRAN change. You have changed the way
> the Jacobian is evaluated. Look at the 3.3 log
>
> SNESSolve             18 1.0 1.2947e+01 1.0 1.22e+09 1.0 0.0e+00 0.0e+00
> 0.0e+00 44 83  0  0  0  93100  0  0  0    94
> SNESFunctionEval      39 1.0 3.3130e+00 1.0 0.00e+00 0.0 0.0e+00 0.0e+00
> 0.0e+00 11  0  0  0  0  24  0  0  0  0     0
> SNESJacobianEval      21 1.0 4.2020e+00 1.0 0.00e+00 0.0 0.0e+00 0.0e+00
> 0.0e+00 14  0  0  0  0  30  0  0  0  0     0
> SNESLineSearch        21 1.0 1.8140e+00 1.0 4.82e+06 1.0 0.0e+00 0.0e+00
> 0.0e+00  6  0  0  0  0  13  0  0  0  0     3
> KSPGMRESOrthog       470 1.0 8.2534e-01 1.0 7.25e+08 1.0 0.0e+00 0.0e+00
> 0.0e+00  3 50  0  0  0   6 59  0  0  0   879
>
> against the dev log
>
> SNESSolve             18 1.0 4.3973e+01 1.0 1.22e+09 1.0 0.0e+00 0.0e+00
> 0.0e+00 40 83  0  0  0  98100  0  0  0    28
> SNESFunctionEval      39 1.0 3.2787e+00 1.0 0.00e+00 0.0 0.0e+00 0.0e+00
> 0.0e+00  3  0  0  0  0   7  0  0  0  0     0
> SNESJacobianEval      21 1.0 3.5318e+01 1.0 0.00e+00 0.0 0.0e+00 0.0e+00
> 0.0e+00 32  0  0  0  0  79  0  0  0  0     0
> SNESLineSearch        21 1.0 1.8007e+00 1.0 4.82e+06 1.0 0.0e+00 0.0e+00
> 0.0e+00  2  0  0  0  0   4  0  0  0  0     3
> KSPGMRESOrthog       470 1.0 8.0665e-01 1.0 7.25e+08 1.0 0.0e+00 0.0e+00
> 0.0e+00  1 50  0  0  0   2 59  0  0  0   899
>
> The huge difference here comes from the Jacobian evaluation, which we do
> not control.
>
>     Matt
>
>
>> 2012/8/22 Aron Ahmadia <aron.ahmadia at kaust.edu.sa>
>>
>>> Can you re-run both codes with: -log_summary please?
>>>
>>> A
>>>
>>> On Wed, Aug 22, 2012 at 10:43 AM, Feng Li <fengshen186 at gmail.com> wrote:
>>> > The performance of petsc_3.3p2 is five times faster than
>>> petsc_dev-r24197.
>>> > Why?
>>> >
>>> > It is tested with pflotran example problem 100x100x100.
>>> > config parameter：--with-mpi=0 --CC=gcc --FC=gfortran --CCFLAGS="-g -O2"
>>> > --FCFLAGS="-g -O2" --blas-lapack-dir=$MKLROOT
>>> >
>>> > detail time:
>>> >                       petsc_3.3p2          petsc_dev24197
>>> > walltime(sec)        28.116                 110.25
>>> >
>>> > Thank you!
>>>
>>> --
>>>
>>> ------------------------------
>>> This message and its contents, including attachments are intended solely
>>> for the original recipient. If you are not the intended recipient or have
>>> received this message in error, please notify me immediately and delete
>>> this message from your computer system. Any unauthorized use or
>>> distribution is prohibited. Please consider the environment before
>>> printing
>>> this email.
>>>
>>
>>
>
>
> --
> 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
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20120822/75abc43a/attachment.html>