[petsc-users] Performance of PETSc TS solver

[Jed Brown]

"Jin, Shuangshuang" <Shuangshuang.Jin at pnnl.gov> writes:

> Hi, Jed,
>
> I followed your suggestion and profiled the IJacobian stage, please see the related profile below:

Cool, all of these are pretty inexpensive, so your time is probably in
computation.  Are all data structures distributed?  Is there any work
that you do redundantly or does each core only compute its local part?

> Summary of Stages:   ----- Time ------  ----- Flops -----  --- Messages ---  -- Message Lengths --  -- Reductions --
>                         Avg     %Total     Avg     %Total   counts   %Total     Avg         %Total   counts   %Total
>  0:      Main Stage: 4.0670e+01  11.5%  1.2019e+11 100.0%  1.082e+07 100.0%  5.757e+02      100.0%  6.386e+04  81.6%
>  1: My IJacobian stage: 3.1379e+02  88.5%  0.0000e+00   0.0%  1.984e+03   0.0%  1.357e-02        0.0%  1.438e+04  18.4%
>
> Event                Count      Time (sec)     Flops                             --- Global ---  --- Stage ---   Total
>                    Max Ratio  Max     Ratio   Max  Ratio  Mess   Avg len Reduct  %T %f %M %L %R  %T %f %M %L %R Mflop/s
> ------------------------------------------------------------------------------------------------------------------------
> --- Event Stage 1: My IJacobian stage
>
> VecSet              1797 1.0 4.7467e-02 1.9 0.00e+00 0.0 0.0e+00 0.0e+00 0.0e+00  0  0  0  0  0   0  0  0  0  0     0
> VecScatterBegin     1796 1.0 4.3967e-01 1.1 0.00e+00 0.0 0.0e+00 0.0e+00 1.8e+03  0  0  0  0  2   0  0  0  0 12     0
> MatAssemblyBegin    1796 1.0 7.2787e+00 1.8 0.00e+00 0.0 0.0e+00 0.0e+00 3.6e+03  2  0  0  0  5   2  0  0  0 25     0
> MatAssemblyEnd      1796 1.0 2.1605e-01 1.1 0.00e+00 0.0 2.0e+03 7.4e+01 1.8e+03  0  0  0  0  2   0  0100100 13     0
>
> Object Type          Creations   Destructions     Memory  Descendants' Mem.
> --- Event Stage 1: My IJacobian stage
>
>               Vector  1798              1         1552     0
>       Vector Scatter  1797           1796      1156624     0
>            Index Set  1798           1798      1370952     0
>
> It seems that IJacobian occupies 88.5% of the total computation time. Anything else can you interpret from the profile which can help me to accelerate the IJacobian computation?
>
> Thanks,
> Shuangshuang
>
>
>
> -----Original Message-----
> From: Jed Brown [mailto:five9a2 at gmail.com] On Behalf Of Jed Brown
> Sent: Tuesday, August 13, 2013 6:50 PM
> To: Jin, Shuangshuang; Shri
> Cc: Barry Smith; petsc-users at mcs.anl.gov
> Subject: RE: [petsc-users] Performance of PETSc TS solver
>
> "Jin, Shuangshuang" <Shuangshuang.Jin at pnnl.gov> writes:
>
>> Hi, Shri,
>>
>> From the log_summary, we can see that the TSJacobianEval/SNESJacobianEval dominates the computation time as you mentioned.
>>
>> Event                Count      Time (sec)     Fl
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20130815/0f56d6c9/attachment.pgp>