[Paraview] AMR reader

[Jean M. Favre]

Randy Hudson wrote:
>  I measured the time spent in the five code blocks indicated, and by far the most time is spent in BLOCK FIVE, ds->GenerateVisibilityArrays().
> 

I have measured that too on a single dataset example. It is a 250 Mb
dataset, 2D, 5 levels, 171 grids, 5 scalars and 1 vector field. It is
stored in a single HFD5 file.

I stuff all levels with node-based data.

Reading 1, 2, 3, 4, 5, then 6 variables gives me a data stuffing time
linearly increasing from 0.35 seconds to 1.2 seconds for all 6 variables

Generating visibility arrays is of constant time (independent of the
number of variables read) and is 4.2 seconds.

so I can only confirm what you see. Don't know that we can speed that up
though.

Jean

-- 
Dr. Jean M. Favre,                            email:favre @ cscs.ch
http://www.cscs.ch/about/Favre.php
CSCS, Swiss National Supercomputing Centre  | Tel:  +41 (91) 610.82.40
Via Cantonale, 6928 Manno, Switzerland      | Fax:  +41 (91) 610.82.82