Chris,
I have been playing around with a vector and scalar data set similar to
what you describe and can not duplicate your problem. The problem David
saw with Regrid dealt with invalid positions.
Do you have an Invalid Positions component?
Do you use Construct to create the grid for Regrid?
Have you had a chance to try a current version of OpenDX?
I also saw a similar performance problem that Randall noted with AutoGrid,
where Regrid performance becomes very slow when nearest is set to something
other than "infinity" is used. Hopefully, someday I can look into that one.
Jeff
At 01:54 PM 6/13/2002 -0400, you wrote:
I was scanning some real old posts about Regrid and the new improved
version that VIS Inc. did but the posts don't seem to mention a bit of
unsavoriness I just discovered. I'm currently using an older SGI version,
so this is the standard Regrid module.
I have a field with a scalar component ("dot") and a vector component
("cross"). I need to regrid this field onto a particular grid. The two
components are bound to the same positions, then I send that field to
Regrid along with the desired output grid. Works fine, but...
If the input data field has "dot" as "data" and "cross" as "cross", the
output has "data" and "cross", regridded, as advertised in the Regrid
docs. But something seemed not quite right about the result, so I tried
reversing the input so that "cross" was "data" and "dot" was "dot". Sure
enough, the values are different in both arrays! Aack!
If I send a field with only "dot" as "data" through Regrid, I get the same
output values as if I send "dot" as "data" along with "cross" as "cross".
That is, the "dot" regridded values agree, since in both cases they were
"data" in and out.
Likewise, sending only "cross" as "data" gives the same output values
(3-vector) as when I send "cross" as "data" and "dot" as "dot", but note
that these "dot" outputs are not the same as those of the previous
paragraph, and these "data" ("cross") values are not the same as those of
the previous paragraph. They are similar in magnitude, but different
numbers. In particular, the vector field looks quite different if glyphed.
So it appears that Regrid is doing something weird internally and the
short-term workaround for me will be to run 2 parallel Regrids and unite
the data components post-regridding.
David, Jeff, anyone? Did you spot this bug when recoding AutoRegrid?
--
Chris Pelkie
Managing Partner
Practical Video LLC
30 West Meadow Drive, Ithaca, NY 14850
