[daip] VBGLU bug

[Cormac Reynolds]

hi,

that seems to work perfectly - thanks very much. As for the table writing, the
version that went slowly had a very large CL table (CLINT= 2 seconds), while
the one you tested with had CLINT=60 seconds.

The CLINT=2 sec version takes about 3.3 hours  to VBGLU in total. The same
dataset with CLINT=10 sec seconds VBGLUed in 17 minutes in total. With CLINT=60
secs the whole thing is done in just 10 minutes, and the table writing part of
that must have been less than two minutes.

So it seems there is some approx. quadratic dependence on the size of the CL
table. The simple solution is to avoid writing such stupidly large CL tables as
I had in the first place.

regards,
Cormac.

-----------------------------------------------------------------
Cormac Reynolds                      |   Phone: +31 (0)521 596507 (New!)
Science Operations & Support Group,  |   Fax: +31 (0)521 596539
Joint Institute for VLBI in Europe,  |   email: reynolds at jive.nl
Postbus 2,                           |
7990 AA Dwingeloo,                   |
The Netherlands.                     |
-----------------------------------------------------------------

On Tue, 31 Jan 2006, Eric Greisen wrote:

> Okay I think I got it.  Attached is a corrected version.  Your 2nd
> data set had significant sections of data where the first data set had
> none.  This triggered an error in the code.
>
> I noticed that the table handling went very fast - did you delete
> tables or has the problem mysteriously vanished.  Amy was starting to
> look at the issue and this absence of times in data set 1 but found in
> data set 2 occurred to her as a source of slowness.
>
> I will send the code in the next message.
>
> Eric Greisen
>