[vtkusers] parallel mesh reading and writing

[Kevin H. Hobbs]

On Tue, 2007-01-23 at 09:19 -0700, Karl Merkley wrote:
> I will admit that I blew right past an XML based format because a of  
> a preconceived notion that XML data would be too big and too slow to  
> read and write large data files (>> 100M elements).    Do you have  
> any data that would correct any misconceptions on my part?
> 
>     Karl

I've primarily used the XML image classes, with a light sprinkling of
polydata so I don't have all the experience you might want to hear
about.

From looking at the files they look like the XML just describes
information about the data blocks such as extent, number of points, and
offset of the data in the file. There are a couple of options on how the
data are encoded. They can be raw or encoded to text, compressed and so
on. I don't think any parsing of XML goes on while reading the data
blocks. Reading the encoded and compressed data is slower than reading
the raw data, but I've never been limited by IO, only by memory and CPU
speeds.

The important part is that they work, which is much faster than things
that should be fast but don't work.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://www.vtk.org/pipermail/vtkusers/attachments/20070123/1571b9d6/attachment.pgp>