[Yt-dev] Projection speed improvement patch
John Wise
jwise at astro.princeton.edu
Sun Nov 8 11:59:04 PST 2009
On 6 Nov 2009, at 10:35, Matthew Turk wrote:
> Hi john,
>
>> Thanks so much for taking a look at my data and examining the
>> memory usage
>> of analyzing a dataset of this size. I'll have to give it another
>> shot on
>> ranger. I can also see how I/O performance is on the Altix here at
>> Princeton, which has a local RAID (just like red).
>
> Awesome. This is all with the hierarchy-opt branch in mercurial, but
> I think I am going to port it back to trunk in the very near future,
> now that I have tested it on a number of different datasets.
The parallel projections work well on ranger now. I ran on 32 cores
just like you did on triton. I projected density, temperature, and
electron fraction, all weighted by density for the same dataset I gave
you.
yt was using slightly more memory on ranger at 2.1GB/core, which isn't
bad at all. This pushed me over the 2GB/core limit on ranger, so I
had to use 8 cores/node instead of 16.
However, it was slower by a factor of 2.5. It took 1084 seconds from
start to finish (including all of the overhead). I had already
created a binary hierarchy beforehand. Ranger in general is slow (I
suspect its interconnect), so maybe it's just a "feature" of ranger.
Somewhat related but -- The Alltoallv call was failing when I compiled
mpi4py with openmpi, but this went away when I compiled it with
mvapich. If you want to see where it failed, I put the trackback at
http://paste.enzotools.org/show/253/
Once I get mpi4py working on the Altix here, I'll post some timings
from that, as well.
Cheers,
John
More information about the yt-dev
mailing list