All,

I've committed a version of stageleader.c in r.556 which won't start processing an image until the previous image is completed. Testing it with 6 CCDs/node (our original values) gets us an image in and out in about 145 seconds. Whichever version of stageleader is used, reducing CCDs/node yields an equal change in latency time; I got a run through with 3 CCDs/node done in about 75 seconds for an image to get in and out.

I'm about to see what I can do to increase the number of cores used.

Cheers,
-Jon


_______________________________________________
LSST-data mailing list
[email protected]
http://www.lsstmail.org/mailman/listinfo/lsst-data

Reply via email to