On Wed, 1 Nov 2006, Jonathan Myers wrote:
Based on a lunch conversation with Jeff, I've added some to the discussion section of the DC1 writeup. Please let me know what you think.
Thanks! I have some text discussing scalability; however, there are some points you provide that I do not cover. I will integrate.
You describe the initial design specifying N images in the pipeline at one time (I simplify); however, with respect to the UML, I'm not sure that was intended. If I substitute "design" with "implementation", then I think we're good. I contend that we should design our distributed computing framework so as to achieve a load average of (slightly more than) 1 per core. To the extent this is still a topic of discussion, I'll not dwell on this distinction too much.
On another topic...here's a question. The flops reported in section 4.2 (Pipeline Performance), .649 GFlops--I take it this is per processor, yes?
thanks! Ray _______________________________________________ LSST-data mailing list [email protected] http://www.lsstmail.org/mailman/listinfo/lsst-data
