<html><head><style type="text/css"><!-- DIV {margin:0px;} --></style></head><body><div style="font-family:times new roman,new york,times,serif;font-size:12pt"><br><div style="font-family: times new roman,new york,times,serif; font-size: 12pt;"><div style="font-family: arial,helvetica,sans-serif; font-size: 10pt;">>> what I understand GPUs are useful only with certain classes of numerical<br>>> problems and discretization schemes, and of course the code must be<br><br>> I think it's fair to say that GPUs are good for graphics-like loads,<br>> or more generally: fairly small data, accessed data-parallel or with <br>> very regular and limited sharing, with high work-per-data.<br><br>From my limited experience I would agree. Getting to the high work-per-data is absolutely key to getting the huge speedups.<br><br>>> I'm part of a group that is purchasing our first beowulf cluster for a<br>>> climate model and an estuary model
using Chombo<br>>> (<a href="http://seesar.lbl.gov/ANAG/chombo/" target="_blank">http://seesar.lbl.gov/ANAG/chombo/</a>). Getting up to speed (ha) on<br><br>> offhand, I'd guess that adaptive grids will be substantially harder <br>> to run efficiently on a GPU than a uniform grid.<br><br>One key thing is that unstructured grid codes don't work as well. The problem is the indirect addressing.<br><br>I know two of the developers at Nvidia and both are CFD gurus - I will ping them to get more details because I know they were looking at this (unstructured vs. structured).<br><br>Jeff<br><br>P.S. I had to do the indentation by hand on this stupid email web-based email tool :)<br><br><br></div></div></div></body></html>