Quote Originally Posted by Final8ty View Post
Double precision point of view is http://www.semiaccurate.com/forums/s...&postcount=769
that's almost trivial compared to other issues. you have to learn to crawl before you can walk. for instance gpugrid, a DC project focused on curing diseases, attempted to make an ATi client based on opencl. we waited 2 months. the 5870 ended up performing as fast as an 8800gtx.

almost 2 years before this nvidia gave away free cards to gpugrid and helped them develop their acemd molecular dynamics software on cuda. gpugrid is currently the fastest bio-med project on boinc.

f@h is in the same situation. they want to make one good opencl client that can run on all hardware but they have to wait on stable compilers which takes time. i know this all sounds like a huge tangent but other developers eager to use gpu's are probably running into the same problems.