Hi! On Fri, Oct 23, 2009 at 10:21:07AM +0100, Christian Nentwich wrote: > ah - I missed the white paper, I will read that later so I can form > a real opinion. I must say, the mere fact that there will be C++ > support and a proper development environment (even if only for > Windows) is a big relief. Working with CUDA at the moment is a > nightmare, straight back to the 80s.
I don't think it's so bad. :) The only annoying thing is that the applications are extremely hard to debug (especially concurrency issues which don't show up in the CPU emulation mode). I'm not sure if Fermi will improve that significantly (at least some way to log debug prints in a ring-buffer or such), but other than that I didn't find the CUDA development particularly difficult. Another great thing would be fine-grained profiling support... > Darren Cook wrote: > >>these articles are still somewhat short on detail, so it's hard to tell. > > > >Yes the linux mag article was a bit empty wasn't it, but did you take a > >look at the 20-page whitepaper: > >http://www.nvidia.com/content/PDF/fermi_white_papers/NVIDIA_Fermi_Compute_Architecture_Whitepaper.pdf I haven't got time to read it yet either, but the thing that took my interest was the multiple kernel support, I think that could achieve much better GPU utilization, but it depends on implementation details. I would love if there would be native support for extremely fast parallel summing of array contents; for my intersection model, that would help immensely, I guess. -- Petr "Pasky" Baudis A lot of people have my books on their bookshelves. That's the problem, they need to read them. -- Don Knuth _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/