next »

[Bad Sector]

In a very rare situation where i decided to write a new blog post, i wrote one titled Playing with CUDA, where i describe my experience with CUDA so far. So far means since yesterday when i installed it in my computer, ate... erm, read the programming guide and wrote this nifty raytracer.

If you're about to play with CUDA too, you can read it as some thoughts (and some parts i got wrong!) about the system. CUDA is a very nice architecture which i'm going to try to learn as good as i can since it provides many opportunities for writing faster tools and probably getting some idea on what could be done in consumer graphics hardware in the years to come.

[Brackhar]

CUDA is great.  I actually worked on the driver team at NVIDIA this past summer.  It's very approachable for doing parallel work, and I'd encourage people to check it out.  The only real tricky part about it is getting the coalesced memory access model down, but that can be overcome with proper workload partitioning.