5 ms·
CUDA is probably the way to go, since especially if you're having to use a gpu, might as well get one of the new nvidia gpu's
by pen2l 11y ago
CUDA is probably the way to go, since especially if you're having to use a gpu, might as well get one of the new nvidia gpu's
- Gladdyu 11y agoCUDA has a nice toolchain, but the point of this is to remove all of the low-level stuff. For performance (even on NVIDIA cards) it doesn't really matter whether you use CUDA or OpenCL [1] + OpenCL runs everywhere as a bonus. [1] http://pds.ewi.tudelft.nl/pubs/papers/icpp2011a.pdf http://pds.ewi.tudelft.nl/pubs/papers/icpp2011a.pdf
- dr_zoidberg 11y agoAdded bonus: OpenCL runs on multicore processors, so you don't even need a GPU (though performance is obviosuly lower).
- lfowles 11y agoNot a whole lot lower though, only in the 10-100x lower range from what I've experimented with (couple generations old card vs couple generations old i7 with a straight up "process some floats" test[0]). Certainly enough to test your kernels against. [0]: https://github.com/krrishnarraj/clpeak https://github.com/krrishnarraj/clpeak
- dr_zoidberg 11y agoGood work, specially on keeping the results from various platforms!
- lfowles 11y agoNot mine! Just the tool I used
- TsiCClawOfLight 11y agoNo thanks, I'll rather use something that works on more than Nvidia. Better or not, that's too proprietary for me.