Hi every one! I am a newbie with cuda, just tried some simple code, but I am quite confident with C programming. I am thinking on doing an academic project which consists of beating my prof. (in terms of time) in calculating several digits (~500 mliions) of the Euler’s number! He implemented a representation of big numbers including basic operations, and used a FFT algorithm. Every thing on the CPU.
From what i know cuda is suitable for this kind of problems that involve lots of calculation with huge numbers and Fourier transformations. However I am not sure if this is all a loss of time, so I would like to make some question!!
- I’ve seen there are libraries included in the sdk witch performs FFTs and linear algebra operations (i.e cufft, cublas, cula…) but my question is, what if i want to use them with big numbers?.. If I implement a big numbers representation I won’t be able to use these libraries will I? Is there an easier way to deal with large numbers on a cuda platform?.. I dont thnk I would be able to write a FFT from zero! :)
-Do you think that this kind of project makes any sense? Or my idea is just wrong?
I will appreciate your suggestions! Thank you in advance guys!
This is the hardware I am using:
Intel i7 920 2.66 Ghz
3x2GB DDR3 RAM (the prof used 12GB!)
NVIDIA 250 GTS 1GB (I know… no double precision!)