[mira_talk] Re: MIRA GPU

Having read part of your thesis I do think we can gain profit here from
using the GPU. I will need to have a closer look on the algorithm,
current SW GPU/SIMD implementations and the CUDA architecture to see how
your algorithm can be ported to the GPU.


Is that architecture valid for both NVidia and ATI cards?
The CUDA architecture is only valid on NVidia cards. At the moment this is the method of choice if you want to go for GPGPU. OpenCL would be better (or plain old x86 coding on the next generation GPUs), but is still in development phase.
On another note, my supervisor and I have decided not to continue to combine MIRA with an existing Smith-Waterman GPU implementation. This is due to the fact that the current implementation do not seem to support backtracking or store the calculated matrix in memory. Therefore an actual alignment can not be shown/made. It would be too much work to rebuild these implementations to make it suitable for MIRA in the time I have available. Due to these limitations we have decided to make our own Smith-Waterman implementation on the GPU which might support backtracking but is not the aim of the implementation. We're trying a new approach to implementing SW on the GPU which hopefuly makes it faster than previous implementations. If backtracking will be supported I will of course keep you informed in case you are interested in combining the GPU with MIRA. 

Cheers,
Frank



--
You have received this mail because you are subscribed to the mira_talk mailing 
list. For information on how to subscribe or unsubscribe, please visit 
http://www.chevreux.org/mira_mailinglists.html

Other related posts: