Here is a paper describing the CUDa implementation of tridiagonal matrix solver:

http://www.jcohen.name/papers/Zhang_Fast_2009.pdf

Thanks for reply.

Yes i have read this paper, but still am unable to figure out how to implement the algorithm, perhaps due to my mechanical background. Can you please explein a little how can i apply recursion to get the algo?

NVidia's cusparse library

https://developer.nvidia.com/cusparse

appears to provide a tridiagonal solver

Indeed, the cusparse library from nvidia already have a tridiagonal solver, so you dont't have to implement it from scratch, just use the one provided in cusparse.

Can you provide me some tutorial on using cuSparse, i have been reading its documentation but am still confused about the implementation, how to get the output, etc...

Finally i have written a sample code using cusparse, thanks Markus and Bogdan for guiding me about cuda libraries i was ignorant about.

I hope my this code shall be sufficient for anyone who hits this question for the same problem i had.

http://musha-shugyo.googlecode.com/files/code.cu

this is my code in cuda-c...

if any doubts, please reply below

(WARNING:i am not a good coder)

Programming Massively Parallel Processors, 2nd Edition: A Hands-On Approach by David B. Kirk and Wen-Mei W. Hwu is now available! Review the full Table of Contents below.

This new edition of the bestselling Programming Massively Parallel Processors: A Hands-On Approach has been thoroughly updated with new and updated content on parallel computing techniques and technologies,

SOME LINKS

http://docs.nvidia.com/cuda/index.html

http://people.maths.ox.ac.uk/gilesm/cuda/

http://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html

http://www.shodor.org/media/content//petascale/materials/UPModules/dynamicProgrammingPartI/

http://www.drdobbs.com/parallel/cuda-supercomputing-for-the-masses-part/207200659

http://heim.ifi.uio.no/~knutm/geilo2008/