0

I am new to CUDA programming. I have to implement recursive algorithm which includes varying number of matrix inversions on GPU . Currently I am using getrfBatched and getriBatched methods in CUBLAS library to get a inversion of an one matrix. I think I can speed up the whole algorithm if I can implement all those matrix inversions parallel . Can anyone suggest me a solution for this??

Thanks

sa1991
  • 47
  • 1
  • 10
  • 1
    If you go to the CUDA registered developer website, you can download source code for batched solver and batched matrix inverse of small matrices, that may help you get started. Look for "Batched Solver" in the download section. – njuffa Mar 23 '15 at 17:54
  • Thank you very much for your responses . I will try that – sa1991 Mar 23 '15 at 19:04

0 Answers0