gauss elimination problem

[dineshparallel]

how to implement the guass elimination(direct method i.e., converting into upper tirangular). if possible provide the code to me. When i try to implement in parallel after successful compilation and entering matrix i am getting the results incorrect compared to serial.The correct results are getting only when the input matrix size is equal to the grain size. The code is implemented by using parallel_for and blocked_range2d.

What is the relation between input matrix size and grain size?

if input matrix size and grain size are equal it consuming more time than serial time.

Please suggest solution.If u r intrested i send the code to u.

Thank you,

[RafSchietekat]

Are you trying to run the outer loop in parallel, perhaps? I don't see a problem with a serialupper-row iteration that nests a parallel lower-row elimination, in terms of correctness anyway. I suppose that somebody may have come up with something more efficient than that, taking considerations like cacheusage into account (TBB normally favours parallelism in the outer loop for a reason), but this seems quite straightforward as an initial approach.