OpenCL GPU P-1 Factoring and ECM Factoring

[xx005fs]

Is there any chance that the CUDA port of P-1 Factoring and ECM would be rewritten into OpenCL versions just yet? Or is there a beta version of the software that is already developed but not fully ready yet? Considering that traditionally AMD GPUs are good at general purpose compute and has higher memory bandwidth than Nvidia counterparts wouldn't it be on par or even faster than Nvidia GPUs doing the same task?

[kriesel]

Originally Posted by xx005fs

Is there any chance that the CUDA port of P-1 Factoring and ECM would be rewritten into OpenCL versions just yet? Or is there a beta version of the software that is already developed but not fully ready yet? Considering that traditionally AMD GPUs are good at general purpose compute and has higher memory bandwidth than Nvidia counterparts wouldn't it be on par or even faster than Nvidia GPUs doing the same task?

Speaking only for myself, the CUDA P-1 program needs a lot of work first before porting. (I've spent gpu-years probing it for bugs and capability limits, with further to go.)
ECM code for gpus is limited to bit lengths much too short to be of use to GIMPS, and ECM is not effective for exponents of relevance to the current GIMPS wavefronts.

See http://www.mersenneforum.org/showthread.php?t=23371 and http://www.mersenneforum.org/forumdisplay.php?f=154
I know of no OpenCL P-1 factoring implementation.
I know of no OpenCL ECM implementation.
I know of no CUDA implementation of the PRP test that will compile for and run on Windows.
If you do, please PM or post where to find any such.

[xx005fs]

Got it.

[kriesel]

See also Preda's recent addition of P-1 to gpuOwL, with stage 1 up front and stage 2 along the way during the PRP computation.