Date: Wed, 15 Aug 2012 08:40:21 +0530 From: Dhiru Kholia <dhiru.kholia@...il.com> To: john-dev@...ts.openwall.com Subject: Re: OpenCL implementation of GPG format On Wed, Aug 15, 2012 at 2:07 AM, magnum <john.magnum@...hmail.com> wrote: > On 2012-08-14 18:25, Dhiru Kholia wrote: >> I have written an OpenCL implementation of GPG format. Currently, the >> speedup is very low (9x). The compiler complains about register >> spilling. According to http://dl.acm.org/citation.cfm?id=2039822 >> speedups upto 100x can be achieved. > > Cool. Have you tried it on bull? Maybe its cards have register files > large enough. Feel free to commit it to magnum-jumbo for easy testing. Code has been committed to magnum-jumbo. Benchmarks from bull system, dhiru@...l:~/OpenCL/src$ ../run/john -fo:gpg-opencl -t -pla=1 OpenCL platform 1: AMD Accelerated Parallel Processing, 2 device(s). Using device 0: Tahiti Benchmarking: OpenPGP / GnuPG Secret Key [OpenCL]... DONE Raw: 63705 c/s real, 83781 c/s virtual dhiru@...l:~/OpenCL/src$ ../run/john -fo:gpg-opencl -t OpenCL platform 0: NVIDIA CUDA, 1 device(s). Using device 0: GeForce GTX 570 Benchmarking: OpenPGP / GnuPG Secret Key [OpenCL]... DONE Raw: 58329 c/s real, 58576 c/s virtual No complaints about register spilling on bull. Still the speedups obtained are quite low. Will try to profile and see where the bottlenecks are. -- Cheers, Dhiru
Powered by blists - more mailing lists
Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.