Date: Sun, 24 May 2015 23:20:22 +0300 From: Solar Designer <solar@...nwall.com> To: john-dev@...ts.openwall.com Subject: Re: PHC: Parallel in OpenCL On Sun, May 24, 2015 at 10:10:44PM +0200, Agnieszka Bielec wrote: > 2015-05-24 21:55 GMT+02:00 Solar Designer <solar@...nwall.com>: > > On Sun, May 24, 2015 at 08:20:16PM +0200, Agnieszka Bielec wrote: > >> I increased the speed slightly: > > > > What changes did you make that resulted in this speed increase? > > As you said, I copied sha512block from cryptsha512_kernel_GCN.cl for > gcn and cryptsha512_kernel_GPU.cl for others > and I included > > #if !gpu_nvidia(DEVICE_INFO) || nvidia_sm_5x(DEVICE_INFO) > #define USE_BITSELECT > #elif gpu_nvidia(DEVICE_INFO) > #define OLD_NVIDIA > #endif > > I don't use USE_BITSELECT or OLD_NVIDIA but it's faster on my laptop FWIW, USE_BITSELECT is meant to help on GCN (so on --dev=1), since GCN has such instruction (and our NVIDIA doesn't; they added it in Maxwell, which is newer than what we have). But this isn't always the case in practice due to source code and compiler shortcomings. In general, if USE_BITSELECT isn't helping on GCN, this suggests there's still room for optimization... and then USE_BITSELECT will likely help. > and I resigned from vectors, was slower with them Alexander
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.