Date: Thu, 21 Mar 2013 00:12:53 +0530 From: Sayantan Datta <std2048@...il.com> To: john-dev@...ts.openwall.com Subject: Re: clLinkProgram() and clCompileProgram() Hi Brian, On Wed, Mar 20, 2013 at 11:41 PM, Brian Wallace <bwall@...nbwall.com> wrote: > I was thinking about doing this with Bitcoin mining(similar concept), > where instead of passing in data, compile the kernels at run time with the > static data for each block to be solved hard coded and possibly optimized. > Could be useful for long hashes with salts that we use once for many hashes. > > https://www.openbwall.com/entry/Optimizing-Bitcoin-Miner-Idea > ------ > bwall@...nbwall.com - NOT to be confused with openwall.com > > On 03/20/2013 01:54 PM, Sayantan Datta wrote: > > Hi all, > > I don't know if you have seen them before but I found two cl* functions > that would allow us to compile and link the opencl binaries at runtime. > Hopefully the linking process will be much faster than rebuilding the > executables from source. > > > http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clLinkProgram.html > > http://www.khronos.org/registry/cl/sdk/1.2/docs/man/xhtml/clCompileProgram.html > > Regards, > Sayantan > > > Yes exactly, I was trying to hardcode the salts for DES. At the moment it takes nearly 20 minutes to build all(4096) the hardcoded kernels on OCed 2500k which is really nasty. Maybe now I can compile the main program once and then compile the hardcoded salts and link them at runtime as required. Regards, Sayantan Content of type "text/html" skipped
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.