Date: Wed, 31 Oct 2012 18:54:43 +0100 From: magnum <john.magnum@...hmail.com> To: john-dev@...ts.openwall.com Subject: Re: Serious bug in md5crypt-cuda - or in driver? On 30 Oct, 2012, at 22:08 , Lukas Odzioba <lukas.odzioba@...il.com> wrote: > 2012/10/30 magnum <john.magnum@...hmail.com>: >> Is there any more error checking that can be made in CUDA than already is the case? > > Hmm, please try this: > > HANDLE_ERROR(cudaGetLastError()); > > Just after the: > > kernel_crypt_r <<< BLOCKS, THREADS >>> (cuda_inbuffer, cuda_outbuffer); Problem solved. The added error handling gave this: "too many resources requested for launch in cryptmd5.cu at line 241" I decided to add a similar check to all CUDA kernels and this quickly solved the three other problems I had too (mscash, wpapsk and pwsafe). All four problematic formats just needed a lower THREADS figure in their respective header file. I did not commit these lower figures though we might consider doing that. I did commit the added error handling to all formats. I also added some info in README-CUDA and a specific hint from HANDLE_ERROR() when this particular error comes up. Thanks, magnum
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.