Date: Mon, 15 Sep 2014 10:51:23 -0400 From: David <john-users@....eml.cc> To: john-users@...ts.openwall.com Subject: Re: Cuda 6.5/OSX 10.9.4 Magnum, you rock, I’m up and running. I tried sm_20 first (which compiled fine), then ran ./john --list=cuda-devices and its output indicated "Compute capability: sm_30” so I recompiled after switching to sm_30 — didn’t seem to make a performance difference. I’ve added this step to my personal wiki… any chance we can add it to the INSTALL doc or official wiki? Thanks again, David On Sep 11, 2014, at 17:20 , magnum <john.magnum@...hmail.com> wrote: > On 2014-09-11 21:47, David wrote: >> When I run make I get: >> >> nvcc fatal : Value 'sm_10' is not defined for option 'gpu-architecture' >> make: *** [cuda_common.o] Error 1 >> make: *** Waiting for unfinished jobs.... >> make: *** [default] Error 2 >> >> Note that I’m running OS X 10.9.4 and just downloaded and installed >> cuda_6.5.14_mac_64.pkg. > > This was discussed somewhere recently although I haven't seen it myself yet. We should add that to autoconf. > > You need to edit line 53 (after ./configure is run) of Makefile. Not sure what hardware you've got but sm_30 should do. Maybe sm_20 still works too. Running "nvcc -h" will probably get you a list of valid architectures. > > Hmm, or maybe just drop the "arch sm_10" statement instead of changing it. Maybe that's what we should default to also. I'll try that right away. > > 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.