Date: Sun, 29 Sep 2013 00:48:26 +0100 From: Dániel Bali <balijanosdaniel@...il.com> To: john-dev@...ts.openwall.com Subject: Re: Daniel's weekly report #15 Hello! By now the patched ELF doesn't freeze the GPU, but it runs with errors. I get the following error: daniel@...l:~/JohnTheRipper/run$ ./john -test --format=crc32-opencl -devices=1 Device 1: Tahiti (AMD Radeon HD 7900 Series) Local worksize (LWS) 128, global worksize (GWS) 16777216 Benchmarking: crc32-opencl, CRC32 [OpenCL (inefficient, development use only)]... Warning: binary() returned misaligned pointer FAILED (cmp_all(1)) Here is the two hex values that are compared in cmp_all: 7ada0bcc, 6e65706f My previous error was forgetting to update numVGPR and numSGPR in the binary. I'm not sure what causes the current error. I used the KernelAnalyzer to produce the assembly from the correctly working OpenCL kernel. I'll be in touch. Regards, Daniel 2013/9/27 Dániel Bali <balijanosdaniel@...il.com> > Hello! > > > 2013/9/26 Solar Designer <solar@...nwall.com> >> >> As far as I can tell, "bull" was automatically rebooted by one of our >> watchdogs. After the reboot, attempts to use the 7970 card work fine. > > > Sorry about that then, I must have missed the reboot. > > I don't have many ideas about what to try next. One idea is to test if the > patching works without JtR. This would mean setting up a minimal host-side > environment that passes the correct buffers and compares the output to the > required results. My other idea would be to change the tests in the format > to check for a single value (0xaaaaaaaa), then make a kernel that does > something different and patch it with this simple assembly code. > > I'll try these, but if anybody has a better idea please let me know. > > Regards, > Daniel > 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.