(07-10-2013, 12:55 PM)epixoip Wrote: For everyone reporting bugs about VirtualCL 1.21: I have not been able to reproduce your issues. But I have found other issues.
First, VCL 1.21 will work with both plus 0.14 and 0.15, using the installation steps detailed in the wiki. It does not work well with 0.14, but it will at least appear that it is working.
With 0.14, I noticed that with -a 0 it only finds about 1/6th of the plains. Or rather, it seems to think it's finished when it's only 1/6th of the way through. I don't know why. At first I thought it was a driver issue since I'm using 13.6, but it finds all of the passwords locally, so that's clearly not the problem. But, it finds all of the plains in 0.15, so I'm not concerned about it. Just realize that you will probably want to use 0.15.
I also identified a minor bug in VCL 1.21. Well, it's a major bug, but it's easily worked around. It's just a rather annoying workaround.
The bug I have identified is that after running a program with VCL, the broker will hang on to the devices and clCreateContext() will fail with CL_DEVICE_NOT_AVAILABLE. Only workaround is to restart the broker daemon after every pass.
I have engaged MOSIX about this bug, but for the time being, you can automate this workaround with a simple shell function:
Code:vclHashcat-plus()
{
/etc/init.d/vcl restart
sleep 5
./vclHashcat-plus64.bin $*
}
As of 0.15b48, vclHashcat-plus64.bin is no longer needed. so if you're using 0.15, replace "vclHashcat-plus64.bin" with "oclHashcat-plus64.bin" in the function.
Just drop that function in your ~/.bashrc and re-source it, then simply invoke the "vclHashcat-plus" function instead of vclHashcat-plus64.bin and all will be well.
If you continue to experience errors:
clCreateContext() -2: manually run "/etc/init.d/vcl restart", wait about 5 seconds, and try again. If you continue to receive this error, restart vcl on the backend as well.
clEnqueueNDRangeKernel() -48: you need to remove your old kernels. Simply run "rm kernels/4098/*.kernel" and try again.
clEnqueueReadBuffer -38: make sure you are using vclHashcat if you're using 0.14. if you receive this error and are using 0.15, let me or atom know.
Good luck!
Hello
Any update on the following bug report to MOSIX? Will ADL every be supported through VCL? Also was the following issue impacting oclHashcat-lit V0.15 as well? AMD has 13.10 beta drivers however I believe you mentioned that it was none-driver specific; not finding the plains is also very peculiar hopefully MOSIX has responded with something positive.