Error cuCtxPopCurrent (cudaHashcat+64.exe)

[jAm]

Hi,
I don't know if it's a bug..
When I run this command line :

Code:

cudaHashcat+64.exe -r rules\d3ad0ne_1.5K.rule -n 512  --gpu-loops 512  -m 1000  hash.txt "dics\full.txt"
cudaHashcat+ v0.03 starting...

ERROR: cuCtxPopCurrent() 999

And program exits.

Why ?
Then I put -n 256 instead of 512 and it works fine.

When I use a smaller rule (like leetspeak.rule) I am able to run the program with -n 512, without any error.
Does the rule file length acts on the -n parameter ?

Thank you.

[Rolf]

What's your GPU ?
I cant test cudahc+ due to driver incompability, but I have an idea that your gpu is just overloaded, that's why -n 512 doesnt work.

[atom]

play with it but dont oversize it

[jAm]

My GPU is GeForce 8400 GS, 239MB, 1400Mhz, 1MCU.
Not so good but works
I will stay on "-n 256 --gpu-loops 512" it's fine for me.

[mastercracker]

I was able to reproduce the same problem by also putting extreme values:

-n 216 --gpu-loops 512

got the error cuCtxPopCurrent() 999

My cards in this case were a GTX260 and a GTS250.

Now another side effect and/or bug. If I put only --gpu-loops 1024. It works fine but end up missing passwords. Definitely don't want that. Maybe it already affects people without knowing...

[atom]

hm, because of the status display it may look like it does not find all passes, but it does. look here:

Quote:root@erebus:~/oclHashcat+-0.02# /root/maskprocessor-0.63/mp64.bin ?d?d?d?d?d?d | ./cudaHashcat+64.bin x -o y
cudaHashcat+ v0.02 starting...

Hashes: 1000000
Bitmaps: 21 bits, 1048576 entries, 0x000fffff mask, 4194304 bytes
Rules: 1
Platform: NVidia compatible platform found
Device #1: GeForce GTX 580, 1535MB, 1700Mhz, 16MCU
Device #2: GeForce GTX 580, 1535MB, 1700Mhz, 16MCU
NOTE: gpu-accel auto-adjusted to: 1
Device #1: Kernel ./kernels/4318/m0000.sm_20.64.cubin
Device #2: Kernel ./kernels/4318/m0000.sm_20.64.cubin

Starting attack in stdin_mode...

p: 131072/1000000, cs: 1, cr: 0, cl: 6, rt: 10172.43ms, s: 12.89k/s
p: 262144/1000000, cs: 1, cr: 0, cl: 6, rt: 9839.46ms, s: 13.32k/s
p: 393216/1000000, cs: 1, cr: 0, cl: 6, rt: 9805.62ms, s: 13.37k/s
p: 524288/1000000, cs: 1, cr: 0, cl: 6, rt: 9909.23ms, s: 13.23k/s
p: 655360/1000000, cs: 1, cr: 0, cl: 6, rt: 9707.79ms, s: 13.50k/s
p: 786432/1000000, cs: 1, cr: 0, cl: 6, rt: 9855.99ms, s: 13.30k/s
p: 917504/1000000, cs: 1, cr: 0, cl: 6, rt: 10075.97ms, s: 13.01k/s


Started: Wed Feb 2 17:04:38 2011
Stopped: Wed Feb 2 17:05:55 2011

but its just a display problem. if you check the output file they are all found. its just missing a final display.

Quote:root@erebus:~/oclHashcat+-0.02# wc -l y
1000000 y

[jAm]

I put a lower number for -n and --gpu-loops parameter, and it works better.

[atom]

maybe i should limit this to like 800 or so in next release

[tatgdi]

Hello Atom,

Please do look at providing some form of automation for the following -n value and --gpuloops-value options discussed above. I find as well depending on the digest length and type it can increase/decrease your attack. When in doubt I have left off the folloiwing 2 parameters and the speed has been descent.

I posted additonal information at the following thread:
http://hashcat.net/forum/thread-293.html

[atom]

no, they are there for a reason. for example if i want to play games or watch movies on the same card while i am cracking i need to lower these number to make room for additional calculations by 3rd applications.