Todo list (improvement suggestions)

[Bitweasil]

If you run into improvements, or things you'd like to see, post them here.

- Detect CUDA out of memory errors on launch and display the amount of free memory & total memory.
- Low end cards (8500, 8600) crash the box on launch with a bluescreen. Reduce thread count to reduce workload on cards with fewer stream processors?

[Bitweasil]

Still getting some "Invalid device symbol" errors on Windows.

[the_drag0n]

making hash algo option (-h) case independant:
you need to type NTLM or MD5 or MD4 lowercase doesnt work.

[synthesis]

I think resume support would be a great feature

[the_drag0n]

if you select hash type MD5 it still says

Compute done: Reference time 36.8 seconds
Stepping rate: 2.7M MD4/s
Search rate: 10882.6M NTLM/s

output does not work for me :

-o ./found.txt

ERROR: Cannot open output file =./found.txt

even if i create found.txt manually

--output does work!

oh yeah and could one manually remove the 4000 hash limit ?

[Bitweasil]

if you select hash type MD5 it still says

Compute done: Reference time 36.8 seconds
Stepping rate: 2.7M MD4/s
Search rate: 10882.6M NTLM/s

output does not work for me :

-o ./found.txt

ERROR: Cannot open output file =./found.txt

even if i create found.txt manually

--output does work!

oh yeah and could one manually remove the 4000 hash limit ?

Interesting, I'll look at -o behavior on Windows. Can't say I really tested it extensively beyond Linux...

The 4000 hash limit cannot currently be removed. It's related to the fact that there is only 64kb of cached constant memory that I use to store hashes. I could store more, but currently, accessing global memory for the checks would be very, very slow.

That said, I do have some ideas up my sleeve that may *dramatically* improve performance on high hash counts. Order of magnitude or more, potentially. InsidePro is doing a much better (faster) job on large hash lists, and I had some brainstorms on how they might be doing it.

[crazydoc]

It would be very interesting to distribute cracking on two or more machines, each with a supported graphics card. For example, I like the way the (Windows-) tool "LCP" does it. There you can split the work in as many parts as you want and the tool will create files containing the info what to do on a certain machine.

[Bitweasil]

It would be very interesting to distribute cracking on two or more machines, each with a supported graphics card. For example, I like the way the (Windows-) tool "LCP" does it. There you can split the work in as many parts as you want and the tool will create files containing the info what to do on a certain machine.

Interesting idea.

Not likely to be implemented terribly soon, though I am working on multi-GPU support (as multi-core GPUs seem to be the way everyone is going right now).

My goal for distributed work is the rainbow tables right now. The brute forcers are an interesting side project.

[the_drag0n]

It would be very interesting to distribute cracking on two or more machines, each with a supported graphics card. For example, I like the way the (Windows-) tool "LCP" does it. There you can split the work in as many parts as you want and the tool will create files containing the info what to do on a certain machine.

that could be done by a resume function. you could do that with bars for single hashes as well by setting the resume word on pc2 to another then on pc1.
resume function would be nice anyway

[Bitweasil]

That said, I do have some ideas up my sleeve that may *dramatically* improve performance on high hash counts. Order of magnitude or more, potentially. InsidePro is doing a much better (faster) job on large hash lists, and I had some brainstorms on how they might be doing it.

Yeah... these are looking, so far, to be pretty sweet.

Still working on it (had to change a bunch of code around, but getting multi-length hash support in as well), it'll likely slow performance on a single hash a bit, but should kick check rates on Very Large Hash Lists (50k or above) into the trillions of checks per second or greater range. My goal is 1 000 000 000 000 checks per second on sufficiently large (10M element or greater) lists. Initial testing indicates this may be possible.

The main limit is that I will require large amounts of video RAM to do this - for a 5M element list, 512MB will be required, and will grow from there with larger lists. I will attempt to degrade gracefully (but still usefully) on lower end cards.

I'm also tightening up the code much further, and making it easier to add more hashes to. Hopefully I will be releasing something this weekend, at least for Linux.