Cryptohaze.com

http://www.cryptohaze.com/releases/CUDA ... .7.tar.bz2

I'm working on getting a SourceForge project set up, but this should suffice for now.

Comments welcome.

Questions welcome if they aren't related to "GAH! UGLY CODE!" - it's due for a total conceptual change & rewrite here soon.

Hello,

Been reading the code:P Of course you're free to use my code:) (as i was already hoping for someone to port it to linux )
I wonder tho, why did you declare the charset and other arrays as shared? as i've seen no improvement in speed while doing that myself (due to the alignment they were stored in registers).

Also, i've got a ati brook+ version of sha1+mssql+mysql done, and im working on the interface atm, maybe we should work together?

Yours,

Vampyr

(And yes i know i overuse smileys)

What else would I declare the character set as?

The biggest advantage of shared memory is that it's multiported (16 banks, all able to be read at once). As the character set calls will be pulling from all over the place, and not a uniform location, the multi-porting makes a significant difference in speed.

I do not believe based on register counts and observed performance that the character set arrays are being put in registers - ATI cards have significantly more registers available, so they may do things differently. I see no indication that these are stored in the register file on nVidia GPUs.

I was hoping to get in touch with you - I tried to find you on gso IRC, but didn't have much luck.

The easiest way to get in touch with me is via #cryptohaze on irc.freenode.net - I believe I saw you log in briefly to GSO, but you did not respond to queries.

I am looking to turn this into an open source cracking project that supports a wide variety of CPU and GPU clients - I would love to have ATI modules available.

Alternately, PM me and I can provide alternate contact info.

Irc it is

I personally prefer constant memory for the char set, and registers for the plaintext/plaintext indexes (somehow they seem to be copied in there), as the char set is not likely to change often during execution of the program, constants seemed a nice place for it to reside;)

At this moment i've set up the V2 of my cracker like this:
shabruter_ati.dll (contains all ati specific functions, called trough a generic interface)
shabruter_nvidia.dll (same for nvidia)
shabruter.exe (interface only)

Using a common interface should allow for hacking up cpu/network support within a week or so.

Maybe we could work out a common API or network interface, so that we can plug in extensions for algorithms and devices into the interface?
So for instance the interface requests if the given extension can support algorithm x, if so, it executes some work, etc.
As my implementation is windows only, i'm trying to look into a way to make this portable to linux

I _really_ dislike the ATI Brook+ SDK tho, as there is no support for arrays, i had to hack up the plaintext generation algorithm to perform its task on dwords (tedious as hell). And there's quite a bit more nasty to that compiler.

Hi i working under linux but i can't compilate CUDA, how it work?


computer@azerty:~/desktop/CUDA-Multiforcer-Release$ make
Makefile:47: ../../common/common.mk: No such file or directory
make: *** No rule to make target `../../common/common.mk'. Stop.
computer@azerty:~/desktop/CUDA-Multiforcer-Release$


thanks

Install the CUDA SDK and put the source in the /projects/ directory of the SDK and it should compile.

Hi thans for the reply but where i can take CUDA SDK?

Go to nvidia.com then go to the menu "Technologies -> CUDA." When that loads, go to the menu "Developing with CUDA" -> "Get CUDA."
Or just click here.

Hey I've hit a snag when trying to make the code. I'm guess I just don't have everything configured correctly any advice would be helpful.

zink@ubuntu:~/NVIDIA_CUDA_SDK/projects/CUDA-Multiforcer-Release$ make
make: *** No rule to make target `-largtable2', needed by `../../bin/linux/release/CUDA-Multiforcer'. Stop

Ehm, my Makefile is kind of nasty. It puts -l arguments in a place that technically isn't correct, it just happens to work on my system.

You might want to redo the make file.

This is the "GAH! CODE UGLY!" category of bugs. I did not release the source as "Hey, this works perfectly" source, more of "Here... enjoy. No promises."