Single hash LM kernel issue

[Rolf]

There's a rugbug in single hash LM kernel in the latest and greatest release of Multiforcer for Windows.

The command line used:

Code: Select all

Cryptohaze-Multiforcer.exe -h LM -c charsets\charsetMicrosoft -f md6.txt -o md6.out --min=1 --max=7 -t 512 -b 360 -m 500

Put any single LM hash in md6.txt file and Multiforcer will report "Terminating due to error: All hashes found! Exiting!".

Multihash LM kernel not affected at all.

I get blazing 652M p/s for two GTX 480.
Elcomderps and their bitsliced LM kernels just got pwnt.

[Bitweasil]

Is md6.txt newline terminated?

Either way, I'll look into this tonight. LM hashes have a new hash file handling class to split it into two parts, so it's definitely not as tested as the old stuff. Thanks for the bug report!

Out of curiosity, what does the ECom LM module do on your box, speed-wise?

[Rolf]

Yes, tried windows and linux(though winedows uses both CR and LF) newline termination, issue still exists.

486M p/s for the same GPUs.
Your Multiforcer is 34.15% faster

[Bitweasil]

Yes, tried windows and linux(though winedows uses both CR and LF) newline termination, issue still exists.

486M p/s for the same GPUs.
Your Multiforcer is 34.15% faster

Awesome. I think that answers the bitslice question nicely... the DES implementation I'm using is pretty much ideal for shared mem.

I'll look into the single hash issue & get it fixed soon. Thanks for finding it!