Best way to use all GPUs

[jamesyoungdigital]

What would be the best way to use all GPUs, giving each instance a keyspace of its own to search?

I was thinking of a shell script like this:

#!/bin/bash
./cuBitCrack -d 0 -s 0x[..] -i list_of_addrs -o ~/cracked.txt &
./cuBitCrack -d 1 -s 0x[..] -i list_of_addrs -o ~/cracked.txt &
./cuBitCrack -d 2 -s 0x[..] -i list_of_addrs -o ~/cracked.txt &
./cuBitCrack -d 3 -s 0x[..] -i list_of_addrs -o ~/cracked.txt &
./cuBitCrack -d 4 -s 0x[..] -i list_of_addrs -o ~/cracked.txt &
./cuBitCrack -d 5 -s 0x[..] -i list_of_addrs -o ~/cracked.txt &
./cuBitCrack -d 6 -s 0x[..] -i list_of_addrs -o ~/cracked.txt &
./cuBitCrack -d 7 -s 0x[..] -i list_of_addrs -o ~/cracked.txt &

But I have some questions: suppose this run was quite successful, and they all wrote to ~/cracked.txt, would that lead to corruption? I know it's in append mode, but might there be a condition where two are trying to append at the same time?

Thanks, how do you run multiple instances and manage them? This script is terrible. I have to run ps aux, and then kill them off one by one when I want to do something else!

[g00dwin555]

You can run multiple instances of the program.
one cmd one graphics processor and one file with the results.

[jamesyoungdigital]

Thanks. So I can just use one file to output results of cracking, and all instances will append to that file correctly?

It would be nice to have them running in the background so I can continue to work on the machine, I'll have a look at the options to see if there's a silent option (like write results to output file when found, don't display progress line/s).

[brichard19]

Theres no silent option right now. I'll open an issue for adding more flexible output/logging options.

[brichard19]

As for multiple GPUs, the only way right now is one process per GPU. It is a much simpler design. In the future there will probably be multi-GPU support out of the box. I guess this will be done by either managing multiple processes or multiple threads.

Beyond that we also need a way to support work distribution among multiple machines since a single machine, or even 10 machines, will not be able to crack the 60-bit puzzles.

[ogronome]

they all wrote to ~/cracked.txt, would that lead to corruption

James Young, what prevents you from pointing to different files?
unix shell is very powerful. You probably need to educate yourself a bit to stop suffering managing several processes.

Beyond that we also need a way to support work distribution among multiple machines since a single machine, or even 10 machines

brichard19, LBC2 ? Great idea. But without "unlogged remote execution feature", please :)

[brichard19]

brichard19, LBC2 ? Great idea. But without "unlogged remote execution feature", please :)

I was thinking more of local clusters/farms. But you're probably right. The puzzles will get so big the only way to solve them is through a community effort.

[voyagerlife]

I have 2 GPUs with a HB SLI Bridge, is it possible now to use both graphics cards as 1 GPU in only one CMD for all addresses?

[voyagerlife]

@brichard19 Apparently there is no support for bitcrack anymore and no answers to the questions...