[FEATURE]: Assign task per videocard

[fuad00]

Description

Let's say i have 1 agent with 16 videocards and the goal is to brute-force tasks in parallel, not agent per task. Goal is to precess small tasks and give result as fast as possible without creating queue

[Brets0150]

As far as I know, Hashcat cannot do this. Hashtopolis is only a wrapper for Hashcat, so if Hashcat cannot do it then Hashtopolis can't do it.

You may consider splitting the GPUs up to multiple systems. Maybe install Proxnox on the system, Build 16 VMs, and passthrough 1 GPU to each VM; thus creating 16 Hashtopolis agents. This maybe okay with a high-performance(lots of ram and CPUs) host system. or it may be awful. ¯_(ツ)_/¯

[fuad00]

Thanks for response!

Virtualization per hashcat probably bad idea

Build 16 VMs

I've tried to run hashcat with different hashmodes in parallel but looks like there some kind of pid lock protection:

I will try to bypass it. If no success maybe i have to try multiple hashtopolis docker containers, but definitely not VMs because of performance

[davidbolvansky]

use —session flag to set different sessions

[fuad00]

—session flag

Thanks for solution! @Brets0150, Looks like hashcat can do this!

Maybe somewhere in hashtopolis code we need to add --session flag with random value in order to split one agent's videocards into their tasks

[s3inlc]

You can configure a hashtopolis agent to only use certain devices (using the -d flag) in the agent specific parameters.

Set up as many agents on the same machine (and in different directories) as you want to have it split into and adjust the agent specific parameters to distribute your GPUs among them. Then each of the agents can work on different chunks/tasks.

[fuad00]

using the -d flag

!solved very uncommon and beautiful solution, thanks!