Multi-GPU miner rebased and up to date with Roma network



I have rebased the multi_gpu branch on GitHub to be placed on top of the Roma release - it should not make any functional difference but it looks better if you look at the Git history.

To use the Multi-GPU functionality you should check out the multi_gpu branch and build it:

git clone epoch
cd epoch
git checkout multi_gpu
make prod-package


Testing this in ubuntu 18.04

sudo apt-get -qq update && sudo apt-get -y upgrade && sudo apt-get -qq -y install git curl autoconf build-essential ncurses-dev libssl-dev erlang libsodium-dev g+±6 gcc-6
sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt install nvidia-cuda-toolkit
git clone -b multi_gpu epoch && cd epoch
make prod-build
cd apps/aecuckoo
make cuda29

This compiled, i will see if it works :slight_smile:


Rebased right now on what will most likely be released as 1.0.1.


hi can someone send me some useful lines of epoch_pow_cuckoo.log? ± 400 to 1000 lines

just put them here:

what else I need is some log lines where you found a block.


So I have cloned the repo again and checked out the 1.0.1 commit ee488f4447902e782b0390fc5a8bdf6e679495b3 , then:

  1. built the prod-build
  2. built the cuda miner
  3. copied the cuda29 binary to the lib/aecuckoo-0.1.0/priv/bin folder.

But I get an error (I KNOW its meaning, it’s not about that) when checking my epoch.yaml that I previously used on a different version of the multi-gpu miner:
Validation failed Position: mining/cuckoominer/instances Reason : No extra properties allowed

Why doesn’t my node understand that I am running a cuda node ?


That looks like you are using a multi_gpu-branch configuration on an ordinary node.

You have to build the multi_gpu branch in order to use instances: N in your configuration.


OK, but will the multi-gpu brnach build also include the updated gossip ? That is the reason I am doing this


Yes, the multi_gpu branch includes the gossip fixes.