Tromp's solvers

I'm not doing any work on OpenCL, but I've given genoil access to my solvers and he's porting them to OpenCL, and testing them on various AMD hardware. However, even if I go open source, it's up to him to decide what to do with his port.

New 0,2 BTC on the way 494e636b63277b808f9ca2fd94aedcaa9f81d21491ea5ef51aba089ca7af84b0

2 Likes

I think if You go open-source NVidia should provide You with a grant :slight_smile:. I would for sure.

1 Like

A big thanks to @superresistant for his sizeable contribution!

Also thanks to @marco!

As others have pointed out, we're 2/3 of the way to the 20BTC goal, so I'll extend the deadline to noon tomorrow. As soon as the goal is reached, I will be committing the solvers to

https://github.com/tromp/equihash

2 Likes

The overnight run of 10000 (144,5) instances yielded 19993 solutions, so it's safe to use a 2 Sol/run average (compared with 1.88 for 200,9), at least for those solvers that discard a negligible fraction of solutions.

1 Like

It was pointed out to me on Slack that a solver is available at

https://github.com/xenoncat

which claims to significantly outperform mine, at 4.4 Sol/s. I read the accompanying documentation and it has much of the same ideas as mine so this seems quite credible.

Since this significantly reduces the value of my solvers, once I confirm the performance, I will offer full refunds to all contributors.

2 Likes

I'm sure many here would like a Cuda solution also ... I for one could use one + ntm only new processors have avx2

1 Like

Compared to xenoncat, whose methods are described in

https://github.com/xenoncat/equihash-xenon/blob/master/notes/algorithm%20description.pdf

my solver differs in having way more buckets, wasting some memory, having simpler pair compression, being multi-threaded, and supporting (144,5).

And of course in not using any assembly.

Oh, and having some cool visualization of bucket size distribution...

3 Likes

I would prefer to not be refunded and still have access to your code. Your code is in C, so I can tinker with it. I don't know assembly. Also, the CUDA support is also valuable.

2 Likes

Once the xenoncat performance claims are confirmed as I expect they will, then I'll offer full (or partial, if you like to support Cuckoo Cycle) refunds, and open source my solvers anyway.

6 Likes

Wow, you are an absolute legend. Thankyou!

2 Likes

Also thanks to xenoncat, whoever he or she may be (my hat off to you) ...

1 Like

Performance of xenoncat is confirmed, but due to API mismatch, correctness of solutions found is not confirmed yet. No doubt that will happen soon, and I'm already preparing my commits...

2 Likes

OK; I've decided to bite the bullet. Full source is available at

https://github.com/tromp/equihash

Just run

git clone git@github.com:tromp/equihash.git

make all

and enjoy. I will be contacting contributors in decreasing order of donation, and asking how they want to be refunded....

Whoever sent 1BTC to OgNasty, please let him know how to handle your refund.

8 Likes

Awesome! You rock, Mr. Tromp.

1 Like

What kind of dependencies are required? Haven't tried to compile yet.

1 Like

You have open sourced your work -- Thank you! -- please keep my small donation.

2 Likes

No dependencies, I think?! Let me know if you find otherwise...

git clone https://github.com/tromp/equihash

works for me.

I get these results:

3 solutions
3 total solutions
1.75user 0.10system 0:01.86elapsed 99%CPU (0avgtext+0avgdata 216080maxresident)k
0inputs+0outputs (0major+7298minor)pagefaults 0swaps
2 Likes

which cpu are you using?