Tromp's solvers

Mining
227

One possible optimization I just tried was further reducing buckets to 2^11.
However any speed advantage of this was cancelled by the need for additional slot encoding, so I’m sticking with 2^12 buckets…

228

I believe the largest bottleneck is now is an efficient workflow for making use of Xenon’s avx/avx2 solvers while still having the ability to use a Stratum Pool or solo mine.

Have you done any bench marking on stratum client integration?

I have gotten some extremely good results by multi-processing and using MySQL/Redis as a local cache to optimize job management and inter-process communication.

5 Sol / s…and it’s PHP 7 @_@

1 Like
229

Hey @tromp, thank you for your work. Is there any plan for an openCL port ?

230

The testnet difficulty is going up due to optimized CPU miners now operating on the testnet!

Compared to last night block 8200 or so difficulty of less then 60!

231

Genoil developed an OpenCL port of my solver, but I don’t know what plans he has for it…

232

Can someone make a Guide for the exact steps to install and run the optimization?

I have no knowledge of coding or programming, I can only follow easy instructions as here: Home · zcash/zcash Wiki · GitHub

Thank you

233

Thank your for the answer. Any update from @Genoil ?

234

@hellcat can you guide me how to check block height and Network Hash/s plz

235

I’m working on an opencl port / build from the ground up. I will post updates with my progress.

Thanks for all of your hard work btw @tromp!

2 Likes
236

Okay great news !! Thank you for your work.

237

Warning: I’ve just started another major rewrite of my solver that should bring some small speedups.

2 Likes
238

Can someone explain what is time Offset=-1 in testnet and what does it has to do with difficulty and block size

239

tromp, you did great work. Doing 6.8 H/s on i7-5930K.

240

How is that possible? I only got 4 Sol/s on a 4Ghz i7?!
Oh, you mean multi-threaded?

1 Like
241

Yes, running with 8 threads (not 12) seems to be the best.

242

noob question: how did you determine the ideal # of threads?

1 Like
243

Enough coding and benching for today.

Will do some serious loop unrolling tomorrow, trying to get closer to 5 Sol/s…

244

Try&error. 8 gave me the best result in terms of H/s.

245

Nicehash, little appreciation to these devs. Btc would do. Show some love.:slight_smile:

1 Like
246

Hi @tromp,
if @xenoncat solver is giving 20sol/s on 8 thread machine,
With 8 instances of equi1 can give 4 * 8 = 32 sol/sec.

Your solver is better placed isn’t it?
Correct me if my understanding is wrong.