dmarcuse / jclminer

The first GPU-accelerated Krist Miner - using JavaCL

This project has been abandoned!

This project has been abandoned in favor of the less buggy, more modular, and generally better turbokrist. No futher changes will be made here and any issues or pull requests will be ignored. Functionality of the current state of this project is not guaranteed.

JCLMiner

The first GPU-accelerated Krist Miner - using JavaCL

JCLMiner is a heavy WIP. With that in mind, you can check out the wiki for instructions on building and using JCLMiner.

jclminer's Issues

Access violation after mining a bit

After mining a while, this happened:

Here's hs_err_pid6028.log.

Multiple "Mining for block" messages

^, most "mining for block" messages are duplicate for me:

However, some of them aren't?

I assume this has something to do with having multiple devices?

EDIT: They only seem to be duplicate after a block success/rejection. When the block changes in the middle of mining, it's correct.

Multi-GPU broken

I'm running the miner with -a, however it only mines on a single GPU. -l finds two GPUs, same signature, of course, but differing IDs (0 and 1).

JCLMiner does not detect block changes while mining

Now that I have someone else mining at the same time I can find bugs like this :P

Linux bugs :DD

So for some reason the miner now magically runs on Linux on my PC...
Of course, as things are, it's buggy as hell, so have fun:

• Profiler crashes the JVM after first test finishes. Crash logs. It still continues but does nothing until it eventually core dumps.
• -a doesn't work. I am on latest beta, so it's not #18. It says it's starting to mine on both GPUs, but I only get the performance of one GPU... or maybe it's not broken at all and the performance really is that crap on Linux.
• Crash on solution submission. It seems to be the same error (caused by clSetKernelArg) as the profiler one, but here it is anyway.

CLException while mining

While mining, I got this exception:

Exception in thread "Thread-37" com.nativelibs4java.opencl.CLException$InvalidMemObject: InvalidMemObject (kernel name = krist_miner_basic, num args = 6, arg index = 2, source = [the kernel source]

It continues mining fine though. (EDIT: sometimes it crashes)
Could this be related to #9?

Miner UI

I'd be happy to write a simple UI to work with the miner. Since all of this is done on the GPU, there shouldn't be any performance impact unlike with CPU mining.

RemoteErrorException class missing in latest beta

I got it to run by removing the exception handling that involved it.

Also, the solution is sometimes rejected, but it works most of the time...

License

What license do you want to use @GeorgeMcKinlay? Personally I'd use the GNU GPL v2 or v3, but if you want something else I'm open to ideas.

Hardware for testing

So far, I've been running all my tests on integrated graphics in my laptop, and my Nvidia GTX 760 in my desktop. If possible, we should also try to test on AMD hardware to ensure compatibility.

Repeated solution rejections

Occasionally, the miner will repeatedly attempt to submit multiple solutions in a row, most/all of which are rejected by a server. This usually happens after a successful submission, leading me to believe that it is due to a bug where the miner does not stop and restart correctly, mining for an old block.

After a few tries the miner will usually fix itself and begin mining normally again with only a minor disruption.

Rewrite Java portion of JCLMiner

The current frontend isn't turning out the way I had anticipated, and is getting difficult to manage and buggy. I am going to be working on rewriting the entire Java portion of JCLMiner, (in a separate branch, of course) making it more extensible and modular.

Vectorize SHA256 implementation for better performance

Modifying the SHA256 implementation to use vector data types and operations instead of scalars should lead to significant performance gains when supported.

Compiler error on Linux

On Linux (Ubuntu 15.10), I'm getting a compiler error while running gradle build. While compileJava runs fine, compileTestJava complains about this line in TestCLHashing.java:

import jdk.nashorn.internal.ir.annotations.Ignore;

Apparently, such a class does not exist. Since I am using OpenJDK 8, perhaps this class is specific to the Oracle JDK?