Skip to content

Latest commit

 

History

History
92 lines (65 loc) · 2.43 KB

README.md

File metadata and controls

92 lines (65 loc) · 2.43 KB
Raw

Golang NanoPoW

NanoPoW is an implementation of the proof-of-work used by Nano. That supports CPU and GPU (currently only OpenCL).

Usage

func main() {
    nanopow.GenerateWork([]byte{/** PREVIOUS HASH **/}, nanopow.V1BaseDifficult)
}

CPU

By default, we always uses CPU. If GPU is available, it will use CPU and GPU, combined.

GPU

By default, we don't support GPU. You need to enable it using the "build tags", when build use:

go build -tags cl

Currently the only available option is OpenCL ("cl" tag).

Multiple GPUs

By default, we use the first detected opencl-compatible GPU. To change this or to use multiple GPUs you can create your own work pool.

Here's an example using all of the GPUs

// Using github.com/inkeliz/go-opencl
platforms, err := opencl.GetPlatforms()
if err != nil {
    return nil, err
}

var platform opencl.Platform
var name string
pool := nanopow.NewPool()
for _, curPlatform := range platforms {
    err = curPlatform.GetInfo(opencl.PlatformName, &name)
    if err != nil {
        return nil, err
    }

    var devices []opencl.Device
    devices, err = curPlatform.GetDevices(opencl.DeviceTypeAll)
    if err != nil {
        return nil, err
    }

    for _, device := range devices {
		gpu, gpuErr := nanopow.NewWorkerGPU(device)
		if gpuErr == nil {
			pool.Workers = append(pool.Workers, gpu)
		} else {
			fmt.Printf("\n⚠️ Unable to use GPU %v", gpuErr)
		}
    }
}

// Generate work
pool.GenerateWork(hash, difficulty)

"Benchmarks"

I don't have an huge amount of data to compare and limited devices to test against. Some devices doesn't supports OpenCL. All times are using the V1 difficulty.

Device OS CPU-Only CPU + OpenCL
Samsung Galaxy S20+ Android ~9.62 seg Untested
Blackberry KeyOne Android ~36.7 seg Untested
R9 3900X + RX 5700XT Windows ~0.66 seg ~0.27 seg

Limitations

Support OpenGL, OpenGLES or Vulkan:

Currently we only support OpenCL, which is not natively supported by some devices. I have some plans to add some support for OpenGL, OpenGLES or Vulkan. Seems that Vulkan Compute, if possible, is the best solution since it's compatible with Windows, Linux and Android.

Contributing

Pull requests are welcome.

License

MIT