Skip to content

Optimize CPU/GPU task scheduling #47

@n13

Description

@n13

GPU miner should just be given new tasks, there's no need to wait for it to finish.

Overall the system as it is is somewhat inefficient as it's treating GPU and CPU the same scheduling tasks to them

CPUs tend to run out of work sometimes.

GPU waits for finishing at 3s which wastes a lot of cycles too - maybe we can find a better system.

Benchmarks are higher than actual throughput because of this - benchmarks are an idealized case where we just crunch numbers and don't get interrupted or have to care about cancellation. In the real world though cancel happens all the time.

There sometimes seem to be breaks in processing too, not sure why this happens.

So overall this issue is: Make real world hash rate closer to bench hash rate (optimal hash rate)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions