CUDA stream optimisation

envidian commented 8 months ago

Adds argument paired_batch_size for specifying fast 1:1 pairing
Adds custom handlers for paired_batch_size
Defines JSON schema for specifying 1:1 combinations and config
Adds use_fast_math support in cmake build (default enabled)
Fix memory leaks in Autodock Vina scoring_function reported by Valgrind
Fix missing CUDA memory initialisation that causes crashes

The original code flow is retained when paired_batch_size is not specified. Please refer the README for using the paired_batch_size argument.

This commit reverts previous request, and makes the code aligned with original code-flow.

envidian commented 8 months ago

Is there any additional information required for this to be reviewed ? @ysyecust Thanks.

ysyecust commented 8 months ago

@envidian Thank you very much for your pull request. If you could add an example of using 1:1 docking in the example folder, I believe it would make it even more perfect and greatly facilitate other users. We would greatly appreciate your consideration of this suggestion.

envidian commented 8 months ago

Thank you. I added https://github.com/dptech-corp/Uni-Dock/commit/5def75620d13b73e371ec7993f903c5bfe65b413 to address the comment. Can you please check ? @ysyecust

dptech-corp / Uni-Dock

CUDA stream optimisation #63