-
Notifications
You must be signed in to change notification settings - Fork 110
FAQ
Best way to start running dockings using AutoDock-GPU in no-time is by following the guideline for users.
Speeding up -- the typically-long -- AutoDock executions by leveraging the embarrassingly parallelism in its LGA implementation.
Preliminary development was part of the OCLADock project.
The AutoDock-GPU implementation involves modifications to the original AutoDock4.2.6 functionality in order to better exploit parallel processing, and the execution performance, without negatively affecting the docking quality. These modifications include the following:
- Arithmetic precision
- Arrangement of data structures
- Genetic selection scheme
- Specification of program arguments
For details, read the article describing the implementation details of AutoDock-GPU.
Definitely! One of the main program's feature is the functionality portability thanks to its OpenCL-based implementation.
While this should be possible, one should also bear in mind that performance portability with OpenCL is not a guarantee!
Devices with a significantly different underlying architecture might not benefit from a typical data-parallelization approach.
An example is OCLADock-FPGA that uses task parallelization for Intel FPGAs. The corresponding publication is available here.
This shoud be possible right away, although this wasn't tested yet.
Yes, similarly as in the original AutoDock4.2.6.
Corresponding development will be integrated into the mainline code soon.
Ideas are always welcome!
If you have ideas or feature request, please add an entry under issues.
If you want to contribute, please read the contributor's guidelines and then create a pull request.
Proceed as indicated here.
Go to Wiki home.
AutoDock for GPUs and other accelerators.
Contents