Summary
This issue is to acknowledge and track the progress of Parcels v4 performance enhancements (a parallel issue in the Parcels repo to be cross-referenced here when live) and its relevance to VirtualShip v1.
More detail
A core part of an upcoming VirtualShip v1 release (#195) will mean the move from Parcels v3 -> v4 (#342) under the hood. In line with the aims of the Parcels v4 release, this overhauls the VirtualShip internals to align more closely with the Pangeo ecosystem and use vectorised operations.
Parcels v4 computation works and is ready for use in VirtualShip simulations, but its performance will continue to be improved as v4 evolves. Latest performance testing suggests that local data ingestion is performant but streaming workflows (e.g. with copernicusmarine) are slower. This is relevant to VirtualShip, where a key feature is the default Digital Twin/streaming nature of the framework. Thus, we should expect to see some performance decrease on initial VirtualShip v1 release. However, ongoing work on the Parcels side to increase performance in v4.x... will feed directly into VirtualShip.
Summary
This issue is to acknowledge and track the progress of Parcels v4 performance enhancements (a parallel issue in the Parcels repo to be cross-referenced here when live) and its relevance to VirtualShip v1.
More detail
A core part of an upcoming VirtualShip v1 release (#195) will mean the move from Parcels v3 -> v4 (#342) under the hood. In line with the aims of the Parcels v4 release, this overhauls the VirtualShip internals to align more closely with the Pangeo ecosystem and use vectorised operations.
Parcels v4 computation works and is ready for use in VirtualShip simulations, but its performance will continue to be improved as v4 evolves. Latest performance testing suggests that local data ingestion is performant but streaming workflows (e.g. with
copernicusmarine) are slower. This is relevant to VirtualShip, where a key feature is the default Digital Twin/streaming nature of the framework. Thus, we should expect to see some performance decrease on initial VirtualShip v1 release. However, ongoing work on the Parcels side to increase performance in v4.x... will feed directly into VirtualShip.