Skip to content

Better document the impact of data dtype on performance #992

@psobolewskiPhD

Description

@psobolewskiPhD

🧰 Task

Our vispy/gpu pipeline only supports int8, uint8, int16, uint16, float32, other dtypes get coerced during slicing.
This has a pretty significant performance implication for large slices of e.g. float64 or in64 data.
See also: napari/napari#1300 (comment)

Importantly, many numpy functions can produce float64 or int64 by default, triggering this issue.

We should raise more awareness of this in the documentation and possibly add an in-depth look at the data handling pipeline from Layer.data to the displayed gpu texture.

Metadata

Metadata

Assignees

No one assigned

    Labels

    documentationImprovements or additions to documentationtask

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions