Skip to content

mapreduce fallback tracking issue #246

@bjarthur

Description

@bjarthur

in trying to understand why count was slow, i came across a comment explaining that it used to fallback to mapreduce but now does not because of internal refactorings of the API. so i dug through this a bit so as to learn how DiskArrays worked, and found that the Generator is hit instead of mapfoldl_impl at least back to julia v1.9. didn't go back further as that is the current minimum version required by DiskArrays.jl.

given enough developer time, it'd be nice to have these fallback method work again. in the meantime, we might consider blocking them out with VERSION as is commonly done in CUDA.jl. probably shouldn't delete them entirely (yet), as they would serve as a template for updating them.

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