Skip to content

Z17 LUT interpolation is 140X slower than full Z17 model #359

@oceancolorcoder

Description

@oceancolorcoder

@ARamsay17 @juanchossn

DALEC_14_2025-02-18_140256Z_L1BQC_L2.log

The attached log shows the partial progress of L1BQC -> L2 for one of four files being processed simultaneously in multithreaded command call.

Each ensemble is run through the MC RhoCorrections.read_Z17_LUT method of the Propagate class with 10 runs per ensemble.

As shown in the logs, each ensemble requires between ~70-80 minutes to process. (The ensemble is 5 minutes of data.)

LUT method occasionally fails individual runs on error, though this is not the source of the slow down:
"One of the requested xi is out of bounds in dimension 3: Unable to use LUT interpolations. Reverting to analytical solution."

From these logs, we also see that running the full analytical Z17 solution takes only 3.6 s, which is about 140X faster than the LUT interpolation.

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