Benchmark with DaCe cpu and gpu backends - Githubissues

NOAA-GFDL / pace

Re-write of FV3GFS weather/climate model in Python

Apache License 2.0

12 stars 11 forks source link

Benchmark with DaCe cpu and gpu backends #50

Closed fmalatino closed 6 months ago

fmalatino commented 6 months ago

Purpose

Branch for benchmarking of PACE. It includes four new configuration files to be used in benchmarking with DaCe backends for use with cpus and gpus.

Code changes:

Addition of four configuration files: baroclinic_c384_cpu/gpu.yaml, baroclinic_c3072_cpu/gpu.yaml
Use of dace_inhibitor decorator in microphysics.py

Requirements changes:

In constraints.txt: dace updated to 0.15.1
In constraints.txt: gridtools-cpp updated to 2.3.1

Infrastructure changes:

N/A

Checklist

Before submitting this PR, please make sure:

[x] You have followed the coding standards guidelines established at Code Review Checklist.
[x] Docstrings and type hints are added to new and updated routines, as appropriate
[x] All relevant documentation has been updated or added (e.g. README, CONTRIBUTING docs)
[x] For each public change and fix in pace-util, HISTORY has been updated
[x] Unit tests are added or updated for non-stencil code changes

FlorianDeconinck commented 6 months ago

The DaCe version has been updated? Did we run model validation on it?

fmalatino commented 6 months ago

The DaCe version has been updated? Did we run model validation on it?

I have to change it to your gcc_dies_on_dacecpu branch, will be in the coming push.