Changing Argo float kernels to use boolean masking

[erikvansebille]

This PR changes the Kernels in the Argo tutorial to use boolean masking, which is easier to read than the np.where() clauses used before.

I'll create a script in the benchmarking repo so that we can test performance between the two types of vectorised if-statements (see also #2197)

Description

Checklist

• Addresses Benchmark use of np.where() versus boolean masking in Parcels kernels and interpolators #2197
• Tests added
• This PR targets the correct branch (main for normal development, v3-support for v3 support)

AI Disclosure

• This PR contains AI-generated content.
  • I have tested any AI-generated content in my PR.
  • I take responsibility for any AI-generated content in my PR.
  • Describe how you used it (e.g., by pasting your prompt):

[erikvansebille]

To inform the difference between np.where() and boolean masking in this Kernel, I created Parcels-code/elphe-hackathon#11

Short summary: they are virtually the same speed. I therefore propose to change the tutorial to use this simpler/more-intuitive kernel

Below is the output for the (old) kernel using np.where(). cumtime is 0.229 seconds

(boolean_masking_profiling) erik ~/Codes/elphe-hackathon/boolean_masking_profiling % python argo_kernel_profiling.py --argo-kernel npwhere
/Users/erik/Codes/elphe-hackathon/boolean_masking_profiling/argo_kernel_profiling.py:9: UserWarning: This is an alpha version of Parcels v4. The API is not stable and may change without deprecation warnings.
  import parcels
INFO: Output files are stored in output_argo.parquet
Integration time: 2024-07-06T22:00:00 100%|█████████████████████████████████████████████████| [00:12<00:00, 40740.27it/s]
         8916903 function calls (8782721 primitive calls) in 16.386 seconds

   Random listing order was used
   List reduced from 4024 to 3 due to restriction <'run_simulation|particleset\\.py:\\d+\\(execute\\)|ArgoVerticalMovement'>

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
      576    0.043    0.000    0.229    0.000 /Users/erik/Codes/elphe-hackathon/boolean_masking_profiling/argo_kernel_profiling.py:34(ArgoVerticalMovement_npwhere)
      2/1    0.000    0.000   12.726   12.726 /Users/erik/Codes/elphe-hackathon/boolean_masking_profiling/argo_kernel_profiling.py:127(run_simulation)

And below is the output for the (new) kernel using boolean masking. cumtime is 0.227 seconds

(boolean_masking_profiling) erik ~/Codes/elphe-hackathon/boolean_masking_profiling % python argo_kernel_profiling.py --argo-kernel boolean
/Users/erik/Codes/elphe-hackathon/boolean_masking_profiling/argo_kernel_profiling.py:9: UserWarning: This is an alpha version of Parcels v4. The API is not stable and may change without deprecation warnings.
  import parcels
INFO: Output files are stored in output_argo.parquet
Integration time: 2024-07-06T22:00:00 100%|█████████████████████████████████████████████████| [00:12<00:00, 42156.98it/s]
         8964778 function calls (8830591 primitive calls) in 16.681 seconds

   Random listing order was used
   List reduced from 4026 to 3 due to restriction <'run_simulation|particleset\\.py:\\d+\\(execute\\)|ArgoVerticalMovement'>

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
      576    0.021    0.000    0.227    0.000 /Users/erik/Codes/elphe-hackathon/boolean_masking_profiling/argo_kernel_profiling.py:85(ArgoVerticalMovement_boolean)
      2/1    0.000    0.000   12.299   12.299 /Users/erik/Codes/elphe-hackathon/boolean_masking_profiling/argo_kernel_profiling.py:127(run_simulation)

[VeckoTheGecko]

Looks good!