GPU batched delta-Eddington shortwave — 291x speedup at 500K columns

[consigcody94]

Summary

The delta-Eddington shortwave radiation computation can run on GPU with one thread per ice column. The 3-band absorption/transmission through snow+ice layers is independent across columns.

Benchmark (RTX 3060)

Ice Columns	CPU	GPU	Speedup	Max Rel Error
10K	<1 ms	0.017 ms	—	4.30e-07
100K	5 ms	0.016 ms	310x	4.49e-07
500K	22 ms	0.076 ms	291x	5.62e-07

Zero NaN, zero failures. The large speedup comes from the exp() calls for layer transmittance — GPU hardware exp is very fast.

Code

$NEW_REPO (noaa_multi_kernel.cu)

[consigcody94]

AI Attribution: This work was developed using Claude Code as an AI-assisted development tool. Code generation and benchmark automation were performed by Claude under my direction. I designed the research methodology, selected mathematical approaches, and validated all results on physical hardware (RTX 3060). All commits carry a Co-Authored-By: Claude tag. When results failed validation (e.g., tensor compression on rte-rrtmgp#394), they were publicly retracted.