v2024.09

What's Changed

• Detect and disallow variable shadowing by @lukeshingles in #115
• Perf: Cache array of level pops in photoionisation nu_edge order by @lukeshingles in #118
• Reduce heap fragmentation during startup and store allcont keep condition in cell cache by @lukeshingles #116
• Default off for VERIFY_WRITTEN_PACKETS_FILES by @lukeshingles in #122

Full Changelog: v2024.08...v2024.09

Performance

The previous release v2024.08 was not benchmarked, but significant gains have been made since v2024.04 with bit-identical output (except for the classic phixs read-in fix in v2024.08).

• W7 (Shingles et al. 2022) NLTE nebular 960 core 2e9pkt 150-410d JUWELS runtime decreased from 18.6k core hrs (published 2022) to 10.4k core hrs (v2024.04 version) to 6.4k core hrs (2.9x faster than 2022).
  • With the Simpson-rule integrator, time decreases to 4.96k core hrs (3.7x faster than 2022)
• subMch Shen+2018 NLTE nebular 960 core 1e9pkt 150-410d runtime decreased from 13.14k core hrs (20210417 version) to 7.30 k core hrs (v2024.04) to 5.36k core hrs (2.45x faster than 20210417).
• sim2010 classic mode 2-120 days 2e7pkt 960 core JUWELS runtime decreased from 5.43k core hrs (ARTIS-CLASSIC) to 3.81k core hrs (20230526 version) to 1.87k core hrs (v2024.04) to 1.18k core hrs (4.6x faster than classic).