Gespeichert in:
| 1. Verfasser: | |
|---|---|
| Format: | Recurso digital |
| Sprache: | Japanisch |
| Veröffentlicht: |
Zenodo
2026
|
| Schlagworte: | |
| Online-Zugang: | https://doi.org/10.5281/zenodo.18939227 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Inhaltsangabe:
- <p># Zenodo v3.1 Description</p> <p>This release contains v3.1 of the Hakozuna allocator paper and artifacts.</p> <p>Contents include:</p> <p>- `main_ja.pdf` / `main_en.pdf`<br>- benchmark tables and reproducibility docs<br>- scripts and result summaries for SSOT (`T=16/R=90`), T/R sweeps, `dist_app`,<br> RSS snapshots, the mimalloc-bench subset, redis-like workloads, and the<br> current Windows-native supplemental benchmark lanes</p> <p>Profiles:</p> <p>- `hz3` (ACE-Alloc): local-heavy workloads and low RSS<br>- `hz4` (message-passing variant): remote-heavy / high-thread scalability</p> <p>Measurement scope:</p> <p>- the paper's main reported measurements were collected on Ubuntu (native) with<br> fixed build flags<br>- Windows-native allocator results are included as supplemental tables and<br> public benchmark summaries, while Ubuntu remains the main evaluation lane<br>- see `docs/benchmarks/` for commands, parameters, and environment details</p> <p>Source code is provided as:</p> <p>- `hakozuna/` (`hz3`)<br>- `hakozuna-mt/` (`hz4`)</p> <p>This release also documents a practical case study of rapid allocator R&D using<br>LLM-assisted design, implementation, and benchmarking.</p> <p>Release shape:</p> <p>- source-only release<br>- no prebuilt binaries<br>- public paper PDFs are tracked in `docs/paper/`</p>