Skip to content

README.md

Latest commit

 

History

History
60 lines (41 loc) · 2.64 KB

README.md

File metadata and controls

60 lines (41 loc) · 2.64 KB

Benchmark Results

Curated Linux Server Long-Safe Run

The server/A100 results intended for the current experiment write-up live in long-safe result directories:

Each directory has a README.md that explains every generated figure, the input CSVs, the filtering/aggregation rules, and the conclusions that the figure supports.

The figures are generated by:

  • experiments/server/long_safe_orchestrator.py
  • experiments/server/long_safe_plots.py

Regenerate from the repository root:

PYTHONPATH=experiments \
/home/wjiang49/conda_envs/fsm4py312/bin/python -c '
from pathlib import Path
from server.long_safe_orchestrator import plot_and_summarize
plot_and_summarize(
    Path("experiments/results/linux_server_cpu/long_safe_20260503_190133"),
    Path("experiments/results/linux_server_a100/long_safe_20260503_190133"),
)
'

The root-level quick-run server outputs were removed after validation. Only the timestamped long-safe directories are curated for GitHub.

A100 Permutation Break-Even

The long-safe A100 permutation chart is historical matched-slice evidence: n=5,000, p=50,000, batch_R=512, before streamed reduction and before the broader break-even sweep. It remains useful because it shows that a GPU logo does not guarantee a win.

The current permutation GPU narrative uses linux_server_a100/permutation_break_even/ for CPU/A100 break-even summaries, streamed-reduction notes, and representative pipeline decomposition.

Speedup in the break-even map is scoped to the matched CPU matrix baseline divided by A100 streamed full end-to-end. Compile is excluded, transfer is included, and kernel-only rows are not used for CPU/A100 speedup decisions.

Presentation Figures

Slides should use the normalized 16:9 summaries in presentation_figures/ when citing server/A100 results. These figures are regenerated by:

python -m experiments.visualization.plot_server_talk_evidence

The MacBook correctness and local evidence tier lives in macbook_air_long/latest/. Use those outputs for reference checks, calibration, power, and local runtime figures; use the server/A100 directories for large-scale claims.

Removed Historical Outputs

The old local benchmark directory was removed after the MacBook correctness/evidence tier and the server long-safe tier became the active story. Do not cite deleted historical figures in slides or the poster.