1 reporting.m5_benchmark

reporting.m5_benchmark

Render a sealed, descriptive M5-versus-CUDA benchmark report.

The renderer admits six already-produced benchmark receipts and their timing table. It neither constructs models nor replays tensors: its only scientific job is to preserve the measured transaction, timing unit, and limits in a portable HTML/SVG report.

1.1 Classes

Name	Description
BenchmarkBundle	Validated input evidence retained by the static report renderer.
BenchmarkCell	One admitted backend/workload receipt with its descriptive timing summary.

1.1.1 BenchmarkBundle

reporting.m5_benchmark.BenchmarkBundle(
    input_root,
    cells,
    observations,
    summary,
    evl_manifest,
    input_hashes,
)

Validated input evidence retained by the static report renderer.

1.1.2 BenchmarkCell

reporting.m5_benchmark.BenchmarkCell(workload, backend, receipt, receipt_path)

One admitted backend/workload receipt with its descriptive timing summary.

1.2 Functions

Name	Description
latency_figure	Plot receipt-supplied descriptive quantiles without treating them as intervals.
load_bundle	Validate the frozen receipt bundle and expose its 180 timed observations.
main	Render the frozen M5 benchmark bundle without executing a benchmark.
render	Render an HTML report and sealed SVG assets from an admitted frozen bundle.
trajectory_figure	Show every timed transaction path without Seaborn aggregation or bootstrap.

1.2.1 latency_figure

reporting.m5_benchmark.latency_figure(cells)

Plot receipt-supplied descriptive quantiles without treating them as intervals.

1.2.2 load_bundle

reporting.m5_benchmark.load_bundle(input_root)

Validate the frozen receipt bundle and expose its 180 timed observations.

1.2.3 main

reporting.m5_benchmark.main()

Render the frozen M5 benchmark bundle without executing a benchmark.

1.2.4 render

reporting.m5_benchmark.render(input_root, output_root, *, readme=None)

Render an HTML report and sealed SVG assets from an admitted frozen bundle.

1.2.5 trajectory_figure

reporting.m5_benchmark.trajectory_figure(observations)

Show every timed transaction path without Seaborn aggregation or bootstrap.

# reporting.m5_benchmark { #aria_nbv.reporting.m5_benchmark } `reporting.m5_benchmark` Render a sealed, descriptive M5-versus-CUDA benchmark report. The renderer admits six already-produced benchmark receipts and their timing table. It neither constructs models nor replays tensors: its only scientific job is to preserve the measured transaction, timing unit, and limits in a portable HTML/SVG report. ## Classes | Name | Description | | --- | --- | | [BenchmarkBundle](#aria_nbv.reporting.m5_benchmark.BenchmarkBundle) | Validated input evidence retained by the static report renderer. | | [BenchmarkCell](#aria_nbv.reporting.m5_benchmark.BenchmarkCell) | One admitted backend/workload receipt with its descriptive timing summary. | ### BenchmarkBundle { #aria_nbv.reporting.m5_benchmark.BenchmarkBundle } ```python reporting.m5_benchmark.BenchmarkBundle( input_root, cells, observations, summary, evl_manifest, input_hashes, ) ``` Validated input evidence retained by the static report renderer. ### BenchmarkCell { #aria_nbv.reporting.m5_benchmark.BenchmarkCell } ```python reporting.m5_benchmark.BenchmarkCell(workload, backend, receipt, receipt_path) ``` One admitted backend/workload receipt with its descriptive timing summary. ## Functions | Name | Description | | --- | --- | | [latency_figure](#aria_nbv.reporting.m5_benchmark.latency_figure) | Plot receipt-supplied descriptive quantiles without treating them as intervals. | | [load_bundle](#aria_nbv.reporting.m5_benchmark.load_bundle) | Validate the frozen receipt bundle and expose its 180 timed observations. | | [main](#aria_nbv.reporting.m5_benchmark.main) | Render the frozen M5 benchmark bundle without executing a benchmark. | | [render](#aria_nbv.reporting.m5_benchmark.render) | Render an HTML report and sealed SVG assets from an admitted frozen bundle. | | [trajectory_figure](#aria_nbv.reporting.m5_benchmark.trajectory_figure) | Show every timed transaction path without Seaborn aggregation or bootstrap. | ### latency_figure { #aria_nbv.reporting.m5_benchmark.latency_figure } ```python reporting.m5_benchmark.latency_figure(cells) ``` Plot receipt-supplied descriptive quantiles without treating them as intervals. ### load_bundle { #aria_nbv.reporting.m5_benchmark.load_bundle } ```python reporting.m5_benchmark.load_bundle(input_root) ``` Validate the frozen receipt bundle and expose its 180 timed observations. ### main { #aria_nbv.reporting.m5_benchmark.main } ```python reporting.m5_benchmark.main() ``` Render the frozen M5 benchmark bundle without executing a benchmark. ### render { #aria_nbv.reporting.m5_benchmark.render } ```python reporting.m5_benchmark.render(input_root, output_root, *, readme=None) ``` Render an HTML report and sealed SVG assets from an admitted frozen bundle. ### trajectory_figure { #aria_nbv.reporting.m5_benchmark.trajectory_figure } ```python reporting.m5_benchmark.trajectory_figure(observations) ``` Show every timed transaction path without Seaborn aggregation or bootstrap.