Skip to content

README.md

Latest commit

 

History

History
82 lines (61 loc) · 2.83 KB

README.md

File metadata and controls

82 lines (61 loc) · 2.83 KB

Scripts Overview

This directory contains repository-level automation helpers. Experiment execution entrypoints (runner.sh, run-one.sh, plot.py) are kept under experiments/ by design.

Available Scripts

../setup.sh (repo root)

Primary one-shot entrypoint for first-time setup and rebuilds. It wraps dependency checks, environment generation, simulator builds, and benchmark builds.

./setup.sh
# or
./setup.sh --rebuild

setup-env.sh

Generates the .zsim-env file at the repository root. The three memory-simulator paths (DRAMSIM3PATH, RAMULATORPATH, RAMULATOR2PATH) are resolved automatically from the repo. PINPATH and HDF5_HOME are discovered from common system prefixes; if missing, the script attempts to download known-good bundles into dependencies/.

./scripts/setup-env.sh
source .zsim-env

This must be run once before building ZSim or running any experiment. See simulator-source/README.md for dependency details.

build-benchmarks.sh

Compiles the shared ptr_chase and traffic_gen benchmarks under benchmarks/. This must be run before any experiment can execute.

This script is intended for the Linux environment used in the paper artifact. In particular, ptr_chase depends on linux/perf_event.h.

./scripts/build-benchmarks.sh

compare-results.sh

Compares two experiment stages by analyzing their processed/bandwidth_latency.csv outputs side by side. Useful for quantifying the performance delta introduced by each interface correction.

./scripts/compare-results.sh <stage-a> <stage-b>
# Example:
./scripts/compare-results.sh 01-baseline 04-model-correct

download-raw.sh

Downloads the externally hosted raw artifacts for a given stage when URLs are published in the stage's raw-manifest.csv. See the Raw Data Policy for details on what is hosted externally.

./scripts/download-raw.sh <stage>
# Example:
./scripts/download-raw.sh 01-baseline

Internal helper:

  • lib/compare_results.py — implementation used by compare-results.sh

Typical Workflow

First time — build everything:

./setup.sh

Run a single stage and plot it:

source .zsim-env
./experiments/runner.sh 01-baseline
./experiments/plot.py experiments/01-baseline/test-raw \
  --config-dir experiments/01-baseline
# → figures land in test-output/01-baseline/figures/

Compare two stages (works against committed CSVs or freshly generated ones):

./scripts/compare-results.sh 01-baseline 04-model-correct

Download published raw artifacts for one stage (if manifest URLs are final):

./scripts/download-raw.sh 01-baseline

-> For details on runner.sh, run-one.sh, and the plotting pipeline see experiments/README.md.