TestPcForProgramers

Python 3BashNumbaNumPymatplotlibmultiprocessingMIT

GitHub: stevenvo780/TestPcForProgramers

---

What it is

TestPcForProgramers is a hardware benchmark suite aimed at developers who want to stress-test and characterise their machines using workloads that reflect real programming tasks: scientific simulation, matrix multiplication, memory throughput, disk I/O, and kernel compilation.

The suite has two layers:

Layer	What it is
N-body gravitational simulator (nbody_simulation.py)	The main CPU stress tool: O(n²) pairwise gravity simulation with Numba JIT, Python multiprocessing, and NumPy BLAS. Real-time matplotlib animation.
Benchmark suite (run_bench_suite.sh + 4 scripts)	Four focused micro-benchmarks with consistent JSON output and history files for longitudinal comparison.

---

N-body gravitational simulator

The flagship tool simulates N-body gravitational interactions under full pairwise O(n²) forces. It is designed to load every CPU core and measure sustained floating-point throughput under a realistic scientific workload.

Parallelisation options (select at runtime):

Mode	Implementation	Typical speedup
Numba JIT	@jit(nopython=True, parallel=True) — auto-parallelises inner loop	3–8× vs pure Python
Python multiprocessing	multiprocessing.Pool — splits body chunks across workers	scales with core count
NumPy / BLAS	vectorised broadcasting — fast on systems with tuned BLAS	good single-core baseline

Integrators:

Integrator	Use case
Runge-Kutta 4	Higher precision, lower energy drift
Euler	Faster iteration, useful for pure throughput benchmarking

Built-in scenarios:

Scenario	Bodies	Purpose
solar	~25	Quick sanity check; short run time
galaxy	200–500	Sustained mid-tier load
benchmark	1000+	Full CPU stress; measures peak parallel throughput

Real-time visualisation: matplotlib animation with movement trails. Disable with --no-viz to maximise CPU allocation to simulation.

---

Benchmark suite

Four focused scripts invoked by run_bench_suite.sh. Each prints a BENCH_JSON: {...} line on stdout and appends results to a JSON history file, enabling longitudinal tracking across hardware changes, kernel updates, or compiler upgrades.

Script	What it measures	History file
kernel-build-bench.sh	Linux kernel compile time (wall clock, user, sys)	kernel-build-history.json
cpu_matmul_bench.py	GEMM peak throughput in GFLOPS (NumPy)	cpu-matmul-history.json
memory_bandwidth_bench.py	Memory-to-memory copy bandwidth in GB/s	memory-bandwidth-history.json
disk_io_bench.py	Sequential write + read throughput in MiB/s	disk-io-history.json

Run all four in sequence:

bash run_bench_suite.sh

---

Installation

N-body simulator

git clone https://github.com/stevenvo780/TestPcForProgramers.git
cd TestPcForProgramers
pip install -r requirements.txt
# requirements: numpy>=1.21.0, matplotlib>=3.5.0, psutil>=5.8.0, numba>=0.56.0

python3 nbody_simulation.py

Benchmark suite

git clone https://github.com/stevenvo780/TestPcForProgramers.git
cd TestPcForProgramers
pip install numpy psutil

# Run full suite
bash run_bench_suite.sh

# Or run individual benchmarks
python3 cpu_matmul_bench.py
python3 memory_bandwidth_bench.py
python3 disk_io_bench.py
bash kernel-build-bench.sh

Virtual environment

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python3 nbody_simulation.py

Numba first-run compilation

The first run with Numba JIT will be slower than subsequent runs — Numba compiles and caches the kernel on the first invocation. Use the benchmark scenario for the first run to warm the cache, then re-run for accurate numbers.

Kernel build benchmark

kernel-build-bench.sh downloads and compiles the Linux kernel source. It requires ~1 GB of disk space and several GB of RAM. Build time on a modern 8-core machine is roughly 10–20 minutes.

---

Usage

N-body simulator flags:

python3 nbody_simulation.py --scenario galaxy --method numba --integrator rk4
python3 nbody_simulation.py --scenario benchmark --method multiprocessing --no-viz

Individual suite scripts:

python3 cpu_matmul_bench.py          # prints GFLOPS + appends to cpu-matmul-history.json
python3 memory_bandwidth_bench.py    # prints GB/s  + appends to memory-bandwidth-history.json
python3 disk_io_bench.py             # prints MiB/s + appends to disk-io-history.json
bash kernel-build-bench.sh           # prints seconds + appends to kernel-build-history.json

---

Stack

Layer	Technology
N-body simulation	Python 3, Numba JIT (@jit nopython+parallel), NumPy BLAS, multiprocessing.Pool
Integration	Runge-Kutta 4 / Euler
Visualisation	matplotlib (real-time animation with trails)
System monitoring	psutil (CPU affinity, core count)
GEMM benchmark	NumPy (BLAS-backed matrix multiply)
Memory benchmark	NumPy array copy, timed with time.perf_counter
Disk benchmark	Python stdlib os + time, sequential file I/O
Kernel benchmark	Bash + make -j$(nproc)
History output	JSON (one record per run, append-only)