Skip to content

BENCHMARK.md

Latest commit

 

History

History
130 lines (89 loc) · 3.56 KB

BENCHMARK.md

File metadata and controls

130 lines (89 loc) · 3.56 KB

Benchmark Report — sixcy v2.0.0

Overview

This document contains benchmark results for the sixcy archive tool. Results are organized by category and clearly labeled with their source version.

Note: This document contains historical data from multiple versions. The primary comparison data below was collected under v0.3.0.

1. Test Environment

Item Value
System Alienware m17 R5 AMD
CPU AMD Ryzen 9 6900HX @ 3.3 GHz
Physical cores 8
Logical processors 16
RAM 16 GB
OS Windows 11

2. sixcy vs 7-Zip Comparison [v0.3.0 Data]

Configuration

Tool Version Codec Notes
sixcy 0.3.0 LZMA Pure Rust via lzma-rs 0.3
7-Zip (not recorded) LZMA2 Level 1, single-threaded

Test file: 10 GiB synthetic binary (highly compressible, repetitive data)

Results

Metric sixcy (LZMA) 7-Zip (LZMA2) Winner
Pack time 13.0 s 34.6 s sixcy (2.7x faster)
Unpack time 47.0 s 8.6 s 7-Zip (5.5x faster)
Archive size 960 KiB 1,527 KiB sixcy (37% smaller)
Pack throughput 0.77 GiB/s 0.29 GiB/s sixcy
Unpack throughput 0.21 GiB/s 1.16 GiB/s 7-Zip

Notes

  • sixcy's faster pack time reflects lzma-rs's efficient implementation for compression.
  • 7-Zip's faster unpack time reflects its mature, hand-optimized C++ decompressor.
  • The smaller archive size for sixcy is due to CAS deduplication of identical chunks.
  • All runs validated (unpacked output matches original).

LZMA Decompression Performance

The lzma-rs 0.3 decompressor is approximately 5-6x slower than 7-Zip's implementation. This is a known characteristic of pure-Rust implementations vs hand-optimized C++. This affects read-heavy workloads but not correctness.

A future release may offer an optional liblzma backend for the decompression path.

3. Compression Ratio on Technical Binary

Test Data [Current - v2.0.0]

Test file: target/release/6cy.exe (Rust release binary)

Codec Level Size Ratio
Original - 3,100,160 B -
Zstd 3 1,420,056 B 2.18:1
LZMA 1 1,889,035 B 1.64:1

For technical binaries, Zstd (level 3, the default) provides better ratio than LZMA at level 1.

4. Interpretation Caveats

  1. Synthetic test data produces extreme compression ratios (≥10,000:1) that do not reflect real-world workloads.

  2. Single-threaded comparison isolates codec efficiency but does not reflect multi-threaded scenarios where 7-Zip with -mmt=on would be significantly faster.

  3. LZMA vs LZMA2 — sixcy uses LZMA (lzma-rs), while 7-Zip uses LZMA2. These are related but different algorithms.

  4. CPU measurements are approximate due to 0.5s sampling interval.

  5. Disk cache effects may influence timings on repeated runs with the same test file.

5. Running Benchmarks

# Build release binary
cargo build --release

# Run RLE pre-filter benchmark on a file
6cy bench your_file.dat

# Compare compression methods
time 6cy pack -o zstd.6cy -i data/ --codec zstd
time 6cy pack -o lzma.6cy -i data/ --codec lzma

6. Reproduction

To reproduce these benchmarks:

  1. Ensure Windows 11 x64 with 16+ GB RAM
  2. Install Rust toolchain (stable 1.70+)
  3. Clone repository and build: cargo build --release
  4. Generate synthetic test data (if needed)
  5. Run comparison tests with timing

Note: Actual results may vary based on system load, disk speed, and other factors.