Base64 Decoding Benchmark

This project benchmarks the performance of Uint8Array.fromBase64 for decoding base64-encoded data using different datasets.

Datasets

• Images: Single base64-encoded image files (bing.txt, googlelogo.txt).
• DNS: Multiple base64-encoded lines from swedenzonebase.txt, decoded individually.
• Emails: Multiple base64-encoded email files, decoded individually and concatenated.

Requirements

• Node.js (version 25.5.0 or later for native Uint8Array.fromBase64 support) or Bun.
• The base64 data files in the base64data/ directory.

Running the Benchmark

With Node.js

node benchmark.js

To reduce GC jitter, use:

node --max-old-space-size=1000 --max-semi-space-size=512 --noconcurrent_sweeping benchmark.js

With Bun

bun run benchmark.js

The script will load all data at startup, then run 100 iterations of each benchmark, reporting average, min, and max times, as well as throughput in GB/s (where 1 GB = 1000³ bytes).

Output Example

Loading data...
Data loaded.
Running benchmark: Decode bing.txt
  Min throughput: 1.500 GB/s
  Average throughput: 1.645 GB/s
  Max throughput: 2.000 GB/s

...

Notes

• If Uint8Array.fromBase64 is not available, it falls back to Buffer.from with base64 decoding.
• Benchmarks measure decoding performance only; data loading is done upfront.