Benchmark encryption & combination (TypeScript via WASM bindings) and partial decryption (Python bindings)

[cygnusv]

No description provided.

[piotr-roslaniec]

Initial benchmark
Out of date - Please see WASM and Python benchmarks below.

• No changes to existing TPKE benchmarks
• Running benchmarks in wasi rather than in the browser. Because of that, these results are not indicative of performance in the browser.
• Raw results
• TLDR of the results: WASM is ~5x slower than Rust implementation
• More detailed benchmarks are coming after proper WASM bindings are shipped WASM bindings #10 Done

[piotr-roslaniec]

WASM bindings benchmark

• Results available here
• WASM in the browser is 14x slower than Rust on desktop
• The worst-case scenario performance-wise is where we use 128 shares: Time to combine these shares is in the ballpark of a few seconds (2.2s in the benchmark)
• TLDR: Combining 128 shares takes up to ~2-3s in the browser

Relevant PR: #17

[piotr-roslaniec]

Python bindings benchmark

• Results available here
• Notice: The code used in the benchmark doesn't perform public verification yet. See this question about "simple vs fast" decryption if it impacts how verification is performed. Edit: After some light reading of the Ferveo source code I concluded that it's not necessary to benchmark this verification.
• TLDR: Partial decryption of 128 shares takes up to 25 ms in Python

Relevant PR: #18

[piotr-roslaniec]

Benchmarks are done. Please reopen should we need to update the benchmark suite.