py2wasm performance is the same as CPython 3.14

[mkokryashkin]

So I've been experimenting with WAMR (wasm-micro-runtime) for a while now, and it has a great feature -- an AOT-compiler for WASM binaries. This compiler tends to speed up the execution of wasm modules orders of magnitude for some scenarions, while still offering all the benefits of WASM/WASI.

With that said, here are the pystone benchmark results for native Python, AOT wasm32-wasi CPython, and AOT py2wasm:

Native Python

Pystone(1.1) time for 50000 passes = 0.131406
This machine benchmarks at 380500 pystones/second

AOT wasm32-wasi CPython

Pystone(1.1) time for 50000 passes = 0.201963
This machine benchmarks at 247570 pystones/second

AOT py2wasm

Pystone(1.1) time for 50000 passes = 0.215168
This machine benchmarks at 232377 pystones/second

It is clearly visible that conventional CPython is faster than py2wasm, although only by a fraction. Moreover, it is consistently better over several runs.

This raises a question of whether py2wasm is a viable option, or the community is better of focusing on performance of the official CPython wasm32-wasi target, rather than on the development of py2wasm.