←
Wasm3616 posts
Wasm3 avatar

Wasm3

@wasm3_engine

🚀 A fast WebAssembly interpreter, and the most universal #wasm runtime. Maintained by @vshymanskyy

◷ Joined January 2020▧ 139 Photos and videos
Wasm3 avatar
Wasm3@wasm3_engine·Aug 24

Wasm3 v0.9.0 is out!

🎉Wasm3 v0.9.0 is out!🎉 It is the first major update in FIVE years. That sounds like a software story. It isn't. Four years ago, I wrote that my house had been destroyed by russians, and that I had to put #Wasm3 into minimal maintenance. The war never went away. 1/4

Post attachment
Wasm3 avatar
Wasm3@wasm3_engine·Aug 25

Tail-call/TCO interpreter approach got more attention afterward. @JoshHaberman 's “Parsing protobuf at 2+GB/s” was inspired by Wasm3, and that line of work helped developing new efficient interpreters. Including the Python's new tail-call interpreter: docs.python.org/3/whatsnew/3… 7/7

Wasm3 avatar
Wasm3@wasm3_engine·Aug 25

That work also helped push embedded-Wasm discussions such as custom page sizes. The standard 64 KiB Wasm page is reasonable on desktops and servers, but painful on small MCUs. Making that constraint visible to the broader spec community mattered. 6/7

Wasm3 avatar
Wasm3@wasm3_engine·Aug 25

But Wasm3’s main goal was never "win every interpreter benchmark." Its real focus was making WebAssembly practical on tiny MCUs and deeply embedded systems with severe RAM, flash, toolchain, and OS constraints. 5/7

Wasm3 avatar
Wasm3@wasm3_engine·Aug 25

That mattered especially for embedded, FaaS, and short-lived workloads—places where large runtimes simply don’t fit. WAMR is a good example: its later “fast interpreter” explicitly cites Wasm3’s architecture as inspiration. 4/7

Wasm3 avatar
Wasm3@wasm3_engine·Aug 25

Before Wasm3, interpreters were often treated as inherently slow or incomplete, while serious performance work focused on JIT/AOT. Wasm3 showed that an interpreter could be fast, tiny, portable, and have unprecedented cold-start behavior. 3/7

Wasm3 avatar
Wasm3@wasm3_engine·Aug 25

On the Wasm3 architecture: we did not invent threaded interpreters, accumulator-style state, or tail-call dispatch. These ideas are much older. What Wasm3 did help do was bring this architecture into the modern WebAssembly conversation. 2/7

Wasm3 avatar
Wasm3@wasm3_engine·Aug 25

⚠️Recently, several #WebAssembly interpreters have appeared with "faster than Wasm3" claims. Obviously - more compact, embeddable interpreters pushing performance forward is healthy for the whole software ecosystem. 1/7

Wasm3 avatar
Wasm3@wasm3_engine·Aug 20

Wasm3 was created in #Kyiv / #Ukraine. Russia attacks my city with drones and ballistic missiles every day. Ordinary people are being killed while trying to keep living, building, and shipping. Fuck russia.