Alon Zakai

Here, removing the useless blocks only requires `--remove-unused-brs --vacuum`. But the reason those are the right passes is just a coincidence - there was simply so much work after inlining...

Re-measuring this now, I see a 1% code size improvement with adding another `-Os` after the first, and only `0.03%` when just adding the two passes mentioned above that fix...

Yes, you are right, it looks like some of those situations need a little more, as before the binary lowering a `drop` exists which is dead code, but interferes with...

Measuring on Dart, I see a small benefit here. Probably it is small as other devirtualization etc. mechanisms already handle most things.

friendly nonurgent ping @tlively

Another way to profile binaries is `wasm-opt --metrics` and `wasm-opt --func-metrics`. You may also want to compare wasm-opt flags between what emcc sends and what you are doing manually. See...

The main benchmarks I've looked at for WasmGC all depend on imported JS strings (and other JS APIs, but strings is the hardest to replace). Is there some kind of...

@peteryongzhong That is open source, yes: basically just compile the [Java Fannkuch benchmark](https://benchmarksgame-team.pages.debian.net/benchmarksgame/program/fannkuchredux-java-8.html) with [J2Wasm](https://github.com/google/j2cl/blob/master/docs/getting-started-j2wasm.md). There are other popular benchmarks in this space that are available too (e.g. we used...

Interesting, that is a slow test, but it was never this slow. This may be affecting users so it seems worth bisecting to find out when this regressed, at least...

No significant change between tot and 3.1.60? Then maybe the change is somehow on CircleCI's timeout, or the ordering of the tests (the timeout is X seconds without output, so...