Alon Zakai
Alon Zakai
Hmm, such changes shouldn't appear when merging in latest upstream. How did you merge? Perhaps something went wrong there.
I believe the backend is fully functional and used in production. @tlively would know best, and about the status of documentation for it.
> Late-binding symbols get a JS stub import that resolves the symbol and then makes an onward call. This breaks JSPI. I wonder if another approach here could be to...
Looks like most of the time is taken up in `inlining-optimizing`, which is not surprising for a large file with many inlining and followup opportunities. But I'm afraid it doesn't...
Hmm, there is no good way, but you can hack the code like this, to remove the part that calls `wasm-opt -O3` ```diff diff --git a/tools/link.py b/tools/link.py index 0dc2dba0d1..9b8bcaa261 100644...
Yes, limitations of local tracking hit us here: we don't see all the contents in the local, just part of the pattern we are looking for. I think, however, that...
It wouldn't fix that particular issue - the problem there wasn't an inner expression with effects like the load in the last example. But it would fix this, and iirc...
IIUC you are proposing that a `WebGPU.wasm` be created, which contains calls in some high-level format to the WebGPU API. That could be linked against a JS module that calls...
A wasm module could import another's exports, and that can work with low level exports in theory. So yes, someone could write a wasm file B that replaces the JS...
fyi, this does reduce, but you need `--text` in the reduction.