Steven Allen
Steven Allen
Specifically, we're looking to correctly set the `memory_access_cost` gas fee: https://github.com/filecoin-project/ref-fvm/blob/280d80503f950a1934e3d60910d659fad685f9c7/fvm/src/gas/price_list.rs#L517. We currently don't charge any additional gas here but we likely will need to do so if/when we support...
On-top-of #484.
> First thought is to check for memory leaks, as the valgrind command has been removed from CI, but seems reasonable otherwise Looks like I'm going to need a container...
We strip debug info (via cargo). I've tried running wasm-opt and didn't notice a huge difference in _size_ but we should see if we can get a performance uplift. In...
Ok, I looked at `wizer` and it's only for running an initialization function then "dumping" the resulting memory back to the wasm file. We could use this to initialize lazy...
My next step is to try to update FVM v2 in forest to see what deps get updated along with it.
We've confirmed that having wasmtime v24 anywhere in the build tree, even if unused, causes this slowdown.
- It's not the `write_core` feature in `object` as far as I can tell. - I'm also seeing some changes in gimli, but v24 uses a different version of that...
I've tracked it down to wasmtime's cranelift feature (does not appear to be feature unification but I may be missing something). I'm now trying with just `wasmtime-cranelift`.
- wasmtime-environ doesn't reproduce. - cranelift-wasm does.