lukewagner
commented
1 year ago I suppose #181 could be considered a (generalized) solution to this (using, e.g., [u8;16] or [f32;4]), with tuple being a short-term workaround. The Canonical ABI would layout a list of such tuples contiguously in linear memory so that core wasm could use v128.load/v128.store directly on the incoming/outgoing lists of vectors.
why component-level valtype doesn't have an equivalent of
v128? if it's intentionally excluded, i guess it warrants a note in the explainer.