target_family="wasm" and target feature
A relaxed dot-product instruction.
This instruction will perform pairwise products of the 8-bit values in
b and then accumulate adjacent pairs into 16-bit results producing a
i16x8 vector. The bytes of
a are always interpreted as signed and
the bytes in
b may be interpreted as signed or unsigned. If the top bit in
b isn’t set then the value is the same regardless of whether it’s signed
The accumulation into 16-bit values may be saturated on some platforms, and on other platforms it may wrap-around on overflow.