anyone had experience where assembly is not actually faster than optimized js when it’s clean like for example when working on preallocated memory in buffer because v8 optimizer is that good?