Benjamin Herrenschmidt writes:> Do we have any indication that it performs better than the C one ?I would expect it to, given that the assembler one has two branches inthe per-byte loop compared to 3 in the C version.Paul.