[net.sources] BCOPY and Hardware Hacking

alan@apollo.UUCP (Alan Lehotsky) (10/14/83)

Concerning the suggested change

>  In the VAX bcopy function posted, it would probably be faster if the
>  constant 65535 was replaced by 65532 wherever it occurs, thus causing
>  all memory references to be longword references if the source and destination
>  are longword-aligned.

I'm fairly certain that the micro-code for VAXEN already does this
for you.  As a former DIGIT (DEC employee), I know that a lot of work
was done to make string moves as fast as possible, and all sorts of
tricks such as getting to quad-word alignment and doing double-bus transfers
over the SBI. [I don't have my hardware manuals handy, so....]