Hi,
On Tue, 12 Dec 2006, Michael Schmitz wrote:
I killed the movep stuff, I'm not sure it still is really a win (maybe at
68000 times it was, but on a 68060 it certainly isn't), some operations
should be even considerably faster now (e.g. odd<->even copy).
The kernel started out on 030, and Andreas wouldn't have written it if it
didn't speed things up there. For emulation and 060 we can probably drop
it. My only rationale for using it was to start from something that had
worked at some time in the past :-)
I don't see a big difference in speed in ARAnyM; need to test on the real
hardware for that.
I guess aranym prefers more compact code, so movep might be faster there.
Looking through the manuals it seems on a 030 a movep is sligtly faster
than four moveb, but on the 040 it's already the other around. If there
are also waitstates due to the fb access the small advantage might be
completely gone, so unless some wants to optimize this badly for a 030, I
guess I prefer the more generic version. :)
Large copy and draw operation are done now using 32bit accesses anyway, so
it would be only single column operations where movep could be useful.
bye, Roman
-
To unsubscribe from this list: send the line "unsubscribe linux-m68k" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
