home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.atari.st.tech
- Path: sparky!uunet!email!nino
- From: nino@vmars.tuwien.ac.at (Marinos Yannikos)
- Subject: Re: 16-bit RAM Access on Falcon ?
- Message-ID: <1992Nov11.110532.16455@email.tuwien.ac.at>
- Sender: news@email.tuwien.ac.at
- Nntp-Posting-Host: quasi.vmars.tuwien.ac.at
- Organization: TU Wien, Vienna, Austria
- References: <1992Nov8.132805.1903@rhrk.uni-kl.de>
- Distribution: eunet
- Date: Wed, 11 Nov 1992 11:05:32 GMT
- Lines: 23
-
- peter@rhrk.uni-kl.de (Gunter Bitz) writes:
- :
- : I don't know what Qindex does, but I testet the following:
- : A short loop (which fits completely in the processor cache) like that:
- : ll:
- : move.w d0,(a0)+
- : dbf d7,ll
- : and another like
- : ll:
- : move.l d0,(a0)+
- : dbf d7,ll
- :
-
- Was the data in a0 longwqord-aligned in the second case? If not, the result
- is not surprising, since it takes 2 accesses to load a longword which is
- not longword-aligned even on full 32-bit machines!
-
- : The result was that filling the same amount of memory with longwords is a bit
- : faster than filling with words, but not twice as fast as it should be on a
- : r e a l 32 bit machine. I'm sorry about that
- :[...]
- : Gunter
- :
-