home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!darwin.sura.net!mips!sdd.hp.com!usc!rpi!utcsri!eecg.toronto.edu!colin
- Newsgroups: comp.sys.amiga.programmer
- From: colin@eecg.toronto.edu (Colin Plumb)
- Subject: Re: CPU cycles
- Message-ID: <1992Jul29.225850.7496@jarvis.csri.toronto.edu>
- Organization: University of Toronto
- References: <jbickers.0bpc@templar.actrix.gen.nz>
- Date: 30 Jul 92 02:58:50 GMT
- Lines: 21
-
- In article <jbickers.0bpc@templar.actrix.gen.nz> jbickers@templar.actrix.gen.nz (John Bickers) writes:
- > I'm having some difficulty working out the relationship between
- > CPU cycles and clock speed. I want to execute a routine at
- > something like 4000 times a second on a 25MHz 68030, so I've tried
- > to figure out how many cycles I'm allowed in the routine (by this
- > I mean CPU instruction cycles, as given in the back of an
- > assembler manual - ie: MOVEQ takes 4 cycles, etc).
-
- Those timings are for a 68000. An '030 or an '040 take fewer cycles per
- instruction as well as cycling faster (especially for long instructions!).
-
- Also, did you remember to add fetch effective address time, and did you
- account for the number of wait states in memory accesses? If
- instructions will fit into the cache, it'll run full-speed, but if it
- won't, there's that fetch, plus the data accesses (there is a small data
- cache, but it's not hard to blow.)
-
- And over a 7.15909 MHz 16-bit bus, memory access takes a *long* time as far
- as a 25 MHz, 32-bit 68030 is concerned...
- --
- -Colin
-