home *** CD-ROM | disk | FTP | other *** search
- Path: mordred.cc.jyu.fi!news
- From: daeron@horus.co.jyu.fi (Aki Laukkanen)
- Newsgroups: comp.sys.amiga.misc
- Subject: Re: Speed: 68040 vs. 68060
- Date: 3 Mar 1996 14:45:22 GMT
- Organization: University of Jyvaskyla, Finland
- Message-ID: <816.6636T687T1238@horus.co.jyu.fi>
- References: <371.6633T989T2700@horus.co.jyu.fi> <1195.6634T1430T809@teclink.net>
- NNTP-Posting-Host: dynamic07.co.jyu.fi
- X-Newsreader: THOR 2.22 (Amiga;TCP/IP) *UNREGISTERED*
-
-
- >Ok, just to clear things up. Your both wrong. The 68060's limit is 2
- >integer instructions + one branch instruction per clock or 1 integer, 1 float
- >and 1 branch instruction per clock. That's 150MIPS peak @ 50MHz. Realistic
-
- Interesting, can a 68060 do simple float operations in one cycle
- (fadd,fsub..)? Then it has gone a long way since 882/040FPU.
-
- >values due to dependency conflicts, branch mispredictions, cache misses, etc.
- >are about half that. (Ref. M68060/D p. 2) (I suppose the original post could
- >have meant 250 VAX MIPS which is a different measure all together.)
-
- And usually compilers don't tend to spread branch and floating point
- instructions around integer instructions just to get peak performance from the
- processor. ;)
-
- --
- Daeron
-
- There is nothing like an ice cream cone in your armpit.
-
-