home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!kithrup!hoptoad!pacbell.com!iggy.GW.Vitalink.COM!cs.widener.edu!eff!sol.ctr.columbia.edu!zaphod.mps.ohio-state.edu!cs.utexas.edu!sun-barr!olivea!charnel!sifon!thunder.mcrcim.mcgill.edu!snorkelwacker.mit.edu!bloom-picayune.mit.edu!athena.mit.edu!jfc
- From: jfc@athena.mit.edu (John F Carr)
- Newsgroups: comp.arch
- Subject: Re: DEC Alpha architecture issues
- Message-ID: <1992Nov22.165752.26572@athena.mit.edu>
- Date: 22 Nov 92 16:57:52 GMT
- References: <1992Nov18.210416.27212@nntpd.lkg.dec.com> <lgnfc3INNqnt@spim.mti.sgi.com> <By0xy0.FsA.2@cs.cmu.edu>
- Sender: news@athena.mit.edu (News system)
- Organization: Massachusetts Institute of Technology
- Lines: 16
- Nntp-Posting-Host: achates.mit.edu
-
- In article <By0xy0.FsA.2@cs.cmu.edu>
- lindsay+@cs.cmu.edu (Donald Lindsay) writes:
-
- >Also, PALcode can be entered from user state - you don't need to do a
- >system call, with its heavy-duty state-change overheads.
-
- A system call does not need to be slow. On an RS/6000, a system call can be
- as fast as 2 cycles. If there is a slow instruction like a multiply or
- divide executing the system call may overlap it and appear to take 0 cycles.
-
- The current Alpha implementation takes more cycles than the RS/6000 to
- switch modes, but that isn't an architectural issue: the 21064 pipeline is
- longer.
-
- --
- John Carr (jfc@athena.mit.edu)
-