Path: utzoo!mnetor!tmsoft!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!malgudi!sunc.osc.edu!xipe.osc.edu!djh
From: djh@xipe.osc.edu (David Heisterberg)
Newsgroups: comp.unix.cray
Subject: Re: Cray YMP instruction timings?
Message-ID: <1247@sunc.osc.edu>
Date: 18 Dec 90 14:21:27 GMT
References: <1212@red.qtp.ufl.edu>
Sender: news@osc.edu
Organization: Ohio SuperComputer Center, Columbus, OH, USA
Lines: 32
Nntp-Posting-Host: xipe.osc.edu

In article <1212@red.qtp.ufl.edu> bernhold@qtp.ufl.edu (David E. Bernholdt) writes:
>1) How long does it take (clock cycles) to execute each for the
>following in scalar mode on a YMP?
>	integer compare
>	integer addition
>	integer multiplication
>	single precision multiplication

There's no integer compare, just use subtract.  48 bit integer
multiplication is done in the floating multiplier (no conversion penalty,
as there is for reciprocal).

The following is from the Symbolic Machine Instructions Reference (SR-0085B):

address add/sub   (32 bit)       2    12ns
address multiply  (32 bit)       5    30ns

scalar/vector integer add/sub    3/3  18/18ns
scalar/vector logical            1/2   6/12ns
scalar/vector shift              3/3  18/18ns
scalar/vector population/parity  7/4  42/24ns
scalar leading zero              6    36ns

floating add/sub                 7  [ 6 for XMP!]  42ns
floating multiply                8  [ 7 for XMP!]  48ns
floating reciprocal             15  [14 for XMP!]  90ns

Dave, you can probably order SR-0085 from OSC.
--
David J. Heisterberg		djh@osc.edu		And you all know
The Ohio Supercomputer Center	djh@ohstpy.bitnet	security Is mortals'
Columbus, Ohio  43212		ohstpy::djh		chiefest enemy.