Path: utzoo!mnetor!uunet!tektronix!sequent!mntgfx!mbutts
From: mbutts@mntgfx.mentor.com (Mike Butts)
Newsgroups: comp.arch
Subject: Re: Performance increase - a suggestion
Message-ID: <1988Feb8.121200.370@mntgfx.mentor.com>
Date: 8 Feb 88 20:11:55 GMT
References: <235@unicom.UUCP> <28200088@ccvaxa> <3536@batcomputer.tn.cornell.edu> <231@m2.mfci.UUCP>
Organization: Mentor Graphics Corporation, Beaverton Oregon
Lines: 49
Keywords: VLIW
Summary: FPS architecture is *not* microcoded

In article <231@m2.mfci.UUCP>, root@mfci.UUCP (SuperUser) writes:
> In article <3536@batcomputer.tn.cornell.edu> kahn@tcgould.tn.cornell.edu (Shahin Kahn) writes:
> >VLIW has been part of Floating Point Systems' machines for more than 
> >10 years.
> 
> Under that definition of VLIW, every microcoded machine ever made
> qualifies.  It's never been that big a trick to make a machine with
> many parallel functional units.  The trick is to be able to keep them
> busy without hand coding, and to provide the right interface so that
> the compiler can express the parallelism that it finds.
> 
> Read Ellis' thesis;  he presents reasonable arguments as to why FPS'
> architectures have hot spots that make it a very difficult target
> for a compiler.
> 

1) Contrary to popular impression, FPS has been compiling directly from
FORTRAN to their 64-bit "LIW" 164 machine since the early '80s.  The code is
of very high quality, and schedules operations onto the multi-field
instructions across a whole basic block.  For example, the memory system
in the FPS 164 is pipelined, and software accounts for the pipeline length,
not a hardware scoreboarder.  The FPS compiler made full use of this pipeline,
scheduling appropriately.  It also has a very effective loop pipeliner.
In fact, I believe Josh Fisher & Co. used 164 hardware for their early work at 
Yale that grew into Multiflow.  

I suppose it's just a consequence of FPS' non-existent paper publishing and 
lack of marketing that people generally still think FPS machines are hand-coded 
array processors.  There is no code between the FORTRAN and the 64-bit LIW, 
and the LIW fields are encoded and at a minicomputer level of expression, 
thus it is not microcode.

Mentor's Compute Engine is also a 64-bit LIW machine, oriented more towards
EDA-type codes, with optimizing, scheduling C, FORTRAN and Pascal compilers that 
make effective use of the machine.  (For historical purposes only - 
not a commercial ;-) )

Don't get me wrong, I have a high opinion of the way Multiflow is developing
VLIW technology, I just feel the early FPS people (I wasn't one of them) should
get more credit for blazing this trail 5-10 years ago.

2) Please give us a reference for Ellis' thesis.


-- 
Mike Butts, Research Engineer         KC7IT           503-626-1302
Mentor Graphics Corp., 8500 SW Creekside Place, Beaverton OR 97005
...!{sequent,tessi,apollo}!mntgfx!mbutts OR  mbutts@pdx.MENTOR.COM
These are my opinions, & not necessarily those of Mentor Graphics.