Re: Can optimization-specific codes exist ? (A Johnstone)
4 Oct 1999 12:08:21 -0400

          From comp.compilers

Related articles
Can optimization-specific codes exist ? (Chango Cho) (1999-09-28)
Re: Can optimization-specific codes exist ? (1999-10-03)
Re: Can optimization-specific codes exist ? (1999-10-04)
Re: Can optimization-specific codes exist ? (1999-10-04)
Re: Can optimization-specific codes exist ? (Tom Lane) (1999-10-06)
Re: Can optimization-specific codes exist ? (1999-10-06)
| List of all articles for this month |

From: (A Johnstone)
Newsgroups: comp.compilers
Date: 4 Oct 1999 12:08:21 -0400
Organization: Royal Holloway, University of London
References: 99-09-118 99-10-011
Keywords: architecture, comment

: [I believe that by "codes" he meant parts of programs, not computer op-codes.
: In any event, the only PDP I can think of that was user microprogrammable
: was the not terribly popular PDP-11/60 and as far as I can tell, the only
: thing that people did with it was some DSP stuff. -John]

A bit off topic, but John is right: only the PDP 11/60 had the
Writeble Control Store option and not many of those were sold. Most
other PDP's were microcoded, but the microcode was not accesible to
the user. Most of the UK shipped machines with the option were used
for a military simulator.

As it happens, I am the Chair of the DEC working party for the UK
Computer Conservation Society a joint effort between the London
Science Museum, the British Computer Society and the Manchester Museum
of Science and Industry. If anybody can tell me where I can find a
PDP11/60 with Writable Control Store that could be donated to the
collection, I'd be very pleased.

Other machines with WCS options include the original VAX 11/780
(again, not popular) and the ICL PERQ aka Three Rivers workstation, an
early incarnation of the window/icon/mouse/pointer concept.

I had a lot of fun building microcoded custom processors when I was a
grad student. I've never really bought the idea that RISC is just
visible microcode. I think you can make a far better case for saying
RISC is what you get if you want something that pipelines easily: that
gives you register-register operations which immediately begets the
need for lots of registers and a three address architecture which
pretty much defines RISC for me. I always thought they should be
called Reduced Addressing Mode Computers, but maybe RAMC isn't as sexy
an acronym as RISC.

VLIW's begin to look like horizontal microcode, but that was
the idea behind their design, after all!
Dr Adrian Johnstone, Senior Lecturer in Computing, Computer Science
Dep, Royal Holloway, University of London, Egham, Surrey, TW20 0EX,
England. Email Tel:+44(0)1784 443425
Fax:+44(0)1784 439786
[Fascinating though this is, it's strayed too far from compilers. -John]

Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.