Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!rutgers!apple!mips!mash
From: mash@mips.COM (John Mashey)
Newsgroups: comp.arch
Subject: Re: Fast conversions, another urban myth?
Keywords: BCD, radix-conversion, COBOL
Message-ID: <27768@winchester.mips.COM>
Date: 18 Sep 89 17:13:01 GMT
References: <832@dms.UUCP> <11488@burdvax.PRC.Unisys.COM>
Reply-To: mash@mips.COM (John Mashey)
Organization: MIPS Computer Systems, Inc.
Lines: 42

In article <11488@burdvax.PRC.Unisys.COM> barry@PRC.Unisys.COM (Barry Traylor) writes:

>I'm afraid I can't state specifics, but the large A Series stuff (A12/15/17)
>does it with a fair amount of silicon and not very much microcode.  There
>is a raging and recurring debate between the A Series types (including
>myself) and the V Series types (V Series stuff is DECIMAL, including
>addressing and peripherals, at least historically) about decimal arithmetic
>vs convert in / binary ops / convert out.  All I can say is that with the
>proper amount of silicon, OUR method beats THEIR method.  

It's good to see this topic discussed a bit more.  Can you share any
info, like frequencies of operations, i.e., info that isn't
implementation-secret?

>2 (I think) BCD convert out ops.  Pretty reduced, I'd say.  And this is an
>instruction set that has been around in pretty much the same form for 20
>years.

Although I think RISC is a much better tradeoff, for various reasons,
and the A-series is fairly far in the other direction (it's not the
number of operations, it's more the complexity of addressing modes,
and other issues), I've long been an admirer of the philosophy behind
the B-5000 (the long-ago ancestor of the A-series).  At least it was the
first major computer line I can think of that really showed a strong
vision of designing the hardware and software closely together.
Also, for the early 60s, it was amazingly far ahead of its time in many ways.
At least a brief look at this architecture is well worth the time
for anyone interested in computer architecture, even if it's not the
direction of most current machines.

As a related topic, wearing my SPEC hat, does anybody out there have
GOOD COBOL benchmarks that have a chance of following SPEC rules:
	1) >= 60 seconds on 8-mips machine, preferably more.
	2) Public-domain, or able to be widely distributable in source form.
	3) Reasonably portable, or be able to made portable
	4) Correct exeuction can be verified mechanically in some way.
I REALLY want to get some such COBOL programs.
-- 
-john mashey	DISCLAIMER: <generic disclaimer, I speak for me only, etc>
UUCP: 	{ames,decwrl,prls,pyramid}!mips!mash  OR  mash@mips.com
DDD:  	408-991-0253 or 408-720-1700, x253
USPS: 	MIPS Computer Systems, 930 E. Arques, Sunnyvale, CA 94086