>From BitGeek@mail.mac.com Sat Jan 22 15:33:32 2000
Received: from a.x.l334.3332.uk.msn.org ([207.55.63.101])
by roundtable.cif.rochester.edu (8.9.3/8.8.8) with ESMTP id PAA05618
for <wearable@cif.rochester.edu>; Sat, 22 Jan 2000 15:33:30 -0500 (EST)
(envelope-from BitGeek@mail.mac.com)
Received: from [10.0.0.2] (127.0.0.1) by a.x.l334.3332.uk.msn.org
with SMTP (Eudora Internet Mail Server 1.3.1); Sat, 22 Jan 2000 12:33:22 -0800
Subject: Re: Crusoe versus StrongArm, Super H, etc.
x-sender: BitGeek@mail.mac.com
x-mailer: Claris Emailer 2.0v3, January 22, 1998
From: BitGeek <BitGeek@mac.com>
To: "Vaughan Pratt" <pratt@CS.Stanford.EDU>, "Wear hard" <wear-hard@haven.org>,
<wearable@cif.rochester.edu>
Mime-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Message-ID: <1263578094-3451658@a.x.l334.3332.uk.msn.org>
On 1/22/00 11:34 AM, Vaughan Pratt said:
>Code morphing for other platforms is of course a possibility, but it
>is clear that the Crusoe processors have been tuned to perform well on
>x86 code.
I would say the "initial Crusoe processors have been tuned..." The two
processors announced have significant hardware differences (implied by
the white paper, anyway). I see no reason to believe that future
processors won't also have significant hardware differences.
An obvious one would be a version that could handle 3 Molecules
simultaneously. Since the code morphing processes goes over chunks of
code (instead of just one instruction at a time) it could easily
slice-and-dice to produce three molecules that can be processes
simultaneously. This would create a processor slightly less than 3 times
as fast as the current generation at the same megahertz without blowing
out the transistor (and power budget) with relativly little design work.
(A lot less than it takes Intel to add another instruction pipeline to
the pentium and deal with the resultant issues in branching and
out-of-order execution. (can I just call that oooe?))
> We have no idea what it is capable of for other ISA's, but
>one thing is clear: morphing for the 128-bit floating point available on
>most other workstation architectures isn't going to work with any decent
>speed without an expansion of the Crusoe's floating-point hardware to
>beyond the 80 bits they have told us about.
I think this is a logical extention of the product line. The code
morphine allows them to pursue whatever works best without worrying about
backward compatibility.
In fact its the pentiums backward compatibility (it has a 386 cpu on
board with the risc cpu) that has hampered it-- that 386 subprocessor was
limited in its design to maintain compatibility with the 286.
The 286 was limited in its design to maintain compatibility with the 186
& 8088.
The 8088 was limited in its design to maintain compatibility with the
8086.
The 8086 was limited in its design to maintain compatibility with the
8085 and 8080.
The 8080 was limited in its design to maintain compatibility with the
8008.
The 8008 was limited in its design to maintain compatibility with the
4004.
The 4004 was arguably the first ever microprocessor.
Now some jumps and changes that broke backward compatibility occured
here, but generally backward compatibility was a high prioirity thruout
the x86 development history and limited what could be done. That's why
the itanium is controversial-- its a VLIW processor, not a CISC one.
BitGeek
From Comp.sys.wearables Newsgroup Archive (CSW)
Maintained by R. Paul McCarty
Archive created with babymail