[Mono-list] Intel and the CLR?

Ben Houston ben@exocortex.org
Mon, 4 Mar 2002 22:36:04 -0500

Hi Piers,

Are we are talking about implementing the MSIL (microsoft intermediate
language) into a chip instead of another version of assembly language
(i.e. x86, 680x0, PowerPC, Java byte code, etc...).  This is a horrible
idea in that it locks the constructs that can be used in MSIL byte code
because one can not upgrade a chip as easily as one can upgrade
software.  I also do not see the point -- one compiles (ie. JIT
compilation) MSIL into x86 assembly (or any other processor assembly
language) anyways before running it anyways.  Thus the only benefit is
that one does not have to JIT code before running it.  But even this
benefit is minimal considering that .NET currently caches the result JIT
binaries between sessions.  I think that Intel's efforts would be better
spent on just plain better processors and compilers (which could be
taken advantage of by MSIL though intelligence JIT'ing) rather that an
MSIL specific processor.

But maybe I am wrong and you do mean a processor that implements the CLR
(common language runtime).  I am not sure how this would work but it
would have to be a really complex chip and the first time there is a bug
found in the CLR code it would be obsolete or it would need a microcode
update.  What language would a CLR-implementing chip run?  MSIL or
something else?

Sorry for being so negative on the idea... :-/  Maybe I am missing
something vital, and if so please enlighten me.

-ben houston
4th Year Cognitive Science/Neuroscience
Carleton University, Ottawa, Canada
( ben@exocortex.org / 613-266-0637 

> -----Original Message-----
> From: mono-list-admin@ximian.com [mailto:mono-list-admin@ximian.com]
> Behalf Of Piers Haken
> Sent: Monday, March 04, 2002 9:57 PM
> To: Duco Fijma; Chris Podurgiel
> Cc: mono-list@ximian.com
> Subject: RE: [Mono-list] Intel and the CLR?
> Didn't sun try to implement the java VM in silicon, or at least design
> processor that could run bytecode more efficiently?
> i think a native-CLR processor is a great idea. the CLS represents
> a good basis for an operating system API and would make a great
> for mobile devices without all the bagage of win32/wince.
> Maybe this would make a good OSS project - an operating system kernel
> based on *BSD/Linux that exposes the CLS API instead of the standard
> unix(2|8) API. It could use only the core parts of the underlying
> (MM/VM/filesystems/networking/drivers, etc...) and add a JIT compiler
> (for architectures that need it) & Garbage Collector. If you
> the security model natively and only allowed the execution of managed
> user code then you could probably implement processes/appdomains and
> all 'user' code in kernel mode to reduce the number of context
> necessary.
> Piers.
> -----Original Message-----
> From: Duco Fijma [mailto:duco@lorentz.xs4all.nl]
> Sent: Monday, March 04, 2002 3:13 PM
> To: Chris Podurgiel
> Cc: mono-list@ximian.com
> Subject: Re: [Mono-list] Intel and the CLR?
> On Mon, 2002-03-04 at 21:14, Chris Podurgiel wrote:
> > --------------------------------------------------
> > I heard at VSLive that Intel plans to put the CLR in silicon. I
> > that will have a positive impact on performance for CLR languages
> > as Visual Basic .NET and C#.
> Plans like these were made since the 60's when somebody invents a new
> language and/or runtime concept. Somewere in the stone age, Lisp was
> invented, having a relative low performance compared to the other
> languages that were popular these days. Guess what: it was proposed to
> build a Lisp interpreter in hardware (and I believe such machines were
> actually built). In the 70's, the same ideas were launched for UCSD
> p-code, a intermediate language closely related to a then-popular
> implementation. In the 80's, silicon graph reduction machines (as
> opposed to the traditional stack based machine) were -at least in the
> scientific world- a hype, caused by the attention lazy functional
> programming languages and other declarative language got. The most
> recent example of the idea is the Java byte code machine.
> While the idea basically sounds good, none of these machines ever
> popular. If you believe in conspiricy theories, you might believe that
> plans like these are a marketing tricks to counter criticism about the
> performance of that nice new runtime concept.
> Duco
> _______________________________________________
> Mono-list maillist  -  Mono-list@ximian.com
> http://lists.ximian.com/mailman/listinfo/mono-list
> _______________________________________________
> Mono-list maillist  -  Mono-list@ximian.com
> http://lists.ximian.com/mailman/listinfo/mono-list