Skip to Content.
Sympa Menu

coq-club - Re: [Coq-Club] Ascii module

coq-club AT inria.fr

Subject: The Coq mailing list

List archive

Re: [Coq-Club] Ascii module


chronological Thread 
  • From: Frederic Blanqui <frederic.blanqui AT inria.fr>
  • To: coq-club AT inria.fr
  • Subject: Re: [Coq-Club] Ascii module
  • Date: Thu, 03 Mar 2011 17:39:37 +0800

You may be interested in knowing that, in the Formes team (http://formes.inria.fr/), we formalized in Coq a big part of the semantics of the ARMv6 processor. In fact, a big part of this formalization is generated automatically from the documentation. See http://www-rocq.inria.fr/~blanqui/rapido11-pdf.html for some details. A public release will certainly be done by the summer but I can provide some files if you are interested. We are also working on the processor SH4. Best regards, Frederic.

Le 03/03/2011 16:32, Stéphane Lescuyer a écrit :
Incidentally, ARM processors were precisely the application I had in
mind when I inquired about hundreds of constructors. That's why I'm
curious about users who possibly tried that many constructors and
their experience with it.

S.

On Thu, Mar 3, 2011 at 4:16 AM, Frederic Blanqui
<frederic.blanqui AT inria.fr>
  wrote:
I don't think that this is so uncommon. For instance, processors often have
dozens or hundreds of instructions. The ARMv6 processor has 147 32-bits
instructions and 73 16-bits instructions... Many term rewrite systems in the
Termination Problem Data Base have dozens of function symbols. Some have
more than several hundred symbols. So, I would say that hundreds of
constructors is not so uncommon. Now, thousands of constructors is perhaps
more uncommon. Frederic.






Archive powered by MhonArc 2.6.16.

Top of Page