🚀 go-pugleaf

golubovsky@mail.ru (Dimitry Golubovsky) writes:

>Hi All,

>I haven't looked into details of each Haskell runtime
>implementation; maybe there exist some papers decribing their
>internals. 

Yes, especially the "G machine" and the "STG machine". Google for these terms
and you will find some information. 

This paper is worth reading:

  "Implementing lazy functional languages on stock hardware: the 
  Spineless Tagless G-machine ", by Simon Peyton Jones

Also there are links to some papers from the GHC website:

  http://haskell.cs.yale.edu/ghc/

look under documentation.

>Has anyone tried to do such a conversion? Maybe such a project already
>exists?

There is also the "Lazy Virtual Machine", which might do a lot of what you
are interested in. See:

   www.cs.uu.nl/~daan/papers/lvm.pdf

Helium (a language like Haskell) uses it:

   http://www.cs.uu.nl/helium/

Cheers,
Bernie.

I know it is not a virtual Haskell machine, but have a look at Peyton
Jones' baby C--, a portable assembly language.

http://www.cminusminus.org/



hth,
Steffen


On 8 Jun 2003 13:56:33 -0700, golubovsky@mail.ru (Dimitry Golubovsky)
wrote:

>Hi All,
>
>I am looking into development of "Haskell virtual machine". Current
>Haskell implementtaions usually compile Haskell source into some code
>in C (containing description of the graph to reduce), and then it is
>passed through a regular C compiler, combined with runtime and then a
>pretty big executable is produced. What I am looking into: to strip
>the runtime into a shared chunk of code (like Java virtual machine is
>the same executable for every Java program), and store the pieces of
>graph (functions) as something like loadable Java classes (binary
>files also shared between different programs) with possibility of
>loading on demand (so these files might be primary output from the
>compiler). I haven't looked into details of each Haskell runtime
>implementation; maybe there exist some papers decribing their
>internals. I also hope that developers of Haskell implementations read
>these newsgroups, so maybe some of them might suggest better ways of
>conversion of their existing runtime into a virtual machine (or
>totally discourage me from doing that).
>
>What might be helpful: information how to understand the structure of
>the C code generated to describe a graph.
>
>I tried to search Google for "Haskell virtual machine", but results
>were mostly related to bridging of Java VM and Haskell what is not I
>am looking for.
>Has anyone tried to do such a conversion? Maybe such a project already
>exists?
>I will appreciate any input.
>PS If you want to contact me via e-mail please write to dimitry at
>golubovsky dot org: the address found in this message's header is
>indeed a spam trap.

Dimitry Golubovsky writes:

 > I am looking into development of "Haskell virtual machine". 

You are aware of <http://haskell.org/hugs/>, right? It may not be
exactly a "virtual machine", but IMHO it's pretty close to one.

Peter

In article <f45cab03.0306081256.169d5a74@posting.google.com>,
golubovsky@mail.ru (Dimitry Golubovsky) wrote:

>I am looking into development of "Haskell virtual machine".

As hinted by the other posts, I have a vague memory that Hugs
http://haskell.org/hugs already works like that. One limitation is that it
cannot execute parallel threads, though. There are some other limitations.

I suggest you inquire in the Haskell and Hugs mailing lists, see
http://haskell.org/.

  Hans Aberg      * Anti-spam: remove "remove." from email address.
                  * Email: Hans Aberg <remove.haberg@member.ams.org>
                  * Home Page: <http://www.math.su.se/~haberg/>
                  * AMS member listing: <http://www.ams.org/cml/>