🚀 go-pugleaf

In article <1992Nov17.162428.15881@coe.montana.edu> uesu03@giac1.oscs.montana.edu (Lou Glassy) writes:
>Is the idea of running code on virtual machines dead?
>Found a couple of old books yesterday on the p-System and 
>MINT.  
>
>The possibility of taking my binary from machine to machine,
>(with possibly different physical architectures), and having
>it run without dinking, is an intriguing one.
>
>Is any work going on in this area?  Or does performance pay
>more than portability?
>
>Thanks in advance,
>
>Lou.
>


Pascal used to do this via its famed "pcode".  The idea was to output an
intermediate code, then have an interpreter engine run on each different
architecture.  flexible, but slow.  We have been interested in this for
process migration on distributed systems.  The real difficulty is that 
we would like to be able to "stop" a process, move it, and re-start it.

Figuring out how to map the "real" machine state to the "psuedo-machine"
state is interesting since different machines have different sets of
registers, flags, etc.  and optimizing the instruction streams for
performance really begins to depend on these machine characteristics.
Imagine running a code for 1/2 hour on a Cray and then moving it to a
Sparc.  I know where to start, but the process is unbelievably messy
and difficult.
-- 
!Robert Hyatt                    Computer and Information Sciences   !
!hyatt@cis.uab.edu               University of Alabama at Birmingham !