Stackless/microthreads merge news

Edward Jason Riedy ejr at lotus.CS.Berkeley.EDU
Wed May 10 00:14:25 EDT 2000


And William Tanksley writes:
 - 
 - I've always liked this, but there's essentially no documentation, and the
 - only source code is a MESS.  Someone's written a parser for these in C++;
 - hopefully they'll write some documentation as well.

The slim binary stuff I know is pretty Oberon-specific...  I'd love
to see Dylan++^H^H^H^H^H^H^HPython 3k use something like it, but then 
it might miss its deadline.  Maybe a load-time fixup to translate it
into a fully threaded code, leaving compilation to future generations.

ObPython:  Jim Fulton's Extension Class mechanism seems to have 
replaced MESS, although I get more use out of CXX.  ;)

 - ANDF is a more standardised version of this (it's also a lot less
 - functional).

ANDF is superficially similar, but quite different underneath.  It's
supposed to work by a sequence of simple macro expansions...  And it
used a really silly method for capturing dependencies, iirc.  The 
remains of ANDF are at http://www.ri.silicomp.fr/andf/index.html.  
It's useful reading just so people know what _not_ to do.

Most importantly, I seem to recall ANDF losing a great deal of 
information that's preserved in an encoding that remembers a bit of 
the source structure.  That could be very important if you're 
interested in dynamic recompilation of floating-point code.  But
my knowledge is just from reading, not from doing, so I may be
wrong about ANDF.

And I was about to point you at the TUNES VM review, but I
recognized your name at the last second.  ;)

Jason



More information about the Python-list mailing list