[TCLCORE] Re: Small patch, big gain ??

[<Miguel S. <mi...@ut...>]

Thanks for your encouragement, I'll keep trying and report any
significant progress I might get. 

I'm trying to apply some of the ideas described in the docs "around"
gforth <http://www.complang.tuwien.ac.at/projects/forth.html>. In
their terminology:  

Dr. Lewis reported in his 1996 paper that he compared "switch
threading" (current implementation) to (I think) "call threading" (a
table of function pointers); I am trying to build an efficient
implementation of "direct token threading" (a table of pointers for
gotos, not calls).  

This should be much faster than call threading, as it avoids the whole
function call circus. However, it _cannot_ be done in ANSI C, it
requires the GNU "Labels as Values" extension (for all I know, other
compilers may have similar extensions ?).

A much bigger project I may tackle in the future is to try another
threading model that avoids bytecodes (faster but larger ...),
essentially compiling into some specially adapted variant of
forth. This does require a modification of the compiler too; I'll need
more stimulants before I even _read_ those files, let alone modify
them ...

Regards

Miguel Sofer

--
The TclCore mailing list is sponsored by Ajuba Solutions
To unsubscribe:  email tcl...@aj... with the 
                 word UNSUBSCRIBE as the subject.