Steven Swerling wrote:
> SASADA Koichi wrote:
>> Hi,
>>
>> Steven Swerling wrote:
>>> I'm a bit confused by the bytecodes too -- from looking at insns.inc, it
>>> appears that there are 78 bytecodes, but the docs on the web show a lot
>>> more beyond the base ones (getlocal_OP_2, send_OP__WC__0_Qfalse_0__WC_,
>>> etc).  I can't find these extra byte codes in the source, not even from
>>> grepping the entire code-base. Are those extra ones dynamicly generated?
>>
>> Basically, you don't need to check opt_XXX instructions.  These
>> instructions are for only optimization.
> 
> But that is the part I'm curious about. I've been reading up on some vm
> inlining techniques and wondering which ones might be added to YARV (I'm
> reading the items recommended in the Strongtalk "syllubus", see
> http://groups-beta.google.com/group/strongtalk-general/browse_frm/thread/83bda45b670b5a5b?tvc=1).
> 
> 
> I couldn't find anything on the web about unification until today --
> "superinstructions" from VMGen. I will try to find some info on that.
> 
> Thanks.
> 
> 
> 
> 
> 
> -- 
> ML: yarv-dev-en quickml.atdot.net
> Info: http://www.atdot.net/~ko1/quickml
> 
A couple of links on vmgen if you don't have them:

http://www.complang.tuwien.ac.at/anton/vmgen/html-docs/
http://www.complang.tuwien.ac.at/anton/vmgen/
http://www.nongnu.org/carbone/
http://savannah.nongnu.org/projects/carbone


By the way, I've actually found some things that benchmark *faster* than
gForth/vmgen (in Forth, not in Ruby!) See

http://maschenwerk.de/HelFORTH/
http://retroforth.org/
http://ronware.org/reva/
http://christophe.lavarenne.free.fr/ff/
http://retroforth.org/glypher/

Finally, for those who wish to tweak for the Intel and AMD chip sets,
there is this outstanding site

http://agner.org/optimize/



--
ML: yarv-dev-en quickml.atdot.net
Info: http://www.atdot.net/~ko1/quickml