Erlang questions on comp.lang.functional.

Shawn Pearce <>
Mon Mar 3 08:44:15 CET 2003


Per Bergqvist <> wrote:
> > >  2. Bytecode interpreted (i.e. not compiled to native code).      
> >                                                                     
> > Isn't this why we have HIPE?  And thus far, I have not seen HIPE    
>> improve                                                               
> > performance over code run in the bytecode interpreter.  Therefore I 
>> have                                                                  
> > to ask:                                                             
> >                                                                     
> > 	- is the rest of erts slow?                                        
> > 	- is the bytecode interpreter really fast?                         
>                                                                       
> erts and beam are quite good.                                         
> HIPE generates good code for the function body but the gain is        
> "hidden" by the overhead costs.                                       
> Until HIPE starts to do loop and whole program optimisations we will  
> not see any major improvement IMHO.                                   

Ok, now you've got me looking forward to the day HiPE starts to do that.  :)
AFAIK, Java is already doing this kind of optimization, it'll be nice
when we can really start to compete at the performance level with other
systems.

Oh wait, Erlang already does.  :-)  The testing I've been doing today
has been knocking my socks off performance wise.  Just simple stuff
with ets, gen_tcp, term_to_binary, yet I'm quite amazed at how fast
Erlang really is.  I can't remember when I had this much fun writing
sorta-complex software.  :)

-- 
Shawn.

  Next Friday will not be your lucky day.  As a matter of fact, you don't
  have a lucky day this year.



More information about the erlang-questions mailing list