I think I'm suggesting that there be an SSE3 version, much
as there is a Mac version.  Presumably installation would
pick the right one to use.  This is one of those things that
seems easy to do as long as you don't have to be the one
that does all the work.

In addition, there could be assembler code for the operations
that gave the best improvement for the effort spent.  Right now
I guess that's   +   -   *   +/   +/ .*  but I don't
really have any insight.

Writing code on the fly is deprecated by Intel, and might not
even be allowed now that processors have the no-execute
flag.

Henry Rich

> -----Original Message-----
> From: [EMAIL PROTECTED] 
> [mailto:[EMAIL PROTECTED] On Behalf Of Miller, Raul D
> Sent: Sunday, September 03, 2006 9:04 PM
> To: Programming forum; Programming forum
> Subject: RE: [Jprogramming] Symmetric inner and outer products
> 
> Henry Rich  wrote:
> > For arithmetic operations in general, it would also be
> > a much better use of coding time to take advantage of
> > the SSE3 instructions.
>  
> Since SSE3 would fail on all CPU lines released before 2004 (and many
> released before 2005), this seems to mean that you are suggesting that
> J's arithmetic core be doubled -- with a side-by-side 
> implementation of
> SSE3 support and non-SSE3 support?
>  
> Perhaps more useful would be some documentation on how to construct
> and call user-defined machine language code from within J.  
> (I'm thinking
> something different from cd here -- I'm thinking: let's 
> generate a sequence
> of literals in J which are the machine code -- how do you 
> call such a thing
> as a function from within J?)
>  
> -- 
> Raul
> 

----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to