[Openmcl-devel] Performance tips

[Erik Pearson]

HI Gary,

On Jun 24, 2006, at 3:20 AM, Gary Byers wrote:

>
>
> On Fri, 23 Jun 2006, Erik Pearson wrote:
>
>> Not knowing how I've implemented this I realize there is not much  
>> for you go
>> to on.
>
> In the absence of specific information, the answer to the question  
> "why
> do things run more slowly than they should ?" winds up being something
> like "er, um, misspent youth ..." and the answer to "why are things so
> fast ?" is often "Good Clean Living!".  While both of those answers  
> might
> be at least somewhat accurate, they're unlikely to be helpful.

Yeah, and too late! Ah, youth ...

>
>>
>> However, all things being equal (i.e. my less than optimal  
>> implementation of
>> the above!), the app did slow down and I'm hoping to rectify this.  
>> I'll try
>> CHUD!
>
> Shark (the GUI profiling tool in CHUD) has options that allow you  
> to export
> its summaries to text files (and/or directly email them); I'd be  
> glad to
> take a look at those results and try to interpret them and/or  
> suggest ways
> of getting further information.

I've set things up according to your directions (needed to download  
CHUD from Apple, and the run CHUD's updater) and now Shark is  
spitting out all sorts of information that I suppose is useful. The  
first thing I notice, though, is that there are a lot of unknown  
functions mixed in with the dppccl. I suppose these are the lisp  
functions? Is there something else that needs to be done to Shark/ 
CHUD to get it to recognize the spatch file?

Ah, ok, so the location of the patch file needed setting in the Shark  
preferences (to my home directory), then exit shark and openmcl, then  
restart, the voila it works great.

So I quickly found one culprit -- some of my code which was doing a  
lot of file reading with each html page rendered -- after caching  
this content the server is much faster -- basically on a par with the  
Allegro version (unofficially). I didn't save the shark diagnostics,  
but it was READ (and subservient functions) which was the slowpoke.

Now onto the other unoptimized and buggy areas of my code...

Great. Thanks!

Erik.

>
>>
>> Thanks,
>> Erik.
>>
> _______________________________________________
> Openmcl-devel mailing list
> Openmcl-devel at clozure.com
> http://clozure.com/mailman/listinfo/openmcl-devel