Re: [clisp-list] Lots o' garbage ...

[Sam S. <sd...@gn...>]

> * In message <200...@va...>
> * On the subject of "[clisp-list] Lots o' garbage ..."
> * Sent on Mon, 16 Sep 2002 16:16:40 -0400
> * Honorable "John K. Hinsdale" <hi...@al...> writes:
>
> I think it is that my library is doing lots of consing and generating
> a LOT of garbage for Lisp's GC to deal with (more on this below) So
> I'm wondering: what's a good way to find the places in my code where
> I'm consing and generating garbage?  Is there a "consing profiler"
> type tool to do that?  I just read the chapter entitled "Speed" in
> Graham's book "ANSI Common Lisp" and it says what functions to avoid
> calling a lot, but I would rather be able to focus in on the hot spots
> instead of having to laboriously rework all my newly debugged code ;)

CLOCC/src/tools/metering.lisp
and CLISP built-in SPACE macro
(see <http://clisp.cons.org/impnotes.html#space>
<http://clisp.cons.org/impnotes/environment-dict.html#space>)

>       %   cumulative   self              self     total           
>      time   seconds   seconds    calls   s/call   s/call  name    
>      29.45     28.07    28.07     1498     0.02     0.04  gar_col_normal
>      22.95     49.95    21.88  4556398     0.00     0.00  gc_mark
>      12.16     61.54    11.59 139359175     0.00     0.00  objsize
>       6.62     67.85     6.31 11624570     0.00     0.00  interpret_bytecode_
>       1.54     69.32     1.47  2372121     0.00     0.00  allocate_iarray
>       1.41     70.66     1.34 25267275     0.00     0.00  funcall
>       1.31     71.91     1.25 10240813     0.00     0.00  funcall_closure
>       0.87     72.74     0.83   355825     0.00     0.00  convert_from_foreign
>       0.86     73.56     0.82 15026462     0.00     0.00  funcall_subr
> 
> Adding the first three lines, it appears to be spending over 60% of
> its time doing GC.

this is ugly.  usually the biggest hog is interpret_bytecode_().

> Also notice the 25 million calls to "funcall" -- I tried to elminate
> that by proclaiming ALL my Oracle package functions as "inline" as
> well as "(declare (optimize (speed 3)))" etc.  I also built CLISP with
> C compilation flags for maximum optimization (-O4)

CLISP does respect INLINE declarations when you compile a file.
CLISP ignores OPTIMIZE.

Alas, CLISP FFI is not too efficient.

Joerg-Cyril posted several interesting rants on the matter, but is yet
to fix this.  Prod him!

-- 
Sam Steingold (http://www.podval.org/~sds) running RedHat7.3 GNU/Linux
<http://www.camera.org> <http://www.iris.org.il> <http://www.memri.org/>
<http://www.mideasttruth.com/> <http://www.palestine-central.com/links.html>
Your mousepad is incompatible with MS Windows - your HD will be reformatted.