[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Chicken-users] Reset counters when profiling Chicken programs

From: Peter Bex
Subject: Re: [Chicken-users] Reset counters when profiling Chicken programs
Date: Fri, 29 Aug 2014 09:13:45 +0200
User-agent: Mutt/

On Fri, Aug 29, 2014 at 09:03:52AM +0200, Sven Hartrumpf wrote:
> Hi all.
> Is there a Chicken function to reset all counters (to 0) when
> a program is run in profile mode?
> Background: I need profiling results, but only after my program
> reached a specific point and until it terminates.

Unfortunately, there's none that I know of.

A nasty trick I sometimes use is to extract "interesting" parts
from a procedure and put it into a differently-named one, so that
it shows up separate from the rest of the procedures.  Depending
on the exact layout of your program, you could try and do that,
like so (or perhaps improved on with some macrology):

(define (proc2 arg1 arg2 ...)

(define (proc arg1 arg2 ...)
    ((> counter counter-limit)
     (proc2 arg1 arg2 ...))
    (else (set! counter (add1 counter))
          ORIGINAL CODE)))

This is extremely ugly, and if you're profiling a large program that
won't really be possible.  The main problem is that the vector into
which profiling information is written is a random name, determined
at compile time.  I don't really know why this is done, but it does
mean you can't easily access it.

The only way around this I can come up with is a user compiler pass.
It could inspect the name and write it to a global variable, which
can then be used by your program to access the profiling vector at
runtime, the same way the compiler's generated profiling
instrumentation does it.  Then you wouldn't need the ugly trick above.

Perhaps we can come up with an API for interfacing with the profiler,
for (beyond) CHICKEN 5.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]