[Openmcl-devel] Profiling using Shark

Osei Poku osei.poku at gmail.com
Mon Feb 11 19:14:00 PST 2008


On Feb 11, 2008, at 6:45 PM, Gary Byers wrote:

> The good news is that ccl:library;chud-metering.lisp should already
> know how to write spatch files for Shark, and that it seems to have
> been conditionalized for x86-64 and ppc64.
>
> The bad news is that - even though CHUD's release notes have declared
> it to be 64-bit aware for the last year or so - the libraries used
> to control profiling from an application are still only available
> in 32-bit versions.  (This was true as of a few months ago;
> I haven't checked to see if more recent versions are available
> and if they address the issue.)

Is this why I am getting this error when I do (require "CHUD-METERING")?

Error opening shared library "/System/Library/PrivateFrameworks/ 
CHUD.Framework/CHUD": dlopen(/System/Library/PrivateFrameworks/ 
CHUD.Framework/CHUD, 10): no suitable image found.  Did find:
         /System/Library/PrivateFrameworks/CHUD.Framework/CHUD: no  
matching architecture in universal wrapper
    [Condition of type SIMPLE-ERROR]


>
>
> I don't know if Shark understand 64-bit spatch files, or if
> the syntax of the spatch files generated by the function in
> chud-metering.lisp matches its expectations.
>

I commented out the chud-framework specific stuff due to the above  
error just to see if a valid spatch file is generated.  That failed  
too.  The file was empty and after further investigation, I discovered  
that the function was returning the same number for both the lower and  
higher bounds.  i.e., there are no functions in here.

CL-USER> (chud::get-readonly-area-bounds)
52777631875072
52777631875072

Finally, I tried to run the print-shark-spatch-record by specifying a  
specific function as an argument but that did not work either.

Any insight would be very appreciated.  Thanks.

Osei



More information about the Openmcl-devel mailing list