Herbert, Stéf, Klaus, John,

thanks very much for your feedback.
I am really excited by this idea of profiling. you count on me to ask  
you to be testers :-)

Cheers,
Alexandre


On Oct 29, 2005, at 11:36 PM, John M McIntosh wrote:

> Well Process>>resume:  in the vm is where process switch. A few  
> years back I wrote  a change set and altered Process to track  
> millisecond clock time, start dispatch, end of dispatch, and total  
> time dispatched. That let you understand the clock time that a  
> process had to run with (not the same as cpu time). One could also  
> track the allocation count for a process by tracking that by the  
> active processor structure when new is invoked.
>
>
> On 29-Oct-05, at 1:00 PM, Klaus D. Witzel wrote:
>
>
>> Hi Alex,
>>
>> perhaps comprehensive system accounting (CSA) is a starter to see  
>> what ideas are already around for everyday's system accounting  
>> (for such long runners as Seaside).
>> - http://www.google.com/search?q=comprehensive+system+accounting+csa
>>
>> In any case I think that accounting for memory and cpu should be  
>> per process (and so I would see my favorite memory integral  
>> again :) but that would require changes to ObjectMemory (object  
>> instantiation + GC) ...
>>
>> Into quite another direction goes the polymorphism metrics, which  
>> addresses the actual receiver and/or target method per call site,  
>> and can serve as a measurement of an application's object- 
>> orientedness, etc. This should also be per process (or, better  
>> perhaps per ClassBox).
>> - http://www.google.com/search?q=call+site+polymorphic
>>
>> Hope that's not too much ;)
>>
>> /Klaus
>>
>> On Sat, 29 Oct 2005 19:04:34 +0200, Alexandre Bergel  
>> <bergel at iam.unibe.ch> wrote:
>>
>>
>>
>>> In the very near future, I plan to work on profiling technics.  
>>> The question I want to ask you is: What are the information we  
>>> want to obtain from a program execution?
>>>
>>> Here are some thoughts:
>>>      - 2 kinds of profiling: time and speed. How long methods of  
>>> my application takes ? How much memory my application requires ?
>>>      - minimize the amount of information by restricting the  
>>> output of the profiler to the information I am really interested  
>>> in. Basically, If I define a seaside application, I am only  
>>> concerned by my application, and not by the seaside framework.
>>>      - Currently, in Squeak,  I can only listen to everything  
>>> (i.e., every method call is measured). It might be interesting to  
>>> limit the measurement to a small part of the system (e.g., objets).
>>>
>>> I would be really interested in hearing your opinion...
>>>
>>> Regards,
>>> Alexandre
>>>
>>> On Oct 29, 2005, at 6:36 PM, tim Rowledge wrote:
>>>
>>>
>>>
>>>>
>>>> On 29-Oct-05, at 8:50 AM, John M McIntosh wrote:
>>>>
>>>>
>>>>
>>>>
>>>>> Let me add a primitive to turn that on or off
>>>>>
>>>>>
>>>>>
>>>> There are already primitives for controlling profiling. Just use  
>>>> those!
>>>> clearProfile,
>>>> dumpProfile,
>>>> startProfiling,
>>>> stopProfiling
>>>>
>>>>
>>>> tim
>>>> --tim Rowledge; tim at rowledge.org; http://www.rowledge.org/tim
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>>
>>
>>
>>
>>
>>
>>
>
> --
> ====================================================================== 
> =====
> John M. McIntosh <johnmci at smalltalkconsulting.com> 1-800-477-2659
> Corporate Smalltalk Consulting Ltd.  http:// 
> www.smalltalkconsulting.com
> ====================================================================== 
> =====
>
>
>

-- 
_,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:
Alexandre Bergel  http://www.iam.unibe.ch/~bergel
^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;.