Process speed and algorithm cost measures [was: Re: About slugginess
of 3.8 unstable]
Stephan Rudlof
sr at evolgo.de
Fri Oct 8 20:03:41 UTC 2004
John and others,
John M McIntosh wrote:
> A long time ago I did a change set that added dispatch clock time to an
> instance variable in
> Process when the VM process scheduler switched processes.
John: Do you remember how much the overhead regarding VM speed has been?
All:
Has anyone counted bytecodes and message sends in the VM loop?
How much has been the overhead in this case?
> This is the
> only way to accurately track this
> information.
Greetings
Stephan
>
> On Oct 4, 2004, at 6:43 PM, danielv at tx.technion.ac.il wrote:
>
>
>>So, real time tests would still be valuable. They give information
>>now, and
>>can be replaced to use CPU time if that becomes available.
>>
>>About making the Squeak scheduler note CPU time - wouldn't it be
>>possible to do this from the image side, by calling the OS "CPU time
>>for
>>process" at every Squeak process rescheduling?
>>
>>Then it doesn't change VM performance, it can be turned on for
>>profiling
>>and off again. Heck, if the call is cheap enough (no idea) compared to
>>Squeak scheduling, it might be worth leaving on.
>>
>>Daniel
>>
>>Stephan Rudlof <sr at evolgo.de> wrote:
>>
>>>Interesting, too, could be a CPU running time for *each* Smalltalk
>>>process: but how to realize without compromising VM performance? Or
>>>wouldn't this be an issue?
>>>
>>>
>>>Greetings
>>>Stephan
>>>
>>>
>>>>...
>>
>>
>>
> --
> ========================================================================
> ===
> John M. McIntosh <johnmci at smalltalkconsulting.com> 1-800-477-2659
> Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com
> ========================================================================
> ===
>
>
>
--
Stephan Rudlof (sr at evolgo.de)
"Genius doesn't work on an assembly line basis.
You can't simply say, 'Today I will be brilliant.'"
-- Kirk, "The Ultimate Computer", stardate 4731.3
More information about the Squeak-dev
mailing list
|