testing needed for branch luben/gc_threshold_adjust

Luben Karavelov karavelov at spnet.net
Fri Nov 26 22:25:08 UTC 2010


 On Fri, 26 Nov 2010 19:57:38 +0100, Nick Wellnhofer 
 <wellnhofer at aevum.de> wrote:
> On 26/11/2010 01:54, Luben Karavelov wrote:
>> - Different programs have different needs. Bigger threshold gives 
>> better
>> throughput, smaller threshold gives better interactive behavior.
>
> More importantly, the threshold should depend on the actual memory
> needs of the program. That's why a static threshold or a threshold
> depending only on the amount of system memory is fundamentally 
> flawed.
>

 True

>> Changes in the branch:
>>
>> - Parrot_sysmem_amount() return the total installed memory
>> - gc_threshold is fraction of total installed memory
>> - gc-threshold commandline parameter is used to specify the 
>> percentage
>> (default 5%). This parameter was already there but it was not used 
>> in
>> the current GC system.
>
> Do we really expect our users to fiddle with that threshold for every
> program they run? I ported the dynamic threshold to GC MS2 two months
> ago:
>
> 
> https://github.com/parrot/parrot/commit/7eb63abc37da66864be49a1ba41919fb110ba903
>
> I'd much rather use that than any static threshold. It worked pretty
> well with the old GC.
>
> Nick

 Yes, dynamic threshold is a better idea, I have tested it a month ago 
 and it
 was working very well. I do not know why it is not yet merged. I have 
 tried to
 locally merge master in it with some conflicts that I could not 
 resolve.

 My branch just fixes the current extremely inconsistent behavior. I 
 also prefer
 your branch to be merged instead of mine.

 Best regards

 Luben



More information about the parrot-dev mailing list