[PATCH v2] benchtest: malloc tcache hotpath benchtest.

[Cupertino Miranda]

Hi Wilco,

I am having seconds thoughts on the RSS argument.
I will simplify it such that it only cares for the hotpath, just like 
you suggest.

Thanks,
Cupertino

On 29-04-2025 20:12, Cupertino Miranda wrote:
> Hi Wilco,
> 
> Thanks for the quick turn around.
> 
> On 29-04-2025 19:52, Wilco Dijkstra wrote:
>> Hi Cupertino,
>>
>> Basically with a completely flat distribution you end up always 
>> overflowing
>> tcache, so it now spends most of the time outside of the tcache 
>> hotpaths...
> Right we should limit to number of slots for each allocation size.
> I did not want to limit to single allocation size, because I wanted to 
> get some RSS measurement too. Does not make sense now, but will once we 
> have multi-size (large) tcache bins.
> 
>> If we want to test just the tcache hotpaths, we don't need even half 
>> this code,
>> it's enough just to do malloc and free in a tight loop for a size that 
>> fits in tcache.
> Damn it, forgot to fixup the patch that would remove it. Apologies !
> Will prepare a new version very quickly.
> 
> Cheers,
> Cupertino
>>
>> Note also the prefill code seems to be still there.
>>
>> Cheers,
>> Wilco
>