[Interest] Benchmarking machine cpu power
Sean Harmer
sean.harmer at kdab.com
Sun Apr 27 13:08:12 CEST 2014
On Sunday 27 April 2014 12:51:50 maitai at virtual-winds.org wrote:
> Thanks for your reply,
>
> That was my intend to write such a small routine that will sample the
> CPU and automatically tune the various options in order to achieve
> something reasonable for the user and machine.
>
> May be a bit off topic, but what do you mean exactly by "minimising
> cache misses"? There is no disk access during calculation phase (which
> is multithreaded).
I meant CPU caches misses. When iterating over data try to make sure it is
contiguous in memory rather than dispersed all over the place.
One example from typical 3D scenegraphs is updating world transformation
matrices and bounding volumes. Ensuring that the matrix data for all nodes is
contiguous (rather than the Nodes and all their other data being contiguous or
scattered over memory) means that when iterating we will have fewer cache
misses and therefore that your CPU will spend more time operating on your data
rather than waiting for it to be transferred from main memory.
This interactive demo gives you a good feel for relative access times of
various cache levels and main memory.
http://www.overbyte.com.au/misc/Lesson3/CacheFun.html
As always, benchmark :)
Have fun,
Sean
--
Dr Sean Harmer | sean.harmer at kdab.com | Managing Director UK
Klarälvdalens Datakonsult AB, a KDAB Group company
Tel. Sweden (HQ) +46-563-540090, USA +1-866-777-KDAB(5322)
KDAB - Qt Experts - Platform-independent software solutions
More information about the Interest
mailing list