olof.bjarnason at gmail.com
Fri Oct 23 13:50:07 CEST 2009
2009/10/23 Stefan Behnel <stefan_ml at behnel.de>
> > Olof Bjarnason wrote:
> > [snip]
> >> A short question after having read through most of this thread, on the
> >> same subject (time-optimizing CPython):
> >> http://mail.python.org/pipermail/python-list/2007-September/098964.html
> >> We are experiencing multi-core processor kernels more and more these
> >> days. But they are all still connected to the main memory, right?
> >> To me that means, even though some algorithm can be split up into
> >> several threads that run on different cores of the processor, that any
> >> algorithm will be memory-speed limited. And memory access is a quite
> >> common operation for most algorithms.
> >> Then one could ask oneself: what is the point of multiple cores, if
> >> memory bandwidth is the bottleneck? Specifically, what makes one
> >> expect any speed gain from parallelizing a sequential algorithm into
> >> four threads, say, when the memory shuffling is the same speed in both
> >> scenarios? (Assuming memory access is much slower than ADDs, JMPs and
> >> such instructions - a quite safe assumption I presume)
> Modern (multi-core) processors have several levels of caches that interact
> with the other cores in different ways.
> You should read up on NUMA.
Thanks that was a good read. Basically NUMA addresses the problems I mention
by creating several primary memories (called banks) - making the motherboard
contain several computers (=cpu+primary memory) instead of one primary
memory and several cores, if I read it correctly.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Python-list