On 18 March 2010 09:57, Francesc Alted <faltet@pytables.org> wrote:
Hi,
Konrad Hinsen has just told me that my article "Why Modern CPUs Are Starving and What Can Be Done About It", which has just released on the March/April issue of "Computing in Science and Engineering", also made into this month's free-access selection on IEEE's ComputingNow portal:
Speak for your own CPUs :). But seriously, congratulations on the wide publication of the article; it's an important issue we often don't think enough about. I'm just a little snarky because this exact issue came up for us recently - a visiting astro speaker put it as "flops are free" - and so I did some tests and found that even without optimizing for memory access, our tasks are already CPU-bound: http://lighthouseinthesky.blogspot.com/2010/03/flops.html In terms of specifics, I was a little surprised you didn't mention FFTW among your software tools that optimize memory access. FFTW's planning scheme seems ideal for ensuring memory locality, as much as possible, during large FFTs. (And in fact I also found that for really large FFTs, reducing padding - memory size - at the cost of a non-power-of-two size was also worth it.)
http://www.computer.org/portal/web/computingnow http://www.computer.org/portal/web/computingnow/0310/whatsnew/cise
On it, I discuss one of my favourite subjects, memory access, and why it is important for nowadays computing. There are also recommendations for people wanting to squeeze the maximum of performance out of their computers. And, although I tried to be as language-agnostic as I could, there can be seen some Python references here and there :-).
Heh. Indeed numexpr is a good tool for this sort of thing; it's an unfortunate fact that simple use of numpy tends to do operations in the pessimal order... Ane
Well, sorry about this semi-OT but I could not resist :-)
-- Francesc Alted _______________________________________________ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion