Re: [Gems-users] CPU accesses to L2 cache

[Dan Gibson <degibson@xxxxxxxx>]

Barnes I'm unsure of -- I don't recall the barnes algorithm and data 
structure well enough. Have a look at the code and see how much memory 
it allocates.
Ocean is simpler... recall that the data structure for ocean is an 
(N+2)x(N+2) matrix. I don't recall the size of the elements (either 4 or 
8 bytes, I'll assume 8 for this calculation). I also don't recall if 
Ocean uses a shadow of its data or updates the data in-place (I'll 
assume a shadow).
Size of ocean working set ~=
Matrix Size * sizeof(matrix element) * Number of Matrices + ~ small 
number of stack vars
66x66 x 8 bytes x 2 matrices + noise
~= 69,696 bytes total ...
divide that by, say, four L1 data caches (as Ocean is essentially 
embarrassingly parallel, has great spatial and temporal locality, and 
very little sharing):
~= 17,424 bytes per cache...

Thus, after the caches are warm, I'd expect to see only a few coherence 
misses.
Regards,
Dan Gibson

Daniele Bordes wrote:
> Thank you very much for your quick reply, Dan. I have a question: how
> can I determine the working set dimension of Ocean and Barnes varying
> input parameters (that is, for instance, grid dimension for Ocean and
> particles number for Barnes)?
>
> Sorry to be a nuisance.
> Thank you very much.
> _______________________________________________
> Gems-users mailing list
> Gems-users@xxxxxxxxxxx
> https://lists.cs.wisc.edu/mailman/listinfo/gems-users
> Use Google to search the GEMS Users mailing list by adding "site:https://lists.cs.wisc.edu/archive/gems-users/"; to your search.
>
>   
--
http://www.cs.wisc.edu/~gibson [esc]:wq!