On Tue, Nov 15, 2005 at 02:12:55PM +0100, Jens M Andreasen wrote:
(good to see you're back on line :-)
> On a related subject: How is level one cache replaced with new data,
> should one (or ones compiler) decide to use some of the prefetch
> instructions available from Intel PII and up? It would make sense to
> fetch the next dataset while doing what has to be done "now". On the
> other hand, overwriting the current dataset is somewhat counter
> productive.
Unless you're hand-coding assembly it's probably wisest to leave this
to the compiler. OTOH, I've no idea how smart gcc/g++ is in this respect.
It could be quite interesting to -S some familiar DSP code and have a
look at the result.
The only place where I've seen prefetch used explicitly is in Brutefir's
sse and 3dnow routines which I recently modified for use in one of my own
projects.
> The layout of the data would influence these decisions, no?
And conversely, considerations relating to cache use (and possible sse
optimisations) may influence your choice of data formats.
-- FAReceived on Tue Nov 15 16:15:07 2005
This archive was generated by hypermail 2.1.8 : Tue Nov 15 2005 - 16:15:07 EET