[f-cpu] cachemm instruction

[nicO <nicolas.boulay@ifrance.com>]

I just reread the instruction description of cachemm : beark !

(compression/decompression flag, virtual memory hint (!))

Even the description isn't so good. (base+offset). What does it means ?
Must we do a loops to count all the cache line ? Must we calculate the
end of the bloc ? 

So i propose to use a base + a mask or just a management paged based.
And how many of this instruction could taken into account (1 ,8 ?) ?

All locks and things like that are far to much implementation dependant.
The same cpu with different implementaton could lose a lot of
performance (imagine a lock then the use of a L1 data cache one-way
instead of the 2-way cache that you use before).

Maybe we could find someway to said what we want to do instead of say
what the processor must do.

We could exprime : this bloc will be used often, this bloc is only used
one time, ...

This instruction manage prefetch and are supposed to fill L0 cache. 

Comments ?

nicO
*************************************************************
To unsubscribe, send an e-mail to majordomo@seul.org with
unsubscribe f-cpu       in the body. http://f-cpu.seul.org/