> On Sat, 28 Mar 1998, Gerard Roudier wrote:
> >
> > Without some very smart algorithm that selects pages to swap out, this
> > 128kB contiguous free area goal seems to me close to impossible without
> > trashing LOTS of really used pages.
>
> No, it seems to be a very good goal, actually. I've had reports of this
> approach getting close to exact the behaviour we want: we don't actually
> want exactly 128kB free, we want something like 5-10% of memory free just
> in case...
>
> I've also got a patch that maxes out at 25% free, just in case, to avoid
> the really bad (but unlikely) cases.
>
> > What's the real need of contiguous memory?
>
> DMA.
The only way that works is scatter/gather support by the device and
virtual mapping if the host wants to see memory as contiguous.
If this is not possible with the device you mention below, I do understand
the problem. But it is painfull to have to be as crappy as W95 in order to
support devices intended for this O/S.
AGP specifications speaks of some relocation page registers in
AGP devices. I hope this will be followed by hardware designers.
> > Modules that need large physically contiguous memory chunks should be
> > loaded very early after system start-up.
>
> Not necessarily. Think video capture etc, that definitely wants memory
> only when it is needed, but that we don't want to allocate memory for all
> the time (it can need quite a bit of memory)
Note that such devices are intended to be used on personnal stations, and
there is room for hacks in Linux in order to become as unreliable as
W95.
As an example, my previous P133 that has worked flawlessly under Linux
for 2 years is now used by my daughters under W95. This breaks every
hours when they play commercial Direct/X games...
Regards, Gerard.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu