On Mon, 30 Dec 2002 11:59:04 +0100 (CET), Jimi wrote:
>I think about limiting to 16 or 32MB, resulting in 2k or 4k SUB-Block average
>possible size per device.
>SUB-Block minimum size is 512 bytes and most of the time that should be enough. Often
>devices do not
even
>need a SUB-Block at all.
I believe I should lower the chunk-size to 256 bytes, so if all 8192 devices would be
present *and* every one
would need some tiny SUB-Block, then 2MB would get eaten up. Sounds good to me.
Also I could limit to 8MB, which would make 1k SUB-Block (average) per device.
Normally it's even below
256 bytes, but you never know what gets invented later on.
This would now eat up 256 GDT-Selectors, which is still a huge amount, but at least
not that insanely high.
If something evil gets invented later (by e.g. M$), this limit may get enlarged of
course. It's just the way how I
manage the SUB-Block Allocation area. In that way I would use 3 separate blocks. One
for the location of the
segment (8 bytes), one for the maximum block and total free chunk counts (4 bytes) and
one freespace
bitmap (32x8 bits->32 bytes), times 128 resulting in around 5-6k allocation area size.
Also each BASE-Block would contain one index (word) and an offset (word), resulting in
a 4 byte requirement
as pointer to the assigned SUB-Block. Also each SUB-Block is limited to 64k (I love
that limit, hehe), anyway
that's way over the top. One could stuff his 10 slot PCI machine up with 10
USB-controller cards (hold e.g. 2
controllers each), which would result in 20 actual controllers, which would result in
theoretically capability of
2540 USB devices (crazy!) and even that would just fill 1/4th of the total HWDevice
space, if it would even
work, because of a) power consumption and b) OS/2 drivers.
cu, Kiewitz
-----------
To unsubscribe yourself from this list, send the following message
to [EMAIL PROTECTED]
unsubscribe acpi-os2
end
