Hi all, can anybody tell me if LBAcache now handles over-formatted floppy
disks properly? As you know, you can use FreeDOS FORMAT (or Linux tools or
whatever else you prefer) to format e.g. to 1680k in an 1440k drive.
In the past, LBAcache bailed out when you had LBAcache FLOP loaded and
tried to use such a disk. For testing, you can load from the prompt:

LBAcache DRV NULL BUF 8 FLOP
(drv null suppresses harddisk caching if you have a problem with it, the
 NORMAL way to use LBAcache would be no DRV ... option at all, letting
 LBAcache autodetect the harddisks)

LBAcache should update geometry properly if you start using a 1680k
floppy after loading it, and it should detect the size change even if you
re-format a 1440k floppy to 1680k while LBAcache is loaded.

So far for the theory. Alas you need a real floppy drive, a free
floppy disk and a FreeDOS kernel (FreeDOS has the nice feature of
being able to use special formats without extra drivers) for this
test and I do not have all of that around right now :-(.

http://www.coli.uni-sb.de/~eric/stuff/soft/ lbacache-14apr2004.zip

Changes: Floppy stuff described above, reduced panicking (now ignores
geometry change function 18 for drives 2..7f/88..ff and ignores Con->Fmt
function 0xec), improved startup message (no longer quotes the command
line, better and redirectable feedback message for DRV option), using
8k instead of 4k element size now (-> only 35k DOS RAM used for 29 MB
cache size, although such huge caches are NOT recommended).
Maybe other small changes which I forgot to mention.

I hope the element size change did not introduce new bugs. If you can
test LBAcache with more than a floppy (e.g. test it with a real or
virtual harddisk) on a test system where data loss is no problem,
please do so. My own test system is not available today, sorry.

A very interesting question is whether the bigger element size reduces
the cache hit ratio in real life situations (whatever your preferred
test is, maybe something like compiling the FreeDOS kernel). If YES,
I would add a command line option to select element sizes of 1 / 2 / 4 / 8
kilobytes (similar to SMARTDRV). If NO, I would just leave LBAcache
compiled to maximum element size of 8k (as it is now, similar to NWCACHE).
Comparison is simple: Run the test with LBAcache 22oct2003 (4k element
size) once and run it again with LBAcache 14apr2004 (of course with a
strategy like 1. make clean 2. LBAcache sync 3. make 4. LBAcache info
to have reproduceable reaults).

The bigger element size will also lower CPU load -> improve disk access
speed in case of cache miss.
However, I hope to find the time to improve cache miss response speed
a lot by changing binsel2.asm findbin algorithm soon anyway.

Would be good if somebody could test the 14apr2004 version SOON
to allow Bernd to include it in the next FreeDOS beta 9 release
candidate. Thanks!


(Pity that my own test system is not ready soon enough...)

Eric


-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
_______________________________________________
Freedos-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/freedos-devel

Reply via email to