In a message to Anthony about "hpfs386", Paul said:

AM> I hate to scare you, but I read recently that something to do with
AM> the chkdsk utility for hpfs386 can trash non hpfs386 formatted
AM> partitions, or something like that.  I also don't think IBM
AM> recommends it for normal useage.

PE> Yeah, I've gotten rid of AUTOCHECK, and I don't run CHKDSK.

Not a nice long term solution...  CHKDSK ___***NEEDS***___ to be run
every time the OS/2 closes prematurely, such as after a crash.  In fact,
with all my partitions being HPFS, OS/2 won't allow me access to a drive
after a crash until CHKDSK has been run on it - normally you do it by
the AUTOCHECK statement, but I took two drives out of this statement
once, and OS/2 said both will remain inaccessible until CHKDSK was run
on them.

PE>  If I want to do that, I will reboot under HPFS16.  As for

Is that safe?

PE> IBM not recommending it, they're as full of shit as the
PE> rest who told me it doesn't improve performance.  BFN.

The reasons I'd heard for it had nothing to do with speed, and
everything to do with data integrity and HPFS386 totally killing a drive
when CHKDSK16 was run on one in some circumstances.

The speed benefits as you saw were directly attributable to the cache
being just the right size to maintain the data of the message logging
operations you were doing almost entirely in the cache - if it were a
different task, say one that required 5Mb of data instead of the 4Mb
that seemed to be a magic number in your trials, then you'd come up
again to the same performance brick wall after 4Mb had been read into
the cache.

8Mb or bigger caches?  What a waste of memory and CPU resources...

'tnt, Anthony.

                                                                           
                      
--- Maximus/2 2.01wb