Remember I was telling you how I switched all my data to FAT so that I
could make use of the 14400k diskcache for FAT (as opposed to the 2048
cache available on HPFS) and it turned out that it was worse?  Well I did
some investigation today, and found that I couldn't even copy 24k of files
to NUL: and then copy them again without it reaccessing the disk?!  I would
have expected read-only to work.

I have 64 meg of memory, all IDE/EIDE, PCI, OS/2 3.0, and observed the
following things:

14400 - no error messages, and what looks like a 16k disk cache.
14336 - same as 14400
14000 - trap-e internal error
13800 - weird messages about device drivers unable to load
13312 - trap-e internal error
12288 - %1 is being used for the cache size.  The total requested
        amount is not available.
        SYS0613 not enough memory is available for the cache to start
        Trap-E internal error
11264 - same as 12288
10240 - same as 12288, except that instead of trap-E'ing, it started.
9216 - fine.        


I am now running with the 9 meg disk cache.  And it is working WONDERFULLY.
 Rescans are sub-1-second.  Compiles are 1.5 seconds.

Any ideas about the 10+ meg disk caches?  BFN.  Paul. 


P.S. More info - I created a 1 meg VDISK to compare my compile times to
purely RAM, and I suddenly got the "not enough memory" and trap-e
again!  So I reduced it down to 8 meg, and got my 1 meg VDISK, and found
that both the VDISK and the 8-meg cache, both took 22 seconds to compile my
application (Tobruk)!  Not bad!  The 2-meg HPFS-16 compile was about 25
seconds (but it also got the advantage of having the compiler and include
files etc in the fat cache).

P.P.S. I am running an IBM Cyrix 686-100 (P120)
@EOT:

---