I had written:

>"Do you have any idea
> > whether the problem is CPU overload, or disk?"
> >
> > It's vaguely possible that Windows ME makes incremental memory
> > allocation very expensive ... though it may
> > be that the relatively small reads (1K-8K) are pessimized.

At 10:18 AM 10/9/2000 -0600, Lawrence N. Green wrote:

>That Sounds like it could be a factor,
>Does anyone know what to do about this?

If the problem is memory allocation, then setting the minimum
working set size might help. There are configuration parameters
for that. Unfortunately, it appears that they are used only
under Windows/NT. However, it's a build time check, so running
the NT kit might work. The relevant code is:

In IBConfig:

SERVER_WORKING_SIZE_MIN <n>
SERVER_WORKING_SIZE_MAX <m>

And no, I don't have a clue what the correct values of m and n
are. You might check the size of the image after it's started
returning rows as a basis for n. The value you supply is the
process size in 1024K hunks.

Inside the engine - value1 is min, value2 is max.

if (hndl)
{
ULONG value1 = *value1_ptr * 1024;
ULONG value2 = *value2_ptr * 1024;

if (!SetProcessWorkingSetSize(hndl, value1, value2))

So, if Windows/ME implements SetProcessWorkingSetSize, and if the
Windows/NT kit works there, and if the problem is incremental memory
allocation, this might be a partial solution.

Sorry not to be more certain...

Regards,

Ann