BTree + CHK Inefficiencies

[John Szakmeister]

On Wed, Aug 4, 2010 at 2:55 PM, John Arbash Meinel
<john at arbash-meinel.com> wrote:
[snip]
> So this cuts 152MB out of peak memory (62% peak vs orig, almost half). I
> did time them, but the numbers are within noise margins at this point.
>
> Note that the peak memory numbers get even better for 64-bit machines.
> The data types are fixed between 32/64bit machines, so we should always
> scale at ~43bytes/record. (vs 234 that I measured before, vs ~400+ for
> 64-bit machines.)
>
> I'm still tweaking it a bit, but the memory numbers are definitely there.

This is awesome stuff John.  Will this help with checkout as well?

Great job!

-John