Since Ken's timings were done, we've improved the performance of both
MemMove and DmWrite. I believe MemMove will now come close to direct
compiler code (there's still the trap overhead). The DmWrite additional
overhead over MemMove is largely unchanged, however, so Ken's general
conclusions still hold.
But to re-iterate a point he makes: even with DmWrite performance still
isn't all that bad, and you should probably only consider copying records
to the dynamic heap if you're in some really tight loop or doing lots
DmWrites.
e.g. the field code does a DmWrite for every character typed, and you don't
ever notice it because typing is so much slower than writing that there's
gobs of time between each character.
--Bob
