Benchmarks: Just a quick update

Java arithmetic tests are done, and they show that the IBM BigDecimal
performs from 3-7 times slower than RPG packed decimal.  The Sun
BigDecimal performs even worse (nearly 20 times slower on a divide). 

 

The JDBC preliminary tests are under way.  Single record UPDATE WHERE is
about 35 times slower than a native UPDATE opcode, but I was surprised
to find that the updatable cursor is even slower. On my machine, I can
do about 2850 native updates a second.  UPDATE WHERE does about 83
updates a second, and a scrollable cursor slows down to 35 per second,
or about 80 times slower than native I/O.

 

Obviously I'll need some help tuning this code (my hope is that prepared
statements will improve these numbers considerably), but that's pretty
abysmal performance.

 

I'll have the entire suite available for download by the end of the
week.

 

Joe