I see Matt has replied, but I'll post this anyway ...

On Mar 6, 6:15 am, "Michael J. Mahon"  wrote:
> The problem with overall speed metrics is that they do not reveal
> how much of the execution time is in the libraries vs. the calling
> code.

True, but in this case both projects are using the same library.
The only difference is the wrappers to interface to the language.

> The table, in fact, suggests that improvements to the graphics
> library were a prime component of the speed changes, and I'm guessing
> that it was not written in Java.

Actually just the opposite, from the dense para above the table:

0.9.2 (85% of C speed, middle column) - "new =93fastjogl=94lib" - the only
lib change listed - "reduces the number of native interface calls" -
which is what the C code is already doing - calling natively. So this
is purely a Java interface optimization.

0.9.3 (98%) - "reducing the number of memory allocations inside the
game loop" - so this huge gain is from Java code changes alone.

0.9.4 (102-106%) - "simple optimizations (vector arrays, optimized
memory handling in loops etc.)" - again Java only changes.

I hope you take a look at the soon to be released Apple Game Server 3.
Brendan has done some amazing (Java!) work there, with a bit of
encouragement from me on the sidelines. ;-)

Cheers,
Nick.
--- SBBSecho 2.12-Win32