[fedora-java] Re: blueMarine and openJDK -- second attempt

[Fabrizio Giudici]

Quick follow-up: I'm trying to grab a Fedora installation DVD. In the meantine,
a guy from JUG Genova was kind enough to try the blueMarine latest snapshot, and
the problem seems not to be there, as you predicted (I'm just waiting to see the
logs to have a confirmation). I'm trying today to close first this issue of
mine; then I'll move on the benchmark stuff. In any case I can anticipate that
the benchmark is completely open source in code and data, I'd just like to write
some more information about how to run it (and maybe trying to reduce as much as
possible the amount of code to download). See you later.

PS I anticipate some unofficial numbers, hoping that the table format is not
screwed out:

	Linux Ubuntu 8.0.4					
W	Java 1.6.0_06		OpenJDK 1.6.0b9		Java 1.5.0_15	
1	0.38	0.28	1.93	0.76	0.44	0.38
2	0.35	0.22	1.44	0.49	0.37	0.33
3	0.34	0.24	1.32	0.58	0.41	0.33
4	0.40	0.25	1.31	0.62	0.41	0.36


These number refers to tests run on the same hardware; the lower the numberm,
the faster the benchmark; three JDKs tested, a pair of numbers for each JDK: the
former for -client, the latter for -server; W means "number of workers" (the
test can be run in parallel, even though tests with 3 and 4 workers are not
meaningful as I'm running on a dual core). You see that where Sun's Java makes
0.38, OpenJDK makes 1.93 (yikes!); with -server the numbers are 0.28 and 0.76
(better, but still 2.5x).