| 1 |
On 10/10/06, David Herron <David.Herron@×××.com> wrote: |
| 2 |
> Because we, at Sun, don't test on Gentoo some on the |
| 3 |
> team were (and are) worried that some niggling incompatibility at the |
| 4 |
> Linux level might cause a behavior difference in Java. |
| 5 |
|
| 6 |
This statement arouses my curiosity. It seems that because you can |
| 7 |
"tweak the Linux level" (meaning: tweak the underlying OS libraries |
| 8 |
and kernel), one would want to test java (compiler and JVM) on this |
| 9 |
platform. |
| 10 |
|
| 11 |
As part of a test suite, you could try and hinder the performance of a |
| 12 |
JVM by modifying compiler flags and kernel parameters. On the same |
| 13 |
token, you could try and maximize the performance (for various goals) |
| 14 |
of a JVM by different modifications. |
| 15 |
|
| 16 |
Then, with these results, a sort of performance matrix could be |
| 17 |
created showing the particular kernel parameters against various |
| 18 |
performance goals (such as flops, or graphics, or EE beans, or servlet |
| 19 |
load, etc). |
| 20 |
|
| 21 |
I am not suggesting Sun should do this work, but I am suggesting that |
| 22 |
Gentoo seems to be an ideal platform (on any/all supported hardware) |
| 23 |
to "tune" java to maximum performance, depending on one's goals. |
| 24 |
|
| 25 |
-gnul |
| 26 |
-- |
| 27 |
gentoo-java@g.o mailing list |
Archives