Hi,
You asked about benchmarks. One of the x86 porters from Blackdown named
Juergen Kreileder tested both the upcoming Blackdown jdk116_v4 against the
OSF Opengroup's jdk116_v1 (native threads) and Netscape-4.06-glibc on a
Dual PPro 233MHz system. Here are his results:
---cut---
Blackdown OSF Netscape-4.06-glibc
Caffeinemark 3: 448 387 175
+ TYA 1.0: 943 886
Volano Mark 1: 598
I couldn't get OSF + vmark working, it always fails with
40 connections so far.
Creating room number 3 ...
thread creation failed
java.lang.OutOfMemoryError
at COM.volano.x.ł(Unknown Source)
at COM.volano.b.<init>(Unknown Source)
at COM.volano.Mark.ů(Unknown Source)
at COM.volano.Mark.main(Unknown Source)
---cut---
So our green_threads implementation is actually faster than their native
threads routine right now (which I expected given the overhead of context
switches with kernel based threads), but this is just their first release
which means they will improve!
If the earlier thread was correct about the OSF having the swing offset
menu problem, then they used our diffs to generate their port (at least as
a beginning basis). That is a bug I introduced by mistake and was not
fixed until V3.
Given their product is heavily based on the hard work of the Blackdown
team, you would think they would share their diffs. Unfortunately, they
have refused all requests to share diffs. So much for being the OpenGroup
(IMHO).
Kevin B. Hendricks
Linux PowerPC / MkLinux JDK Porting Team
[EMAIL PROTECTED]
ps. By the way, our v4 should come out soon with a couple of importqant bug
fixes and more to come!
----------------------------------------------------------
Kevin B. Hendricks
Associate Professor, Operations & Information Technology
School of Business, College of William & Mary
Williamsburg, VA 23187, [EMAIL PROTECTED]
http://business.tyler.wm.edu
