Date: Wed, 5 Feb 2003 14:19:22 -0800 From: Bill Huey (Hui) <billh@gnuppy.monkey.org> To: Brent Verner <brent@rcfile.org> Cc: freebsd-java@freebsd.org, "Bill Huey (Hui)" <billh@gnuppy.monkey.org> Subject: Re: patchset 2 report (a love story) Message-ID: <20030205221922.GA3198@gnuppy.monkey.org> In-Reply-To: <20030205121006.GA69850@rcfile.org> References: <20030205121006.GA69850@rcfile.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Feb 05, 2003 at 07:10:06AM -0500, Brent Verner wrote: > After a long 6 hour compile that ended in a failure on a plugin > target, I had a fresh native 1.4.1 image (with HotSpot no less!) > for my 4.7-stable box. I've been running a couple of tomcat apps > pretty heavily for the past few hours, and things seem stable, aside > from a few DEBUG: signal messages[1]. Aside from apparent stability, > the performance improvement over native 1.3.1 is about 600%. > Nice work. That's good to hear. Why do you think the performance is much better in this system verses 1.3.1 ? > On the HotSpot note, is there any work being done on getting 1.3.1's > HotSpot working? AFAIR, it would not build with 1.3.1-patchset-7. I've been doing other work, but I'm probably going to start up work again on this matter. I'd like to hand this work over to somebody in some formal way so that I can move onto other projects and still have things be completely functional. There is a tiny framework I'm doing is going to allow for both HotSpot versions, 1.3.x and 1.4.x to run properly without the use of signal driven ucontext manipulations for both -stable and -current. That'll be good for all of the FreeBSD folks and Alexey can then revert the removal of some critical asserts to the GC system concerning safepoint timeout. Dan Eischen's will have to commit the kernel/UTS related ucontext stuff (don't know what's happened to that) before we can use the signal driven stuff in -current without modification. Currently that's all turned off since it's broken. That'll be the proper interface to a threads ucontext until the thread control structure is pubically available in some formal capacity when KSEs get into place. > [1] one of these is signal 11, so if this is _the_ signal 11ôI'm > familiar with, and you want to know how to duplicate, I could > probably distill a testcase. Pester me so that I'll do this work soon. I've been in RTOS land for a while and having really felt pressure to do any more HotSpot work from anything external to me. :) bill To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-java" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030205221922.GA3198>