[rvm-core] New DaCapo regressions
|
View:
New views
8 Messages
—
Rating Filter:
Alert me
|
| < Prev | 1 - 2 | Next > |
 |
Re: [rvm-core] New DaCapo regressions
by Filip Pizlo-2
::
Rate this Message:
Agreed. Also - I don't understand the argument for having to move to
0.98 in order to work on Harmony. Ian, can you elaborate? -F On Feb 11, 2009, at 23:57 , Daniel Frampton wrote: > I have backed out classpath 0.98. > > We really have to give native threads some time to settle in terms of > both stability and performance. Work on classpath 0.98 plus harmony > can go ahead on a branch if it needs to happen right now. > > Cheers, > Daniel. > > On Wed, Feb 11, 2009 at 11:25 PM, Ian Rogers > <ian.rogers@...> wrote: >> 2009/2/11 David P Grove <groved@...>: >>> In my opinion there are too many things broken right now for any >>> forward >>> development to be happening. >>> >>> Either we back out some of the easy to revert changes (eg >>> classpath 0.98, >>> possibly the native thread optimizations) or we put a total code >>> freeze on >>> for anything except bug fixes until some level of stability is >>> restored. >>> >>> Ian, you are underestimating how much time it typically takes for >>> me to get >>> a new version of GNU Classpath working on AIX. An easy upgrade >>> takes a >>> day....some of the more annoying ones in the past have taken a >>> week or more >>> to get straightened out. Turnaround time is a serious issue in >>> debugging >>> this. Each attempt to build classpath takes 60-90 minutes on AIX >>> due to the >>> very slow configure step and this isn't a week where I have a lot >>> of spare >>> cycles. >>> >>> --dave >> >> I won't fix Harmony on Classpath 0.97, fixing the class library is >> tricky enough without worrying about the code being stale. Dave, if >> you can handle the AIX issues it seems the machines aren't sane. I'd >> suggest we just take the performance hit with Classpath 0.98 and bump >> the timeouts. Most of the regressions I see relate to native threads >> in particular RVM-784 that Fil is looking into. I think the priority >> has to be to get RVM-784 closed and Fil has a hopeful fix in r15419. >> >> Ian >> >> ------------------------------------------------------------------------------ >> Create and Deploy Rich Internet Apps outside the browser with >> Adobe(R)AIR(TM) >> software. With Adobe AIR, Ajax developers can use existing skills >> and code to >> build responsive, highly engaging applications that combine the >> power of local >> resources and data with the reach of the web. Download the Adobe >> AIR SDK and >> Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com >> _______________________________________________ >> Jikesrvm-core mailing list >> Jikesrvm-core@... >> https://lists.sourceforge.net/lists/listinfo/jikesrvm-core >> > > ------------------------------------------------------------------------------ > Create and Deploy Rich Internet Apps outside the browser with > Adobe(R)AIR(TM) > software. With Adobe AIR, Ajax developers can use existing skills > and code to > build responsive, highly engaging applications that combine the > power of local > resources and data with the reach of the web. Download the Adobe AIR > SDK and > Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com > _______________________________________________ > Jikesrvm-core mailing list > Jikesrvm-core@... > https://lists.sourceforge.net/lists/listinfo/jikesrvm-core ------------------------------------------------------------------------------ Create and Deploy Rich Internet Apps outside the browser with Adobe(R)AIR(TM) software. With Adobe AIR, Ajax developers can use existing skills and code to build responsive, highly engaging applications that combine the power of local resources and data with the reach of the web. Download the Adobe AIR SDK and Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com _______________________________________________ Jikesrvm-core mailing list Jikesrvm-core@... https://lists.sourceforge.net/lists/listinfo/jikesrvm-core |
|
|
Re: [rvm-core] New DaCapo regressions
by David P Grove
::
Rate this Message:
Ian Rogers <ian.rogers@...> wrote on 02/11/2009 07:25:55 AM: |
|
|
Re: [rvm-core] New DaCapo regressions
by David P Grove
::
Rate this Message:
David P Grove/Watson/IBM@IBMUS wrote on 02/11/2009 08:23:44 AM: |
|
|
Re: [rvm-core] New DaCapo regressions
by Daniel Frampton
::
Rate this Message:
Excellent! That is great news!
I guess this means we should be fine to switch over to classpath 0.98 in a week or two. What do people think? I am hoping that by then we will have a good idea of the stability and performance of native threads. Cheers, Daniel. On Thu, Feb 12, 2009 at 10:03 AM, David P Grove <groved@...> wrote: > David P Grove/Watson/IBM@IBMUS wrote on 02/11/2009 08:23:44 AM: >> >> There's unlikely to be anything deeply "wrong" with the machine >> (outside of a potential /tmp file space problem I fixed last night). >> The issues are almost always that AIX is not Linux, the GNU >> Classpath build process is complex, and we're the only ones who ever >> test it on AIX (as far as I've been able to tell over the years). >> > > And we get lucky...the two fixes Ian and I already made to avoid building > non-essential components plus cleaning junk out of /tmp on piccolo so that > the configure doesn't fail in bizarre ways due to /tmp being 100% full are > enough to get classpath 0.98 to build on AIX and prototype-opt to pass the > basic tests. > > --dave > > ------------------------------------------------------------------------------ > Create and Deploy Rich Internet Apps outside the browser with > Adobe(R)AIR(TM) > software. With Adobe AIR, Ajax developers can use existing skills and code > to > build responsive, highly engaging applications that combine the power of > local > resources and data with the reach of the web. Download the Adobe AIR SDK and > Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com > _______________________________________________ > Jikesrvm-core mailing list > Jikesrvm-core@... > https://lists.sourceforge.net/lists/listinfo/jikesrvm-core > > ------------------------------------------------------------------------------ Create and Deploy Rich Internet Apps outside the browser with Adobe(R)AIR(TM) software. With Adobe AIR, Ajax developers can use existing skills and code to build responsive, highly engaging applications that combine the power of local resources and data with the reach of the web. Download the Adobe AIR SDK and Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com _______________________________________________ Jikesrvm-core mailing list Jikesrvm-core@... https://lists.sourceforge.net/lists/listinfo/jikesrvm-core |
|
|
Re: [rvm-core] New DaCapo regressions
by Ian Rogers (nabble)
::
Rate this Message:
2009/2/11 Daniel Frampton <zyridium@...>:
> Excellent! That is great news! > > I guess this means we should be fine to switch over to classpath 0.98 > in a week or two. What do people think? > > I am hoping that by then we will have a good idea of the stability and > performance of native threads. > > Cheers, > Daniel. I'd really hope we were on Classpath 0.98 now and that it hadn't been backed out. The problem here is that PNT has regressed us all over the place, not Classpath (other than a performance regression on luindex). The worst regression at the moment is Harmony. Ideally we'd have a dual implementation for the PNT code and then we could be running green threads along side PNT - I know there are many reasons for this but history and for flexibilities sake changes should whenever possible be purely additive (ie not take away - like green thread ability has now gone). This would also mean less work for people reliant on the green thread code (like me). The class library interface is made up of: Classpath + patch set + stuff we override in library interface for Classpath + stuff we override in library interface for Common For Harmony the same is true just substitute Classpath for Harmony. With 0.97 the patch set is huge, making it hard to work with. What we override in library interface common is shared between Classpath and Harmony. The prospect of having to work not on two sets of code (made up of 4 components) but three sets is completely unappealing. It would be like working on a source code repository that you knew had changed without first running "svn update". What's also annoying is that twice in as many weeks we've talked about (and have) regressing the code base because of AIX. AIX is becoming something of a barrier to forward progress. The faster we can get a machine in the open to test on the better. I think its worth reiterating the case for Harmony. Classpath is incomplete and won't run SPEC JVM 2008. It also lacks performance (hence our contributions to Classpath in this area). It is not sensible for Jikes RVM developers to also take on the challenge of maintaining Classpath. Harmony needs to be our default class library implementation, the sooner the better. We have cared more in the PNT merge about AIX than having a performant, conformant VM on what is the architecture used by the majority of people. I have a strong suspicion that Harmony won't build on AIX (this seems to be the case for a majority of software) so it will be a barrier to forward progress again. Anyway, PNT shouldn't have been merged without testing Harmony, doing so was short sighted. There seems little point fixing Harmony on a branch and then waiting weeks to merge Classpath 0.98 and Harmony with the trunk. We're failing Harmony regression tests on trunk at the moment. It will be a hassle to move these tests on to a branch. We're already out of sync on other branches, like quarantine. I'm concerned currently because I expect Daniel and Fil's cycles to work on this to reduce drastically next week. A merge has been pushed in and then left for other people to clean up. Ian > On Thu, Feb 12, 2009 at 10:03 AM, David P Grove <groved@...> wrote: >> David P Grove/Watson/IBM@IBMUS wrote on 02/11/2009 08:23:44 AM: >>> >>> There's unlikely to be anything deeply "wrong" with the machine >>> (outside of a potential /tmp file space problem I fixed last night). >>> The issues are almost always that AIX is not Linux, the GNU >>> Classpath build process is complex, and we're the only ones who ever >>> test it on AIX (as far as I've been able to tell over the years). >>> >> >> And we get lucky...the two fixes Ian and I already made to avoid building >> non-essential components plus cleaning junk out of /tmp on piccolo so that >> the configure doesn't fail in bizarre ways due to /tmp being 100% full are >> enough to get classpath 0.98 to build on AIX and prototype-opt to pass the >> basic tests. >> >> --dave >> >> ------------------------------------------------------------------------------ >> Create and Deploy Rich Internet Apps outside the browser with >> Adobe(R)AIR(TM) >> software. With Adobe AIR, Ajax developers can use existing skills and code >> to >> build responsive, highly engaging applications that combine the power of >> local >> resources and data with the reach of the web. Download the Adobe AIR SDK and >> Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com >> _______________________________________________ >> Jikesrvm-core mailing list >> Jikesrvm-core@... >> https://lists.sourceforge.net/lists/listinfo/jikesrvm-core >> >> > > ------------------------------------------------------------------------------ > Create and Deploy Rich Internet Apps outside the browser with Adobe(R)AIR(TM) > software. With Adobe AIR, Ajax developers can use existing skills and code to > build responsive, highly engaging applications that combine the power of local > resources and data with the reach of the web. Download the Adobe AIR SDK and > Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com > _______________________________________________ > Jikesrvm-core mailing list > Jikesrvm-core@... > https://lists.sourceforge.net/lists/listinfo/jikesrvm-core > ------------------------------------------------------------------------------ Create and Deploy Rich Internet Apps outside the browser with Adobe(R)AIR(TM) software. With Adobe AIR, Ajax developers can use existing skills and code to build responsive, highly engaging applications that combine the power of local resources and data with the reach of the web. Download the Adobe AIR SDK and Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com _______________________________________________ Jikesrvm-core mailing list Jikesrvm-core@... https://lists.sourceforge.net/lists/listinfo/jikesrvm-core |
|
|
Re: [rvm-core] New DaCapo regressions
by David P Grove
::
Rate this Message:
I don't think there is any debate about the long term strategic goals; we're getting caught up in the staging and trivial details. |
|
|
Re: [rvm-core] New DaCapo regressions
by David P Grove
::
Rate this Message:
|
|
|
Re: [rvm-core] New DaCapo regressions
by Ian Rogers (nabble)
::
Rate this Message:
2009/2/12 David P Grove <groved@...>:
> I don't think there is any debate about the long term strategic goals; we're > getting caught up in the staging and trivial details. > > (a) Native threads will be the preferred choice for almost all of our users. > It will enable more programs to be run on Jikes RVM since it sidesteps all > of the problems with native code & thread models; it should also yield > higher performance on SMP/multi-core machines since the OS will understand > what the JVM is doing it should be a simpler and easier to maintain VM > implementation once the various cleanups get completed. > (b) Harmony is likely to be the primary class library for almost all of our > users in the medium term. Long term, I'd expect Open JDK to be the class > libraries of choice. Sure, it's a nightmare to build. > (c) linux-ia32 is the primary platform for almost all of our users. I think this is increasingly x86 64. I also think Windows IA32 may be more appealing than Linux IA32. > Rather then spend lots of time arguing about which one of half a dozen > important things gets fixed today vs. fixed in 3 days, can we just let the > people who are doing the work do it. > > --dave Sure, I was in the process of fixing Harmony against trunk then people started backing out commits. The move to Classpath 0.98 wasn't just a delete and bump versions change, RVM-432 needed rewriting (as logged by JIRA). Imo temporarily increasing the luindex timeout would have been more expedient than moving development to a branch that's outside of the regression test infrastructure. Ian ------------------------------------------------------------------------------ Create and Deploy Rich Internet Apps outside the browser with Adobe(R)AIR(TM) software. With Adobe AIR, Ajax developers can use existing skills and code to build responsive, highly engaging applications that combine the power of local resources and data with the reach of the web. Download the Adobe AIR SDK and Ajax docs to start building applications today-http://p.sf.net/sfu/adobe-com _______________________________________________ Jikesrvm-core mailing list Jikesrvm-core@... https://lists.sourceforge.net/lists/listinfo/jikesrvm-core |
| < Prev | 1 - 2 | Next > |
| Free embeddable forum powered by Nabble | Forum Help |
