December 2009 - Isis-fish-devel - mailman.cloud.codelutin.com

Fwd: Re: question: java
by Eric Chatellier 04 Dec '09

04 Dec '09

-------- Message original -------- Sujet: Re: question: java Date : Wed, 02 Dec 2009 14:30:51 +0100 De : Tina ODAKA <Tina.Odaka(a)ifremer.fr> Organisation : IFREMER Pour : Eric Chatellier <chatellier(a)codelutin.com>, Benjamin Poussin <poussin(a)codelutin.com>, Jean Couteau <couteau(a)codelutin.com> Copie à : bpre(a)ifremer.fr <bpre(a)ifremer.fr>, Marc Simon <smarc(a)sgi.com>, COTTY <Pierre.Cotty(a)ifremer.fr>, aboudot(a)sgi.com, MAHEVAS <Stephanie.Mahevas(a)ifremer.fr> p.s. for your problem of that you can only use '2G'. i started to talk with colleagues from sgi so that you can 'write' on memory space, by using '/tmpfs' we did not do enough test yet on this subject, but once the tests are done, i will let you know where you can 'write' your matrix data you need but which exeed 2G so that you have to write on 'disk' regards, tina Tina ODAKA a écrit : > hello. > > I went to see the people from intel what they think about your idia > that C is faster on intel machine than java code. > > they say, 10 years ago, may be it was true. > but with their new SSE4.2 system and icc compiler (for 'c', or ifort > compiler for fortran) > now it is not true anymore, since with SSE you can for example multiple > 4 values at a time, but with java compiler that is not possible. > > regards, tina > > Eric Chatellier a écrit : >> Le 12/11/2009 09:49, Tina ODAKA a écrit : >>> hello eric jean benjamin, >> Hi, >> >> Sorry for the long response delay :-( >>> i was talking with sgi about that your code is slow. >>> they asked me one thing >>>> JAVA : >>>> Ce language se base sur une machine virtuelle qui fait la génération >>>> des instruction à la volée. Les langages interprétés ne sont pas >>>> performant par définition. Il doit être possible de générer une >>>> version "compilée" du code Java qui doit apporter de meilleurs >>>> performances. >>> i guess your code is 'version "compilée"', isn't it? >> IsisFish code is "pre-compiled" into java bytecode [1]. >> This bytecode is interpreted by the Java virtual machine at runtime. >> >> Next, Java use JIT compilation [2], to improve runtime performance. >> >> So, Java isn't slow, and run as fast as native compiled programs, >> or maybe even faster ;-) >>> tina >> [1] : http://en.wikipedia.org/wiki/Bytecode >> [2] : http://en.wikipedia.org/wiki/Just-in-time_compilation >> > -- =================================================== Tina Odaka RIC - IDM - IFREMER Tel: +33 (0)2 98 22 41 85 Fax: +33 (0)2 98 22 45 46 email: Tina.Odaka(a)ifremer.fr http://www.ifremer.fr/pcim ==================================================

1 0

[Fwd: Re: Problem with ISIS simulations on Caparmor]
by Jean Couteau 04 Dec '09

04 Dec '09

1 0

Problem with ISIS simulations on Caparmor
by Jean Couteau 04 Dec '09

04 Dec '09

Tina, Last friday, I launched 3 simulations plans of 1500 jobs each (1multi-job of 1500 sub-jobs each). The first one went to the end, but the next two were either killed in the middle or never started. Do you know why those jobs were killed or did not started ? And is there a way so that we can do such things without any problems. Best regards Jean Couteau

2 3

Suppression du dossier "isis-export"
by Eric Chatellier 02 Dec '09

02 Dec '09

Bonjour, À partir de IsisFish 3.3.0.0 le dossier "isis-export" (situé a coté de isis-database-3) n'existera plus. Il ne semblait pas être utilisé pas beaucoup de monde, et fesait doublon avec les exports dans chaque simulation ce qui doublait l'espace disque utilisé. Concernant le dev, le dossier "isis-export" servait aussi a stocker les résultats : - de la seconde passe de l'analyse de sensibilité - les fichier .Rdata Peut-on les déplacer dans "isis-database-3/simulations" ? Pour la seconde passe ça fera apparaître en plus une simulation qui n'en ai pas vraiment une dans la liste des anciennes simulations. -- Éric <chatellier(a)codelutin.com> Tel: 02 40 50 29 28 http://www.codelutin.com

3 4

Re: [Isis-fish-devel] [Rd] Bug in R evaluating a huge instruction (PR#14096)
by Jean Couteau 02 Dec '09

02 Dec '09

> FWIW the restriction is only on the console input - if you send it > directly, there is no limit (I have never heard of nuiton-j2r but if > you use rJava to interface from Java to R there is no limit - the > string you parse+evaluate can be of any length). Hi Simon, nuiton-j2r is in fact an abstraction layer to rjava or rserve so that it can switch between one or other implementation using simple configuration (http://maven-site.nuiton.org/nuiton-j2r/ sorry but this is in french). And in fact I encounter the problem using it. What is done is basically creating the R instruction from java and sending it to R using the eval method from the Rengine object (as we use jri in this specific application). The R instruction should create an object a of class morris, and I try then to get back the a$X object (the X item of the morris object) but I get back a null object without any reason, that is why I tried copying the instruction directly in R and thus getting this problem. So that is strange if you say that there is no limit using rjava/jri. Cheers, Jean > > Cheers, > Simon > > >> Duncan Murdoch wrote: >>> On 01/12/2009 12:50 AM, Jean Couteau wrote: >>>> Thanks for your time Duncan, >>>> >>>> I join here the instruction that is not correct, hoping that might >>>> help you. The file is encoded in utf-8 so you should not have any >>>> problem reading it. >>>> >>>> I doubt to that it is an R bug too, but with all my tests i am less >>>> and less sure of that. >>> >>> I have no problem sourcing that file, but I do get an error if I try >>> to cut and paste it, because it's a single line of 4890 characters, >>> and that's too long. So this is an R limitation, but it's one with >>> an easy workaround: just add some line breaks into your source. >>> >>> Duncan Murdoch >>> >>> >>>> >>>> Best regards, >>>> Jean Couteau >>>>> Your message has encoding problems, so it's not readable. Could >>>>> you put the code online somewhere where we could download it in >>>>> its original form? I doubt if this is an R bug, but I can't point >>>>> out the problem in your code (or confirm that it really is an R >>>>> bug) without an undamaged copy of the code. >>>>> >>>>> Duncan Murdoch >>>>> >>>>> >>>>> >>>> >>> >>> >> >> ______________________________________________ >> R-devel(a)r-project.org mailing list >> https://stat.ethz.ch/mailman/listinfo/r-devel >> >> > >

1 0

Re: [Isis-fish-devel] [Rd] Bug in R evaluating a huge instruction (PR#14096)
by Jean Couteau 02 Dec '09

02 Dec '09

Thank you very much for your input Duncan and Romain. I will try to do what you are advising. The R instruction is generated from java using jri (and an abstraction layer we built upon it : nuiton-j2r), i will try add this limitation to nuiton-j2r so that it split the instruction if it is too long. Best regards, Jean Duncan Murdoch wrote: > On 01/12/2009 12:50 AM, Jean Couteau wrote: >> Thanks for your time Duncan, >> >> I join here the instruction that is not correct, hoping that might >> help you. The file is encoded in utf-8 so you should not have any >> problem reading it. >> >> I doubt to that it is an R bug too, but with all my tests i am less >> and less sure of that. > > I have no problem sourcing that file, but I do get an error if I try > to cut and paste it, because it's a single line of 4890 characters, > and that's too long. So this is an R limitation, but it's one with an > easy workaround: just add some line breaks into your source. > > Duncan Murdoch > > >> >> Best regards, >> Jean Couteau >>> Your message has encoding problems, so it's not readable. Could you >>> put the code online somewhere where we could download it in its >>> original form? I doubt if this is an R bug, but I can't point out >>> the problem in your code (or confirm that it really is an R bug) >>> without an undamaged copy of the code. >>> >>> Duncan Murdoch >>> >>> >>> >> > >

1 0

Bug in R evaluating a huge instruction
by Jean Couteau 01 Dec '09

01 Dec '09

Hello, I do not know if i am facing a bug or a mistake (and if it is a bug i do not know in which category I can put it on the bug tracker), so I post my message here. I am trying to use the Morris function from the sensitivity package (but that is not the problem). I created a (big) instruction that should generate a morris object in which there is a data.frame i am interested in. My instruction is : a<-morris(model=NULL,factors=c("PecheAmateurs décembre.proportionMetier","PechePetitMetierMer mai.proportionMetier","PecheAmateurs mai.proportionMetier","PecheChalutiere janvier.minInactivityDays","PechePetitMetierEtang octobre.minInactivityDays","PechePetitMetierEtang décembre.minInactivityDays","FiletMer.standardisationFactor","PecheAmateurs juin.minInactivityDays","PechePetitMetierEtang novembre.minInactivityDays","PechePetitMetierMer septembre.minInactivityDays","PechePetitMetierMer avril.minInactivityDays","PecheAmateurs août.proportionMetier","PecheChalutiere novembre.minInactivityDays","PechePetitMetierMer février.proportionMetier","PecheChalutiere juillet.minInactivityDays","PechePetitMetierMer octobre.proportionMetier","CSar Groupe 5.reproductionRate","PechePetitMetierMer août.proportionMetier","PechePetitMetierMer janvier.minInactivityDays","PechePetitMetierEtang mai.minInactivityDays","PecheAmateurs octobre.minInactivityDays","PecheAmateurs septembre.minInactivityDays","PecheChalutiere juin.minInactivityDays","PecheChalutiere avril.minInactivityDays","PecheAmateurs juillet.proportionMetier","PecheChalutiere mai.minInactivityDays","PechePetitMetierEtang septembre.minInactivityDays","PechePetitMetierMer janvier.proportionMetier","PecheAmateurs septembre.proportionMetier","PechePetitMetierEtang février.minInactivityDays","PechePetitMetierEtang mars.minInactivityDays","PechePetitMetierEtang juillet.minInactivityDays","PechePetitMetierMer septembre.proportionMetier","CSar.capturability","PechePetitMetierMer juin.proportionMetier","PecheChalutiere mars.minInactivityDays","PecheChalutiere septembre.minInactivityDays","PecheAmateurs janvier.minInactivityDays","PechePetitMetierMer août.minInactivityDays","PechePetitMetierMer juillet.proportionMetier","PecheChalutiere décembre.minInactivityDays","PecheAmateurs avril.minInactivityDays","PecheAmateurs mars.proportionMetier","PechePetitMetierMer novembre.minInactivityDays","PechePetitMetierEtang août.minInactivityDays","CSar.recruitmentDistribution","PecheChalutiere février.minInactivityDays","CSar.naturalDeathRate","PecheAmateurs octobre.proportionMetier","PechePetitMetierMer mai.minInactivityDays","PechePetitMetierMer mars.proportionMetier","PecheAmateurs août.minInactivityDays","FiletEtang.standardisationFactor","PecheChalutiere août.minInactivityDays","PecheAmateurs janvier.proportionMetier","PecheAmateurs décembre.minInactivityDays","PechePetitMetierEtang avril.minInactivityDays","PecheAmateurs mars.minInactivityDays","PechePetitMetierMer décembre.minInactivityDays","PecheAmateurs mai.minInactivityDays","PechePetitMetierEtang juin.minInactivityDays","Harpon.standardisationFactor","PecheAmateurs novembre.minInactivityDays","PecheAmateurs novembre.proportionMetier","PechePetitMetierMer mars.minInactivityDays","Ligneemb.standardisationFactor","PecheAmateurs juin.proportionMetier","PecheAmateurs avril.proportionMetier","PechePetitMetierMer juillet.minInactivityDays","PechePetitMetierMer avril.proportionMetier","PechePetitMetierMer octobre.minInactivityDays","PechePetitMetierMer février.minInactivityDays","PecheAmateurs juillet.minInactivityDays","PechePetitMetierMer décembre.proportionMetier","PecheAmateurs février.proportionMetier","PecheChalutiere octobre.minInactivityDays","CSar Groupe 4.reproductionRate","CSar Groupe 7.reproductionRate","Lignecote.standardisationFactor","CSar.meanWeight","CSar Groupe 6.reproductionRate","PechePetitMetierMer juin.minInactivityDays","PechePetitMetierMer novembre.proportionMetier","PechePetitMetierEtang janvier.minInactivityDays","PecheAmateurs février.minInactivityDays","CSar.growth"),r=10,design=list(type="oat",levels=c(5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5),grid.jump=c(2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2)),binf=c(0.0,0.0,0.0,11.0,10.0,12.0,0.06,10.0,11.0,0.0,0.0,0.0,10.0,0.0,11.0,0.0,1120.0,0.0,10.0,22.0,12.0,10.0,10.0,11.0,0.0,11.0,15.0,0.0,0.0,23.0,16.0,22.0,0.0,0.0,0.0,11.0,10.0,18.0,0.0,0.0,11.0,11.0,0.0,3.0,22.0,0.0,8.0,0.0,0.0,0.0,0.0,8.0,0.1456,11.0,0.0,18.0,17.0,11.0,7.0,11.0,20.0,0.0464,16.0,0.0,1.0,0.1515,0.0,0.0,0.0,0.0,1.0,11.0,9.0,0.0,0.0,11.0,1120.0,1120.0,0.0749,0.0,1120.0,0.0,0.0,13.0,16.0,0.0),bsup=c(1.0,1.0,1.0,15.0,14.0,16.0,0.09,14.0,15.0,2.0,2.0,1.0,14.0,1.0,15.0,1.0,1680.0,1.0,14.0,26.0,16.0,14.0,14.0,14.0,1.0,15.0,19.0,1.0,1.0,27.0,20.0,26.0,1.0,1.0,1.0,15.0,14.0,22.0,2.0,1.0,15.0,15.0,1.0,7.0,26.0,1.0,12.0,1.0,1.0,2.0,1.0,12.0,0.2184,15.0,1.0,22.0,21.0,15.0,11.0,15.0,24.0,0.0696,20.0,1.0,5.0,0.22,1.0,1.0,2.0,1.0,5.0,15.0,13.0,1.0,1.0,15.0,1680.0,1680.0,0.11244,1.0,1680.0,2.0,1.0,17.0,20.0,1.0)) I checked several times and this intruction seams to have the right syntax but R asks me to finish my instruction (I need to add two parenthesis at the end so that it can be evaluated, but do not give me the right results). I replaced the factors vector by its size (the morris method allows that) and the evaluation is ok (with right results). I tryed also replacing the binf and bsup vectors by a single value (authorized by the morris function also), leaving the factors vector, and the evaluation is also ok (so it is not the factors vector that is creating the problem). This seems really odd to me and I cannot see where is the problem so I am thinking about a R bug (even if the sensitivity package is not loaded, i cannot enter the instruction, thus it is not a sensitivity problem). Did somebody already encountered such a bug (if it is a bug) ? I am using R 2.9.2 under linux/Xubuntu 9.10 and this has also been tried on R 2.9.0 on Windows 7. I can provide more info if needed. Jean Couteau Code Lutin

2 1