Hi André, There is an optional MPI support since OTB 5.6, you can activate it at configure time (you have to set OTB_USE_MPI=ON when configuring OTB build). This speedup any applications implementing a streamed pipeline (that is, a lot!). Other applications might implement specific MPI related directives in pragmas.
We are working on the MPI support inside LSGRM, however this feature is not currently fully finalized! You can test it, and give us feedback, but keep in mind that its not really finished yet. (By the way, I strongly recommend you to update the code with the last LSGRM commit) Cheers, Rémi Le mercredi 7 septembre 2016 18:11:07 UTC+2, André Stumpf a écrit : > > Hi again, > > I finally launched a test which is still running (smoothly so far). > It seems that parts of the tile-based processing run rather independent > and could be easily parallelized. > I also noticed several directives for what seems to be MPI support: > > +#ifdef OTB_USE_MPI > ... > +#endif > > Is this already functional and if so how do I switch it on. > Something like this? > > cmake -DCMAKE_CXX_FLAGS=OTB_USE_MPI -std=c++11 > > Kind regards, > Andre > > > > > > On Monday, September 5, 2016 at 11:58:30 AM UTC+2, André Stumpf wrote: >> >> Thanks Julien and Remi,..that did the trick: >> >> ./otbcli_LSGRM -help >> This is the GenericRegionMerging application, version 5.6.1 >> This application allows to use the Large Scale Generic Region Merging >> library (LSGRM) and provides currently 3 homogeneity criteria: Euclidean >> Distance, Full Lambda Schedule and Baatz & Schape criterion. >> >> Complete documentation: >> http://www.orfeo-toolbox.org/Applications/GenericRegionMerging.html >> >> Parameters: >> -progress <boolean> Report progress >> MISSING -in <string> Input Image (mandatory) >> MISSING -out <string> [pixel] Ouput Label Image >> [pixel=uint8/uint16/int16/uint32/int32/float/double] (default value is >> uint32) (mandatory) >> -criterion <string> Homogeneity criterion to >> use [bs/ed/fls] (mandatory, default value is bs) >> -criterion.bs.cw <float> Weight for the spectral >> homogeneity (optional, on by default, default value is 0.5) >> -criterion.bs.sw <float> Weight for the spatial >> homogeneity (optional, on by default, default value is 0.5) >> MISSING -threshold <float> Threshold for the >> criterion (mandatory) >> -niter <int32> Maximum number of >> iterations (optional, on by default, default value is 75) >> MISSING -tmpdir <string> Directory for temporary >> files (mandatory) >> -tiling <string> Tiling layout for the >> large scale segmentation [auto/user/none] (mandatory, default value is auto) >> -tiling.user.sizex <int32> Tiles width (mandatory) >> -tiling.user.sizey <int32> Tiles height (mandatory) >> -tiling.user.nfirstiter <int32> Number of first >> iterations (mandatory) >> -memory <int32> Restrict memory use >> (mb) (optional, off by default) >> -inxml <string> Load otb application >> from xml file (optional, off by default) >> >> Examples: >> None >> >> ...I'll run a few tests and report back. >> >> Best, >> André >> >> On Monday, September 5, 2016 at 10:41:11 AM UTC+2, remicres wrote: >>> >>> Fixed >>> >>> Le lundi 5 septembre 2016 10:31:38 UTC+2, Julien Michel a écrit : >>>> >>>> >>>> In lsgrm/include/lsgrmGraphOperations.txx, ligne 527, replace "bool >>>> margin = true)" by "bool margin)" >>>> >>>> Le 05/09/2016 à 10:25, André Stumpf a écrit : >>>> > Hi Rémi, >>>> > >>>> > I tried to recompile with lsgrm in Modules/Remote. I switched on >>>> C++11 >>>> > support but still get the following compilation error: >>>> > >>>> > | >>>> > Infile included >>>> > >>>> from***/OTB/otb/Modules/Remote/lsgrm/include/lsgrmGraphOperations.h:172:0, >>>> > from***/OTB/otb/Modules/Remote/lsgrm/include/lsgrmSplitter.h:4, >>>> > from***/OTB/otb/Modules/Remote/lsgrm/include/lsgrmController.h:4, >>>> > from***/OTB/otb/Modules/Remote/lsgrm/app/otbLSGRM.cxx:18: >>>> > >>>> ***/OTB/otb/Modules/Remote/lsgrm/include/lsgrmGraphOperations.txx:527:45:error:redeclaration >>>> >>>> >>>> > of >>>> > >>>> ‘template<classTSegmenter>voidlsgrm::InsertNodesFromTile(typenameTSegmenter::GraphType&,lsgrm::ProcessingTile&,bool)’may >>>> >>>> >>>> > nothave defaultarguments [-fpermissive] >>>> > ProcessingTile&tile,boolmargin =true) >>>> > | >>>> > >>>> > Do you have any idea what could be the reason? >>>> > >>>> > André >>>> > >>>> > >>>> > >>>> > On Friday, September 2, 2016 at 5:18:28 PM UTC+2, André Stumpf wrote: >>>> > >>>> > Thanks Rémi, this looks promising. I will give it a try next week >>>> > and report back. >>>> > >>>> > Bon weekend, >>>> > André >>>> > >>>> > On Friday, September 2, 2016 at 4:42:30 PM UTC+2, remicres wrote: >>>> > >>>> > Hello André, >>>> > >>>> > You might be interested by the forthcoming LSGRM external >>>> > module, which is available here: >>>> > http://tully.ups-tlse.fr/cressonr/lsgrm/tree/master >>>> > <http://tully.ups-tlse.fr/cressonr/lsgrm/tree/master> in >>>> beta >>>> > version. >>>> > This is a fork of Pierre's repository that is nearly >>>> > operational (see below), and we are still working on it. >>>> > >>>> > This external module in dependent of the GRM external module >>>> > (the classic one). So you must have these two both aside in >>>> > Modules/Remote/ >>>> > >>>> > Keep us updated, your feedback is appreciated :) >>>> > >>>> > Rémi >>>> > >>>> > Le vendredi 2 septembre 2016 11:22:13 UTC+2, André Stumpf a >>>> > écrit : >>>> > >>>> > Thanks for the hint, but mmh, I get a 404 when I try to >>>> > look at the source code. >>>> > http://tully.ups-tlse.fr/lassallep/lss/tree/master >>>> > <http://tully.ups-tlse.fr/lassallep/lss/tree/master> >>>> > >>>> > Any plans to integrate this into OTB? >>>> > >>>> > I saw that there is also a parallel version here: >>>> > http://tully.ups-tlse.fr/lassallep/lsgrm/tree/master >>>> > <http://tully.ups-tlse.fr/lassallep/lsgrm/tree/master> >>>> > But I'm not sure if this comprise the tile-based >>>> processing. >>>> > >>>> > Cheers, >>>> > André >>>> > >>>> > On Friday, September 2, 2016 at 9:15:21 AM UTC+2, Julien >>>> > Michel wrote: >>>> > >>>> > I think you actually need this : >>>> > http://tully.ups-tlse.fr/lassallep/lss >>>> > <http://tully.ups-tlse.fr/lassallep/lss> >>>> > >>>> > Regards, >>>> > >>>> > Julien >>>> > >>>> > Le 01/09/2016 à 18:42, André Stumpf a écrit : >>>> > > Hi, >>>> > > >>>> > > I just managed to compile the remote module for GRM >>>> > segmentation and >>>> > > ran a few tests. >>>> > > Unfortunately the memory consumption is quite >>>> > extraordinary (~16 GB >>>> > > for a 5000x5000 image with 4 bands). >>>> > > Since the paper which the implementation is based >>>> on >>>> > [1] describes a >>>> > > tile-based framework I'm wondering if the >>>> > tiling-merging mechanism is >>>> > > already included in the implementation (and I'm >>>> > doing something wrong) >>>> > > or if the algorithm is meant to be used in >>>> > combination with other >>>> > > tools that handle the tiling-merging? >>>> > > >>>> > > Many thanks in advance for any help, >>>> > > André >>>> > > >>>> > > [1] Lassalle, Pierre, et al. "A Scalable Tile-Based >>>> > Framework for >>>> > > Region-Merging Segmentation." /IEEE Transactions on >>>> > Geoscience and >>>> > > Remote Sensing/ 53.10 (2015): 5473-5485. >>>> > > >>>> > > >>>> > > P.S.: The application help returns a help link >>>> which >>>> > is broken: >>>> > > >>>> > >>>> http://www.orfeo-toolbox.org/Applications/GenericRegionMerging.html >>>> > < >>>> http://www.orfeo-toolbox.org/Applications/GenericRegionMerging.html> >>>> > >>>> > > >>>> > > -- >>>> > > -- >>>> > > Check the OTB FAQ at >>>> > > http://www.orfeo-toolbox.org/FAQ.html >>>> > <http://www.orfeo-toolbox.org/FAQ.html> >>>> > > >>>> > > You received this message because you are >>>> subscribed >>>> > to the Google >>>> > > Groups "otb-users" group. >>>> > > To post to this group, send email to >>>> > [email protected] >>>> > > To unsubscribe from this group, send email to >>>> > > [email protected] >>>> > > For more options, visit this group at >>>> > > http://groups.google.com/group/otb-users?hl=en >>>> > <http://groups.google.com/group/otb-users?hl=en> >>>> > > --- >>>> > > You received this message because you are >>>> subscribed >>>> > to the Google >>>> > > Groups "otb-users" group. >>>> > > To unsubscribe from this group and stop receiving >>>> > emails from it, send >>>> > > an email to [email protected] >>>> > > <mailto:[email protected]>. >>>> > > For more options, visit >>>> > https://groups.google.com/d/optout >>>> > <https://groups.google.com/d/optout>. >>>> > >>>> > >>>> > -- >>>> > Julien MICHEL >>>> > CNES - DCT/SI/AP - BPI 1219 >>>> > 18, avenue Edouard Belin >>>> > 31401 Toulouse Cedex 09 - France >>>> > Tel: +33 561 282 894 - Fax: +33 561 283 109 >>>> > >>>> > -- >>>> > -- >>>> > Check the OTB FAQ at >>>> > http://www.orfeo-toolbox.org/FAQ.html >>>> > >>>> > You received this message because you are subscribed to the Google >>>> > Groups "otb-users" group. >>>> > To post to this group, send email to [email protected] >>>> > To unsubscribe from this group, send email to >>>> > [email protected] >>>> > For more options, visit this group at >>>> > http://groups.google.com/group/otb-users?hl=en >>>> > --- >>>> > You received this message because you are subscribed to the Google >>>> > Groups "otb-users" group. >>>> > To unsubscribe from this group and stop receiving emails from it, >>>> send >>>> > an email to [email protected] >>>> > <mailto:[email protected]>. >>>> > For more options, visit https://groups.google.com/d/optout. >>>> >>>> >>>> -- >>>> Julien MICHEL >>>> CNES - DCT/SI/AP - BPI 1219 >>>> 18, avenue Edouard Belin >>>> 31401 Toulouse Cedex 09 - France >>>> Tel: +33 561 282 894 - Fax: +33 561 283 109 >>>> >>>> -- -- Check the OTB FAQ at http://www.orfeo-toolbox.org/FAQ.html You received this message because you are subscribed to the Google Groups "otb-users" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/otb-users?hl=en --- You received this message because you are subscribed to the Google Groups "otb-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
