GPU Computing with Flow Cytometry
1
0
Entering edit mode
@kevin-schiesser-5594
Last seen 9.6 years ago
Hi y'all, Does Bioconductor support GPU/multi-core computing for flow cytometry data analysis? For instance, in the flowMerge package there is a pFlowClust option. When I load the flowMerge library, and call flowClust(), I get the output 'Using the serial version of flowClust'. I have a quad-core processor (Intel(R) Core(TM) i5-3210M CPU @ 2.50GHz) and a geForce GT 650M video card (CUDA enabled) running on Debian Testing 3.2.0-4-amd64. What does pFlowClust look for to choose serial or parallel? I ask because on my lowly laptop the runtime for flowClust(dat, K=1:10) prohibits sound troubleshooting for a flowFrame with 9 stains. I would like to optimize the hardware I have access to, but I am running out of ideas. Any thoughts? Much thanks, Kevin
flowMerge flowMerge • 2.0k views
ADD COMMENT
0
Entering edit mode
@kevin-schiesser-5611
Last seen 9.6 years ago
On 11/15/2012 02:52 PM, Kevin Schiesser wrote: > Hi y'all, > > Does Bioconductor support GPU/multi-core computing for flow cytometry > data analysis? > > For instance, in the flowMerge package there is a pFlowClust option. > When I load the flowMerge library, and call flowClust(), I get the > output 'Using the serial version of flowClust'. I have a quad-core > processor (Intel(R) Core(TM) i5-3210M CPU @ 2.50GHz) and a geForce GT > 650M video card (CUDA enabled) running on Debian Testing > 3.2.0-4-amd64. What does pFlowClust look for to choose serial or > parallel? > > I ask because on my lowly laptop the runtime for flowClust(dat, > K=1:10) prohibits sound troubleshooting for a flowFrame with 9 stains. > I would like to optimize the hardware I have access to, but I am > running out of ideas. Any thoughts? > > Much thanks, > Kevin > Nevermind... I found http://www.findthatdoc.com/search-39323504-hPDF/download-documents- flowclustflowmergeslides-pdf.htm to keep me busy for a while... Thanks Greg! Still, GPU computing would be a very nice option for the flowSuiteOfSoftwares. -k
ADD COMMENT
0
Entering edit mode
Dear Kevin I am no expert of the package you are enquiring about, but in general you might be interested in substituting the generic BLAS on your machine with Intel's MKL (non free, but available with no costs for research via Intel) in order to enable native parallelization of most of the plausible functions, with no need to modify your code. Here are some tests: http://planetflux.adamwilson.us/2012/05/compiling-r-with-intels-math- kernel.html You can get inspiration on the way to install MKL and force its use by R, by reading this blog http://www.r-bloggers.com/compiling-64-bit-r-2-10-1-with-mkl-in-linux/ The use of your GPU is a completely different issue, and despite having a CUDA enabled GPU myself I gave up on it for lack of time and frustration... but maybe you are smarter and luckier than me. You can start reading from this page and then follow the links: http://brainarray.mbni.med.umich.edu/Brainarray/rgpgpu/ I hope this helps. Good luck and best regards, Marco -- Dr Marco Manca University of Maastricht Faculty of Health, Medicine and Life Sciences (FHML) Cardiovascular Research Institute (CARIM) Mailing address: PO Box 616, 6200 MD Maastricht (The Netherlands) Visiting address: UNS40 West building - 5th floor Room5.544, Universiteit Singel 40, 6229 HX Maastricht E-mail: m.manca at maastrichtuniversity.nl Office telephone: +31(0)433884289 Personal mobile: +31(0)626441205 Twitter: @markomanka ********************************************************************** *********************************************** This email and any files transmitted with it are confidential and solely for the use of the intended recipient. It may contain material protected by privacy or doctor-patient /consultant-client privilege. If you are not the intended recipient or the person responsible for delivering to the intended recipient, be advised that you have received this email in error and that any use is STRICTLY PROHIBITED. If you have received this email in error please notify us by telephone on +31626441205 Dr Marco MANCA ********************************************************************** *********************************************** ________________________________________ Da: bioconductor-bounces at r-project.org [bioconductor-bounces at r-project.org] per conto di Kevin Schiesser [schiesserk at medsfgh.ucsf.edu] Inviato: venerd? 16 novembre 2012 0.18 A: kevin.schiesser at ucsf.edu Cc: bioconductor at r-project.org Oggetto: Re: [BioC] GPU Computing with Flow Cytometry On 11/15/2012 02:52 PM, Kevin Schiesser wrote: > Hi y'all, > > Does Bioconductor support GPU/multi-core computing for flow cytometry > data analysis? > > For instance, in the flowMerge package there is a pFlowClust option. > When I load the flowMerge library, and call flowClust(), I get the > output 'Using the serial version of flowClust'. I have a quad-core > processor (Intel(R) Core(TM) i5-3210M CPU @ 2.50GHz) and a geForce GT > 650M video card (CUDA enabled) running on Debian Testing > 3.2.0-4-amd64. What does pFlowClust look for to choose serial or > parallel? > > I ask because on my lowly laptop the runtime for flowClust(dat, > K=1:10) prohibits sound troubleshooting for a flowFrame with 9 stains. > I would like to optimize the hardware I have access to, but I am > running out of ideas. Any thoughts? > > Much thanks, > Kevin > Nevermind... I found http://www.findthatdoc.com/search-39323504-hPDF/download-documents- flowclustflowmergeslides-pdf.htm to keep me busy for a while... Thanks Greg! Still, GPU computing would be a very nice option for the flowSuiteOfSoftwares. -k _______________________________________________ Bioconductor mailing list Bioconductor at r-project.org https://stat.ethz.ch/mailman/listinfo/bioconductor Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor
ADD REPLY
0
Entering edit mode
anyone else tried Andrew Cron's GPUmix wrapper from R? I was under the impression that both Mike West (who I think is Cron's advisor) and Marc Suchard were doing a lot of work applying GPUs to multi-label (i.e. 6, 7, 8, 9+ antibodies) FC. http://www.stat.duke.edu/gpustatsci/software.html http://www.ficcs.org/wp-content/uploads/downloads/2010/12/ficcs.pdf On Thu, Nov 15, 2012 at 11:26 PM, Manca Marco (PATH) < m.manca@maastrichtuniversity.nl> wrote: > Dear Kevin > > I am no expert of the package you are enquiring about, but in general you > might be interested in substituting the generic BLAS on your machine with > Intel's MKL (non free, but available with no costs for research via Intel) > in order to enable native parallelization of most of the plausible > functions, with no need to modify your code. Here are some tests: > > http://planetflux.adamwilson.us/2012/05/compiling-r-with-intels- math-kernel.html > > You can get inspiration on the way to install MKL and force its use by R, > by reading this blog > http://www.r-bloggers.com/compiling-64-bit-r-2-10-1-with-mkl-in- linux/ > > The use of your GPU is a completely different issue, and despite having a > CUDA enabled GPU myself I gave up on it for lack of time and frustration... > but maybe you are smarter and luckier than me. > You can start reading from this page and then follow the links: > http://brainarray.mbni.med.umich.edu/Brainarray/rgpgpu/ > > I hope this helps. Good luck and best regards, > Marco > > > -- > Dr Marco Manca > University of Maastricht > Faculty of Health, Medicine and Life Sciences (FHML) > Cardiovascular Research Institute (CARIM) > > Mailing address: PO Box 616, 6200 MD Maastricht (The Netherlands) > Visiting address: UNS40 West building - 5th floor Room5.544, Universiteit > Singel 40, 6229 HX Maastricht > > E-mail: m.manca@maastrichtuniversity.nl > Office telephone: +31(0)433884289 > Personal mobile: +31(0)626441205 > Twitter: @markomanka > > > > ******************************************************************** ************************************************* > > This email and any files transmitted with it are confidential and solely > for the use of the intended recipient. > > It may contain material protected by privacy or > doctor-patient/consultant-client privilege. If you are not the intended > recipient or the person responsible for > > delivering to the intended recipient, be advised that you have received > this email in error and that any use is STRICTLY PROHIBITED. > > If you have received this email in error please notify us by telephone on > +31626441205 Dr Marco MANCA > > > ******************************************************************** ************************************************* > ________________________________________ > Da: bioconductor-bounces@r-project.org [bioconductor- bounces@r-project.org] > per conto di Kevin Schiesser [schiesserk@medsfgh.ucsf.edu] > Inviato: venerdì 16 novembre 2012 0.18 > A: kevin.schiesser@ucsf.edu > Cc: bioconductor@r-project.org > Oggetto: Re: [BioC] GPU Computing with Flow Cytometry > > On 11/15/2012 02:52 PM, Kevin Schiesser wrote: > > Hi y'all, > > > > Does Bioconductor support GPU/multi-core computing for flow cytometry > > data analysis? > > > > For instance, in the flowMerge package there is a pFlowClust option. > > When I load the flowMerge library, and call flowClust(), I get the > > output 'Using the serial version of flowClust'. I have a quad-core > > processor (Intel(R) Core(TM) i5-3210M CPU @ 2.50GHz) and a geForce GT > > 650M video card (CUDA enabled) running on Debian Testing > > 3.2.0-4-amd64. What does pFlowClust look for to choose serial or > > parallel? > > > > I ask because on my lowly laptop the runtime for flowClust(dat, > > K=1:10) prohibits sound troubleshooting for a flowFrame with 9 stains. > > I would like to optimize the hardware I have access to, but I am > > running out of ideas. Any thoughts? > > > > Much thanks, > > Kevin > > > Nevermind... I found > > http://www.findthatdoc.com/search-39323504-hPDF/download-documents- flowclustflowmergeslides-pdf.htm > to keep me busy for a while... Thanks Greg! > > Still, GPU computing would be a very nice option for the > flowSuiteOfSoftwares. -k > > _______________________________________________ > Bioconductor mailing list > Bioconductor@r-project.org > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor > > _______________________________________________ > Bioconductor mailing list > Bioconductor@r-project.org > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor > -- *A model is a lie that helps you see the truth.* * * Howard Skipper<http: cancerres.aacrjournals.org="" content="" 31="" 9="" 1173.full.pdf=""> [[alternative HTML version deleted]]
ADD REPLY
0
Entering edit mode
Manca, I am not able to compile R 2.15.1 linked to mkl 11.0 on Debian successfully. Do I have to compile flowMerge from source as well? Any further suggestions would be appreciated. Thanks much, Kevin Thus far, I have tried... #get Debian patched source >apt-get source r-base-core #edit /r-base-core-12.15.1/debian/rules to read... 217 dh_testdir 218 R_PAPERSIZE=letter \ 219 R_BROWSER=xdg-open \ 220 PAGER=/usr/bin/pager \ 221 PERL=/usr/bin/perl \ 222 R_UNZIPCMD=/usr/bin/unzip \ 223 R_ZIPCMD=/usr/bin/zip \ 224 R_PRINTCMD=/usr/bin/lpr \ 225 LIBnn=lib \ 226 AWK=/usr/bin/awk \ 227 CC=$(compiler) \ 228 CXX=${cxxcompiler} \ 229 ${fortrancompiler} \ 230 FC=${f95compiler} \ 231 CFLAGS="$(cflags)" \ 232 CXXFLAGS="$(cxxflags)" \ 233 FFLAGS="$(fcflags)" \ 234 FCFLAGS="$(fcflags)" \ 235 CPPFLAGS= \ 236 LDFLAGS="-L/opt/intel/composer_xe_2013.1.117/mkl/lib/intel64/, -Bdirect" \ 237 SHLIB_LDFLAGS="-lpthread" \ 238 MAIN_LDFLAGS="-lpthread" \ 239 MKL_LIB_PATH=/opt/intel/composer_xe_2013.1.117/mkl/lib/intel64/ \ 240 FLIBS= \ 241 rsharedir=/usr/share/R/share \ 242 rincludedir=/usr/share/R/include \ 243 rdocdir=/usr/share/R/doc \ 244 LD_LIBRARY_PATH=/opt/intel/composer_xe_2013.1.117/mkl/lib/intel64/ \ 245 ./configure --prefix=/usr \ 246 --with-cairo \ 247 --with-jpeglib \ 248 --with-readline \ 249 --with-tcltk \ 250 --with-system-bzlib \ 251 --with-system-pcre \ 252 --with-system-zlib \ 253 --mandir=/usr/share/man \ 254 --infodir=/usr/share/info \ 255 --datadir=/usr/share/R/share \ 256 --includedir=/usr/share/R/include \ 257 $(atlas) \ 258 $(lapack) \ 259 --enable-R-profiling \ 260 --enable-R-shlib \ 261 --enable-memory-profiling \ 262 --with-recommended-packages \ 263 --build $(buildarch) \ 264 --with-blas="-L/opt/intel/composer_xe_2013.1.117/mkl/lib/intel64/ -Wl, --start-group /opt/intel/composer_xe_2013.1.117/mkl/lib/intel64/libmkl_gf_lp64.a \ /opt/intel/composer_xe_2013.1.117/mkl/lib/intel64 /libmkl_gnu_thread.a \ /opt/intel/composer_xe_2013.1.117/mkl/lib/intel64/mkl_core.a \ -Wl, --end-group -lgomp -lpthread" \ 265 CLFAGS="-O3 -funroll-loops -march=native" \ 266 CXXFLAGS="-O3 -funroll-loops -march=native" \ 267 FFLAGS="-O3 -funroll-loops -march=native" \ 268 FCFLAGS="-O3 -funroll-loops -march=native" \ 269 --with-lapack When I run the flowMerge function pFlowClust (with snowfall) R is running off BLAS still (evident by giving >lsof | grep blas R 6885 seahorse mem REG 8,2 257648 7621081 /usr/lib/libgslcblas.so.0.0.0 R 6885 seahorse mem REG 8,2 655336 7615156 /usr/lib/libblas/libblas.so.3.0 R 6885 23358 seahorse mem REG 8,2 257648 7621081 /usr/lib/libgslcblas.so.0.0.0 R 6885 23358 seahorse mem REG 8,2 655336 7615156 /usr/lib/libblas/libblas.so.3.0 R 23392 seahorse mem REG 8,2 257648 7621081 /usr/lib/libgslcblas.so.0.0.0 R 23392 seahorse mem REG 8,2 655336 7615156 /usr/lib/libblas/libblas.so.3.0 R 23399 seahorse mem REG 8,2 257648 7621081 /usr/lib/libgslcblas.so.0.0.0 R 23399 seahorse mem REG 8,2 655336 7615156 /usr/lib/libblas/libblas.so.3.0 >top | grep 6885 6885 seahorse 20 0 2095m 1.8g 9012 S 0.3 23.8 0:21.35 R On 11/16/2012 10:33 AM, Tim Triche, Jr. wrote: > anyone else tried Andrew Cron's GPUmix wrapper from R? I was under > the impression that both Mike West (who I think is Cron's advisor) and > Marc Suchard were doing a lot of work applying GPUs to multi-label > (i.e. 6, 7, 8, 9+ antibodies) FC. > > http://www.stat.duke.edu/gpustatsci/software.html > > http://www.ficcs.org/wp-content/uploads/downloads/2010/12/ficcs.pdf > > > > > On Thu, Nov 15, 2012 at 11:26 PM, Manca Marco (PATH) > <m.manca@maastrichtuniversity.nl> <mailto:m.manca@maastrichtuniversity.nl>> wrote: > > Dear Kevin > > I am no expert of the package you are enquiring about, but in > general you might be interested in substituting the generic BLAS > on your machine with Intel's MKL (non free, but available with no > costs for research via Intel) in order to enable native > parallelization of most of the plausible functions, with no need > to modify your code. Here are some tests: > http://planetflux.adamwilson.us/2012/05/compiling-r-with-intels- math-kernel.html > > You can get inspiration on the way to install MKL and force its > use by R, by reading this blog > http://www.r-bloggers.com/compiling-64-bit-r-2-10-1-with-mkl-in- linux/ > > The use of your GPU is a completely different issue, and despite > having a CUDA enabled GPU myself I gave up on it for lack of time > and frustration... but maybe you are smarter and luckier than me. > You can start reading from this page and then follow the links: > http://brainarray.mbni.med.umich.edu/Brainarray/rgpgpu/ > > I hope this helps. Good luck and best regards, > Marco > > > -- > Dr Marco Manca > University of Maastricht > Faculty of Health, Medicine and Life Sciences (FHML) > Cardiovascular Research Institute (CARIM) > > Mailing address: PO Box 616, 6200 MD Maastricht (The Netherlands) > Visiting address: UNS40 West building - 5th floor Room5.544, > Universiteit Singel 40, 6229 HX Maastricht > > E-mail: m.manca@maastrichtuniversity.nl > <mailto:m.manca@maastrichtuniversity.nl> > Office telephone: +31(0)433884289 <tel:%2b31%280%29433884289> > Personal mobile: +31(0)626441205 <tel:%2b31%280%29626441205> > Twitter: @markomanka > > > **************************************************************** ***************************************************** > > This email and any files transmitted with it are confidential and > solely for the use of the intended recipient. > > It may contain material protected by privacy or > doctor-patient/consultant-client privilege. If you are not the > intended recipient or the person responsible for > > delivering to the intended recipient, be advised that you have > received this email in error and that any use is STRICTLY PROHIBITED. > > If you have received this email in error please notify us by > telephone on +31626441205 <tel:%2b31626441205> Dr Marco MANCA > > **************************************************************** ***************************************************** > ________________________________________ > Da: bioconductor-bounces@r-project.org > <mailto:bioconductor-bounces@r-project.org> > [bioconductor-bounces@r-project.org > <mailto:bioconductor-bounces@r-project.org>] per conto di Kevin > Schiesser [schiesserk@medsfgh.ucsf.edu > <mailto:schiesserk@medsfgh.ucsf.edu>] > Inviato: venerdì 16 novembre 2012 0.18 > A: kevin.schiesser@ucsf.edu <mailto:kevin.schiesser@ucsf.edu> > Cc: bioconductor@r-project.org <mailto:bioconductor@r-project.org> > Oggetto: Re: [BioC] GPU Computing with Flow Cytometry > > On 11/15/2012 02:52 PM, Kevin Schiesser wrote: > > Hi y'all, > > > > Does Bioconductor support GPU/multi-core computing for flow > cytometry > > data analysis? > > > > For instance, in the flowMerge package there is a pFlowClust option. > > When I load the flowMerge library, and call flowClust(), I get the > > output 'Using the serial version of flowClust'. I have a quad- core > > processor (Intel(R) Core(TM) i5-3210M CPU @ 2.50GHz) and a > geForce GT > > 650M video card (CUDA enabled) running on Debian Testing > > 3.2.0-4-amd64. What does pFlowClust look for to choose serial or > > parallel? > > > > I ask because on my lowly laptop the runtime for flowClust(dat, > > K=1:10) prohibits sound troubleshooting for a flowFrame with 9 > stains. > > I would like to optimize the hardware I have access to, but I am > > running out of ideas. Any thoughts? > > > > Much thanks, > > Kevin > > > Nevermind... I found > http://www.findthatdoc.com/search-39323504-hPDF/download- documents-flowclustflowmergeslides-pdf.htm > to keep me busy for a while... Thanks Greg! > > Still, GPU computing would be a very nice option for the > flowSuiteOfSoftwares. -k > > _______________________________________________ > Bioconductor mailing list > Bioconductor@r-project.org <mailto:bioconductor@r-project.org> > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor > > _______________________________________________ > Bioconductor mailing list > Bioconductor@r-project.org <mailto:bioconductor@r-project.org> > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor > > > > > -- > /A model is a lie that helps you see the truth./ > / > / > Howard Skipper > <http: cancerres.aacrjournals.org="" content="" 31="" 9="" 1173.full.pdf=""> > [[alternative HTML version deleted]]
ADD REPLY

Login before adding your answer.

Traffic: 832 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6