Re: programming for CUDA processing

[dpoburko]

Hi All,

    This is a really broad question. I am wondering if people are
programming Image J plugins to farm out work to CUDA enabled NVIDIA
graphics cards. I'm just starting to get a feel of what CUDA is and can
do, so I'm wondering if anyone has any positive or negative experiences
with ImageJ and CUDA processing?

Thanks,
Damon

[Johan Henriksson-2]

Damon Poburko wrote:

> Hi All,
>
>    This is a really broad question. I am wondering if people are
> programming Image J plugins to farm out work to CUDA enabled NVIDIA
> graphics cards. I'm just starting to get a feel of what CUDA is and
> can do, so I'm wondering if anyone has any positive or negative
> experiences with ImageJ and CUDA processing?
>
> Thanks,
> Damon

we've been playing with opencl. despite that it is such a close
relative, it is very immature, the compiler has bugs, the profiler
crashes and performance is not always that great. never assume that GPU
code is faster than CPU code; benchmark carefully.

/Johan

[Daniel James White]

There are now java bindings for openCL (which is somewhat like CUDA,  
see http://en.wikipedia.org/wiki/OpenCL )

http://code.google.com/p/javacl/

works on OSX 10.6 at least.

Dan


Begin forwarded message:

> Date:    Wed, 10 Mar 2010 21:04:10 +0100
> From:    Johan Henriksson <[hidden email]>
> Subject: Re: programming for CUDA processing
>
> Damon Poburko wrote:
>> Hi All,
>>
>>   This is a really broad question. I am wondering if people are
>> programming Image J plugins to farm out work to CUDA enabled NVIDIA
>> graphics cards. I'm just starting to get a feel of what CUDA is and
>> can do, so I'm wondering if anyone has any positive or negative
>> experiences with ImageJ and CUDA processing?
>>
>> Thanks,
>> Damon
> we've been playing with opencl. despite that it is such a close
> relative, it is very immature, the compiler has bugs, the profiler
> crashes and performance is not always that great. never assume that  
> GPU
> code is faster than CPU code; benchmark carefully.
>
> /Johan

Dr. Daniel James White BSc. (Hons.) PhD
Senior Microscopist / Image Visualisation, Processing and Analysis
Light Microscopy and Image Processing Facilities
Max Planck Institute of Molecular Cell Biology and Genetics
Pfotenhauerstrasse 108
01307 DRESDEN
Germany

[phone] (German Mobile)
[phone] (Work phone at MPI-CBG)
[phone] (Fax MPI-CBG LMF)

http://www.bioimagexd.net  BioImageXD
http://pacific.mpi-cbg.de                Fiji -  is just ImageJ (Batteries Included)
http://www.chalkie.org.uk                Dan's Homepages
https://ifn.mpi-cbg.de  Dresden Imaging Facility Network
dan (at) chalkie.org.uk
( white (at) mpi-cbg.de )

[Johan Henriksson-2]

that's the one we have switched to now (we had our own bindings before).
unfortunately it doesn't work on linux for us (crashes) but I don't
think it will be too hard to fix if you understand the low-level stuff.

/Johan


On Thu, 2010-03-11 at 10:29 +0100, Daniel James White wrote:

> There are now java bindings for openCL (which is somewhat like CUDA,  
> see http://en.wikipedia.org/wiki/OpenCL )
>
> http://code.google.com/p/javacl/
>
> works on OSX 10.6 at least.
>
> Dan
>
>
> Begin forwarded message:
>
> > Date:    Wed, 10 Mar 2010 21:04:10 +0100
> > From:    Johan Henriksson <[hidden email]>
> > Subject: Re: programming for CUDA processing
> >
> > Damon Poburko wrote:
> >> Hi All,
> >>
> >>   This is a really broad question. I am wondering if people are
> >> programming Image J plugins to farm out work to CUDA enabled NVIDIA
> >> graphics cards. I'm just starting to get a feel of what CUDA is and
> >> can do, so I'm wondering if anyone has any positive or negative
> >> experiences with ImageJ and CUDA processing?
> >>
> >> Thanks,
> >> Damon
> > we've been playing with opencl. despite that it is such a close
> > relative, it is very immature, the compiler has bugs, the profiler
> > crashes and performance is not always that great. never assume that  
> > GPU
> > code is faster than CPU code; benchmark carefully.
> >
> > /Johan
>
> Dr. Daniel James White BSc. (Hons.) PhD
> Senior Microscopist / Image Visualisation, Processing and Analysis
> Light Microscopy and Image Processing Facilities
> Max Planck Institute of Molecular Cell Biology and Genetics
> Pfotenhauerstrasse 108
> 01307 DRESDEN
> Germany
>
> [phone] (German Mobile)
> [phone] (Work phone at MPI-CBG)
> [phone] (Fax MPI-CBG LMF)
>
> http://www.bioimagexd.net  BioImageXD
> http://pacific.mpi-cbg.de                Fiji -  is just ImageJ (Batteries Included)
> http://www.chalkie.org.uk                Dan's Homepages
> https://ifn.mpi-cbg.de  Dresden Imaging Facility Network
> dan (at) chalkie.org.uk
> ( white (at) mpi-cbg.de )
>
>
>
>
>
>
>
>
>
>
>

[Stephan Preibisch]

Graphic Card Programming under Java should(!) be possible natively using the Fragment and Vertex Shader support of Java3D, i.e. using GLSL (OpenGL Shading Language). However, this is of course not CUDA and thus more complicated to design and implement.
But on the plus side it is independent of the Graphic Card Manufacturer, Java based and according to professional developers direct shader programming is much faster than CUDA anyways...

Some Java examples are here:
http://www.koders.com/info.aspx?c=ProjectInfo&pid=MTCTQ59VNG186N9WYAL9UEYQ3E&s=SpotLightState

GLSL Wiki Page:
http://en.wikipedia.org/wiki/GLSL

If somebody digs more into that it would be very interesting to hear the results :)

Stephan P.

-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Johan Henriksson
Sent: Thursday, March 11, 2010 11:15 AM
To: [hidden email]
Subject: Re: programming for CUDA processing

that's the one we have switched to now (we had our own bindings before).
unfortunately it doesn't work on linux for us (crashes) but I don't
think it will be too hard to fix if you understand the low-level stuff.

/Johan


On Thu, 2010-03-11 at 10:29 +0100, Daniel James White wrote:

> There are now java bindings for openCL (which is somewhat like CUDA,  
> see http://en.wikipedia.org/wiki/OpenCL )
>
> http://code.google.com/p/javacl/
>
> works on OSX 10.6 at least.
>
> Dan
>
>
> Begin forwarded message:
>
> > Date:    Wed, 10 Mar 2010 21:04:10 +0100
> > From:    Johan Henriksson <[hidden email]>
> > Subject: Re: programming for CUDA processing
> >
> > Damon Poburko wrote:
> >> Hi All,
> >>
> >>   This is a really broad question. I am wondering if people are
> >> programming Image J plugins to farm out work to CUDA enabled NVIDIA
> >> graphics cards. I'm just starting to get a feel of what CUDA is and
> >> can do, so I'm wondering if anyone has any positive or negative
> >> experiences with ImageJ and CUDA processing?
> >>
> >> Thanks,
> >> Damon
> > we've been playing with opencl. despite that it is such a close
> > relative, it is very immature, the compiler has bugs, the profiler
> > crashes and performance is not always that great. never assume that  
> > GPU
> > code is faster than CPU code; benchmark carefully.
> >
> > /Johan
>
> Dr. Daniel James White BSc. (Hons.) PhD
> Senior Microscopist / Image Visualisation, Processing and Analysis
> Light Microscopy and Image Processing Facilities
> Max Planck Institute of Molecular Cell Biology and Genetics
> Pfotenhauerstrasse 108
> 01307 DRESDEN
> Germany
>
> [phone] (German Mobile)
> [phone] (Work phone at MPI-CBG)
> [phone] (Fax MPI-CBG LMF)
>
> http://www.bioimagexd.net  BioImageXD
> http://pacific.mpi-cbg.de                Fiji -  is just ImageJ (Batteries Included)
> http://www.chalkie.org.uk                Dan's Homepages
> https://ifn.mpi-cbg.de  Dresden Imaging Facility Network
> dan (at) chalkie.org.uk
> ( white (at) mpi-cbg.de )
>
>
>
>
>
>
>
>
>
>
>