Text archives Help
- From: Carson Brownlee <
>
- To:
- Subject: [Manta] Re: Re: Re: Re: Re: Re: Re: Where does one find ... FindManta.cmake?
- Date: Thu, 14 Jan 2010 21:05:12 -0700
thanks for the reply Austin,
ah ok I assumed Optix was implemented with C based CUDA API calls. A
single code path would be desirable though it would have questionable
use if there is already a GPU on the machine (my program is faster than
software mesa but much slower than GPU optimized OpenGL).
Carson
On Thu, 2010-01-14 at 20:43 -0700, arobison wrote:
>
OpenCL is just a front-end language and API. OptiX is similar in that it
>
provides a C API similar to OpenGL and takes, as program input, PTX. PTX is
>
the low level virtual assembly language for the CUDA architecture, though
>
OptiX uses it mostly as an intermediate representation and compiler target,
>
much like LLVM's IR. OpenCL's language can be compiled to PTX, as can C
>
for CUDA. We provide wrappers with the OptiX SDK to produce suitable PTX
>
from C with nvcc. The implementation of a CPU fallback path is independent
>
of the language used to program OptiX: PTX is the input language and is
>
then compiled down, internally, to run on a GPU, or in the future, a CPU.
>
I'm sure you could easily port your program that translates GL calls into
>
Manta calls to translate to OptiX as well. That would enable you to target
>
GPUs and CPUs with a single code path (once the fallback has been
>
implemented).
>
>
~Austin
>
>
On Thu, 14 Jan 2010 16:53:14 -0700, Carson Brownlee
>
<
>
>
wrote:
>
> hi Austin,
>
> I don't suppose you could tell me if they're switching to OpenCL or
>
> writing a separate CPU implementation of Optix? I too would be very
>
> interested in a CPU Optix as I built a program that takes OpenGL calls
>
> and converts them to Manta for rendering on compute clusters.
>
>
>
> thanks,
>
> Carson
>
>
>
> On Thu, 2010-01-14 at 16:48 -0700, Austin Robison wrote:
>
>> A CPU fallback path for OptiX is one of our most requested features;
>
>> we're working on it =)
>
>>
>
>> (GeForce support is coming very soon as well)
>
>>
>
>> ~Austin
>
>>
>
>> On Jan 14, 2010, at 4:15 PM, Carson Brownlee
>
>> <
>
>
>>
>
>> wrote:
>
>>
>
>> > the manta-project requires the same specification of build and source
>
>> > directories as your implementation. There is no make install style
>
>> > installation that would install to /usr/include/manta and /usr/lib or
>
>> > anything of that sort that I know of. This could easily be done if
>
>> > you
>
>> > decide on a default installation location however it's just a matter
>
>
>> > of
>
>> > copying the src folders and build/include folders to a default
>
include
>
>> > directory and then the build/lib over to a default lib directory.
>
>> >
>
>> > as for Optix, this probably isn't much of an option if you're
>
>> > looking at
>
>> > using ray tracing to begin with because it's running on cpu based
>
>> > compute clusters without gpus which I think Dave is looking at.
>
>> > Perhaps that will change if they switch to OpenCL.
>
>> >
>
>> > Carson
>
>> >
>
>> >
>
>> > On Thu, 2010-01-14 at 16:54 -0500, David E DeMarle wrote:
>
>> >> I'll take an out of date version for now.
>
>> >>
>
>> >> I am specifically looking to improve upon what is in vtkManta's
>
>> >> cmakelist.txt.
>
>> >> There the user has to manually specify both the Manta source and
>
>> >> build
>
>> >> directories. My first thought was to add a @DEFINE@ to
>
>> >> vtkMantaConfigure.h.in which tells where the source is (as for vtk),
>
>> >> which would make it so the user only has to specify the build tree,
>
>> >> but y'all might already do soemthing better.
>
>> >>
>
>> >> Two more questions while I have people's attention:
>
>> >> 1) Is there doxygen for the Manta class hierarchy?
>
>> >> 2) Has/does anyone use Manta on windows? We are pursuing changes to
>
>> >> allow that, but if y'all have it working somehow we don't want to
>
>> >> duplicate effort.
>
>> >>
>
>> >> David E DeMarle
>
>> >> Kitware, Inc.
>
>> >> R&D Engineer
>
>> >> 28 Corporate Drive
>
>> >> Clifton Park, NY 12065-8662
>
>> >> Phone: 518-371-3971 x109
>
>> >>
>
>> >>
>
>> >>
>
>> >> On Thu, Jan 14, 2010 at 4:45 PM, Thiago Ize
>
>> >> <
>
>
>> >>
>
>> >> wrote:
>
>> >>> Yeah, we had some svn server changes that ended up mucking with
>
>> >>> permissions.
>
>> >>> Abe would need to make that directory public or Abe could copy
>
>> >>> that into the
>
>> >>> Manta repository. I have a copy of Manta-project on my machine
>
>> >>> that I can
>
>> >>> pass to you, but it might be an older version than what Abe has.
>
>> >>>
>
>> >>> Thiago
>
>> >>>
>
>> >>> David E DeMarle wrote:
>
>> >>>>
>
>> >>>> http://mantawiki.sci.utah.edu/manta/index.php/Manta_project
>
>> >>>>
>
>> >>>> says that there is a sample external project, which includes a
>
>> >>>> FindManta.cmake, here:
>
>> >>>>
>
>> >>>> svn co
>
https://code.sci.utah.edu/svn/people/abe/code/Manta-project/
>
>> >>>> Manta-project
>
>> >>>>
>
>> >>>> but I can't svn checkout that...
>
>> >>>>
>
>> >>>> How do I get it?
>
>> >>>>
>
>> >>>> (hmm, maybe someone could point me to a
>
>> >>>> FindFindManta.cmake.cmake ...)
>
>> >>>>
>
>> >>>> David E DeMarle
>
>> >>>> Kitware, Inc.
>
>> >>>> R&D Engineer
>
>> >>>> 28 Corporate Drive
>
>> >>>> Clifton Park, NY 12065-8662
>
>> >>>> Phone: 518-371-3971 x109
>
>> >>>>
>
>> >>>
Archive powered by MHonArc 2.6.16.