AMD OpenCL Presentations: SPH and Device Fission

OpenCL logo

AMD has published two OpenCL presentations about smoothed particle hydrodynamics (SPH) and a new extension called cl_khr_device_fission:

1 – Smoothed Particle Hydrodynamics

You can download the presentation here: Smoothed Particle Hydrodynamics, Application Example – 31-page PDF, .

AMD OpenCL whitepapers: Smoothed Particle Hydrodynamics

AMD OpenCL whitepapers: Smoothed Particle Hydrodynamics

2 – Device Fission Extension

You can download the presentation here: Device Fission Extensions for OpenCL – 26-page PDF.

Detailed specification of the device fission extension can be found here: cl_ext_device_fission @ OpenCL resgistry.

This extension provides an interface for sub-dividing an OpenCL device
into multiple sub-devices. There are a number of cases in which a typical
user would like to subdivide a device:

1. To reserve part of the device for use for high priority /
latency-sensitive tasks
2. To more directly control the assignment of work to individual compute
3. To subdivide compute devices along some shared hardware feature like a

Typically these are areas where some level of additional control is required
to get optimal performance beyond that provided by standard OpenCL 1.1 APIs.
Proper use of this interface assumes some detailed knowledge of the devices
in question.

AMD OpenCL whitepapers: Device Fission Extension

AMD OpenCL whitepapers: Device Fission Extension

AMD OpenCL whitepapers: Device Fission Extension



  • Ahmad

    Is that mean AMD is working on OpenCL based physx?

  • @Ahmad: Bullet is amd supported opensource physic engine (they added opencl to it.)

  • AMD’s Ho talked about status quo of OpenCL in Bullet last week:
    “The plan for OpenCL rigid body support as shown seen in the OpenCL branch is something we are still discussing. The current state of it is experimental rather than production.

    However, I have just committed some further updates to the OpenCL cloth support to include interop with OpenGL. This aspect of OpenCL support is very much in production code.”

    Also there is an SVN update for Bullet @

  • Stefem

    Bullet is also supported by NVIDIA, the Bullet devs has implemented some OpenCL based solver not AMD

  • mike

    AMD has vowed to keep any of it’s physics implementations with it’s cards open sourced and not proprietory at all. AMD is a very strong supporter of Bullet and soon Bullet will run on all video cards and PhysX still only on Nvidia. I wonder which will die off and which will be embraced by programmers? … hmmmm; because ultimately, if you could choose half a market or a full market, you choose the full market. I can see the end of the road for PhysX already.

  • mareknr

    Mike: Can you write here some web link to support your arguments? As I know leader of Bullet said that he didn’t know about AMD support and that NVIDIA is great technology partner so where did you read this:

    “AMD is a very strong supporter of Bullet”

    Eric Coumans(author of Buller) said this:

    “Bullet’s GPU acceleration via OpenCL will work with any compliant drivers, we use NVIDIA GeForce cards for our development and even use code from their OpenCL SDK, they are a great technology partner.”

  • mareknr

    You cabn read some more info here:

  • I don’t think AMD is working in it’s own version of physx, the SPH implementation is to show the power of OpenCL against CUDA I guess.