AMD has focused on the open standard, cross-platform OpenCL standard for GPU-Compute functionality with support for acceleration on both GPUs and CPUs (including embedded and hand-held devices). OpenCL allows the GPU to help the CPU do the computing or data crunching, to enable faster and more efficient processing.
SiSoftware has posted OpenCL benchmarks for GPU based acceleration and compared it to both CUDA and ATI Stream.
Using the latest OpenCL 1.0 Beta 4, the benchmarks show great results: performance parity with CUDA (in some cases OpenCL is faster) and 50% faster than native CAL/STREAM.
The take away conclusion: There is no reason not to port CUDA code to OpenCL now!
AMD announced the release of the ATI Stream v2.2 SDK with full OpenCL 1.1 support using both CPU and GPU. The new SDK release also delivers expanded support for operating systems, compilers, and additional hardware.
The interesting non-technical parts from the official press release: “Availability of the ATI Stream SDK v2.2 with OpenCL 1.1 support is a great example of how CPU and GPU technology continues to mature and usher in next-generation computing experiences, where voice, touch, gesture and facial recognition capabilities are common, everyday features,” said Patricia Harrell, director of Stream Computing, AMD. “The enhancements in the ATI Stream SDK v2.2 are especially important due to the support for OpenCL 1.1, which is integral to the forthcoming AMD Fusion family of APUs. These tools allow the developer community to take advantage of heterogeneous computing architectures both today and tomorrow.”
AMD today announced the first OpenGL ES 2.0 driver for desktop computing environments specifically to support the WebGL standard for plugin-free 3D content on the web. Apple (Safari), Google (Chrome), Mozilla (Firefox), and Opera (Opera) are the current browser contributors/supporters of the WebGL standard.
The new driver also enables software developers to use desktop PC equipped with Radeon or FirePro graphics (as opposed to embedded systems) to create applications based on OpenGL ES 2.0 for smartphones and tablets e.g. iOS4 and Android devices. This will make it easier for developers to port software applications between PCs and handhelds, avoiding exclusive use emulators or translation layers to get OpenGL ES 2.0 code up and running.
The beta OpenGL ES 2.0 driver is available now and will be supported on all currently available AMD graphics products introduced since 2008, including ATI Radeon desktop, ATI Mobility Radeon, and ATI FirePro professional graphics cards
Updated 11 am Jul 26: From Stumbling Ahead blog:
This is not an emulator or some sort of layer on top of OpenGL, but a full implementation of OpenGL ES accessed through EGL. Why is this important? How will this make your PC experience better? There are three reasons.
First, OpenGL ES on all modern AMD graphics desktops gives developers a unified environment. Game developers can write one 3D pipeline that can run on an HTC Evo, an Apple iPad, and on desktop and laptop PCs.
Second, this change will make web experiences faster and richer in the immediate future. WebGL is based on OpenGL ES. Web browsers will be able to use OpenGL ES directly on AMD hardware instead of having to translate every call to some other API first.
Third, mobile developers can create mobile content first directly on a PC without having to wrangle with SDKs or emulators to make sure their code functions correctly on OpenGL ES. This makes developers lives easier and speeds up the whole development process.
Look for OpenGL ES to make a big impact in bridging the gap with mobile devices and making 3D graphics more accessible.
AMD Professional Workstation Graphics team members will be presenting at a number of courses on accelerated rendering, graphics and visualization at SIGGRAPH 2010.
AMD announced that their Open Physics Initiative now offers game developers the open source, free Bullet Physics as the default rigid body physics system combined with Pixelux’s DMM2 ( Digital Molecular Matter) material physics engine. Developers can now design and interact with rigid body systems familiar to them and easily add DMM objects incrementally enabling them to bend and break based on real physical properties.
The Free PC version of DMM2 has no license fee for development or production deployment and includes all the features of the premium version including GPU acceleration. Free PC DMM2 is expected to be made available shortly to interested developers. All of the Bullet Physics implementations described above can be run on any OpenCL- or DirectCompute-capable platform
gDEBugger is an OpenGL debugger and profiler which runs on Windows, Linux and Mac OS X. The new v5.5 release adds AMD GPU Performance Counters integration, displaying AMD (ATI) graphic hardware and driver performance counters inside gDEBugger’s Performance Graph and Performance Dashboard Views, allowing developers to optimize OpenGL application performance on ATI FirePro and Radeon graphics hardware.
AMD Developer Centeral has posted the ATI Stream OpenCL Technical Overview Video Series. The series of 5 videos provides ATI Stream developers an overview of the OpenCL API and OpenCL C programming language.
Here is a summary of the videos by AMD’s Justin Hensley:
The first production release of ATI Stream SDK with OpenCL 1.0 support is out for Windows XP, Vista, 7 as well as openSuse 11.0 and Ubuntu 9.04. The ATI implementation of OpenCL lets developers use combined CPU and GPU power for accelerating applications. This release supports all FirePro workstation cards, as well as the consumer Radeon HD 4XXX, HD 5XXX, and Mobility HD 4XXX .
What's new in ATI Stream SDK 2.0?
First production release of ATI Stream SDK with OpenCL 1.0 support.
New: Support for OpenCL ICD (Installable Client Driver).
New: Support for atomic functions for 32-bit integers.
New: Microsoft Visual Studio 2008-integrated ATI Stream Profiler performance analysis tool.
Preview: Support for OpenCL / OpenGL interoperability.
Preview: Support for OpenCL / Microsoft DirectX 10 interoperability.
Preview: Support for double-precision floating point basic arithmetic in OpenCL C kernels.
Updated OpenCL runtime to conditionally load ATI CAL runtime libraries to allow execution on compatible CPUs without ATI Catalyst installed.
Updated OpenCL runtime to allow simultaneous use of OpenCL and ATI CAL APIs in a single user application.
Updated cl.hpp from the Khronos OpenCL working group release.
Various OpenCL compiler and runtime fixes and enhancements
SmallptGPU is a small and simple Path Tracer demo written in OpenCL in order to test the performance of this new standard. Path tracing is essentially a form of ray tracing whereby each ray is recursively traced along a path until it reaches a light emitting source where the light contribution along the path is calculated. This recursive tracing helps for solving the lighting equation more accurately than conventional ray tracing (definition courtesy of Wikipedia).
SmallptGPU was originally written for Linux using the ATI OpenCL SDK beta4. But there are now Windows 32 & 64 bit builds in this thread 4th post from top. Since it is OpenCL, the code should work on any platform/implementation.
The following video shows the demo running on a Radeon 4870. You see the progressive rendering raytracing technique in action.
Keep in mind that a Radeon 5970 should be at least 4 times faster. Moreover an OpenCL renderer should scale across as many cards as you can cram onto a board.
The ATI Stream Quarterly Newsletter is now online. With the recent release of the OpenCL GPU Beta as part of the ATI Stream SDK v2.0 Beta Program, this quarterly is packed full of OpenCL information and resources.
Here is a summary of what you will find:
OpenCL CPU+GPU Beta Release
Introductory Tutorial to OpenCL with Benedict Gaster
AMD Developer Inside Track: Introduction to OpenCL with Michael Houston
Image Convolution Using OpenCL – A Step-by-Step Tutorial
OpenCL Tutorial – N-Body Simulation
Spotlight Application: Distributed RC5 Encryption with ATI Stream
AMD and SiSoftware Collaborate on OpenCL Industry Benchmark Suite
Tips and Tricks: Porting CUDA Applications to OpenCL
Coming Soon! OpenCL Technical Overview Video Series
Coming in December! CAPS to release AMD CAL/IL Backend for HMPP
Available Now! ATI Stream Development Platforms from Colfax and Exxact
Developer Training Program: OpenCL Course from VizExperts
Related GPGPU benchmarking suite released: The new Sandra 2010 benchmark suite for GPGPU computing enables testing of ATI Stream, Cuda, OpenCL and DX11 Compute Shaders. SiSoft published some initial benchmarks and the showstopper was the performance of the new Radeon 5870 running OpenCL. Quote from the test results page: “Pummels everything into dust with fantastic performance, power and cost efficiency. The very best!”
FireUser.com is a community resource for CAD, visualization, 3D, video and engineering professionals to learn about the latest acceleration and display technologies and news with a focus on the AMD FirePro workstation graphics line.
