Issue link: https://digital.copcomm.com/i/30784
n n n n Workstations•Graphics systems for the Mac, PC, and Linux with support for as many as 16 GPUs. Rendering companies also Cutting-edge software is taking advantage of the GPU, including GenArts’ Sapphire plug-ins for lighting effects, one of the early plug-in companies with GPU-accelerated products. sRGB color space, and professionals learned to live within their means. Wide-gamut displays extend the number of colors closer to a range defined as Adobe RGB. Portrait Displays’ Mi- chael James develops color management tools for monitors. He says, “Wide gamut is coming faster than we thought. Brand manufacturers want better color to differentiate their prod- ucts on the retail shelf. Today, the cost differ- ence between sRGB and wide gamut is based only on volume. I’m beginning to think that color will be the big change in the computer industry over the next few years.” And, as James indicates, color is fast on its way to becoming a mainstream technology and not one that is reserved for the high-end markets. Professionals in the visual arts fields are definitely going to be moving to wide- gamut displays, and, interestingly enough, many enthusiasts will make the move as well. It’s About the Software Traditionally, software lags hardware consider- ably, causing a vicious cycle for the industry as people wait for their software to catch up with the hardware before they go shopping. The companies building graphics processors have begged the software graphics giants, such as Adobe and Autodesk, to take advantage of their hardware for such tasks as 3D modeling, rendering, image processing, video transcod- ing, and anything else that requires processors to do a whole lot of the same thing over and over. Graphics processors have been put to work simulating weather conditions, and they could be pretty handy for calculating finite- element analysis and fluid dynamics. In general, software companies are slow to make radical changes in their software specifi- cally to accommodate hardware because they don’t want to devote a whole lot of effort to a major project that will only benefit people using the very latest products. Software devel- opers want their software to run on as many systems as possible. That’s always going to be true to a certain extent. But there are changes 32 April 2011 happening presently as the benefits of multi- ple processors become obvious—obvious, but not easy. New programming paradigms have been developed in order to take advantage of the many graphics processors that live on a single graphics chip, and the CPU processors work in tandem with the graphics processor. Nvidia has given the world CUDA, while Apple developed OpenCL and gave it to the Khronos Group for universal deployment. And, Microsoft has developed DirectCom- pute within its DirectX API for Windows- based applications. There are several companies that have been using CUDA in their development. Most of them will tell you that they’re taking advantage of CUDA for two reasons: to get an imme- diate performance boost from Nvidia’s GPUs and to learn how to use GPU processors. 2010 was a good year for putting the GPU to work. Adobe has built the Mercury Engine using CUDA to take advantage of the GPU and speed up Adobe’s After Effects, Photoshop, and Premiere. GPU compute is being paired with 64-bit compute to enable crazy amounts of tasks layered on the time- lines. Further, GPUs speed color correction, special effects, wire removal, transcoding, and a lot more. Adobe is hardly the only vendor work- ing toward this end. For the past few years at IBC, there has been an obvious trend away from dedicated workstations for creative tasks, such as finishing, compositing, and color cor- rection. This year, as the trend continues, the GPU is playing a larger role. Autodesk has taken advantage of the GPU, and the com- pany is moving its high-end system products into software for mainstream workstations. The plug-in companies, meanwhile, have been early in. GenArts’ widely used Sapphire package of lighting effects is available for the Mac version of Smoke, and is GPU-acceler- ated. Blackmagic Design, which acquired da Vinci Systems in 2009, is taking advantage of the GPU for color correction, and is selling have been early comers, taking advantage of multiple proces- sors, be they CPUs or GPUs. Rendering is an inherently par- allel task. Programs are designed to color and shine up each and every pixel in a scene as fast as possible. Upstart renderer V-Ray (Chaos Group) has been gaining ground with its GPU-based rendering tools, and the relentless people at Nvidia have con- verted several companies to using its technolo- gies for GPU rendering using its OptiX or iRay engines. Nvidia has devoted considerable effort into making it easy for software developers to use the company’s technology. Early adopters include Bunkspeed and Lightworks Design, both fast renderers used in CAD. But, a transi- tion is coming as OpenCL comes online. Rendering is actually something of a stalking horse for GPU compute. It’s an obvious-use case, and there is a lot of rendering technology around—open source as well as proprietary. Throughout 2010 and now into 2011, you’ll see new applications taking advantage of GPU compute as a result of what has been learned by the early pioneers of rendering. Up to now, there’s been a lot of emphasis on GPU compute as Nvidia has evangelized the use of its graphics processors for HPC (high- performance computing). At SIGGRAPH 2010, the OpenCL group within Khronos had met several critical requirements for developers. And, AMD has rolled OpenCL into its Stream heterogeneous computing initiative, and it is- sued an SDK and tools this past August. AMD has put considerable weight behind OpenCL because it needs the standard to compete against Nvidia’s CUDA. In reality, the indus- try needs a range of programming resources to take advantage of CPUs and GPUs. OpenCL is being developed to harness both GPUs and CPUs for heterogeneous computing. Microsoft’s DirectX and DirectCompute play a similar role. With most of the com- pute world still working on Windows, there is plenty of reason to develop with Microsoft’s dev tools. With DirectX 11, Microsoft’s Di- rectCompute API also gives developers access to the CPU and GPU, as well. Software developers, in fact, are looking at all the tools available. Autodesk CTO Jeff Kowalski tells us that his company isn’t interested in picking sides. Rather, he sees CUDA, OpenCL, and DirectCompute as