From desktops to smartphones: OpenGL ground-up makeover
The annual Siggraph technical conference is being rigged up, and so is the next-generation OpenGL applications programming interface for graphics that the Khronos Group undertakes to develop from scratch.
Apple, AMD, and Microsoft all have proprietary graphics APIs in developer review. "We're putting out a call for work on a ground-up redesign of OpenGL to be a cross-platform standard for high-efficiency access to graphics and compute chips," Neil Trevett, president of Khronos, an ad hoc graphics industry standards group, told us.
The future spec will be the first to unify the desktop-oriented OpenGL with its embedded cousin, Open GL ES, which is widely used in SoCs for smartphones, tablets, and embedded systems. Graphics chip vendors are already at work on the effort, along with representatives from Apple, Google, and some major computer game developers.
"There's more energy and momentum here than when we first started OpenGL ES," said Trevett. "For the first time, we've engaged the top makers of games engines such as Epic, Unity, and Valve."
Today's OpenGL applications "behave quite differently" on different chip and operating system platforms. "Different drivers take different decisions on how to manage things and dispatch work across processors." The new API developers "want more predictable, explicit control."
The new API will also include tools for developers who lack the sophistication to work directly with OpenGL. "Not every developer has the skills to take advantage of a low-level API, so we will provide a way to access OpenGL that's as easy to use as OpenGL ES is today," he said. "We don't know what those tools are yet, but we are working on them."
Apple is about to release Metal, a proprietary API for the Imagination Technologies graphics cores used in the A-series chips in iPhones and iPads. Microsoft is set to release DirectX 12, its next-generation graphics API for Windows. AMD is about to ship Mantle, a proprietary API for its Radeon graphics chips that promises better performance, generally targeting gamers. "With more mobile, web, cloud, and car platforms becoming 3D capable, developers will go crazy if each has its own API," Trevett said.
The target space for the next OpenGL is very large, ranging from high-end PC computer games to smartphones and embedded systems such as car GPS devices. Therefore, it may use extensions and profiles to provide features only needed for some areas, he said.
On the bleeding edge of graphics, the now famous Oculus VR is also taking part in the OpenGL effort.
Virtual reality goggles like the Oculus Rift "could be very fundamental and another way of consuming graphics," Trevett said. "There are still fundamental latency and accurate-tracking issues, and we have latency in the graphics pipeline we need to reduce to help VR vendors over the next couple years."
The new API also targets multi-core processors and hybrid chips that use CPU and GPU cores. Intel is also a part of the effort, though it has not said whether it sees the new API as a target for its Xeon Phi multi-core x86 chip.
The new OpenGL will work below the level of the AMD-led Heterogeneous Systems Architecture Foundation (HSA), developing interfaces for SoCs using GPU and CPU cores. "Khronos has developer-facing APIs, while HSA is doing the next level down—hardware interfaces and mechanisms."
Eventually, Khronos might work on intermediate-level interfaces to link OpenGL to HSA interfaces.
More than 2 billion devices ship each year using some flavour of the OpenGL, GL ES, or WebGL APIs. Source: Khronos
The Khronos announcement is targeted in part at attendees of this week's Siggraph computer graphics technical conference. Trevett said he expects the event to speak to the interactions between cloud and client-based graphics rendering. "This area will get even more interesting, enabling new use cases."
Another hot topic at Siggraph is visual computing. "We tend to focus on moving from data to imagery, but visual computing is also taking camera images and generating data with image analysis, so vision processing like we see with the Google Project Tango is becoming significant."
- Rick Merritt
|Related Articles||Editor's Choice|