Consumers aren't likely to be talking about the work of the Heterogeneous System Architecture (HSA) Foundation. But the group of computing leaders say that a new specification announced today will generate much better computing performance for a new generation of electronic devices.
The HSA says its 1.0 specification for a new design for processing data will lead to faster and more power-efficient personal computers, tablets, smartphones, and cloud servers. It goes hand in hand with hUMA, a new way for processors to access the memory inside an accelerated processing unit (APU), a single chip that combines both a microprocessor and graphics.
The problem is that it isn't easy for programmers to harness the power of the graphics processing unit (GPU) inside an APU. The HSA has been designed to fix this problem, making graphics an equal partner with the CPU (central processing unit) and other processors, such as a digital signal processor, inside a computing system.
Different types of processors have in the past required separate chips. AMD wants to simplify this by enabling them to be packaged inside the same system-on-chip (SoC).
"Through HSA, we are working to ensure that end users of technology live in a world of new, incredible applications that run fast at low power," said Phil Rogers, president of the HSA Foundation, in a statement. "The Foundation members have been collaborating on this project since we joined together in June 2012, and we are thrilled to be delivering the fruit of that labor today."
GPUs can be used for non-graphics computing tasks, but it often takes too long to route requests for data through a CPU. Most developers don’t want to deal with the difficulty of optimizing their code for this kind of work. But a new technique, dubbed “heterogenous queuing,” allows applications to directly communicate with the GPU, treating it as an equal partner alongside a CPU when it comes to accessing data quickly. That means an application won’t have to wait for the CPU when what it really needs to is to access the GPU.
With HSA and heterogeneous queuing, the GPU doesn’t have to wait for the CPU to feed it data. It can spawn its own tasks on its own.
To date, programmers have been unable to harness the whole power of an APU. But with heterogeneous queuing and HSA, AMD says that it will be easier to better utilize the system’s resources, improving computing for a variety of tasks. It will be useful in gaming, natural user interfaces, accelerated video and image editing, and accelerated media playback. These kinds of tasks place far more stress on the system than previous kinds of computing tasks.
Intel and Nvidia haven’t supported HSA. The newly-approved specification allows for the programmability of heterogeneous processors, the portability of programming code, and interoperability across different vendor devices. The spec defines how hardware operates and addresses the needs of the ecosystem of software, tool, and compiler developers. And it addresses how applications interact with HSA platforms.
The group has supporting statements from Wen-Mei Hwu, a computer engineering professor at the University of Illinois at Urbana-Champaign; Jon Peddie, a computer graphics expert at Jon Peddie Research; and Patrick Moorhead, analyst at Moor Insights & Strategy.
“I anticipate a lot of interesting use cases -- from video chat apps and search to TV shows and movies," Moorhead said. "App developers should also find it easier to harness all of the processors together.”
Supporters at an HSA event in San Jose, Calif., include AMD, ARM, Imagination Technologies, MediaTek, Qualcomm and Samsung. Currently, more than 40 companies and 17 universities support HSA.
“HSA 1.0 is an idea whose time has come," said Manju Hegde, corporate vice president at AMD, in a statement. "It gives developers easier access to the power-efficient performance on today’s rich SoCs than ever before, freeing them to find creative solutions to [computing's] toughest challenges. AMD intends to bring processors which incorporate the architecture described in the specification to market in 2015 and help lead the industry into the new era of heterogeneous computing.”