When the HSA Foundation was officially introduced back in 2012, I remember being on a conference call with our EVP of marketing Tony King-Smith and Patrick Moorhead*, founder and president of Moor Insights & Strategy.
Patrick was working on a whitepaper (you can read it here) and was talking to the founding members to get more background information on how and why the foundation had been created. Right from the start, one of interesting things that Patrick pointed out was the wave of positive feedback that was pouring in from everyone involved in computing, whether it was mobile and embedded, desktop or HPC.
The buzz in the air was real and palpable.
The HSA Foundation united AMD, ARM, Imagination, MediaTek, Qualcomm, Samsung and Texas Instruments in a quest to build interoperable hardware and software that would address the high-efficiency requirements of next-generation platforms.
Yesterday the HSA Foundation launched version 1.0 of the specification during an event held at the Fairmont Hotel in San Jose, California (read the press release here). All board members were there to discuss software, the ecosystem and applications in the mobile, desktop PC and high-performance computing.
To accompany the launch, I have produced some examples of how system designers can use our processor IP to implement power-efficient heterogeneous platforms:
- MIPS I6400 CPU: This new 64-bit CPU is designed for true heterogeneous processing for mobile and embedded applications; at the CPU level, it scales from one to four threads – and system architects can put up to six CPUs in a cluster. Furthermore, an I6400 processor is designed to scale up to 64 clusters, extending to over 1500 threads. Every CPU inside the cluster can be managed individually, allowing for fine-grain control of frequency and power.
- All PowerVR Rogue GPUs feature a scalar SIMD architecture that treats general purpose compute as a first class citizen, while not forgetting what makes a shader core great for graphics. The latest PowerVR Series7XT GPUs feature a scalable 64-bit floating point co-processor per cluster for customers who want to use our IP in HPC applications. We’ve also made several GPU compute setup and cache throughput improvements resulting in up to 300% better parallel processing performance.
- Ensigma RPUs continue the trend of moving more processing on chip. Designed for full system integration, these radio processors include a powerful DSP architecture that is capable of handling multiple wireless communications standards on a single chip, including all digital TV and radio formats as well as 802.11 Wi-Fi and Bluetooth.
One superb illustration of how these different processors come together in a true heterogeneous platform is our PowerVR Vision IP platform. Suitable for applications such as mobile and wearable, automotive (ADAS), surveillance or augmented reality, this platform includes multiple processing units, all working together efficiently to handle multimedia-related tasks.
For example, the PowerVR ISP can send metadata to the CPU, video encoder or the graphics processor to speed up certain tasks; developers can also distribute computer vision algorithms more efficiently across several units (e.g. heavy parallel-oriented code can run on the GPU, while other optional algorithms can be executed on the CPU SIMD engine).
On the software side, we’ve recently introduced the PowerVR imaging framework for Android. This includes a set of extensions to the OpenCL and EGL APIs that enable efficient zero-copy sampling of YUV and RGB camera data. The framework also includes low-level functions for integrating these extensions within the camera Hardware Abstraction Layer (HAL).
If you want to know more about the HSA Foundation, you can visit their website. Make sure you also follow us on Twitter (@ImaginationPR, @MIPSGuru, @PowerVRInsider, @GPUcompute) for the latest news and announcements from Imagination.
* Patrick Moorhead is currently ranked one of the top technology analysts in the world and has extensive experience in covering the computing market, comprising personal computing, graphics or servers. You can read more about him here and follow him on Twitter (@PatrickMoorhead).