HSA Announces Publication of New Guide to Heterogeneous System Architecture

[btarunr]

The Heterogeneous System Architecture (HSA) Foundation today announced publication of Heterogeneous System Architecture: A New Compute Platform Infrastructure (1st Edition), edited by Dr. Wen-Mei Hwu. The book, published by Elsevier Publishing (found here: here), offers a practical guide to understanding HSA, a standardized platform design that unlocks the performance and power efficiency of parallel computing engines found in most modern electronic devices.

"Heterogeneous computing is a key enabler of the next generation of compute environments, wherein entire systems will interconnect autonomously and in real time," said HSA Foundation President Dr. John Glossner. "Developers who are skilled in the use of this platform will have the upper hand in terms of design time, IP portability, power efficiency and performance."

To support these developers, the HSA Foundation working groups are rapidly standardizing tools and APIs for debug and profiling, creating guidelines for incorporating IP from multiple vendors into the same SoC, and much more. The Foundation released the v1.0 specification in March, and soon thereafter, companies including AMD, ARM, Imagination Technologies and MediaTek previewed their plans for rolling out the world's first products based on HSA.

"The HSA guidebook will help proliferate the platform among students, programmers and developers worldwide," said Dr. Hwu. "This publication will help them quickly learn more about HSA concepts, fundamentals, and practices, including techniques for creating virtual parallel systems, as well as compiling and simulating designs."
Through the new book, software application developers, computer science researchers, and students in computer architecture, distributed computing, and software engineering courses will learn:

• How performance-bound programming algorithms and application types can be significantly optimized by using HSA hardware and software features;
• Ideal mapping of processing resources from CPUs to many other heterogeneous processors, in compliance with HSA specifications ;
• Clear and concise explanations of key HSA concepts and fundamentals provided by expert HSA specification contributors.

The book begins with an overview of the evolution of heterogeneous parallel processing and its historic challenges. Later chapters provide a deeper perspective on topics such as runtime, memory model, queuing, context switching, the architected queuing language, simulators, and tool chains. The publication also includes three real world examples that clearly demonstrate how HSA can deliver significantly higher performance thru C++ based applications.

Contributing authors include HSA Foundation members and experts from both academia and industry. Some of these distinguished authors include: Yeh-Ching Chung, Benedict R. Gaster, Juan Gómez-Luna, Derek Hower, Lee Howes, Shih-Hao Hung, Thomas B. Jablin, David Kaeli, Phil Rogers, Ben Sander, I-Jui (Ray) Sung.

View at TechPowerUp Main Site

 

[Easy Rhino]

This is the most IEEE sounding press release ever.

 

[chlamchowder]

IMO, the biggest problem with HSA is that it makes the most sense with integrated graphics, but right now discrete cards with separate memory pools provide much more compute power.

 

[happita]

IMO, the biggest problem with HSA is that it makes the most sense with integrated graphics, but right now discrete cards with separate memory pools provide much more compute power.

There's a reason it makes sense, its because the integrated graphics market accounts for a huge chunk of the pie when it comes to PCs (include laptops, ultrabooks, tablets, etc.). HSA has far-reaching applications outside of discrete graphics cards for desktop PCs. One other desktop application is AMD's very own APUs, which they helped integrate into that architecture years ago. The mobile SoC industry is also a very big one where we could see HSA leveraging some assets in order to improve the performance smartphones deliver in the next few years. Definitely some exciting times ahead.

 

[chlamchowder]

There's a reason it makes sense, its because the integrated graphics market accounts for a huge chunk of the pie when it comes to PCs (include laptops, ultrabooks, tablets, etc.). HSA has far-reaching applications outside of discrete graphics cards for desktop PCs. One other desktop application is AMD's very own APUs, which they helped integrate into that architecture years ago. The mobile SoC industry is also a very big one where we could see HSA leveraging some assets in order to improve the performance smartphones deliver in the next few years. Definitely some exciting times ahead.

Yes, it is a huge chunk of the pie. But I'm not sure how much motivation there is to code for it.

Performance on IGPs is low, probably to the point where just copying the data over to a discrete card, doing the computation there, and copying the result back is faster. Also, getting something working (and working well) on a GPU is really difficult. I'd rather put time into coding for a card with dedicated memory (that will also work with IGPs), than try to use HSA and have something that doesn't work (or doesn't work well) with discrete cards.