AMD要在HSA处理器上成功？努力不能停

异质系统架构(Heterogeneous Systems Architecture，HSA)基金会在那个对未来的处理器 SoC 设计工程师与相关软件非常重要的新技术领域取得了不错进展，但该技术要取得成功恐怕还有一段很长的路要走。 日前 AMD 宣布将在明年1月推出首款采用 HSA 技术的处理器；该架构能让中央处理器(CPU)与绘图处理核心共享资源。AMD以Linux为基础开发了支持新架构的软件，但到目前为止并未有其它支持HSA的处理器或操作系统计划公布。 包 括ARM、Imagination、联发科(Mediatek)、高通(Qualcomm)与三星( Samsung)等移动处理器大厂与核心供货商，也都是HSA基金会的成员；在一场日前于美国举行的AMD开发者大会上，一位来自Imagination 的高层表示，该公司总有一天会支持HSA架构，但他并未透露相关产品的推出时间表。 ARM 首席技术官Mike Muller的态度更为保守，他在会中表示：“HSA对中型系统来说很好，但我不确定它是否能适用最小型的系统，我也不认为它能支持内含上万颗处理器的高性能运算。单一系统的架构不会扩展到整个应用空间。” 苹果(Apple)、Google与微软(Microsoft)则对他们在iOS、Android与Chrome OS，还有Windows等操作系统对HSA的支持三缄其口；目前设计工程师虽然能撰写驱动程序以在Windows等操作系统支持该架构，但并没有更理想的解决方案。 而最与众不同的是AMD的竞争对手英特尔(Intel)与Nvidia，这两家公司与苹果都在移动与台式计算机系统支持 OpenCL；由Khronos Group开发的 OpenCL 规格是针对各种绘图处理器(GPU)之平行编程任务所开发的较高层级应用程序接口(API)。 OpenCL 已经存在好一段时间，但才刚开始加入一些HSA基金会正在定义的共享内存功能。一位Khronos 与Nvidia的代表就指出，HSA、OpenCL以及Nvidia专有的Cuda软件环境，其实是朝平行路线发展，并提供越来越类似的功能。 Cuda是Nvidia在高性能运算领域的一次成功出击，曾进驻全球前五百大超级计算机；而由于很早就进军高性能运算领域，市场上有越来越多采用Nvidia配备Cuda之GPU的系统。但这样的情况也可能会改变。 高性能运算领域专家、美国伊利诺大学(University of Illinois)计算机科学教授Wen-mei Hwu表示：“超级计算机开发商每三到四年就会去找新东西、选定其中某个，然后在几年后实际运用。”最新版的Cuda已经可以支持部分HSA的平面内存 (flat-memory)功能，但Cuda并不怎么适合一些超级计算机应用程序，这就为HSA基金会与AMD带来机会。 至于英特尔则在高阶运算领域有自己的方案，其 Xeon Phi 处理器采用60个简化的 x86核心来处里平行任务，搭配该公司自家开发的工具； Phi 已经开始获得一些全球五百大超级计算机开发商的青睐(而且是在短时间内)。 本文授权编译自EE Times，版权所有，谢绝转载 本文下一页：HSA的甜蜜点在哪？

相关阅读：
• 联发科新八核芯片将成导入HSA技术试金石
• Tensilica加入HSA基金会，助力嵌入式异构计算标准建立
• AMD发布X86/ARM架构混搭处理器，目标笔电和平板nECesmc

{pagination} HSA将扩展至GPU以外的处理器 HSA的甜蜜点在于能解决主流运算领域一些非常实际的问题，而该领域有越来越多看起来像是配备各种核心的智能手机芯片。 “我们一直很谨慎开发能支持各种加速处理器的(HSA)架构，”HSA基金会总经理、专长绘图处里的AMD院士Phil Rogers表示：“我们是以GPU为第一步，而其它种类的加速处理器是可以预期的，包括DSP、DMA引擎、译码器以及加密处理器；只要是合理的，每一 家公司都将会转向开发加速处理器。” HSA为程序开发者解决了非常实际的问题；如Rogers所言，当处理器的CPU与GPU核心拥有各自的内存时，要在处理器SoC上操作主流应用程序会有许多困难：“人们已经开发了利用两组处理器核心的应用程序，但问题在于它们是用很不自然的方式才能运作。” Rogers表示，如果虚拟内存空间中存在只有CPU能理解的“指针(pointers)”，要横跨现有的不同资料架构几乎是不可能的。因此今日的移动应用程序设计工程师，通常得重新打造资料架构，以保持追踪不同 CPU 与 GPU 核心的资料。 在 专题演说中，Rogers提到了可利用HSA带来之额外性能与简易性的各种应用程序，例如在Hadoop与Map Reduce平台上执行的大型资料中心、特定的数据库任务、石油天然气探勘，以及电信营运商与网络资料中心的媒体转码(media transcoding)。 因此HSA能为处理器与应用程序开发商带来实际的优势，还有AMD──如果该公司能藉由某种方式 提供超越英特尔与Nvidia方案的独特功能。而由于在电子学方面还有许多问题存在，一切只能等待。AMD早在2006年为了GPU收购ATI时，就看好 微处理器的未来是混合式核心SoC；七年后，该公司仍坚持初衷，但要实现该愿景，还有好一段路得走。 本文授权编译自EE Times，版权所有，谢绝转载 编译：Judith Cheng 参考英文原文：AMD Faces Long Road to HSA Success，by Rick Merritt

相关阅读：
• 联发科新八核芯片将成导入HSA技术试金石
• Tensilica加入HSA基金会，助力嵌入式异构计算标准建立
• AMD发布X86/ARM架构混搭处理器，目标笔电和平板nECesmc

{pagination} AMD Faces Long Road to HSA Success Rick Merritt SAN JOSE, Calif. — The Heterogeneous Systems Architecture (HSA) Foundation is making good progress in an area important for future designers of SoCs and the software that rides on them, but it still faces a long road to success. This week, Advanced Micro Devices said it will ship the first processor to use HSA techniques of letting CPU and graphics cores share resources in January. AMD developed support under Linux for the techniques, but no other plans have been announced so far for processors or operating systems supporting the techniques. The chips (and probably the operating systems) will come at some point. ARM, Imagination, Mediatek, Qualcomm, Samsung, and other top mobile SoC and core designers are part of HSA. During a panel discussion at AMD's developer conference here, an Imagination executive said the company will support HSA on all its cores eventually, but he would not give say which cores would ship when. ARM CTO Mike Muller was more reserved. "HSA is good for midrange systems, but I'm not sure if it scales to the smallest systems, and I don't think it does scale to high-performance computing with tens of thousands of processors," he said during the panel discussion. "One system's architecture won't scale across the whole space." Apple, Google, and Microsoft have remained mum about their plans to support HSA in iOS, Android, Chrome OS, and Windows. Designers can write drivers to support the techniques in operating systems such as Windows, but that's not an ideal approach. The big standouts are AMD's archrivals: Intel and Nvidia. They and Apple back OpenCL in mobile and desktop systems. The Khronos Group's OpenCL spec is a higher-level API for programming parallel tasks on any GPU. It has been around for a while but is just starting to build in some of the shared-memory goodies the HSA Foundation is defining. A representative for Khronos and Nvidia said HSA, OpenCL, and Nvidia's proprietary Cuda software environment are on parallel paths and offering increasingly similar features. Cuda has been a hit for Nvidia in high-performance computing, including in the top 500 supercomputers. Thanks to an early move into this market, a growing share of these systems use Nvidia's GPUs with Cuda. But that could change. "Every three or four years, supercomputer developers look around for what's new, seize on something, and really milk it for a few years," said Wen-mei Hwu, a professor of computer science at the University of Illinois and an expert in the field. The latest versions of Cuda already support some HSA flat-memory features. But Cuda is not well suited for some supercomputer apps, and that could open a door for the HSA Foundation and AMD. Intel does Phi, AMD beyond GPUs Intel has its own offering at the high end. Its Xeon Phi uses more than 60 simplified x86 cores to handle parallel jobs, using tools Intel has developed. Phi is just starting to gain traction (quite quickly) among the top 500. The sweet spot for HSA is solving very real problems in the mainstream computing space, where processors are increasingly looking like smartphone chips with a variety of cores. "We've been careful to develop an [HSA] architecture that works for all accelerators," said Phil Rogers, president of the HSA Foundation and an AMD fellow with a specialty in graphics. "We started first with GPUs, but other accelerators are expected, including DSPs, DMA engines, codecs, and crypto processors. Each company will migrate to accelerators as it makes sense." The programmer's problem HSA solves is very real. Rogers talked about the difficulties handling mainstream apps on SoCs where CPU and GPU cores have their own private pools of memory. "People already developing apps that use both sets of cores, but the problem is they have to perform unnatural acts to make it work." It's nearly impossible to "walk through the data structures that exist if they have pointers that are in the virtual memory space only the CPU understands." Thus today's mobile programmers typically recraft data structures to keep track of information in various CPUs and GPUs. In addition to creating extra complexity and work, the approach sacrifices some performance. In his keynote, Rodgers talked about all sorts of apps that could use the extra performance and simplicity HSA promises. They include big data jobs run on Hadoop and Map Reduce, certain database tasks, oil and gas discovery, and media transcoding for carriers and web datacenters. So HSA promises real advantages for developers of SoCs and apps -- and for AMD if it can deliver the features in a way that leapfrogs Intel and Nvidia offerings. As with so many issues in electronics, it comes down to a question of when. AMD saw the future of microprocessors was in mixed-core SoCs back in 2006, when it purchased ATI Technologies for its GPUs. Seven years later, it is still pursuing the vision, and it still faces a long road to deliver on it.