David Brooks

Bio: David Brooks is an academic researcher from Harvard University. The author has contributed to research in topics: Hardware acceleration & Computer science. The author has an hindex of 54, co-authored 242 publications receiving 15012 citations. Previous affiliations of David Brooks include IBM & Princeton University.

Wattch: a framework for architectural-level power analysis and optimizations

Abstract: Power dissipation and thermal issues are increasingly significant in modern processors. As a result, it is crucial that power/performance tradeoffs be made more visible to chip architects and even compiler writers, in addition to circuit designers. Most existing power analysis tools achieve high accuracy by calculating power estimates for designs only after layout or floorplanning are complete. In addition to being available only late in the design process, such tools are often quite slow, which compounds the difficulty of running them for a large space of design possibilities.This paper presents Wattch, a framework for analyzing and optimizing microprocessor power dissipation at the architecture-level. Wattch is 1000X or more faster than existing layout-level power tools, and yet maintains accuracy within 10% of their estimates as verified using industry tools on leading-edge designs. This paper presents several validations of Wattch's accuracy. In addition, we present three examples that demonstrate how architects or compiler writers might use Wattch to evaluate power consumption in their design process.We see Wattch as a complement to existing lower-level tools; it allows architects to explore and cull the design space early on, using faster, higher-level tools. It also opens up the field of power-efficient computing to a wider range of researchers by providing a power evaluation methodology within the portable and familiar SimpleScalar framework.

Dynamic thermal management for high-performance microprocessors

Abstract: With the increasing clock rate and transistor count of today's microprocessors, power dissipation is becoming a critical component of system design complexity. Thermal and power-delivery issues are becoming especially critical for high-performance computing systems. In this work, we investigate dynamic thermal management as a technique to control CPU power dissipation. With the increasing usage of clock gating techniques, the average power dissipation typically seen by common applications is becoming much less than the chip's rated maximum power dissipation. However system designers still must design thermal heat sinks to withstand the worse-case scenario. We define and investigate the major components of any dynamic thermal management scheme. Specifically we explore the tradeoffs between several mechanisms for responding to periods of thermal trauma and we consider the effects of hardware and software implementations. With approximate dynamic thermal management, the CPU can be designed for a much lower maximum power rating, with minimal performance impact for typical applications.

System level analysis of fast, per-core DVFS using on-chip switching regulators

Abstract: Portable, embedded systems place ever-increasing demands on high-performance, low-power microprocessor design. Dynamic voltage and frequency scaling (DVFS) is a well-known technique to reduce energy in digital systems, but the effectiveness of DVFS is hampered by slow voltage transitions that occur on the order of tens of microseconds. In addition, the recent trend towards chip-multiprocessors (CMP) executing multi-threaded workloads with heterogeneous behavior motivates the need for per-core DVFS control mechanisms. Voltage regulators that are integrated onto the same chip as the microprocessor core provide the benefit of both nanosecond-scale voltage switching and per-core voltage control. We show that these characteristics provide significant energy-saving opportunities compared to traditional off-chip regulators. However, the implementation of on-chip regulators presents many challenges including regulator efficiency and output voltage transient characteristics, which are significantly impacted by the system-level application of the regulator. In this paper, we describe and model these costs, and perform a comprehensive analysis of a CMP system with on-chip integrated regulators. We conclude that on-chip regulators can significantly improve DVFS effectiveness and lead to overall system energy savings in a CMP, but architects must carefully account for overheads and costs when designing next-generation DVFS systems and algorithms.

Minerva: enabling low-power, highly-accurate deep neural network accelerators

Abstract: The continued success of Deep Neural Networks (DNNs) in classification tasks has sparked a trend of accelerating their execution with specialized hardware. While published designs easily give an order of magnitude improvement over general-purpose hardware, few look beyond an initial implementation. This paper presents Minerva, a highly automated co-design approach across the algorithm, architecture, and circuit levels to optimize DNN hardware accelerators. Compared to an established fixed-point accelerator baseline, we show that fine-grained, heterogeneous datatype optimization reduces power by 1.5×; aggressive, inline predication and pruning of small activity values further reduces power by 2.0×; and active hardware fault detection coupled with domain-aware error mitigation eliminates an additional 2.7× through lowering SRAM voltages. Across five datasets, these optimizations provide a collective average of 8.1× power reduction over an accelerator baseline without compromising DNN model accuracy. Minerva enables highly accurate, ultra-low power DNN accelerators (in the range of tens of milliwatts), making it feasible to deploy DNNs in power-constrained IoT and mobile devices.

Applied Machine Learning at Facebook: A Datacenter Infrastructure Perspective

Abstract: Machine learning sits at the core of many essential products and services at Facebook. This paper describes the hardware and software infrastructure that supports machine learning at global scale. Facebook's machine learning workloads are extremely diverse: services require many different types of models in practice. This diversity has implications at all layers in the system stack. In addition, a sizable fraction of all data stored at Facebook flows through machine learning pipelines, presenting significant challenges in delivering data to high-performance distributed training flows. Computational requirements are also intense, leveraging both GPU and CPU platforms for training and abundant CPU capacity for real-time inference. Addressing these and other emerging challenges continues to require diverse efforts that span machine learning algorithms, software, and hardware design.

Machine learning

Abstract: Machine Learning is the study of methods for programming computers to learn. Computers are applied to a wide range of tasks, and for most of these it is relatively easy for programmers to design and implement the necessary software. However, there are many tasks for which this is difficult or impossible. These can be divided into four general categories. First, there are problems for which there exist no human experts. For example, in modern automated manufacturing facilities, there is a need to predict machine failures before they occur by analyzing sensor readings. Because the machines are new, there are no human experts who can be interviewed by a programmer to provide the knowledge necessary to build a computer system. A machine learning system can study recorded data and subsequent machine failures and learn prediction rules. Second, there are problems where human experts exist, but where they are unable to explain their expertise. This is the case in many perceptual tasks, such as speech recognition, hand-writing recognition, and natural language understanding. Virtually all humans exhibit expert-level abilities on these tasks, but none of them can describe the detailed steps that they follow as they perform them. Fortunately, humans can provide machines with examples of the inputs and correct outputs for these tasks, so machine learning algorithms can learn to map the inputs to the outputs. Third, there are problems where phenomena are changing rapidly. In finance, for example, people would like to predict the future behavior of the stock market, of consumer purchases, or of exchange rates. These behaviors change frequently, so that even if a programmer could construct a good predictive computer program, it would need to be rewritten frequently. A learning program can relieve the programmer of this burden by constantly modifying and tuning a set of learned prediction rules. Fourth, there are applications that need to be customized for each computer user separately. Consider, for example, a program to filter unwanted electronic mail messages. Different users will need different filters. It is unreasonable to expect each user to program his or her own rules, and it is infeasible to provide every user with a software engineer to keep the rules up-to-date. A machine learning system can learn which mail messages the user rejects and maintain the filtering rules automatically. Machine learning addresses many of the same research questions as the fields of statistics, data mining, and psychology, but with differences of emphasis. Statistics focuses on understanding the phenomena that have generated the data, often with the goal of testing different hypotheses about those phenomena. Data mining seeks to find patterns in the data that are understandable by people. Psychological studies of human learning aspire to understand the mechanisms underlying the various learning behaviors exhibited by people (concept learning, skill acquisition, strategy change, etc.).

Cloud computing: state-of-the-art and research challenges

Abstract: Cloud computing has recently emerged as a new paradigm for hosting and delivering services over the Internet. Cloud computing is attractive to business owners as it eliminates the requirement for users to plan ahead for provisioning, and allows enterprises to start from the small and increase resources only when there is a rise in service demand. However, despite the fact that cloud computing offers huge opportunities to the IT industry, the development of cloud computing technology is currently at its infancy, with many issues still to be addressed. In this paper, we present a survey of cloud computing, highlighting its key concepts, architectural principles, state-of-the-art implementation as well as research challenges. The aim of this paper is to provide a better understanding of the design challenges of cloud computing and identify important research directions in this increasingly important area.

In-Datacenter Performance Analysis of a Tensor Processing Unit

Abstract: Many architects believe that major improvements in cost-energy-performance must now come from domain-specific hardware. This paper evaluates a custom ASIC---called a Tensor Processing Unit (TPU)---deployed in datacenters since 2015 that accelerates the inference phase of neural networks (NN). The heart of the TPU is a 65,536 8-bit MAC matrix multiply unit that offers a peak throughput of 92 TeraOps/second (TOPS) and a large (28 MiB) software-managed on-chip memory. The TPU's deterministic execution model is a better match to the 99th-percentile response-time requirement of our NN applications than are the time-varying optimizations of CPUs and GPUs (caches, out-of-order execution, multithreading, multiprocessing, prefetching, ...) that help average throughput more than guaranteed latency. The lack of such features helps explain why, despite having myriad MACs and a big memory, the TPU is relatively small and low power. We compare the TPU to a server-class Intel Haswell CPU and an Nvidia K80 GPU, which are contemporaries deployed in the same datacenters. Our workload, written in the high-level TensorFlow framework, uses production NN applications (MLPs, CNNs, and LSTMs) that represent 95% of our datacenters' NN inference demand. Despite low utilization for some applications, the TPU is on average about 15X - 30X faster than its contemporary GPU or CPU, with TOPS/Watt about 30X - 80X higher. Moreover, using the GPU's GDDR5 memory in the TPU would triple achieved TOPS and raise TOPS/Watt to nearly 70X the GPU and 200X the CPU.

Wattch: a framework for architectural-level power analysis and optimizations

Abstract: Power dissipation and thermal issues are increasingly significant in modern processors. As a result, it is crucial that power/performance tradeoffs be made more visible to chip architects and even compiler writers, in addition to circuit designers. Most existing power analysis tools achieve high accuracy by calculating power estimates for designs only after layout or floorplanning are complete. In addition to being available only late in the design process, such tools are often quite slow, which compounds the difficulty of running them for a large space of design possibilities.This paper presents Wattch, a framework for analyzing and optimizing microprocessor power dissipation at the architecture-level. Wattch is 1000X or more faster than existing layout-level power tools, and yet maintains accuracy within 10% of their estimates as verified using industry tools on leading-edge designs. This paper presents several validations of Wattch's accuracy. In addition, we present three examples that demonstrate how architects or compiler writers might use Wattch to evaluate power consumption in their design process.We see Wattch as a complement to existing lower-level tools; it allows architects to explore and cull the design space early on, using faster, higher-level tools. It also opens up the field of power-efficient computing to a wider range of researchers by providing a power evaluation methodology within the portable and familiar SimpleScalar framework.

In-Datacenter Performance Analysis of a Tensor Processing Unit

Abstract: Many architects believe that major improvements in cost-energy-performance must now come from domain-specific hardware. This paper evaluates a custom ASIC---called a Tensor Processing Unit (TPU) --- deployed in datacenters since 2015 that accelerates the inference phase of neural networks (NN). The heart of the TPU is a 65,536 8-bit MAC matrix multiply unit that offers a peak throughput of 92 TeraOps/second (TOPS) and a large (28 MiB) software-managed on-chip memory. The TPU's deterministic execution model is a better match to the 99th-percentile response-time requirement of our NN applications than are the time-varying optimizations of CPUs and GPUs that help average throughput more than guaranteed latency. The lack of such features helps explain why, despite having myriad MACs and a big memory, the TPU is relatively small and low power. We compare the TPU to a server-class Intel Haswell CPU and an Nvidia K80 GPU, which are contemporaries deployed in the same datacenters. Our workload, written in the high-level TensorFlow framework, uses production NN applications (MLPs, CNNs, and LSTMs) that represent 95% of our datacenters' NN inference demand. Despite low utilization for some applications, the TPU is on average about 15X -- 30X faster than its contemporary GPU or CPU, with TOPS/Watt about 30X -- 80X higher. Moreover, using the CPU's GDDR5 memory in the TPU would triple achieved TOPS and raise TOPS/Watt to nearly 70X the GPU and 200X the CPU.