Today’s performance analysis frameworks for deep learning accelerators suffer from two significant limitations. First, although modern convolutional neural networks (CNNs) consist of many types of layers other than convolution, especially during training, these frameworks largely focus on convolution layers only. Second, these frameworks are generally targeted towards inference and lack support for training operations. This work proposes a novel open-source performance analysis framework, SimDIT, for general ASIC-based systolic hardware accelerator platforms. The modeling effort of SimDIT comprehensively covers convolution and non-convolution operations of both CNN inference and training on a highly parameterizable hardware substrate. SimDIT is integrated with a backend silicon implementation flow and provides detailed end-to-end performance statistics (i.e., data access cost, cycle counts, energy, and power) for executing CNN inference and training workloads. SimDIT-enabled performance analysis reveals that on a 64×64 processing array, non-convolution operations constitute 59.5% of total runtime for ResNet-50 training workload. In addition, by optimally distributing available off-chip DRAM bandwidth and on-chip SRAM resources, SimDIT achieves 18× performance improvement over a generic static resource allocation for ResNet-50 inference.

With the ever increasing prevalence of neural networks and the upheaval from the language models, it is time to rethink neural acceleration. Up to this point, the broader research community, including ourselves, has disproportionately focused on GEneral Matrix Multiplication (GEMM) operations. The supporting argument was that the large majority of the neural operations are GEMM. This argument guided the research in Neural Processing Units (NPUs) for the last decade. However, scant attention was paid to non-GEMM operations and they are rather overlooked. As deep learning evolved and progressed, these operations have grown in diversity and also large variety of structural patterns have emerged that interweave them with the GEMM operations. However, conventional NPU designs have taken rather simplistic approaches by supporting these operations through either a number of dedicated blocks or fall back to general-purpose processors.

While (I) serverless computing is emerging as a popular form of cloud execution, datacenters are going through major changes: (II) storage dissaggregation in the system infrastructure level and (III) integration of domain-specific accelerators in the hardware level. Each of these three trends individually provide significant benefits; however, when combined the benefits diminish. On the convergence of these trends, the paper makes the observation that for serverless functions, the overhead of accessing dissaggregated storage overshadows the gains from accelerators. Therefore, to benefit from all these trends in conjunction, we propose In-Storage Domain-Specific Acceleration for Serverless Computing (dubbed DSCS-Serverless1). The idea contributes a server-less model that utilizes a programmable accelerator embedded within computational storage to unlock the potential of acceleration in disaggregated datacenters. Our results with eight applications show that integrating a comparatively small accelerator within the storage (DSCS-Serverless) that fits within the storage's power constraints (25 Watts), significantly outperforms a traditional disaggregated system that utilizes NVIDIA RTX 2080 Ti GPU (250 Watts). Further, the work highlights that disaggregation, serverless model, and the limited power budget for computation in storage device require a different design than the conventional practices of integrating microprocessors and FPGAs. This insight is in contrast with current practices of designing computational storage devices that are yet to address the challenges associated with the shifts in datacenters. In comparison with two such conventional designs that use ARM cores or a Xilinx FPGA, DSCS-Serverless provides 3.7× and 1.7× end-to-end application speedup, 4.3× and 1.9× energy reduction, and 3.2× and 2.3× better cost efficiency, respectively.

Representative modeling of I/O activity is crucial when designing large-scale distributed storage systems. Particularly important use cases are counterfactual "what-if" analyses that assess the impact of anticipated or hypothetical new storage policies or hardware prior to deployment. We propose Thesios, a methodology to accurately synthesize such hypothetical full-resolution I/O traces by carefully combining down-sampled I/O traces collected from multiple disks attached to multiple storage servers. Applying this approach to real-world traces that are already routinely sampled at Google, we show that our synthesized traces achieve 95--99.5% accuracy in read/write request numbers, 90--97% accuracy in utilization, and 80--99.8% accuracy in read latency compared to metrics collected from actual disks. We demonstrate how Thesios enables diverse counterfactual I/O trace synthesis and analyses of hypothetical policy, hardware, and server changes through four case studies: (1) studying the effects of changing disk's utilization, fullness, and capacity, (2) evaluating new data placement policy, (3) analyzing the impact on power and performance of deploying disks with reduced rotations-per-minute (RPM), and (4) understanding the impact of increased buffer cache size on a storage server. Without Thesios, such counterfactual analyses would require costly and potentially risky A/B experiments in production.

There has been an arms race for devising accelerators for deep learning in recent years. However, real-world applications are not only neural networks but often span across multiple domains, e.g., database queries, compression, encryption, video coding, signal processing, and traditional machine learning, which may or may not contain deep learning. The sole focus on this single domain is sub-optimal as it misses the potential to proliferate and promote cross-domain multi-acceleration as there is an opportunity to harness the power of chaining heterogeneous Domain-Specific Architectures (DSAs) in modern datacenter applications. However, there is a catch as the data motion overhead can outweigh the benefits from all these chained heterogeneous accelerators. We dub the data restructuring and communication overhead of executing a single application using a chain of accelerators [1] as the data motion overhead. In a stark contrast with most works on DSAs that deal with accelerating compute kernels, this work focuses on accelerating data motion within a chain of heterogeneous DSAs in a multi-accelerator datacenter. To that end, this paper introduces Data Motion Acceleration (DMX) for (1) reducing data movement, (2) accelerating data restructuring, and (3) enabling interoperability between heterogeneous accelerators from different domains through a cross-stack hardware-software solution. The results with five end-to-end applications show that utilizing DMX offers up to 8.2 ×, 13.6 ×, and 5.2 × improvement in latency, throughput, and energy efficiency in a multi-accelerator system, respectively.

Parameterizable machine learning (ML) accelerators are the product of recent breakthroughs in ML. To fully enable their design space exploration (DSE), we propose a physical-design-driven, learning-based prediction framework for hardware-accelerated deep neural network (DNN) and non-DNN ML algorithms. It adopts a unified approach that combines power, performance, and area (PPA) analysis with frontend performance simulation, thereby achieving a realistic estimation of both backend PPA and system metrics such as runtime and energy. In addition, our framework includes a fully automated DSE technique, which optimizes backend and system metrics through an automated search of architectural and backend parameters. Experimental studies show that our approach consistently predicts backend PPA and system metrics with an average 7% or less prediction error for the ASIC implementation of two deep learning accelerator platforms, VTA and VeriGOOD-ML, in both a commercial 12 nm process and a research-oriented 45 nm process.

Modern heterogeneous CPUs incorporate hardware accelerators to enable domain-specialized execution and achieve improved efficiency. A well-known class among them, spatial accelerators, are designed with reconfigurability to accelerate a wide range of compute-heavy and data-parallel applications. Unlike CPU cores, however, they tend to require specialized compilers and software stacks, libraries, or languages to operate and cannot be utilized with ease by all applications. As a result, the accelerator's large pool of compute and memory resources sit wastefully idle when it is not explicitly programmed. Our goal is to dismantle this CPU-accelerator barrier by monitoring CPU threads for acceleration opportunities during execution and, if viable, dynamically reconfigure the accelerator to allow transparent offloading. We develop MESA (Microarchitecture Extensions for Spatial Architecture Generation), a hardware block on the CPU that translates machine code to build an accelerator configuration specialized for the running program. While such a dynamic translation/reconfiguration approach is challenging, it has a key advantage over ahead-of-time compilers: access to runtime information, revealing not only dynamic dependencies but also performance characteristics. MESA maintains a real-time performance model of the program mapped on the accelerator in the form of a spatial dataflow graph with nodes weighted by operation latency and edges weighted by data transfer latency. Features of this dataflow graph are continuously updated with runtime information captured by performance counters, allowing a feedback loop of optimization, reconfiguration, and acceleration. This performance model allows MESA to identify the accelerator's critical paths and pinpoint its bottlenecks, upon which we implement in hardware a data-driven instruction mapping algorithm that locally minimizes latency. Backed by a synthesized RTL implementation, we evaluate the feasibility of our microarchitectural solution with different accelerator configurations. Across the Rodinia benchmarks, results demonstrate an average 1.3× speedup in performance and 1.8× gain in energy efficiency against a multicore CPU baseline.

Self-attention is a key enabler of state-of-art accuracy for various transformer-based Natural Language Processing models. This attention mechanism calculates a correlation score for each word with respect to the other words in a sentence. Commonly, only a small subset of words highly correlates with the word under attention, which is only determined at runtime. As such, a significant amount of computation is inconsequential due to low attention scores and can potentially be pruned. The main challenge is finding the threshold for the scores below which subsequent computation will be inconsequential. Although such a threshold is discrete, this paper formulates its search through a soft differentiable regularizer integrated into the loss function of the training. This formulation piggy backs on the back-propagation training to analytically co-optimize the threshold and the weights simultaneously, striking a formally optimal balance between accuracy and computation pruning. To best utilize this mathematical innovation, we devise a bit-serial architecture, dubbed LeOPArd, for transformer language models with bit-level early termination microarchitectural mechanism. We evaluate our design across 43 back-end tasks for MemN2N, BERT, ALBERT, GPT-2, and Vision transformer models. Post-layout results show that, on average, LeOPArd yields 1.9×and 3.9×speedup and energy reduction, respectively, while keeping the average accuracy virtually intact (< 0.2% degradation).

Domain-specific accelerators obtain performance benefits by restricting their algorithmic domain. These accelerators utilize specialized languages constrained to particular hardware, thus trading off expressiveness for high performance. The pendulum has swung from one hardware for all domains (general-purpose processors) to one hardware per individual domain. The middle-ground on this spectrum-which provides a unified computational stack across multiple, but not all, domains- is an emerging and open research challenge. This paper sets out to explore this region and its associated tradeoff between expressiveness and performance by defining a cross-domain stack, dubbed PolyMath. This stack defines a high-level cross-domain language (CDL), called PMLang, that in a modular and reusable manner encapsulates mathematical properties to be expressive across multiple domains-Robotics, Graph Analytics, Digital Signal Processing, Deep Learning, and Data Analytics. PMLang is backed by a recursively-defined intermediate representation allowing simultaneous access to all levels of operation granularity, called sr DFG. Accelerator-specific or domain-specific IRs commonly capture operations in the granularity that best fits a set of Domain-Specific Architectures (DSAs). In contrast, the recursive nature of the sr DFG enables simultaneous access to all the granularities of computation for every operation, thus forming an ideal bridge for converting to various DSA-specific IRs across multiple domains. Our stack unlocks multi-acceleration for end-to-end applications that cross the boundary of multiple domains each comprising different data and compute patterns. Evaluations show that by using PolyMath it is possible to harness accelerators across the five domains to realize an average speedup of 3.3× over a Xeon CPU along with 18.1× reduction in energy. In comparison to Jetson Xavier and Titan XP, cross-domain acceleration offers 1.7× and 7.2× improvement in performance-per-watt, respectively. We measure the cross-domain expressiveness and performance tradeoff by comparing each benchmark against its hand-optimized implementation to achieve 83.9% and 76.8% of the optimal performance for single-domain algorithms and end-to-end applications. For the two case studies of end-to-end applications (comprising algorithms from multiple domains), results show that accelerating all kernels offers an additional 2.0× speedup over CPU, 6.1× improvement in performance-per-watt over Titan Xp, and 2.8× speedup over Jetson Xavier compared to only the one most effective single-domain kernel being accelerated. Finally, we examine the utility and expressiveness of PolyMath through a user study, which shows, on average, PolyMath requires 1.9× less time to implement algorithms from two different domains with 2.5× fewer lines of code relative to Python.

This paper introduces VeriGOOD-ML, an automated methodology for generating Verilog with no human in the loop, starting from a high-level description of a machine learning (ML) algorithm in a standard format such as ONNX. The Verilog RTL is then translated through a back-end design flow to GDSII, driven by a design planning approach that is well tailored to the macro-intensive nature of ML platforms. VeriGOOD-ML uses three approaches to build ML hardware: the TABLA platform uses a dataflow architecture that is well suited to non-DNN ML algorithms; the GeneSys platform, with a systolic array and a SIMD array, is optimized for implementing DNNs; and the Axiline approach synthesizes small ML algorithms by hardcoding the structure of the algorithm into hardware, thus trading off flexibility for performance and power. The overall approach explores the design space of platform configurations and Pareto-optimal-PPA back-end implementations to yield designs that represent different tradeoffs at the algorithmic level between area, power, performance, and execution time. The overall methodology, from architecture to back-end design to hardware implementation, is described in this paper, and the results of VeriGOOD-ML are demonstrated on a set of ML benchmarks.

Albeit low-power, mixed-signal circuitry suffers from significant overhead of Analog to Digital (A/D) conversion, limited range for information encoding, and susceptibility to noise. This paper aims to address these challenges by offering and leveraging the following mathematical insight regarding vector dot-product---the basic operator in Deep Neural Networks (DNNs). This operator can be reformulated as a wide regrouping of spatially parallel low-bitwidth calculations that are interleaved across the bit partitions of multiple elements of the vectors. As such, the computational building block of our accelerator becomes a wide bit-interleaved analog vector unit comprising a collection of low-bitwidth multiply-accumulate modules that operate in the analog domain and share a single A/D converter(ADC). This bit-partitioning results in a lower-resolution ADC while the wide regrouping alleviates the need for A/D conversion per operation, amortizing its cost across multiple bit-partitions of the vector elements. Moreover, the low-bitwidth modules require smaller encoding range and also provide larger margins for noise mitigation. We also utilize the switched-capacitor design for our bit-level reformulation of DNN operations. The proposed switched-capacitor circuitry performs the regrouped multiplications in the charge domain and accumulates the results of the group in its capacitors over multiple cycles. The capacitive accumulation combined with wide bit-partitioned regrouping reduces the rate of A/D conversions, further improving the overall efficiency of the design.

Conventional neural accelerators rely on isolated self-sufficient functional units that perform an atomic operation while communicating the results through an operand delivery-aggregation logic. Each single unit processes all the bits of their operands atomically and produce all the bits of the results in isolation. This paper explores a different design style, where each unit is only responsible for a slice of the bit-level operations to interleave and combine the benefits of bit-level parallelism with the abundant data-level parallelism in deep neural networks. A dynamic collection of these units cooperate at runtime to generate bits of the results, collectively. Such cooperation requires extracting new grouping between the bits, which is only possible if the operands and operations are vectorizable. The abundance of Data-Level Parallelism and mostly repeated execution patterns, provides a unique opportunity to define and leverage this new dimension of Bit-Parallel Vector Composability. This design intersperses bit parallelism within data-level parallelism and dynamically interweaves the two together. As such, the building block of our neural accelerator is a Composable Vector Unit that is a collection of Narrower-Bitwidth Vector Engines, which are dynamically composed or decomposed at the bit granularity. Using six diverse CNN and LSTM deep networks, we evaluate this design style across four design points: with and without algorithmic bitwidth heterogeneity and with and without availability of a high-bandwidth off-chip memory. Across these four design points, Bit-Parallel Vector Composability brings (1.4× to 3.5×) speedup and (1.1× to 2.7×) energy reduction. We also comprehensively compare our design style to the Nvidia's RTX 2080 TI GPU, which also supports INT-4 execution. The benefits range between 28.0× and 33.7× improvement in Performance-per-Watt.

Deep Neural Networks (DNNs) have reinvigorated real-world applications that rely on learning patterns of data and are permeating into different industries and markets. Cloud infrastructure and accelerators that offer INFerence-as-a-Service (INFaaS) have become the enabler of this rather quick and invasive shift in the industry. To that end, mostly accelerator-based INFaaS (Google's TPU [1], NVIDIA T4 [2], Microsoft Brainwave [3], etc.) has become the backbone of many real-life applications. However, as the demand for such services grows, merely scaling-out the number of accelerators is not economically cost-effective. Although multi-tenancy has propelled datacenter scalability, it has not been a primary factor in designing DNN accelerators due to the arms race for higher speed and efficiency. This paper sets out to explore this timely requirement of multi-tenancy through a new dimension: dynamic architecture fission. To that end, we define Planaria1that can dynamically fission (break) into multiple smaller yet full-fledged DNN engines at runtime. This microarchitectural capability enables spatially co-locating multiple DNN inference services on the same hardware, offering simultaneous multi-tenant DNN acceleration. To realize this dynamic reconfigurability, we first devise breakable omni-directional systolic arrays for DNN acceleration that allows omni-directional flow of data. Second, it uses this capability and a unique organization of on-chip memory, interconnection, and compute resources to enable fission in systolic array based DNN accelerators. Architecture fission and its associated flexibility enables an extra degree of freedom for task scheduling, that even allows breaking the accelerator with regard to the server load, DNN topology, and task priority. As such, it can simultaneously co-locate DNNs to enhance utilization, throughput, QoS, and fairness. We compare the proposed design to PREMA [4], a recent effort that offers multi-tenancy by time-multiplexing the DNN accelerator across multiple tasks. We use the same frequency, the same amount of compute and memory resources for both accelerators. The results show significant benefits with (soft, medium, hard) QoS requirements, in throughput (7.4×, 7.2×, 12.2×), SLA satisfaction rate (45%, 15%, 16%), and fairness (2.1×, 2.3×, 1.9×).

Generative Adversarial Networks (GANs) are a frontier in deep learning. GANs consist of two models: generative and discriminative. While the discriminative model uses the conventional convolution, the generative model depends on a fundamentally different operator, called transposed convolution. This operator initially inserts a large number of zeros in its input and then slides a window over this expanded input. This zero-insertion step leads to a large number of ineffectual operations and creates distinct patterns of computation across the sliding windows. The ineffectual operations along with the variation in computation patterns lead to significant resource underutilization when using conventional convolution hardware. To alleviate these sources of inefficiency, this paper devises FlexiGAN, an end-to-end solution, that generates an optimized synthesizable FPGA accelerator from a high-level GAN specification. FlexiGAN is coupled with a novel template architecture that aims to harness the benefits of both MIMD and SIMD execution models to avoid ineffectual operations. To this end, the proposed architecture separates data retrieval and data processing units at the finest granularity of each compute engine. Leveraging this separation enables the architecture to use a succinct set of operations to cope with the irregularities of transposed convolution. At the same time, it significantly reduces the on-chip memory usage, which is generally limited in FPGAs. We evaluate our end-to-end solution by generating FPGA accelerators for a variety of GANs. These generated accelerators provide 2.4× higher performance than an optimized conventional convolution design. In addition, FlexiGAN, on average, yields 2.8× (up to 3.7×) improvements in Performance-per-Watt over a Titan X GPU.