arXiv subjects

Tushar Krishna

Publications and source records attributed to Tushar Krishna.

At least 91 records · Page 5Linked to original sources

arXiv2023

VEGETA: Vertically-Integrated Extensions for Sparse/Dense GEMM Tile Acceleration on CPUs

Deep Learning (DL) acceleration support in CPUs has recently gained a lot of traction, with several companies (Arm, Intel, IBM) announcing products with specialized matrix engines accessible via GEMM instructions. CPUs are pervasive and need to handle diverse requirements across DL workloads running in edge/HPC/cloud platforms. Therefore, as DL workloads embrace sparsity to reduce the computations and memory size of models, it is also imperative for CPUs to add support for sparsity to avoid under-utilization of the dense matrix engine and inefficient usage of the caches and registers. This work presents VEGETA, a set of ISA and microarchitecture extensions over dense matrix engines to support flexible structured sparsity for CPUs, enabling programmable support for diverse DL models with varying degrees of sparsity. Compared to the state-of-the-art (SOTA) dense matrix engine in CPUs, a VEGETA engine provides 1.09x, 2.20x, 3.74x, and 3.28x speed-ups when running 4:4 (dense), 2:4, 1:4, and unstructured (95%) sparse DNN layers.

cs.AR↗

arXiv2023

Flexagon: A Multi-Dataflow Sparse-Sparse Matrix Multiplication Accelerator for Efficient DNN Processing

Sparsity is a growing trend in modern DNN models. Existing Sparse-Sparse Matrix Multiplication (SpMSpM) accelerators are tailored to a particular SpMSpM dataflow (i.e., Inner Product, Outer Product or Gustavsons), that determines their overall efficiency. We demonstrate that this static decision inherently results in a suboptimal dynamic solution. This is because different SpMSpM kernels show varying features (i.e., dimensions, sparsity pattern, sparsity degree), which makes each dataflow better suited to different data sets. In this work we present Flexagon, the first SpMSpM reconfigurable accelerator that is capable of performing SpMSpM computation by using the particular dataflow that best matches each case. Flexagon accelerator is based on a novel Merger-Reduction Network (MRN) that unifies the concept of reducing and merging in the same substrate, increasing efficiency. Additionally, Flexagon also includes a 3-tier memory hierarchy, specifically tailored to the different access characteristics of the input and output compressed matrices. Using detailed cycle-level simulation of contemporary DNN models from a variety of application domains, we show that Flexagon achieves average performance benefits of 4.59x, 1.71x, and 1.35x with respect to the state-of-the-art SIGMA-like, Sparch-like and GAMMA-like accelerators (265% , 67% and 18%, respectively, in terms of average performance/area efficiency).

cs.AR↗

arXiv2022

Demystifying Map Space Exploration for NPUs

Map Space Exploration is the problem of finding optimized mappings of a Deep Neural Network (DNN) model on an accelerator. It is known to be extremely computationally expensive, and there has been active research looking at both heuristics and learning-based methods to make the problem computationally tractable. However, while there are dozens of mappers out there (all empirically claiming to find better mappings than others), the research community lacks systematic insights on how different search techniques navigate the map-space and how different mapping axes contribute to the accelerator's performance and efficiency. Such insights are crucial to developing mapping frameworks for emerging DNNs that are increasingly irregular (due to neural architecture search) and sparse, making the corresponding map spaces much more complex. In this work, rather than proposing yet another mapper, we do a first-of-its-kind apples-to-apples comparison of search techniques leveraged by different mappers. Next, we extract the learnings from our study and propose two new techniques that can augment existing mappers -- warm-start and sparsity-aware -- that demonstrate speedups, scalability, and robustness across diverse DNN models.

cs.LG↗

arXiv2022

FLAT: An Optimized Dataflow for Mitigating Attention Bottlenecks

Attention mechanisms, primarily designed to capture pairwise correlations between words, have become the backbone of machine learning, expanding beyond natural language processing into other domains. This growth in adaptation comes at the cost of prohibitively large memory requirements and computational complexity, especially at higher number of input elements. This limitation is due to inherently limited data reuse opportunities and quadratic growth in memory footprints, leading to severe memory-boundedness and limited scalability of input elements. This work addresses these challenges by devising a tailored dataflow optimization, called FLAT, for attention mechanisms without altering their functionality. This dataflow processes costly attention operations through a unique fusion mechanism, transforming the memory footprint quadratic growth to merely a linear one. To realize the full potential of this bespoke mechanism, we propose a tiling approach to enhance the data reuse across attention operations. Our method both mitigates the off-chip bandwidth bottleneck as well as reduces the on-chip memory requirement. FLAT delivers 1.94x (1.76x) speedup and 49% and (42%) of energy savings compared to the state-of-the-art Edge (Cloud) accelerators with no customized dataflow optimization. When on-chip resources are scarce (20 KB-200 KB), FLAT yields, on average, 1.5x end-to-end latency reduction across a diverse range of conventional attention-based models with input sequence lengths ranging from 512-token to 64K-token. Our evaluations demonstrate that state-of-the-art DNN dataflow applied to attention operations reach the efficiency limit for inputs above 512 elements. In contrast, FLAT unblocks transformer models for inputs with up to 64K elements

cs.LG↗

arXiv2022

Training Recipe for N:M Structured Sparsity with Decaying Pruning Mask

Sparsity has become one of the promising methods to compress and accelerate Deep Neural Networks (DNNs). Among different categories of sparsity, structured sparsity has gained more attention due to its efficient execution on modern accelerators. Particularly, N:M sparsity is attractive because there are already hardware accelerator architectures that can leverage certain forms of N:M structured sparsity to yield higher compute-efficiency. In this work, we focus on N:M sparsity and extensively study and evaluate various training recipes for N:M sparsity in terms of the trade-off between model accuracy and compute cost (FLOPs). Building upon this study, we propose two new decay-based pruning methods, namely "pruning mask decay" and "sparse structure decay". Our evaluations indicate that these proposed methods consistently deliver state-of-the-art (SOTA) model accuracy, comparable to unstructured sparsity, on a Transformer-based model for a translation task. The increase in the accuracy of the sparse model using the new training recipes comes at the cost of marginal increase in the total training compute (FLOPs).

cs.LG↗

arXiv2022

Impact of RoCE Congestion Control Policies on Distributed Training of DNNs

RDMA over Converged Ethernet (RoCE) has gained significant attraction for datacenter networks due to its compatibility with conventional Ethernet-based fabric. However, the RDMA protocol is efficient only on (nearly) lossless networks, emphasizing the vital role of congestion control on RoCE networks. Unfortunately, the native RoCE congestion control scheme, based on Priority Flow Control (PFC), suffers from many drawbacks such as unfairness, head-of-line-blocking, and deadlock. Therefore, in recent years many schemes have been proposed to provide additional congestion control for RoCE networks to minimize PFC drawbacks. However, these schemes are proposed for general datacenter environments. In contrast to the general datacenters that are built using commodity hardware and run general-purpose workloads, high-performance distributed training platforms deploy high-end accelerators and network components and exclusively run training workloads using collectives (All-Reduce, All-To-All) communication libraries for communication. Furthermore, these platforms usually have a private network, separating their communication traffic from the rest of the datacenter traffic. Scalable topology-aware collective algorithms are inherently designed to avoid incast patterns and balance traffic optimally. These distinct features necessitate revisiting previously proposed congestion control schemes for general-purpose datacenter environments. In this paper, we thoroughly analyze some of the SOTA RoCE congestion control schemes vs. PFC when running on distributed training platforms. Our results indicate that previously proposed RoCE congestion control schemes have little impact on the end-to-end performance of training workloads, motivating the necessity of designing an optimized, yet low-overhead, congestion control scheme based on the characteristics of distributed training platforms and workloads.

cs.NI↗

arXiv2022

Themis: A Network Bandwidth-Aware Collective Scheduling Policy for Distributed Training of DL Models

Distributed training is a solution to reduce DNN training time by splitting the task across multiple NPUs (e.g., GPU/TPU). However, distributed training adds communication overhead between the NPUs in order to synchronize the gradients and/or activation, depending on the parallelization strategy. In next-generation platforms for training at scale, NPUs will be connected through multi-dimensional networks with diverse, heterogeneous bandwidths. This work identifies a looming challenge of keeping all network dimensions busy and maximizing the network BW within the hybrid environment if we leverage scheduling techniques for collective communication on systems today. We propose Themis, a novel collective scheduling scheme that dynamically schedules collectives (divided into chunks) to balance the communication loads across all dimensions, further improving the network BW utilization. Our results show that on average, Themis can improve the network BW utilization of the single All-Reduce by 1.72X (2.70X max), and improve the end-to-end training iteration performance of real workloads such as ResNet-152, GNMT, DLRM, and Transformer-1T by 1.49X (2.25X max), 1.30X (1.78X max), 1.30X (1.77X max), and 1.25X (1.53X max), respectively.

cs.DC↗

arXiv2022

DNNFuser: Generative Pre-Trained Transformer as a Generalized Mapper for Layer Fusion in DNN Accelerators

Dataflow/mapping decides the compute and energy efficiency of DNN accelerators. Many mappers have been proposed to tackle the intra-layer map-space. However, mappers for inter-layer map-space (aka layer-fusion map-space), have been rarely discussed. In this work, we propose a mapper, DNNFuser, specifically focusing on this layer-fusion map-space. While existing SOTA DNN mapping explorations rely on search-based mappers, this is the first work, to the best of our knowledge, to propose a one-shot inference-based mapper. We leverage Transformer as our DNN architecture to learn layer-fusion optimization as a sequence modeling problem. Further, the trained DNNFuser can generalize its knowledge and infer new solutions for unseen conditions. Within one inference pass, DNNFuser can infer solutions with compatible performance to the ones found by a highly optimized search-based mapper while being 66x-127x faster.

cs.LG↗

arXiv2022

A Formalism of DNN Accelerator Flexibility

The high efficiency of domain-specific hardware accelerators for machine learning (ML) has come from specialization, with the trade-off of less configurability/ flexibility. There is growing interest in developing flexible ML accelerators to make them future-proof to the rapid evolution of Deep Neural Networks (DNNs). However, the notion of accelerator flexibility has always been used in an informal manner, restricting computer architects from conducting systematic apples-to-apples design-space exploration (DSE) across trillions of choices. In this work, we formally define accelerator flexibility and show how it can be integrated for DSE. Specifically, we capture DNN accelerator flexibility across four axes: tiling, ordering, parallelization, and array shape. We categorize existing accelerators into 16 classes based on their axes of flexibility support, and define a precise quantification of the degree of flexibility of an accelerator across each axis. We leverage these to develop a novel flexibility-aware DSE framework. We demonstrate how this can be used to perform first-of-their-kind evaluations, including an isolation study to identify the individual impact of the flexibility axes. We demonstrate that adding flexibility features to a hypothetical DNN accelerator designed in 2014 improves runtime on future (i.e., present-day) DNNs by 11.8x geomean.

cs.AR↗

arXiv2022

Enabling Compute-Communication Overlap in Distributed Deep Learning Training Platforms

Deep Learning (DL) training platforms are built by interconnecting multiple DL accelerators (e.g., GPU/TPU) via fast, customized interconnects with 100s of gigabytes (GBs) of bandwidth. However, as we identify in this work, driving this bandwidth is quite challenging. This is because there is a pernicious balance between using the accelerator's compute and memory for both DL computations and communication. This work makes two key contributions. First, via real system measurements and detailed modeling, we provide an understanding of compute and memory bandwidth demands for DL compute and comms. Second, we propose a novel DL collective communication accelerator called Accelerator Collectives Engine (ACE) that sits alongside the compute and networking engines at the accelerator endpoint. ACE frees up the endpoint's compute and memory resources for DL compute, which in turn reduces the required memory BW by 3.5X on average to drive the same network BW compared to state-of-the-art baselines. For modern DL workloads and different network sizes, ACE, on average, increases the effective network bandwidth utilization by 1.44X (up to 2.67X), resulting in an average of 1.41X (up to 1.51X), 1.12X (up to 1.17X), and 1.13X (up to 1.19X) speedup in iteration time for ResNet-50, GNMT and DLRM when compared to the best baseline configuration, respectively.

cs.AR↗

arXiv2022

Self-Adaptive Reconfigurable Arrays (SARA): Using ML to Assist Scaling GEMM Acceleration

With increasing diversity in Deep Neural Network(DNN) models in terms of layer shapes and sizes, the research community has been investigating flexible/reconfigurable accelerator substrates. This line of research has opened up two challenges. The first is to determine the appropriate amount of flexibility within an accelerator array that that can trade-off the performance benefits versus the area overheads of the reconfigurability. The second is being able to determine the right configuration of the array for the current DNN model and/or layer and reconfigure the accelerator at runtime. This work introduces a new class of accelerators that we call Self Adaptive Reconfigurable Array (SARA). SARA architectures comprise of both a reconfigurable array and a hardware unit capable of determining an optimized configuration for the array at runtime. We demonstrate an instance of SARA with an accelerator we call SAGAR, which introduces a novel reconfigurable systolic array that can be configured to work as a distributed collection of smaller arrays of various sizes or as a single array with flexible aspect ratios. We also develop a novel recommendation neural network called ADAPTNET which recommends an array configuration and dataflow for the current layer parameters. ADAPTNET runs on an integrated custom hardware ADAPTNETX that runs ADAPTNET at runtime and reconfigures the array, making the entire accelerator self-sufficient. SAGAR is capable of providing the same mapping flexibility as a collection of 1024 4x4 arrays working as a distributed system while achieving 3.5x more power efficiency and 3.2x higher compute density Furthermore, the runtime achieved on the recommended parameters from ADAPTNET is 99.93% of the best achievable runtime.

cs.AR↗

arXiv2022

Understanding the Design-Space of Sparse/Dense Multiphase GNN dataflows on Spatial Accelerators

Graph Neural Networks (GNNs) have garnered a lot of recent interest because of their success in learning representations from graph-structured data across several critical applications in cloud and HPC. Owing to their unique compute and memory characteristics that come from an interplay between dense and sparse phases of computations, the emergence of reconfigurable dataflow (aka spatial) accelerators offers promise for acceleration by mapping optimized dataflows (i.e., computation order and parallelism) for both phases. The goal of this work is to characterize and understand the design-space of dataflow choices for running GNNs on spatial accelerators in order for mappers or design-space exploration tools to optimize the dataflow based on the workload. Specifically, we propose a taxonomy to describe all possible choices for mapping the dense and sparse phases of GNN inference, spatially and temporally over a spatial accelerator, capturing both the intra-phase dataflow and the inter-phase (pipelined) dataflow. Using this taxonomy, we do deep-dives into the cost and benefits of several dataflows and perform case studies on implications of hardware parameters for dataflows and value of flexibility to support pipelined execution.

cs.DC↗

arXiv2022

MAGMA: An Optimization Framework for Mapping Multiple DNNs on Multiple Accelerator Cores

As Deep Learning continues to drive a variety of applications in edge and cloud data centers, there is a growing trend towards building large accelerators with several sub-accelerator cores/chiplets. This work looks at the problem of supporting multi-tenancy on such accelerators. In particular, we focus on the problem of mapping jobs from several DNNs simultaneously on an accelerator. Given the extremely large search space, we formulate the search as an optimization problem and develop an optimization framework called M3E. In addition, we develop a specialized optimization algorithm called MAGMA with custom operators to enable structured sample-efficient exploration. We quantitatively compare MAGMA with several state-of-the-art methods, black-box optimization, and reinforcement learning methods across different accelerator settings (large/small accelerators) and different sub-accelerator configurations (homogeneous/heterogeneous), and observe MAGMA can consistently find better mappings.

cs.AR↗

arXiv2022

DiGamma: Domain-aware Genetic Algorithm for HW-Mapping Co-optimization for DNN Accelerators

The design of DNN accelerators includes two key parts: HW resource configuration and mapping strategy. Intensive research has been conducted to optimize each of them independently. Unfortunately, optimizing for both together is extremely challenging due to the extremely large cross-coupled search space. To address this, in this paper, we propose a HW-Mapping co-optimization framework, an efficient encoding of the immense design space constructed by HW and Mapping, and a domain-aware genetic algorithm, named DiGamma, with specialized operators for improving search efficiency. We evaluate DiGamma with seven popular DNNs models with different properties. Our evaluations show DiGamma can achieve (geomean) 3.0x and 10.0x speedup, comparing to the best-performing baseline optimization algorithms, in edge and cloud settings.

cs.NE↗

arXiv2022

Enabling Flexibility for Sparse Tensor Acceleration via Heterogeneity

Recently, numerous sparse hardware accelerators for Deep Neural Networks (DNNs), Graph Neural Networks (GNNs), and scientific computing applications have been proposed. A common characteristic among all of these accelerators is that they target tensor algebra (typically matrix multiplications); yet dozens of new accelerators are proposed for every new application. The motivation is that the size and sparsity of the workloads heavily influence which architecture is best for memory and computation efficiency. To satisfy the growing demand of efficient computations across a spectrum of workloads on large data centers, we propose deploying a flexible 'heterogeneous' accelerator, which contains many 'sub-accelerators' (smaller specialized accelerators) working together. To this end, we propose: (1) HARD TACO, a quick and productive C++ to RTL design flow to generate many types of sub-accelerators for sparse and dense computations for fair design-space exploration, (2) AESPA, a heterogeneous sparse accelerator design template constructed with the sub-accelerators generated from HARD TACO, and (3) a suite of scheduling strategies to map tensor kernels onto heterogeneous sparse accelerators with high efficiency and utilization. AESPA with optimized scheduling achieves 1.96X higher performance, and 7.9X better energy-delay product (EDP) than a Homogeneous EIE-like accelerator with our diverse workload suite.

cs.AR↗

arXiv2021

Union: A Unified HW-SW Co-Design Ecosystem in MLIR for Evaluating Tensor Operations on Spatial Accelerators

To meet the extreme compute demands for deep learning across commercial and scientific applications, dataflow accelerators are becoming increasingly popular. While these "domain-specific" accelerators are not fully programmable like CPUs and GPUs, they retain varying levels of flexibility with respect to data orchestration, i.e., dataflow and tiling optimizations to enhance efficiency. There are several challenges when designing new algorithms and mapping approaches to execute the algorithms for a target problem on new hardware. Previous works have addressed these challenges individually. To address this challenge as a whole, in this work, we present a HW-SW co-design ecosystem for spatial accelerators called Union within the popular MLIR compiler infrastructure. Our framework allows exploring different algorithms and their mappings on several accelerator cost models. Union also includes a plug-and-play library of accelerator cost models and mappers which can easily be extended. The algorithms and accelerator cost models are connected via a novel mapping abstraction that captures the map space of spatial accelerators which can be systematically pruned based on constraints from the hardware, workload, and mapper. We demonstrate the value of Union for the community with several case studies which examine offloading different tensor operations(CONV/GEMM/Tensor Contraction) on diverse accelerator architectures using different mapping schemes.

cs.AR↗

arXiv2021

RASA: Efficient Register-Aware Systolic Array Matrix Engine for CPU

As AI-based applications become pervasive, CPU vendors are starting to incorporate matrix engines within the datapath to boost efficiency. Systolic arrays have been the premier architectural choice as matrix engines in offload accelerators. However, we demonstrate that incorporating them inside CPUs can introduce under-utilization and stalls due to limited register storage to amortize the fill and drain times of the array. To address this, we propose RASA, Register-Aware Systolic Array. We develop techniques to divide an execution stage into several sub-stages and overlap instructions to hide overheads and run them concurrently. RASA-based designs improve performance significantly with negligible area and power overhead.

cs.AR↗

arXiv2021

AIRCHITECT: Learning Custom Architecture Design and Mapping Space

Design space exploration is an important but costly step involved in the design/deployment of custom architectures to squeeze out maximum possible performance and energy efficiency. Conventionally, optimizations require iterative sampling of the design space using simulation or heuristic tools. In this paper we investigate the possibility of learning the optimization task using machine learning and hence using the learnt model to predict optimal parameters for the design and mapping space of custom architectures, bypassing any exploration step. We use three case studies involving the optimal array design, SRAM buffer sizing, mapping, and schedule determination for systolic-array-based custom architecture design and mapping space. Within the purview of these case studies, we show that it is possible to capture the design space and train a model to "generalize" prediction the optimal design and mapping parameters when queried with workload and design constraints. We perform systematic design-aware and statistical analysis of the optimization space for our case studies and highlight the patterns in the design space. We formulate the architecture design and mapping as a machine learning problem that allows us to leverage existing ML models for training and inference. We design and train a custom network architecture called AIRCHITECT, which is capable of learning the architecture design space with as high as 94.3% test accuracy and predicting optimal configurations which achieve on average (GeoMean) of 99.9% the best possible performance on a test dataset with $10^5$ GEMM workloads.

cs.LG↗