اشترك بالحزمة الذهبية واحصل على وصول غير محدود شمرا أكاديميا
تسجيل مستخدم جديدUnion: A Unified HW-SW Co-Design Ecosystem in MLIR for Evaluating Tensor Operations on Spatial Accelerators
184
0
0.0
(
0
)
اسأل ChatGPT حول البحث
ﻻ يوجد ملخص باللغة العربية
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.
قيم البحث
اقرأ أيضاً
Hardware/Software (HW/SW) co-designed processors provide a promising solution to the power and complexity problems of the modern microprocessors by keeping their hardware simple. Moreover, they employ several runtime optimizations to improve the perf
ormance. One of the most potent optimizations, vectorization, has been utilized by modern microprocessors, to exploit the data level parallelism through SIMD accelerators. Due to their hardware simplicity, these accelerators have evolved in terms of width from 64-bit vectors in Intel MMX to 512-bit wide vector units in Intel Xeon Phi and AVX-512. Although SIMD accelerators are simple in terms of hardware design, code generation for them has always been a challenge. Moreover, increasing vector lengths with each new generation add to this complexity. This paper explores the scalability of SIMD accelerators from the code generation point of view. We discover that the SIMD accelerators remain underutilized at higher vector lengths mainly due to: a) reduced dynamic instruction stream coverage for vectorization and b) increase in permutations. Both of these factors can be attributed to the rigidness of the SIMD architecture. We propose a novel SIMD architecture that possesses the flexibility needed to support higher vector lengths. Furthermore, we propose Variable Length Vectorization and Selective Writing in a HW/SW co-designed environment to transparently target the flexibility of the proposed architecture. We evaluate our proposals using a set of SPECFP2006 and Physicsbench applications. Our experimental results show an average dynamic instruction reduction of 31% and 40% and an average speed up of 13% and 10% for SPECFP2006 and Physicsbench respectively, for 512-bit vector length, over the scalar baseline code.
Tiled spatial architectures have proved to be an effective solution to build large-scale DNN accelerators. In particular, interconnections between tiles are critical for high performance in these tile-based architectures. In this work, we identify th
e inefficiency of the widely used traditional on-chip networks and the opportunity of software-hardware co-design. We propose METRO with the basic idea of decoupling the traffic scheduling policies from hardware fabrics and moving them to the software level. METRO contains two modules working in synergy: METRO software scheduling framework to coordinate the traffics and METRO hardware facilities to deliver the data based on software configurations. We evaluate the co-design using different flit sizes for synthetic study, illustrating its effectiveness under various hardware resource constraints, in addition to a wide range of DNN models selected from real-world workloads. The results show that METRO achieves 56.3% communication speedup on average and up to 73.6% overall processing time reduction compared with traditional on-chip network designs.
End-to-end performance estimation and measurement of deep neural network (DNN) systems become more important with increasing complexity of DNN systems consisting of hardware and software components. The methodology proposed in this paper aims at a re
duced turn-around time for evaluating different design choices of hardware and software components of DNN systems. This reduction is achieved by moving the performance estimation from the implementation phase to the concept phase by employing virtual hardware models instead of gathering measurement results from physical prototypes. Deep learning compilers introduce hardware-specific transformations and are, therefore, considered a part of the design flow of virtual system models to extract end-to-end performance estimations. To validate the run-time accuracy of the proposed methodology, a system processing the DilatedVGG DNN is realized both as virtual system model and as hardware implementation. The results show that up to 92 % accuracy can be reached in predicting the processing time of the DNN inference.
Many FPGAs vendors have recently included embedded processors in their devices, like Xilinx with ARM-Cortex A cores, together with programmable logic cells. These devices are known as Programmable System on Chip (PSoC). Their ARM cores (embedded in t
he processing system or PS) communicates with the programmable logic cells (PL) using ARM-standard AXI buses. In this paper we analyses the performance of exhaustive data transfers between PS and PL for a Xilinx Zynq FPGA in a co-design real scenario for Convolutional Neural Networks (CNN) accelerator, which processes, in dedicated hardware, a stream of visual information from a neuromorphic visual sensor for classification. In the PS side, a Linux operating system is running, which recollects visual events from the neuromorphic sensor into a normalized frame, and then it transfers these frames to the accelerator of multi-layered CNNs, and read results, using an AXI-DMA bus in a per-layer way. As these kind of accelerators try to process information as quick as possible, data bandwidth becomes critical and maintaining a good balanced data throughput rate requires some considerations. We present and evaluate several data partitioning techniques to improve the balance between RX and TX transfer and two different ways of transfers management: through a polling routine at the userlevel of the OS, and through a dedicated interrupt-based kernellevel driver. We demonstrate that for longer enough packets, the kernel-level driver solution gets better timing in computing a CNN classification example. Main advantage of using kernel-level driver is to have safer solutions and to have tasks scheduling in the OS to manage other important processes for our application, like frames collection from sensors and their normalization.
Transfer learning in natural language processing (NLP), as realized using models like BERT (Bi-directional Encoder Representation from Transformer), has significantly improved language representation with models that can tackle challenging language p
roblems. Consequently, these applications are driving the requirements of future systems. Thus, we focus on BERT, one of the most popular NLP transfer learning algorithms, to identify how its algorithmic behavior can guide future accelerator design. To this end, we carefully profile BERT training and identify key algorithmic behaviors which are worthy of attention in accelerator design. We observe that while computations which manifest as matrix multiplication dominate BERTs overall runtime, as in many convolutional neural networks, memory-intensive computations also feature prominently. We characterize these computations, which have received little attention so far. Further, we also identify heterogeneity in compute-intensive BERT computations and discuss software and possible hardware mechanisms to further optimize these computations. Finally, we discuss implications of these behaviors as networks get larger and use distributed training environments, and how techniques such as micro-batching and mixed-precision training scale. Overall, our analysis identifies holistic solutions to optimize systems for BERT-like models.
الأسئلة المقترحة
سجل دخول لتتمكن من نشر تعليقات
التعليقات
جاري جلب التعليقات
سجل دخول لتتمكن من متابعة معايير البحث التي قمت باختيارها
