Proceedings of the 44th Annual International Symposium on Computer Architecture

Conference Information
Name: Proceedings of the 44th Annual International Symposium on Computer Architecture

Latest articles from this conference

Jiecao Yu, Andrew Lukefahr, David Palframan, Ganesh Dasika, Reetuparna Das, Scott Mahlke
ACM SIGARCH Computer Architecture News, Volume 45, pp 548-560;

As the size of Deep Neural Networks (DNNs) continues to grow to increase accuracy and solve more complex problems, their energy footprint also scales. Weight pruning reduces DNN model size and the computation by removing redundant weights. However, we implemented weight pruning for several popular networks on a variety of hardware platforms and observed surprising results. For many networks, the network sparsity caused by weight pruning will actually hurt the overall performance despite large reductions in the model size and required multiply-accumulate operations. Also, encoding the sparse format of pruned networks incurs additional storage space overhead. To overcome these challenges, we propose Scalpel that customizes DNN pruning to the underlying hardware by matching the pruned network structure to the data-parallel hardware organization. Scalpel consists of two techniques: SIMD-aware weight pruning and node pruning. For low-parallelism hardware (e.g., microcontroller), SIMD-aware weight pruning maintains weights in aligned fixed-size groups to fully utilize the SIMD units. For high-parallelism hardware (e.g., GPU), node pruning removes redundant nodes, not redundant weights, thereby reducing computation without sacrificing the dense matrix format. For hardware with moderate parallelism (e.g., desktop CPU), SIMD-aware weight pruning and node pruning are synergistically applied together. Across the microcontroller, CPU and GPU, Scalpel achieves mean speedups of 3.54x, 2.61x, and 1.25x while reducing the model sizes by 88%, 82%, and 53%. In comparison, traditional weight pruning achieves mean speedups of 1.90x, 1.06x, 0.41x across the three platforms.
Amro Awad, Yipeng Wang, Deborah Shands, Yan Solihin
ACM SIGARCH Computer Architecture News, Volume 45, pp 107-119;

Trustworthy software requires strong privacy and security guarantees from a secure trust base in hardware. While chipmakers provide hardware support for basic security and privacy primitives such as enclaves and memory encryption. these primitives do not address hiding of the memory access pattern, information about which may enable attacks on the system or reveal characteristics of sensitive user data. State-of-the-art approaches to protecting the access pattern are largely based on Oblivious RAM (ORAM). Unfortunately, current ORAM implementations suffer from very significant practicality and overhead concerns, including roughly an order of magnitude slowdown, more than 100% memory capacity overheads, and the potential for system deadlock. Memory technology trends are moving towards 3D and 2.5D integration, enabling significant logic capabilities and sophisticated memory interfaces. Leveraging the trends, we propose a new approach to access pattern obfuscation, called ObfusMem. ObfusMem adds the memory to the trusted computing base and incorporates cryptographic engines within the memory. ObfusMem encrypts commands and addresses on the memory bus, hence the access pattern is cryptographically obfuscated from external observers. Our evaluation shows that ObfusMem incurs an overhead of 10.9% on average, which is about an order of magnitude faster than ORAM implementations. Furthermore, ObfusMem does not incur capacity overheads and does not amplify writes. We analyze and compare the security protections provided by ObfusMem and ORAM, and highlight their differences.
, Longfei Wang, Weize Yu, Selçuk Köse, Ulya R. Karpuzcu
ACM SIGARCH Computer Architecture News, Volume 45, pp 120-132;

Tailoring the operating voltage to fine-grain temporal changes in the power and performance needs of the workload can effectively enhance power efficiency. Therefore, power-limited computing platforms of today widely deploy integrated (i.e., on-chip) voltage regulation which enables fast fine-grain voltage control. Voltage regulators convert and distribute power from an external energy source to the processor. Unfortunately, power conversion loss is inevitable and projected integrated regulator designs are unlikely to eliminate this loss even asymptotically. Reconfigurable power delivery by selective shut-down, i.e., gating, of distributed on-chip regulators in response to spatio-temporal changes in power demand can sustain operation at the minimum conversion loss. However, even the minimum conversion loss is sizable, and as conversion loss gets dissipated as heat, on-chip regulators can easily cause thermal emergencies due to their small footprint. Although reconfigurable distributed on-chip power delivery is emerging as a new design paradigm to enforce sustained operation at minimum possible power conversion loss, thermal implications have been overlooked at the architectural level. This paper hence provides a thermal characterization. We introduce ThermoGater, an architectural governor for a collection of practical, thermally-aware regulator gating policies to mitigate (if not prevent) regulator-induced thermal emergencies, which also consider potential implications for voltage noise. Practical ThermoGater policies can not only sustain minimum power conversion loss throughout execution effectively, but also keep the maximum temperature (thermal gradient) across chip within 0.6°C (0.3°C) on average in comparison to thermally-optimal oracular regulator gating, while the maximum voltage noise stays within 1.0% of the best case voltage noise profile.
Muhammad Shoaib Bin Altaf, David A. Wood
ACM SIGARCH Computer Architecture News, Volume 45, pp 375-388;

With the end of Dennard scaling, architects have increasingly turned to special-purpose hardware accelerators to improve the performance and energy efficiency for some applications. Unfortunately, accelerators don't always live up to their expectations and may under-perform in some situations. Understanding the factors which effect the performance of an accelerator is crucial for both architects and programmers early in the design stage. Detailed models can be highly accurate, but often require low-level details which are not available until late in the design cycle. In contrast, simple analytical models can provide useful insights by abstracting away low-level system details. In this paper, we propose LogCA---a high-level performance model for hardware accelerators. LogCA helps both programmers and architects identify performance bounds and design bottlenecks early in the design cycle, and provide insight into which optimizations may alleviate these bottlenecks. We validate our model across a variety of kernels, ranging from sub-linear to super-linear complexities on both on-chip and off-chip accelerators. We also describe the utility of LogCA using two retrospective case studies. First, we discuss the evolution of interface design in SUN/Oracle's encryption accelerators. Second, we discuss the evolution of memory interface design in three different GPU architectures. In both cases, we show that the adopted design optimizations for these machines are similar to LogCA's suggested optimizations. We argue that architects and programmers can use insights from these retrospective studies for improving future designs.
Chang Hyun Park, Taekyung Heo, Jungi Jeong, Jaehyuk Huh
ACM SIGARCH Computer Architecture News, Volume 45, pp 444-456;

To mitigate excessive TLB misses in large memory applications, techniques such as large pages, variable length segments, and HW coalescing, increase the coverage of limited hardware translation entries by exploiting the contiguous memory allocation. However, recent studies show that in non-uniform memory systems, using large pages often leads to performance degradation, or allocating large chunks of memory becomes more difficult due to memory fragmentation. Although each of the prior techniques favors its own best chunk size, diverse contiguity of memory allocation in real systems cannot always provide the optimal chunk of each technique. Under such fragmented and diverse memory allocations, this paper proposes a novel HW-SW hybrid translation architecture, which can adapt to different memory mappings efficiently. In the proposed hybrid coalescing technique, the operating system encodes memory contiguity information in a subset of page table entries, called anchor entries. During address translation through TLBs, an anchor entry provides translation for contiguous pages following the anchor entry. As a smaller number of anchor entries can cover a large portion of virtual address space, the efficiency of TLB can be significantly improved. The most important benefit of hybrid coalescing is its ability to change the coverage of the anchor entry dynamically, reflecting the current allocation contiguity status. By using the contiguity information directly set by the operating system, the technique can provide scalable translation coverage improvements with minor hardware changes, while allowing the flexibility of memory allocation. Our experimental results show that across diverse allocation scenarios with different distributions of contiguous memory chunks, the proposed scheme can effectively reap the potential translation coverage improvement from the existing contiguity.
Rajat Kateja, Anirudh Badam, Sriram Govindan, Bikash Sharma, Greg Ganger
ACM SIGARCH Computer Architecture News, Volume 45, pp 613-626;

Non-Volatile Memories (NVMs) can significantly improve the performance of data-intensive applications. A popular form of NVM is Battery-backed DRAM, which is available and in use today with DRAMs latency and without the endurance problems of emerging NVM technologies. Modern servers can be provisioned with up-to 4 TB of DRAM, and provisioning battery backup to write out such large memories is hard because of the large battery sizes and the added hardware and cooling costs. We present Viyojit, a system that exploits the skew in write working sets of applications to provision substantially smaller batteries while still ensuring durability for the entire DRAM capacity. Viyojit achieves this by bounding the number of dirty pages in DRAM based on the provisioned battery capacity and proactively writing out infrequently written pages to an SSD. Even for write-heavy workloads with less skew than we observe in analysis of real data center traces, Viyojit reduces the required battery capacity to 11% of the original size, with a performance overhead of 7-25%. Thus, Viyojit frees battery-backed DRAM from stunted growth of battery capacities and enables servers with terabytes of battery-backed DRAM.
Mario Drumond, Alexandros Daglis, Nooshin Mirzadeh, Dmitrii Ustiugov, Javier Picorel, Babak Falsafi, Boris Grot, Dionisios Pnevmatikatos
ACM SIGARCH Computer Architecture News, Volume 45, pp 639-651;

The increasing demand for extracting value out of ever-growing data poses an ongoing challenge to system designers, a task only made trickier by the end of Dennard scaling. As the performance density of traditional CPU-centric architectures stagnates, advancing compute capabilities necessitates novel architectural approaches. Near-memory processing (NMP) architectures are reemerging as promising candidates to improve computing efficiency through tight coupling of logic and memory. NMP architectures are especially fitting for data analytics, as they provide immense bandwidth to memory-resident data and dramatically reduce data movement, the main source of energy consumption. Modern data analytics operators are optimized for CPU execution and hence rely on large caches and employ random memory accesses. In the context of NMP, such random accesses result in wasteful DRAM row buffer activations that account for a significant fraction of the total memory access energy. In addition, utilizing NMP's ample bandwidth with fine-grained random accesses requires complex hardware that cannot be accommodated under NMP's tight area and power constraints. Our thesis is that efficient NMP calls for an algorithm-hardware co-design that favors algorithms with sequential accesses to enable simple hardware that accesses memory in streams. We introduce an instance of such a co-designed NMP architecture for data analytics, the Mondrian Data Engine. Compared to a CPU-centric and a baseline NMP system, the Mondrian Data Engine improves the performance of basic data analytics operators by up to 49x and 5x, and efficiency by up to 28x and 5x, respectively.
Jaeha Kung, Yun Long, Duckhwan Kim, Saibal Mukhopadhyay
ACM SIGARCH Computer Architecture News, Volume 45, pp 403-415;

The fast and energy-efficient simulation of dynamical systems defined by coupled ordinary/partial differential equations has emerged as an important problem. The accelerated simulation of coupled ODE/PDE is critical for analysis of physical systems as well as computing with dynamical systems. This paper presents a fast and programmable accelerator for simulating dynamical systems. The computing model of the proposed platform is based on multilayer cellular nonlinear network (CeNN) augmented with nonlinear function evaluation engines. The platform can be programmed to accelerate wide classes of ODEs/PDEs by modulating the connectivity within the multilayer CeNN engine. An innovative hardware architecture including data reuse, memory hierarchy, and near-memory processing is designed to accelerate the augmented multilayer CeNN. A dataflow model is presented which is supported by optimized memory hierarchy for efficient function evaluation. The proposed solver is designed and synthesized in 15nm technology for the hardware analysis. The performance is evaluated and compared to GPU nodes when solving wide classes of differential equations and the power consumption is analyzed to show orders of magnitude improvement in energy efficiency.
Raghu Prabhakar, Yaqi Zhang, David Koeplinger, Matt Feldman, Tian Zhao, Stefan Hadjis, Ardavan Pedram, Christos Kozyrakis, Kunle Olukotun
ACM SIGARCH Computer Architecture News, Volume 45, pp 389-402;

Reconfigurable architectures have gained popularity in recent years as they allow the design of energy-efficient accelerators. Fine-grain fabrics (e.g. FPGAs) have traditionally suffered from performance and power inefficiencies due to bit-level reconfigurable abstractions. Both fine-grain and coarse-grain architectures (e.g. CGRAs) traditionally require low level programming and suffer from long compilation times. We address both challenges with Plasticine, a new spatially reconfigurable architecture designed to efficiently execute applications composed of parallel patterns. Parallel patterns have emerged from recent research on parallel programming as powerful, high-level abstractions that can elegantly capture data locality, memory access patterns, and parallelism across a wide range of dense and sparse applications. We motivate Plasticine by first observing key application characteristics captured by parallel patterns that are amenable to hardware acceleration, such as hierarchical parallelism, data locality, memory access patterns, and control flow. Based on these observations, we architect Plasticine as a collection of Pattern Compute Units and Pattern Memory Units. Pattern Compute Units are multi-stage pipelines of reconfigurable SIMD functional units that can efficiently execute nested patterns. Data locality is exploited in Pattern Memory Units using banked scratchpad memories and configurable address decoders. Multiple on-chip address generators and scatter-gather engines make efficient use of DRAM bandwidth by supporting a large number of outstanding memory requests, memory coalescing, and burst mode for dense accesses. Plasticine has an area footprint of 113 mm2 in a 28nm process, and consumes a maximum power of 49 W at a 1 GHz clock. Using a cycle-accurate simulator, we demonstrate that Plasticine provides an improvement of up to 76.9x in performance-per-Watt over a conventional FPGA over a wide range of dense and sparse applications.
Gunjae Koo, Yunho Oh, Won Woo Ro, Murali Annavaram
ACM SIGARCH Computer Architecture News, Volume 45, pp 307-319;

Long latency of memory operation is a prominent performance bottleneck in graphics processing units (GPUs). The small data cache that must be shared across dozens of warps (a collection of threads) creates significant cache contention and premature data eviction. Prior works have recognized this problem and proposed warp throttling which reduces the number of active warps contending for cache space. In this paper we discover that individual load instructions in a warp exhibit four different types of data locality behavior: (1) data brought by a warp load instruction is used only once, which is classified as streaming data (2) data brought by a warp load is reused multiple times within the same warp, called intra-warp locality (3) data brought by a warp is reused multiple times but across different warps, called inter-warp locality (4) and some data exhibit both a mix of intra- and inter-warp locality. Furthermore, each load instruction exhibits consistently the same locality type across all warps within a GPU kernel. Based on this discovery we argue that cache management must be done using per-load locality type information, rather than applying warp-wide cache management policies. We propose Access Pattern-aware Cache Management (APCM), which dynamically detects the locality type of each load instruction by monitoring the accesses from one exemplary warp. APCM then uses the detected locality type to selectively apply cache bypassing and cache pinning of data based on load locality characterization. Using an extensive set of simulations we show that APCM improves performance of GPUs by 34% for cache sensitive applications while saving 27% of energy consumption over baseline GPU.
Back to Top Top