International Technology Roadmap for Semiconductors 2003の要求清浄度について － シリコンウエハ表面と雰囲気環境に要求される清浄度, 分析方法の現状について －

Processing-in-memory (PIM) is a promising solution to address the "memory wall" challenges for future computer systems. Prior proposed PIM architectures put additional computation logic in or near memory. The emerging metal-oxide resistive random access memory (ReRAM) has showed its potential to be used for main memory. Moreover, with its crossbar array structure, ReRAM can perform matrix-vector multiplication efficiently, and has been widely studied to accelerate neural network (NN) applications. In this work, we propose a novel PIM architecture, called PRIME, to accelerate NN applications in ReRAM based main memory. In PRIME, a portion of ReRAM crossbar arrays can be configured as accelerators for NN applications or as normal memory for a larger memory space. We provide microarchitecture and circuit designs to enable the morphable functions with an insignificant area overhead. We also design a software/hardware interface for software developers to implement various NNs on PRIME. Benefiting from both the PIM architecture and the efficiency of using ReRAM for NN computation, PRIME distinguishes itself from prior work on NN acceleration, with significant performance improvement and energy saving. Our experimental results show that, compared with a state-of-the-art neural processing unit design, PRIME improves the performance by ~2360× and the energy consumption by ~895×, across the evaluated machine learning benchmarks.

PRIME: a novel processing-in-memory architecture for neural network computation in ReRAM-based main memory

New storage-class memory (SCM) technologies, such as phase-change memory, STT-RAM, and memristors, promise user-level access to non-volatile storage through regular memory instructions. These memory devices enable fast user-mode access to persistence, allowing regular in-memory data structures to survive system crashes.In this paper, we present Mnemosyne, a simple interface for programming with persistent memory. Mnemosyne addresses two challenges: how to create and manage such memory, and how to ensure consistency in the presence of failures. Without additional mechanisms, a system failure may leave data structures in SCM in an invalid state, crashing the program the next time it starts.In Mnemosyne, programmers declare global persistent data with the keyword "pstatic" or allocate it dynamically. Mnemosyne provides primitives for directly modifying persistent variables and supports consistent updates through a lightweight transaction mechanism. Compared to past work on disk-based persistent memory, Mnemosyne reduces latency to storage by writing data directly to memory at the granularity of an update rather than writing memory pages back to disk through the file system. In tests emulating the performance characteristics of forthcoming SCMs, we show that Mnemosyne can persist data as fast as 3 microseconds. Furthermore, it provides a 35 percent performance increase when applied in the OpenLDAP directory server. In microbenchmark studies we find that Mnemosyne can be up to 1400% faster than alternative persistence strategies, such as Berkeley DB or Boost serialization, that are designed for disks.

/pdf/mnemosyne-lightweight-persistent-memory-x1ts5a6ers.pdf

Mnemosyne: lightweight persistent memory

The computing demands of future data-intensive applications will greatly exceed the capabilities of current electronics, and are unlikely to be met by isolated improvements in transistors, data storage technologies or integrated circuit architectures alone. Instead, transformative nanosystems, which use new nanotechnologies to simultaneously realize improved devices and new integrated circuit architectures, are required. Here we present a prototype of such a transformative nanosystem. It consists of more than one million resistive random-access memory cells and more than two million carbon-nanotube field-effect transistors-promising new nanotechnologies for use in energy-efficient digital logic circuits and for dense data storage-fabricated on vertically stacked layers in a single chip. Unlike conventional integrated circuit architectures, the layered fabrication realizes a three-dimensional integrated circuit architecture with fine-grained and dense vertical connectivity between layers of computing, data storage, and input and output (in this instance, sensing). As a result, our nanosystem can capture massive amounts of data every second, store it directly on-chip, perform in situ processing of the captured data, and produce 'highly processed' information. As a working prototype, our nanosystem senses and classifies ambient gases. Furthermore, because the layers are fabricated on top of silicon logic circuitry, our nanosystem is compatible with existing infrastructure for silicon-based technologies. Such complex nano-electronic systems will be essential for future high-performance and highly energy-efficient electronic systems.

Three-dimensional integration of nanotechnologies for computing and data storage on a single chip

Phase-Change Memory (PCM) technology has received substantial attention recently. Because PCM is byte-addressable and exhibits access times in the nanosecond range, it can be used in main memory designs. In fact, PCM has higher density and lower idle power consumption than DRAM. Unfortunately, PCM is also slower than DRAM and has limited endurance. For these reasons, researchers have proposed memory systems that combine a small amount of DRAM and a large amount of PCM. In this paper, we propose a new hybrid design that features a hardware-driven page placement policy. The policy relies on the memory controller (MC) to monitor access patterns, migrate pages between DRAM and PCM, and translate the memory addresses coming from the cores. Periodically, the operating system updates its page mappings based on the translation information used by the MC. Detailed simulations of 27 workloads show that our system is more robust and exhibits lower energy-delay2 than state-of-the-art hybrid systems.

Page placement in hybrid memory systems

An apparatus, system, and method are disclosed for managing eviction of data. A cache write module stores data on a non-volatile storage device sequentially using a log-based storage structure having a head region and a tail region. A direct cache module caches data on the non-volatile storage device using the log-based storage structure. The data is associated with storage operations between a host and a backing store storage device. An eviction module evicts data of at least one region in succession from the log-based storage structure starting with the tail region and progressing toward the head region.

Apparatus, system, and method for managing eviction of data

Power consumption and DRAM latencies are serious concerns in modern chip-multiprocessor (CMP or multi-core) based compute systems. The management of the DRAM row buffer can significantly impact both power consumption and latency. Modern DRAM systems read data from cell arrays and populate a row buffer as large as 8 KB on a memory request. But only a small fraction of these bits are ever returned back to the CPU. This ends up wasting energy and time to read (and subsequently write back) bits which are used rarely. Traditionally, an open-page policy has been used for uni-processor systems and it has worked well because of spatial and temporal locality in the access stream. In future multi-core processors, the possibly independent access streams of each core are interleaved, thus destroying the available locality and significantly under-utilizing the contents of the row buffer. In this work, we attempt to improve row-buffer utilization for future multi-core systems. The schemes presented here are motivated by our observations that a large number of accesses within heavily accessed OS pages are to small, contiguous "chunks" of cache blocks. Thus, the co-location of chunks (from different OS pages) in a row-buffer will improve the overall utilization of the row buffer contents, and consequently reduce memory energy consumption and access time. Such co-location can be achieved in many ways, notably involving a reduction in OS page size and software or hardware assisted migration of data within DRAM. We explore these mechanisms and discuss the trade-offs involved along with energy and performance improvements from each scheme. On average, for applications with room for improvement, our best performing scheme increases performance by 9% (max. 18%) and reduces memory energy consumption by 15% (max. 70%).

/pdf/micro-pages-increasing-dram-efficiency-with-locality-aware-5ohqy1wdbl.pdf

Micro-pages: increasing DRAM efficiency with locality-aware data placement

The IO performance of storage devices has accelerated from hundreds of IOPS five years ago, to hundreds of thousands of IOPS today, and tens of millions of IOPS projected in five years. This sharp evolution is primarily due to the introduction of NAND-flash devices and their data parallel design. In this work, we demonstrate that the block layer within the operating system, originally designed to handle thousands of IOPS, has become a bottleneck to overall storage system performance, specially on the high NUMA-factor processors systems that are becoming commonplace. We describe the design of a next generation block layer that is capable of handling tens of millions of IOPS on a multi-core system equipped with a single storage device. Our experiments show that our design scales graciously with the number of cores, even on NUMA systems with multiple sockets.

Linux block IO: introducing multi-queue SSD access on multi-core systems

Historically, improvements in GPU-based high performance computing have been tightly coupled to transistor scaling. As Moore's law slows down, and the number of transistors per die no longer grows at historical rates, the performance curve of single monolithic GPUs will ultimately plateau. However, the need for higher performing GPUs continues to exist in many domains. To address this need, in this paper we demonstrate that package-level integration of multiple GPU modules to build larger logical GPUs can enable continuous performance scaling beyond Moore's law. Specifically, we propose partitioning GPUs into easily manufacturable basic GPU Modules (GPMs), and integrating them on package using high bandwidth and power efficient signaling technologies. We lay out the details and evaluate the feasibility of a basic Multi-Chip-Module GPU (MCM-GPU) design. We then propose three architectural optimizations that significantly improve GPM data locality and minimize the sensitivity on inter-GPM bandwidth. Our evaluation shows that the optimized MCM-GPU achieves 22.8% speedup and 5x inter-GPM bandwidth reduction when compared to the basic MCM-GPU architecture. Most importantly, the optimized MCM-GPU design is 45.5% faster than the largest implementable monolithic GPU, and performs within 10% of a hypothetical (and unbuildable) monolithic GPU. Lastly we show that our optimized MCM-GPU is 26.8% faster than an equally equipped Multi-GPU system with the same total number of SMs and DRAM bandwidth.

MCM-GPU: Multi-Chip-Module GPUs for Continued Performance Scalability

The method includes receiving an object operation from an application at a hardware device manager. The object operation includes an object identifier. The method includes performing the object operation directly on a storage device. A physical address for the object corresponding to the object identifier is mapped directly to the object identifier in an index managed by the hardware device manager.

David Nellans

Papers

Apparatus, system, and method for managing eviction of data

Micro-pages: increasing DRAM efficiency with locality-aware data placement

Linux block IO: introducing multi-queue SSD access on multi-core systems

MCM-GPU: Multi-Chip-Module GPUs for Continued Performance Scalability

Object-based memory storage