Recent studies from several hyperscalars pinpoint to embedding layers as the most memory-intensive deep learning (DL) algorithm being deployed in today's datacenters. This paper addresses the memory capacity and bandwidth challenges of embedding layers and the associated tensor operations. We present our vertically integrated hardware/software co-design, which includes a custom DIMM module enhanced with near-memory processing cores tailored for DL tensor operations. These custom DIMMs are populated inside a GPU-centric system interconnect as a remote memory pool, allowing GPUs to utilize for scalable memory bandwidth and capacity expansion. A prototype implementation of our proposal on real DL systems shows an average 6.2-17.6× performance improvement on state-of-the-art DNN-based recommender systems.

TensorDIMM: A Practical Near-Memory Processing Architecture for Embeddings and Tensor Operations in Deep Learning

Many modern and emerging applications must process increasingly large volumes of data. Unfortunately, prevalent computing paradigms are not designed to efficiently handle such large-scale data: The energy and performance costs to move this data between the memory subsystem and the CPU now dominate the total costs of computation. This forces system architects and designers to fundamentally rethink how to design computers. Processing-in-memory (PIM) is a computing paradigm that avoids most data movement costs by bringing computation to the data. New opportunities in modern memory systems are enabling architectures that can perform varying degrees of processing inside the memory subsystem. However, many practical system-level issues must be tackled to construct PIM architectures, including enabling workloads and programmers to easily take advantage of PIM. This article examines three key domains of work toward the practical construction and widespread adoption of PIM architectures. First, we describe our work on systematically identifying opportunities for PIM in real applications and quantify potential gains for popular emerging applications (e.g., machine learning, data analytics, genome analysis). Second, we aim to solve several key issues in programming these applications for PIM architectures. Third, we describe challenges that remain for the widespread adoption of PIM.

Processing-in-memory: A workload-driven perspective

Emerging applications such as deep neural network demand high off-chip memory bandwidth. However, under stringent physical constraints of chip packages and system boards, it becomes very expensive to further increase the bandwidth of off-chip memory. Besides, transferring data across the memory hierarchy constitutes a large fraction of total energy consumption of systems, and the fraction has steadily increased with the stagnant technology scaling and poor data reuse characteristics of such emerging applications. To cost-effectively increase the bandwidth and energy efficiency, researchers began to reconsider the past processing-in-memory (PIM) architectures and advance them further, especially exploiting recent integration technologies such as 2.5D/3D stacking. Albeit the recent advances, no major memory manufacturer has developed even a proof-of-concept silicon yet, not to mention a product. This is because the past PIM architectures often require changes in host processors and/or application code which memory manufacturers cannot easily govern. In this paper, elegantly tackling the aforementioned challenges, we propose an innovative yet practical PIM architecture. To demonstrate its practicality and effectiveness at the system level, we implement it with a 20nm DRAM technology, integrate it with an unmodified commercial processor, develop the necessary software stack, and run existing applications without changing their source code. Our evaluation at the system level shows that our PIM improves the performance of memory-bound neural network kernels and applications by 11.2× and 3.5×, respectively. Atop the performance improvement, PIM also reduces the energy per bit transfer by 3.5×, and the overall energy efficiency of the system running the applications by 3.2×.

Hardware Architecture and Software Stack for PIM Based on Commercial DRAM Technology : Industrial Product

Recent studies from several hyperscalars pinpoint to embedding layers as the most memory-intensive deep learning (DL) algorithm being deployed in today's datacenters. This paper addresses the memory capacity and bandwidth challenges of embedding layers and the associated tensor operations. We present our vertically integrated hardware/software co-design, which includes a custom DIMM module enhanced with near-data processing cores tailored for DL tensor operations. These custom DIMMs are populated inside a GPU-centric system interconnect as a remote memory pool, allowing GPUs to utilize for scalable memory bandwidth and capacity expansion. A prototype implementation of our proposal on real DL systems shows an average 6.2-17.6x performance improvement on state-of-the-art recommender systems.

Ongoing climate change calls for fast and accurate weather and climate modeling. However, when solving large-scale weather prediction simulations, state-of-the-art CPU and GPU implementations suffer from limited performance and high energy consumption. These implementations are dominated by complex irregular memory access patterns and low arithmetic intensity that pose fundamental challenges to acceleration. To overcome these challenges, we propose and evaluate the use of near-memory acceleration using a reconfigurable fabric with high-bandwidth memory (HBM). We focus on compound stencils that are fundamental kernels in weather prediction models. By using high-level synthesis techniques, we develop NERO, an FPGA+HBM-based accelerator connected through IBM CAPI2 (Coherent Accelerator Processor Interface) to an IBM POWER9 host system. Our experimental results show that NERO outperforms a 16-core POWER9 system by 4.2x and 8.3x when running two different compound stencil kernels. NERO reduces the energy consumption by 22x and 29x for the same two kernels over the POWER9 system with an energy efficiency of 1.5 GFLOPS/Watt and 17.3 GFLOPS/Watt. We conclude that employing near-memory acceleration solutions for weather prediction modeling is promising as a means to achieve both high performance and high energy efficiency.

NERO: A Near High-Bandwidth Memory Stencil Accelerator for Weather Prediction Modeling

The physical memory capacity of servers is expected to increase drastically with the deployment of the forthcoming non-volatile memory technologies. This is a welcomed improvement for the emerging data-intensive applications. For such servers to be cost-effective, nonetheless, we must cost-effectively increase compute throughput and memory bandwidth commensurate with the increase in memory capacity without compromising the application readiness. Tackling this challenge, we present Memory Channel Network (MCN) architecture in this paper. Specifically, first, we propose an MCN DIMM, an extension of a buffered DIMM where a small but capable processor called MCN processor is integrated with a buffer device on the DIMM for near-memory processing. Second, we implement device drivers to give the host and MCN processors in a server an illusion that they are independent heterogeneous nodes connected through an Ethernet link. These allow the host and MCN processors in a server to run a given data-intensive application together based on popular distributed computing frameworks such as MPI and Spark without any change in the host processor hardware and its application software, while offering the benefits of high-bandwidth and low-latency communication between the host and MCN processors over the memory channels. As such, MCN can serve as an application-transparent framework which can seamlessly unify the near-memory processing within a server and the distributed computing across such servers for data-intensive applications. Our simulation running the full software stack shows that a server with 8 MCN DIMMs offers 4.56 x higher throughput and consume 47.5% less energy than a cluster with 9 conventional nodes connected through Ethernet links, as it facilitates up to 8.17 x higher aggregate DRAM bandwidth utilization. Lastly, we demonstrate the feasibility of MCN with an IBM POWER8 system and an experimental buffered DIMM.

Oliver O'Halloran

Papers

Application-transparent near-memory processing architecture with memory channel network