Related papers: Mechanism to Mitigate AVX-Induced Frequency Reduct…

Structured sparsity has been proposed as an efficient way to prune the complexity of modern Machine Learning (ML) applications and to simplify the handling of sparse data in hardware. The acceleration of ML models - for both training and…

Intel SGX is known to be vulnerable to a class of practical attacks exploiting memory access pattern side-channels, notably page-fault attacks and cache timing attacks. A promising hardening scheme is to wrap applications in hardware…

This article describes the ARM Scalable Vector Extension (SVE). Several goals guided the design of the architecture. First was the need to extend the vector processing capability associated with the ARM AArch64 execution state to better…

Transformers are central in modern natural language processing and computer vision applications. Despite recent works devoted to reducing the quadratic cost of such models (as a function of the sequence length), dealing with ultra long…

Graphics Processing Units (GPUs) support dynamic voltage and frequency scaling (DVFS) in order to balance computational performance and energy consumption. However, there still lacks simple and accurate performance estimation of a given GPU…

Single Instruction, Multiple Data (SIMD) vectorization is a major driver of performance in current architectures, and is mandatory for achieving good performance with codes that are limited by instruction throughput. We investigate the…

Xilinx's AI Engine is a recent industry example of energy-efficient vector processing that includes novel support for 2D SIMD datapaths and shuffle interconnection network. The current approach to programming the AI Engine relies on a C/C++…

Many virtual machines exist for sensor nodes with only a few KB RAM and tens to a few hundred KB flash memory. They pack an impressive set of features, but suffer from a slowdown of one to two orders of magnitude compared to optimised…

Modern GPUs increasingly rely on specialized and asynchronous hardware units to deliver high performance. Yet these units are often underutilized because today's GPU software stacks still organize programming and execution around a…

In High Performance Computing, systems are evaluated based on their computational throughput. However, performance in contemporary server processors is primarily limited by power and thermal constraints. Ensuring operation within a given…

Modern processors dynamically control their operating frequency to optimize resource utilization, maximize energy savings, and conform to system-defined constraints. If, during the execution of a software workload, the running average of…

Structured sparsity has been proposed as an efficient way to prune the complexity of Machine Learning (ML) applications and to simplify the handling of sparse data in hardware. Accelerating ML models, whether for training, or inference,…

The technologies of heterogeneous multi-core architectures, co-location, and virtualization can be used to reduce server power consumption and improve system utilization, which are three important technologies for data centers. This article…

Modern processors have instructions to process 16 bytes or more at once. These instructions are called SIMD, for single instruction, multiple data. Recent advances have leveraged SIMD instructions to accelerate parsing of common Internet…

Vector processing has become commonplace in today's CPU microarchitectures. Vector instructions improve performance and energy which is crucial for resource-constraint mobile devices. The research community currently lacks a comprehensive…

In the rapidly evolving domain of high-performance computing (HPC), heterogeneous architectures such as the SX-Aurora TSUBASA (SX-AT) system architecture, which integrate diverse processor types, present both opportunities and challenges…

Most modern processors contain vector units that simultaneously perform the same arithmetic operation over multiple sets of operands. The ability of compilers to automatically vectorize code is critical to effectively using these units.…

Large-scale machine learning workloads increasingly rely on multi-GPU systems, yet their performance is often limited by an overlooked component: the CPU. Through a detailed study of modern large language model (LLM) inference and serving…

FlashAttention-style online softmax enables exact attention computation with linear memory by streaming score tiles through on-chip memory and maintaining a running maximum and normalizer. However, as attention kernels approach peak…

This paper summarizes our work on experimental characterization and analysis of reduced-voltage operation in modern DRAM chips, which was published in SIGMETRICS 2017, and examines the work's significance and future potential. We take a…