Related papers: TileLink: Generating Efficient Compute-Communicati…

Deep Learning (DL) training platforms are built by interconnecting multiple DL accelerators (e.g., GPU/TPU) via fast, customized interconnects with 100s of gigabytes (GBs) of bandwidth. However, as we identify in this work, driving this…

Dense linear algebra kernels are critical for wireless applications, and the oncoming proliferation of 5G only amplifies their importance. Many such matrix algorithms are inductive, and exhibit ample amounts of fine-grain ordered…

Mixture-of-Experts-based (MoE-based) diffusion models demonstrate remarkable scalability in high-fidelity image generation, yet their reliance on expert parallelism introduces critical communication bottlenecks. State-of-the-art methods…

AcceleratedKernels.jl is introduced as a backend-agnostic library for parallel computing in Julia, natively targeting NVIDIA, AMD, Intel, and Apple accelerators via a unique transpilation architecture. Written in a unified, compact…

Distributed applications based on micro-services in edge computing are becoming increasingly popular due to the rapid evolution of mobile networks. While Kubernetes is the default framework when it comes to orchestrating and managing…

We show in this work that memory intensive computations can result in severe performance problems due to off-chip memory access and CPU-GPU context switch overheads in a wide range of deep learning models. For this problem, current…

Polyhedral compilers perform optimizations such as tiling and parallelization; when doing both, they usually generate code that executes "barrier-synchronized wavefronts" of tiles. We present a system to express and generate code for hybrid…

Optimizing deep learning models is generally performed in two steps: (i) high-level graph optimizations such as kernel fusion and (ii) low level kernel optimizations such as those found in vendor libraries. This approach often leaves…

Nucleus decompositions have been shown to be a useful tool for finding dense subgraphs. The coreness value of a clique represents its density based on the number of other cliques it is adjacent to. One useful output of nucleus decomposition…

Serving Large Language Models (LLMs) is critical for AI-powered applications, yet it demands substantial computational resources, particularly in memory bandwidth and computational throughput. Low-precision computation has emerged as a key…

Transaction processing systems are the crux for modern data-center applications, yet current multi-node systems are slow due to network overheads. This paper advocates for Compute Express Link (CXL) as a network alternative, which enables…

Distributed data processing systems have advanced through models that expose more and more opportunities for concurrency within a computation. The scheduling of these increasingly sophisticated models has become the bottleneck for improved…

Tree kernels are fundamental tools that have been leveraged in many applications, particularly those based on machine learning for Natural Language Processing tasks. In this paper, we devise a parallel implementation of the sequential…

This paper proposes ScalePool, a novel cluster architecture designed to interconnect numerous accelerators using unified hardware interconnects rather than traditional long-distance networking. ScalePool integrates Accelerator-Centric Links…

Interconnection is crucial for computing systems. However, the current interconnection performance between processors and devices, such as memory devices and accelerators, significantly lags behind their computing performance, severely…

Compilers convert between representations -- usually, from higher-level, human writable code to lower-level, machine-readable code. A compiler backend is the portion of the compiler containing optimizations and code generation routines for…

There are existing standard solvers for tackling discrete optimization problems. However, in practice, it is uncommon to apply them directly to the large input space typical of this class of problems. Rather, the input is preprocessed to…

In the realm of Large Language Model (LLM) inference, the inherent structure of transformer models coupled with the multi-GPU tensor parallelism strategy leads to a sequential execution of computation and communication. This results in…

Large Language Models (LLMs) built on transformer architectures have transformed natural language processing, achieving remarkable performance across diverse applications. While distributed inference frameworks enable practical deployment…

Custom CUDA kernel development is essential for maximizing GPU utilization in large-scale distributed LLM training and inference, yet manually writing kernels that jointly leverage both computation and communication remains a…