Related papers: Managing Multi Instance GPUs for High Throughput a…

Efficiently harnessing GPU compute is critical to improving user experience and reducing operational costs in large language model (LLM) services. However, current inference engine schedulers overlook the attention backend's sensitivity to…

We present a cross-architecture evaluation of production LLM inference on AMD Instinct MI325X GPUs, benchmarking four models spanning 235B to 1 trillion parameters across three architectural families (MoE+MLA, Dense+GQA, MoE+GQA) on an…

We detail the performance optimizations made in rocHPL, AMD's open-source implementation of the High-Performance Linpack (HPL) benchmark targeting accelerated node architectures designed for exascale systems such as the Frontier…

Collocating deep learning training tasks improves GPU utilization but risks resource contention, severe slowdowns, and out-of-memory (OOM) failures. Accurate memory estimation is essential for robust collocation, and GPU utilization…

Transformers and large language models~(LLMs) have seen rapid adoption in all domains. Their sizes have exploded to hundreds of billions of parameters and keep increasing. Under these circumstances, the training of transformers is very…

Large scale-free graphs are famously difficult to process efficiently: the skewed vertex degree distribution makes it difficult to obtain balanced partitioning. Our research instead aims to turn this into an advantage by partitioning the…

The significant resource demands in LLM serving prompts production clusters to fully utilize heterogeneous hardware by partitioning LLM models across a mix of high-end and low-end GPUs. However, existing parallelization approaches often…

As the increasing complexity of Neural Network(NN) models leads to high demands for computation, AMD introduces a heterogeneous programmable system-on-chip (SoC), i.e., Versal ACAP architectures featured with programmable logic (PL), CPUs,…

General Matrix Multiplication (GEMM) is a fundamental operation in many scientific workloads, signal processing, and particularly deep learning. It is often a bottleneck for performance and energy efficiency, especially in edge environments…

Large Language Models (LLMs) have achieved impressive results across various tasks, yet their high computational demands pose deployment challenges, especially on consumer-grade hardware. Mixture of Experts (MoE) models provide an efficient…

Fine-grained workload and resource balancing is the key to high performance for regular and irregular computations on the GPUs. In this dissertation, we conduct an extensive survey of existing load-balancing techniques to build an…

Modern heterogeneous supercomputing systems are comprised of CPUs, GPUs, and high-speed network interconnects. Communication libraries supporting efficient data transfers involving memory buffers from the GPU memory typically require the…

FPGAs are a promising platform for accelerating Deep Learning (DL) applications, due to their high performance, low power consumption, and reconfigurability. Recently, the leading FPGA vendors have enhanced their architectures to more…

We investigate the energy efficiency of a library designed for parallel computations with sparse matrices. The library leverages high-performance, energy-efficient Graphics Processing Unit (GPU) accelerators to enable large-scale scientific…

In modern GPU inference, cache efficiency remains a major bottleneck, and heuristic policies such as \textsc{LRU} can perform far worse than the offline optimum. Existing learning-based caching systems improve hit rates mainly through…

Process mapping asks to assign vertices of a task graph to processing elements of a supercomputer such that the computational workload is balanced while the communication cost is minimized. Motivated by the recent success of GPU-based graph…

The edge computing paradigm has emerged to handle cloud computing issues such as scalability, security and low response time among others. This new computing trend heavily relies on ubiquitous embedded systems on the edge. Performance and…

Multi-Instance GPU (MIG) is a new feature introduced by NVIDIA A100 GPUs that partitions one physical GPU into multiple GPU instances. With MIG, A100 can be the most cost-efficient GPU ever for serving Deep Neural Networks (DNNs). However,…

While GPUs are responsible for training the vast majority of state-of-the-art deep learning models, the implications of their architecture are often overlooked when designing new deep learning (DL) models. As a consequence, modifying a DL…

To address increasing compute demand from recent multi-model workloads with heavy models like large language models, we propose to deploy heterogeneous chiplet-based multi-chip module (MCM)-based accelerators. We develop an advanced…