相关论文: Garbage Collection for Multicore NUMA Machines

Large inter-GPU all-reduce operations, prevalent throughout deep learning, are bottlenecked by communication costs. Emerging heterogeneous architectures are comprised of complex nodes, often containing $4$ GPUs and dozens to hundreds of CPU…

Even with generational improvements in DRAM technology, memory access latency still remains the major bottleneck for application accelerators, primarily due to limitations in memory interface IPs which cannot fully account for variations in…

Cost of serving large language models (LLM) is high, but the expensive and scarce GPUs are poorly efficient when generating tokens sequentially, unless the batch of sequences is enlarged. However, the batch size is limited by some…

We propose, implement, and experimentally evaluate a runtime middleware to support high-throughput execution on hybrid cluster machines of large-scale analysis applications. A hybrid cluster machine consists of computation nodes which have…

The last improvements in programming languages, programming models, and frameworks have focused on abstracting the users from many programming issues. Among others, recent programming frameworks include simpler syntax, automatic memory…

In this paper, we present a framework for moving compute and data between processing elements in a distributed heterogeneous system. The implementation of the framework is based on the LLVM compiler toolchain combined with the UCX…

As multimodal and AI-driven services exchange hundreds of megabytes per request, existing IPC runtimes spend a growing share of CPU cycles on memory copies. Although both hardware and software mechanisms are exploring memory offloading,…

The emergence of symmetric multi-processing (SMP) systems with non-uniform memory access (NUMA) has prompted extensive research on process and data placement to mitigate the performance impact of NUMA on applications. However, existing…

Hardware heterogeneity is here to stay for high-performance computing. Large-scale systems are currently equipped with multiple GPU accelerators per compute node and are expected to incorporate more specialized hardware in the future. This…

Garbage Collection in concurrent data structures, especially lock-free ones, pose multiple design and consistency challenges. In this instance, we consider the case of concurrent sets. A set is a collection of elements, where the elements…

Limits on power dissipation have pushed CPUs to grow in parallel processing capabilities rather than clock rate, leading to the rise of "manycore" or GPU-like processors. In order to achieve the best performance, applications must be able…

This paper presents an open-source kernel-level heterogeneous memory characterization framework (MemScope) for embedded systems. MemScope enables precise characterization of the temporal behavior of available memory modules under…

We present a new adaptive parallel algorithm for the challenging problem of multi-dimensional numerical integration on massively parallel architectures. Adaptive algorithms have demonstrated the best performance, but efficient many-core…

To satisfy the compute and memory demands of deep neural networks, neural processing units (NPUs) are widely being utilized for accelerating deep learning algorithms. Similar to how GPUs have evolved from a slave device into a mainstream…

Although the Transformer is currently the best-performing architecture in the homogeneous configuration (self-attention only) in Neural Machine Translation, many State-of-the-Art models in Natural Language Processing are made of a…

Large Language Models (LLMs) are increasingly deployed on edge devices with Neural Processing Units (NPUs), yet the decode phase remains memory-intensive, limiting performance. Processing-in-Memory (PIM) offers a promising solution, but…

Modern supercomputers are increasingly requiring the presence of accelerators and co-processors. However, it has not been easy to achieve good performance on such heterogeneous clusters. The key challenge has been to ensure good load…

Modern multi-socket architectures exhibit non-uniform memory access (NUMA) behavior, where access by a core to data cached locally on a socket is much faster than access to data cached on a remote socket. Prior work offers several efficient…

Nowadays, latency-critical, high-performance applications are parallelized even on power-constrained client systems to improve performance. However, an important scenario of fine-grained tasking on simultaneous multithreading CPU cores in…

Scalable persistent memory (PM) has opened up new opportunities for building indexes that operate and persist data directly on the memory bus, potentially enabling instant recovery, low latency and high throughput. When real PM hardware…