Related papers: ML-Triton, A Multi-Level Compilation and Language …

Specializing large language models (LLMs) for local deployment in domain-specific use cases is necessary for strong performance while meeting latency and privacy constraints. However, conventional task-specific adaptation approaches do not…

The rapid adoption of Large Language Models (LLMs) has made GPU inference efficiency an increasingly critical system concern. The runtime of LLM workloads is largely dominated by tile-based kernels, particularly General Matrix…

Driven by the increasing demand for low-latency and real-time processing, machine learning applications are steadily migrating toward edge computing platforms, where Field-Programmable Gate Arrays (FPGAs) are widely adopted for their energy…

LLM-guided compiler optimization has recently shown promise, but existing approaches rely on a single large LLM throughout search, making them expensive and excluding smaller models. We pose the research question: whether heterogeneous LLMs…

Scaling long-context ability is essential for Large Language Models (LLMs). To amortize the memory consumption across multiple devices in long-context training, inter-data partitioning (a.k.a. Data Parallelism) and intra-data partitioning…

Concurrent computation and communication (C3) is a pervasive paradigm in ML and other domains, making its performance optimization crucial. In this paper, we carefully characterize C3 in ML on GPUs, which are most widely deployed for ML…

Multimodal large language model (MLLM) inference splits into two phases with opposing hardware demands: vision encoding is compute-bound, while language generation is memory-bandwidth-bound. We show that under standard transformer KV…

The rapid growth of deep learning has driven exponential increases in model parameters and computational demands. NVIDIA GPUs and their CUDA-based software ecosystem provide robust support for parallel computing, significantly alleviating…

This report presents some early results on code generation targeting tensor cores on NVIDIA GPUs using the MLIR compiler infrastructure. The state-of-the-art in high-performance deep learning today is primarily driven by manually optimized…

Modern high-performance computing architectures (Multicore, GPU, Manycore) are based on tightly-coupled clusters of processing elements, physically implemented as rectangular tiles. Their size and aspect ratio strongly impact the achievable…

Fluid-flow devices with low dissipation, but high contact area, are of importance in many applications. A well-known strategy to design such devices is multi-scale topology optimization (MTO), where optimal microstructures are designed…

The billion-scale Large Language Models (LLMs) need deployment on expensive server-grade GPUs with large-storage HBMs and abundant computation capability. As LLM-assisted services become popular, achieving cost-effective LLM inference on…

Modern LLM serving now spans multi-stage pipelines including RAG retrieval and KV cache reuse, each with distinct compute, memory, and latency demands. Inference engines expose a large configuration space with no systematic navigation…

The auto-regressive decoding of Large Language Models (LLMs) results in significant overheads in their hardware performance. While recent research has investigated various speculative decoding techniques for multi-token generation, these…

High-Level Synthesis (HLS) enables rapid prototyping of complex hardware designs by translating C or C++ code to low-level RTL code. However, the testing and evaluation of HLS designs still typically rely on slow RTL-level simulators that…

We present a single-node, multi-GPU programmable graph processing library that allows programmers to easily extend single-GPU graph algorithms to achieve scalable performance on large graphs with billions of edges. Directly using the…

Various layer-skipping methods have been proposed to accelerate token generation in large language models (LLMs). However, limited attention has been paid to a fundamental question: How do computational demands vary across the generation of…

A tridiagonal matrix algorithm (TDMA), Pipelined-TDMA, is developed for multi-GPU systems to resolve the scalability bottlenecks caused by the sequential structure of conventional divide-and-conquer TDMA. The proposed method pipelines…

Efficient inference of LLMs remains a crucial challenge, with two main phases: a compute-intensive prompt computation and a memory-intensive token generation. Despite existing batching and scheduling techniques, token generation phases fail…

Graphics processors, or GPUs, have recently been widely used as accelerators in the shared environments such as clusters and clouds. In such shared environments, many kernels are submitted to GPUs from different users, and throughput is an…