Related papers: A Case Study in CUDA Kernel Fusion: Implementing F…

Large Language Models (LLMs) encounter severe memory inefficiencies during long-context inference due to conventional handling of key-value (KV) caches. In this work, we introduce a novel integration of PagedAttention with PyTorch's…

Recently, numerous efficient Transformers have been proposed to reduce the quadratic computational complexity of standard Transformers caused by the Softmax attention. However, most of them simply swap Softmax with an efficient attention…

We develop hybrid memory architectures for general-purpose sequence processing neural networks, that combine key-value memory using softmax attention (KV-memory) with fast weight memory through dynamic synaptic modulation (FW-memory) -- the…

Existing attention accelerators often trade exact softmax semantics, depend on fused Tensor Core kernels, or incur sequential depth that limits FP32 throughput on long sequences. We present \textbf{ELSA}, an algorithmic reformulation of…

Transformer-based models have emerged as one of the most widely used architectures for natural language processing, natural language generation, and image generation. The size of the state-of-the-art models has increased steadily reaching…

The efficiency of attention is critical because its time complexity grows quadratically with sequence length. SageAttention2 addresses this by utilizing quantization to accelerate matrix multiplications (Matmul) in attention. To further…

Processing 3D data efficiently has always been a challenge. Spatial operations on large-scale point clouds, stored as sparse data, require extra cost. Attracted by the success of transformers, researchers are using multi-head attention for…

All current LLM serving systems place the GPU at the center, from production-level attention-FFN disaggregation to NVIDIA's Rubin GPU-LPU heterogeneous platform. Even academic PIM/PNM proposals still treat the GPU as the central hub for…

An existing hybrid MPI-OpenMP scheme is augmented with a CUDA-based fine grain parallelization approach for multidimensional distributed Fourier transforms, in a well-characterized pseudospectral fluid turbulence code. Basics of the hybrid…

Window-based transformers excel in large-scale point cloud understanding by capturing context-aware representations with affordable attention computation in a more localized manner. However, the sparse nature of point clouds leads to a…

Each request in LLM inference goes through two phases: compute-bound prefill and memory-bandwidth-bound decode. To improve GPU utilization, recent systems use hybrid batching that combines the prefill and decode phases of different requests…

Matrix-accelerated stencil computation is a hot research topic, yet its application to three-dimensional (3D) high-order stencils and HPC remains underexplored. With the emergence of matrix units on multicore CPUs, we analyze matrix-based…

Scaling inference for large language models (LLMs) is increasingly constrained by limited GPU memory, especially due to growing key-value (KV) caches required for long-context generation. While existing approaches offload KV caches to CPU…

The Transformer architecture has revolutionized deep learning, delivering the state-of-the-art performance in areas such as natural language processing, computer vision, and time series prediction. However, its core component,…

In this paper, we propose CUDA-L2, a system that combines large language models (LLMs) and reinforcement learning (RL) to automatically optimize Half-precision General Matrix Multiply (HGEMM) CUDA kernels. Using CUDA execution speed as the…

Modern GPUs are able to perform significantly more arithmetic operations than transfers of a single word to or from global memory. Hence, many GPU kernels are limited by memory bandwidth and cannot exploit the arithmetic power of GPUs.…

Neural architectures and hardware accelerators have been two driving forces for the progress in deep learning. Previous works typically attempt to optimize hardware given a fixed model architecture or model architecture given fixed…

Recent efforts recognize the power of scale in 3D learning (e.g. PTv3) and attention mechanisms (e.g. FlashAttention). However, current point cloud backbones fail to holistically unify geometric locality, attention mechanisms, and GPU…

Late-interaction retrieval (ColBERT, ColPali) scores a query against a document with the MaxSim operator: for every query token, the maximum similarity over the document tokens, summed over query tokens. The standard implementation…

In this paper we focus on the integration of high-performance numerical libraries in ab initio codes and the portability of performance and scalability. The target of our work is FLEUR, a software for electronic structure calculations…