Related papers: Evaluating CUDA Tile for AI Workloads on Hopper an…

While graph-based dynamic programming (DP) is a cornerstone of genomics and network analytics, its efficiency is hampered by fundamentally conflicting computational patterns. Matrix-centric DP drives regular, compute-bound network…

The proliferation of large language models (LLMs) is accelerating the integration of multimodal assistants into edge devices, where inference is executed under stringent latency and energy constraints, often exacerbated by intermittent…

Modern cloud platforms increasingly host large-scale deep learning (DL) workloads, demanding high-throughput, low-latency GPU scheduling. However, the growing heterogeneity of GPU clusters and limited visibility into application…

CUDA (formerly an abbreviation of Compute Unified Device Architecture) is a parallel computing platform and API model created by Nvidia allowing software developers to use a CUDA-enabled graphics processing unit (GPU) for general purpose…

Large language models (LLMs) require substantial computational resources, leading to significant carbon emissions and operational costs. Although training is energy-intensive, the long-term environmental burden arises from inference,…

Neural networks (NNs) are growing in importance and complexity. A neural network's performance (and energy efficiency) can be bound either by computation or memory resources. The processing-in-memory (PIM) paradigm, where computation is…

Experience shows that on today's high performance systems the utilization of different acceleration cards in conjunction with a high utilization of all other parts of the system is difficult. Future architectures, like exascale clusters,…

General matrix multiplication (GEMM) is the computational backbone of modern AI workloads, and its efficiency is critically dependent on effective tiling strategies. Conventional approaches employ symmetric tile buffering, where the…

We present VitaLLM, a mixed precision accelerator that enables ternary weight large language models to run efficiently on edge devices. The design combines two compute cores, a multiplier free TINT core for ternary-INT projections and a…

Porting deep learning algorithms to new hardware accelerators requires developers to repeatedly apply the same low-level optimizations -- quantization, memory access coalescing, tile size tuning, and architecture-specific workarounds -- to…

Efficient GPU programming is crucial for achieving high performance in deep learning (DL) applications. The performance of GPU programs depends on how data is parallelized across threads and arranged within memory subsystems. The mapping…

Fine-tuning Large Language Models (LLMs) on consumer-grade GPUs is highly cost-effective, yet constrained by limited GPU memory and slow PCIe interconnects. Pipeline parallelism combined with CPU offloading mitigates these hardware…

We present tritonBLAS, a fast and deterministic analytical model that uses architectural parameters like the cache hierarchy, and relative code and data placement to generate performant GPU GEMM kernels. tritonBLAS explicitly models the…

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…

This paper focuses on evaluating the performance impact of different data layouts on the GPU-accelerated IDW interpolation. First, we redesign and improve our previous GPU implementation that was performed by exploiting the feature CUDA…

Finite element schemes based on discontinuous Galerkin methods possess features amenable to massively parallel computing accelerated with general purpose graphics processing units (GPUs). However, the computational performance of such…

Neural network training entails heavy computation with obvious bottlenecks. The Compute Unified Device Architecture (CUDA) programming model allows us to accelerate computation by passing the processing workload from the CPU to the graphics…

Unstructured meshes present challenges in scientific data analysis due to irregular distribution and complex connectivity. Computing and storing connectivity information is a major bottleneck for visualization algorithms, affecting both…

Human texture perception is a weighted average of multi-sensory inputs: visual and tactile. While the visual sensing mechanism extracts global features, the tactile mechanism complements it by extracting local features. The lack of coupled…

Kepler GTX Titan Black and Kepler Tesla K40 are still the best GPUs for high performance computing, although Maxwell GPUs such as GTX 980 are available in the market. Hence, we measure the performance of our lattice QCD codes using the…