high performance computing on graphics processing units: hgpu.org

Visual-to-Code generation, which transforms scientific plots, vector graphics, and webpages into executable scripts, demands a level of pixel-precise alignment that standard Multimodal Large Language Models (MLLMs) fail to achieve through Supervised Fine-Tuning (SFT) alone. While Reinforcement Learning (RL) offers a theoretical pathway to bridge this gap, its application is hindered by two fundamental obstacles: (1) […]

We report a viability result: an entire Llama-class decoder, written in pure Rust and compiled to PTX by NVIDIA’s own experimental first-party Rust→PTX backend (cuda-oxide), runs FP4-quantized weights on a consumer NVIDIA RTX 5070 Ti and generates coherent English text. On a real TinyLlama-1.1B model quantized to MXFP4, the engine sustains roughly 181 tokens/s of […]

Efficient inter-GPU communication is very important for scalable distributed deep learning, yet the internal behaviour of NVIDIA’s Collective Communication Library (NCCL) at the GPU kernel level remains largely unexplored. Existing profiling tools observe only host-side call boundaries, giving no visibility into the individual send, receive, and reduce steps that constitute each collective operation.This thesis aims […]

In the AI-driven era, NVIDIA CUDA libraries have become indispensable for accelerating compute-intensive tasks, yet their security assessment remains critically understudied due to closed-source code and unique programming paradigms. Existing efforts primarily target CUDA compiler vulnerabilities (e.g., NVCC), but overlook broader library-specific risks. This paper addresses the challenges of fuzzing CUDA libraries: (1) context-dependent API […]

Emerging programming models like Triton enable developers to better exploit modern accelerators, but translating legacy code to Triton remains challenging. While Large Language Models (LLMs) show promise for code translation, they often generate incorrect or suboptimal implementations due to a lack of precise parallelization reasoning. We present TritonPilot, a compiler-assisted LLM framework for translating C […]

Large Language Models (LLMs) show promise for code compilation tasks, but applying them to runtime performance tuning is difficult due to complex microarchitectural effects and noisy runtime measurements. We present AutoPass, a multi-agent framework for compiler performance tuning that uses compiler and runtime evidence to guide LLM-generated optimization decisions. Rather than treating the compiler as […]

Agentic kernel optimization automates manual GPU kernel tuning via iterative generation, validation, and profiling with reasoning LLMs, casting the optimization task as feedback-guided search. However, our workload characterization reveals three system-level inefficiencies that limit search efficiency: (1) long generation latency due to LLM reasoning, (2) insufficient profiling feedback, and (3) underutilized validation/profiling resources. Our key […]

We present KernelPro, a closed-loop multi-agent system that automatically generates, profiles, and iteratively optimizes GPU kernel code by integrating large language model (LLM) code generation with hardware profiler feedback and pluggable bottleneck detection tools. KernelPro introduces four contributions: (1) a semantic feedback operator that encodes expert heuristics as pluggable micro-profiling tools, transforming raw hardware metrics […]

LLM-based coding agents need higher-level operational knowledge about a repository (which files house which subsystems, how to run the test suite, which workflows have historically led to wrong fixes) that does not exist in the code itself. Engineers typically maintain AGENTS.md files to supply this context as instructions for coding agents, but whether they help […]

Benchmarks for LLM-generated GPU kernels (KernelBench, TritonBench, GEAK) score correctness through fixed-shape, small-sample allclose-style checks. The number of inputs varies between benchmarks. The shape, dtype, and tolerance are fixed for each kernel. We test that oracle empirically. We construct a controlled corpus of 24 Triton and CPU stand-in kernels (15 correct controls and 9 LLM-style […]

GPU kernel optimization represents a paradigm where functional correctness is assumed and execution efficiency is the objective. We present daVinci-kernel, a reinforcement learning framework that couples skill discovery with skill exploitation through a dynamically evolving skill library. daVinci-kernel jointly trains three agents sharing one LLM backbone: a Skill Selection Agent that retrieves relevant techniques via […]

High-Performance Computing (HPC) applications increasingly depend on GPUs, yet developing optimized kernels across evolving GPU architectures remains a major productivity bottleneck. With a tile-based programming model, Triton, a Python-based domain-specific language from the AI ecosystem, presents a compelling opportunity to simplify high-performance GPU kernel development for HPC. However, its tight coupling with Python creates significant […]