@hgpu profile picture
HGPU group @hgpu

High performance computing on graphics processing units (GPU): AMD/ATI, nVidia, Intel Xeon Phi, CUDA, OpenCL, OpenGL, GPGPU, HPC

Joined May 2011
  • Tweets 10,685
  • Following 119
  • Followers 3,852
10 Photos and videos
Towards Feedback-to-Plan Decisions for Self-Evolving LLM Agents in CUDA Kernel Generation #CUDA #PTX #LLM #Package hgpu.org/?p=30834

Towards Feedback-to-Plan Decisions for Self-Evolving LLM Agents in CUDA Kernel Generation

Large language models (LLMs) have shown strong empirical gains as self-evolving agents for CUDA kernel generation, driven by feedback-conditioned planning across generations. However, how planning …

hgpu.org
2
11
788
CodegenBench: Can LLMs Write Efficient Code Across Architectures? #CUDA #LLM #HPC #Package hgpu.org/?p=30833

CodegenBench: Can LLMs Write Efficient Code Across Architectures?

While large language models (LLMs) have been extensively evaluated on code generation tasks for general-purpose programming and GPU-accelerated environments (e.g., PyTorch, CUDA), their capabilitie…

hgpu.org
1
10
1,151
KForge: LLM-Driven Cross-Platform Kernel Generation for AI Accelerators #CUDA #PTX #Triton #LLM #CodeGeneration #Intel hgpu.org/?p=30832

KForge: LLM-Driven Cross-Platform Kernel Generation for AI Accelerators

Production inference increasingly targets a heterogeneous mix of accelerators. Agentic pipelines interleave reasoning, tool calls, and multi-agent coordination, each with distinct compute and memor…

hgpu.org
2
13
732
MusaCoder: Native GPU Kernel Generation with Full-Stack Training on Moore Threads GPU #CUDA #LLM hgpu.org/?p=30831

MusaCoder: Native GPU Kernel Generation with Full-Stack Training on Moore Threads GPU

Native GPU kernel generation turns high-level tensor programs into executable, efficient low-level code. Existing Large Language Models (LLMs) struggle with this task, while execution-based reinfor…

hgpu.org
5
22
2,117
Autonomous heterogeneous catalyst discovery with a self-evolving multi-agent digital twin #Chemistry #LLM hgpu.org/?p=30830

Autonomous heterogeneous catalyst discovery with a self-evolving multi-agent digital twin

Theoretical heterogeneous catalysis promises rapid catalyst discovery, yet computational and machine-learning predictions often deviate from experiment and stay confined to narrow material families…

hgpu.org
1
1
215
Do you remember when you joined X? I do! #MyXAnniversary
1
157
Pretraining large language models with MXFP4 on Native FP4 Hardware #ROCm #LLM #MXFP4 #FP4 hgpu.org/?p=30811

Pretraining large language models with MXFP4 on Native FP4 Hardware

Why does full-pipeline FP4 training of large language models often diverge, even when forward activations and activation gradients remain stable? We address this question through a controlled study…

hgpu.org
6
313
Analyzing the Impact of Kernel Fusion on GPU Tensor Operation Performance: A Systematic Performance Study #CUDA #Package hgpu.org/?p=30810

Analyzing the Impact of Kernel Fusion on GPU Tensor Operation Performance: A Systematic Performance...

Large numbers of small tensor kernels are executed by GPUs in modern deep learning frameworks, where total performance is frequently constrained by memory bandwidth and kernel launch overheads. Sys…

hgpu.org
10
22
1,667
CUDABeaver: Benchmarking LLM-Based Automated CUDA Debugging #CUDA #LLM #Benchmarking hgpu.org/?p=30809

CUDABeaver: Benchmarking LLM-Based Automated CUDA Debugging

Debugging CUDA programs has long been challenging because failures often arise from subtle interactions among hardware behavior, compiler decisions, memory hierarchy, and asynchronous execution. Mo…

hgpu.org
2
3
234
Source-to-Source Transformations for GPU Code Generation #CUDA #CodeGeneration hgpu.org/?p=30808

Source-to-Source Transformations for GPU Code Generation

GPUs have become essential in modern high performance computing, but programming them correctly remains a significant challenge. This difficulty arises from subtle concurrency bugs that result from…

hgpu.org
10
752
KernelBenchX: A Comprehensive Benchmark for Evaluating LLM-Generated GPU Kernels #LLM #Benchmarking #Package hgpu.org/?p=30812

KernelBenchX: A Comprehensive Benchmark for Evaluating LLM-Generated GPU Kernels

LLM-based Triton kernel generation has attracted significant interest, yet a fundamental empirical question remains unanswered: where does this capability break down, and why? We present KernelBenc…

hgpu.org
3
15
1,222
CUDAHercules: Benchmarking Hardware-Aware Expert-level CUDA Optimization for LLMs #CUDA #LLM #Benchmarking hgpu.org/?p=30807

CUDAHercules: Benchmarking Hardware-Aware Expert-level CUDA Optimization for LLMs

Large language models show promise for automated CUDA programming, however even the strongest coding models (e.g., Claude-Opus-4.6) may still fall short of expert-level, architecture-aware optimiza…

hgpu.org
2
11
857
DITRON: Distributed Multi-level Tiling Compiler for Parallel Tensor Programs #Triton #LLM #Package hgpu.org/?p=30797

DITRON: Distributed Multi-level Tiling Compiler for Parallel Tensor Programs

The scaling of large language models (LLMs) is currently bottlenecked by the rigidity of distributed programming. While high-performance libraries like CuBLAS and NCCL provide optimized primitives,…

hgpu.org
3
12
936
Microbenchmark-Driven Analytical Performance Modeling Across Modern GPU Architectures #CUDA #AMD #Benchmarking hgpu.org/?p=30796

Microbenchmark-Driven Analytical Performance Modeling Across Modern GPU Architectures

Rapidly evolving GPU architectures featuring complex memory hierarchies, matrix units, and varied precision formats continue to widen the gap between theoretical peaks and achievable performance. W…

hgpu.org
1
7
471
CuBridge: An LLM-Based Framework for Understanding and Reconstructing High-Performance Attention Kernels #CUDA #LLM hgpu.org/?p=30795

CuBridge: An LLM-Based Framework for Understanding and Reconstructing High-Performance Attention...

Efficient CUDA implementations of attention mechanisms are critical to modern deep learning systems, yet supporting diverse and evolving attention variants remains challenging. Existing frameworks …

hgpu.org
3
13
844
KEET: Explaining Performance of GPU Kernels Using LLM Agents #CUDA #LLM #Performance hgpu.org/?p=30794

KEET: Explaining Performance of GPU Kernels Using LLM Agents

Performance profiles of GPU kernels generated by tools such as Nsight Compute are rich in detail but are often challenging to interpret. To achieve the best performance possible on a given GPU arch…

hgpu.org
5
18
1,318
Kerncap: Automated Kernel Extraction and Isolation for AMD GPUs #Triton #AMD #LLM #Package hgpu.org/?p=30793

Kerncap: Automated Kernel Extraction and Isolation for AMD GPUs

Iterative GPU kernel tuning is bottlenecked by the scale of the applications that host the kernels. Rapid iteration requires isolating the kernel so it can be edited, recompiled, and validated with…

hgpu.org
3
262
Evaluating CUDA Tile for AI Workloads on Hopper and Blackwell GPUs #Triton #CUDA #CUBLAS #LLM #Performance #Package hgpu.org/?p=30763

Evaluating CUDA Tile for AI Workloads on Hopper and Blackwell GPUs

NVIDIA’s CUDA Tile (CuTile) introduces a Python-based, tile-centric abstraction for GPU kernel development that aims to simplify programming while retaining Tensor Core and Tensor Memory Acce…

hgpu.org
1
3
8
373
ARGUS: Agentic GPU Optimization Guided by Data-Flow Invariants #Triton #AMD #LLM #DSL hgpu.org/?p=30764

ARGUS: Agentic GPU Optimization Guided by Data-Flow Invariants

LLM-based coding agents can generate functionally correct GPU kernels, yet their performance remains far below hand-optimized libraries on critical computations such as matrix multiplication, atten…

hgpu.org
1
3
16
1,123
A Human–Machine Collaborative Tuning Framework for Triton Kernel Optimization on SIMD Platforms #Triton #AutoTuning #EvolutionaryComputations hgpu.org/?p=30762

A Human–Machine Collaborative Tuning Framework for Triton Kernel Optimization on SIMD Platforms

Single Instruction, Multiple Data (SIMD) technology enhances performance through parallel data processing on CPUs. SIMD platforms are widely adopted across domains ranging from high-performance com…

hgpu.org
1
4
272
Load more