New framework uses ML to speed up HPC performance autotuning

Researchers have developed a new framework to accelerate the process of performance autotuning in High Performance Computing (HPC) systems. This framework utilizes a machine learning-based ensemble LLVM Intermediate Representation (IR) ranker called Neural Configuration Scorer (NCS). NCS improves efficiency by ranking the performance of IRs, thereby reducing tuning overhead and avoiding suboptimal evaluations. The system leverages transfer learning to achieve similar performance gains with significantly fewer evaluations compared to existing methods. AI

IMPACT This research could lead to more efficient and faster performance tuning for complex HPC systems, potentially accelerating scientific discovery and computational tasks.

RANK_REASON The cluster describes a research paper detailing a new framework for accelerating autotuning in HPC systems using machine learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• high-performance computing
• LLVM
• Md Arafat Hossain
• Neural Configuration Scorer

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →