LEO: Tracing GPU Stall Root Causes via Cross-Vendor Backward Slicing
Fuente:
arXiv
Salvato in:
| Autori principali: | , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915948518178816 |
|---|---|
| author | Xia, Yuning Mellor-Crummey, John |
| author_facet | Xia, Yuning Mellor-Crummey, John |
| contents | More than half of the Top 500 supercomputers employ GPUs as accelerators. On GPU-accelerated platforms, developers face a key diagnostic gap: profilers show source lines where stalls occur, but not why they occur. Furthermore, the same kernel may have different stalls and underlying causes on different GPUs. This paper presents LEO, a root-cause analyzer for NVIDIA, AMD, and Intel GPUs that performs backward slicing from stalled instructions, considering dependencies arising from registers as well as vendor-specific synchronization mechanisms. LEO attributes GPU stalls to source instructions with the goal of explaining root causes of these inefficiencies. Across 21 workloads on three GPU platforms, LEO-guided optimizations deliver geometric-mean speedups of 1.73$\times$--1.82$\times$. Our case studies show that (1) the same kernel may require different optimizations for different GPU architectures, and (2) LEO's structured diagnostics improve code optimization with large language models relative to code-only and raw-stall-count baselines. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_20032 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | LEO: Tracing GPU Stall Root Causes via Cross-Vendor Backward Slicing Xia, Yuning Mellor-Crummey, John Distributed, Parallel, and Cluster Computing Performance More than half of the Top 500 supercomputers employ GPUs as accelerators. On GPU-accelerated platforms, developers face a key diagnostic gap: profilers show source lines where stalls occur, but not why they occur. Furthermore, the same kernel may have different stalls and underlying causes on different GPUs. This paper presents LEO, a root-cause analyzer for NVIDIA, AMD, and Intel GPUs that performs backward slicing from stalled instructions, considering dependencies arising from registers as well as vendor-specific synchronization mechanisms. LEO attributes GPU stalls to source instructions with the goal of explaining root causes of these inefficiencies. Across 21 workloads on three GPU platforms, LEO-guided optimizations deliver geometric-mean speedups of 1.73$\times$--1.82$\times$. Our case studies show that (1) the same kernel may require different optimizations for different GPU architectures, and (2) LEO's structured diagnostics improve code optimization with large language models relative to code-only and raw-stall-count baselines. |
| title | LEO: Tracing GPU Stall Root Causes via Cross-Vendor Backward Slicing |
| topic | Distributed, Parallel, and Cluster Computing Performance |
| url | https://arxiv.org/abs/2604.20032 |