CUDA 'sticky errors' and undefined behavior explained
blelbach · x · 2026-08-29
Explaining CUDA errors that bypass Python try-catch blocks (like illegal memory access), a developer clarified the terminology and mechanics:
- Sticky errors: The term for hard-to-contain failures.
- Cause: In modern architectures with unified memory, device-side undefined behavior can corrupt host-side state (e.g., the CUDA context), making them uncatchable at the Python layer.
More from Infra
- Magnitude open-source inference engine beats Ollama by over 2x with auto-tuned kernels — nickbaumann_ · 2026-10-15
- Nvidia to announce $28B in AI investments in Europe this week, report says — georgemillo · 2026-10-15
- A $10,811 Cloudflare Bill From an Unused AI-Written Test Project: Refunds Aren't a Safety Net — sujingshen · 2026-10-15
- Retraining-Free Method Turns Transformers Into Elastic Models Adapting to Any Compute Budget — VISLab-Amsterdam · 2026-10-15
- vLLM's MTP3 validation: AIME25 up to 0.9667, prefill kernels up to 84% lower latency — vllm_project · 2026-10-15
- Open-weights s1-llm-auto-router picks the right LLM in 26ms on a CPU — airesearch12 · 2026-10-15