CUDA code now runs on Apple Silicon with zero source code changes
petewoodbridge · x · 2026-08-26
Someone has enabled CUDA code to run on Apple Silicon without any source code modifications.
Technical Implementation:
Built a translator chain using GPT-5.6 Sol that rewrites CUDA to Metal step-by-step through existing compilers.
Performance:
- Workload: Heavy 3D fluid simulation (particle sorting, neighbor list building, atomics), not a toy demo.
- Apple GPU: 6 seconds.
- Same code on CPU: 60 seconds.
- Overhead: Zero performance loss. Unlike typical cross-compilation that eats 30–50% throughput, this solution wastes none, with the GPU fully saturated.
More from Infra
- OpenAI shows Doom running on custom inference chip ported via Codex — BenBajarin · 2026-08-26
- Idea: Turning AI Data Centers into Water Parks via Waste Heat — MechanicalDan1 · 2026-08-26
- Leasing a 256GB M5 Ultra Mac Studio now costs barely more than a Claude Max sub — pcuenq · 2026-08-26
- Stagnant US Grid Demand Led to Brain Drain and Lack of Innovation — wandb · 2026-08-26
- Merge Launches Workforce: A Unified Governance Layer for Enterprise AI Access — shensi · 2026-08-26
- Stripe Tests Provisioning API for Agentic App Stores — jeff_weinstein · 2026-08-26