In the news
Scaling Decision Optimization to 100 Million Variables and Beyond with mPDLP in NVIDIA cuOpt
NVIDIA Developer · Published · 3 min read
In 30 seconds
- What happened
- NVIDIA cuOpt released mPDLP, a multi-GPU linear programming solver distributing problems across NVLink-connected GPUs, handling up to 2.1 billion nonzeros.
- Why it matters
- Supply chain planners and energy modelers solving large optimization problems that exceed single-GPU memory or time constraints need this capability.
- Watch out
- Performance gains depend heavily on matrix sparsity structure and edge-cut ratios; communication overhead can limit speedup on smaller problems below ten million nonzeros.
- eval
The patterns behind this
Each one covers how the technique works, when it earns its cost, and where it breaks.
The Agent Architect
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.