--- title: "Optimization" description: "13 posts about Optimization, summarised, each linking to the original." --- # Optimization > 13 posts about Optimization, summarised, each linking to the original. ## Articles ### [Scaling Decision Optimization to 100 Million Variables and Beyond with mPDLP in NVIDIA cuOpt](https://yomu.fyi/post/scaling-decision-optimization-to-100-million-variables-and-beyond-with.md) - Company: [NVIDIA](https://yomu.fyi/company/nvidia-developer-blog.md) - Author: Tanya Lenz - Published: Oct 7, 2026 - Type: Announcement Supply chain and energy grid planning problems now exceed the memory and time limits of single-GPU solvers. NVIDIA cuOpt introduces mPDLP, a multi-GPU primal-dual hybrid gradient solver that distributes linear programming problems across NVLink-connected GPUs. The algorithm’s hot loop is a sparse matrix-vector multiplication, a memory-bound operation that benefits from the increased bandwidth of multiple GPUs. Min-cut partitioning reduces communication overhead by keeping tightly connected dependencies on the same GPU, enabling up to 6x lower peak memory per GPU. On the benchmark problem zib03 with over 104 million nonzeros, mPDLP achieved a solve rate nearly 10x faster than the previous year, and partners Kinaxis and PSR demonstrated 3.3x and 5x speedups on real-world models. ### [Faster Rates for Federated Variational Inequalities](https://yomu.fyi/post/faster-rates-for-federated-variational-inequalities.md) - Company: [Apple](https://yomu.fyi/company/apple.md) - Published: Sep 28, 2026 - Type: Problem & solution This paper studies federated optimization for solving stochastic variational inequalities (VIs), a field with growing attention but a significant performance gap compared to federated convex optimization. The authors address this gap by establishing a series of improved convergence rates. They first refine the analysis of the classical Local Extra SGD algorithm, providing tighter guarantees for smooth and monotone VIs. However, they identify an inherent limitation of Local Extra SGD: excessive client drift. To overcome this, they propose the Local Inexact Proximal Point Algorithm with Extra Step (LIPPAX), which mitigates drift and achieves better rates under bounded Hessian, bounded operator, and low-variance regimes. The work also extends these improved guarantees to federated composite variational inequalities. ### [Open-Sourcing Rebalancer: A Generic, High-Performance Library for Solving Assignment Problems](https://yomu.fyi/post/open-sourcing-rebalancer-a-generic-high-performance-library-for-solvin.md) - Company: [Meta](https://yomu.fyi/company/meta.md) - Author: By Richard Barnes, Neeraj Kumar, Pol Mauri Ruiz - Published: Sep 21, 2026 - Type: Announcement Meta has open-sourced Rebalancer, a generic library for solving assignment problems that has been used internally for over nine years. Rebalancer separates problem specification from solution, using a three-step approach with modeling constructs like dimensions, partitions, scopes, and utilization, and a high-level spec API for common objectives and constraints. Problems are translated into an expression graph, and Rebalancer offers two solving techniques: an optimal solver using mixed integer programming (MIP) with commercial or open-source solvers, and a local search solver that evaluates millions of moves per second for large-scale problems. The local search solver handles problems with over a million objects and 5k bins, with an average solve time of 171 seconds. Rebalancer is used for roughly 40 million assignment problems daily across Meta’s infrastructure, and the open-source release includes a debugging UI called Rebalancer Explorer. The accompanying paper was published at OSDI’24. ### [Pruning LLMs Like a Physicist: Block Removal as an Ising Optimization Problem](https://yomu.fyi/post/pruning-llms-like-a-physicist-block-removal-as-an-ising-optimization-p.md) - Company: [Hugging Face](https://yomu.fyi/company/hugging-face.md) - Author: Antonio Tiene, Ali Hashemi, David Jansen, Roman Rausch - Published: Sep 21, 2026 - Type: Problem & solution Removing entire transformer blocks is a cheap way to speed up large language models, yielding predictable inference gains and stacking with quantization and other compression methods. The challenge is that block choices interact, making it a combinatorial problem rather than a simple ranking task. The authors reformulate block selection as a constrained binary optimization (CBO) problem, mapping it onto an Ising glass, a disordered spin system with all-to-all interactions and a fixed number of removed blocks. They note that mean-field methods, which treat each block independently, are inadequate for this interacting system. Block removal composes with quantization, low-rank compression, width pruning, and knowledge-distillation-based healing, fitting into a larger pipeline rather than competing with it. The paper is titled ‘LLM Compression by Block Removal with Constrained Binary Optimization.’ ### [Saving another 100TB of RAM with math (and Rust)](https://yomu.fyi/post/saving-another-100tb-of-ram-with-math-and-rust.md) - Company: [Cloudflare](https://yomu.fyi/company/cloudflare.md) - Author: Ivan Babrou - Published: Sep 18, 2026 - Type: Explainer Cloudflare describes how excessive memory use in pingora-ketama, the consistent-hashing component of its Pingora Backend Router, was investigated as a resource-efficiency problem at global scale. Consistent hashing maps servers and cache keys onto a circular hash space, but a single hash per server creates uneven ranges, while multiple hashes improve balance; with 160 points per server, the coefficient of variation falls to about 8% in the 100-server example. The work produced a v2 ring with compacted storage, faster sorting, and a configurable base number of hashes per node, while retaining the existing v1 ring and supporting simultaneous use during migration. After the old ring was decommissioned, measured memory usage for PBR fell by 100TB globally, and the changes are available in pingora-ketama behind an unadvertised Cargo feature. ### [Dense vs. MoE Models: Active Parameters, Throughput, and When to Choose Each](https://yomu.fyi/post/dense-vs-moe-models-active-parameters-throughput-and-when-to-choose-ea.md) - Company: [NVIDIA](https://yomu.fyi/company/nvidia-developer-blog.md) - Author: Elizabeth Goodman - Published: Sep 15, 2026 - Type: Explainer Dense models activate every parameter for every token, while Mixture-of-Experts (MoE) models route each token through a selected subset of feed-forward experts. This changes the relationship between capacity, memory, and compute: total parameters determine the memory required to load a model, whereas active parameters largely determine per-token work. At equal total size, MoE models often provide higher throughput, especially at batch size one, but routing, simultaneous expert residency, and reduced KV-cache room add serving complexity; latency advantages narrow at high concurrency. The comparison uses Nemotron 3.5 Lightning, a roughly 30B-parameter Mamba-2 plus MoE plus Attention hybrid with 3B active parameters, alongside dense and MoE examples, and concludes that dense models suit simpler serving and fine-tuning while MoE models suit throughput-oriented deployments with manageable memory and operational complexity. ### [How NVIDIA Groq 3 LPX Deterministic Execution Drives Power-Efficient High-Interactivity Inference on NVIDIA Vera Rubin](https://yomu.fyi/post/how-nvidia-groq-3-lpx-deterministic-execution-drives-power-efficient-h.md) - Company: [NVIDIA](https://yomu.fyi/company/nvidia-developer-blog.md) - Author: Tanya Lenz - Published: Sep 15, 2026 - Type: Explainer AI factories are constrained by site power, making performance per watt especially important for high-interactivity inference. The NVIDIA Groq 3 LPX rack uses deterministic execution: its compiler schedules computation, data movement, and clock-cycle timing across 256 LPU chips, then derives a cycle-level current-demand curve. Preemptive Power (PEP) prepares the power-delivery network for upcoming spikes, while Clock Period Synthesis (CPS) lengthens or shortens selected cycles to reduce di/dt, voltage droop, and the voltage guardband. Internal testing reported more than 60% less voltage drop, while the source estimates a high-single-digit baseline-voltage reduction and potentially low-double-digit workload-power reduction; paired with Vera Rubin NVL72, the platform is described as delivering up to 35x higher throughput per megawatt than GB200 NVL72 for 2T+ parameter models at long context and high interactivity. ### [To Infinity and Beyond: ThunderKittens Now on NVIDIA Vera Rubin NVL72!](https://yomu.fyi/post/to-infinity-and-beyond-thunderkittens-now-on-nvidia-vera-rubin-nvl72.md) - Company: [Together AI](https://yomu.fyi/company/together-ai.md) - Published: Sep 10, 2026 - Type: Explainer Together’s kernels team ported ThunderKittens to NVIDIA Vera Rubin NVL72 and investigated why existing Blackwell GEMMs underuse the platform. The work targets NVFP4 and FP8 GEMMs: although Rubin preserves Blackwell’s programming model, the original kernels reach only about 42.1% and 44.4% of roofline because tensor cores consume operands twice as fast as the old pipeline supplies them. The team integrates Rubin’s 64-byte K step, exclusive tensor memory extending to 576 columns, oversized shared memory up to 328 KiB, B-side collector buffers, and earlier A-release signaling into its NVFP4 kernel. The resulting kernel exceeds 22 PFLOPS and is described as competitive with cuBLAS and CuTe DSL, while the post also compares Rubin’s specifications with NVIDIA HGX B200. ### [Uber Reveals 2025 “Airport of the Year” Award Winners](https://yomu.fyi/post/uber-reveals-2025-airport-of-the-year-award-winners.md) - Company: [Uber](https://yomu.fyi/company/uber.md) - Published: Jun 2, 2026 - Type: Announcement Uber’s 2025 Airport Awards recognize airports in San Antonio, San Diego, Dallas, and Pittsburgh for improving rideshare reliability, infrastructure, and rider experience. San Antonio International Airport won Quickest Pickup, reporting an average on-demand pickup time of just over four minutes after implementing a nearby staging lot, advanced dispatch technology including ExpressMatch, defined pickup zones, and efficient loading practices. San Diego International Airport and Dallas Love Field received Experience Obsessed recognition for redesigned pickup arrangements, while San Diego added a pedestrian-island setup and secondary staging lot and Dallas replaced a difficult garage location with a terminal-adjacent area. Pittsburgh International Airport won Smartest Space for pull-through loading bays, coordinated signage, Rematch and Prematch technology, and a larger driver waiting lot designed to improve matching, reliability, and pickup speed. ### [Welcoming Tatira Andrade, our new Executive Assistant](https://yomu.fyi/post/welcoming-tatira-andrade-our-new-executive-assistant.md) - Company: [Resend](https://yomu.fyi/company/resend.md) - Published: May 20, 2026 - Type: Announcement Tatira Andrade is joining Resend as an Executive Assistant, bringing nearly 16 years of experience in executive support, complex logistics, and process optimization. Her background includes working with companies such as Nubank, where she helped leadership teams streamline daily operations and scale workflows efficiently, while her interests in travel, geography, languages, and organization align with the role. Andrade says she is motivated by making processes and experiences more seamless; she favors Gemini, uses Command + Shift + 4 for screenshots, and currently names Montreal in autumn as a favorite destination. Her advice to ambitious executive assistants is to keep learning, build technology and language skills, stay updated, and remain genuinely curious about their companies and colleagues. ### [Beyond Prediction: Solving the Multiple Knapsack Problem at Scale: How Uber Optimizes Incentives](https://yomu.fyi/post/beyond-prediction-solving-the-multiple-knapsack-problem-at-scale-how-u.md) - Company: [Uber](https://yomu.fyi/company/uber.md) - Published: May 14, 2026 - Type: Explainer Uber’s Tarot platform frames incentive allocation as a Multiple Knapsack Problem (MKP), replacing FIFO contention handling and reactive budget management. It combines segmentation, budget pacing, uplift predictions, multi-objective value scoring, and a plug-and-play optimizer coordinated by an orchestrator. For each run, Tarot forms user-lever opportunities, predicts incremental ROI and cost, reconciles available budget, and selects assignments under team, business, and concurrency constraints before publishing them to domain services. The initial HiGHS linear programming approach took more than 24 hours to find a feasible solution for 100,000 users and one lever, while CP-SAT solved the same dataset in minutes and handled millions of users across multiple levers within stated SLA timeframes. The result is a production architecture that moves beyond heuristics by combining constraint programming with budget pacing, while future work targets automated exploration and model feedback. ### [Zero-Growth Stack, Real Gains: How Stack Allocation Can Save 10% CPU in Go](https://yomu.fyi/post/zero-growth-stack-real-gains-how-stack-allocation-can-save-10-cpu-in-g.md) - Company: [Uber](https://yomu.fyi/company/uber.md) - Published: May 7, 2026 - Type: Benchmark At Uber, repeated Go goroutine stack expansion consumed substantial CPU in a service whose stack-memory usage remained low, making memory a viable trade-off. Go starts goroutines with 2KB stacks and doubles them when needed, copying the previous stack; the post also notes that adaptive stack sizing and stack shrinking can add runtime cost. Uber patched the runtime, disabled adaptive sizing and shrinking, and used profile-guided analysis of pprof traces plus assembly-derived function stack usage to choose a static power-of-two starting size. For the profiled service, the analyzer identified about 19KB of required stack usage, leading to a 32KB setting instead of 2KB. After deployment, stack-growth CPU fell from close to 10% to below 1%, while most instances used about 50MB of stack and a few reached 200MB. ### [Transformation in action: Why AI deployment depth makes the difference](https://yomu.fyi/post/transformation-in-action-why-ai-deployment-depth-makes-the-difference.md) - Company: [Intercom](https://yomu.fyi/company/intercom.md) - Author: Declan Ivory - Published: Mar 20, 2026 - Type: Explainer AI adoption is widespread in customer service, but the report identifies a deployment gap between organizations experimenting with the technology and those integrating it deeply into operations. Based on a survey of 2,470 global support professionals, it finds that 82% of senior leaders invested in AI in the past year and 87% plan to invest in 2026, while only 10% report mature deployment. Mature teams put AI at the core of workflows, give it meaningful responsibility, and use continuous improvement to achieve higher-quality, more consistent support; 43% report that outcome, nearly double the rate for teams at initial deployment. Lightspeed illustrates the approach by combining tailored training, executive alignment, content investment, and AI operations discipline, reaching 72% Fin resolution without human intervention across 43,000-plus monthly requests and 12-plus languages.