Search for: All records

Creators/Authors contains: "Ye, Zihao"

Note: When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external site maintained by the publisher. Some full text articles may not yet be available without a charge during the embargo (administrative interval).
What is a DOI Number?

Some links on this page may take you to non-federal websites. Their policies may differ from this site.

  1. We introduce Mirage Persistent Kernel (MPK), the first compiler and runtime system that automatically transforms multi-GPU model inference into a single high-performance mega-kernel. MPK introduces an SM-level graph representation that captures data dependencies at the granularity of individual streaming multiprocessors (SMs), enabling cross-operator software pipelining, fine-grained kernel overlap, and other previously infeasible GPU optimizations. The MPK compiler lowers tensor programs into highly optimized SM-level task graphs and generates optimized CUDA implementations for all tasks, while the MPK in-kernel parallel runtime executes these tasks within a single mega-kernel using decentralized scheduling across SMs. Together, these components provide end-to-end kernel fusion with minimal developer effort, while preserving the flexibility of existing programming models. Our evaluation shows that MPK significantly outperforms existing kernel-per-operator LLM serving systems by reducing end-to-end inference latency by up to 1.7×, pushing LLM inference performance close to hardware limits. MPK is publicly available at https://github.com/mirage-project/mirage. 
    more » « less
    Free, publicly-accessible full text available July 6, 2027
  2. Storing sunlight in a compact and rechargeable form remains a central challenge for solar energy utilization. Molecular solar thermal (MOST) energy storage systems, which harness photon energy and release it as heat on demand, provide a direct approach, but have long failed to meet practical benchmarks. Inspired by the architecture of DNA, we report a pyrimidone-based MOST system that stores energy in the strained Dewar photoisomer upon excitation at 300 nm. Designed with sustainability in mind, the system operates solvent-free and remains compatible with aqueous environments while overcoming one of the field’s greatest hurdles: the controlled extraction and transfer of stored heat. When catalyzed by acid, the Dewar isomer releases enough heat to boil water (~0.5 mL). These advances help point the way toward decentralized solar heat storage and off-grid energy solutions. 
    more » « less
    Free, publicly-accessible full text available February 12, 2027
  3. Free, publicly-accessible full text available October 22, 2026