L
Lingjia Tang
Researcher at University of Michigan
Publications - 77
Citations - 5361
Lingjia Tang is an academic researcher from University of Michigan. The author has contributed to research in topics: Server & Quality of service. The author has an hindex of 31, co-authored 73 publications receiving 4071 citations. Previous affiliations of Lingjia Tang include University of Virginia & University of California, San Diego.
Papers
More filters
Journal ArticleDOI
Reining in Long Tails in Warehouse-Scale Computers with Quick Voltage Boosting Using Adrenaline
Chang-Hong Hsu,Yunqi Zhang,Michael A. Laurenzano,David Meisner,Thomas F. Wenisch,Ronald G. Dreslinski,Jason Mars,Lingjia Tang +7 more
TL;DR: This work proposes Adrenaline, an approach to leverage finer-granularity (tens of nanoseconds) voltage boosting to effectively rein in the tail latency with query-level precision, and demonstrates the effectiveness of the methodology under various workload configurations.
Posted Content
Rethinking Numerical Representations for Deep Neural Networks
Parker Hill,Babak Zamirai,Shengshuo Lu,Yu-Wei Chao,Michael A. Laurenzano,Mehrzad Samadi,Marios C. Papaefthymiou,Scott Mahlke,Thomas F. Wenisch,Jia Deng,Lingjia Tang,Jason Mars +11 more
TL;DR: This work explores unconventional narrow-precision floating-point representations as it relates to inference accuracy and efficiency to steer the improved design of future DNN platforms and presents a novel technique that drastically reduces the time required to derive the optimal precision configuration.
Patent
Runtime compiler environment with dynamic co-located code execution
TL;DR: In this paper, a co-designed compiler and runtime system that virtualizes a selected set of edges in a host program, where these edges provide hooks through which the runtime system may redirect execution into an intermediate representation utilized to optimize introspective and extrospective processes.
Journal ArticleDOI
PowerChop: identifying and managing non-critical units in hybrid processor architectures
TL;DR: This work introduces PowerChop, a novel technique that leverages the unique capabilities of HW/SW co-designed hybrid processors to enact unit-level power management at the application phase level, and significantly decreases power consumption.
Patent
System and methods for sharing memory subsystem resources among datacenter applications
TL;DR: In this article, the authors present a system and methods for mapping applications onto system resource of a computing platform using control circuitry, using a request to run a plurality of applications on a computing platforms having a pluralityof system resources.