Heike Jagode
University of Tennessee-Knoxville, TN·Computer Science
Modern computers mix many kinds of hardware: ordinary processors, graphics chips, AI accelerators, and fast networks that link machines together. To make scientific software run quickly, reliably and with less energy, researchers need precise measurements of what that hardware is actually doing.
- Fields
Computer Science
Engineering
- Topics
high-performance computing
performance monitoring
hardware performance counters
- Grants
- 2 known grants, through Aug 2027
What they work on
- Add performance and power monitoring for AI hardware to the MINCER infrastructure so researchers can make informed choices about optimizing and configuring AI systems.
- Collect network measurements through MINCER that support research on network anomaly detection, resilience and energy efficiency.
- Integrate MINCER monitoring with the Chameleon, ACCESS and Open Science Grid platforms to study cloud and edge computing and the link between data transfer and computation.
- Extend the PAPI performance-counter library under SPADE to monitor new and emerging hardware across the computing stack, including vendor accelerators.
- Build a C++ framework that analyzes both performance and accuracy, and test the trade-offs of low-precision floating-point arithmetic in scientific and machine learning applications.
More about the research
These two projects build open-source tools that collect such measurements, from single chips up to large distributed systems.
Why it matters
Scientists who rely on simulations and machine learning, in areas the grants name such as climate science and drug discovery, need their code to run efficiently on fast-changing hardware. Better measurement tools could help them find bottlenecks, cut energy use, improve network security and judge when lower-precision arithmetic is safe to use. Because the software is open source, smaller institutions could gain access to the same performance analysis used at major supercomputing centers.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Ryan Johnson
University of South Dakota, SD·Computer Science
Add computing capacity, including GPU-accelerated nodes, to the Lawrence high performance computing cluster to speed research in clean energy, sustainability, PTSD and climate change across South Dakota.
Bo Yuan
Rutgers University-New Brunswick, NJ·Engineering·Early-career
Develop design principles for building domain expertise into learning-based robot motion planners, improving their reliability and performance in complex, resource-limited real-world settings.
Mahmut Kandemir
Pennsylvania State University, PA·Computer Science
Characterize diverse HPC workloads on real and simulated computational storage devices to identify which computations benefit from running inside storage, and what performance tradeoffs result.
The school
University of Tennessee-Knoxville: 16 labs on PI Pages, $387M a year on research, 426 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.