Akhil Langer
Senior software engineer, NVSHMEM lead and HPC runtime systems
Summary
Work
Education
Projects
Writing
Enhancing Application Portability and Compatibility across New Platforms Using NVIDIA Magnum IO NVSHMEM 3.0
September 1, 2024NVIDIA technical blog post detailing the NVSHMEM 3.0 release: multi-node/multi-interconnect support, host-device ABI backward compatibility, CPU-assisted IBGDA, and object-oriented symmetric heap framework.
Scaling Scientific Computing with NVSHMEM
August 1, 2020NVIDIA technical blog post describing NVSHMEM memory model, case studies (LBANN, Kokkos CGSolve, Jacobi stencil) and scaling results demonstrating NVSHMEM benefits for scientific workloads.
An Initial Assessment of NVSHMEM for High Performance Computing
January 1, 2020Conference paper evaluating NVSHMEM at scale (matrix multiplication and Jacobi solver) on Summit, assessing usability, functionality, and scalability of NVSHMEM for GPU clusters.
Parallel algorithms for two-stage stochastic optimization
January 1, 2015Ph.D. dissertation developing scalable algorithms for two-stage stochastic programs, including improvements to Benders decomposition, parallel branch-and-bound solvers, and application to large-scale allocation problems.
A Text Based Drug Query System for Mobile Phones
January 1, 2014Peer-reviewed paper describing a mobile text-based drug query system that returns biomedical answers and doctor-locator functionality for noisy SMS queries.
Enabling massive parallelism for two-stage stochastic integer optimizations: a branch and bound based approach
December 1, 2011M.S. thesis describing branch-and-bound approaches and performance optimizations to enable parallel solution of stochastic integer programs; applications include aircraft allocation problems.