Lingchuan Meng received his PhD degree in Computer Science from Drexel University in 2015. He has been advised by Dr.Jeremy Johnson. His research interests are in the intersection of symbolic computation, automatic performance tuning and compiler techniques. After graduation, he joined Qualcomm Research Silicon Valley as a senior engineer (LinkedIn profile).
||Software/hardware generation for DSP algorithms by Drexel University, CMU and UIUC
- Project Goal: Automatic library generation for modular FFTs, truncated Fourier transforms (TFT), convolution and polynomial arithmetic
- Techniques: FFT and convolution algorithms, Polynomial arithmetic, Domain-specific language and compiler techniques, Code generation and optimization, Automatic vectorization and parallelization
- Contributions: Automatically generated fixed-size codes and general-size libraries outperform best hand-tuned implementation by 10 to 20 times, as a result of:
- New general-radix and parallel algorithms for the TFT and inverse TFT (ITFT).
- High-level transforms and algorithms expressed in a domain-specific language (SPL).
- Automatic derivation of recursive library structures with symbolic variables.
- Automatic scalar code optimizations, including loop unrolling, merging, interchange, etc.
- Automatic vectorization (SSE+AVX) and parallelization (OpenMP) through rewriting systems.
- Search engine using dynamic programming heuristic for optimal breakdown trees.
- Highly optimized vectorized modular arithmetic implementations.
- Press Coverage: Teaching Computers How to Write Fast Software, NSF Discovery
||High-Assurance Cyber Military Systems intiated by DARPA
- Project Goal: To create technology for the construction of high-assurance cyber-physical systems
- Techniques: Automatic theorem proving, Applicative Common Lisp (ACL)
- Contributions: Assisted the design and development of a new pilot course on automatic theorem proveing. Helped develop theorems for basic vector properties using first order logic.
- Press Coverage: DARPA: Nobody's safe on the Internet, CBS 60 Minutes
||GNU Multiple Precision Arithmetic Library
- Project Goal: To improve the performance of large integer arithmetic in GMP
- Techniques: Arbitrary precision arithmetic, Schönhage–Strassen algorithm, Discrete weighted transform, performance tuning, micro-benchmarking
- Contributions: Integer multiplication using floating point FFT outperforms GMP by up to 25%, as a result of:
- Floating-point FFT-based large integer multiplication
- Formal error bound analysis on FP FFTs
- Automatically generated general-radix and non power-of-2 FFT implementations
- Improved FFT threshold setting algorithm based on micro benchmarking
|Human bone scan analysis and classification, Drexel University College of Medicine
- Project Goal: To retrieve features from the micro-CT scans of human bones and classify future samples with machine learning techniques
- Techniques: Java multimedia, convex hull algorithms, support vector machine, data mining and analysis
- Designed and implemented a user-guided medical system for bone feature retrieval and classification in Java.
- Efficient implementation of the convex hall algorithms.
- The resulting system discovered previously unknown bone features for efficient classification.
|Swarm intelligence, ISR at University of Maryland
- Project Goal: To develop formation control algorithms for flocking of UAVs under energy consumption constraints
- Techniques: Matlab, control theory, potential field algorithms
- Improved the existing algorithm by introducing new terms that smooth sensing transition of the UAVs to further reduce energy consumption.
- Implemented the algorithm and designed simulations in Matlab.
||L. Meng. Automatic Library Generation and Performance Tuning for Modular Polynomial Multiplication. Doctoral thesis 2015
||L. Meng, J. Johnson. High Performance Implementation of the ITFT. PASCO 2014
||L. Meng, J. Johnson. High Performance Implementation of the TFT. ISSAC 2014
||L. Meng, J. Johnson. Auto. Parallel Library Gen. for General-Size Modular FFT Algorithms. CASC 2013
||L. Meng, J. Johnson. Towards Parallel General-Size Library Gen. for Poly. Mul.. ACM Communications in Computer Algebra 47(3), 2013
||L. Meng, Y. Voronenko, J. Johnson, et al. Spiral-Generated Modular FFT Algorithms. PASCO 2010
Teaching and TA
A major part of his teaching/TA experience is about using Maple to teach engineering students to create math models and write programs. Other courses he's guest lectured/assisted before include:
Honors and Awards
||Kobe University, Kobe, Japan.
||Duke University, Durham, North Carolana.
||Konrad Zuse-Zentrum für Informationstechnik Berlin (ZIB), Berlin, Germany.
||Northeastern University, Boston, MA.
||US Naval Academy, Annapolis, MD.
||University of Waterloo, Waterloo, GA.
||Microsoft Research in Redmond, WA.
||External reviewer. Grenoble, France.
||Emory University, Atlanta, GA.
||École de technologie supérieure, Montreal, Canada.
He's built and been maintaining the alumni website for his alma mater Harbin Institute of Technology. He also served as a representative of the Computer Science Department at the EGA during 2011-12.
Last Updated: Jan. 2016