High Performance Computing

GPU-Accelerated Deep Learning Optimization

Research on leveraging GPU architectures to optimize neural network training and inference through custom kernels, memory hierarchies, and distributed acceleration strategies.

RESEARCH GAP FRONTIERS

Exascale Computing System Architecture Design

Investigates next-generation supercomputer architectures capable of achieving exascale performance through novel processor designs, interconnects, and power efficiency mechanisms.

RESEARCH GAP FRONTIERS

Heterogeneous Computing Load Balancing Algorithms

Develops dynamic load distribution strategies for systems combining CPUs, GPUs, and specialized accelerators to maximize resource utilization and minimize execution time.

RESEARCH GAP FRONTIERS

Quantum-Classical Hybrid Computing Architectures

Explores integration frameworks for quantum processors with classical HPC systems to solve optimization and simulation problems beyond classical capabilities.

RESEARCH GAP FRONTIERS

Machine Learning-Driven HPC Job Scheduling

Applies reinforcement learning and neural networks to optimize job scheduling, resource allocation, and workload prediction in large-scale computing clusters.

RESEARCH GAP FRONTIERS

Advanced MPI Collective Communication Patterns

Optimizes message passing interface collective operations through topology-aware algorithms and novel communication protocols for modern interconnect architectures.

RESEARCH GAP FRONTIERS

Memory Hierarchy and Cache Optimization

Studies novel cache replacement policies, prefetching strategies, and memory layouts to reduce latency and improve bandwidth utilization in HPC systems.

RESEARCH GAP FRONTIERS

Domain-Specific Language Compilation for HPC

Designs compiler frameworks and code generation techniques for domain-specific languages targeting heterogeneous HPC architectures with automatic optimization.

Fault Tolerance and Resilience Mechanisms

Develops checkpoint-restart strategies, algorithmic fault tolerance, and redundancy techniques to maintain correctness in large-scale distributed computing systems.

Tensor Processing Units Utilization Strategies

Investigates methods to effectively utilize specialized tensor processors for machine learning and scientific computing workloads at scale.

Power-Aware Performance Optimization Frameworks

Develops techniques that dynamically adjust processor frequency, voltage, and parallelism levels to minimize energy consumption while maintaining performance targets.

Distributed Graph Processing at Scale

Creates algorithms and systems for processing billion-node graphs using distributed HPC platforms with efficient partitioning and communication patterns.

OpenMP Advanced Task Scheduling Strategies

Develops sophisticated task scheduling policies within OpenMP frameworks to optimize load balancing and data locality on multi-core systems.

InfiniBand Network Protocol Optimization

Optimizes InfiniBand interconnect performance through adaptive routing, congestion control, and quality-of-service mechanisms for HPC applications.

Stencil Computation Auto-Tuning Methods

Develops automated parameter selection and code generation techniques for optimizing stencil-based computations across diverse hardware platforms.

In-Situ Data Analytics for Scientific Computing

Creates integrated analysis frameworks that process massive simulation data in-memory during execution to reduce I/O bottlenecks and storage requirements.

Sparse Linear Algebra Kernel Optimization

Investigates high-performance implementations of sparse matrix operations through novel data structures and computation patterns for iterative solvers.

AI-Based Performance Prediction Modeling

Develops machine learning models to predict application performance on diverse HPC hardware configurations without explicit execution.

Roofline Model-Based Performance Analysis

Applies roofline analysis frameworks to identify performance bottlenecks and guide optimization strategies for memory-bound and compute-bound kernels.

Neuromorphic Computing Architecture Integration

Explores integration of neuromorphic processors with conventional HPC systems to accelerate brain-inspired computing algorithms and simulations.

Photonic Interconnect Technologies for HPC

Investigates optical interconnect architectures as alternatives to electrical networks to reduce latency and power consumption in exascale systems.

Data Movement Minimization Algorithms

Develops computation-communication overlapping and data locality optimization techniques to reduce costly data movement in memory hierarchies.

Floating Point Precision Trade-Off Analysis

Studies mixed-precision arithmetic techniques and reduced-precision computation methods to accelerate applications while maintaining numerical accuracy.

Collaborative Filtering at Petascale

Designs distributed algorithms for large-scale recommendation systems using HPC infrastructure with emphasis on scalability and communication efficiency.

I/O and Storage System Co-Design

Optimizes I/O patterns, file systems, and burst buffer hierarchies to match application requirements and system capabilities for high-throughput data handling.

Molecular Dynamics Simulation Acceleration

Develops high-performance computing techniques for molecular dynamics simulations including force calculation optimization and particle decomposition strategies.

Climate Model High-Resolution Computation

Creates optimized algorithms and implementations for running ultra-high-resolution climate simulations on exascale systems with improved physical accuracy.

Finite Element Method Solver Optimization

Optimizes finite element implementations through adaptive mesh refinement, preconditioner selection, and parallel iterative solver strategies.

Container Orchestration for HPC Workloads

Develops container platforms and orchestration systems designed specifically for HPC requirements including low-overhead virtualization and efficient resource management.

Compiler Vectorization and Loop Transformation

Advances automatic vectorization techniques and polyhedral loop optimization methods to generate high-performance code for modern vector processors.

Machine Learning Inference at Scale

Develops systems and algorithms for distributed inference of large language models and foundation models across HPC clusters.

Parallel Prefix and Scan Algorithm Optimization

Investigates efficient parallel implementation of prefix and scan operations fundamental to many HPC applications on modern architectures.

Network Function Virtualization at Scale

Creates high-performance packet processing and network function virtualization frameworks leveraging HPC techniques for data plane acceleration.

Numerical Precision and Stability Analysis

Studies numerical stability of parallel algorithms and develops techniques to reduce rounding errors in high-performance floating-point computation.

Spectral Element Method High-Order Computing

Optimizes high-order spectral element discretization methods for efficient computation on modern parallel architectures with improved accuracy.

Reconfigurable Computing FPGA Optimization

Develops high-level synthesis and optimization techniques for implementing HPC algorithms on FPGA devices with dynamic reconfiguration capabilities.

Multi-GPU Programming Model Development

Creates abstraction layers and programming models for efficiently managing computation and communication across multiple GPU devices in clusters.

Lattice Boltzmann Method GPU Acceleration

Implements optimized lattice Boltzmann algorithms on GPU architectures for computational fluid dynamics simulations at exascale.

Transactional Memory Scalability Enhancement

Advances hardware and software transactional memory systems to improve programmability and scalability of parallel programs on many-core systems.

Adaptive Runtime System for HPC

Designs runtime systems that dynamically adjust execution strategies based on application behavior and system state for improved performance.

Approximation Algorithms for Billion-Scale Problems

Develops scalable approximation and heuristic algorithms for optimization problems with billions of variables suitable for HPC platforms.

Accelerator-Aware Algorithm Redesign

Fundamentally restructures classical algorithms to match accelerator hardware characteristics including memory access patterns and parallelism constraints.

Blockchain Consensus at High Performance

Develops high-throughput consensus mechanisms and distributed ledger protocols optimized for execution on HPC infrastructure.

Visualization and Data Reduction Techniques

Creates advanced visualization and lossy compression methods to effectively present and store exascale simulation results.

Surrogate Model-Based Optimization

Develops machine learning surrogate models to replace expensive simulations in optimization workflows accelerating design space exploration.

Adaptive Precision Numerical Methods

Creates algorithms that dynamically adjust computational precision throughout execution to balance accuracy and performance requirements.

Software-Defined Infrastructure for HPC

Develops software-defined networking and storage architectures tailored for HPC workloads with dynamic resource configuration capabilities.

Quantum Circuit Simulation Performance

Optimizes classical simulation of quantum circuits on HPC systems using novel tensor contraction and state representation methods.

Ensemble Weather Prediction System Scaling

Optimizes algorithms and systems for running massive ensemble weather forecasting simulations across distributed HPC infrastructure.

Serverless Computing for Batch HPC Workloads

Adapts serverless computing paradigms to manage embarrassingly parallel HPC workloads with elastic resource allocation and automatic scaling.

Asynchronous Many-Task Runtime Systems Design

Development of scalable runtime systems that efficiently schedule and execute millions of fine-grained asynchronous tasks across distributed HPC architectures.

Reduced-Order Modeling for Real-Time Simulation

Investigation of model reduction techniques to accelerate complex scientific simulations while maintaining accuracy for real-time decision-making applications.

Graph Neural Network Training at Exascale

Study of distributed algorithms and communication optimization for training graph neural networks on billion-node graphs using exascale systems.

Memory-Centric Computing Architecture Innovation

Design of novel processor and memory architectures that prioritize data locality and minimize computational latency for memory-intensive workloads.

Bayesian Uncertainty Quantification in HPC

Development of scalable Bayesian inference methods for quantifying uncertainties in large-scale scientific computing simulations and predictions.

Time-Stepper Acceleration and Multigrid Methods

Advancement of multigrid and multilevel techniques to accelerate convergence in time-dependent partial differential equation solvers on massively parallel systems.

Topology-Aware Task Mapping and Scheduling

Development of algorithms that exploit network topology information to optimize task placement and reduce communication overhead in irregular applications.

Automated Kernel Fusion and Loop Nest Optimization

Creation of compiler techniques that automatically fuse compute kernels and optimize loop structures to maximize cache reuse and instruction-level parallelism.

Dynamic Load Balancing for Irregular Applications

Investigation of adaptive load balancing strategies for applications with unpredictable computational patterns and irregular data access across distributed systems.

Machine Learning Model Compression for Edge HPC

Study of techniques to compress and optimize machine learning models for deployment on edge devices connected to HPC infrastructure.

Parallel I/O Patterns for Data-Intensive Computing

Analysis and optimization of I/O access patterns to maximize throughput in data-intensive HPC applications with complex data dependencies.

Accelerated Linear Algebra Using Tensor Cores

Development of novel algorithms leveraging specialized tensor hardware to accelerate dense and sparse linear algebra operations.

Straggler Mitigation in Distributed Computing

Design of techniques to identify and compensate for slow-running tasks that delay overall job completion in large-scale distributed HPC systems.

Coupled Physics Simulation Framework Optimization

Creation of efficient coupling strategies and communication patterns for multi-physics simulations running on heterogeneous accelerated architectures.

Inverse Problem Solving with Machine Learning

Application of deep learning and physics-informed neural networks to solve large-scale inverse problems efficiently on HPC systems.

NUMA-Aware Memory Management Strategies

Development of memory allocation and data placement strategies that minimize remote memory access latency on non-uniform memory access architectures.

Synthetic Data Generation for HPC Benchmarking

Creation of realistic synthetic workloads and datasets that accurately characterize HPC application behavior for performance analysis and optimization.

Application-Aware Network Traffic Engineering

Design of network routing and scheduling algorithms that adapt to application communication patterns to optimize bandwidth utilization.

Federated Learning Across Distributed HPC Clusters

Study of privacy-preserving distributed machine learning techniques that coordinate training across geographically separated HPC facilities.

Octree Adaptive Mesh Refinement Parallelization

Development of efficient parallel algorithms for managing and load-balancing octree-based adaptive mesh refinement on large-scale systems.

GPU Memory Hierarchy Optimization Framework

Creation of tools and techniques to optimize data movement through GPU memory hierarchies including shared memory, L1, L2, and global memory.

Variational Method Acceleration for PDE Solving

Investigation of variational formulations and iterative methods that converge faster for solving partial differential equations on parallel architectures.

Approximate Computing with Precision Control

Study of controlled approximation techniques that trade numerical precision for performance gains while maintaining solution quality bounds.

Collective Communication Optimization for Scale

Research into hierarchical and bandwidth-optimal algorithms for collective operations that scale efficiently to millions of processes.

Privacy-Preserving Data Analytics at Petascale

Development of differential privacy and secure multi-party computation techniques for analyzing sensitive data on HPC systems.

OpenACC Advanced Compiler Code Generation

Enhancement of OpenACC compiler backends to generate optimized code for diverse accelerator architectures with advanced optimization strategies.

Monolithic Domain Decomposition Solver Scaling

Development of coupled domain decomposition methods that scale to hundreds of thousands of subdomains for multiscale simulations.

Real-Time Anomaly Detection in HPC Systems

Creation of machine learning-based monitoring systems that detect hardware failures and performance anomalies in real-time across HPC infrastructure.

Discontinuous Galerkin Method GPU Implementation

Investigation of efficient GPU implementations of discontinuous Galerkin methods for wave propagation and hyperbolic conservation laws.

Distributed Key-Value Store Optimization

Design of scalable distributed data structures and key-value stores optimized for HPC workflows with extreme throughput requirements.

Implicit Function Representation for Visualization

Study of neural implicit representations and their efficient computation for real-time visualization of large-scale scientific data.

Persistent Memory Architecture Integration

Investigation of programming models and algorithms that leverage persistent memory technologies to improve I/O performance and fault tolerance.

Transfer Learning for Scientific Computing

Application of transfer learning techniques from machine learning to accelerate simulations by leveraging pre-trained models from related problems.

Hierarchical Markov Chain Monte Carlo Sampling

Development of multi-level sampling algorithms for uncertainty quantification that scale efficiently to large numbers of uncertain parameters.

Cross-Platform Code Portability Abstraction Layer

Creation of unified programming abstractions that enable efficient code generation across CPUs, GPUs, and emerging accelerator architectures.

Machine Learning Surrogate Modeling Framework

Development of machine learning surrogate models that replace expensive simulations while maintaining accuracy for design optimization workflows.

Algorithmic Differentiation in HPC Applications

Integration of automatic differentiation techniques into large-scale HPC applications for gradient-based optimization and sensitivity analysis.

Sparse Tensor Decomposition at Scale

Research into parallel algorithms for decomposing high-dimensional sparse tensors efficiently across distributed memory systems.

Self-Tuning Parallel Algorithms Framework

Creation of frameworks that automatically tune algorithmic parameters and data layouts based on runtime performance feedback.

Predictive Failure Detection in HPC Networks

Development of predictive models using historical network data to forecast failures and enable proactive maintenance in HPC infrastructure.

Mixed-Precision Iterative Refinement Methods

Design of iterative refinement algorithms that exploit different precision levels to accelerate convergence while maintaining accuracy.

Coarse-Grained Parallelism Task Graph Execution

Investigation of efficient execution models for large-scale task graphs representing complex workflows with coarse-grained dependencies.

Sparsity-Aware Neural Network Acceleration

Study of hardware and software techniques to exploit sparsity patterns in neural networks for faster inference on HPC systems.

Computational Electromagnetics Solver Optimization

Optimization of iterative solvers and preconditioners for large-scale electromagnetic simulations on heterogeneous parallel architectures.

Context-Aware Cache Replacement Policy Design

Development of intelligent cache replacement policies that adapt to application-specific memory access patterns for improved hit rates.

Graph Partitioning for Communication Minimization

Research into advanced graph partitioning algorithms that minimize inter-partition communication in irregular domain decomposition schemes.

Machine Learning-Based Compiler Optimization

Application of machine learning to automatically select optimal compiler flags and transformations for diverse computational kernels.

Scalable Eigenvalue Problem Solver Development

Creation of advanced eigenvalue solver algorithms that scale to billions of unknowns for materials science and quantum chemistry applications.

Network-Aware Data Caching Strategy

Design of intelligent caching systems that exploit network topology and bandwidth information to optimize data placement and prefetching.

Transient Dynamics Simulation GPU Scaling

Investigation of techniques to efficiently scale explicit transient dynamics simulations across multiple GPUs with minimal communication overhead.

FPGA-Based Custom Dataflow Architecture Design

Research on designing reconfigurable dataflow architectures using FPGAs to optimize computation patterns for specialized HPC workloads.

Asynchronous Communication Pattern Optimization

Investigation of efficient asynchronous communication strategies to reduce synchronization overhead in distributed HPC systems.

Reduced-Order Modeling for Scientific Computing

Development of reduced-order models to accelerate large-scale scientific simulations through efficient data compression and approximation.

ARM-Based HPC Cluster Architecture Optimization

Exploration of ARM processors for HPC systems to achieve power efficiency and cost-effectiveness at exascale levels.

Dynamic Memory Scheduling in Accelerated Systems

Research on runtime memory management strategies to optimize memory bandwidth utilization across CPU and accelerator hierarchies.

Graph Neural Network Distributed Training

Study of distributed training algorithms for graph neural networks on heterogeneous HPC clusters with communication efficiency focus.

Topology-Aware Process Placement Strategies

Development of intelligent process mapping algorithms that leverage system topology to minimize communication latency and maximize locality.

Stream Processing for Real-Time Analytics

Investigation of high-throughput streaming frameworks for processing continuous data streams in HPC environments with sub-millisecond latency.

OpenCL Kernel Optimization for Diverse Accelerators

Research on automatic and manual kernel optimization techniques to achieve performance portability across heterogeneous GPU and FPGA platforms.

Collective I/O Optimization for Metadata Operations

Study of coordinated I/O strategies for minimizing metadata bottlenecks in parallel file systems at exascale.

Machine Learning Model Ensemble Acceleration

Investigation of techniques for efficiently training and deploying ensemble machine learning models on distributed HPC systems.

Hierarchical Task Scheduling for Dynamic Workloads

Development of multi-level task scheduling frameworks that adapt to dynamic computational demands and system heterogeneity.

Seismic Imaging Wave Propagation Simulation

Acceleration of full-waveform inversion and reverse-time migration algorithms for high-resolution seismic imaging using HPC.

Heterogeneous Memory System Management

Research on runtime systems for managing multi-tier memory hierarchies including HBM, DDR, and persistent memory in HPC systems.

Sparse Tensor Contraction Optimization

Development of efficient algorithms for sparse tensor operations critical to machine learning and scientific computing workflows.

Time-Stepping Solver Performance Tuning

Optimization of explicit and implicit time integration methods for large-scale PDEs to maximize throughput and minimize wall-clock time.

Quantum-Ready Classical Algorithm Development

Design of classical algorithms structured to leverage quantum acceleration while maintaining performance on current HPC systems.

Bandwidth-Aware Algorithm Redesign Methodology

Systematic approach to reformulating computational algorithms to reduce memory bandwidth requirements and improve arithmetic intensity.

Computational Fluid Dynamics Solver Acceleration

Optimization of CFD simulation kernels for massively parallel systems including turbulence modeling and multiphase flow capabilities.

Checkpoint-Restart Strategy Optimization Framework

Research on intelligent checkpoint selection and compression strategies to minimize downtime overhead in long-running HPC simulations.

Deep Learning Compiler Backend Design

Development of specialized compiler backends for efficiently mapping deep learning operations onto diverse HPC hardware accelerators.

Monte Carlo Variance Reduction Techniques

Investigation of advanced variance reduction methods for accelerating Monte Carlo simulations in HPC environments.

Parallel I/O Format and Serialization Optimization

Research on efficient data formats and serialization protocols optimized for high-bandwidth parallel I/O in scientific applications.

Energy-Performance Trade-Off Analysis Framework

Development of analytical models and tools for characterizing energy-performance Pareto frontiers in HPC system design.

Lattice QCD Simulation High-Order Optimization

Acceleration of lattice quantum chromodynamics simulations through advanced kernel fusion and algorithm optimization techniques.

Communication-Avoiding Matrix Factorization

Design of communication-reducing algorithms for LU, QR, and Cholesky factorizations applicable to petascale linear algebra.

Coarse-Grained Parallelism Exploitation Methods

Strategies for identifying and exploiting task-level parallelism across multiple algorithm stages in scientific computing applications.

Software-Defined Networking for HPC Clusters

Implementation of programmable network switches and controllers optimized for MPI collective operations and inter-node communication.

Multigrid Solver Multilevel Acceleration Design

Optimization of multigrid algorithms for extreme-scale systems with focus on communication efficiency and convergence acceleration.

Synthetic Benchmarking Suite for HPC Evaluation

Creation of comprehensive synthetic workloads to characterize and predict performance of diverse HPC applications and systems.

Particle-in-Cell Plasma Simulation Optimization

Acceleration of PIC simulations through optimized particle management, load balancing, and electromagnetic field computations.

Automated Performance Model Calibration Techniques

Development of machine learning methods for automated calibration of analytical performance models against empirical measurements.

Hybrid Precision Iterative Refinement Methods

Study of algorithms combining multiple precision levels to achieve high-accuracy results while maximizing computational throughput.

Nonlinear Equation System Solver Parallelization

Research on scalable Newton and Newton-Krylov methods for solving large-scale nonlinear systems in scientific simulations.

Cross-Platform Code Generation for Accelerators

Development of compiler techniques for generating efficient code from high-level abstractions across multiple accelerator types.

Bayesian Optimization for Hyperparameter Tuning

Application of Bayesian methods to efficiently explore HPC parameter spaces and identify optimal configurations automatically.

Adaptive Mesh Refinement Load Balancing

Research on dynamic load balancing strategies for AMR applications with time-varying computational domains and heterogeneous workloads.

Conjugate Gradient Solver Iteration Acceleration

Optimization of CG and GMRES solvers through preconditioning strategies and communication-avoiding variants for large sparse systems.

Symbolic Computation for Scientific Software

Integration of symbolic algebra systems with HPC to enable automatic code generation and specialized kernel compilation.

Exascale Storage System Architecture Modeling

Design and evaluation of hierarchical storage systems architecture capable of sustaining exascale data movement requirements.

Machine Learning-Based Performance Prediction Models

Development of neural network models for predicting application performance across diverse HPC system configurations.

Temporal Locality Enhancement for Cache Systems

Research on software and hardware techniques for improving temporal data reuse patterns to enhance cache utilization.

Aerodynamic Shape Optimization Framework

Development of gradient-based optimization loops coupling CFD solvers with design parameterization for aerodynamic applications.

Time Series Analysis at Extreme Scale

Scalable algorithms for time series forecasting, anomaly detection, and pattern recognition on streaming data from large systems.

Approximate Computing for Scientific Simulation

Investigation of controlled approximation strategies to trade accuracy for performance in computationally intensive simulations.

Petri Net Modeling for Workflow Simulation

Use of formal verification methods to analyze and optimize complex HPC workflow execution on heterogeneous platforms.

Network Bandwidth Utilization Maximization

Research on congestion-aware routing and traffic engineering to maximize throughput in high-dimensional interconnection networks.

Metabolic Network Analysis Acceleration Methods

Optimization of flux balance analysis and constraint-based modeling algorithms for systems biology simulations.

GPU Memory Coalescing Pattern Analysis

Research on optimizing memory access patterns to maximize bandwidth utilization and minimize warp divergence in GPU kernels through intelligent coalescing strategies.

Asynchronous Task-Based Runtime Systems

Development of advanced runtime systems that manage asynchronous task execution across heterogeneous architectures with intelligent work-stealing and load balancing.

Bandwidth-Limited Algorithm Transformation

Techniques for algorithmic restructuring to overcome bandwidth limitations in memory-bound applications through computation reordering and data access patterns.

Hierarchical Communication Topology Optimization

Research on optimizing multi-level communication hierarchies in large-scale clusters to reduce latency and congestion in collective operations.

Dynamic Voltage and Frequency Scaling Prediction

Machine learning approaches to predict optimal DVFS settings in real-time for energy-efficient HPC execution without sacrificing performance.

Polyhedral Compilation Framework Enhancement

Advanced polyhedral model techniques for automatic loop transformation, parallelization, and optimization of complex nested loop structures.

Non-Blocking Collective Communication Implementation

Design and optimization of non-blocking collective communication primitives to overlap computation with communication in distributed applications.

Checkpoint Restart Overhead Reduction Methods

Techniques for minimizing checkpoint and restart overhead through incremental checkpointing, compression, and intelligent replica placement strategies.

Irregular Data Access Pattern Optimization

Methods for optimizing applications with irregular memory access patterns through data gathering, compression, and adaptive scheduling techniques.

Topology-Aware Task Placement Algorithm

Algorithms for mapping computational tasks onto cluster topologies while minimizing communication distance and maximizing locality benefits.

High-Order Time Integration Scheme Development

Research on developing and optimizing high-order temporal integration schemes for improved accuracy and stability in large-scale simulations.

Non-Uniform Memory Access Latency Mitigation

Strategies for addressing NUMA latency effects through data replication, migration policies, and first-touch memory allocation optimization.

Adaptive Mesh Refinement Load Distribution

Load balancing techniques specifically designed for AMR simulations that dynamically evolve computational grids with varying resolution requirements.

Iterative Solver Convergence Acceleration

Research on preconditioning techniques, multigrid methods, and Krylov subspace acceleration for faster convergence of large linear systems.

Hardware Performance Counter Data Analysis

Machine learning methods for analyzing hardware performance counter data to identify bottlenecks and guide automatic program optimization.

Distributed Deep Neural Network Training

Optimization techniques for distributed training of large neural networks across multiple GPUs and nodes with efficient gradient aggregation.

Algorithmic Resilience Against Silent Data Corruption

Development of algorithms inherently resilient to silent data corruption through redundant computation, checksums, and self-validating methods.

Cross-Node Interconnect Congestion Mitigation

Techniques for detecting and reducing network congestion in high-speed interconnects through adaptive routing and traffic engineering.

Sparse Matrix Format Adaptation and Selection

Automatic selection and conversion between sparse matrix formats optimized for specific computational kernels and hardware accelerators.

Fine-Grain Data Locality Analysis Framework

Tools and frameworks for detecting and quantifying data locality opportunities at instruction granularity to guide memory optimization.

Quantum Error Correction Circuit Optimization

Research on optimizing quantum error correction circuits and their classical simulation for hybrid quantum-classical HPC systems.

Massive-Scale Graph Partitioning Strategy

Partitioning algorithms for billion-vertex graphs that minimize edge cuts while maintaining load balance across distributed systems.

Autotuning Parameter Space Exploration Methods

Advanced search strategies including Bayesian optimization and active learning for efficient exploration of HPC kernel parameter spaces.

All-to-All Communication Pattern Optimization

Optimization of all-to-all and all-reduce operations through novel communication schedules and pipelined algorithms for scalable systems.

Reduced-Order Model Integration Strategy

Techniques for integrating reduced-order models within HPC simulations to accelerate computation while maintaining accuracy bounds.

Register Allocation Pressure Analysis

Compiler analysis techniques for quantifying and reducing register pressure in GPU kernels through intelligent variable mapping.

Portable SIMD Code Generation Strategy

Methods for generating efficient SIMD code across diverse architectures including AVX-512, ARM NEON, and custom accelerators.

Simulation-Based Performance Prediction Model

Architecture simulation frameworks for predicting HPC application performance on future systems before physical hardware availability.

Overlapping Communication and Computation Scheduling

Runtime scheduling strategies that maximize overlap between network communication and local computation to hide latency in distributed codes.

Integer Linear Programming Optimization Formulation

ILP formulations for HPC optimization problems including scheduling, mapping, and partitioning with exact solver implementations.

Heterogeneous Data Type Precision Management

Systems for managing mixed-precision computation with different floating-point formats to balance accuracy and performance.

Memory Access Pattern Prefetching Strategy

Hardware and software prefetching techniques using learned access patterns to predict and retrieve data before compute units demand it.

Distributed Matrix Multiplication Algorithm Family

Development and optimization of 2.5D and 3D matrix multiplication algorithms minimizing communication for exascale systems.

Synchronization Bottleneck Identification Tool

Tools for detecting and analyzing synchronization bottlenecks in parallel applications through trace analysis and critical path methods.

Implicit-Explicit Time Stepping Integration

Research on IMEX schemes that combine explicit and implicit time steps for efficient simulation of multiscale phenomena.

FPGA-CPU Co-Processing Framework Development

Framework for seamless integration of FPGA accelerators with CPU computation including data movement and synchronization.

Ensemble Computation Resource Management

Systems for managing resources and scheduling of ensemble computations where multiple simulations run with parametric variations.

Loop Tiling and Blocking Configuration Automation

Automatic determination of optimal tile sizes and blocking factors for loop transformations based on hardware characteristics.

Multi-Level Cache Coherence Protocol Design

Novel cache coherence protocols for many-core systems that reduce traffic overhead and improve scalability.

Particle-in-Cell Simulation Optimization

GPU and accelerator optimization of PIC methods for plasma simulations through efficient particle binning and field interpolation.

Dynamic Load Imbalance Correction Strategy

Runtime strategies for detecting and correcting load imbalances through dynamic load balancing with migration overhead minimization.

Automatic Differentiation HPC Compilation

Compilation techniques for efficient automatic differentiation in large-scale machine learning and scientific computing applications.

Heterogeneous Memory Hierarchy Exploitation

Strategies for exploiting systems with multiple memory types including HBM, DRAM, and NVMe to optimize application performance.

Algorithmic Triangle Counting Acceleration

GPU-accelerated algorithms for massive-scale triangle counting in graphs with optimization for different sparsity patterns.

Software Pipeline Optimization for Kernels

Compiler and architecture-level techniques for software pipelining to maximize instruction-level parallelism in compute kernels.

Lightweight Fault Injection Testing Framework

Framework for efficiently injecting and studying the impact of various fault types on HPC application resilience.

Compiler-Guided Data Placement Optimization

Compiler techniques for guiding optimal data placement decisions across heterogeneous memory hierarchies.

Unified Memory Architecture Coherence Protocol Design

Research focuses on developing efficient cache coherence protocols for unified memory systems spanning heterogeneous accelerators and multi-socket architectures to minimize memory latency and maximize bandwidth utilization.

Sparse Tensor Computation and Contraction Optimization

This category investigates advanced algorithms and compiler techniques for optimizing sparse tensor operations and multi-way contractions critical to machine learning and scientific computing workloads at scale.

Machine Learning-Based Algorithm Selection

Systems using machine learning to automatically select optimal algorithms for computational kernels based on problem characteristics.

Bayesian Performance Modeling for HPC Workload Characterization

Research develops probabilistic models using Bayesian inference to predict system performance characteristics and identify bottlenecks in complex HPC applications across heterogeneous computing environments.

Dynamic Graph Repartitioning for Evolving Distributed Networks

This area explores adaptive graph partitioning strategies that maintain load balance and minimize communication overhead for streaming and time-evolving graph problems in large-scale distributed systems.