IPDPS 2025 Advance Program

THURSDAY - June 5, 2025

DAYS • Thursday • Friday • Saturday

08:30-09:00 – Plenary

Welcome to IPDPS'25: Introductions and Highlights

09:00-10:00 – Plenary

Keynote Address

Hava Siegelmann
Distributed Collaborative AI with Applications to Drones

10:00-10:30 - Coffee Break I - Hall

10:30-11:50

1. Serverless Computing

Room A

• TOSS: Tiering of Serverless Snapshots for Memory-Efficient Serverless Computing 
• Ekko: Fully Decentralized Scheduling for Serverless Edge Computing 
• It Takes Two to Tango: Serverless Workflow Serving via Bilaterally Engaged Resource Adaptation 
• Tide: A Runtime Management Framework for the Things-Edge-Cloud Computing Continuum 

2. Multithreading and Scheduling
Room B

• PISA: An Adversarial Approach To Comparing Task Graph Scheduling Algorithms 
• Enhancing OmpSs-2 Suspendable Tasks by Combining Operating System and User-Level Threads with C++ Coroutines 
• Optimizing Fine-Grained Parallelism Through Dynamic Load Balancing on Multi-Socket Many-Core Systems 
• CALock: Multi-granularity locking in dynamic hierarchies 

3. High-Performance AI
Room C

• AdAPT-S: Effective DNN Pruning via Unified Accuracy and Performance Tuning 
• An Efficient Adaptive Dual-Threshold SVM Based on Heterogeneous Collaboration 
• Accelerating Tensor-train Decomposition on Graph Neural Networks 

11:50-13:30 – Lunch I – Hall

13:30-14:30

4. Emerging Systems and Architectures
Room A

• Energy-Optimal and Low-Depth Algorithmic Primitives for Spatial Dataflow Architectures 
• AQUA: Hardware-Agnostic Qubit Allocation for Quantum Multi-Programming 
• Distributed Construction of Demand-Aware Datacenter Networks 

5. Graph Algorithms I
Room B

• ComSpark: A Holistic Approach for Scalable Clique Counting 
• Less is More: Faster Maximum Clique Search by Work-Avoidance 
• ALGAS: A Low-latency GPU-Accelerated Approximate Nearest Neighbor Search System 

6. AI and Applications
Room C

• AI and HPC Applications on Leadership Computing Platforms: Performance and Scalability Studies 
• Accelerate Coastal Ocean Circulation Model with AI Surrogate 
• FastCHGNet: Training one Universal Interatomic Potential to 1.5 Hours with 32 GPUs 

14:40-15:40

7. Design Space Exploration
Room A

• Compiler, Runtime, and Hardware Parameters Design Space Exploration 
• Using performance projection for design-space exploration on future HPC CPU architectures 
• Pallas: a generic trace format for large HPC trace analysis 

8. Graph Algorithms II
Room B

• Tera-Scale Multilevel Graph Partitioning 
• A Bidirectional GPU Algorithm for Computing Maximum Matchings in Bipartite Graphs 
• Edge-Disjoint Spanning Trees on Star Products 

9. AI for Systems
Room C

• IOAgent: Democratizing Trustworthy HPC I/O Performance Diagnosis Capability via LLMs 
• Graph Neural Network-based Latency Prediction for Stream Processing Task 
• P3Forecast: Personalized Privacy-Preserving Cloud Workload Prediction based on Federated Generative Adversarial Networks 

15:40-16:10 – Coffee Break II – Hall

16:10-18:15 - Plenary

Best Paper Nominees

• FlexRLHF: A Flexible Placement and Parallelism Framework for Efficient RLHF Training 
• Enabling Efficient Error-controlled Lossy Compression for Unstructured Scientific Data 
• PolyMorphous: An MLIR-Based Polyhedral Compiler with Loop Transformation Primitives 
• Parallel scheduling of task graphs with minimal memory requirements 
• The Artificial Scientist: in-transit Machine Learning of Plasma Simulations 

FRIDAY - JUNE 6, 2025

DAYS • Thursday • Friday • Saturday

See full paper information here with authors & affiliations and link to abstract.

IEEE CS Charles Babbage Award Announcement

08:50-09:50 – Plenary

Babbage Keynote Address

Srinivas Aluru
The Power of Parallelism: Accelerating Discovery in the Biosciences 

09:50-10:20 – Coffee Break III – Hall

10:20-11:20

10. Data Flow and Scheduling
Room A

• Automatically Inferring Detailed & Interpretable Workflow Scaling Models for Better Scheduling 
• CELLO: Co-designing Schedule and Hybrid Implicit/Explicit Buffer for Complex Tensor Reuse 
• Locality Aware Process Remapping for Distributed-Memory Graph Workloads 

11. HPC for Biology
Room B

• A Work-Optimal Parallel Algorithm for Aligning Sequences to Genome Graphs 
• An Asynchronous Distributed-Memory Parallel Algorithm for k-mer Counting 
• Pandemics in silico: Scaling Agent-based Simulations on Realistic Social Contact Networks 

12. Federated Learning
Room C

• Air-FedGA: A Grouping Asynchronous Federated Learning Mechanism Exploiting Over-the-air Computation 
• IP-FL: Incentive-driven Personalization in Federated Learning 
• SEAFL: Enhancing Efficiency in Semi-Asynchronous Federated Learning through Adaptive Aggregation and Selective Training 

11:30-12:30

13. Compilation and Code Generation
Room A

• Leveraging Compilation Statistics for Compiler Phase Ordering 
• PCEBench: A Multi-dimensional Benchmark for Evaluating Large Language Models in Parallel Code Generation 
• Gensor: A Graph-based Construction Tensor Compilation Method for Deep Learning 

14. Data and Signal Processing
Room B

• SPRT²: Scalable, Parallel, and Real-Time fMRI Data Analysis on Heterogeneous Architectures 
• The Tensor-Core Beamformer: A High-Speed Signal-Processing Library for Multidisciplinary Use 
• Parallel-in-Time Kalman Smoothing Using Orthogonal Transformations 

15. Security and Privacy
Room C

• FLAME: Federated Learning for Attack Mitigation and Evasion 
• Hybrid-Granularity Parallelism Support for Fast Transaction Processing in Blockchain-based Federated Learning 
• Pair-then-Aggregate: Simplified and Efficient Parallel Programming Paradigm for Secure Multi-party Computation 

12:30-14:00 – Lunch II – Hall

14:00-15:00

16. File System Performance
Room A

• Be Aware of Metadata Corruption in Parallel File System: It Can Be Silent and Catastrophic 
• LaOvl: Lifecycle-Aware Overlay File System for Efficient Container I/O in Cloud Computing 
• KVAccel: A Novel Write Accelerator for LSM-Tree-Based KV Stores with Host-SSD Collaboration 

17. HPC Applications I
Room B

• Accelerating the Dutch Atmospheric Large-Eddy Simulation (DALES) model with OpenACC 
• Automated MPI-X code generation for scalable finite-difference solvers 
• A GPU-Accelerated Distributed Algorithm for Optimal Power Flow in Distribution Systems 

18. Latency and Performance for ML
Room C

• PredTOP: Latency Predictor for Distributed Deep Learning Training with Operator Parallelism 
• Reducing the End-to-End Latency of DNN-based Recommendation Systems Deployed in GPU Pools 
• Improving Accuracy and Efficiency of Graph Embedding Training with Fine-Grained Parameter Management 

15:10-16:10

19. Storage and I/O
Room A

• A Deep Look into The Temporal I/O Behavior of HPC Applications 
• VerifyIO: Verifying Adherence to Parallel I/O Consistency Semantics 
• AdapTBF: Decentralized Bandwidth Control via Adaptive Token Borrowing for HPC Storage 

20. HPC Applications II
Room B

• A New Spin on the Fast Multipole Method for GPUs: Rethinking the Far-Field Operators 
• Large Scale Finite-Temperature Real-time Time Dependent Density Functional Theory Calculation with Hybrid Functional on ARM and GPU Systems 
• Improving Parallel Scalability for Molecular Dynamics Simulations in the Exascale Era 

21. Performance Profiling and Characterization
Room C

• Phase-based Frequency Scaling for Energy-efficient Heterogeneous Computing 
• GNNPerf: Towards Effective Performance Profiling and Analysis across GNN Frameworks 
• DeepBAT: Performance and Cost Optimization of Serverless Inference Using Transformers 

16:10-17:30

Conference Poster Reception
Hall

17:30-18:30

PhD Forum Posters
Hall

SATURDAY - JUNE 7, 2025

DAYS • Thursday • Friday • Saturday

See full paper information here with authors & affiliations and link to abstract.

08:30-08:50 – Plenary

Awards and IPDPS'26 Announcement

08:50-09:50 – Plenary

Keynote Address

David Keyes
For What the Bell Tolls 

09:50-10:20 – Coffee Break V – Hall

10:20-11:20

22. Compression and Data Reduction I
Room A

• To Compress or Not To Compress: Energy and Runtime Trade-Offs in Lossy Compressed I/O 
• Fast and Effective Lossy Compression on GPUs and CPUs with Guaranteed Error Bounds 
• A Memory-efficient and Computation-balanced Lossy Compressor on Wafer-Scale Engine 

23. Matrix Multiplication
Room B

• BRP-SpMM: Block-Row Partition Based Sparse Matrix Multiplication with Tensor and CUDA Cores 
• Graph Input-Aware Matrix Multiplication for Pruned Graph Neural Network Acceleration 
• NM-SpMM: Accelerating Matrix Multiplication Using N:M Sparsity with GPGPU 

24. Communication
Room C

• Unified Designs of Multi-rail-aware MPI Allreduce and Alltoall Operations Across Diverse GPU and Interconnect Systems 
• HiCCL: A Hierarchical Collective Communication Library 
• NBLFQ: a lock-free MPMC queue optimized for low contention 

11:30-12:30

25. Compression and Data Reduction II
Room A

• Improving the Efficiency of Interpolation-Based Scientific Data Compressors with Adaptive Quantization Index Prediction 
• An Adaptive Two-Stage Algorithm for Error-Bounded Scientific Data Compression 
• Achieving Better Benefits via Flexible Feature Matching in Post-Deduplication Delta Compression 

26. Linear Solvers
Room B

• Scalable and portable LU factorization with partial pivoting on top of runtime systems 
• Accelerating Sparse Linear Solvers on Intelligence Processing Units 
• Adaptive s-step GMRES with randomized and truncated low-synchronization orthogonalization 

27. Memory and Networking
Room C

• Performance Characterization of CXL Memory and Its Use Cases 
• RXT: Reflexive Address Translation for Pointer-Chasing Workloads 
• CoRD: Converged RDMA Dataplane 

12:30-14:00 – Lunch III – Hall

14:00-15:00

28. Compression and Data Reduction III
Room A

• Accelerating Graph Neural Networks Using a Novel Computation-Friendly Matrix Compression Format 
• HPDR: High-Performance Portable Scientific Data Reduction Framework 
• Sensitivity and Impacts on Parallel Compression of Prediction of Lossy Compression Ratios for Scientific Data 

29. Graph Processing
Room B

• TaijiGraph: An Out-of-core Graph Processing System Enhanced with Computational Storage 
• CORD: Parallelizing Query Processing across Multiple Computational Storage Devices 
• Matcha: A Language and Compiler for Backtracking-based Subgraph Matching 

30. Transformers
Room C

• FATHOM: Fast Attention Through Optimizing Memory 
• Longer Attention Span: Increasing Transformer Context Length with Sparse Graph Processing Techniques 
• Characterizing the Behavior and Impact of KV Caching on Transformer Inferences under Concurrency 

15:10-16:10

31. Data and Image Processing
Room A

• SymProp: Scaling Sparse Symmetric Tucker Decomposition via Symmetry Propagation 
• Accelerating Homotopy Continuation with GPUs: Application to Trifocal Pose Estimation 
• Enhanced JPEG Decoding using PIM Architectures with Parallel MCU Processing 

32. Error Prediction and Fault Tolerance
Room B

• An Effective Uncorrectable Memory Error Prediction Framework by Exploiting UPH Indicator in Production Environment 
• Fine-Grained Global Search for Inputs Triggering Floating-Point Exceptions in GPU Programs 
• GuardianOMP: A framework for highly productive fault tolerance via OpenMP task-level replication 

33. High-Performance Inference
Room C

• InkStream: Instantaneous GNN Inference on Dynamic Graphs via Incremental Update 
• GIFTS: Efficient \underline{G}CN \underline{I}nference \underline{F}ramework on Py\underline{T}orch-CPU via Exploring the \underline{S}parsity 
• MeanCache: User-Centric Semantic Caching for LLM Web Services