Managing Parallelism 2025
2025-10-20..2025-10-24
Berkeley, United States
[Parallel and Distributed Algorithms for Inference and Optimization (Simons Institute, Oct. 21–24, 2013), 18th USENIX Symposium on Operating Systems Design and Implementation (OSDI '24), 20th USENIX Symposium on Operating Systems Design and Implementation (OSDI '26), OSDI '25, USENIX ATC '22, USENIX ATC '25]
who it's for
parallel algorithm researchers, GPU systems engineers, high-performance computing practitioners, distributed systems researchers, compiler developers, database and transactional systems researchers, machine learning infrastructure engineers, security and cryptography engineers
topics
tensor core computation model; GPU memory and data-movement optimization; static cost semantics for CUDA; sparse tensor kernels for attention and MoE; automatic parallelism management and scheduling; graph algorithms with ANN indices; RDMA-driven distributed graph processing; impossibility results for sharded transactions; multi-SSD scalable parallel algorithms; dynamic GPU data structures and memory management; parallel clustering and density peaks methods; load balancing with duration prediction
| title | author | length |
|---|---|---|
| AMPC | Jakub Lacki | 01h 17m |
Introduces AMPC = MPC + distributed hash table and claims adaptive reads let many graph tasks drop from polylog rounds to constant rounds by doing short… | ||
| The FIDS Theorems: Tensions between Multinode and Multicore Performance in Transactional Systems | Naama Ben-David | 36m |
Paper proves a concrete incompatibility: no sharded, asynchronous transactional system can simultaneously provide serializability, invisible reads and… | ||
| The Parallel Batch-Dynamic Model with Asynchronous Reads | Quanquan Liu | 44m |
Introduces a method that adds asynchronous linearizable reads to a parallel batch‑dynamic model by recording and traversing an explicit dependency graph of… | ||
| Parallel Efficient Clustering with Graph-Based Approximate Nearest Neighbor Search | Julian Shun | 34m |
PECANN implements parallel Density Peaks Clustering for high-dimensional data using graph-based ANN indices and a doubling search that doubles k until a… | ||
| Designing Parallel Data structures | Yan Gu | 35m |
Provides a parallel k-d tree that builds λ levels together by sampling and sieves points into subtrees, cutting I/O to near-optimal and achieving polylog span… | ||
| Scalable Zero-Knowledge Proof Made Easy | TB Schardl | 36m |
Zera extends OpenCilk/Kitsune/Tapir/LLVM to offload nested task-parallel C++ to GPUs, synthesize reductions, insert UVM prefetches and pool small allocations… | ||
| The challenge of managing parallelism and data-movement for tensor computations on GPUs | Saday Sadayappan | 01h 04m |
Claims that dense tensor computations like matrix multiplication and convolution exhibit very regular, structured parallelism so finding a load-balanced… | ||
| Computational model and algorithms for tensor core units | Rezaul Chowdhury | 35m |
Defines a simple TCU model (m,l): a RAM plus one tensor core that performs multiply‑accumulate on an n×√m tall matrix times a √m×√m square matrix (no min‑plus… | ||
| Cost semantics and static resource analysis for CUDA | Stefan Muller | 31m |
Defines a formal cost semantics for CUDA and a static analysis tool RaCUDA that uses automated amortized resource analysis and quantitative Hoare logic (via… | ||
| Unstructured Sparsity Meets Tensor Cores: Lessons from Sparse Attention and MoE | Aparna Chandramowlishwaran | 37m |
Demonstrates that unstructured sparsity in attention and Mixture-of-Experts must be made structured to hit GPU tensor cores, introducing a binary sparse block… | ||
| Parallelism in modern sparse solvers | Aydin Buluç | 01h 13m |
Claims concrete speedups and workable tradeoffs: reordering is now ~25% of runtime so approximate minimum degree (AMD) with a relaxed MIS-2 selection and… | ||
| Automatic Parallelism Management | Sam Westrick | 35m |
A research compiler called MaPLe compiler implements an automatic parallelism manager that delays task creation and amortizes spawn cost using a… | ||
| Register Tiling for Unstructured Sparsity in Neural Network Inference | Maryam Mehri Dehnavi | 28m |
PATCH produces tile-based unstructured sparsity masks using Gumbel-Softmax mask learning over tile choices and 2:4 patterns, POST-trains models for an overall… | ||
| Scalable Parallel Algorithms | Yihan Sun | 37m |
SIGMOD paper implements a work-efficient, easy-to-code k-core decomposition that unifies offline and online peeling, uses ParlayLib on a 96-core machine and… | ||
| Scaling Parallel Algorithms to Massive Datasets using Multi-SSD Machines | Laxman Dhulipala | 35m |
A 64-core Threadripper machine outfitted with 31 PCIe SSDs and about 1/2 TB DRAM saturates at ~135 GB/s aggregate I/O; a user-space library built around… | ||
| Parallel Graph Algorithms | Jeremy Fineman | 01h 12m |
Survey of concrete parallel methods for directed reachability and SSSP that states costs and limits: Spencer's algorithm increases outdegree via two-hop… | ||
| Breaking 3-Factor Approximation for Correlation Clustering in Polylogarithmic Rounds | Nairen Cao | 33m |
The paper gives a parallel PRAM/MPC algorithm that breaks the 3× barrier, achieving 2.4+ε approximation in polylogarithmic rounds (poly(1/ε)·log^O(1) n) by… | ||
| Load Balancing with Duration Predictions | Yossi Azar | 33m |
Proposes an online algorithm for temporary jobs on unrelated machines that uses job duration estimates but not exact times, proving a competitive ratio of… | ||
| Parallel Cluster-BFS and Applications to Shortest Paths | Letong Wang | 29m |
Parallel cluster-BFS implements simultaneous BFS from a cluster by combining bitwise parallelism with thread-level parallelism, with work O(d m (k/w + 1)) and… | ||
| Combining and Aggregating for Fast Synchronization | Panagiota Fatourou | 42m |
Software implements fetch-and-add in software by splitting P threads across m aggregators (k ≈ P/m per aggregator), using per-aggregator AggSum and a global… | ||
| Approximation hardness and online algorithms for resource scheduling | Rathish Das | 35m |
Proves concrete hardness: a gap-preserving reduction shows no (3/2 − ε)-approximation for makespan under integer resource demands unless P=NP, and the natural… | ||
| Dynamic Data Structures and Memory Management on GPUs | Prashant Pandey | 38m |
Library demonstrates an on-GPU memory manager that performs variable-size allocations and frees without host intervention and backs resizeable, shrinkable data… | ||
| On Interplanetary and Relativistic Distributed Computing | Siddhartha Jayanti | 59m |
Develops a formal model tying Einsteinian relativity to asynchronous distributed systems, formalizes the relativity of simultaneity as a constraint on event… | ||
| nothing matches | ||