CS 533 Parallel Computer Architectures

Spring 2018

Course Project Info

Project Proposals

Due In Class

Format: One page with the title, names of the team members (3 is best), and explanation of what you want to do

Specify if you will need accounts in a parallel machine, and if so, where?

Example Topics:

Examples from SPRING 2014:

• Communication cost of algorithms in Xeon Phi
• Processing in memory of SIMD operation
• Parallel sorting accelerator
• Autotuning sparse computations
• Measurements of thread scheduling behavior
• Runtime system for data parallelism
• Accelerating record and deterministic replay

Examples from SPRING 2013:

• Exploring architectural enhancements for fine-grain synchronization
• Efficacy on 2 Tier networks
• Quantum COmputing Simulation on CUDA
• Comparison of current programming models
• Memory access determinism in parallel HPC applications
• Network on chip with integrated and configurable cache

Examples from SPRING 2012:

• Study the lock contention in a machine
• Study a problem in the memory hierarchy of a machine
• Address trace generation study
• Operation systems problems
• Parallel simulation
• Effectiveness of an architectural feature in a parallel machine
• Interaction of compiler-architecture
• Interaction of OS-architecture
• Interaction of application-architecture
• Parallel applications (non-trivial ones only!)

Examples from SPRING 2011:

• Study of the scalability of threading and other OS overheads
• Parallel fault-detection architecture
• Replaying MPI I/O
• Parallelization of a numerical application
• Memory subsystem design
• Parallelization of a ray tracing application
• Parallelization of a graph problem
• Development of a parallel system using FPGAs
• Study of the utilization of the cache hierarchy
• Development of software-managed cache coherence

Examples from SPRING 2010:

• Software for deterministic replay of parallel programs
• Parallelization of computationally-intensive image processing algorithms
• Clustered checkpointing schemes for scalable architectures
• Study power-effient chip multiprocessors
• GP-GPU scheduling
• Performance scalability of parallel programs
• Hardware-assisted network I/O management for parallel programs
• Computing in a distributed memory cluster
• Supporting deterministic multi-threading

Examples from SPRING 2009:

• Implementation of a State Exploration Tool
• Implementation of invalidate/update hybrid coherence protocols
• Parallel heap allocator
• Parallelization of the construction of a KD Tree
• Cache-aware version of Intel's TBB
• Fault tolerance in MPI and Shared-memory machines
• Parallelization of the Pathfinder algorithm
• Analysis of the synchronization in an operating system

Examples from SPRING 2008:

• Efficient and Scalable On-chip interconnect networks for thousand-core shared memory machines
• Implementating LogTM-based transactional memory in the M5 simulator
• Signature based data race detection
• A suite of micro-benchmarks for multicore architectures
• A programming model and architecture for scalability in many-core computing
• Clustered processor design for massively parallel manycore accelereator architectures
• Optimizing directory-based cache coherence for scalable CMPs
• Analyzing the behavior of applications on UltraSPARC T1 processor
• Parallelizing linear equation solver
• Exploring Architectural support for speculative multithreading
• Parallelization of a cosmological initial condition generator
• Using transactional-memory: where locks are no good
• Shared Bus Fabric for a Hierarchical Design of multicore processor
• Fluid dynamic simulator acceleration on a GPU cluster

Examples from SPRING 2005:

• Transient Fault Detection with Simultaneous Redundant Threads
• Thread-Level Speculation with Out-of-Order Spawn on a Simultaneous Multithreaded Processor
• Parallelizing Alpha-beta search for optimum performance
• Parallel Ray Tracing
• Implementation and Analysis of Adaptive Routing Strategies for Interconnection Networks on Massively Parallel Systems
• A Parallel Decoded for Ogg Vorbis
• Parallel Logic Simulation
• Implementation of a Charm++ Machine Layer on RDMA Over Ethernet

Examples from SPRING 2004:

• Data Locality Schemes for NUMA
• Parallelization of Security Algorithms
• A Parallel MP3 Decoder for Embedded Systems
• Development of a Linux Driver for Efficient I/O Rollback
• Parallel Genetic Algorithms
• Evaluation of Cache Coherence Protocol Traffic
• Support for Fault-Tolerance in an SMT Processor

Examples from SPRING 2002:

• Distributed Shared Memory Versus Message Passing on Clusters
• Parallel Game Tree Search Algorithms
• Developing and Evaluating Runtime Support for Heterogenous Dynamic Parallelization
• Parallelization of a Quantum Computing Simulator
• Performance of Speculative Locks on Shared Memory Multiprocessors
• Sphinx parallelization
• Parallel Chess
• Barnes Hut Parallelization
• Study of Scalable Directory-Based Cache COherence Schemes
• Fault Tolerant Cache COherence Protocol
• Checkpointing Strategies for Message Passing
• Adaptability of Protocols
• Optimizing collective communication
• Design and evaluation of resource sharing policies in SMTs

Simulator

One possibility is SESC, a multiprocessor simulator that you can use for your projects. Enjoy it!

Project Midterm Reports:

Due in Class

These are some guidelines for the CS533 midterm project report:

- About 8-10 pages

- Describe:

1. Problem you are addressing
2. How you solve it
3. What you have done so far
• code
• machine
• algorithms
• results
4. What you plan to do until the final report

Final project:

The Report is Due after classes finish

Build on top of the midterm report. Use the same structure. Think of it as a conference paper.

In addition, we will have one session of final project presentations. Each team will present for 15 minutes. Please bring a laptop with your presentation. Please be on time and do not take more than the allotted time. Expect a couple of questions at the end.

Students are encouraged to attend the presentations of all the other students.

SCHEDULE: