Home


Dr. Tuck's group focuses on the design of advanced multi-core architectures, compilers and run-time systems.  

We need systems that can scale in terms of performance and power efficiency. As more processing elements are added to chips, they should provide a commensurate benefit in performance that remains power efficient.  One way to ensure that systems scale is through effective exploitation of parallelism from the micro-architecture up to the run-time system.  This is no small task and needs advances in many areas.

There are several themes we are currently exploring that cross-cut the boundaries of computer architecture, compiler design, and the run-time systems:
  • Helper computing
  • Architectural support for program profiling and optimization
  • Speculative Parallelization and Optimization
  • Customized/configurable cores
  • Software Reliability and Fault Tolerance

News

  • Efficient Queueing for Fine-grained Threading My team will be presenting two papers on this theme at upcoming conferences.  In HPCA-17, we will present HAQu (haiku), a hardware accelerated queuing primitive.  We've used this ...
    Posted Nov 23, 2010, 7:26 AM by James Tuck
  • Posh 1.0 available Posh is a speculatively parallelizing compiler built inside GCC.  It is now available for download, and detailed instructions on obtaining and installing it are found here.
    Posted May 18, 2010, 11:07 AM by James Tuck
  • MMT is in the news MMT has been reported on HPCWire, Sciencedaily, EScienceNews, R&DMagazine, Local News.
    Posted May 18, 2010, 2:50 PM by James Tuck
Showing posts 1 - 3 of 3. View more »