Instruction level parallelism (ILP) limitations have forced processor manufacturers to develop multi-core platforms with the expectation that programs will be able to exploit thread level parallelism (TLP). Multi-core programming shifts the burden of locating additional performance away from computer hardware to the software developers, who often attempt high-level redesigns focused on exposing thread level parallelism, as well as explore aggressive optimizations for sequential codes.
Precise dynamic analysis can provide useful guidance for program optimization efforts, including efforts to find and extract thread level parallelism. Unfortunately, finding regions of code amenable to further optimization efforts requires analyzing traces that can quickly grow in size. Analysis of large dynamic traces (e.g. one billion instructions or more) is often impractical for commodity hardware.
An ideal representation for dynamic trace data would provide compression. However, decompressing large software traces, even if decompressed data is never permanently stored, would make many analysis impractical. A better solution would allow analysis of the compressed data, without a costly decompression step. Prior works have developed trace compressors that generate an analyzable representation, but often limit the precision or scope of analyses.
Zero-suppressed binary decision diagram (ZDDs) exhibit many of the desired properties of an ideal trace representation. This thesis shows: (1) dynamic trace data may be represented by zero-suppressed binary decision diagrams (ZDDs); (2) ZDDs allow many analyses to scale; (3) encoding traces as ZDDs can be performed in a reasonable amount of time; and, (4) ZDD-based analyses, such as irrelevant instruction detection and potential coarse-grained thread level parallelism extraction, can reveal a number of performance opportunities in sequential programs.
|Commitee:||Chang, Bor-Yuh Evan, Moseley, Tipp, Siek, Jeremy G., Somenzi, Fabio|
|School:||University of Colorado at Boulder|
|School Location:||United States -- Colorado|
|Source:||DAI-B 72/11, Dissertation Abstracts International|
|Subjects:||Electrical engineering, Computer science|
|Keywords:||Aggressive optimizations, Dynamic program analysis, Multicore programming, Parallelism, Sequential codes, Zero-suppressed BDDs|
Copyright in each Dissertation and Thesis is retained by the author. All Rights Reserved
The supplemental file or files you are about to download were provided to ProQuest by the author as part of a
dissertation or thesis. The supplemental files are provided "AS IS" without warranty. ProQuest is not responsible for the
content, format or impact on the supplemental file(s) on our system. in some cases, the file type may be unknown or
may be a .exe file. We recommend caution as you open such files.
Copyright of the original materials contained in the supplemental file is retained by the author and your access to the
supplemental files is subject to the ProQuest Terms and Conditions of use.
Depending on the size of the file(s) you are downloading, the system may take some time to download them. Please be