2010
DOI: 10.1007/978-3-642-13374-9_21
|View full text |Cite
|
Sign up to set email alerts
|

Effective Source-to-Source Outlining to Support Whole Program Empirical Optimization

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
37
0

Year Published

2010
2010
2024
2024

Publication Types

Select...
4
3
1

Relationship

2
6

Authors

Journals

citations
Cited by 50 publications
(37 citation statements)
references
References 14 publications
0
37
0
Order By: Relevance
“…The generated OpenMP versions were further compiled using our own OpenMP translator, which is a ROSE-based OpenMP 3.0 implementation targeting the GCC OpenMP runtime library (GOMP) [1]. Internally, our OpenMP translator invokes the ROSE outliner [19] to outline code portions to generate parallel tasks. GCC 4.4.1 was used as the backend compiler with an optimization option -O3.…”
Section: Preliminary Resultsmentioning
confidence: 99%
See 1 more Smart Citation
“…The generated OpenMP versions were further compiled using our own OpenMP translator, which is a ROSE-based OpenMP 3.0 implementation targeting the GCC OpenMP runtime library (GOMP) [1]. Internally, our OpenMP translator invokes the ROSE outliner [19] to outline code portions to generate parallel tasks. GCC 4.4.1 was used as the backend compiler with an optimization option -O3.…”
Section: Preliminary Resultsmentioning
confidence: 99%
“…), class hierarchy analysis and dependence analysis. Representative program translations developed with ROSE are partial redundancy elimination, constant folding, inlining, outlining [19] (separating out a portion of code as a function), and loop transformations [34] (a loop optimizer supporting aggressive loop optimizations such as fusion, fission, interchange, unrolling and blocking).…”
Section: The Rose Compiler Infrastructurementioning
confidence: 99%
“…Currently, the code-sections are outlined manually. In the future, we intend to automate this process using the ROSE compiler framework [12]. Each node running the application keeps track of the best code-variant it has seen thus far in the tuning process.…”
Section: System Designmentioning
confidence: 99%
“…), class hierarchy, data dependence and system dependence. Representative program optimization and translation interfaces cover partial redundancy elimination, constant folding, inlining, outlining [4], and loop transformations. Through a mechanism called persistent attribute, the ROSE AST also allows user-defined data to be attached to any node as a way to extend the IR to store any additional information.…”
Section: The Rose Compilermentioning
confidence: 99%
“…We have developed a set of AST transformation functions to support these common tasks. For example, the ROSE outliner [4] is a general-purpose tool to extract code portions from both C and C++ to create functions. It automatically handles variable passing according to variable scope and use information.…”
Section: Openmp Translation and Runtime Supportmentioning
confidence: 99%