Using MPI communication patterns to guide source code transformations

Robert Preissl; Martin Schulz; Dieter Kranzlmüller; Bronis R. De Supinski; Daniel J. Quinlan

doi:10.1007/978-3-540-69389-5_29

Using MPI communication patterns to guide source code transformations

Robert Preissl, Martin Schulz, Dieter Kranzlmüller, Bronis R. De Supinski, Daniel J. Quinlan

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

23 Scopus citations

Abstract

Optimizing the performance of HPC software requires a high-level understanding of communication patterns as well as their relation to source code structures. We describe an algorithm to detect communication patterns in parallel traces and show how these patterns can guide static code analysis. First, we detect patterns that identify potential bottlenecks in MPI communication traces. Next, we associate the patterns with the corresponding nodes in an abstract syntaxtree using the ROSE compiler framework. Finally we perform static analysis on the annotated control flow and system dependence graphs to guide transformations such as code motion or the automatic introduction of MPI collectives.

Original language	English
Title of host publication	Computational Science - ICCS 2008 - 8th International Conference, Proceedings
Pages	253-260
Number of pages	8
Edition	PART 3
DOIs	https://doi.org/10.1007/978-3-540-69389-5_29
State	Published - 2008
Externally published	Yes
Event	8th International Conference on Computational Science, ICCS 2008 - Krakow, Poland Duration: 23 Jun 2008 → 25 Jun 2008

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Number	PART 3
Volume	5103 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	8th International Conference on Computational Science, ICCS 2008
Country/Territory	Poland
City	Krakow
Period	23/06/08 → 25/06/08

Access to Document

10.1007/978-3-540-69389-5_29

Cite this

Preissl, R., Schulz, M., Kranzlmüller, D., De Supinski, B. R., & Quinlan, D. J. (2008). Using MPI communication patterns to guide source code transformations. In Computational Science - ICCS 2008 - 8th International Conference, Proceedings (PART 3 ed., pp. 253-260). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5103 LNCS, No. PART 3). https://doi.org/10.1007/978-3-540-69389-5_29

Preissl, Robert ; Schulz, Martin ; Kranzlmüller, Dieter et al. / Using MPI communication patterns to guide source code transformations. Computational Science - ICCS 2008 - 8th International Conference, Proceedings. PART 3. ed. 2008. pp. 253-260 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 3).

@inproceedings{5aa79e8969244adb8c29ce2fd4da5d0f,

title = "Using MPI communication patterns to guide source code transformations",

abstract = "Optimizing the performance of HPC software requires a high-level understanding of communication patterns as well as their relation to source code structures. We describe an algorithm to detect communication patterns in parallel traces and show how these patterns can guide static code analysis. First, we detect patterns that identify potential bottlenecks in MPI communication traces. Next, we associate the patterns with the corresponding nodes in an abstract syntaxtree using the ROSE compiler framework. Finally we perform static analysis on the annotated control flow and system dependence graphs to guide transformations such as code motion or the automatic introduction of MPI collectives.",

author = "Robert Preissl and Martin Schulz and Dieter Kranzlm{\"u}ller and {De Supinski}, {Bronis R.} and Quinlan, {Daniel J.}",

year = "2008",

doi = "10.1007/978-3-540-69389-5_29",

language = "English",

isbn = "3540693882",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

number = "PART 3",

pages = "253--260",

booktitle = "Computational Science - ICCS 2008 - 8th International Conference, Proceedings",

edition = "PART 3",

note = "8th International Conference on Computational Science, ICCS 2008 ; Conference date: 23-06-2008 Through 25-06-2008",

}

Preissl, R, Schulz, M, Kranzlmüller, D, De Supinski, BR & Quinlan, DJ 2008, Using MPI communication patterns to guide source code transformations. in Computational Science - ICCS 2008 - 8th International Conference, Proceedings. PART 3 edn, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 3, vol. 5103 LNCS, pp. 253-260, 8th International Conference on Computational Science, ICCS 2008, Krakow, Poland, 23/06/08. https://doi.org/10.1007/978-3-540-69389-5_29

Using MPI communication patterns to guide source code transformations. / Preissl, Robert; Schulz, Martin; Kranzlmüller, Dieter et al.
Computational Science - ICCS 2008 - 8th International Conference, Proceedings. PART 3. ed. 2008. p. 253-260 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5103 LNCS, No. PART 3).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Using MPI communication patterns to guide source code transformations

AU - Preissl, Robert

AU - Schulz, Martin

AU - Kranzlmüller, Dieter

AU - De Supinski, Bronis R.

AU - Quinlan, Daniel J.

PY - 2008

Y1 - 2008

N2 - Optimizing the performance of HPC software requires a high-level understanding of communication patterns as well as their relation to source code structures. We describe an algorithm to detect communication patterns in parallel traces and show how these patterns can guide static code analysis. First, we detect patterns that identify potential bottlenecks in MPI communication traces. Next, we associate the patterns with the corresponding nodes in an abstract syntaxtree using the ROSE compiler framework. Finally we perform static analysis on the annotated control flow and system dependence graphs to guide transformations such as code motion or the automatic introduction of MPI collectives.

AB - Optimizing the performance of HPC software requires a high-level understanding of communication patterns as well as their relation to source code structures. We describe an algorithm to detect communication patterns in parallel traces and show how these patterns can guide static code analysis. First, we detect patterns that identify potential bottlenecks in MPI communication traces. Next, we associate the patterns with the corresponding nodes in an abstract syntaxtree using the ROSE compiler framework. Finally we perform static analysis on the annotated control flow and system dependence graphs to guide transformations such as code motion or the automatic introduction of MPI collectives.

UR - http://www.scopus.com/inward/record.url?scp=47749130103&partnerID=8YFLogxK

U2 - 10.1007/978-3-540-69389-5_29

DO - 10.1007/978-3-540-69389-5_29

M3 - Conference contribution

AN - SCOPUS:47749130103

SN - 3540693882

SN - 9783540693888

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 253

EP - 260

BT - Computational Science - ICCS 2008 - 8th International Conference, Proceedings

T2 - 8th International Conference on Computational Science, ICCS 2008

Y2 - 23 June 2008 through 25 June 2008

ER -

Preissl R, Schulz M, Kranzlmüller D, De Supinski BR, Quinlan DJ. Using MPI communication patterns to guide source code transformations. In Computational Science - ICCS 2008 - 8th International Conference, Proceedings. PART 3 ed. 2008. p. 253-260. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 3). doi: 10.1007/978-3-540-69389-5_29

Using MPI communication patterns to guide source code transformations

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this