Macrotask-level unlimited speculative execution on multiprocessors

Hayato Yamana, Mitsuhisa Sato, Yuetsu Kodama, Hirofumi Sakane, Shunichi Sakai, Yoshinori Yamaguchi

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Citations (Scopus)

Abstract

The purpose of this paper is to propose a new fast execution scheme of FORTRAN programs. The proposed scheme enables the fast initiation of macrotask when its data dependences are satisfied even if the control flow has not been reached. The previous schemes to parallelize a program including conditional branches have a number of problems - 1) Though the theoretical speedup ratio is up to N when N conditional branches are jumped on either a VLIW or a superscalar machine, the number of N is restricted up to the number of ALU's on a chip, 2) Since conventional control schemes use a few processors to control macrotasks, the overhead to control them is large. The proposed scheme solves these problems - 1) The proposed scheme enables speculative execution between coarse grain tasks, i.e., macrotasks, on multiprocessors by jumping many conditional branches, 2) A distributed control scheme is proposed and implemented on the EM-4 multiprocessor to decrease the control overhead of macrotasks. Preliminary evaluations show that the control overhead of the proposed scheme is smaller than that of the other control schemes. Moreover, it is confirmed that the distributed control can be implemented by using software when the average macrotask execution time is larger than 14.4μs on the EM-4 multiprocessor.

Original languageEnglish
Title of host publicationProceedings of the International Conference on Supercomputing
Place of PublicationNew York, NY, United States
PublisherACM
Pages328-337
Number of pages10
Publication statusPublished - 1995
Externally publishedYes
EventProceedings of the 1995 Conference on Supercomputing - Barcelona, Spain
Duration: 1995 Jul 31995 Jul 7

Other

OtherProceedings of the 1995 Conference on Supercomputing
CityBarcelona, Spain
Period95/7/395/7/7

Fingerprint

Flow control

ASJC Scopus subject areas

  • Computer Science(all)

Cite this

Yamana, H., Sato, M., Kodama, Y., Sakane, H., Sakai, S., & Yamaguchi, Y. (1995). Macrotask-level unlimited speculative execution on multiprocessors. In Proceedings of the International Conference on Supercomputing (pp. 328-337). New York, NY, United States: ACM.

Macrotask-level unlimited speculative execution on multiprocessors. / Yamana, Hayato; Sato, Mitsuhisa; Kodama, Yuetsu; Sakane, Hirofumi; Sakai, Shunichi; Yamaguchi, Yoshinori.

Proceedings of the International Conference on Supercomputing. New York, NY, United States : ACM, 1995. p. 328-337.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Yamana, H, Sato, M, Kodama, Y, Sakane, H, Sakai, S & Yamaguchi, Y 1995, Macrotask-level unlimited speculative execution on multiprocessors. in Proceedings of the International Conference on Supercomputing. ACM, New York, NY, United States, pp. 328-337, Proceedings of the 1995 Conference on Supercomputing, Barcelona, Spain, 95/7/3.
Yamana H, Sato M, Kodama Y, Sakane H, Sakai S, Yamaguchi Y. Macrotask-level unlimited speculative execution on multiprocessors. In Proceedings of the International Conference on Supercomputing. New York, NY, United States: ACM. 1995. p. 328-337
Yamana, Hayato ; Sato, Mitsuhisa ; Kodama, Yuetsu ; Sakane, Hirofumi ; Sakai, Shunichi ; Yamaguchi, Yoshinori. / Macrotask-level unlimited speculative execution on multiprocessors. Proceedings of the International Conference on Supercomputing. New York, NY, United States : ACM, 1995. pp. 328-337
@inproceedings{7f55c0239d8449bb81f56e2c5a2dcc1f,
title = "Macrotask-level unlimited speculative execution on multiprocessors",
abstract = "The purpose of this paper is to propose a new fast execution scheme of FORTRAN programs. The proposed scheme enables the fast initiation of macrotask when its data dependences are satisfied even if the control flow has not been reached. The previous schemes to parallelize a program including conditional branches have a number of problems - 1) Though the theoretical speedup ratio is up to N when N conditional branches are jumped on either a VLIW or a superscalar machine, the number of N is restricted up to the number of ALU's on a chip, 2) Since conventional control schemes use a few processors to control macrotasks, the overhead to control them is large. The proposed scheme solves these problems - 1) The proposed scheme enables speculative execution between coarse grain tasks, i.e., macrotasks, on multiprocessors by jumping many conditional branches, 2) A distributed control scheme is proposed and implemented on the EM-4 multiprocessor to decrease the control overhead of macrotasks. Preliminary evaluations show that the control overhead of the proposed scheme is smaller than that of the other control schemes. Moreover, it is confirmed that the distributed control can be implemented by using software when the average macrotask execution time is larger than 14.4μs on the EM-4 multiprocessor.",
author = "Hayato Yamana and Mitsuhisa Sato and Yuetsu Kodama and Hirofumi Sakane and Shunichi Sakai and Yoshinori Yamaguchi",
year = "1995",
language = "English",
pages = "328--337",
booktitle = "Proceedings of the International Conference on Supercomputing",
publisher = "ACM",

}

TY - GEN

T1 - Macrotask-level unlimited speculative execution on multiprocessors

AU - Yamana, Hayato

AU - Sato, Mitsuhisa

AU - Kodama, Yuetsu

AU - Sakane, Hirofumi

AU - Sakai, Shunichi

AU - Yamaguchi, Yoshinori

PY - 1995

Y1 - 1995

N2 - The purpose of this paper is to propose a new fast execution scheme of FORTRAN programs. The proposed scheme enables the fast initiation of macrotask when its data dependences are satisfied even if the control flow has not been reached. The previous schemes to parallelize a program including conditional branches have a number of problems - 1) Though the theoretical speedup ratio is up to N when N conditional branches are jumped on either a VLIW or a superscalar machine, the number of N is restricted up to the number of ALU's on a chip, 2) Since conventional control schemes use a few processors to control macrotasks, the overhead to control them is large. The proposed scheme solves these problems - 1) The proposed scheme enables speculative execution between coarse grain tasks, i.e., macrotasks, on multiprocessors by jumping many conditional branches, 2) A distributed control scheme is proposed and implemented on the EM-4 multiprocessor to decrease the control overhead of macrotasks. Preliminary evaluations show that the control overhead of the proposed scheme is smaller than that of the other control schemes. Moreover, it is confirmed that the distributed control can be implemented by using software when the average macrotask execution time is larger than 14.4μs on the EM-4 multiprocessor.

AB - The purpose of this paper is to propose a new fast execution scheme of FORTRAN programs. The proposed scheme enables the fast initiation of macrotask when its data dependences are satisfied even if the control flow has not been reached. The previous schemes to parallelize a program including conditional branches have a number of problems - 1) Though the theoretical speedup ratio is up to N when N conditional branches are jumped on either a VLIW or a superscalar machine, the number of N is restricted up to the number of ALU's on a chip, 2) Since conventional control schemes use a few processors to control macrotasks, the overhead to control them is large. The proposed scheme solves these problems - 1) The proposed scheme enables speculative execution between coarse grain tasks, i.e., macrotasks, on multiprocessors by jumping many conditional branches, 2) A distributed control scheme is proposed and implemented on the EM-4 multiprocessor to decrease the control overhead of macrotasks. Preliminary evaluations show that the control overhead of the proposed scheme is smaller than that of the other control schemes. Moreover, it is confirmed that the distributed control can be implemented by using software when the average macrotask execution time is larger than 14.4μs on the EM-4 multiprocessor.

UR - http://www.scopus.com/inward/record.url?scp=0029202416&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0029202416&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:0029202416

SP - 328

EP - 337

BT - Proceedings of the International Conference on Supercomputing

PB - ACM

CY - New York, NY, United States

ER -