Abstract
Library and language support for scheduling non-blocking tasks has greatly improved, as have lightweight (user) threading packages. However, there is a significant gap between the two developments. In previous work—and in today’s software packages—lightweight thread creation incurs much larger overheads than tasking libraries, even on tasks that end up never blocking. This limitation can be removed. To that end, we describe an extension to the Intel Cilk Plus runtime system, Concurrent Cilk, where tasks are lazily promoted to threads. Concurrent Cilk removes the overhead of thread creation on threads which end up calling no blocking operations, and is the first system to do so for C/C++ with legacy support (standard calling conventions and stack representations). We demonstrate that Concurrent Cilk adds negligible overhead to existing Cilk programs, while its promoted threads remain more efficient than OS threads in terms of context-switch overhead and blocking communication. Further, it enables development of blocking data structures that create non-fork-join dependence graphs—which can expose more parallelism, and better supports data-driven computations waiting on results from remote devices.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
- 2.
Cilk is a work first system, which means that the thread that executes spawn f will begin executing f immediately; it is the continuation of spawn that is exposed for stealing.
- 3.
Ray tracing follows an imaginary line from each pixel in the image into the scene to see what objects are encountered, rather than starting with the objects and drawing (rasterizing) them onto the screen.
- 4.
In other words, manually converting the application to continuation passing style (CPS).
- 5.
Here, and in the rest of this paper, we omit the prefix
which is found in most of the symbols in CilkPlus, and our fork, Concurrent Cilk https://github.com/iu-parfunc/concurrent_cilk. - 6.
A Cilk worker represents a thread local state which sits on top of an OS level thread.
- 7.
The original proof of Cilk’s space and time bounds relies on the critical path of the computation remaining always accessible in this way. Non-uniform probabilities in work-stealing are a concern to some authors of Cilk.
- 8.
However, the specific, narrow case of linear, synchronous dataflow graphs is addressed by recent work on extending Cilk with pipeline parallelism via a new looping construct [10].
- 9.
Using all four cores of an Intel Westmere processor (i5-2400 at 3.10 GHz), 4 Gb memory, Linux 2.6.32, GHC 7.4.2 and Go 1.0.3.
which is found in most of the symbols in CilkPlus, and our fork, Concurrent Cilk References
Intel Cilk Plus. http://software.intel.com/en-us/articles/intel-cilk-plus/
Intel Cilk Plus Application Binary Interface Specification. https://www.cilkplus.org/sites/default/files/open_specifications/CilkPlusABI_1.1.pdf
Agrawal, K., Leiserson, C., Sukha, J.: Executing task graphs using work-stealing. In: IPDPS, pp. 1–12, April 2010
Blumofe, R.D., Joerg, C.F., Kuszmaul, B.C., Leiserson, C.E., Randall, K.H., Zhou, Y.: Cilk: an efficient multithreaded runtime system. SIGPLAN Not. 30, 207–216 (1995)
Fluet, M., Rainey, M., Reppy, J., Shaw, A., Xiao, Y.: Manticore: a heterogeneous parallel language. In: 2007 Workshop on Declarative Aspects of Multicore Programming, DAMP 2007, pp. 37–44. ACM, New York (2007)
Frigo, M., Leiserson, C.E., Randall, K.H.: The implementation of the cilk-5 multithreaded language. SIGPLAN Not. 33(5), 212–223 (1998)
Goldstein, S.C., Schauser, K.E., Culler, D.E.: Lazy threads: implementing a fast parallel call. J. Parallel Distrib. Comput. 37(1), 5–20 (1996)
Google. The Go Programming Language. https://golang.org
Haynes, C.T., Friedman, D.P., Wand, M.: Obtaining coroutines with continuations. Comput. Lang. 11(3.4), 143–153 (1986)
Lee, I., Angelina, T., Leiserson, C.E., Schardl, T.B., Sukha, J., Zhang, Z.: On-the-fly pipeline parallelism. In: Proceedings of the 25th ACM Symposium on Parallelism in Algorithms and Architectures, pp. 140–151. ACM (2013)
Leijen, D., Schulte, W., Burckhardt, S.: The design of a task parallel library. SIGPLAN Not. 44, 227–242 (2009)
Marlow, S., Jones, S.P., Thaller, W.: Extending the haskell foreign function interface with concurrency. In: Proceedings of the ACM SIGPLAN Workshop on Haskell, pp. 22–32. ACM (2004)
Marlow, S., Peyton Jones, S., Singh, S.: Runtime support for multicore haskell. In: International Conference on Functional Programming, ICFP 2009, pp. 65–78. ACM, New York (2009)
Michael, M.M., Scott, M.L.: Simple, fast, and practical non-blocking and blocking concurrent queue algorithms. In: Proceedings of the Fifteenth Annual ACM Symposium on Principles of Distributed Computing, PODC 1996, pp. 267–275. ACM, New York (1996)
Reinders, J.: Intel Threading Building Blocks: Outfitting C++ for Multi-core Processor Parallelism. O’Reilly Media, Sebastopol (2007)
Reppy, J.H.: Concurrent ML: design, application and semantics. In: Lauer, P.E. (ed.) Functional Programming, Concurrency, Simulation and Automated Reasoning. LNCS, vol. 693, pp. 165–198. Springer, Heidelberg (1993)
Rompf, T., Maier, I., Odersky, M.: Implementing first-class polymorphic delimited continuations by a type-directed selective cps-transform. SIGPLAN Not. 44, 317–328 (2009)
Sivaramakrishnan, K., Ziarek, L., Prasad, R., Jagannathan, S.: Lightweight asynchrony using parasitic threads. In: Workshop on Declarative Aspects of Multicore Programming, DAMP 2010, pp. 63–72. ACM, New York (2010)
von Behren, R., Condit, J., Zhou, F., Necula, G.C., Brewer, E.: Capriccio: scalable threads for internet services. SIGOPS Oper. Syst. Rev. 37(5), 268–281 (2003)
Wheeler, K.B., Murphy, R.C., Thain, D.: Qthreads: an api for programming with millions of lightweight threads. In: IPDPS, pp. 1–8. IEEE (2008)
Acknowledgements
This material is based in part upon work supported by the Department of Energy under Award Number DE-SC0008809, and by the National Science Foundation under Grant No. 1337242.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2016 Springer International Publishing Switzerland
About this paper
Cite this paper
Zakian, C.S., Zakian, T.A.K., Kulkarni, A., Chamith, B., Newton, R.R. (2016). Concurrent Cilk: Lazy Promotion from Tasks to Threads in C/C++ . In: Shen, X., Mueller, F., Tuck, J. (eds) Languages and Compilers for Parallel Computing. LCPC 2015. Lecture Notes in Computer Science(), vol 9519. Springer, Cham. https://doi.org/10.1007/978-3-319-29778-1_5
Download citation
DOI: https://doi.org/10.1007/978-3-319-29778-1_5
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-29777-4
Online ISBN: 978-3-319-29778-1
eBook Packages: Computer ScienceComputer Science (R0)