Efficient automated marshaling of C++ data structures for MPI applications

Tansey, Wesley; Tilevich, Eli

doi:10.1109/ipdps.2008.4536307

Cited by 10 publications

(9 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In some cases additional annotations [Hericko et al, 2003, Cisternino et al, 2005, Tansey and Eli, 2008 are required in order to identify data to serialize or to change serialization format.…”

Section: Serializationmentioning

confidence: 99%

Spray: programming with a persistent distributed heap

Grandi

2016

Companion Proceedings of the 2016 ACM SIGPLAN International Conference on Systems, Programming, Languages and Applications: Sof

View full text Add to dashboard Cite

We introduce a programming paradigm for distributed applications based on a persistent distributed heap. A proof-of-concept implementation is provided as a JavaScript library, together with several examples that embody popular patterns for web applications.To those I love, especially to my grandmother Rita, my brother Leonardo, and Carlotta. AcknowledgementsThis work would have been much harder, and surely longer, without the contributions and help I received from many people. I would especially like to thank my supervisors Vincenzo Gervasi and Antonio Cisternino who encouraged me to investigate persistence and distribution in programming languages. The conversations I had with them laid the groundwork of my thesis and helped me during its development.I believe that a valuable contribution to this work comes from Vincenzo Gervasi, who gave me advice on how to write a scientific paper and suggested I adopt a shallow approach to persistence.I thank Andrea Canciani, who gave me his previous research proposal about persistence. It was a good starting point for my study on the stateof-the-art of persistence. I am also particularly grateful for the assistance given by Simone Zenzaro, who helped me to understand the Abstract State Machine method.I would like to thank my friends who closely shared with me the time spent working on the development of a thesis. I would especially thank Daniele Virgilio, who supported me at the beginning of this work with several hints about how to face a thesis, Marco Ponza, who proved one more time to be a priceless adventure companion, Nicola Corti, who always made coffee for me and whose laugh is contagious, and Alessandro Lenzi, who was an amusing desk neighbour.Thanks to the "Il Parcheggione" juggling group, especially the President Cristiana, Federica, Massimo and Valentino, with whom I enjoy my free time. Thanks to the "Food Bytes" community, with whom I enjoyed tasty food and had a good time, and to "Geckosoft", namely Fabio and Davide, who supported me at the end of this work. I also greatly appreciate the support of all the other friends I have in Pisa.My family and all my Venetian friends deserve special credits for supporting me through the entire development of this thesis.Finally, I would like to express my gratitude to Carlotta, who has always believed in me and accompanied me to the end of this path in my life. ContentsIntroduction 1 Persistence and distribution of state are two important aspects in modern distributed applications. Although they may be considered two orthogonal concepts, because persistence concerns the extent in time of state, while distribution concerns its extent in space, they are in fact closely connected.In distributed applications, these two aspects are combined in order to let users access the application from different locations at different times. In general, these applications preserve their state after being used and restore it the next time they are accessed, without regard of locations. However, the state of an application may change ...

show abstract

Section: Serializationmentioning

confidence: 99%

Spray: programming with a persistent distributed heap

Grandi

2016

Companion Proceedings of the 2016 ACM SIGPLAN International Conference on Systems, Programming, Languages and Applications: Sof

View full text Add to dashboard Cite

show abstract

“…For each dynamic array attribute, some approaches [18,11,15,17,6] assume that there is a matching attribute storing its size. To be able to use these approaches for generating the de-/serialization methods for the computer architecture simulator, we need to make each element of array heapSize an attribute as shown in Figure 8 -(b).…”

Section: Application Of Ser++mentioning

confidence: 99%

“…However, implementation of de-/serialization methods of user defined classes is still realized manually with these frameworks. Second, there are tools for generating de-/serialization code, which parse the class structure and generate the functions accordingly [1,4,6,11,15,17,18]. However, these generators have restrictions on how arrays are implemented and do not support serialization of void/function pointers.…”

Section: Related Workmentioning

confidence: 99%

“…In the literature, there is a vast amount of research concerning automatic generation of de-/serialization functions for classes implemented in C++ [1,6,18,11,17]. Still practical, de/serialization code cannot be generated incrementally -without modifying the source code-with the existing approaches for the following reasons: Restriction on the implementation of dynamic arrays: To facilitate the serialization of arrays, existing approaches rely on developers to provide the size of the arrays or dynamic arrays to be implemented with special wrappers.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Exploiting points-to maps for de-/serialization code generation

Ciraci

Villa

2013

Proceedings of the 28th Annual ACM Symposium on Applied Computing

View full text Add to dashboard Cite

Serialization code generators for C++ have restrictions on the implementation of dynamic arrays and void/function pointers. If the target program is not implemented with these restrictions, developers have to manually change the source code to facilitate serialization code generation. Unfortunately, such changes hamper the benefits of code generation, and they are not localized. This paper presents the de-/serialization code generator Ser++ that does not restrict the implementation of these pointer types and, hence, eliminates the need to adapt the source code for serialization code generation.Ser++ can be considered an aspect weaver that i) traces the pointers, ii) identifies the statements in which properties regarding the serialization of pointer attributes can be extracted and, finally, iii) weaves the code to store these properties at runtime. It generates the de-/serialization functions in such a way that they serialize the pointer attributes according to the stored values of the properties. We have successfully used Ser++ to generate de-/serialization methods for a computer architecture and a power-flow simulator, without any modifications to the existing source code.

show abstract

“…These are exemplified by Tansey & Tilevich's tool for generating datatypes for C++ classes [6]. Our approach differs as it transforms packing code to datatypes.…”

Section: Related Workmentioning

confidence: 99%

Automatic datatype generation and optimization

Kjølstad

Hoefler

Snir

2012

Proceedings of the 17th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming

View full text Add to dashboard Cite

Many high performance applications spend considerable time packing noncontiguous data into contiguous communication buffers. MPI Datatypes provide an alternative by describing noncontiguous data layouts. This allows sophisticated hardware to retrieve data directly from application data structures. However, packing codes in real-world applications are often complex and specifying equivalent datatypes is difficult, time-consuming, and error prone. We present an algorithm that automates the transformation. We have implemented the algorithm in a tool that transforms packing code to MPI Datatypes, and evaluated it by transforming 90 packing codes from the NAS Parallel Benchmarks. The transformation allows easy porting of applications to new machines that benefit from datatypes, thus improving programmer productivity.

show abstract

Efficient automated marshaling of C++ data structures for MPI applications

Abstract: We

Cited by 10 publications

References 19 publications

Spray: programming with a persistent distributed heap

Spray: programming with a persistent distributed heap

Exploiting points-to maps for de-/serialization code generation

Automatic datatype generation and optimization

Contact Info

Product

Resources

About