Efficient, portable implementation of asynchronous multi-place programs

Bikshandi, Ganesh; Castaños, José G.; Kodali, Sreedhar; Nandivada, V. Krishna; Peshansky, Igor; Saraswat, Vijay; Sur, Sayantan; Varma, Pradeep; Wen, Tong

doi:10.1145/1504176.1504215

Cited by 17 publications

(5 citation statements)

References 20 publications

(21 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…SPMDization. There have been many prior works [7,10,16,52] that translate fork-join style code to SPMD code. We use a similar approach and generate MPI SPMD code from Green-Marl.…”

Section: Related Workmentioning

confidence: 99%

DisGCo

Rajendran

Nandivada

2020

ACM Trans. Archit. Code Optim.

Self Cite

View full text Add to dashboard Cite

Graph algorithms are widely used in various applications. Their programmability and performance have garnered a lot of interest among the researchers. Being able to run these graph analytics programs on distributed systems is an important requirement. Green-Marl is a popular Domain Specific Language (DSL) for coding graph algorithms and is known for its simplicity. However, the existing Green-Marl compiler for distributed systems (Green-Marl to Pregel) can only compile limited types of Green-Marl programs (in Pregel canonical form). This severely restricts the types of parallel Green-Marl programs that can be executed on distributed systems. We present DisGCo, the first compiler to translate any general Green-Marl program to equivalent MPI program that can run on distributed systems. Translating Green-Marl programs to MPI (SPMD/MPMD style of computation, distributed memory) presents many other exciting challenges, besides the issues related to differences in syntax, as Green-Marl gives the programmer a unified view of the whole memory and allows the parallel and serial code to be intermixed. We first present the set of challenges involved in translating Green-Marl programs to MPI and then present a systematic approach to do the translation. We also present a few optimization techniques to improve the performance of our generated programs. DisGCo is the first graph DSL compiler that can handle all syntactic capabilities of a practical graph DSL like Green-Marl and generate code that can run on distributed systems. Our preliminary evaluation of DisGCo shows that our generated programs are scalable. Further, compared to the state-of-the-art DH-Falcon compiler that translates a subset of Falcon programs to MPI, our generated codes exhibit a geomean speedup of 17.32×.

show abstract

“…SPMDization. There have been many prior works [7,10,16,52] that translate fork-join style code to SPMD code. We use a similar approach and generate MPI SPMD code from Green-Marl.…”

Section: Related Workmentioning

confidence: 99%

DisGCo

Rajendran

Nandivada

2020

ACM Trans. Archit. Code Optim.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Yonezawa et al [Yonezawa et al(2006)Yonezawa, Wada, and Aida] aim at reducing the barrier synchronization operations, by generating efficient communication code for data transfer operations in a distributed application. Similarly, Bikshandi et al [Bikshandi et al(2009)Bikshandi, Castanos, Kodali, Nandivada, Peshansky, Saraswat, Sur, Varma, and Wen] propose methods to efficiently execute outer-most finish operations. Nagarajan and Gupta [Nagarajan and Gupta(2010)] use speculative execution to reduce the overheads associated with barriers.…”

Section: Related Workmentioning

confidence: 99%

DCAFE: Dynamic load-balanced loop Chunking & Aggressive Finish Elimination for Recursive Task Parallel Programs

Gupta¹,

Sinha²,

Nandivada³

2015

Preprint

Self Cite

View full text Add to dashboard Cite

In this paper, we present two symbiotic optimizations to optimize recursive task parallel (RTP) programs by reducing the task creation and termination overheads. Our first optimization Aggressive Finish-Elimination (AFE) helps reduce the redundant join operations to a large extent. The second optimization Dynamic Load-Balanced loop Chunking (DLBC) extends the prior work on loop chunking to decide on the number of parallel tasks based on the number of available worker threads, at runtime. Further, we discuss the impact of exceptions on our optimizations and extend them to handle RTP programs that may throw exceptions. We implemented DCAFE (= DLBC+AFE) in the X10v2.3 compiler and tested it over a set of benchmark kernels on two different hardwares (a 16-core Intel system and a 64-core AMD system). With respect to the base X10 compiler extended with loop-chunking of Nandivada et al [Nandivada et al.(2013)Nandivada, Shirako, Zhao, and Sarkar] (LC), DCAFE achieved a geometric mean speed up of 5.75× and 4.16× on the Intel and AMD system, respectively. We also present an evaluation with respect to the energy consumption on the Intel system and show that on average, compared to the LC versions, the DCAFE versions consume 71.2% less energy.

show abstract

“…After that, infect is used to start the actual application code on the previously allocated claim. The actual application code that is spread onto infected resources for subsequent parallel execution is called i-let (standing for invaslet 2 ). Once the execution on all cores finishes, the number of cores inside the claim can be altered by calling invade or retreat to either expand or shrink the application's claim.…”

Section: Invasive Computingmentioning

confidence: 99%

“…Fortress [12] is based on functional programming and features implicit parallelism, which does not support the invasive computing concept of explicit resource-and loadawareness. Chapel [5] and X10 [2,4] on the other hand offer already basic language constructs, which enable a programmer to explicitly assign tasks to resources. Chapel was mainly designed with High Performance Computing (HPC) in mind, thus we selected X10 as the base for an implementation of our language for resource-aware programming.…”

Section: Existing Workmentioning

confidence: 99%

Resource-aware programming and simulation of MPSoC architectures through extension of X10

Hannig

Roloff

Snelting

et al. 2011

Proceedings of the 14th International Workshop on Software and Compilers for Embedded Systems

View full text Add to dashboard Cite

The efficient use of future MPSoCs with 1000 or more processor cores requires new means of resource-aware programming to deal with increasing imperfections such as process variation, fault rates, aging effects, and power as well as thermal problems. In this paper, we apply a new approach called invasive computing that enables an application programmer to spread computations to processors deliberately and on purpose at certain points of the program. Such decisions can be made depending on the degree of application parallelism and the state of the underlying resources such as utilization, load, and temperature. The introduced programming constructs for resource-aware programming are embedded into the parallel computing language X10 as developed by IBM using a library-based approach. Moreover, we show how individual heterogeneous MPSoC architectures may be modeled for subsequent functional simulation by defining compute resources such as processors themselves by lightweight threads that are executed in parallel together with the application threads by the X10 run-time system. Thus, the state changes of each hardware resource may be simulated including temperature, aging, and other useful monitor functionality to provide a first high-level programming test-bed for invasive computing.

show abstract

Efficient, portable implementation of asynchronous multi-place programs

Cited by 17 publications

References 20 publications

DisGCo

DisGCo

DCAFE: Dynamic load-balanced loop Chunking & Aggressive Finish Elimination for Recursive Task Parallel Programs

Resource-aware programming and simulation of MPSoC architectures through extension of X10

Contact Info

Product

Resources

About