| Title | : | DisGCo: A Compiler for Distributed Graph Programs |
| Speaker | : | Anchu R S (IITM) |
| Details | : | Thu, 28 Feb, 2019 4:00 PM @ Turing Hall |
| Abstract: | : | 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, as the real-world graphs can be very large and may
not fit in the memory of a single machine. Green-Marl is a popularly used Domain Specific Language
(DSL) for coding graph algorithms and is known for its simplicity. However, currently, no compiler can
compile the complete set of Green-Marl programs, to be run on distributed systems. We present DisGCo,
the first compiler to translate any general Green-Marl program to an efficient MPI program that can run
on distributed systems. Besides the issues related to differences in syntax, since Green-Marl gives the programmer a unified view of the whole memory and allows the parallel and serial code to be inter-mixed, translating Green- Marl programs to MPI (SPMD/MPMD style of computation, distributed memory) throws up many other interesting challenges. 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. 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 15.51×. |
