Google
First, we identify a taxonomy of key algorithm variants. Then we develop a template architecture (PolyGraph) that is flexible across these variants while being�...
Our accelerator – PolyGraph – combines the above mech- anisms in a programmable multicore specialized for many variants of graph processing, and is well-suited�...
PolyGraph: exposing the value of flexibility for graph processing accelerators. Authors: Vidushi Dadu. Vidushi Dadu. University of California, Los Angeles. View�...
With dynamic flexibility per-phase, performance further improves by up to 50%. PolyGraph: Exposing the Value of Flexibility forGraph Processing�...
Polygraph [14] generalized prior accelerator designs to perform any of their algorithmic variants and optimize work efficiency based on dataset characteristics.
Video for PolyGraph: Exposing the Value of Flexibility for Graph Processing Accelerators.
May 31, 2021[ISCA' 21 Full] PolyGraph: Exposing the Value of Flexibility in Graph Processing Accelerators.
Duration: 19:53
Posted: May 31, 2021
Jul 1, 2021Title: Polygraph: Exposing the Value of Flexibility for Graph Processing Accelerators. Because of the importance of graph workloads and the�...
Polygraph: Exposing the value of flexibility for graph processing accelerators. V Dadu, S Liu, T Nowatzki. 2021 ACM/IEEE 48th Annual International Symposium�...
Overall we find that flexibility is essential; PolyGraph outperforms similarly provisioned GPUs by mean 49.6� (up to 275�), and the best prior accelerator by�...
Exposing the Value of Flexibility in Graph Processing Accelerators ... Then we develop a template architecture (PolyGraph) that is flexible across�...