AdaMemScheduling

Generalizable and Scalable Memory-Aware Scheduling: the Power of Adaptivity Methods

Code

Code List

Note that some parameters should be modified befor using. Input/output path is always needed.

scheduler.cpp

Main scheduler methods in the paper, each method corresponds to a namespace. The cost of the topological order can be evaluated by namespace Topsort. Related to all Experiments.

dm-tree.cpp

Randomly generate a tree-based dag, related to tree-based datasets.

dm-scav.cpp

Generator node-deleted graphs for scalability test, related to Exp-9.

dm-scae.cpp

Generator edge-deleted graphs for scalability test, related to Exp-9.

dp.cpp

State compression dynamic programming. Use to calculate the optimal solution on small graphs. Related to Exp-1.

Usage

Step1: Modify the parameters in code, including input/output path, threshold and so on.

Step2: Compile and run.

g++ code.cpp -o code -std=c++17 -O2
./code

Data

Because some datasets used in the paper are too large to be uploaded to GitHub, we have prepared a cloud storage for storing most of the data. However, for Exp-8 and Exp-9, the data is still too large and the cloud storage space is insufficient, and we are unable to upload all the data. A more reasonable approach is to use the generator in the code to generate it. The download links are as follows:

Google Drive

OneDrive