Skip to content

Replication package for the paper: "Performance Prediction From Source Code Is Task and Domain Specific"

License

Notifications You must be signed in to change notification settings

ubisoft/ubisoft-laforge-predictionperformance

Repository files navigation

Replication package for "Performance Prediction From Source Code Is Task and Domain Specific"

The paper will appear in the proceedings of ICPC'23, RENE track.

All the results are available in the results/ folder.

You can find the data in the data/ folder. Those are compressed with 7z. Please unzip before usage.

The replication instructions are the following:

Install requirements.txt for Python 3.8.

Scenario a: Prediction of the execution time of a program directly:

  • astnn_train_reg.py
  • To reproduce Table 1 and Figure 6, run:
    • python astnn_train_reg.py --n_epoch=20 --n_min_samples=100 --split_method=p
    • python astnn_train_reg.py --n_epoch=20 --n_min_samples=100 --split_method=r
  • astnn_test_reg.py to make plots

Scenario b: Prediction of the task solved by a program:

  • astnn_train_class.py
  • To reproduce Figure 7 and 8, run:
    • python astnn_train_class.py --n_epoch=10 --n_min_samples=100
    • astnn_class4reg.py
    • astnn_similarity_vs_reg.py

Scenario c: For a pair of programs, prediction of which performs better:

  • astnn_train.py
  • To reproduce Table 2 and Figure 9, run:
    • python astnn_train.py --n_epoch=5 --N=50000 --pairs_name=cf_cpp_pairs_1.5_1.1_False --n_min_samples=100 --split_method=p
    • python astnn_train.py --n_epoch=5 --N=50000 --pairs_name=cf_cpp_pairs_1.5_1.1_False --n_min_samples=100 --split_method=rp

© [2023] Ubisoft Entertainment. All Rights Reserved.

About

Replication package for the paper: "Performance Prediction From Source Code Is Task and Domain Specific"

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published