Using machine learning to partition streaming programs
Problem
Automatically partitioning streaming parallel programs for multicore systems is difficult because the ideal mapping depends on both the program and the underlying architecture.
Why it matters
Hand-tuning program partitioning is slow and architecture-specific, limiting performance of parallel software.
What the AI did
A machine learning predictor trained offline predicts the ideal partition structure, allowing rapid search of the program space without executing code.
Results
On a 4-core platform, achieved 60% of the best performance found by compiling/executing 3000+ partitions per program; average 1.90× speedup over tuned StreamIt partitioning.
Impact
Shows a learned predictor can automatically rival exhaustive compile-and-execute tuning for multicore partitioning.
Record
| Venue | ACM Transactions on Architecture and Code Optimization |
|---|---|
| Published | 2013 |
| Work type | article |
| Authors | Zheng Wang, Michael O’Boyle |
| Citations | 46 (OpenAlex, at import time) |
| References | 50 |
| Model or method | not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1145/2512436 |
| Open access | bronze |
| Keywords | machine learning, compiler, streaming programs, multicore, partitioning |
Importance rating
real but narrower advance
Compiler-based multicore partitioning affects parallel software performance; abstract shows 1.90x speedups on StreamIt apps, but impact is limited to compiler/tooling domain with no broader scale claimed.
More in Software & Coding
| # | Paper | Year | Citations | Importance |
|---|---|---|---|---|
| 1 | ImageNet classification with deep convolutional neural networks | 2017 | 106,345 |
8
|
| 2 | Deep Residual Learning for Image Recognition | 2016 | 221,134 |
8
|
| 3 | Faster sorting algorithms discovered using deep reinforcement learning | 2023 | 175 |
8
|
| 4 | A deep learning‐based approach for software vulnerability detection using code metrics | 2022 | 42 |
7
|
| 5 | Deep Learning-Based Attack Detection and Classification in Android Devices | 2023 | 41 |
7
|