Good news! Using machine learning for improved compiler optimization!
The issue: "But performance models for machine code are handwritten by a relatively small group of experts and are not properly validated. As a consequence, the simulated performance measurements often deviate from real-life results. "
"“Intel’s [compiler] documents are neither error-free nor complete, and Intel will omit certain things, because it’s proprietary,” ... “However, when you use data, you don’t need to know the documentation. If there’s something hidden you can learn it directly from the data.”"
Tool predicts how fast code will run on a chip | MIT News: MIT researchers have invented a machine-learning tool and dataset of profiled basic blocks that helps better predict how fast computer chips will execute code from various applications.
No comments:
Post a Comment