Advanced Compilation Technology for the AI Stack

AI models are energy hungry and growing fast, while the semiconductor scaling advantage is approaching a cliff. We’ve been relying on Moore’s Law for a long time, but the gain in Power, Performance, Area (PPA) is much slower now. For this reason, learning to compile AI models for efficient hardware execution is crucial. An AI compiler converts an input neural net into an efficient program that runs on target hardware, while optimizing for latency, performance, and power. Qualcomm has developed polyhedral compiler technology that allows for optimizing more complex, larger AI models in a reasonable amount of time.

In this webinar you will find out:

• Why AI compilers are key to making machine learning ubiquitous
• The reasons why compiler optimization techniques need to scale with the explosion of AI
• Some of our innovations in polyhedral compiler technology for AI and our state-of-the-art results
• How we are enabling AI models to scale across a variety of use cases and run efficiently