GNU Cauldron BoF: GCC for AI - topics wanted
Call for topics for the 'GCC for AI' Birds-of-a-Feather session at GNU Cauldron.
Pietra Ferreira is organizing a Birds-of-a-Feather (BoF) session on ‘GCC for AI’ at the upcoming GNU Cauldron and is soliciting topic suggestions from the community. Potential areas include ML-guided optimizations, codegen for AI workloads, and GCC’s positioning relative to MLIR-based toolchains.
- proposer
Soliciting community input for topics to be covered in the 'GCC for AI' Birds-of-a-Feather session at GNU Cauldron.
“Some topics I'm already considering: - ML-guided optimization passes and heuristics (e.g. inlining, register allocation) - Codegen and vectorization for AI/ML workloads (matmul, tensor ops) - Target-specific codegen for AI accelerators - Where GCC fits relative to MLIR-based toolchains for AI workloads”
In Details
This is a call for participation in a Birds-of-a-Feather (BoF) session at the GNU Cauldron conference, focusing on the intersection of GCC and Artificial Intelligence development. The organizer is seeking community input on specific areas like machine learning for compiler optimizations (e.g., guiding pass selection or heuristics), specialized code generation for AI/ML hardware accelerators and workloads (like tensor operations), and how GCC compares to alternative toolchains, particularly those based on MLIR, for AI development.
- GNU Cauldron
- An annual conference for GNU project developers and users.
- BoF
- Birds-of-a-Feather session, an informal meeting at a conference for people with a common interest.
- AI
- Artificial Intelligence, the field of computer science dedicated to creating systems that can perform tasks typically requiring human intelligence.
- ML
- Machine Learning, a subset of AI that focuses on algorithms that allow systems to learn from data.
- optimization passes
- Individual stages within a compiler that analyze and transform the intermediate representation of the code to improve its performance or size.
- codegen
- Code generation, the process of producing machine code or assembly code from an intermediate representation.
- MLIR
- Multi-Level Intermediate Representation, a compiler infrastructure project designed to represent code at various levels of abstraction.