Computing, Electronics and Digital Systems
[Closed]
Work description
- Study MLIR compilation flows that allow transforming neural network representations into viable representations for mapping in parallel computing; - Study the current LLVM / MLIR ecosystem and determine flaws and/or potential re-usable flows that allow creating out-of-tree MLIR dialects; - Propose methodology for creating out-of-tree dialects; - Proof-of-concept implementation of MLIR compilation flow using the developed out-of-tree methodology; - Write the grant activity report.
Academic Qualifications
Student in Informatics Engineering or similar
Minimum profile required
- Expert in a programming language;- Fluency in English (spoken and written).
Preference factors
- Experience in compilers (parsing, and IRs), and LLVM and/or MLIR ecosystem; - Basic knowledge of assembly generation; - Fluency in Portuguese (spoken and written).
Application Period
Since 14 Dec 2023 to 28 Dec 2023
[Closed]
Centre
Human-Centered Computing and Information Science