Session Type
Workshop
Date & Time
Tuesday, April 9, 2024, 9:00 AM - 3:45 PM
Name
MLIR Workshop
Location Name
PSC I-III
Description

MLIR is a modular, composable compiler infrastructure framework that can be used to build a wide range of compilers and tools. The MLIR workshop co-located with LLVM Dev Mtg brings together developers and users to discuss new developments, uses and explorations of MLIR. MLIR content will also be featured throughout the LLVM Developers’ Meeting, but the workshop provides an opportunity for work in progress talks and novel prototypes.

The workshop will include technical talks, roundtable discussions, and ample opportunities for networking and collaboration with other MLIR developers.

Cost to attend:
Corporate

$125 (register by February 29)  |  $175 (after February 29)

Individual/Academic/Nonprofit
$100 
(register by February 29)  |  $150 (after February 29)

 Student 
$75 (register by February 29)  |  $125 (after February 29)

Talk Order

9:00 Towards a high-performance AI compiler with upstream MLIR
Renato Golin (Intel Labs), Lorenzo Chelini, Adam Siemieniuk, Kavitha Madhu, Niranjan Hasabnis, Hans Pabst, Evangelos Georganas, and Alexander Heinecke

9:30 Control-Flow: Breaking Free of the Basic Block
Jeff Niu (Modular), Mehdi Amini (Nvidia)

10:00 Coffee Break (30min)

10:30 XBLang: an extensible MLIR frontend
Fabian Mora-Cordero, Sunita Chandrasekaran, Johannes Doerfert, Felipe Cabarcas

11:00 A Case Study of the MLIR Data-flow Analysis Framework for Automatic Differentiation
Mai Jacob Peng (McGill University), Oleksandr Zinenko (Google DeepMind),
William S. Moses (University of Illinois Urbana Champaign), Christophe Dubach (McGill University)

11:30 Test-case reduction tools for MLIR
Kunwar Grover (AMD), Jacques Pienaar (Google)

12:00 Lunch Break

13:35 Linalg vs. MDH – A Comparison of Two MLIR Dialects
Jens Hunloh (University of Muenster), Lars Hunloh, Richard Schulze, Sergei Gorlatch, Ari Rasch, Tobias Grosser (University of Cambridge)

13:55 MLIR Auto-Scheduler: A Framework for Automatic Code Optimization in MLIR
Iheb Nassim Aouadj (New York University Abu Dhabi), Riyadh Baghdadi

14:15 Break (30min)

14:45 Roundtables

15:45 End