This Sunday (2/28), there are two workshops collocated with CGO that have MLIR-related talks:
C4ML (Compilers for Machine Learning):
- Uday Bondhugula, Polymage Labs & Indian Institute of Science: Polyhedral Building Blocks for High-Performance Code Generation in MLIR.
- Xiaoyong Liu, Alibaba: An MLIR-Based end-to-end dynamic shape compiler.
- Chao Liu & Wen-Heng (Jack) Chung, AMD: Realize implicit GEMM-based convolutions on AMD GPU using MLIR.
Fifth LLVM Performance Workshop:
- Vinay M, Ranjith Kumar H, Siddharth Tiwary and Prashantha Nr: Classical Loop Nest Transformation Framework on MLIR.
- Vinay Madhusudan, Ranjith Kumar and Prashanth Nr: Moving LLVM’s code generator to MLIR framework.
- Prashantha Nr and Ranjith Kumar: LTO and Data Layout Optimisations in MLIR.
- Gokcen Kestor: COMET: Domain Specific Compilation for Heterogenous Targets
Registration is $60 ($30 for IEEE/ACM members).
It seems that some of these could be great topics of discussion within our Open Design meeting. I may try to ask the presenters if they are interested in repeating their talk for us
Let’s add two other talks in the main conference:
And I missed that on Monday: @lorenzo_chelini presents Progressive Raising in Multi-level IR (Progressive Raising in Multi-level IR – Google Research)
We’ll have quite a folks floating around at CGO and C4ML (in talks & gather town), so looking forward to seeing folks there!
Does anybody know if workshop talks have been recorded? Or, at least, if presentations can be made available?
(Holding workshops over the weekend is okay-ish for in-person conferences but makes life hard for attendees in the remote setting…)
They were yes (well there was an option to opt-out if I recall correctly, so all may not be posted).
Although I prefer in-person and easier to not get distracted, I found remote easier over weekend as I didn’t spend an extra day traveling (so it took 1 day of my weekend rather than both) and I could fulfill some of my regular responsibilities at home. Instead during the week remote is more difficult for me, as folks know I’m at my desk still