This Thursday, June 8th (9am California Time, 16:00 UTC), @Hardcode84 will present the progress on python compiler based on Numba and MLIR.
The goal of this project is to create compiler for numeric python code which provides efficient code generation for CPUs and portable and efficient offload capabilities for GPUs.
Numba is used as a frontend and mapped to a set of custom MLIR dialects: Ntensor - dialect for numpy arrays and functions, which supports mutable tensors, broadcasting on dynamic shapes, ArrayAPI-style compute-follows-data approach and lowering to linalg-on-tensors/memrefs. GPUX/gpu_runtime - extension to exisitng GPU dialect which allows more precise control over offloading scope, device/queue selection, kernel loading and supports lowering to SYCL runtime.
If you missed it, we had a first presentation two years ago on this project (slides - recording).
Zoom Meeting Link is unchanged, the presentation will be recorded and posted here and on our talks page on the website as usual.
You can also subscribe to this Google calendar to keep informed about incoming meetings.
Meeting ID: 851 5109 0498
Passcode: 828404
