if you were wondering what a now page was, well, you should check it out and make one too.
I will be at EuroLLVM 2025 in Berlin, presenting our work on further integrating IRDL into the MLIR ecosystem by providing C++ bindings for the dialect-defining dialect.
I’m looking into compiler optimizations for rendering pipelines. The machine learning world has taken the field of “graphics card abuse” to its absolute limit, and there’s no reason its rendering cousin shouldn’t be able to leverage what they’ve done. Cooperative matrix and vector operations give shaders the ability to access the tensor cores that have somehow hopped (heh) their way into GPUs these days – opening the opportunity for new optimizations.
Over the years, the software engineering curriculum has accrued plenty of cruft, and it waterboards students with a deluge of techniques both current and ancient that ends up vastly irrelevant to real-world work and drastically underequips students for their future roles in industry. I hope to simplify the scope of SWE101, reorganizing it under a singular lens:
Software engineering is not a coding problem, it is a people problem.