13: High performance Clojure numerics with Chris Nuernberger

Chris Nuernberger talks about his work on tvm-clj, unsigned bytes on the JVM, efficient copying of data, neural networks, GPUs. Chris has a wide background across many different areas of computer science and software engineering. He first got into GPGPU programming around 2008 with a research group at CU trying to auto-optimize a simple linear algebra expression (y = Ax + b). He first got into LISP with cmucl after a bout of RSI (pain in hands and fingers) convinced him that mainstream languages at the time (C++, python, C#, Java, ML) didn’t offer enough leverage for the types of projects that he was involved in. He is currently a partner at company named Tech Ascent based in Boulder, Colorado. tvm-clj Next Gen Numeric Compilers TVM Time tech.datatype Cortex TVM Halide XGBoost Smile