Improving OpenCL On CPUs, Building Linux

Back in April there was an LLVM European Conference in London where several interesting technical discussions happened. Among the topics covered were auto-vectorization with LLVM, building Linux with LLVM, and using LLVM to improve the performance of OpenCL on CPUs.

The slides and videos of the two-day LLVM London conference were published a few weeks ago at LLVM.org. (I've just been a bit behind in catching up with the non-immediate news items for Phoronix and yesterday's writing of the Google C++ re-factoring reminded me to get out the rest.) For those into the LLVM compiler infrastructure, some of the interesting topics worth looking at the slides or videos would be:

- Argonne National Laboratory talking about auto-vectorization with LLVM (PDF).

- The Google engineer's plans to refactor C++ with Clang to make it more fun.

- The LLVM linker, lld. (PDF)

- Mark Charlebois talking again about building the Linux kernel with LLVM/Clang. (PDF)

- Work towards improving performance of OpenCL on CPUs. (PDF) This work comes out of Saarland University in Germany, the same place where the OpenCL LLVM back-end was born and a new BSD-licensed OpenCL driver.