Beyond shutting down the Clear Linux project, various Linux driver maintainers let go that have even led to some Intel drivers being “orphaned” in the Linux kernel, there is another open-source project that has ended at Intel with the developers departing the company. Though at least this project has found a new open-source home under the NumPy umbrella.
Two years ago Intel published a blazing fast AVX-512 sorting library, x86-simd-sort. The x86-simd-sort library would go on to be added to NumPy for faster number sorting, new algorithms added to it over time along with even greater performance optimizations for multiple times faster performance than alternatives. PyTorch also has been making use of x86-simd-sort due to its speedy performance alongside the popular NumPy library as well as the GRAPE C++ library for parallel graph processing, and a modified version into OpenJDK.
The open-source x86-simd-sort library is used by notable projects and helps illustrate the importance of AVX-512 and even AVX2 for speeding up sorting. It’s great for showcasing on the latest Intel processors as well as AMD. Unfortunately, it is the latest to run into trouble amid the ongoing restructuring at Intel.
The developers of x86-simd-sort were let go from Intel and apparently Intel has no appetite for maintaining such an open-source project anymore. But the good news is that x86-simd-sort will continue on as a community open-source project at NumPy.
Recently this NumPy ticket was opened over needing a new home for x86-simd-sort. The request began with:
“Maintainers of x86-simd-sort no longer work for Intel and we are uncertain of Intel’s support of this library going forward. There are two options:
– Transfer the repository to Numpy.
– Fork and vendor the library under NumPy namespace.”
Fortunately, Intel agreed to transfer the existing x86-simd-sort repository from the Intel organization so it’s now new de facto home is with NumPy.
The x86-simd-sort repository moving forward is now at numpy/x86-simd-sort on GitHub. It’s unfortunate that the developers behind it are no longer employed by Intel and that Intel will apparently not be investing in this notable open-source project any longer, but at least the open-source code can continue on now under the NumPy umbrella.
