Intel has stopped supporting its Xeon Phi Knights Mill and Knights Landing accelerators in the LLVM/Clang 19 compiler, signaling the end of MIC architecture support in the Aurora supercomputer. This decision comes after the GCC compiler also ended support for Intel’s Xeon Phi accelerators, likely due to Intel discontinuing hardware support. The deprecation of Xeon Phi support began in LLVM/Clang 18 and will be fully removed with the release of LLVM 19 in September. Intel cited reducing maintenance efforts and simplifying development and support processes as reasons for the removal.
This move aligns with a broader trend among compilers focusing on AI and HPC programs rather than older architectures like Xeon Phi. The Aurora supercomputer, powered by the Intel codename Old Bridge Compute GPU, faced hardware issues, a malfunctioning cooling system, and stability problems that prevented it from reaching its full potential.
In conclusion, Intel’s decision to drop support for Xeon Phi accelerators in the LLVM/Clang 19 compiler signifies a shift towards newer technologies and a focus on AI and HPC programs. This move is in line with industry trends and aims to streamline development and support processes. The Aurora supercomputer, despite being powered by the Intel Old Bridge Compute GPU, faced various hardware and stability issues that hindered its performance. Intel’s move reflects a broader industry shift away from older architectures like Xeon Phi towards more advanced technologies.
Article Source
https://www.tomshardware.com/tech-industry/intel-throws-in-the-towel-on-the-processors-that-killed-the-first-aurora-supercomputer-knights-mill-and-knights-landing-support-removed-from-llvm