LLVM Clang Shows Off Great Performance Advantage On NVIDIA GH200's Neoverse-V2 Cores
Clang was typically yielding the fastest binaries tested on this NVIDIA GH200 server compared to GCC 13.2 as shipped by Ubuntu 23.10 ARM64.
Across nearly five dozen benchmarks tested, running the Clang 17 binaries were around 9% faster than GCC 13 on average for this Ubuntu Linux AArch64 testing on the GH200. Though the performance advantage of using Clang varies a lot depending upon the workload, so ultimately it's important to test your own particular workload(s) to evaluate the performance advantages or not to using Clang. There's also other factors to consider in choosing between compilers besides just the performance of the generated binary. In any event for those always curious around GCC vs. Clang performance, these numbers from the NVIDIA GH200 with the 72-core Neoverse-V2 CPU cores was quite interesting and some of the most advantageous yet for the LLVM stack. Again though given the popularity of Clang on mobile devices and its work by Apple as the default compiler and now under heavy ARM focus with Apple Silicon, these results shouldn't come as too much of a shocker.
If you enjoyed this article consider joining Phoronix Premium to view this site ad-free, multi-page articles on a single page, and other benefits. PayPal or Stripe tips are also graciously accepted. Thanks for your support.