AMD AOCC 3.0 Compiler Performance With The EPYC 75F3 - Making Fast Even Faster
AOCC 3.0 was also yielding much better performance than GCC on Ubuntu 20.04 LTS for FinanceBench.
Similar to some of the other image and video encode benchmarks, using AOCC 3.0 for building out the ASTC texture encoder was a sizable win as well compared to GCC on Ubuntu 20.04 LTS.
Microsoft's ONNX Runtime was also frequently delivering the best performance on AOCC 3.0 for this AMD EPYC Milan 2P server of the compilers tested.
Across the span of dozens of benchmarks, using AOCC 3.0 allowed for roughly 6% better performance overall compared to the default GCC 9.3 compiler of Ubuntu 20.04 LTS. It really depends upon the particular workload(s) of most interest to your own needs for how well using the AMD Optimizing C/C++ Compiler can pay off -- especially as shown in some of the image and video encoding tests as well as some deep learning workloads like Intel oneDNN and Microsoft ONNX, AOCC can deliver hefty improvements. Of course, it will be really interesting to see how that AOCC 3.0 performance compares to GCC 11 once the ongoing Znver3 patches settle down and when any more Znver3 work lands in upstream LLVM/Clang. Stay tuned.
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.