GCC 11 vs. LLVM Clang 12 Compilers On The AMD EPYC 7763

Written by Michael Larabel in Software on 4 May 2021 at 01:18 PM EDT. Page 1 of 4. 7 Comments.

For those wondering how the recent releases of the GCC 11 and LLVM 12 (Clang 12) open-source compilers are competing on AMD Zen 3, here are some recently conducted benchmarks looking at that showdown on an AMD EPYC 7763 1P server.

Now that both Clang 12 and GCC 11 stable are out, fresh stable compiler benchmarks are being carried out on various AArch64 and x86_64 systems. These EPYC 7763 1P tests were wrapped up last month while waiting for that GCC 11.1 stable debut and as a result were using the GCC 11.0.1 20210413 snapshot of the time. Clang 12.0.0 was in its latest release form. Both compilers were running in their release modes.

During this GCC 11 vs. Clang 12 benchmarking, the CFLAGS/CXXFLAGS set were "-O3 -march=native" and a wide variety of open-source C/C++ benchmarks were carried out in looking primarily at the performance of the resulting binaries from each compiler.

The system used for this round of compiler benchmarking was the AMD EPYC 7763 1P with the Supermicro H12SSL-i motherboard, 128GB of RAM, 3.8TB Micron 9300 NVMe SSD, and running Ubuntu 20.04 LTS with the Linux 5.12 Git kernel.

The GCC 11 compiler was still holding its ground against Clang in some of the FFTW benchmarks.

The MrBayes results were meanwhile neck-and-neck for that molecular biology workload.

The WebP image encode performance was a toss-up depending upon the encode settings used as to which compiler was faster on this AMD Zen 3 server.

Similarly, the LZ4 compression benchmarks were mixed.

Related Articles