LLVM Clang 3.7 vs. GCC Compiler Benchmarks On Linux

Written by Michael Larabel in Software on 31 August 2015 at 11:20 AM EDT. Page 1 of 4. 10 Comments.

With the official release of LLVM 3.7 being imminent, here are some fresh compiler benchmarks comparing its performance on Linux x86_64 to that of LLVM Clang 3.6 as well as GCC 4.9 and GCC 5.2.

If you haven't already, first read about the new features in LLVM / Clang 3.7 as outlined last week on Phoronix. The LLVM 3.7.0 release has already been tagged and its official debut is expected as soon as later today.

Since last week I've been running some fresh Linux 64-bit compiler benchmarks on Ubuntu with an Intel Xeon rig testing:

- GCC 4.9.2 as shipped by Ubuntu 15.04.

- GCC 5.2.0 as the latest stable GNU Compiler Collection release and built from source in a release mode.

- LLVM Clang 3.6.0 as packaged on Ubuntu 15.04.

- LLVM Clang 3.7.0 SVN built in a release/optimized mode and obtained from the 3.7 branch on 26 August.

All tests were done from an Ubuntu 15.04 x86_64 host with Intel Xeon E5-2687W v3 Haswell processor. The same compiler flags were all passed to each compiler and other variables maintained during testing of this GCC vs. Clang compiler benchmarking with the open-source Phoronix Test Suite software.

Intel Xeon Compiler Tests

Originally I was also going to feature some OpenMP benchmarks in this compiler comparison since LLVM/Clang 3.7 now has OpenMP 3.1 support, but with these tests and using the latest LLVM OpenMP library, I was still running into some issues even when setting the appropriate compiler flag. I'm still investigating the issue so for now all of the tests in this article are not using OpenMP.

As a side note, daily LLVM/Clang SVN benchmarks using the daily LLVM APT snapshot repository continue to be done in a fully-automated manner each morning on multiple systems over at LinuxBenchmarking.com.


Related Articles