GCC Compiler Optimization Benchmarks With A 16-Thread Haswell-EP

The latest Linux benchmarks to share of our two new Intel "Haswell-EP" Xeons are some compiler optimization performance tests with the Intel Xeon E5-1680 v3 running with Fedora 21.

In addition to running some GCC vs. LLVM Clang on Fedora 21 compiler benchmarks and some early GCC 5 compiler benchmarking with the eight-core Intel Xeon CPU plus Hyper Threading, I also ran some reference compiler optimization benchmarks with this hardware/software stack. The tests for this one-page article are similar to the many other Phoronix compiler optimization articles and are namely being put out there for those curious about the modern impact of different compiler flags for a high-end Intel Xeon system with Fedora Linux.

Using GCC 4.9.2, benchmarks were done with the CFLAGS/CXXFLAGS being set to -O0 for no optimizations, -O2 for the middle of the road in terms of optimizations, -O3 for great optimizations, -O3 -march=native for also allowing the generated code to target the Haswell Xeon CPU, -O3 -march=native -flto for also including link-time optimizations, and lastly was running -Ofast -march=native for applying all of the fast optimizations.

All of these compiler benchmarks were carried out via the Phoronix Test Suite benchmarking software.
These test results don't come as too much of a surprise compared to our past compiler optimization articles running at similar levels, but it is nice for showing off the incredible speed of the new Intel Xeon CPUs when allowing GCC 4.9 to take full advantage of the processor's available instruction set extensions, etc. You can find these full test results on OpenBenchmarking.org.