GCC 4.8 vs. LLVM/Clang 3.3 On Intel's Core i7 4770K

For the Jacobi Successive Over-Relaxation test in SciMark2, the performance is noticeably superior on LLVM/Clang.

For the Himeno scientific workload, LLVM/Clang is slower than GCC and unfortunately with the new LLVM/Clang 3.3 the performance had regressed on the Core i7 4770K CPU rather than improve.

The area where LLVM/Clang tends to always have a big advantage over the GNU Compiler Collection is with its amazing compilation speed.

GCC remains faster than LLVM/Clang at running the multi-threaded C-Ray ray-tracer and the 4.8 release made it even faster than its already good standing on GCC 4.7 -- compared to Clang.

In other common computational test profiles used by OpenBenchmarking.org and the Phoronix Test Suite, there isn't too much to see out of the LLVM/Clang vs. GCC comparison on the Intel Core i7 4770K "Haswell" CPU.

Overall, GCC 4.8 is leading in terms of delivering the fastest binaries in the vast range of Linux benchmarks, but LLVM/Clang 3.3 remains very competitive. In some workloads, LLVM/Clang 3.3 is outperforming GCC but for the rest it's at least generally running close to GCC, the long-standing open-source compiler choice for Linux. Where LLVM/Clang continues to win is with its blazingly fast compilation speeds and the LLVM compiler infrastructure's use in other innovative use-cases as commonly covered on Phoronix.