GCC vs. LLVM Clang On NVIDIA's Tegra K1 Quad-Core Cortex-A15

It's worth noting that LLVM Clang 3.4 and also within LLVM Clang SVN head there still isn't full support for OpenMP, which will lead the compiler to performing poorly for many scientific tests that use OpenMP for exploiting the multi-core potential out of modern systems... Intel developers (among others) have been working on Clang OpenMP support, including their own Git branch of the support, but nothing that's been mainlined to master at the time of writing this article.

With the Himeno Poisson Pressure Solver, LLVM Clang 3.4 was much faster than GCC 4.8.2.

As with x86, under ARM the LLVM Clang compiler continues to live by its promise of offering super-fast compile times compared to the GNU Compiler Collection.