GCC vs. LLVM Clang On NVIDIA's Tegra K1 Quad-Core Cortex-A15
With the SciMark collection of C-based computational tests, LLVM's Clang 3.4 compiler was running measurably ahead of GCC 4.8.2 on the NVIDIA Jetson TK1 board with Tegra K1 quad-core ARM Cortex-A15 SoC.
The fast performance out of Clang for the Monte Carlo and Fast Fourier Transform sub-tests is what led to SciMark 2's high composite score for the LLVM-based compiler.
With the Sparse Matrix Multiply and Dense LU Matrix Factorization, the performance between Clang 3.4 and GCC 4.8.2 was very close.
Clang finished out the SciMark 2 tests with another win, this time for the Jacobi Successive Over-Relaxation sub-test.