Clang Fights GCC On AMD's Athlon AM1 APU With Jaguar Cores

Written by Michael Larabel in Software on 22 April 2014 at 02:00 AM EDT. Page 2 of 4. 9 Comments.
AMD Athlon 5350 - GCC 4.8 vs. LLVM Clang 3.4 Compilers
AMD Athlon 5350 - GCC 4.8 vs. LLVM Clang 3.4 Compilers

The SciMark computational performance between GCC 4.8.2 and LLVM Clang 3.4 started out at roughly the same speed...

AMD Athlon 5350 - GCC 4.8 vs. LLVM Clang 3.4 Compilers
AMD Athlon 5350 - GCC 4.8 vs. LLVM Clang 3.4 Compilers
AMD Athlon 5350 - GCC 4.8 vs. LLVM Clang 3.4 Compilers

However, in the final SciMark tests of the Sparse Matrix Multiply, Dense LU Matrix Factorization, and Jacobi Successive Over-Relaxation sub-tests, Clang 3.4 speeded ahead of GCC 4.8.2. There were noticeable performance advantages to LLVM's Clang on this AM1 Kabini APU for these workloads over GCC.

AMD Athlon 5350 - GCC 4.8 vs. LLVM Clang 3.4 Compilers

One item to just to point out for those that may be new to the LLVM/Clang world... This compiler still sadly lacks support for OpenMP. Thus applications that utilize OpenMP for multi-threading are left slower on Clang until the OpenMP support is merged. There's out-of-tree LLVM/Clang OpenMP support but nothing that's in the mainline code as of right now.


Related Articles