I also did a little benchmarks. Here is the results: Time for multiplying two 1000x1000 matrics: clang 2.8 (-O3) 2.4s gcc 4.4.5 (-O3) 2.5s ldc2 llvm-2.9 (-O3 -release) 2.7s gdc2 4.4.6 (-O3 -frelease -finline) 2.9s dmd 2.052 (-O -release -inline) 4.9s 1500x1500: