Show HN: CPU Mflops in 6 languages

Some numbers for latest gcc vs clang

    gcc-5-O3:   			MFLOPS(4)       =  3607.9350
    gcc-6-O3:   			MFLOPS(4)       =  3781.8053
    gcc-6-Os:   			MFLOPS(4)       =  1655.9644
    gcc-7-O3:   			MFLOPS(4)       =  3909.7981
    gcc-7-Os:   			MFLOPS(4)       =  1700.6888
    gcc-7-Os-native-align16:   	        MFLOPS(4)       =  2911.5069
    gcc-7-O3-native:                    MFLOPS(4)       =  6305.7652
    gcc-7-O3-native-align4:             MFLOPS(4)       =  6428.9799
    gcc-7-O3-native-align8:             MFLOPS(4)       =  6253.1456
    gcc-7-O3-native-align16:            MFLOPS(4)       =  6669.3797 (best)

    clang-5-O3:   			MFLOPS(4)       =  3967.8234
    clang-5-Os:   			MFLOPS(4)       =  4003.6235
    clang-5-O3-native:   		MFLOPS(4)       =  4091.8200
    clang-6-O3:   			MFLOPS(4)       =  3991.1970
    clang-6-Os:   			MFLOPS(4)       =  3988.9485
    clang-6-O3-native:   		MFLOPS(4)       =  4049.5218
    clang-6-Os-native:   		MFLOPS(4)       =  4182.5567 (best)
    clang-7-O3:  		 	MFLOPS(4)       =  3810.0356
    clang-7-Os:   			MFLOPS(4)       =  3932.1634
    clang-7-O3-native:   		MFLOPS(4)       =  4114.0933
    clang-7-Os-native:   		MFLOPS(4)       =  3984.8512

    gcc-native: -O3 -Wa,-q -march=native
    gcc-native-alignN: -O3 -Wa,-q -falign-loops=N -falign-functions=N -march=native

Biggest impact: SSE over O3/Os over alignment, clang's -Os is sometimes faster than -O3. Could not find clang's alignment tuning. This is on an haswell optimized libc: darwin. linux glibc would be much slower.