Arithmetic broadcasting in Julia 5x slower than MATLAB

General Usage Performance
2

Is matlab automatically threading this? You can probably make this faster by using LoopVectorization. I see the following at 5x faster than your original f!

function f2!(D,A,B,C)
	@turbo D .= max.(A.*B.+C, 0.2)
end

and using @tturbo (threaded version of @turbo) gives another 4x speedup.

function f2!(D,A,B,C)
	@tturbo D .= max.(A.*B.+C, 0.2)
end
7 Likes