It calls fib(N-2) twice ( fib(3)), fib(N-3) three ( fib(4)) times, fib(N-4) five ( fib(5)) times, fib(N-5) eight times, …, fib(1) a total of fib(N + 1 - 1) times. (Julia packages ZMQ and JSON are required for the interface to work, as stated in the Wolfram documentation.)įib(N) is slow because we have an exponential number of function calls. Only 8 milliseconds was used (including FFI latency), more than 300 times faster than the pure Mathematica version.įinally, the Julia session is closed by DeleteObject (* defined Julia function callable from Mathematica *)ĪbsoluteTiming] (* Actual calculation *) Now, using the Julia interface, juliaSession = StartExternalSession (* Start Julia session *)įibJulia = ExternalEvaluate[juliaSession, Here’s a short example, using Julia to speed up calculation of Fibonacci numbers, using an inefficient recursive algorithm just for illustration.įirst, here’s the pure Mathematica implementation for calculating the 30th Fibonacci number, fib := If + fib] Ģ.7 seconds was used to get the result 832040.
I had a try and found the interface quite convenient.
Mathematica added a built-in Julia interface in version 12 last year.
0 kommentar(er)
