Julia is regarded as one of most promising scientific programming languages recently. Its syntax is similar to Matlab and performs well on numerical computing.
Since Julia also has JIT technique, it is faster than Python+Numpy and only 2~3x slower than C, if everything is programmed in a decent way.
I was lucky to know this language when it was still at baby step, and I occasionally found a benchmark on the performance over Matlab which uses while loops and for loops everywhere, however, it still showed its capability on computing. It is also worthy to observe the summation for loops with pisum in its benchmark, it has a comparable speed with -O3 switched-on C program, which means Matlab already has a JIT to optimize for maybe nested for.
Like Matlab’s MEX interface, Julia also can use ccall for calling from shared library, which would be a easy hack on porting codes to Julia, but as far as I know, only C base types including struct can be converted forward and backward, C++ still is not supported officially.
For packages, community has contributed a lot, especially in optimization and data science, statistics and so on, but unfortunately, a large portion of the existing packages do not have sufficient support and maintenance.
We quoted from Cpp.jl
Julia can call C code with no overhead, but it does not natively support C++. However, the C++ ABI is essentially “C plus some extra conventions,” of which the most noteworthy is name mangling. Name mangling is used to support function overloading, a key C++ (and Julia) feature. Infamously, different compilers use different mangling conventions, and this has lead to more than a few headaches. However, in recent years there has been a greater push for standardization of the C++ ABI, and there is good documentation available on calling conventions of different compilers.
In order to get C++ codes working with Matlab as well as Julia, we probably have to use Cpp.jl to generate mangled names or simply declaring extern "C" in order to disable name mangling.
#ifdef __cplusplus
extern "C"
#endif
void some_function(...);#include "header.h"
void some_function(...) {
//C++ codes here.
}
compile with gcc but linking against C++ library as lstdc++, or simply use g++.
gcc -c -fPIC libsome_func.o
gcc -shared -o libsome_func.so -lstdc++