Working on algorithms to plot algebraic curves and surfaces (based on equations) I discovered that no matter how good the algorithm was, the main problem always will be the speed of compute/evaluate the algebraic expression.
To evaluate math expressions, Analitza does an excellent job: the API is stable, has a lots of features, is fast for evaluations, its semantics are based on a good enough standard (MathML), etc.
Although Analitza is amazing in many ways, I always felt that with the use of JIT compilation, then it would rocks harder, better and stronger: The idea would be translate the algebraic expression into machine code and evaluate the machine code instead of the Analitza version. This, indeed, would be a little hard, but for such obvious benefits: it worth the effort!
So, I used LLVM to implement the JIT compilation from Analitza expression into LLVM IR and then into machine code. Later I could use the machine code version to perform the evaluations, the results were very good and promising. I did a benchmark of 3 lambda expressions and compared their evaluations over 3 different domains ... 3 images worth more than a million of words ;)
The AnalitzaJIT module is still on heavy development, the basic bits and the foundations are there to grow up until have feature parity with the Analitza core module ... we will succeed on this goal!