I saw this paper recently: w20263.pdf 262.05 KB The codes are here: https://github.com/jesusfv/Comparison-Programming-Languages-Economics I run the Fortran code and Julia code on my machine, my results are consistent with the findings of this paper. Howeve…

Dear all I am Jesus Fernandez-Villaverde, one of the authors of the paper discussed here. The published paper is here: https://www.sciencedirect.com/science/article/pii/S0165188915000883 It has some changes with respect to the working paper that some of you are discussing. Nothing of importance t…

On my computer, both the original Julia code and your improved version are faster than the C and Fortran code compiled with GCC. However, a version I previously compiled with the Intel compiler is considerably faster than both (0.9s Intel vs 1.6s Julia). I no longer have the Intel compiler installe…

Once again I’m focusing on Python Numba vs Julia :slight_smile: I’m surprised by the results, however… Python Numba 1.3s Julia 1.94s Julia Libm 1.53s Julia Libm+Inbound 1.55s (no noticeable difference) Details are here: Julia notebook Python notebook

There is this line maxDifference = maximum(abs.(mValueFunctionNew-mValueFunction)) This allocates memory just to find the maximum. An alternative would be reduce((x_max, x) -> max(x_max, abs(x[1] - x[2])), 0.0, zip(mValueFunctionNew, mValueFunction)) and here valueProvisional = (1-β)*Base.M…

We can optimize it and make it crazily fast but the point is that a general economist would likely write the code just like the original version and would not know how (or don’t want to spend time on) optimizing it. [image] saschatimme: reduce((x_max, x) -> max(x_max, abs(x[1] - x[2])), 0.0, zi…

maximum accepts a function as first argument for these cases. For example, to compute the maximum absolute value of a vector v without extra allocations one can do: maximum(abs, v) Or with generators: maximum(abs(x) for x in v) To get the absolute value of the difference between two arrays we ne…

Interesting. I may try the next time I’m on an Intel computer. (Although icc/ifort is still pretty fast on AMD.) Mostly though, I’d want Julia to beat Numba. It surprises me how much faster Julia is than code compiled with Clang. It’s also much faster than code compiled with Flang, but that’s a mu…

I am getting 1.7 seconds with your code, making it much slower than the rest.

[image] Elrod: I am getting 1.7 seconds with your code, making it much slower than the rest. which code are you referring to?

Numba. I didn’t run it IPython, but from the interpreter: start = time.time() main_func2() end = time.time() print(end - start) #1.7 This was the same computer as my other benchmarks, where the original Julia ran in 1.25s. This is from a fresh install of Anaconda for Python 3.6.

A Comparison of Programming Languages in Economics

Community

StefanKarpinski February 11, 2018, 9:35pm 23

These really should be as fast as the loop version so there’s some missed optimization opportunity here. Might be worth opening an issue about it.

Topic		Replies	Views
Optimize benchmark-code Performance question	11	840	December 10, 2019
Julia slower than Matlab & Python? No Performance economics , tensorflow , matlab , pytorch	119	18604	March 16, 2021
Nice blogpost comparing Julia w/ 6 other languages to solve a Logistic DE New to Julia	24	1777	May 26, 2022
This post claims that Julia is still 4-5 times slower than cython Performance	38	5003	February 21, 2022
Choosing a numerical programming language for economic research: Julia, Community blog , blog-post	135	7961	May 30, 2023

A Comparison of Programming Languages in Economics

Related topics