why knnsearch function in matlab2016a and 2020b shows different efficiency？

like this code，in matlab 2020b t1=10s in matlab 2016 a t1=70s，why？is 2020b knnsearch applies parallel programing？

for M=1500*1500:5000:1500*1500
rate_arr_matlab_cpp8t=[];
rate_arr_cpp_cpp8t=[];
a=(randi([-50000,58020],M,2));
b=(randi([-50000,58020],M,2));
tic
[index,dist_min]=knnsearch(a,b,'k',15);
t1=toc;
matlab=[matlab,t3];
end

[Walter Roberson]

There have been a number of performance improvements in MATLAB between R2016 and R2020b.

[俊凯 王]

is there apply parallel computing like openmp or something else in 2020b knnsearch function？

Answer 1

[Walter Roberson]

R2019a release notes specifically mention performance improvements for knnsearch .

[俊凯 王]

let me see see

[俊凯 王]

thanks

[俊凯 王]

where is the link about performance improvements for knnsearch

[俊凯 王]

i can't find detail reference in R2019a release notes ,could you show me the detail link ?

[Walter Roberson]

https://www.mathworks.com/help/stats/release-notes.html?category=machine_learning_algorithms&rntext=&startrelease=R2019a&endrelease=R2019a&groupby=release&sortby=descending&searchHighlight=