sparse() is really slow

[ViralBShah]

On the most simple case, julia sparse() is 20 times slower than matlab. I think the code can be written in a better way for sure, but I'm not sure if it will give us 20x improvement. This does not even include the cases where the inputs have to be sorted.

Matlab:

a = 1:100000;
tic; s = sparse(a,a,a); toc
Elapsed time is 0.010585 seconds.

Julia:

julia> x = 1:100000;
julia> tic(); sparse(x,x,x); toc();
elapsed time: 0.19214701652526855 seconds

[JeffBezanson]

Matlab's sparse takes longer if I and J are random; I wonder if an issorted() check would help. And there are many other operations in there too. This will have to be broken down.

[ViralBShah]

If I use x = [1:100000], then we are faster than matlab as of commit c9c671b. The ranges have to eventually be expanded for storing in the sparse matrix data structure, and hence the version with ranges will be slower.

julia> tic(); s = sparse(x,x,x,100000,100000); toc();
elapsed time: 0.0044701099395752 seconds

[JeffBezanson]

I improved integer min and max, and now we don't need the 4th and 5th arguments to sparse to get that performance.
Pretty amazing to actually get more than a 20x speedup and now beat them :)