Well, I am sure that these algorithms can be parallelized, using 28.io,
or other vendor solutions.
However, honestly, today, you would have to spend tons of processors,
watts, intelligence, to perform a matrix multiplication over a 4000x4000
matrix with JSONiq or XQUERY 3.0.
I can compute such a multiplication within a second using BLAS over a
single threaded process.
I don't think anyone(*) is suggesting that XQuery or JSONiq should be
anyone's number 1 choice for doing a matrix multiplication. The key thing
is to make the language powerful enough so that if you have an application
doing 300 tasks, the fact that one of them involves matrix mutliplication
shouldn't stop you.
Just like XSLT isn't one's obvious choice of graphics programming
language, but it's sure handy that you can use it to generate an SVG
histogram or pie chart.
(*) well, I exclude the XQuery fanatics from this....