Any news on this PR?

It failed on the xtest_extended_xmath.average test.

/home/travis/build/QuantStack/xtensor/test/test_extended_xmath_reducers.cpp:810: Failure
Value of: xt::allclose(av, py_av)
  Actual: false
Expected: true
[  FAILED  ] xtest_extended_xmath.average (0 ms)

Yes we have a vicious bug in release when TBB is enable and all the tests are run. (Running only the failing tests with the same build options makes them magically pass).
We have been quite busy these two last weeks with workshops and conferences, we should have more time to investigate this now.

What is the status on this? I just tested it on my machine and the performance gain for calculating the variance is about factor 100 on my machine

The status has not evolved since June, we still have some weird bug on the CI when TBB is enabled. (Now we also have some conflicts due to the recent refactoring in xtensor).

I am also interested in this, we use a lot of strided views.

Santa may be early this year? Two thumbs up for speedy views!!!

Well we still have random issues with TBB (and now OpenMP) :/