Nima Sarajpoor

@YannCabanes > we can read: > "Without "parallel=True" in the jit-decorator, the prange statement is equivalent to range." Nice catch! I totally overlooked this part. > As we do not...

> @NimaSarajpoor can you make sure that we do not degrade performance by doing so? I will run the performance again after changing `prange` to `range`. @YannCabanes Thanks for your...

@YannCabanes Not yet. I was busy with resolving issue #413 as I couldn't run pytest. I solved the issue and I will resume the investigation in a few days.

So, I removed the redundant inner `prange` and I also pass `parallel=True` to decorator `@njit` where appropriate (i.e. where we have a `prange`). I am providing the performances below. **btw,...

@rtavenar @GillesVandewiele @YannCabanes FYI: (1) It seems a `comma` was missing in `tslearn/metrics/__init__.py`! (2) ALL tests in `tslearn/tests` are passing in my local repo.

I think it is not ready. I just reviewed the changes and I noticed that I added `parallel=True` in function `def _njit_itakura_mask` but I forgot to use `prange` in the...

> When we write "parallel=True", is it possible to be sure that the computations are really performed in parallel and to know how many processors are running in parallel? >...

Thanks for the info. Yeah. I do a similar thing. I posted the performance code in one of the previous comments in which you can see the approach I took...

@GillesVandewiele @rtavenar I think I am getting the same issue #413

@rtavenar @GillesVandewiele After banging my head several times to a wall, I think the following steps helped me solve the issue! ``` $pip install pytest-cython $CC=gcc python setup.py build_ext --inplace...