When compiled with VML support, the number of threads is set to 1
for VML core, and to the number of cores for the native pthreads
implementation. This leads to better performance. Closes#39.
The ``set_num_threads`` now returns the number of previous thread
setting, as stated in the docstrings.
This are special registers in VM because they need to be unique for each
thread. Creating them in initialization time is a big loss of space,
because the NumExpr instances are cached for speed.
We are loosing a bit of performance with this, but I think the memory
savings pay for this.
The problem was that the internal buffers in expressions need to be refreshed when the
number of threads changes. I delete now the cache for expression each time the number of
threads is changed.
The bug.py has been migrated to a test case in the test suite.
Added `print_versions()` function for easily query the versions of software
that numexpr relies on.
Documentation updated accordingly. Announcement updated as well and
almost ready for release 1.2.
Francesc Alted