I think you are right, it's the assignment itself which is slow. Merged loop is only a tad quicker.
On Thursday, September 8, 2016 at 6:04:41 PM UTC+1, Christian Gollwitzer wrote: > > Why nested loops are so slow in Python? Is it because new contexts are > > created? > > For more details, see > > http://stackoverflow.com/questions/26611043/numpy-vs-cython-nested-loop-so-slow > > http://stackoverflow.com/questions/39371021/efficient-loop-over-numpy-array > > The answers are right there, too - what are you mising? CPython is an > interpreter. For every seemingly simple operation like a[i]=j, a > function pointer is called, the variables contain polymorphic types > which are morphed, etc. The same thing in compiled code is a single > machine instruction. Speed differences of ~100x are normal between > compiled and interpreted code over all dynamic languages. > > Christian -- https://mail.python.org/mailman/listinfo/python-list
