This is an enhancement request.
Is it possible for numexpr to store result in place, similar to numpy's "out"
in some operations, example: "np.add(ar1,ar2,out=ar1)". This way, there is no
temporary memory consumed during numexpr evaluation.
Desired example:
ne.evaluate("ar1 % 5",out=ar1)
I'm expecting "VmPeak" not to increase before the execution and after since
result is stored in-place.
In very large array calculation, where the array size is almost equivalent to
the physical memory of machine, then this enhancement would greatly help.
Original issue reported on code.google.com by johnqdel...@gmail.com on 22 May 2013 at 5:16
Original issue reported on code.google.com by
johnqdel...@gmail.com
on 22 May 2013 at 5:16