I'm trying to record how memory usage is modified while new objects are created.
The easiest scenario is to create a numpy array and see how does it changes. So my goal is to measure memory status before and after and see how it changes. However, I'm getting the following:
>>> print(psutil.virtual_memory()); a=np.zeros(1024*1024*10, dtype=np.uint8);print(psutil.virtual_memory())
svmem(total=16482779136, available=7692648448, percent=53.3, used=7079854080, free=2953998336, active=9224863744, inactive=3692097536, buffers=516026368, cached=5932900352, shared=1366462464, slab=387022848)
svmem(total=16482779136, available=7692648448, percent=53.3, used=7079854080, free=2953998336, active=9224863744, inactive=3692097536, buffers=516026368, cached=5932900352, shared=1366462464, slab=387022848)
But before and after creating the numpy array, the memory status is exactly the same.
This is a toy example where I was expecting to see a difference of 1010241024=10MB of memory, consumed by the numpy array a, but they are the same.
I'm running it on Ubuntu 18.04 LTS, Python 3.7.7 and psutil 5.8.0.
Thanks in advance.
Memory allocation of Python is smart. Keep memory in advance as heap, and assign what they need. Lacking heap, keep memory again. I said like that roughly. In other words, general method can't measure change of memory. Therefore, there is such a way. You must be exciting? Enjoyment to deal with raw memory. Sadness of fragment pointers. Great!