Does PIL resize() hold the GIL?

Question

Example:

image = Image.open('foo.png')
# releases the GIL?
resized = image.resize((800, 600), Image.ANTIALIAS)
# reacquires the GIL?

Obviously the variable assignment needs to hold the GIL, but it's difficult to break that up into two lines. :)

If there are two threads doing image resizes, can those resizes run on two different cores?

Answer 1

Looking at the source of 1.1.7, it doesn't appear to release the GIL for _resize.

The functions that release the GIL seem to be:

PyImaging_CreateWindowWin32 (createwindow on Win32)
PyImaging_EventLoopWin32 (eventloop on Win32)
pyCMSdoTransform (apply)
_buildTransform (buildTransform)
_buildProofTransform (buildProofTransform)
_encode_to_file (encode_to_file)

Answer 2

From the Python wiki on the GIL:

Note that potentially blocking or long-running operations, such as I/O, image processing, and NumPy number crunching, happen outside the GIL. Therefore it is only in multithreaded programs that spend a lot of time inside the GIL, interpreting CPython bytecode, that the GIL becomes a bottleneck.

PIL uses C extensions to do most of its heavy lifting. So the actual image resizing should take advantage of multi-threads if applicable.

If you are asking about resizing multiple images concurrently, I recommend looking into using Python's native multiprocessing library. This should achieve the desired effect of using multiple cores.