[Pixman] How pixman_implementation_t is getting free

Andrea Canciani ranma42 at gmail.com
Tue Nov 2 04:38:56 PDT 2010

2010/11/2 kb pachauri <kb.pachauri at gmail.com>:
> Actually what i mean for freeing memory is .. freeing the opencl context,
> releasing command queue, all opencl kernels etc...
> sorry for not too clear in my words..
> I dont think this will be freed until i free the context..

Then you might look this thread:

Both Siarhei and I have been working towards thread-safe and valgrind-clean
pixman, but nothing to handle it is in master yet.

Siarhei directly hooks on compiler attributes that allow a function to be
called upon initialization or termination. He only did it for the constructor,
but the destructor can be implemented with the same approach.

My wip/simpleops branch provides a pixman_fini which is (supposed to be)
automatically called upon library termination, so any deallocation for
pixman implementations would go there.

Please notice that my wip/simpleops adds a dependency to the simpleops
library, which is currently under development and not released yet and
(being in very early stages of development) afaik only tested on my laptop
(so it probably won't work on win32 and might have problems with
non-gcc compilers).


More information about the Pixman mailing list