Fallback to CPU execution if RSoV compilation fails
Bug: 30964317
With a few additional changes:
* Cleaned up the RSoV script initialization code to get rid of
memory leaks;
* Updated the RSoVAllocation destructor to destroy the buffer but
not the uninitialized image.
* Updated allocation initialization code to correctly calculate
allocation size when rows are not aligned to the default minimum
alignment (16 bytes).
* Added a utility function to check if a script is completely backed
by CPU, and updated the CPU context lookup callback function to
use this utility to correctly return a CPU driver-specific script
for a HAL script.
Test: RSoVTest and RsTest
Change-Id: If85b8f01bde3190a9397aaedce8b9808d09716e6
6 files changed