amdgpu: reduce amdgpu_winsys_bo size
This MR reduces struct amdgpu_winsys_bo
' size: 208 bytes -> 168 bytes.
This improves performance in some SPECviewperf13 (tested on top of !7078 (merged)). For instance in subtest10, fps
change is: 230 -> 280.
A sysprof capture shows that the percentage of total samples in amdgpu_cs_add_buffer
decreases from 23% to 16%.
The LIST_FOR_EACH_ENTRY_PTR
from the 1st commit is a bit ugly... let me know if you prefer to move this to list.h
.