mirror of
https://gitlab.freedesktop.org/wlroots/wlroots.git
synced 2026-06-13 14:32:57 -04:00
When uploading shm buffers to our internal backing texture, we first allocate suitable chunk from our staging buffer accessible from both CPU and GPU, and then we queue a GPU-side copy from the staging buffer to the texture. This is a lot of copying, especially on iGPUs where everything is system memory anyway. Instead, use VK_EXT_host_image_copy when available for formats that are reported as having optimal device access. This allows us to copy directly to the target texture from CPU, eliminating the queued GPU-side copy. To keep things simple we keep this texture in GENERAL for the entire duration for now. |
||
|---|---|---|
| .. | ||
| backend | ||
| interfaces | ||
| render | ||
| types | ||
| util | ||
| wlr | ||
| xcursor | ||
| xwayland | ||
| meson.build | ||