[idd] select linear RGB24 packing automatically

Replace the texture-shaped RGB24 output with a linear raw buffer and
keep its packing, damage translation, and buffer copies inside the
RGB24 effect.

Benchmark full-frame native and packed processing and retain the faster
path for each source format. Preserve logical damage rectangles for
client updates and alternating framebuffer repair.

Return compute outputs to COMMON for COPY queue handoff and refresh
cached framebuffer sizes when switching packed and native layouts.
This commit is contained in:
Geoffrey McRae
2026-08-03 20:35:49 +10:00
parent fa2769b332
commit 6109501eed
16 changed files with 987 additions and 232 deletions

View File

@@ -63,9 +63,9 @@ private:
DWORD m_lastShapeId = 0;
std::atomic<UINT> m_sdrWhiteLevel { KVMFR_SDR_WHITE_LEVEL_DEFAULT };
// Output-space damage from the previous published frame. The shared-memory
// frame buffers alternate, so this must be copied along with the current
// damage to bring the older target buffer up to date.
// Logical output-space damage from the previous published frame. The
// shared-memory frame buffers alternate, so this must be copied along with
// the current damage to bring the older target buffer up to date.
RECT m_dirtyRects[LG_MAX_DIRTY_RECTS] = {};
unsigned m_nbDirtyRects = 0;