Push HWC cache slot generation down into CompositionEngine

Stop caching buffers inside SurfaceFlinger and remove the tight coupling
between SurfaceFlinger's ClientCache and the Hardware Composer cache.
This allows a better seperation of responsibility, where buffer cache
management is not split between HwcSlotGenerator and HwcBufferCache, but
is instead solely handled by HwcBufferCache.

Note that FramebufferSurface and VirtualDisplaySurface no longer use
HwcBufferCache, but instead use their own cache slot management that
is solely based on BufferQueue slot numbers.

Also do minor refactoring in FramebufferSurface to simplify code.

Bug: 258196272
Test: started and stopped multiple YouTube videos on adt4 and verified
no change in graphic buffer usage
Test: atest HwcBufferCacheTest
Test: atest OutputPrepareTest

Change-Id: Ica7955ab4bc70e3c70207390e36dff73a2fc4949

1 files changed, 8 insertions, 6 deletions

diff --git a/services/surfaceflinger/SurfaceFlinger.cpp b/services/surfaceflinger/SurfaceFlinger.cpp
index bd3c0f12cc..a7b59a16ed 100644
--- a/services/surfaceflinger/SurfaceFlinger.cpp
+++ b/services/surfaceflinger/SurfaceFlinger.cpp
@@ -112,6 +112,7 @@
 #include <ui/DisplayIdentification.h>
 #include "BackgroundExecutor.h"
 #include "Client.h"
+#include "ClientCache.h"
 #include "Colorizer.h"
 #include "Display/DisplayMap.h"
 #include "DisplayDevice.h"
@@ -2255,6 +2256,8 @@ void SurfaceFlinger::composite(TimePoint frameTime, VsyncId vsyncId)
         });
     }
 
+    refreshArgs.bufferIdsToUncache = std::move(mBufferIdsToUncache);
+
     refreshArgs.layersWithQueuedFrames.reserve(mLayersWithQueuedFrames.size());
     for (auto layer : mLayersWithQueuedFrames) {
         if (auto layerFE = layer->getCompositionEngineLayerFE())
@@ -3999,10 +4002,6 @@ status_t SurfaceFlinger::setTransactionState(
                     getExternalTextureFromBufferData(*resolvedState.state.bufferData,
                                                      layerName.c_str(), transactionId);
             mBufferCountTracker.increment(resolvedState.state.surface->localBinder());
-            if (layer) {
-                resolvedState.hwcBufferSlot =
-                        layer->getHwcCacheSlot(resolvedState.state.bufferData->cachedBuffer);
-            }
         }
     }
 
@@ -4078,7 +4077,10 @@ bool SurfaceFlinger::applyTransactionState(const FrameTimelineInfo& frameTimelin
     }
 
     if (uncacheBuffer.isValid()) {
-        ClientCache::getInstance().erase(uncacheBuffer);
+        sp<GraphicBuffer> buffer = ClientCache::getInstance().erase(uncacheBuffer);
+        if (buffer != nullptr) {
+            mBufferIdsToUncache.push_back(buffer->getId());
+        }
     }
 
     // If a synchronous transaction is explicitly requested without any changes, force a transaction
@@ -4469,7 +4471,7 @@ uint32_t SurfaceFlinger::setClientStateLocked(const FrameTimelineInfo& frameTime
     if (what & layer_state_t::eBufferChanged) {
         if (layer->setBuffer(composerState.externalTexture, *s.bufferData, postTime,
                              desiredPresentTime, isAutoTimestamp, dequeueBufferTimestamp,
-                             frameTimelineInfo, composerState.hwcBufferSlot)) {
+                             frameTimelineInfo)) {
             flags |= eTraversalNeeded;
         }
     } else if (frameTimelineInfo.vsyncId != FrameTimelineInfo::INVALID_VSYNC_ID) {