LineageOS/android_vendor_qcom_opensource_display-commonsys-intf
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

hwc: cache layers on framebuffer

Browse Source 
In the cases where all the layers that are marked as HWC_FRAMEBUFFER did not
change, we can mark them as HWC_OVERLAY since their contents are already on the
framebuffer and do not need any additional processing.

This optimization shows power improvement for video playback.

Bug: 7623741
Change-Id: Ia178a926e6f56a3ec9291250a22a66f212c30b14
Signed-off-by: Iliyan Malchev <malchev@google.com>
This commit is contained in:
Naseer Ahmed
2012-10-17 00:32:50 -04:00
parent 09549f6d75
commit b1c7632ec5
3 changed files with 109 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

73
libhwcomposer/hwc_utils.cpp
View File

@@ -60,6 +60,7 @@ void initContext(hwc_context_t *ctx)
ctx->mMDP.hasOverlay = qdutils::MDPVersion::getInstance().hasOverlay();
ctx->mMDP.panel = qdutils::MDPVersion::getInstance().getPanelType();
ctx->mExtDisplay = new ExternalDisplay(ctx);
ctx->mLayerCache = new LayerCache();
MDPComp::init(ctx);
pthread_mutex_init(&(ctx->vstate.lock), NULL);
@@ -208,7 +209,7 @@ int hwc_sync(hwc_context_t *ctx, hwc_display_contents_1_t* list, int dpy) {
int ret = 0;
#ifdef USE_FENCE_SYNC
struct mdp_buf_sync data;
int acquireFd[4];
int acquireFd[MAX_NUM_LAYERS];
int count = 0;
int releaseFd = -1;
int fbFd = -1;
@@ -219,7 +220,8 @@ int hwc_sync(hwc_context_t *ctx, hwc_display_contents_1_t* list, int dpy) {
for(uint32_t i = 0; i < list->numHwLayers; i++) {
if((list->hwLayers[i].compositionType == HWC_OVERLAY ||
list->hwLayers[i].compositionType == HWC_FRAMEBUFFER_TARGET) &&
list->hwLayers[i].acquireFenceFd != -1) {
list->hwLayers[i].acquireFenceFd != -1 &&
(list->hwLayers[i].flags & HWC_MDPCOMP)) {
acquireFd[count++] = list->hwLayers[i].acquireFenceFd;
}
}
@@ -251,4 +253,71 @@ int hwc_sync(hwc_context_t *ctx, hwc_display_contents_1_t* list, int dpy) {
return ret;
}
void LayerCache::resetLayerCache(int num) {
for(uint32_t i = 0; i < MAX_NUM_LAYERS; i++) {
hnd[i] = NULL;
}
numHwLayers = num;
}
void LayerCache::updateLayerCache(hwc_display_contents_1_t* list) {
int numFbLayers = 0;
int numCacheableLayers = 0;
canUseLayerCache = false;
//Bail if geometry changed or num of layers changed
if(list->flags & HWC_GEOMETRY_CHANGED ||
list->numHwLayers != numHwLayers ) {
resetLayerCache(list->numHwLayers);
return;
}
for(uint32_t i = 0; i < list->numHwLayers; i++) {
//Bail on skip layers
if(list->hwLayers[i].flags & HWC_SKIP_LAYER) {
resetLayerCache(list->numHwLayers);
return;
}
if(list->hwLayers[i].compositionType == HWC_FRAMEBUFFER) {
numFbLayers++;
if(hnd[i] == NULL) {
hnd[i] = list->hwLayers[i].handle;
} else if (hnd[i] ==
list->hwLayers[i].handle) {
numCacheableLayers++;
} else {
hnd[i] = NULL;
return;
}
} else {
hnd[i] = NULL;
}
}
if(numFbLayers == numCacheableLayers)
canUseLayerCache = true;
//XXX: The marking part is separate, if MDP comp wants
// to use it in the future. Right now getting MDP comp
// to use this is more trouble than it is worth.
markCachedLayersAsOverlay(list);
}
void LayerCache::markCachedLayersAsOverlay(hwc_display_contents_1_t* list) {
//This optimization only works if ALL the layer handles
//that were on the framebuffer didn't change.
if(canUseLayerCache){
for(uint32_t i = 0; i < list->numHwLayers; i++) {
if (list->hwLayers[i].handle &&
list->hwLayers[i].handle == hnd[i] &&
list->hwLayers[i].compositionType != HWC_FRAMEBUFFER_TARGET)
{
list->hwLayers[i].compositionType = HWC_OVERLAY;
}
}
}
}
};//namespace