cry/Hyprland
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

protocols: add Fifo-v1 and commit-timing-v1 (#12052)

Browse source 
* protocols: add Fifo-v1

introduce fifo-v1

* fifo: only present locked surfaces

dont present to unlocked surfaces and commit pending states from the
fifo protocol.

* fifo: cformat

cformat

* protocols: add committiming and surface state queue

introduce CSurfaceStateQueue and commit-timing-v1

* fifo: schedule a frame if waiting on barrier

if we are waiting on a barrier the state doesnt commit until the next
refresh cycle meaning the monitor might have no pending damage and we
never get onPresented to unlock the barrier, moment 22. so schedule a
frame.

* fifo: properly check monitor intersection

check for m_enteredoutputs or monitor intersection if client hasnt bound
one yet, and dont fifo lock it until the surface is mapped.

* buffer: try to merge states before committing them

try to merge states before committing them meaning way less churn and
surface commits if a surface sends multiple small ones while we wait for
buffer readyness from either fifo locks or simply fences.

* buffer: dont commit states past the buffer

certain changes are relative to the buffer attached, cant go beyond it
and apply those onto the next buffer.

* buffer: set the lockmask directly

cant use .lock since the state hasnt been queued yet, set the lockmask
directly when exporting buffer fence.

* fifo: dont fifo lock on tearing

dont fifo lock on tearing.

* buffer: queue the state directly

queue the state directly and use the .lock function instead of directly
modify the lockMask on the state.

* buffer: revert creating texture at commit time

fifo barriers introduces such long wait that upon commit time a
race happends with current xdg configure implentation that the buffer
and image is actually destroyed when entering commitState, doing it at
buffer creation time with EGL_PRESERVED_KHR means it sticks around until
we are done. so revert 82759d4 and 32f3233 for now.

* buffer: rename enum and lockreasons

eLockReason and LOCK_REASON_NONE.

* fifo: workaround direct scanout lock

workaround cursor commits causing fifo to get forever locked, this
entire thing needs to be worked out.
This commit is contained in:
Tom Englund 2025-11-06 14:25:49 +01:00 • committed by GitHub
parent c757fd375c
commit 8e8bfbb0b1
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
26 changed files with 750 additions and 120 deletions
Download patch file Download diff file Expand all files Collapse all files

80
src/protocols/core/Compositor.cpp
View file

@ -91,10 +91,12 @@ CWLSurfaceResource::CWLSurfaceResource(SP<CWlSurface> resource_) : m_resource(re
if (buf && buf->m_buffer) {
m_pending.buffer = CHLBufferReference(buf->m_buffer.lock());
m_pending.texture = buf->m_buffer->m_texture;
m_pending.size = buf->m_buffer->size;
m_pending.bufferSize = buf->m_buffer->size;
} else {
m_pending.buffer = {};
m_pending.buffer = {};
m_pending.texture.reset();
m_pending.size = Vector2D{};
m_pending.bufferSize = Vector2D{};
}
@ -134,23 +136,33 @@ CWLSurfaceResource::CWLSurfaceResource(SP<CWlSurface> resource_) : m_resource(re
commitState(m_pending);
// remove any pending states.
while (!m_pendingStates.empty()) {
m_pendingStates.pop();
}
m_pendingWaiting = false;
m_stateQueue.clear();
m_pending.reset();
return;
}
// save state while we wait for buffer to become ready to read
const auto& state = m_pendingStates.emplace(makeUnique<SSurfaceState>(m_pending));
// save state while we wait for buffer to become ready
auto state = m_stateQueue.enqueue(makeUnique<SSurfaceState>(m_pending));
m_pending.reset();
if (!m_pendingWaiting) {
m_pendingWaiting = true;
scheduleState(state);
// fifo and fences first
m_events.stateCommit.emit(state);
if (state->buffer && state->buffer->type() == Aquamarine::BUFFER_TYPE_DMABUF && state->buffer->dmabuf().success && !state->updated.bits.acquire) {
state->buffer->m_syncFd = dc<CDMABuffer*>(state->buffer.m_buffer.get())->exportSyncFile();
if (state->buffer->m_syncFd.isValid())
m_stateQueue.lock(state, LOCK_REASON_FENCE);
}
// now for timer.
m_events.stateCommit2.emit(state);
if (state->rejected) {
m_stateQueue.dropState(state);
return;
}
scheduleState(state);
});
m_resource->setDamage([this](CWlSurface* r, int32_t x, int32_t y, int32_t w, int32_t h) {
@ -479,43 +491,25 @@ CBox CWLSurfaceResource::extends() {
}
void CWLSurfaceResource::scheduleState(WP<SSurfaceState> state) {
auto whenReadable = [this, surf = m_self, state] {
if (!surf || state.expired() || m_pendingStates.empty())
auto whenReadable = [this, surf = m_self](auto state, auto reason) {
if (!surf || !state)
return;
while (!m_pendingStates.empty() && m_pendingStates.front() != state) {
commitState(*m_pendingStates.front());
m_pendingStates.pop();
}
commitState(*m_pendingStates.front());
m_pendingStates.pop();
// If more states are queued, schedule next state
if (!m_pendingStates.empty()) {
scheduleState(m_pendingStates.front());
} else {
m_pendingWaiting = false;
}
m_stateQueue.unlock(state, reason);
};
if (state->updated.bits.acquire) {
// wait on acquire point for this surface, from explicit sync protocol
state->acquire.addWaiter(std::move(whenReadable));
state->acquire.addWaiter([state, whenReadable]() { whenReadable(state, LOCK_REASON_FENCE); });
} else if (state->buffer && state->buffer->isSynchronous()) {
// synchronous (shm) buffers can be read immediately
whenReadable();
} else if (state->buffer && state->buffer->type() == Aquamarine::BUFFER_TYPE_DMABUF && state->buffer->dmabuf().success) {
m_stateQueue.unlock(state);
} else if (state->buffer && state->buffer->m_syncFd.isValid()) {
// async buffer and is dmabuf, then we can wait on implicit fences
auto syncFd = dc<CDMABuffer*>(state->buffer.m_buffer.get())->exportSyncFile();
if (syncFd.isValid())
g_pEventLoopManager->doOnReadable(std::move(syncFd), std::move(whenReadable));
else
whenReadable();
g_pEventLoopManager->doOnReadable(std::move(state->buffer->m_syncFd), [state, whenReadable]() { whenReadable(state, LOCK_REASON_FENCE); });
} else {
// state commit without a buffer.
whenReadable();
m_stateQueue.unlock(state);
}
}
@ -526,8 +520,6 @@ void CWLSurfaceResource::commitState(SSurfaceState& state) {
if (m_current.buffer) {
if (m_current.buffer->isSynchronous())
m_current.updateSynchronousTexture(lastTexture);
else if (!m_current.buffer->isSynchronous() && state.updated.bits.buffer) // only get a new texture when a new buffer arrived
m_current.texture = m_current.buffer->createTexture();
// if the surface is a cursor, update the shm buffer
// TODO: don't update the entire texture
@ -558,6 +550,13 @@ void CWLSurfaceResource::commitState(SSurfaceState& state) {
nullptr);
}
if (m_current.updated.bits.damage) {
// damage is always relative to the current commit
m_current.updated.bits.damage = false;
m_current.damage.clear();
m_current.bufferDamage.clear();
}
// release the buffer if it's synchronous (SHM) as updateSynchronousTexture() has copied the buffer data to a GPU tex
// if it doesn't have a role, we can't release it yet, in case it gets turned into a cursor.
if (m_current.buffer && m_current.buffer->isSynchronous() && m_role->role() != SURFACE_ROLE_UNASSIGNED)
@ -667,7 +666,8 @@ CWLCompositorResource::CWLCompositorResource(SP<CWlCompositor> resource_) : m_re
return;
}
RESOURCE->m_self = RESOURCE;
RESOURCE->m_self = RESOURCE;
RESOURCE->m_stateQueue = CSurfaceStateQueue(RESOURCE);
LOGM(LOG, "New wl_surface with id {} at {:x}", id, (uintptr_t)RESOURCE.get());