2020-05-04 04:24:59 +02:00
|
|
|
|
using OpenTK.Graphics.OpenGL;
|
|
|
|
|
using Ryujinx.Graphics.GAL;
|
|
|
|
|
using System;
|
|
|
|
|
using System.Collections.Generic;
|
|
|
|
|
using System.Threading;
|
|
|
|
|
|
|
|
|
|
namespace Ryujinx.Graphics.OpenGL.Queries
|
|
|
|
|
{
|
|
|
|
|
class CounterQueue : IDisposable
|
|
|
|
|
{
|
|
|
|
|
private const int QueryPoolInitialSize = 100;
|
|
|
|
|
|
|
|
|
|
public CounterType Type { get; }
|
|
|
|
|
public bool Disposed { get; private set; }
|
|
|
|
|
|
|
|
|
|
private Queue<CounterQueueEvent> _events = new Queue<CounterQueueEvent>();
|
|
|
|
|
private CounterQueueEvent _current;
|
|
|
|
|
|
|
|
|
|
private ulong _accumulatedCounter;
|
Add a Multithreading layer for the GAL, multi-thread shader compilation at runtime (#2501)
* Initial Implementation
About as fast as nvidia GL multithreading, can be improved with faster command queuing.
* Struct based command list
Speeds up a bit. Still a lot of time lost to resource copy.
* Do shader init while the render thread is active.
* Introduce circular span pool V1
Ideally should be able to use structs instead of references for storing these spans on commands. Will try that next.
* Refactor SpanRef some more
Use a struct to represent SpanRef, rather than a reference.
* Flush buffers on background thread
* Use a span for UpdateRenderScale.
Much faster than copying the array.
* Calculate command size using reflection
* WIP parallel shaders
* Some minor optimisation
* Only 2 max refs per command now.
The command with 3 refs is gone. :relieved:
* Don't cast on the GPU side
* Remove redundant casts, force sync on window present
* Fix Shader Cache
* Fix host shader save.
* Fixup to work with new renderer stuff
* Make command Run static, use array of delegates as lookup
Profile says this takes less time than the previous way.
* Bring up to date
* Add settings toggle. Fix Muiltithreading Off mode.
* Fix warning.
* Release tracking lock for flushes
* Fix Conditional Render fast path with threaded gal
* Make handle iteration safe when releasing the lock
This is mostly temporary.
* Attempt to set backend threading on driver
Only really works on nvidia before launching a game.
* Fix race condition with BufferModifiedRangeList, exceptions in tracking actions
* Update buffer set commands
* Some cleanup
* Only use stutter workaround when using opengl renderer non-threaded
* Add host-conditional reservation of counter events
There has always been the possibility that conditional rendering could use a query object just as it is disposed by the counter queue. This change makes it so that when the host decides to use host conditional rendering, the query object is reserved so that it cannot be deleted. Counter events can optionally start reserved, as the threaded implementation can reserve them before the backend creates them, and there would otherwise be a short amount of time where the counter queue could dispose the event before a call to reserve it could be made.
* Address Feedback
* Make counter flush tracked again.
Hopefully does not cause any issues this time.
* Wait for FlushTo on the main queue thread.
Currently assumes only one thread will want to FlushTo (in this case, the GPU thread)
* Add SDL2 headless integration
* Add HLE macro commands.
Co-authored-by: Mary <mary@mary.zone>
2021-08-27 00:31:29 +02:00
|
|
|
|
private int _waiterCount;
|
2020-05-04 04:24:59 +02:00
|
|
|
|
|
|
|
|
|
private object _lock = new object();
|
|
|
|
|
|
|
|
|
|
private Queue<BufferedQuery> _queryPool;
|
|
|
|
|
private AutoResetEvent _queuedEvent = new AutoResetEvent(false);
|
|
|
|
|
private AutoResetEvent _wakeSignal = new AutoResetEvent(false);
|
Add a Multithreading layer for the GAL, multi-thread shader compilation at runtime (#2501)
* Initial Implementation
About as fast as nvidia GL multithreading, can be improved with faster command queuing.
* Struct based command list
Speeds up a bit. Still a lot of time lost to resource copy.
* Do shader init while the render thread is active.
* Introduce circular span pool V1
Ideally should be able to use structs instead of references for storing these spans on commands. Will try that next.
* Refactor SpanRef some more
Use a struct to represent SpanRef, rather than a reference.
* Flush buffers on background thread
* Use a span for UpdateRenderScale.
Much faster than copying the array.
* Calculate command size using reflection
* WIP parallel shaders
* Some minor optimisation
* Only 2 max refs per command now.
The command with 3 refs is gone. :relieved:
* Don't cast on the GPU side
* Remove redundant casts, force sync on window present
* Fix Shader Cache
* Fix host shader save.
* Fixup to work with new renderer stuff
* Make command Run static, use array of delegates as lookup
Profile says this takes less time than the previous way.
* Bring up to date
* Add settings toggle. Fix Muiltithreading Off mode.
* Fix warning.
* Release tracking lock for flushes
* Fix Conditional Render fast path with threaded gal
* Make handle iteration safe when releasing the lock
This is mostly temporary.
* Attempt to set backend threading on driver
Only really works on nvidia before launching a game.
* Fix race condition with BufferModifiedRangeList, exceptions in tracking actions
* Update buffer set commands
* Some cleanup
* Only use stutter workaround when using opengl renderer non-threaded
* Add host-conditional reservation of counter events
There has always been the possibility that conditional rendering could use a query object just as it is disposed by the counter queue. This change makes it so that when the host decides to use host conditional rendering, the query object is reserved so that it cannot be deleted. Counter events can optionally start reserved, as the threaded implementation can reserve them before the backend creates them, and there would otherwise be a short amount of time where the counter queue could dispose the event before a call to reserve it could be made.
* Address Feedback
* Make counter flush tracked again.
Hopefully does not cause any issues this time.
* Wait for FlushTo on the main queue thread.
Currently assumes only one thread will want to FlushTo (in this case, the GPU thread)
* Add SDL2 headless integration
* Add HLE macro commands.
Co-authored-by: Mary <mary@mary.zone>
2021-08-27 00:31:29 +02:00
|
|
|
|
private AutoResetEvent _eventConsumed = new AutoResetEvent(false);
|
2020-05-04 04:24:59 +02:00
|
|
|
|
|
|
|
|
|
private Thread _consumerThread;
|
|
|
|
|
|
|
|
|
|
internal CounterQueue(CounterType type)
|
|
|
|
|
{
|
|
|
|
|
Type = type;
|
|
|
|
|
|
|
|
|
|
QueryTarget glType = GetTarget(Type);
|
|
|
|
|
|
|
|
|
|
_queryPool = new Queue<BufferedQuery>(QueryPoolInitialSize);
|
|
|
|
|
for (int i = 0; i < QueryPoolInitialSize; i++)
|
|
|
|
|
{
|
|
|
|
|
_queryPool.Enqueue(new BufferedQuery(glType));
|
|
|
|
|
}
|
|
|
|
|
|
2020-12-03 19:42:59 +01:00
|
|
|
|
_current = new CounterQueueEvent(this, glType, 0);
|
2020-05-04 04:24:59 +02:00
|
|
|
|
|
|
|
|
|
_consumerThread = new Thread(EventConsumer);
|
|
|
|
|
_consumerThread.Start();
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
private void EventConsumer()
|
|
|
|
|
{
|
|
|
|
|
while (!Disposed)
|
|
|
|
|
{
|
|
|
|
|
CounterQueueEvent evt = null;
|
|
|
|
|
lock (_lock)
|
|
|
|
|
{
|
|
|
|
|
if (_events.Count > 0)
|
|
|
|
|
{
|
|
|
|
|
evt = _events.Dequeue();
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (evt == null)
|
|
|
|
|
{
|
|
|
|
|
_queuedEvent.WaitOne(); // No more events to go through, wait for more.
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
Add a Multithreading layer for the GAL, multi-thread shader compilation at runtime (#2501)
* Initial Implementation
About as fast as nvidia GL multithreading, can be improved with faster command queuing.
* Struct based command list
Speeds up a bit. Still a lot of time lost to resource copy.
* Do shader init while the render thread is active.
* Introduce circular span pool V1
Ideally should be able to use structs instead of references for storing these spans on commands. Will try that next.
* Refactor SpanRef some more
Use a struct to represent SpanRef, rather than a reference.
* Flush buffers on background thread
* Use a span for UpdateRenderScale.
Much faster than copying the array.
* Calculate command size using reflection
* WIP parallel shaders
* Some minor optimisation
* Only 2 max refs per command now.
The command with 3 refs is gone. :relieved:
* Don't cast on the GPU side
* Remove redundant casts, force sync on window present
* Fix Shader Cache
* Fix host shader save.
* Fixup to work with new renderer stuff
* Make command Run static, use array of delegates as lookup
Profile says this takes less time than the previous way.
* Bring up to date
* Add settings toggle. Fix Muiltithreading Off mode.
* Fix warning.
* Release tracking lock for flushes
* Fix Conditional Render fast path with threaded gal
* Make handle iteration safe when releasing the lock
This is mostly temporary.
* Attempt to set backend threading on driver
Only really works on nvidia before launching a game.
* Fix race condition with BufferModifiedRangeList, exceptions in tracking actions
* Update buffer set commands
* Some cleanup
* Only use stutter workaround when using opengl renderer non-threaded
* Add host-conditional reservation of counter events
There has always been the possibility that conditional rendering could use a query object just as it is disposed by the counter queue. This change makes it so that when the host decides to use host conditional rendering, the query object is reserved so that it cannot be deleted. Counter events can optionally start reserved, as the threaded implementation can reserve them before the backend creates them, and there would otherwise be a short amount of time where the counter queue could dispose the event before a call to reserve it could be made.
* Address Feedback
* Make counter flush tracked again.
Hopefully does not cause any issues this time.
* Wait for FlushTo on the main queue thread.
Currently assumes only one thread will want to FlushTo (in this case, the GPU thread)
* Add SDL2 headless integration
* Add HLE macro commands.
Co-authored-by: Mary <mary@mary.zone>
2021-08-27 00:31:29 +02:00
|
|
|
|
// Spin-wait rather than sleeping if there are any waiters, by passing null instead of the wake signal.
|
|
|
|
|
evt.TryConsume(ref _accumulatedCounter, true, _waiterCount == 0 ? _wakeSignal : null);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if (_waiterCount > 0)
|
|
|
|
|
{
|
|
|
|
|
_eventConsumed.Set();
|
2020-05-04 04:24:59 +02:00
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
internal BufferedQuery GetQueryObject()
|
|
|
|
|
{
|
|
|
|
|
// Creating/disposing query objects on a context we're sharing with will cause issues.
|
|
|
|
|
// So instead, make a lot of query objects on the main thread and reuse them.
|
|
|
|
|
|
|
|
|
|
lock (_lock)
|
|
|
|
|
{
|
|
|
|
|
if (_queryPool.Count > 0)
|
|
|
|
|
{
|
|
|
|
|
BufferedQuery result = _queryPool.Dequeue();
|
|
|
|
|
return result;
|
|
|
|
|
}
|
|
|
|
|
else
|
|
|
|
|
{
|
|
|
|
|
return new BufferedQuery(GetTarget(Type));
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
internal void ReturnQueryObject(BufferedQuery query)
|
|
|
|
|
{
|
|
|
|
|
lock (_lock)
|
|
|
|
|
{
|
|
|
|
|
_queryPool.Enqueue(query);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
Add a Multithreading layer for the GAL, multi-thread shader compilation at runtime (#2501)
* Initial Implementation
About as fast as nvidia GL multithreading, can be improved with faster command queuing.
* Struct based command list
Speeds up a bit. Still a lot of time lost to resource copy.
* Do shader init while the render thread is active.
* Introduce circular span pool V1
Ideally should be able to use structs instead of references for storing these spans on commands. Will try that next.
* Refactor SpanRef some more
Use a struct to represent SpanRef, rather than a reference.
* Flush buffers on background thread
* Use a span for UpdateRenderScale.
Much faster than copying the array.
* Calculate command size using reflection
* WIP parallel shaders
* Some minor optimisation
* Only 2 max refs per command now.
The command with 3 refs is gone. :relieved:
* Don't cast on the GPU side
* Remove redundant casts, force sync on window present
* Fix Shader Cache
* Fix host shader save.
* Fixup to work with new renderer stuff
* Make command Run static, use array of delegates as lookup
Profile says this takes less time than the previous way.
* Bring up to date
* Add settings toggle. Fix Muiltithreading Off mode.
* Fix warning.
* Release tracking lock for flushes
* Fix Conditional Render fast path with threaded gal
* Make handle iteration safe when releasing the lock
This is mostly temporary.
* Attempt to set backend threading on driver
Only really works on nvidia before launching a game.
* Fix race condition with BufferModifiedRangeList, exceptions in tracking actions
* Update buffer set commands
* Some cleanup
* Only use stutter workaround when using opengl renderer non-threaded
* Add host-conditional reservation of counter events
There has always been the possibility that conditional rendering could use a query object just as it is disposed by the counter queue. This change makes it so that when the host decides to use host conditional rendering, the query object is reserved so that it cannot be deleted. Counter events can optionally start reserved, as the threaded implementation can reserve them before the backend creates them, and there would otherwise be a short amount of time where the counter queue could dispose the event before a call to reserve it could be made.
* Address Feedback
* Make counter flush tracked again.
Hopefully does not cause any issues this time.
* Wait for FlushTo on the main queue thread.
Currently assumes only one thread will want to FlushTo (in this case, the GPU thread)
* Add SDL2 headless integration
* Add HLE macro commands.
Co-authored-by: Mary <mary@mary.zone>
2021-08-27 00:31:29 +02:00
|
|
|
|
public CounterQueueEvent QueueReport(EventHandler<ulong> resultHandler, ulong lastDrawIndex, bool hostReserved)
|
2020-05-04 04:24:59 +02:00
|
|
|
|
{
|
|
|
|
|
CounterQueueEvent result;
|
2020-12-03 19:42:59 +01:00
|
|
|
|
ulong draws = lastDrawIndex - _current.DrawIndex;
|
2020-05-04 04:24:59 +02:00
|
|
|
|
|
|
|
|
|
lock (_lock)
|
|
|
|
|
{
|
2020-12-03 19:42:59 +01:00
|
|
|
|
// A query's result only matters if more than one draw was performed during it.
|
|
|
|
|
// Otherwise, dummy it out and return 0 immediately.
|
|
|
|
|
|
Add a Multithreading layer for the GAL, multi-thread shader compilation at runtime (#2501)
* Initial Implementation
About as fast as nvidia GL multithreading, can be improved with faster command queuing.
* Struct based command list
Speeds up a bit. Still a lot of time lost to resource copy.
* Do shader init while the render thread is active.
* Introduce circular span pool V1
Ideally should be able to use structs instead of references for storing these spans on commands. Will try that next.
* Refactor SpanRef some more
Use a struct to represent SpanRef, rather than a reference.
* Flush buffers on background thread
* Use a span for UpdateRenderScale.
Much faster than copying the array.
* Calculate command size using reflection
* WIP parallel shaders
* Some minor optimisation
* Only 2 max refs per command now.
The command with 3 refs is gone. :relieved:
* Don't cast on the GPU side
* Remove redundant casts, force sync on window present
* Fix Shader Cache
* Fix host shader save.
* Fixup to work with new renderer stuff
* Make command Run static, use array of delegates as lookup
Profile says this takes less time than the previous way.
* Bring up to date
* Add settings toggle. Fix Muiltithreading Off mode.
* Fix warning.
* Release tracking lock for flushes
* Fix Conditional Render fast path with threaded gal
* Make handle iteration safe when releasing the lock
This is mostly temporary.
* Attempt to set backend threading on driver
Only really works on nvidia before launching a game.
* Fix race condition with BufferModifiedRangeList, exceptions in tracking actions
* Update buffer set commands
* Some cleanup
* Only use stutter workaround when using opengl renderer non-threaded
* Add host-conditional reservation of counter events
There has always been the possibility that conditional rendering could use a query object just as it is disposed by the counter queue. This change makes it so that when the host decides to use host conditional rendering, the query object is reserved so that it cannot be deleted. Counter events can optionally start reserved, as the threaded implementation can reserve them before the backend creates them, and there would otherwise be a short amount of time where the counter queue could dispose the event before a call to reserve it could be made.
* Address Feedback
* Make counter flush tracked again.
Hopefully does not cause any issues this time.
* Wait for FlushTo on the main queue thread.
Currently assumes only one thread will want to FlushTo (in this case, the GPU thread)
* Add SDL2 headless integration
* Add HLE macro commands.
Co-authored-by: Mary <mary@mary.zone>
2021-08-27 00:31:29 +02:00
|
|
|
|
if (hostReserved)
|
|
|
|
|
{
|
|
|
|
|
// This counter event is guaranteed to be available for host conditional rendering.
|
|
|
|
|
_current.ReserveForHostAccess();
|
|
|
|
|
}
|
|
|
|
|
|
Fix deadlock for GPU counter report when 0 draws are done (#3019)
This fixes a rare bug where reporting a counter for a region containing 0 draws could deadlock the GPU. If this write overlaps with a tracking action, then the GPU could end up waiting on something that it's meant to do in the future, so it would just get stuck.
Before, this reported immediately and wrote the result to guest memory (tracked) from the backend thread.
The backend thread cannot be allowed to trigger read actions that wait on the GPU, as it will end up waiting on itself, and never advancing.
In the case of backend multithreading's `SyncMap`, it would try to wait for a backend sync object that does not yet exist, as the sync object would exist according to the GPU and tracking, but it has not yet been created by the backend (and never will be, since it's stuck).
The fix is to queue the 0 draw event just like any other, its _bufferMap value is just forced to 0.
This affects games that use Conditional Rendering: SMO, Splatoon 2, MK8. Was generally indicated by a red message in log saying that the query result timed out after 5000 tries, but not always the cause.
2022-01-21 23:19:15 +01:00
|
|
|
|
_current.Complete(draws > 0);
|
|
|
|
|
_events.Enqueue(_current);
|
2020-12-03 19:42:59 +01:00
|
|
|
|
|
Fix deadlock for GPU counter report when 0 draws are done (#3019)
This fixes a rare bug where reporting a counter for a region containing 0 draws could deadlock the GPU. If this write overlaps with a tracking action, then the GPU could end up waiting on something that it's meant to do in the future, so it would just get stuck.
Before, this reported immediately and wrote the result to guest memory (tracked) from the backend thread.
The backend thread cannot be allowed to trigger read actions that wait on the GPU, as it will end up waiting on itself, and never advancing.
In the case of backend multithreading's `SyncMap`, it would try to wait for a backend sync object that does not yet exist, as the sync object would exist according to the GPU and tracking, but it has not yet been created by the backend (and never will be, since it's stuck).
The fix is to queue the 0 draw event just like any other, its _bufferMap value is just forced to 0.
This affects games that use Conditional Rendering: SMO, Splatoon 2, MK8. Was generally indicated by a red message in log saying that the query result timed out after 5000 tries, but not always the cause.
2022-01-21 23:19:15 +01:00
|
|
|
|
_current.OnResult += resultHandler;
|
2020-05-04 04:24:59 +02:00
|
|
|
|
|
|
|
|
|
result = _current;
|
|
|
|
|
|
2020-12-03 19:42:59 +01:00
|
|
|
|
_current = new CounterQueueEvent(this, GetTarget(Type), lastDrawIndex);
|
2020-05-04 04:24:59 +02:00
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
_queuedEvent.Set();
|
|
|
|
|
|
|
|
|
|
return result;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
public void QueueReset()
|
|
|
|
|
{
|
|
|
|
|
lock (_lock)
|
|
|
|
|
{
|
|
|
|
|
_current.Clear();
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
private static QueryTarget GetTarget(CounterType type)
|
|
|
|
|
{
|
|
|
|
|
switch (type)
|
|
|
|
|
{
|
|
|
|
|
case CounterType.SamplesPassed: return QueryTarget.SamplesPassed;
|
|
|
|
|
case CounterType.PrimitivesGenerated: return QueryTarget.PrimitivesGenerated;
|
|
|
|
|
case CounterType.TransformFeedbackPrimitivesWritten: return QueryTarget.TransformFeedbackPrimitivesWritten;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return QueryTarget.SamplesPassed;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
public void Flush(bool blocking)
|
|
|
|
|
{
|
|
|
|
|
if (!blocking)
|
|
|
|
|
{
|
|
|
|
|
// Just wake the consumer thread - it will update the queries.
|
|
|
|
|
_wakeSignal.Set();
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
lock (_lock)
|
|
|
|
|
{
|
|
|
|
|
// Tell the queue to process all events.
|
|
|
|
|
while (_events.Count > 0)
|
|
|
|
|
{
|
|
|
|
|
CounterQueueEvent flush = _events.Peek();
|
|
|
|
|
if (!flush.TryConsume(ref _accumulatedCounter, true))
|
|
|
|
|
{
|
|
|
|
|
return; // If not blocking, then return when we encounter an event that is not ready yet.
|
|
|
|
|
}
|
|
|
|
|
_events.Dequeue();
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
public void FlushTo(CounterQueueEvent evt)
|
|
|
|
|
{
|
Add a Multithreading layer for the GAL, multi-thread shader compilation at runtime (#2501)
* Initial Implementation
About as fast as nvidia GL multithreading, can be improved with faster command queuing.
* Struct based command list
Speeds up a bit. Still a lot of time lost to resource copy.
* Do shader init while the render thread is active.
* Introduce circular span pool V1
Ideally should be able to use structs instead of references for storing these spans on commands. Will try that next.
* Refactor SpanRef some more
Use a struct to represent SpanRef, rather than a reference.
* Flush buffers on background thread
* Use a span for UpdateRenderScale.
Much faster than copying the array.
* Calculate command size using reflection
* WIP parallel shaders
* Some minor optimisation
* Only 2 max refs per command now.
The command with 3 refs is gone. :relieved:
* Don't cast on the GPU side
* Remove redundant casts, force sync on window present
* Fix Shader Cache
* Fix host shader save.
* Fixup to work with new renderer stuff
* Make command Run static, use array of delegates as lookup
Profile says this takes less time than the previous way.
* Bring up to date
* Add settings toggle. Fix Muiltithreading Off mode.
* Fix warning.
* Release tracking lock for flushes
* Fix Conditional Render fast path with threaded gal
* Make handle iteration safe when releasing the lock
This is mostly temporary.
* Attempt to set backend threading on driver
Only really works on nvidia before launching a game.
* Fix race condition with BufferModifiedRangeList, exceptions in tracking actions
* Update buffer set commands
* Some cleanup
* Only use stutter workaround when using opengl renderer non-threaded
* Add host-conditional reservation of counter events
There has always been the possibility that conditional rendering could use a query object just as it is disposed by the counter queue. This change makes it so that when the host decides to use host conditional rendering, the query object is reserved so that it cannot be deleted. Counter events can optionally start reserved, as the threaded implementation can reserve them before the backend creates them, and there would otherwise be a short amount of time where the counter queue could dispose the event before a call to reserve it could be made.
* Address Feedback
* Make counter flush tracked again.
Hopefully does not cause any issues this time.
* Wait for FlushTo on the main queue thread.
Currently assumes only one thread will want to FlushTo (in this case, the GPU thread)
* Add SDL2 headless integration
* Add HLE macro commands.
Co-authored-by: Mary <mary@mary.zone>
2021-08-27 00:31:29 +02:00
|
|
|
|
// Flush the counter queue on the main thread.
|
2020-05-04 04:24:59 +02:00
|
|
|
|
|
Add a Multithreading layer for the GAL, multi-thread shader compilation at runtime (#2501)
* Initial Implementation
About as fast as nvidia GL multithreading, can be improved with faster command queuing.
* Struct based command list
Speeds up a bit. Still a lot of time lost to resource copy.
* Do shader init while the render thread is active.
* Introduce circular span pool V1
Ideally should be able to use structs instead of references for storing these spans on commands. Will try that next.
* Refactor SpanRef some more
Use a struct to represent SpanRef, rather than a reference.
* Flush buffers on background thread
* Use a span for UpdateRenderScale.
Much faster than copying the array.
* Calculate command size using reflection
* WIP parallel shaders
* Some minor optimisation
* Only 2 max refs per command now.
The command with 3 refs is gone. :relieved:
* Don't cast on the GPU side
* Remove redundant casts, force sync on window present
* Fix Shader Cache
* Fix host shader save.
* Fixup to work with new renderer stuff
* Make command Run static, use array of delegates as lookup
Profile says this takes less time than the previous way.
* Bring up to date
* Add settings toggle. Fix Muiltithreading Off mode.
* Fix warning.
* Release tracking lock for flushes
* Fix Conditional Render fast path with threaded gal
* Make handle iteration safe when releasing the lock
This is mostly temporary.
* Attempt to set backend threading on driver
Only really works on nvidia before launching a game.
* Fix race condition with BufferModifiedRangeList, exceptions in tracking actions
* Update buffer set commands
* Some cleanup
* Only use stutter workaround when using opengl renderer non-threaded
* Add host-conditional reservation of counter events
There has always been the possibility that conditional rendering could use a query object just as it is disposed by the counter queue. This change makes it so that when the host decides to use host conditional rendering, the query object is reserved so that it cannot be deleted. Counter events can optionally start reserved, as the threaded implementation can reserve them before the backend creates them, and there would otherwise be a short amount of time where the counter queue could dispose the event before a call to reserve it could be made.
* Address Feedback
* Make counter flush tracked again.
Hopefully does not cause any issues this time.
* Wait for FlushTo on the main queue thread.
Currently assumes only one thread will want to FlushTo (in this case, the GPU thread)
* Add SDL2 headless integration
* Add HLE macro commands.
Co-authored-by: Mary <mary@mary.zone>
2021-08-27 00:31:29 +02:00
|
|
|
|
Interlocked.Increment(ref _waiterCount);
|
2020-05-04 04:24:59 +02:00
|
|
|
|
|
Add a Multithreading layer for the GAL, multi-thread shader compilation at runtime (#2501)
* Initial Implementation
About as fast as nvidia GL multithreading, can be improved with faster command queuing.
* Struct based command list
Speeds up a bit. Still a lot of time lost to resource copy.
* Do shader init while the render thread is active.
* Introduce circular span pool V1
Ideally should be able to use structs instead of references for storing these spans on commands. Will try that next.
* Refactor SpanRef some more
Use a struct to represent SpanRef, rather than a reference.
* Flush buffers on background thread
* Use a span for UpdateRenderScale.
Much faster than copying the array.
* Calculate command size using reflection
* WIP parallel shaders
* Some minor optimisation
* Only 2 max refs per command now.
The command with 3 refs is gone. :relieved:
* Don't cast on the GPU side
* Remove redundant casts, force sync on window present
* Fix Shader Cache
* Fix host shader save.
* Fixup to work with new renderer stuff
* Make command Run static, use array of delegates as lookup
Profile says this takes less time than the previous way.
* Bring up to date
* Add settings toggle. Fix Muiltithreading Off mode.
* Fix warning.
* Release tracking lock for flushes
* Fix Conditional Render fast path with threaded gal
* Make handle iteration safe when releasing the lock
This is mostly temporary.
* Attempt to set backend threading on driver
Only really works on nvidia before launching a game.
* Fix race condition with BufferModifiedRangeList, exceptions in tracking actions
* Update buffer set commands
* Some cleanup
* Only use stutter workaround when using opengl renderer non-threaded
* Add host-conditional reservation of counter events
There has always been the possibility that conditional rendering could use a query object just as it is disposed by the counter queue. This change makes it so that when the host decides to use host conditional rendering, the query object is reserved so that it cannot be deleted. Counter events can optionally start reserved, as the threaded implementation can reserve them before the backend creates them, and there would otherwise be a short amount of time where the counter queue could dispose the event before a call to reserve it could be made.
* Address Feedback
* Make counter flush tracked again.
Hopefully does not cause any issues this time.
* Wait for FlushTo on the main queue thread.
Currently assumes only one thread will want to FlushTo (in this case, the GPU thread)
* Add SDL2 headless integration
* Add HLE macro commands.
Co-authored-by: Mary <mary@mary.zone>
2021-08-27 00:31:29 +02:00
|
|
|
|
_wakeSignal.Set();
|
|
|
|
|
|
|
|
|
|
while (!evt.Disposed)
|
|
|
|
|
{
|
|
|
|
|
_eventConsumed.WaitOne(1);
|
2020-05-04 04:24:59 +02:00
|
|
|
|
}
|
Add a Multithreading layer for the GAL, multi-thread shader compilation at runtime (#2501)
* Initial Implementation
About as fast as nvidia GL multithreading, can be improved with faster command queuing.
* Struct based command list
Speeds up a bit. Still a lot of time lost to resource copy.
* Do shader init while the render thread is active.
* Introduce circular span pool V1
Ideally should be able to use structs instead of references for storing these spans on commands. Will try that next.
* Refactor SpanRef some more
Use a struct to represent SpanRef, rather than a reference.
* Flush buffers on background thread
* Use a span for UpdateRenderScale.
Much faster than copying the array.
* Calculate command size using reflection
* WIP parallel shaders
* Some minor optimisation
* Only 2 max refs per command now.
The command with 3 refs is gone. :relieved:
* Don't cast on the GPU side
* Remove redundant casts, force sync on window present
* Fix Shader Cache
* Fix host shader save.
* Fixup to work with new renderer stuff
* Make command Run static, use array of delegates as lookup
Profile says this takes less time than the previous way.
* Bring up to date
* Add settings toggle. Fix Muiltithreading Off mode.
* Fix warning.
* Release tracking lock for flushes
* Fix Conditional Render fast path with threaded gal
* Make handle iteration safe when releasing the lock
This is mostly temporary.
* Attempt to set backend threading on driver
Only really works on nvidia before launching a game.
* Fix race condition with BufferModifiedRangeList, exceptions in tracking actions
* Update buffer set commands
* Some cleanup
* Only use stutter workaround when using opengl renderer non-threaded
* Add host-conditional reservation of counter events
There has always been the possibility that conditional rendering could use a query object just as it is disposed by the counter queue. This change makes it so that when the host decides to use host conditional rendering, the query object is reserved so that it cannot be deleted. Counter events can optionally start reserved, as the threaded implementation can reserve them before the backend creates them, and there would otherwise be a short amount of time where the counter queue could dispose the event before a call to reserve it could be made.
* Address Feedback
* Make counter flush tracked again.
Hopefully does not cause any issues this time.
* Wait for FlushTo on the main queue thread.
Currently assumes only one thread will want to FlushTo (in this case, the GPU thread)
* Add SDL2 headless integration
* Add HLE macro commands.
Co-authored-by: Mary <mary@mary.zone>
2021-08-27 00:31:29 +02:00
|
|
|
|
|
|
|
|
|
Interlocked.Decrement(ref _waiterCount);
|
2020-05-04 04:24:59 +02:00
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
public void Dispose()
|
|
|
|
|
{
|
|
|
|
|
lock (_lock)
|
|
|
|
|
{
|
|
|
|
|
while (_events.Count > 0)
|
|
|
|
|
{
|
|
|
|
|
CounterQueueEvent evt = _events.Dequeue();
|
|
|
|
|
|
|
|
|
|
evt.Dispose();
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
Disposed = true;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
_queuedEvent.Set();
|
|
|
|
|
|
|
|
|
|
_consumerThread.Join();
|
|
|
|
|
|
|
|
|
|
foreach (BufferedQuery query in _queryPool)
|
|
|
|
|
{
|
|
|
|
|
query.Dispose();
|
|
|
|
|
}
|
Add a Multithreading layer for the GAL, multi-thread shader compilation at runtime (#2501)
* Initial Implementation
About as fast as nvidia GL multithreading, can be improved with faster command queuing.
* Struct based command list
Speeds up a bit. Still a lot of time lost to resource copy.
* Do shader init while the render thread is active.
* Introduce circular span pool V1
Ideally should be able to use structs instead of references for storing these spans on commands. Will try that next.
* Refactor SpanRef some more
Use a struct to represent SpanRef, rather than a reference.
* Flush buffers on background thread
* Use a span for UpdateRenderScale.
Much faster than copying the array.
* Calculate command size using reflection
* WIP parallel shaders
* Some minor optimisation
* Only 2 max refs per command now.
The command with 3 refs is gone. :relieved:
* Don't cast on the GPU side
* Remove redundant casts, force sync on window present
* Fix Shader Cache
* Fix host shader save.
* Fixup to work with new renderer stuff
* Make command Run static, use array of delegates as lookup
Profile says this takes less time than the previous way.
* Bring up to date
* Add settings toggle. Fix Muiltithreading Off mode.
* Fix warning.
* Release tracking lock for flushes
* Fix Conditional Render fast path with threaded gal
* Make handle iteration safe when releasing the lock
This is mostly temporary.
* Attempt to set backend threading on driver
Only really works on nvidia before launching a game.
* Fix race condition with BufferModifiedRangeList, exceptions in tracking actions
* Update buffer set commands
* Some cleanup
* Only use stutter workaround when using opengl renderer non-threaded
* Add host-conditional reservation of counter events
There has always been the possibility that conditional rendering could use a query object just as it is disposed by the counter queue. This change makes it so that when the host decides to use host conditional rendering, the query object is reserved so that it cannot be deleted. Counter events can optionally start reserved, as the threaded implementation can reserve them before the backend creates them, and there would otherwise be a short amount of time where the counter queue could dispose the event before a call to reserve it could be made.
* Address Feedback
* Make counter flush tracked again.
Hopefully does not cause any issues this time.
* Wait for FlushTo on the main queue thread.
Currently assumes only one thread will want to FlushTo (in this case, the GPU thread)
* Add SDL2 headless integration
* Add HLE macro commands.
Co-authored-by: Mary <mary@mary.zone>
2021-08-27 00:31:29 +02:00
|
|
|
|
|
|
|
|
|
_queuedEvent.Dispose();
|
|
|
|
|
_wakeSignal.Dispose();
|
|
|
|
|
_eventConsumed.Dispose();
|
2020-05-04 04:24:59 +02:00
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|