Cocoa Port: Reduce threading overhead for framebuffer fetches and SPU_Emulate_user().

- Framebuffer fetches no longer run on a CocoaDSThread, but instead uses a pthread directly. This can be done since framebuffer fetching only serves one function and always receives the same execution message, making a full CocoaDSThread a waste.
- SPU_Emulate_user() is no longer called on a separate thread, and instead is called in the emulation thread directly. For the typical SPU use case (SPU Sound Synchronization w/ Advanced SPU Logic), SPU_Emulator_user() becomes negligible, and so the threading overhead becomes unnecessary. In the use case where Dual SPU Synch/Asynch is used, Advanced SPU Logic is almost always disabled with it, and so the penalty of calling SPU_Emulator_user() on the emulation thread will be more than compensated by the performance increase of turning off Advanced SPU Logic.
- Also do some code cleanup/refactoring here and there.

desmume/src/frontend/cocoa/cocoa_GPU.h

@@ -39,7 +39,7 @@ class GPUEventHandlerOSX;
 typedef std::map<CGDirectDisplayID, CVDisplayLinkRef> DisplayLinksActiveMap;
 typedef std::map<CGDirectDisplayID, int64_t> DisplayLinkFlushTimeLimitMap;
 
-@interface MacClientSharedObject : CocoaDSThread
+@interface MacClientSharedObject : NSObject
 {
 	GPUClientFetchObject *GPUFetchObject;
 	pthread_rwlock_t *_rwlockFramebuffer[2];
@@ -50,6 +50,13 @@ typedef std::map<CGDirectDisplayID, int64_t> DisplayLinkFlushTimeLimitMap;
 	
 	DisplayLinksActiveMap _displayLinksActiveList;
 	DisplayLinkFlushTimeLimitMap _displayLinkFlushTimeList;
+	
+	OSSpinLock spinlockFetchSignal;
+	BOOL _isFetchSignalled;
+	uint8_t _fetchIndex;
+	pthread_t _threadFetch;
+	pthread_cond_t _condSignalFetch;
+	pthread_mutex_t _mutexFetchExecute;
 }
 
 @property (assign, nonatomic) GPUClientFetchObject *GPUFetchObject;
@@ -69,6 +76,9 @@ typedef std::map<CGDirectDisplayID, int64_t> DisplayLinkFlushTimeLimitMap;
 - (void) displayLinkStartUsingID:(CGDirectDisplayID)displayID;
 - (void) displayLinkListUpdate;
 
+- (void) signalFetchAtIndex:(uint8_t)index;
+- (void) runFetchLoop;
+
 @end
 
 @interface CocoaDSGPU : NSObject
@@ -131,6 +141,8 @@ extern "C"
 {
 #endif
 
+static void* RunFetchThread(void *arg);
+
 CVReturn MacDisplayLinkCallback(CVDisplayLinkRef displayLink,
 								const CVTimeStamp *inNow,
 								const CVTimeStamp *inOutputTime,

desmume/src/frontend/cocoa/cocoa_GPU.mm

@@ -887,6 +887,13 @@ public:
 	_displayLinkFlushTimeList.clear();
 	[self displayLinkListUpdate];
 	
+	spinlockFetchSignal = OS_SPINLOCK_INIT;
+	_isFetchSignalled = NO;
+	_fetchIndex = 0;
+	pthread_cond_init(&_condSignalFetch, NULL);
+	pthread_create(&_threadFetch, NULL, &RunFetchThread, self);
+	pthread_mutex_init(&_mutexFetchExecute, NULL);
+		
 	[[NSNotificationCenter defaultCenter] addObserver:self
 											 selector:@selector(respondToScreenChange:)
 												 name:@"NSApplicationDidChangeScreenParametersNotification"
@@ -897,6 +904,13 @@ public:
 
 - (void)dealloc
 {
+	pthread_cancel(_threadFetch);
+	pthread_join(_threadFetch, NULL);
+	_threadFetch = NULL;
+	
+	pthread_cond_destroy(&_condSignalFetch);
+	pthread_mutex_destroy(&_mutexFetchExecute);
+	
 	pthread_mutex_lock(&_mutexFlushVideo);
 	
 	for (DisplayLinksActiveMap::iterator it = _displayLinksActiveList.begin(); it != _displayLinksActiveList.end(); ++it)
@@ -914,6 +928,7 @@ public:
 	}
 	
 	pthread_mutex_unlock(&_mutexFlushVideo);
+	pthread_mutex_destroy(&_mutexFlushVideo);
 	
 	pthread_mutex_t *currentMutex = _mutexOutputList;
 	
@@ -931,28 +946,10 @@ public:
 	
 	pthread_rwlock_destroy(_rwlockFramebuffer[0]);
 	pthread_rwlock_destroy(_rwlockFramebuffer[1]);
-	pthread_mutex_destroy(&_mutexFlushVideo);
 	
 	[super dealloc];
 }
 
-- (void)handlePortMessage:(NSPortMessage *)portMessage
-{
-	NSInteger message = (NSInteger)[portMessage msgid];
-	NSArray *messageComponents = [portMessage components];
-	
-	switch (message)
-	{
-		case MESSAGE_FETCH_AND_PUSH_VIDEO:
-			[self handleFetchFromBufferIndexAndPushVideo:[messageComponents objectAtIndex:0]];
-			break;
-			
-		default:
-			[super handlePortMessage:portMessage];
-			break;
-	}
-}
-
 - (void) handleFetchFromBufferIndexAndPushVideo:(NSData *)indexData
 {
 	const NSInteger index = *(NSInteger *)[indexData bytes];
@@ -1013,7 +1010,7 @@ public:
 	
 	for (CocoaDSOutput *cdsOutput in _cdsOutputList)
 	{
-		if ([cdsOutput isKindOfClass:[CocoaDSDisplay class]])
+		if ([cdsOutput isKindOfClass:[CocoaDSDisplayVideo class]])
 		{
 			[CocoaDSUtil messageSendOneWay:[cdsOutput receivePort] msgID:MESSAGE_RECEIVE_GPU_FRAME];
 		}
@@ -1036,9 +1033,9 @@ public:
 	
 	for (CocoaDSOutput *cdsOutput in _cdsOutputList)
 	{
-		if ([cdsOutput isKindOfClass:[CocoaDSDisplay class]])
+		if ([cdsOutput isKindOfClass:[CocoaDSDisplayVideo class]])
 		{
-			ClientDisplay3DView *cdv = [(CocoaDSDisplay *)cdsOutput clientDisplayView];
+			ClientDisplay3DView *cdv = [(CocoaDSDisplayVideo *)cdsOutput clientDisplayView];
 			cdv->FinishFrameAtIndex(bufferIndex);
 		}
 	}
@@ -1189,6 +1186,36 @@ public:
 	pthread_mutex_unlock(&_mutexFlushVideo);
 }
 
+- (void) signalFetchAtIndex:(uint8_t)index
+{
+	pthread_mutex_lock(&_mutexFetchExecute);
+	
+	_fetchIndex = index;
+	_isFetchSignalled = YES;
+	pthread_cond_signal(&_condSignalFetch);
+	
+	pthread_mutex_unlock(&_mutexFetchExecute);
+}
+
+- (void) runFetchLoop
+{
+	do
+	{
+		pthread_mutex_lock(&_mutexFetchExecute);
+		
+		while (!_isFetchSignalled)
+		{
+			pthread_cond_wait(&_condSignalFetch, &_mutexFetchExecute);
+		}
+		_isFetchSignalled = NO;
+		
+		GPUFetchObject->FetchFromBufferIndex(_fetchIndex);
+		[self pushVideoDataToAllDisplayViews];
+		
+		pthread_mutex_unlock(&_mutexFetchExecute);
+	} while(true);
+}
+
 - (void) respondToScreenChange:(NSNotification *)aNotification
 {
 	[self displayLinkListUpdate];
@@ -1256,7 +1283,7 @@ void GPUEventHandlerOSX::DidFrameEnd(bool isFrameSkipped, const NDSDisplayInfo &
 #if !defined(PORT_VERSION_OPENEMU)
 	if (!isFrameSkipped)
 	{
-		[CocoaDSUtil messageSendOneWayWithInteger:[sharedViewObject receivePort] msgID:MESSAGE_FETCH_AND_PUSH_VIDEO integerValue:latestDisplayInfo.bufferIndex];
+		[sharedViewObject signalFetchAtIndex:latestDisplayInfo.bufferIndex];
 	}
 #endif
 }
@@ -1317,6 +1344,14 @@ CGLContextObj OSXOpenGLRendererContext = NULL;
 CGLPBufferObj OSXOpenGLRendererPBuffer = NULL;
 #pragma GCC diagnostic pop
 
+static void* RunFetchThread(void *arg)
+{
+	MacClientSharedObject *sharedData = (MacClientSharedObject *)arg;
+	[sharedData runFetchLoop];
+	
+	return NULL;
+}
+
 CVReturn MacDisplayLinkCallback(CVDisplayLinkRef displayLink,
 								const CVTimeStamp *inNow,
 								const CVTimeStamp *inOutputTime,

desmume/src/frontend/cocoa/cocoa_core.mm

@@ -35,6 +35,7 @@
 #include "../../gdbstub.h"
 #include "../../slot1.h"
 #include "../../slot2.h"
+#include "../../SPU.h"
 #undef BOOL
 
 // Need to include assert.h this way so that GDB stub will work
@@ -1031,6 +1032,7 @@ static void* RunCoreThread(void *arg)
 		// Execute the frame and increment the frame counter.
 		pthread_rwlock_wrlock(&param->rwlockCoreExecute);
 		NDS_exec<false>();
+		SPU_Emulate_user();
 		execControl->FetchOutputPostNDSExec();
 		pthread_rwlock_unlock(&param->rwlockCoreExecute);
 		
@@ -1069,7 +1071,7 @@ static void* RunCoreThread(void *arg)
 						[(CocoaDSDisplay *)cdsOutput setNDSFrameInfo:ndsFrameInfo];
 					}
 					
-					if ( ![cdsOutput isKindOfClass:[CocoaDSDisplay class]] || (framesToSkip == 0) )
+					if ( ![cdsOutput isKindOfClass:[CocoaDSSpeaker class]] && (![cdsOutput isKindOfClass:[CocoaDSDisplay class]] || (framesToSkip == 0)) )
 					{
 						[cdsOutput doCoreEmuFrame];
 					}

desmume/src/frontend/cocoa/cocoa_output.h

@@ -96,10 +96,6 @@
 
 @interface CocoaDSDisplay : CocoaDSOutput
 {
-	ClientDisplay3DView *_cdv;
-	ClientDisplayViewProperties _intermediateViewProps;
-	NSSize displaySize;
-	
 	uint32_t _receivedFrameIndex;
 	uint32_t _currentReceivedFrameIndex;
 	uint32_t _receivedFrameCount;
@@ -108,28 +104,21 @@
 	
 	OSSpinLock spinlockReceivedFrameIndex;
 	OSSpinLock spinlockNDSFrameInfo;
-	OSSpinLock spinlockViewProperties;
 }
 
-@property (assign, nonatomic) ClientDisplay3DView *clientDisplayView;
-@property (readonly) NSSize displaySize;
-
-- (void) commitViewProperties:(const ClientDisplayViewProperties &)viewProps;
-
 - (void) handleReceiveGPUFrame;
-- (void) handleChangeViewProperties;
-- (void) handleRequestScreenshot:(NSData *)fileURLStringData fileTypeData:(NSData *)fileTypeData;
-- (void) handleCopyToPasteboard;
 
 - (void) takeFrameCount;
 - (void) setNDSFrameInfo:(const NDSFrameInfo &)ndsFrameInfo;
-- (NSImage *) image;
-- (NSBitmapImageRep *) bitmapImageRep;
 
 @end
 
 @interface CocoaDSDisplayVideo : CocoaDSDisplay
 {
+	ClientDisplay3DView *_cdv;
+	ClientDisplayViewProperties _intermediateViewProps;
+	
+	OSSpinLock spinlockViewProperties;
 	OSSpinLock spinlockIsHUDVisible;
 	OSSpinLock spinlockUseVerticalSync;
 	OSSpinLock spinlockVideoFiltersPreferGPU;
@@ -140,6 +129,7 @@
 	OSSpinLock spinlockDisplayID;
 }
 
+@property (assign, nonatomic) ClientDisplay3DView *clientDisplayView;
 @property (readonly, nonatomic) BOOL canFilterOnGPU;
 @property (readonly, nonatomic) BOOL willFilterOnGPU;
 @property (assign) BOOL isHUDVisible;
@@ -168,12 +158,19 @@
 @property (assign) NSInteger outputFilter;
 @property (assign) NSInteger pixelScaler;
 
+- (void) commitViewProperties:(const ClientDisplayViewProperties &)viewProps;
+
+- (void) handleChangeViewProperties;
 - (void) handleReceiveGPUFrame;
 - (void) handleReloadReprocessRedraw;
 - (void) handleReprocessRedraw;
 - (void) handleRedraw;
+- (void) handleCopyToPasteboard;
+- (void) handleRequestScreenshot:(NSData *)fileURLStringData fileTypeData:(NSData *)fileTypeData;
 
 - (void) setScaleFactor:(float)theScaleFactor;
 - (void) hudUpdate;
+- (NSImage *) image;
+- (NSBitmapImageRep *) bitmapImageRep;
 
 @end

desmume/src/frontend/cocoa/cocoa_output.mm

@@ -501,9 +501,6 @@
 
 @implementation CocoaDSDisplay
 
-@dynamic clientDisplayView;
-@dynamic displaySize;
-
 - (id)init
 {
 	self = [super init];
@@ -514,9 +511,7 @@
 	
 	spinlockReceivedFrameIndex = OS_SPINLOCK_INIT;
 	spinlockNDSFrameInfo = OS_SPINLOCK_INIT;
-	spinlockViewProperties = OS_SPINLOCK_INIT;
 	
-	_cdv = NULL;
 	_ndsFrameInfo.clear();
 	
 	_receivedFrameIndex = 0;
@@ -531,38 +526,9 @@
 	[super dealloc];
 }
 
-- (void) setClientDisplayView:(ClientDisplay3DView *)clientDisplayView
-{
-	_cdv = clientDisplayView;
-}
-
-- (ClientDisplay3DView *) clientDisplayView
-{
-	return _cdv;
-}
-
-- (void) commitViewProperties:(const ClientDisplayViewProperties &)viewProps
-{
-	OSSpinLockLock(&spinlockViewProperties);
-	_intermediateViewProps = viewProps;
-	OSSpinLockUnlock(&spinlockViewProperties);
-	
-	[self handleChangeViewProperties];
-}
-
-- (NSSize) displaySize
-{
-	pthread_rwlock_rdlock(self.rwlockProducer);
-	NSSize size = NSMakeSize((CGFloat)GPU->GetCustomFramebufferWidth(), (_cdv->GetMode() == ClientDisplayMode_Dual) ? (CGFloat)(GPU->GetCustomFramebufferHeight() * 2): (CGFloat)GPU->GetCustomFramebufferHeight());
-	pthread_rwlock_unlock(self.rwlockProducer);
-	
-	return size;
-}
-
 - (void)handlePortMessage:(NSPortMessage *)portMessage
 {
 	NSInteger message = (NSInteger)[portMessage msgid];
-	NSArray *messageComponents = [portMessage components];
 	
 	switch (message)
 	{
@@ -570,18 +536,6 @@
 			[self handleReceiveGPUFrame];
 			break;
 		
-		case MESSAGE_CHANGE_VIEW_PROPERTIES:
-			[self handleChangeViewProperties];
-			break;
-			
-		case MESSAGE_REQUEST_SCREENSHOT:
-			[self handleRequestScreenshot:[messageComponents objectAtIndex:0] fileTypeData:[messageComponents objectAtIndex:1]];
-			break;
-			
-		case MESSAGE_COPY_TO_PASTEBOARD:
-			[self handleCopyToPasteboard];
-			break;
-		
 		default:
 			[super handlePortMessage:portMessage];
 			break;
@@ -595,46 +549,6 @@
 	OSSpinLockUnlock(&spinlockReceivedFrameIndex);
 }
 
-- (void) handleChangeViewProperties
-{
-	OSSpinLockLock(&spinlockViewProperties);
-	_cdv->CommitViewProperties(_intermediateViewProps);
-	OSSpinLockUnlock(&spinlockViewProperties);
-	
-	_cdv->SetupViewProperties();
-}
-
-- (void) handleRequestScreenshot:(NSData *)fileURLStringData fileTypeData:(NSData *)fileTypeData
-{
-	NSString *fileURLString = [[NSString alloc] initWithData:fileURLStringData encoding:NSUTF8StringEncoding];
-	NSURL *fileURL = [NSURL URLWithString:fileURLString];
-	NSBitmapImageFileType fileType = *(NSBitmapImageFileType *)[fileTypeData bytes];	
-	
-	NSDictionary *userInfo = [[NSDictionary alloc] initWithObjectsAndKeys:
-							  fileURL, @"fileURL", 
-							  [NSNumber numberWithInteger:(NSInteger)fileType], @"fileType",
-							  [self image], @"screenshotImage",
-							  nil];
-	
-	[[NSNotificationCenter defaultCenter] postNotificationOnMainThreadName:@"org.desmume.DeSmuME.requestScreenshotDidFinish" object:self userInfo:userInfo];
-	[userInfo release];
-	
-	[fileURLString release];
-}
-
-- (void) handleCopyToPasteboard
-{
-	NSImage *screenshot = [self image];
-	if (screenshot == nil)
-	{
-		return;
-	}
-	
-	NSPasteboard *pboard = [NSPasteboard generalPasteboard];
-	[pboard declareTypes:[NSArray arrayWithObjects:NSTIFFPboardType, nil] owner:self];
-	[pboard setData:[screenshot TIFFRepresentationUsingCompression:NSTIFFCompressionLZW factor:1.0f] forType:NSTIFFPboardType];
-}
-
 - (void) takeFrameCount
 {
 	OSSpinLockLock(&spinlockReceivedFrameIndex);
@@ -650,88 +564,11 @@
 	OSSpinLockUnlock(&spinlockNDSFrameInfo);
 }
 
-- (NSImage *) image
-{
-	NSImage *newImage = [[NSImage alloc] initWithSize:[self displaySize]];
-	if (newImage == nil)
-	{
-		return newImage;
-	}
-	 
-	// Render the frame in an NSBitmapImageRep
-	NSBitmapImageRep *newImageRep = [self bitmapImageRep];
-	if (newImageRep == nil)
-	{
-		[newImage release];
-		newImage = nil;
-		return newImage;
-	}
-	
-	// Attach the rendered frame to the NSImageRep
-	[newImage addRepresentation:newImageRep];
-	
-	return [newImage autorelease];
-}
-
-- (NSBitmapImageRep *) bitmapImageRep
-{
-	GPUClientFetchObject &fetchObjMutable = (GPUClientFetchObject &)_cdv->GetFetchObject();
-	NDSDisplayInfo &displayInfoMutable = (NDSDisplayInfo &)fetchObjMutable.GetFetchDisplayInfoForBufferIndex(fetchObjMutable.GetLastFetchIndex());
-	
-	NSUInteger w = (NSUInteger)displayInfoMutable.customWidth;
-	NSUInteger h = (_cdv->GetMode() == ClientDisplayMode_Dual) ? (NSUInteger)(displayInfoMutable.customHeight * 2) : (NSUInteger)displayInfoMutable.customHeight;
-	
-	NSBitmapImageRep *imageRep = [[NSBitmapImageRep alloc] initWithBitmapDataPlanes:NULL
-																		 pixelsWide:w
-																		 pixelsHigh:h
-																	  bitsPerSample:8
-																	samplesPerPixel:4
-																		   hasAlpha:YES
-																		   isPlanar:NO
-																	 colorSpaceName:NSCalibratedRGBColorSpace
-																		bytesPerRow:w * 4
-																	   bitsPerPixel:32];
-	
-	if(imageRep == nil)
-	{
-		return imageRep;
-	}
-	
-	void *displayBuffer = displayInfoMutable.masterCustomBuffer;
-	uint32_t *bitmapData = (uint32_t *)[imageRep bitmapData];
-	
-	pthread_rwlock_wrlock(self.rwlockProducer);
-	
-	GPU->PostprocessDisplay(NDSDisplayID_Main,  displayInfoMutable);
-	GPU->PostprocessDisplay(NDSDisplayID_Touch, displayInfoMutable);
-	GPU->ResolveDisplayToCustomFramebuffer(NDSDisplayID_Main,  displayInfoMutable);
-	GPU->ResolveDisplayToCustomFramebuffer(NDSDisplayID_Touch, displayInfoMutable);
-	
-	if (displayInfoMutable.pixelBytes == 2)
-	{
-		ColorspaceConvertBuffer555To8888Opaque<false, true>((u16 *)displayBuffer, bitmapData, (w * h));
-	}
-	else if (displayInfoMutable.pixelBytes == 4)
-	{
-		memcpy(bitmapData, displayBuffer, w * h * sizeof(uint32_t));
-	}
-	
-	pthread_rwlock_unlock(self.rwlockProducer);
-	
-#ifdef MSB_FIRST
-	for (size_t i = 0; i < w * h; i++)
-	{
-		bitmapData[i] = LE_TO_LOCAL_32(bitmapData[i]);
-	}
-#endif
-	
-	return [imageRep autorelease];
-}
-
 @end
 
 @implementation CocoaDSDisplayVideo
 
+@dynamic clientDisplayView;
 @dynamic canFilterOnGPU;
 @dynamic willFilterOnGPU;
 @dynamic isHUDVisible;
@@ -768,6 +605,9 @@
 		return self;
 	}
 	
+	_cdv = NULL;
+	
+	spinlockViewProperties = OS_SPINLOCK_INIT;
 	spinlockIsHUDVisible = OS_SPINLOCK_INIT;
 	spinlockUseVerticalSync = OS_SPINLOCK_INIT;
 	spinlockVideoFiltersPreferGPU = OS_SPINLOCK_INIT;
@@ -785,6 +625,25 @@
 	[super dealloc];
 }
 
+- (void) setClientDisplayView:(ClientDisplay3DView *)clientDisplayView
+{
+	_cdv = clientDisplayView;
+}
+
+- (ClientDisplay3DView *) clientDisplayView
+{
+	return _cdv;
+}
+
+- (void) commitViewProperties:(const ClientDisplayViewProperties &)viewProps
+{
+	OSSpinLockLock(&spinlockViewProperties);
+	_intermediateViewProps = viewProps;
+	OSSpinLockUnlock(&spinlockViewProperties);
+	
+	[self handleChangeViewProperties];
+}
+
 - (BOOL) canFilterOnGPU
 {
 	return (_cdv->CanFilterOnGPU()) ? YES : NO;
@@ -1198,9 +1057,14 @@
 - (void)handlePortMessage:(NSPortMessage *)portMessage
 {
 	NSInteger message = (NSInteger)[portMessage msgid];
+	NSArray *messageComponents = [portMessage components];
 	
 	switch (message)
 	{
+		case MESSAGE_CHANGE_VIEW_PROPERTIES:
+			[self handleChangeViewProperties];
+			break;
+			
 		case MESSAGE_RELOAD_REPROCESS_REDRAW:
 			[self handleReloadReprocessRedraw];
 			break;
@@ -1213,6 +1077,14 @@
 			[self handleRedraw];
 			break;
 			
+		case MESSAGE_COPY_TO_PASTEBOARD:
+			[self handleCopyToPasteboard];
+			break;
+			
+		case MESSAGE_REQUEST_SCREENSHOT:
+			[self handleRequestScreenshot:[messageComponents objectAtIndex:0] fileTypeData:[messageComponents objectAtIndex:1]];
+			break;
+			
 		default:
 			[super handlePortMessage:portMessage];
 			break;
@@ -1226,6 +1098,15 @@
 	_cdv->HandleEmulatorFrameEndEvent();
 }
 
+- (void) handleChangeViewProperties
+{
+	OSSpinLockLock(&spinlockViewProperties);
+	_cdv->CommitViewProperties(_intermediateViewProps);
+	OSSpinLockUnlock(&spinlockViewProperties);
+	
+	_cdv->SetupViewProperties();
+}
+
 - (void) handleReceiveGPUFrame
 {
 	[super handleReceiveGPUFrame];
@@ -1257,6 +1138,37 @@
 	_cdv->UpdateView();
 }
 
+- (void) handleCopyToPasteboard
+{
+	NSImage *screenshot = [self image];
+	if (screenshot == nil)
+	{
+		return;
+	}
+	
+	NSPasteboard *pboard = [NSPasteboard generalPasteboard];
+	[pboard declareTypes:[NSArray arrayWithObjects:NSTIFFPboardType, nil] owner:self];
+	[pboard setData:[screenshot TIFFRepresentationUsingCompression:NSTIFFCompressionLZW factor:1.0f] forType:NSTIFFPboardType];
+}
+
+- (void) handleRequestScreenshot:(NSData *)fileURLStringData fileTypeData:(NSData *)fileTypeData
+{
+	NSString *fileURLString = [[NSString alloc] initWithData:fileURLStringData encoding:NSUTF8StringEncoding];
+	NSURL *fileURL = [NSURL URLWithString:fileURLString];
+	NSBitmapImageFileType fileType = *(NSBitmapImageFileType *)[fileTypeData bytes];
+	
+	NSDictionary *userInfo = [[NSDictionary alloc] initWithObjectsAndKeys:
+							  fileURL, @"fileURL",
+							  [NSNumber numberWithInteger:(NSInteger)fileType], @"fileType",
+							  [self image], @"screenshotImage",
+							  nil];
+	
+	[[NSNotificationCenter defaultCenter] postNotificationOnMainThreadName:@"org.desmume.DeSmuME.requestScreenshotDidFinish" object:self userInfo:userInfo];
+	[userInfo release];
+	
+	[fileURLString release];
+}
+
 - (void) setScaleFactor:(float)theScaleFactor
 {
 	OSSpinLockLock(&spinlockIsHUDVisible);
@@ -1276,4 +1188,86 @@
 	OSSpinLockUnlock(&spinlockNDSFrameInfo);
 }
 
+- (NSImage *) image
+{
+	pthread_rwlock_rdlock(self.rwlockProducer);
+	NSSize displaySize = NSMakeSize((CGFloat)GPU->GetCustomFramebufferWidth(), (_cdv->GetMode() == ClientDisplayMode_Dual) ? (CGFloat)(GPU->GetCustomFramebufferHeight() * 2): (CGFloat)GPU->GetCustomFramebufferHeight());
+	pthread_rwlock_unlock(self.rwlockProducer);
+	
+	NSImage *newImage = [[NSImage alloc] initWithSize:displaySize];
+	if (newImage == nil)
+	{
+		return newImage;
+	}
+	
+	// Render the frame in an NSBitmapImageRep
+	NSBitmapImageRep *newImageRep = [self bitmapImageRep];
+	if (newImageRep == nil)
+	{
+		[newImage release];
+		newImage = nil;
+		return newImage;
+	}
+	
+	// Attach the rendered frame to the NSImageRep
+	[newImage addRepresentation:newImageRep];
+	
+	return [newImage autorelease];
+}
+
+- (NSBitmapImageRep *) bitmapImageRep
+{
+	GPUClientFetchObject &fetchObjMutable = (GPUClientFetchObject &)_cdv->GetFetchObject();
+	NDSDisplayInfo &displayInfoMutable = (NDSDisplayInfo &)fetchObjMutable.GetFetchDisplayInfoForBufferIndex(fetchObjMutable.GetLastFetchIndex());
+	
+	NSUInteger w = (NSUInteger)displayInfoMutable.customWidth;
+	NSUInteger h = (_cdv->GetMode() == ClientDisplayMode_Dual) ? (NSUInteger)(displayInfoMutable.customHeight * 2) : (NSUInteger)displayInfoMutable.customHeight;
+	
+	NSBitmapImageRep *imageRep = [[NSBitmapImageRep alloc] initWithBitmapDataPlanes:NULL
+																		 pixelsWide:w
+																		 pixelsHigh:h
+																	  bitsPerSample:8
+																	samplesPerPixel:4
+																		   hasAlpha:YES
+																		   isPlanar:NO
+																	 colorSpaceName:NSCalibratedRGBColorSpace
+																		bytesPerRow:w * 4
+																	   bitsPerPixel:32];
+	
+	if (imageRep == nil)
+	{
+		return imageRep;
+	}
+	
+	void *displayBuffer = displayInfoMutable.masterCustomBuffer;
+	uint32_t *bitmapData = (uint32_t *)[imageRep bitmapData];
+	
+	pthread_rwlock_wrlock(self.rwlockProducer);
+	
+	GPU->PostprocessDisplay(NDSDisplayID_Main,  displayInfoMutable);
+	GPU->PostprocessDisplay(NDSDisplayID_Touch, displayInfoMutable);
+	GPU->ResolveDisplayToCustomFramebuffer(NDSDisplayID_Main,  displayInfoMutable);
+	GPU->ResolveDisplayToCustomFramebuffer(NDSDisplayID_Touch, displayInfoMutable);
+	
+	if (displayInfoMutable.pixelBytes == 2)
+	{
+		ColorspaceConvertBuffer555To8888Opaque<false, true>((u16 *)displayBuffer, bitmapData, (w * h));
+	}
+	else if (displayInfoMutable.pixelBytes == 4)
+	{
+		memcpy(bitmapData, displayBuffer, w * h * sizeof(uint32_t));
+	}
+	
+	pthread_rwlock_unlock(self.rwlockProducer);
+	
+#ifdef MSB_FIRST
+	for (size_t i = 0; i < w * h; i++)
+	{
+		bitmapData[i] = LE_TO_LOCAL_32(bitmapData[i]);
+	}
+#endif
+	
+	return [imageRep autorelease];
+}
+
 @end

desmume/src/frontend/cocoa/userinterface/DisplayWindowController.mm

@@ -1323,7 +1323,7 @@ static std::unordered_map<NSScreen *, DisplayWindowController *> _screenMap; //
 	
 	// Set up the video output thread.
 	CocoaDSDisplayVideo *newDisplayOutput = [[[CocoaDSDisplayVideo alloc] init] autorelease];
-	[newDisplayOutput setClientDisplayView:[view clientDisplay3DView]];
+	[newDisplayOutput setClientDisplayView:[newView clientDisplay3DView]];
 	
 	ClientDisplayView *cdv = [newDisplayOutput clientDisplayView];
 	NSString *fontPath = [[NSBundle mainBundle] pathForResource:@"SourceSansPro-Bold" ofType:@"otf"];

desmume/src/frontend/cocoa/userinterface/EmuControllerDelegate.mm

@@ -810,32 +810,32 @@
 {
 	const float vol = [self currentVolumeValue];
 	[self setCurrentVolumeValue:vol];
-	[CocoaDSUtil messageSendOneWayWithFloat:[cdsSpeaker receivePort] msgID:MESSAGE_SET_VOLUME floatValue:vol];
+	[[self cdsSpeaker] setVolume:vol];
 }
 
 - (IBAction) changeAudioEngine:(id)sender
 {
-	[CocoaDSUtil messageSendOneWayWithInteger:[cdsSpeaker receivePort] msgID:MESSAGE_SET_AUDIO_PROCESS_METHOD integerValue:[CocoaDSUtil getIBActionSenderTag:sender]];
+	[[self cdsSpeaker] setAudioOutputEngine:[CocoaDSUtil getIBActionSenderTag:sender]];
 }
 
 - (IBAction) changeSpuAdvancedLogic:(id)sender
 {
-	[CocoaDSUtil messageSendOneWayWithBool:[cdsSpeaker receivePort] msgID:MESSAGE_SET_SPU_ADVANCED_LOGIC boolValue:[CocoaDSUtil getIBActionSenderButtonStateBool:sender]];
+	[[self cdsSpeaker] setSpuAdvancedLogic:[CocoaDSUtil getIBActionSenderButtonStateBool:sender]];
 }
 
 - (IBAction) changeSpuInterpolationMode:(id)sender
 {
-	[CocoaDSUtil messageSendOneWayWithInteger:[cdsSpeaker receivePort] msgID:MESSAGE_SET_SPU_INTERPOLATION_MODE integerValue:[CocoaDSUtil getIBActionSenderTag:sender]];
+	[[self cdsSpeaker] setSpuInterpolationMode:[CocoaDSUtil getIBActionSenderTag:sender]];
 }
 
 - (IBAction) changeSpuSyncMode:(id)sender
 {
-	[CocoaDSUtil messageSendOneWayWithInteger:[cdsSpeaker receivePort] msgID:MESSAGE_SET_SPU_SYNC_MODE integerValue:[CocoaDSUtil getIBActionSenderTag:sender]];
+	[[self cdsSpeaker] setSpuSyncMode:[CocoaDSUtil getIBActionSenderTag:sender]];
 }
 
 - (IBAction) changeSpuSyncMethod:(id)sender
 {
-	[CocoaDSUtil messageSendOneWayWithInteger:[cdsSpeaker receivePort] msgID:MESSAGE_SET_SPU_SYNC_METHOD integerValue:[CocoaDSUtil getIBActionSenderTag:sender]];
+	[[self cdsSpeaker] setSpuSyncMethod:[CocoaDSUtil getIBActionSenderTag:sender]];
 }
 
 - (IBAction) toggleAllDisplays:(id)sender
@@ -1533,7 +1533,7 @@
 	}
 
 	[self setCurrentVolumeValue:vol];
-	[CocoaDSUtil messageSendOneWayWithFloat:[cdsSpeaker receivePort] msgID:MESSAGE_SET_VOLUME floatValue:vol];
+	[[self cdsSpeaker] setVolume:vol];
 }
 
 - (void) cmdToggleGPUState:(NSValue *)cmdAttrValue

desmume/src/frontend/cocoa/userinterface/MacMetalDisplayView.mm

@@ -1127,7 +1127,7 @@
 {
 	const VideoFilterAttributes vfAttr = VideoFilter::GetAttributesByID(filterID);
 	
-	id<MTLCommandBuffer> cb = [[sharedData commandQueue] commandBufferWithUnretainedReferences];
+	id<MTLCommandBuffer> cb = [self newCommandBuffer];
 	id<MTLBlitCommandEncoder> dummyEncoder = [cb blitCommandEncoder];
 	[dummyEncoder endEncoding];
 	[cb commit];
@@ -1145,7 +1145,7 @@
 																  options:MTLResourceStorageModeManaged
 															  deallocator:nil]];
 	
-	cb = [[sharedData commandQueue] commandBufferWithUnretainedReferences];
+	cb = [self newCommandBuffer];
 	dummyEncoder = [cb blitCommandEncoder];
 	[dummyEncoder endEncoding];
 	[cb commit];
@@ -1285,7 +1285,7 @@
 		VideoFilter *vfMain  = _cdv->GetPixelScalerObject(NDSDisplayID_Main);
 		VideoFilter *vfTouch = _cdv->GetPixelScalerObject(NDSDisplayID_Touch);
 		
-		id<MTLCommandBuffer> cb = [[sharedData commandQueue] commandBufferWithUnretainedReferences];
+		id<MTLCommandBuffer> cb = [self newCommandBuffer];
 		id<MTLComputeCommandEncoder> cce = [cb computeCommandEncoder];
 		
 		// Run the video source filters and the pixel scalers
@@ -1591,7 +1591,7 @@
 		// Now that everything is set up, go ahead and draw everything.
 		id<CAMetalDrawable> layerDrawable = [self nextDrawable];
 		[colorAttachment0Desc setTexture:[layerDrawable texture]];
-		id<MTLCommandBuffer> cb = [[sharedData commandQueue] commandBufferWithUnretainedReferences];
+		id<MTLCommandBuffer> cb = [self newCommandBuffer];
 		id<MTLRenderCommandEncoder> ce = [cb renderCommandEncoderWithDescriptor:_outputRenderPassDesc];
 		
 		if (_needEncodeViewport)
@@ -1804,8 +1804,11 @@ void MacMetalFetchObject::FetchFromBufferIndex(const u8 index)
 	MacClientSharedObject *sharedViewObject = (MacClientSharedObject *)this->_clientData;
 	this->_useDirectToCPUFilterPipeline = ([sharedViewObject numberViewsUsingDirectToCPUFiltering] > 0);
 	
+	@autoreleasepool
+	{
 		[(MetalDisplayViewSharedData *)this->_clientData fetchFromBufferIndex:index];
 	}
+}
 
 void MacMetalFetchObject::_FetchNativeDisplayByID(const NDSDisplayID displayID, const u8 bufferIndex)
 {	

desmume/src/frontend/cocoa/userinterface/appDelegate.mm

@@ -183,13 +183,6 @@
 	
 	// Init the DS emulation core.
 	CocoaDSCore *newCore = [[[CocoaDSCore alloc] init] autorelease];
-	MacClientSharedObject *sharedViewObject = [[newCore cdsGPU] sharedData];
-	
-	[NSThread detachNewThreadSelector:@selector(runThread:) toTarget:sharedViewObject withObject:nil];
-	while ([sharedViewObject thread] == nil)
-	{
-		[NSThread sleepUntilDate:[NSDate dateWithTimeIntervalSinceNow:0.01]];
-	}
 	
 	// Init the DS controller.
 	[[newCore cdsController] setDelegate:emuControl];
@@ -205,13 +198,6 @@
 	[slot2WindowDelegate setHidManager:[inputManager hidManager]];
 	[slot2WindowDelegate setAutoSelectedDeviceText:[[slot2WindowDelegate deviceManager] autoSelectedDeviceName]];
 	
-	// Start up the threads for our outputs.
-	[NSThread detachNewThreadSelector:@selector(runThread:) toTarget:newSpeaker withObject:nil];
-	while ([newSpeaker thread] == nil)
-	{
-		[NSThread sleepUntilDate:[NSDate dateWithTimeIntervalSinceNow:0.01]];
-	}
-	
 	// Set up all the object controllers.
 	[cdsCoreController setContent:newCore];
 	[romInfoPanelController setContent:[CocoaDSRom romNotLoadedBindings]];