Report memory required in final render dialog.

Fix broken state of ember during failed render with strips.

Data/Version History.txt

@@ -1,3 +1,15 @@
+--User changes
+	Use actual strips count when computing memory requirements in the final render dialog.
+	Output CL_DEVICE_ADDRESS_BITS when passing --openclinfo to command line programs.
+	Warn if single/total allocations are greater than the max allowed for OpenCL in the final render dialog.
+	
+--Bug Fixes
+	Fix bad values left in an ember when a render with strips > 1 fails in the final render dialog.
+
+--Code Changes
+	Break RendererBase::MemoryRequired() into two functions, it and HistMemRequired() and return a tuple.
+	Add GlobalMemSize() and MaxAllocSize() functions in OpenCLWrapper.
+
 0.4.1.6 Beta 11/29/2014
 --User Changes
 	None.

Source/Ember/RendererBase.cpp

@@ -116,12 +116,10 @@ void RendererBase::ChangeVal(std::function<void(void)> func, eProcessAction acti
 }
 
 /// <summary>
-/// Return the amount of memory needed to render the current ember.
+/// Return the amount of memory needed for the histogram.
-/// Optionally include the memory needed for the final output image.
 /// </summary>
-/// <param name="includeFinal">If true include the memory needed for the final output image, else don't.</param>
+/// <returns>The memory required for the histogram to render the current ember</returns>
-/// <returns>The memory required to render the current ember</returns>
+size_t RendererBase::HistMemoryRequired(size_t strips)
-size_t RendererBase::MemoryRequired(size_t strips, bool includeFinal)
 {
 	bool newFilterAlloc = false;
 
@@ -130,9 +128,24 @@ size_t RendererBase::MemoryRequired(size_t strips, bool includeFinal)
 	ComputeBounds();
 
 	//Because ComputeBounds() was called, this includes gutter.
-	size_t histSize = (SuperSize() * HistBucketSize()) / strips;
+	return (SuperSize() * HistBucketSize()) / strips;
+}
 
-	return (histSize * 2) + (includeFinal ? FinalBufferSize() : 0);//Multiply hist by 2 to account for the density filtering buffer which is the same size as the histogram.
+/// <summary>
+/// Return a pair whose first member contains the amount of memory needed for the histogram,
+/// and whose second member contains the total the amount of memory needed to render the current ember.
+/// Optionally include the memory needed for the final output image in pair.second.
+/// </summary>
+/// <param name="includeFinal">If true include the memory needed for the final output image, else don't.</param>
+/// <returns>The histogram memory required in first, and the total memory required in second</returns>
+pair<size_t, size_t> RendererBase::MemoryRequired(size_t strips, bool includeFinal)
+{
+	pair<size_t, size_t> p;
+
+	p.first = HistMemoryRequired(strips);
+	p.second = (p.first * 2) + (includeFinal ? FinalBufferSize() : 0);//Multiply hist by 2 to account for the density filtering buffer which is the same size as the histogram.
+
+	return p;
 }
 
 /// <summary>

Source/Ember/RendererBase.h

@@ -98,7 +98,8 @@ public:
 
 	//Non-virtual processing functions.
 	void ChangeVal(std::function<void(void)> func, eProcessAction action);
-	size_t MemoryRequired(size_t strips, bool includeFinal);
+	size_t HistMemoryRequired(size_t strips);
+	pair<size_t, size_t> MemoryRequired(size_t strips, bool includeFinal);
 	vector<QTIsaac<ISAAC_SIZE, ISAAC_INT>> RandVec();
 	bool PrepFinalAccumVector(vector<byte>& pixels);
 

Source/EmberCL/OpenCLWrapper.cpp

@@ -554,13 +554,13 @@ int OpenCLWrapper::FindImageIndex(const string& name, bool shared)
 	{
 		for (size_t i = 0; i < m_GLImages.size(); i++)
 			if (m_GLImages[i].m_Name == name)
-				return i;
+				return int(i);
 	}
 	else
 	{
 		for (size_t i = 0; i < m_Images.size(); i++)
 			if (m_Images[i].m_Name == name)
-				return i;
+				return int(i);
 	}
 
 	return -1;
@@ -966,7 +966,7 @@ bool OpenCLWrapper::RunKernel(uint kernelIndex, uint totalGridWidth, uint totalG
 /// <param name="name">The device field/feature to query</param>
 /// <returns>The value of the field</returns>
 template<typename T>
-T OpenCLWrapper::GetInfo(size_t platform, size_t device, cl_device_info name)
+T OpenCLWrapper::GetInfo(size_t platform, size_t device, cl_device_info name) const
 {
 	T val;
 
@@ -1094,6 +1094,7 @@ string OpenCLWrapper::DumpInfo()
 			os << "CL_DEVICE_MAX_READ_IMAGE_ARGS: "		  << GetInfo<cl_uint> (platform, device, CL_DEVICE_MAX_READ_IMAGE_ARGS)		  << endl;
 			os << "CL_DEVICE_MAX_WRITE_IMAGE_ARGS: "	  << GetInfo<cl_uint> (platform, device, CL_DEVICE_MAX_WRITE_IMAGE_ARGS)	  << endl;
 			os << "CL_DEVICE_MAX_MEM_ALLOC_SIZE: "		  << GetInfo<cl_ulong>(platform, device, CL_DEVICE_MAX_MEM_ALLOC_SIZE)		  << endl;
+			os << "CL_DEVICE_ADDRESS_BITS: "			  << GetInfo<cl_uint> (platform, device, CL_DEVICE_ADDRESS_BITS)			  << endl;
 
 			os << "CL_DEVICE_GLOBAL_MEM_CACHE_TYPE: "	  << GetInfo<cl_uint> (platform, device, CL_DEVICE_GLOBAL_MEM_CACHE_TYPE)	  << endl;
 			os << "CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE: " << GetInfo<cl_uint> (platform, device, CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE) << endl;
@@ -1126,7 +1127,9 @@ bool OpenCLWrapper::Shared() const { return m_Shared; }
 cl::Context OpenCLWrapper::Context() const { return m_Context; }
 uint OpenCLWrapper::PlatformIndex() const { return m_PlatformIndex; }
 uint OpenCLWrapper::DeviceIndex() const { return m_DeviceIndex; }
+size_t OpenCLWrapper::GlobalMemSize() const { return GetInfo<cl_ulong>(PlatformIndex(), DeviceIndex(), CL_DEVICE_GLOBAL_MEM_SIZE); }
 uint OpenCLWrapper::LocalMemSize() const { return m_LocalMemSize; }
+size_t OpenCLWrapper::MaxAllocSize() const { return GetInfo<cl_ulong>(PlatformIndex(), DeviceIndex(), CL_DEVICE_MAX_MEM_ALLOC_SIZE); }
 
 /// <summary>
 /// Makes the even grid dims.

Source/EmberCL/OpenCLWrapper.h

@@ -175,7 +175,7 @@ public:
 
 	//Info.
 	template<typename T>
-	T GetInfo(size_t platform, size_t device, cl_device_info name);
+	T GetInfo(size_t platform, size_t device, cl_device_info name) const;
 	string PlatformName(size_t platform);
 	vector<string> PlatformNames();
 	string DeviceName(size_t platform, size_t device);
@@ -190,6 +190,8 @@ public:
 	uint PlatformIndex() const;
 	uint DeviceIndex() const;
 	uint LocalMemSize() const;
+	size_t GlobalMemSize() const;
+	size_t MaxAllocSize() const;
 
 	static void MakeEvenGridDims(uint blockW, uint blockH, uint& gridW, uint& gridH);
 

Source/EmberCommon/EmberCommon.h

@@ -281,7 +281,7 @@ static bool StripsRender(RendererBase* renderer, Ember<T>& ember, vector<byte>&
 	std::function<void(size_t strip)> perStripError,
 	std::function<void(Ember<T>& finalEmber)> allStripsFinished)
 {
-	bool success = true;
+	bool success = false;
 	size_t origHeight, realHeight = ember.m_FinalRasH;
 	T centerY = ember.m_CenterY;
 	T floatStripH = T(ember.m_FinalRasH) / T(strips);
@@ -321,31 +321,28 @@ static bool StripsRender(RendererBase* renderer, Ember<T>& ember, vector<byte>&
 			renderer->SetEmber(ember);//Set one final time after modifications for strips.
 		}
 
-		if ((renderer->Run(finalImage, time, 0, false, stripOffset) != RENDER_OK) || renderer->Aborted() || finalImage.empty())
+		if ((renderer->Run(finalImage, time, 0, false, stripOffset) == RENDER_OK) && !renderer->Aborted() && !finalImage.empty())
-		{
-			perStripError(strip);
-			success = false;
-			break;
-		}
-		else
 		{
 			perStripFinish(strip);
 		}
+		else
+		{
+			perStripError(strip);
+			break;
+		}
 
 		if (strip == strips - 1)
-		{
+			success = true;
+	}
+
 	//Restore the ember values to their original values.
-			if (strips > 1)
-			{
 	ember.m_Quality /= strips;
 	ember.m_FinalRasH = realHeight;
 	ember.m_CenterY = centerY;
 	renderer->SetEmber(ember);//Further processing will require the dimensions to match the original ember, so re-assign.
-			}
 
+	if (success)
 		allStripsFinished(ember);
-		}
-	}
 
 	Memset(finalImage);
 

Source/EmberRender/EmberRender.cpp

@@ -330,6 +330,7 @@ int _tmain(int argc, _TCHAR* argv[])
 	//This must be done in the application and not in the EmberCL DLL.
 #ifdef WIN32
 	_putenv_s("GPU_MAX_ALLOC_PERCENT", "100");
+	//_putenv_s("GPU_FORCE_64BIT_PTR", "1");
 #else
 	putenv(const_cast<char*>("GPU_MAX_ALLOC_PERCENT=100"));
 #endif

Source/Fractorium/FinalRenderDialog.cpp

@@ -708,10 +708,48 @@ bool FractoriumFinalRenderDialog::SetMemory()
 {
 	if (isVisible() && CreateControllerFromGUI(true))
 	{
-		pair<size_t, size_t> p = m_Controller->SyncAndComputeMemory();
+		bool error = false;
+		tuple<size_t, size_t, size_t> p = m_Controller->SyncAndComputeMemory();
+
+		ui.FinalRenderParamsTable->item(m_MemoryCellIndex, 1)->setText(ToString(get<1>(p)));
+		ui.FinalRenderParamsTable->item(m_ItersCellIndex, 1)->setText(ToString(get<2>(p)));
+
+		if (OpenCL())
+		{
+			if (!m_Wrapper.Ok() || PlatformIndex() != m_Wrapper.PlatformIndex() || DeviceIndex() != m_Wrapper.DeviceIndex())
+				m_Wrapper.Init(PlatformIndex(), DeviceIndex());
+
+			if (m_Wrapper.Ok())
+			{
+				size_t histSize = get<0>(p);
+				size_t totalSize = get<1>(p);
+				size_t maxAlloc = m_Wrapper.MaxAllocSize();
+				size_t totalAvail = m_Wrapper.GlobalMemSize();
+				QString s;
+
+				if (histSize > maxAlloc)
+				{
+					s = "Histogram/Accumulator memory size of " + ToString(histSize) +
+						" is greater than the max OpenCL allocation size of " + ToString(maxAlloc);
+				}
+
+				if (totalSize > totalAvail)
+				{
+					s += "\n\nTotal required memory size of " + ToString(totalSize) +
+						" is greater than the max OpenCL available memory of " + ToString(totalAvail);
+				}
+
+				if (!s.isEmpty())
+				{
+					error = true;
+					ui.FinalRenderTextOutput->setText(s + ".\n\nRendering will most likely fail.");
+				}
+			}
+		}
+
+		if (!error)
+			ui.FinalRenderTextOutput->clear();
 
-		ui.FinalRenderParamsTable->item(m_MemoryCellIndex, 1)->setText(ToString(p.first));
-		ui.FinalRenderParamsTable->item(m_ItersCellIndex, 1)->setText(ToString(p.second));
 		return true;
 	}
 

Source/Fractorium/FinalRenderEmberController.cpp

@@ -151,8 +151,9 @@ FinalRenderEmberController<T>::FinalRenderEmberController(FractoriumFinalRenderD
 	{
 		m_Run = true;
 		m_TotalTimer.Tic();//Begin timing for progress of all operations.
-		size_t i;
 		m_GuiState = m_FinalRenderDialog->State();//Cache render settings from the GUI before running.
+		
+		size_t i;
 		bool doAll = m_GuiState.m_DoAll && m_EmberFile.Size() > 1;
 		uint currentStripForProgress = 0;//Sort of a hack to get the strip value to the progress function.
 		QString path = doAll ? ComposePath(QString::fromStdString(m_EmberFile.m_Embers[0].m_Name)) : ComposePath(Name());
@@ -566,8 +567,9 @@ void FinalRenderEmberController<T>::ResetProgress(bool total)
 /// </summary>
 /// <returns>If successful, memory required in bytes, else zero.</returns>
 template <typename T>
-pair<size_t, size_t> FinalRenderEmberController<T>::SyncAndComputeMemory()
+tuple<size_t, size_t, size_t> FinalRenderEmberController<T>::SyncAndComputeMemory()
 {
+	size_t iterCount;
 	pair<size_t, size_t> p(0, 0);
 
 	if (m_Renderer.get())
@@ -587,11 +589,12 @@ pair<size_t, size_t> FinalRenderEmberController<T>::SyncAndComputeMemory()
 		m_Renderer->ComputeCamera();
 		CancelPreviewRender();
 		m_FinalPreviewRenderFunc();
-		p.first = m_Renderer->MemoryRequired(m_FinalRenderDialog->Strips(), true);
+
-		p.second = m_Renderer->TotalIterCount(strips);
+		p = m_Renderer->MemoryRequired(strips, true);
+		iterCount = m_Renderer->TotalIterCount(strips);
 	}
 
-	return p;
+	return tuple<size_t, size_t, size_t>(p.first, p.second, iterCount);
 }
 
 /// <summary>

Source/Fractorium/FinalRenderEmberController.h

@@ -63,7 +63,7 @@ public:
 	virtual void SyncGuiToEmbers(size_t widthOverride = 0, size_t heightOverride = 0) { }
 	virtual void SyncCurrentToSizeSpinners(bool scale, bool size) { }
 	virtual void ResetProgress(bool total = true) { }
-	virtual pair<size_t, size_t> SyncAndComputeMemory() { return pair<size_t, size_t>(0, 0); }
+	virtual tuple<size_t, size_t, size_t> SyncAndComputeMemory() { return tuple<size_t, size_t, size_t>(0, 0, 0); }
 	virtual double OriginalAspect() { return 1; }
 	virtual QString ComposePath(const QString& name) { return ""; }
 
@@ -121,7 +121,7 @@ public:
 	virtual void SyncGuiToEmbers(size_t widthOverride = 0, size_t heightOverride = 0) override;
 	virtual void SyncCurrentToSizeSpinners(bool scale, bool size) override;
 	virtual void ResetProgress(bool total = true)  override;
-	virtual pair<size_t, size_t> SyncAndComputeMemory() override;
+	virtual tuple<size_t, size_t, size_t> SyncAndComputeMemory() override;
 	virtual double OriginalAspect() override { return double(m_Ember->m_OrigFinalRasW) / m_Ember->m_OrigFinalRasH; }
 	virtual QString Name() const override { return QString::fromStdString(m_Ember->m_Name); }
 	virtual QString ComposePath(const QString& name) override;