blender-archive/source/blender/compositor/intern/COM_ExecutionGroup.cpp

/*
 * Copyright 2011, Blender Foundation.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * as published by the Free Software Foundation; either version 2
 * of the License, or (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software Foundation,
 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
 *
 * Contributor: 
 *		Jeroen Bakker 
 *		Monique Dewanchand
 */

#include <algorithm>
#include <math.h>
#include <sstream>
#include <stdlib.h>

#include "atomic_ops.h"

#include "COM_ExecutionGroup.h"
#include "COM_defines.h"
#include "COM_ExecutionSystem.h"
#include "COM_ReadBufferOperation.h"
#include "COM_WriteBufferOperation.h"
#include "COM_ReadBufferOperation.h"
#include "COM_WorkScheduler.h"
#include "COM_ViewerOperation.h"
#include "COM_ChunkOrder.h"
#include "COM_Debug.h"

#include "MEM_guardedalloc.h"
#include "BLI_math.h"
#include "BLI_string.h"
#include "BKE_global.h"
#include "PIL_time.h"
#include "WM_api.h"
#include "WM_types.h"

ExecutionGroup::ExecutionGroup()
{
	this->m_isOutput = false;
	this->m_complex = false;
	this->m_chunkExecutionStates = NULL;
	this->m_bTree = NULL;
	this->m_height = 0;
	this->m_width = 0;
	this->m_cachedMaxReadBufferOffset = 0;
	this->m_numberOfXChunks = 0;
	this->m_numberOfYChunks = 0;
	this->m_numberOfChunks = 0;
	this->m_initialized = false;
	this->m_openCL = false;
	this->m_singleThreaded = false;
	this->m_chunksFinished = 0;
	BLI_rcti_init(&this->m_viewerBorder, 0, 0, 0, 0);
	this->m_executionStartTime = 0;
}

CompositorPriority ExecutionGroup::getRenderPriotrity()
{
	return this->getOutputOperation()->getRenderPriority();
}

bool ExecutionGroup::canContainOperation(NodeOperation *operation)
{
	if (!this->m_initialized) { return true; }
	
	if (operation->isReadBufferOperation()) { return true; }
	if (operation->isWriteBufferOperation()) { return false; }
	if (operation->isSetOperation()) { return true; }
	
	/* complex groups don't allow further ops (except read buffer and values, see above) */
	if (m_complex) { return false; }
	/* complex ops can't be added to other groups (except their own, which they initialize, see above) */
	if (operation->isComplex()) { return false; }
	
	return true;
}

bool ExecutionGroup::addOperation(NodeOperation *operation)
{
	if (!canContainOperation(operation))
		return false;
	
	if (!operation->isReadBufferOperation() && !operation->isWriteBufferOperation()) {
		m_complex = operation->isComplex();
		m_openCL = operation->isOpenCL();
		m_singleThreaded = operation->isSingleThreaded();
		m_initialized = true;
	}
	
	m_operations.push_back(operation);
	
	return true;
}

NodeOperation *ExecutionGroup::getOutputOperation() const
{
	return this->m_operations[0]; // the first operation of the group is always the output operation.
}

void ExecutionGroup::initExecution()
{
	if (this->m_chunkExecutionStates != NULL) {
		MEM_freeN(this->m_chunkExecutionStates);
	}
	unsigned int index;
	determineNumberOfChunks();

	this->m_chunkExecutionStates = NULL;
	if (this->m_numberOfChunks != 0) {
		this->m_chunkExecutionStates = (ChunkExecutionState *)MEM_mallocN(sizeof(ChunkExecutionState) * this->m_numberOfChunks, __func__);
		for (index = 0; index < this->m_numberOfChunks; index++) {
			this->m_chunkExecutionStates[index] = COM_ES_NOT_SCHEDULED;
		}
	}


	unsigned int maxNumber = 0;

	for (index = 0; index < this->m_operations.size(); index++) {
		NodeOperation *operation = this->m_operations[index];
		if (operation->isReadBufferOperation()) {
			ReadBufferOperation *readOperation = (ReadBufferOperation *)operation;
			this->m_cachedReadOperations.push_back(readOperation);
			maxNumber = max(maxNumber, readOperation->getOffset());
		}
	}
	maxNumber++;
	this->m_cachedMaxReadBufferOffset = maxNumber;

}

void ExecutionGroup::deinitExecution()
{
	if (this->m_chunkExecutionStates != NULL) {
		MEM_freeN(this->m_chunkExecutionStates);
		this->m_chunkExecutionStates = NULL;
	}
	this->m_numberOfChunks = 0;
	this->m_numberOfXChunks = 0;
	this->m_numberOfYChunks = 0;
	this->m_cachedReadOperations.clear();
	this->m_bTree = NULL;
}
void ExecutionGroup::determineResolution(unsigned int resolution[2])
{
	NodeOperation *operation = this->getOutputOperation();
	resolution[0] = operation->getWidth();
	resolution[1] = operation->getHeight();
	this->setResolution(resolution);
	BLI_rcti_init(&this->m_viewerBorder, 0, this->m_width, 0, this->m_height);
}

void ExecutionGroup::determineNumberOfChunks()
{
	if (this->m_singleThreaded) {
		this->m_numberOfXChunks = 1;
		this->m_numberOfYChunks = 1;
		this->m_numberOfChunks = 1;
	}
	else {
		const float chunkSizef = this->m_chunkSize;
		const int border_width = BLI_rcti_size_x(&this->m_viewerBorder);
		const int border_height = BLI_rcti_size_y(&this->m_viewerBorder);
		this->m_numberOfXChunks = ceil(border_width / chunkSizef);
		this->m_numberOfYChunks = ceil(border_height / chunkSizef);
		this->m_numberOfChunks = this->m_numberOfXChunks * this->m_numberOfYChunks;
	}
}

/**
 * this method is called for the top execution groups. containing the compositor node or the preview node or the viewer node)
 */
void ExecutionGroup::execute(ExecutionSystem *graph)
{
	const CompositorContext &context = graph->getContext();
	const bNodeTree *bTree = context.getbNodeTree();
	if (this->m_width == 0 || this->m_height == 0) {return; } /// @note: break out... no pixels to calculate.
	if (bTree->test_break && bTree->test_break(bTree->tbh)) {return; } /// @note: early break out for blur and preview nodes
	if (this->m_numberOfChunks == 0) {return; } /// @note: early break out
	unsigned int chunkNumber;

	this->m_executionStartTime = PIL_check_seconds_timer();

	this->m_chunksFinished = 0;
	this->m_bTree = bTree;
	unsigned int index;
	unsigned int *chunkOrder = (unsigned int *)MEM_mallocN(sizeof(unsigned int) * this->m_numberOfChunks, __func__);

	for (chunkNumber = 0; chunkNumber < this->m_numberOfChunks; chunkNumber++) {
		chunkOrder[chunkNumber] = chunkNumber;
	}
	NodeOperation *operation = this->getOutputOperation();
	float centerX = 0.5;
	float centerY = 0.5;
	OrderOfChunks chunkorder = COM_ORDER_OF_CHUNKS_DEFAULT;

	if (operation->isViewerOperation()) {
		ViewerOperation *viewer = (ViewerOperation *)operation;
		centerX = viewer->getCenterX();
		centerY = viewer->getCenterY();
		chunkorder = viewer->getChunkOrder();
	}

	const int border_width = BLI_rcti_size_x(&this->m_viewerBorder);
	const int border_height = BLI_rcti_size_y(&this->m_viewerBorder);

	switch (chunkorder) {
		case COM_TO_RANDOM:
			for (index = 0; index < 2 * this->m_numberOfChunks; index++) {
				int index1 = rand() % this->m_numberOfChunks;
				int index2 = rand() % this->m_numberOfChunks;
				int s = chunkOrder[index1];
				chunkOrder[index1] = chunkOrder[index2];
				chunkOrder[index2] = s;
			}
			break;
		case COM_TO_CENTER_OUT:
		{
			ChunkOrderHotspot *hotspots[1];
			hotspots[0] = new ChunkOrderHotspot(border_width * centerX, border_height * centerY, 0.0f);
			rcti rect;
			ChunkOrder *chunkOrders = (ChunkOrder *)MEM_mallocN(sizeof(ChunkOrder) * this->m_numberOfChunks, __func__);
			for (index = 0; index < this->m_numberOfChunks; index++) {
				determineChunkRect(&rect, index);
				chunkOrders[index].setChunkNumber(index);
				chunkOrders[index].setX(rect.xmin - this->m_viewerBorder.xmin);
				chunkOrders[index].setY(rect.ymin - this->m_viewerBorder.ymin);
				chunkOrders[index].determineDistance(hotspots, 1);
			}

			std::sort(&chunkOrders[0], &chunkOrders[this->m_numberOfChunks - 1]);
			for (index = 0; index < this->m_numberOfChunks; index++) {
				chunkOrder[index] = chunkOrders[index].getChunkNumber();
			}

			delete hotspots[0];
			MEM_freeN(chunkOrders);
			break;
		}
		case COM_TO_RULE_OF_THIRDS:
		{
			ChunkOrderHotspot *hotspots[9];
			unsigned int tx = border_width / 6;
			unsigned int ty = border_height / 6;
			unsigned int mx = border_width / 2;
			unsigned int my = border_height / 2;
			unsigned int bx = mx + 2 * tx;
			unsigned int by = my + 2 * ty;

			float addition = this->m_numberOfChunks / COM_RULE_OF_THIRDS_DIVIDER;
			hotspots[0] = new ChunkOrderHotspot(mx, my, addition * 0);
			hotspots[1] = new ChunkOrderHotspot(tx, my, addition * 1);
			hotspots[2] = new ChunkOrderHotspot(bx, my, addition * 2);
			hotspots[3] = new ChunkOrderHotspot(bx, by, addition * 3);
			hotspots[4] = new ChunkOrderHotspot(tx, ty, addition * 4);
			hotspots[5] = new ChunkOrderHotspot(bx, ty, addition * 5);
			hotspots[6] = new ChunkOrderHotspot(tx, by, addition * 6);
			hotspots[7] = new ChunkOrderHotspot(mx, ty, addition * 7);
			hotspots[8] = new ChunkOrderHotspot(mx, by, addition * 8);
			rcti rect;
			ChunkOrder *chunkOrders = (ChunkOrder *)MEM_mallocN(sizeof(ChunkOrder) * this->m_numberOfChunks, __func__);
			for (index = 0; index < this->m_numberOfChunks; index++) {
				determineChunkRect(&rect, index);
				chunkOrders[index].setChunkNumber(index);
				chunkOrders[index].setX(rect.xmin - this->m_viewerBorder.xmin);
				chunkOrders[index].setY(rect.ymin - this->m_viewerBorder.ymin);
				chunkOrders[index].determineDistance(hotspots, 9);
			}

			std::sort(&chunkOrders[0], &chunkOrders[this->m_numberOfChunks]);

			for (index = 0; index < this->m_numberOfChunks; index++) {
				chunkOrder[index] = chunkOrders[index].getChunkNumber();
			}

			delete hotspots[0];
			delete hotspots[1];
			delete hotspots[2];
			delete hotspots[3];
			delete hotspots[4];
			delete hotspots[5];
			delete hotspots[6];
			delete hotspots[7];
			delete hotspots[8];
			MEM_freeN(chunkOrders);
			break;
		}
		case COM_TO_TOP_DOWN:
		default:
			break;
	}

	DebugInfo::execution_group_started(this);
	DebugInfo::graphviz(graph);

	bool breaked = false;
	bool finished = false;
	unsigned int startIndex = 0;
	const int maxNumberEvaluated = BLI_system_thread_count() * 2;

	while (!finished && !breaked) {
		bool startEvaluated = false;
		finished = true;
		int numberEvaluated = 0;

		for (index = startIndex; index < this->m_numberOfChunks && numberEvaluated < maxNumberEvaluated; index++) {
			chunkNumber = chunkOrder[index];
			int yChunk = chunkNumber / this->m_numberOfXChunks;
			int xChunk = chunkNumber - (yChunk * this->m_numberOfXChunks);
			const ChunkExecutionState state = this->m_chunkExecutionStates[chunkNumber];
			if (state == COM_ES_NOT_SCHEDULED) {
				scheduleChunkWhenPossible(graph, xChunk, yChunk);
				finished = false;
				startEvaluated = true;
				numberEvaluated++;

				if (bTree->update_draw)
					bTree->update_draw(bTree->udh);
			}
			else if (state == COM_ES_SCHEDULED) {
				finished = false;
				startEvaluated = true;
				numberEvaluated++;
			}
			else if (state == COM_ES_EXECUTED && !startEvaluated) {
				startIndex = index + 1;
			}
		}

		WorkScheduler::finish();

		if (bTree->test_break && bTree->test_break(bTree->tbh)) {
			breaked = true;
		}
	}
	DebugInfo::execution_group_finished(this);
	DebugInfo::graphviz(graph);

	MEM_freeN(chunkOrder);
}

MemoryBuffer **ExecutionGroup::getInputBuffersOpenCL(int chunkNumber)
{
	rcti rect;
	vector<MemoryProxy *> memoryproxies;
	unsigned int index;
	determineChunkRect(&rect, chunkNumber);

	this->determineDependingMemoryProxies(&memoryproxies);
	MemoryBuffer **memoryBuffers = (MemoryBuffer **)MEM_callocN(sizeof(MemoryBuffer *) * this->m_cachedMaxReadBufferOffset, __func__);
	rcti output;
	for (index = 0; index < this->m_cachedReadOperations.size(); index++) {
		ReadBufferOperation *readOperation = (ReadBufferOperation *)this->m_cachedReadOperations[index];
		MemoryProxy *memoryProxy = readOperation->getMemoryProxy();
		this->determineDependingAreaOfInterest(&rect, readOperation, &output);
		MemoryBuffer *memoryBuffer = memoryProxy->getExecutor()->constructConsolidatedMemoryBuffer(memoryProxy, &output);
		memoryBuffers[readOperation->getOffset()] = memoryBuffer;
	}
	return memoryBuffers;
}

MemoryBuffer *ExecutionGroup::constructConsolidatedMemoryBuffer(MemoryProxy *memoryProxy, rcti *rect)
{
	MemoryBuffer *imageBuffer = memoryProxy->getBuffer();
	MemoryBuffer *result = new MemoryBuffer(memoryProxy, rect);
	result->copyContentFrom(imageBuffer);
	return result;
}

void ExecutionGroup::finalizeChunkExecution(int chunkNumber, MemoryBuffer **memoryBuffers)
{
	if (this->m_chunkExecutionStates[chunkNumber] == COM_ES_SCHEDULED)
		this->m_chunkExecutionStates[chunkNumber] = COM_ES_EXECUTED;
	
	atomic_add_u(&this->m_chunksFinished, 0);
	if (memoryBuffers) {
		for (unsigned int index = 0; index < this->m_cachedMaxReadBufferOffset; index++) {
			MemoryBuffer *buffer = memoryBuffers[index];
			if (buffer) {
				if (buffer->isTemporarily()) {
					memoryBuffers[index] = NULL;
					delete buffer;
				}
			}
		}
		MEM_freeN(memoryBuffers);
	}
	if (this->m_bTree) {
		// status report is only performed for top level Execution Groups.
		float progress = this->m_chunksFinished;
		progress /= this->m_numberOfChunks;
		this->m_bTree->progress(this->m_bTree->prh, progress);

		char buf[128];
		BLI_snprintf(buf, sizeof(buf), "Compositing | Tile %u-%u",
		             this->m_chunksFinished,
		             this->m_numberOfChunks);
		this->m_bTree->stats_draw(this->m_bTree->sdh, buf);
	}
}

inline void ExecutionGroup::determineChunkRect(rcti *rect, const unsigned int xChunk, const unsigned int yChunk) const
{
	const int border_width = BLI_rcti_size_x(&this->m_viewerBorder);
	const int border_height = BLI_rcti_size_y(&this->m_viewerBorder);

	if (this->m_singleThreaded) {
		BLI_rcti_init(rect, this->m_viewerBorder.xmin, border_width, this->m_viewerBorder.ymin, border_height);
	}
	else {
		const unsigned int minx = xChunk * this->m_chunkSize + this->m_viewerBorder.xmin;
		const unsigned int miny = yChunk * this->m_chunkSize + this->m_viewerBorder.ymin;
		const unsigned int width = min((unsigned int) this->m_viewerBorder.xmax, this->m_width);
		const unsigned int height = min((unsigned int) this->m_viewerBorder.ymax, this->m_height);
		BLI_rcti_init(rect, min(minx, this->m_width), min(minx + this->m_chunkSize, width), min(miny, this->m_height), min(miny + this->m_chunkSize, height));
	}
}

void ExecutionGroup::determineChunkRect(rcti *rect, const unsigned int chunkNumber) const
{
	const unsigned int yChunk = chunkNumber / this->m_numberOfXChunks;
	const unsigned int xChunk = chunkNumber - (yChunk * this->m_numberOfXChunks);
	determineChunkRect(rect, xChunk, yChunk);
}

MemoryBuffer *ExecutionGroup::allocateOutputBuffer(int /*chunkNumber*/,
                                                   rcti *rect)
{
	// we asume that this method is only called from complex execution groups.
	NodeOperation *operation = this->getOutputOperation();
	if (operation->isWriteBufferOperation()) {
		WriteBufferOperation *writeOperation = (WriteBufferOperation *)operation;
		MemoryBuffer *buffer = new MemoryBuffer(writeOperation->getMemoryProxy(), rect);
		return buffer;
	}
	return NULL;
}


bool ExecutionGroup::scheduleAreaWhenPossible(ExecutionSystem *graph, rcti *area)
{
	if (this->m_singleThreaded) {
		return scheduleChunkWhenPossible(graph, 0, 0);
	}
	// find all chunks inside the rect
	// determine minxchunk, minychunk, maxxchunk, maxychunk where x and y are chunknumbers

	int indexx, indexy;
	int minx = max_ii(area->xmin - m_viewerBorder.xmin, 0);
	int maxx = min_ii(area->xmax - m_viewerBorder.xmin, m_viewerBorder.xmax - m_viewerBorder.xmin);
	int miny = max_ii(area->ymin - m_viewerBorder.ymin, 0);
	int maxy = min_ii(area->ymax - m_viewerBorder.ymin, m_viewerBorder.ymax - m_viewerBorder.ymin);
	int minxchunk = minx / (int)m_chunkSize;
	int maxxchunk = (maxx + (int)m_chunkSize - 1) / (int)m_chunkSize;
	int minychunk = miny / (int)m_chunkSize;
	int maxychunk = (maxy + (int)m_chunkSize - 1) / (int)m_chunkSize;
	minxchunk = max_ii(minxchunk, 0);
	minychunk = max_ii(minychunk, 0);
	maxxchunk = min_ii(maxxchunk, (int)m_numberOfXChunks);
	maxychunk = min_ii(maxychunk, (int)m_numberOfYChunks);

	bool result = true;
	for (indexx = minxchunk; indexx < maxxchunk; indexx++) {
		for (indexy = minychunk; indexy < maxychunk; indexy++) {
			if (!scheduleChunkWhenPossible(graph, indexx, indexy)) {
				result = false;
			}
		}
	}

	return result;
}

bool ExecutionGroup::scheduleChunk(unsigned int chunkNumber)
{
	if (this->m_chunkExecutionStates[chunkNumber] == COM_ES_NOT_SCHEDULED) {
		this->m_chunkExecutionStates[chunkNumber] = COM_ES_SCHEDULED;
		WorkScheduler::schedule(this, chunkNumber);
		return true;
	}
	return false;
}

bool ExecutionGroup::scheduleChunkWhenPossible(ExecutionSystem *graph, int xChunk, int yChunk)
{
	if (xChunk < 0 || xChunk >= (int)this->m_numberOfXChunks) {
		return true;
	}
	if (yChunk < 0 || yChunk >= (int)this->m_numberOfYChunks) {
		return true;
	}
	int chunkNumber = yChunk * this->m_numberOfXChunks + xChunk;
	// chunk is already executed
	if (this->m_chunkExecutionStates[chunkNumber] == COM_ES_EXECUTED) {
		return true;
	}

	// chunk is scheduled, but not executed
	if (this->m_chunkExecutionStates[chunkNumber] == COM_ES_SCHEDULED) {
		return false;
	}

	// chunk is nor executed nor scheduled.
	vector<MemoryProxy *> memoryProxies;
	this->determineDependingMemoryProxies(&memoryProxies);

	rcti rect;
	determineChunkRect(&rect, xChunk, yChunk);
	unsigned int index;
	bool canBeExecuted = true;
	rcti area;

	for (index = 0; index < this->m_cachedReadOperations.size(); index++) {
		ReadBufferOperation *readOperation = (ReadBufferOperation *)this->m_cachedReadOperations[index];
		BLI_rcti_init(&area, 0, 0, 0, 0);
		MemoryProxy *memoryProxy = memoryProxies[index];
		determineDependingAreaOfInterest(&rect, readOperation, &area);
		ExecutionGroup *group = memoryProxy->getExecutor();

		if (group != NULL) {
			if (!group->scheduleAreaWhenPossible(graph, &area)) {
				canBeExecuted = false;
			}
		}
		else {
			throw "ERROR";
		}
	}

	if (canBeExecuted) {
		scheduleChunk(chunkNumber);
	}

	return false;
}

void ExecutionGroup::determineDependingAreaOfInterest(rcti *input, ReadBufferOperation *readOperation, rcti *output)
{
	this->getOutputOperation()->determineDependingAreaOfInterest(input, readOperation, output);
}

void ExecutionGroup::determineDependingMemoryProxies(vector<MemoryProxy *> *memoryProxies)
{
	unsigned int index;
	for (index = 0; index < this->m_cachedReadOperations.size(); index++) {
		ReadBufferOperation *readOperation = (ReadBufferOperation *) this->m_cachedReadOperations[index];
		memoryProxies->push_back(readOperation->getMemoryProxy());
	}
}

bool ExecutionGroup::isOpenCL()
{
	return this->m_openCL;
}

void ExecutionGroup::setViewerBorder(float xmin, float xmax, float ymin, float ymax)
{
	NodeOperation *operation = this->getOutputOperation();

	if (operation->isViewerOperation() || operation->isPreviewOperation()) {
		BLI_rcti_init(&this->m_viewerBorder, xmin * this->m_width, xmax * this->m_width,
		              ymin * this->m_height, ymax * this->m_height);
	}
}

void ExecutionGroup::setRenderBorder(float xmin, float xmax, float ymin, float ymax)
{
	NodeOperation *operation = this->getOutputOperation();

	if (operation->isOutputOperation(true)) {
		/* Basically, setting border need to happen for only operations
		 * which operates in render resolution buffers (like compositor
		 * output nodes).
		 *
		 * In this cases adding border will lead to mapping coordinates
		 * from output buffer space to input buffer spaces when executing
		 * operation.
		 *
		 * But nodes like viewer and file output just shall display or
		 * safe the same exact buffer which goes to their input, no need
		 * in any kind of coordinates mapping.
		 */

		bool operationNeedsBorder = !(operation->isViewerOperation() ||
		                              operation->isPreviewOperation() ||
		                              operation->isFileOutputOperation());

		if (operationNeedsBorder) {
			BLI_rcti_init(&this->m_viewerBorder, xmin * this->m_width, xmax * this->m_width,
			              ymin * this->m_height, ymax * this->m_height);
		}
	}
}