Cycles: merge of changes from tomato branch.

Regular rendering now works tiled, and supports save buffers to save memory
during render and cache render results.

Brick texture node by Thomas.
http://wiki.blender.org/index.php/Doc:2.6/Manual/Render/Cycles/Nodes/Textures#Brick_Texture

Image texture Blended Box Mapping.
http://wiki.blender.org/index.php/Doc:2.6/Manual/Render/Cycles/Nodes/Textures#Image_Texture
http://mango.blender.org/production/blended_box/

Various bug fixes by Sergey and Campbell.
* Fix for reading freed memory in some node setups.
* Fix incorrect memory read when synchronizing mesh motion.
* Fix crash appearing when direct light usage is different on different layers.
* Fix for vector pass gives wrong result in some circumstances.
* Fix for wrong resolution used for rendering Render Layer node.
* Option to cancel rendering when doing initial synchronization.
* No more texture limit when using CPU render.
* Many fixes for new tiled rendering.

9 files changed, 412 insertions, 166 deletions

diff --git a/intern/cycles/device/CMakeLists.txt b/intern/cycles/device/CMakeLists.txt
index 17072d230bb..6038abd815e 100644
--- a/intern/cycles/device/CMakeLists.txt
+++ b/intern/cycles/device/CMakeLists.txt
@@ -17,6 +17,7 @@ set(SRC
 	device_multi.cpp
 	device_network.cpp
 	device_opencl.cpp
+	device_task.cpp
 )
 
 set(SRC_HEADERS
@@ -24,6 +25,7 @@ set(SRC_HEADERS
 	device_memory.h
 	device_intern.h
 	device_network.h
+	device_task.h
 )
 
 add_definitions(-DGLEW_STATIC)
diff --git a/intern/cycles/device/device.cpp b/intern/cycles/device/device.cpp
index 33040f287d1..9a4d364a9b8 100644
--- a/intern/cycles/device/device.cpp
+++ b/intern/cycles/device/device.cpp
@@ -33,65 +33,6 @@
 
 CCL_NAMESPACE_BEGIN
 
-/* Device Task */
-
-DeviceTask::DeviceTask(Type type_)
-: type(type_), x(0), y(0), w(0), h(0), rng_state(0), rgba(0), buffer(0),
-  sample(0), resolution(0),
-  shader_input(0), shader_output(0),
-  shader_eval_type(0), shader_x(0), shader_w(0)
-{
-}
-
-void DeviceTask::split_max_size(list<DeviceTask>& tasks, int max_size)
-{
-	int num;
-
-	if(type == SHADER) {
-		num = (shader_w + max_size - 1)/max_size;
-	}
-	else {
-		max_size = max(1, max_size/w);
-		num = (h + max_size - 1)/max_size;
-	}
-
-	split(tasks, num);
-}
-
-void DeviceTask::split(list<DeviceTask>& tasks, int num)
-{
-	if(type == SHADER) {
-		num = min(shader_w, num);
-
-		for(int i = 0; i < num; i++) {
-			int tx = shader_x + (shader_w/num)*i;
-			int tw = (i == num-1)? shader_w - i*(shader_w/num): shader_w/num;
-
-			DeviceTask task = *this;
-
-			task.shader_x = tx;
-			task.shader_w = tw;
-
-			tasks.push_back(task);
-		}
-	}
-	else {
-		num = min(h, num);
-
-		for(int i = 0; i < num; i++) {
-			int ty = y + (h/num)*i;
-			int th = (i == num-1)? h - i*(h/num): h/num;
-
-			DeviceTask task = *this;
-
-			task.y = ty;
-			task.h = th;
-
-			tasks.push_back(task);
-		}
-	}
-}
-
 /* Device */
 
 void Device::pixels_alloc(device_memory& mem)
diff --git a/intern/cycles/device/device.h b/intern/cycles/device/device.h
index b17abac2a1b..2ee2e044618 100644
--- a/intern/cycles/device/device.h
+++ b/intern/cycles/device/device.h
@@ -22,10 +22,10 @@
 #include <stdlib.h>
 
 #include "device_memory.h"
+#include "device_task.h"
 
 #include "util_list.h"
 #include "util_string.h"
-#include "util_task.h"
 #include "util_thread.h"
 #include "util_types.h"
 #include "util_vector.h"
@@ -33,6 +33,7 @@
 CCL_NAMESPACE_BEGIN
 
 class Progress;
+class RenderTile;
 
 /* Device Types */
 
@@ -67,32 +68,6 @@ public:
 	}
 };
 
-/* Device Task */
-
-class DeviceTask : public Task {
-public:
-	typedef enum { PATH_TRACE, TONEMAP, SHADER } Type;
-	Type type;
-
-	int x, y, w, h;
-	device_ptr rng_state;
-	device_ptr rgba;
-	device_ptr buffer;
-	int sample;
-	int resolution;
-	int offset, stride;
-
-	device_ptr shader_input;
-	device_ptr shader_output;
-	int shader_eval_type;
-	int shader_x, shader_w;
-
-	DeviceTask(Type type = PATH_TRACE);
-
-	void split(list<DeviceTask>& tasks, int num);
-	void split_max_size(list<DeviceTask>& tasks, int max_size);
-};
-
 /* Device */
 
 class Device {
@@ -150,6 +125,10 @@ public:
 	void server_run();
 #endif
 
+	/* multi device */
+	virtual void map_tile(Device *sub_device, RenderTile& tile) {}
+	virtual int device_number(Device *sub_device) { return 0; }
+
 	/* static */
 	static Device *create(DeviceInfo& info, bool background = true, int threads = 0);
 
diff --git a/intern/cycles/device/device_cpu.cpp b/intern/cycles/device/device_cpu.cpp
index 070b20aec49..4c54671b0d0 100644
--- a/intern/cycles/device/device_cpu.cpp
+++ b/intern/cycles/device/device_cpu.cpp
@@ -27,6 +27,8 @@
 
 #include "osl_shader.h"
 
+#include "buffers.h"
+
 #include "util_debug.h"
 #include "util_foreach.h"
 #include "util_function.h"
@@ -141,28 +143,56 @@ public:
 			OSLShader::thread_init(kg);
 #endif
 
-#ifdef WITH_OPTIMIZED_KERNEL
-		if(system_cpu_support_optimized()) {
-			for(int y = task.y; y < task.y + task.h; y++) {
-				for(int x = task.x; x < task.x + task.w; x++)
-					kernel_cpu_optimized_path_trace(kg, (float*)task.buffer, (unsigned int*)task.rng_state,
-						task.sample, x, y, task.offset, task.stride);
+		RenderTile tile;
+		
+		while(task.acquire_tile(this, tile)) {
+			float *render_buffer = (float*)tile.buffer;
+			uint *rng_state = (uint*)tile.rng_state;
+			int start_sample = tile.start_sample;
+			int end_sample = tile.start_sample + tile.num_samples;
 
-				if(task_pool.cancelled())
-					break;
+#ifdef WITH_OPTIMIZED_KERNEL
+			if(system_cpu_support_optimized()) {
+				for(int sample = start_sample; sample < end_sample; sample++) {
+					if (task.get_cancel() || task_pool.cancelled())
+						break;
+
+					for(int y = tile.y; y < tile.y + tile.h; y++) {
+						for(int x = tile.x; x < tile.x + tile.w; x++) {
+							kernel_cpu_optimized_path_trace(kg, render_buffer, rng_state,
+								sample, x, y, tile.offset, tile.stride);
+						}
+					}
+
+					tile.sample = sample + 1;
+
+					task.update_progress(tile);
+				}
 			}
-		}
-		else
+			else
 #endif
-		{
-			for(int y = task.y; y < task.y + task.h; y++) {
-				for(int x = task.x; x < task.x + task.w; x++)
-					kernel_cpu_path_trace(kg, (float*)task.buffer, (unsigned int*)task.rng_state,
-						task.sample, x, y, task.offset, task.stride);
-
-				if(task_pool.cancelled())
-					break;
+			{
+				for(int sample = start_sample; sample < end_sample; sample++) {
+					if (task.get_cancel() || task_pool.cancelled())
+						break;
+
+					for(int y = tile.y; y < tile.y + tile.h; y++) {
+						for(int x = tile.x; x < tile.x + tile.w; x++) {
+							kernel_cpu_path_trace(kg, render_buffer, rng_state,
+								sample, x, y, tile.offset, tile.stride);
+						}
+					}
+
+					tile.sample = sample + 1;
+
+					task.update_progress(tile);
+				}
 			}
+
+			task.release_tile(tile);
+
+			if(task_pool.cancelled())
+				break;
 		}
 
 #ifdef WITH_OSL
@@ -228,8 +258,7 @@ public:
 		/* split task into smaller ones, more than number of threads for uneven
 		 * workloads where some parts of the image render slower than others */
 		list<DeviceTask> tasks;
-
-		task.split(tasks, TaskScheduler::num_threads()*10);
+		task.split(tasks, TaskScheduler::num_threads()+1);
 
 		foreach(DeviceTask& task, tasks)
 			task_pool.push(new CPUDeviceTask(this, task));
diff --git a/intern/cycles/device/device_cuda.cpp b/intern/cycles/device/device_cuda.cpp
index 357f99145b2..c8dcfdc2f3d 100644
--- a/intern/cycles/device/device_cuda.cpp
+++ b/intern/cycles/device/device_cuda.cpp
@@ -23,6 +23,8 @@
 #include "device.h"
 #include "device_intern.h"
 
+#include "buffers.h"
+
 #include "util_cuda.h"
 #include "util_debug.h"
 #include "util_map.h"
@@ -37,6 +39,7 @@ CCL_NAMESPACE_BEGIN
 class CUDADevice : public Device
 {
 public:
+	TaskPool task_pool;
 	CUdevice cuDevice;
 	CUcontext cuContext;
 	CUmodule cuModule;
@@ -192,6 +195,8 @@ public:
 
 	~CUDADevice()
 	{
+		task_pool.stop();
+
 		cuda_push_context();
 		cuda_assert(cuCtxDetach(cuContext))
 	}
@@ -466,13 +471,13 @@ public:
 		}
 	}
 
-	void path_trace(DeviceTask& task)
+	void path_trace(RenderTile& rtile, int sample)
 	{
 		cuda_push_context();
 
 		CUfunction cuPathTrace;
-		CUdeviceptr d_buffer = cuda_device_ptr(task.buffer);
-		CUdeviceptr d_rng_state = cuda_device_ptr(task.rng_state);
+		CUdeviceptr d_buffer = cuda_device_ptr(rtile.buffer);
+		CUdeviceptr d_rng_state = cuda_device_ptr(rtile.rng_state);
 
 		/* get kernel function */
 		cuda_assert(cuModuleGetFunction(&cuPathTrace, cuModule, "kernel_cuda_path_trace"))
@@ -486,29 +491,28 @@ public:
 		cuda_assert(cuParamSetv(cuPathTrace, offset, &d_rng_state, sizeof(d_rng_state)))
 		offset += sizeof(d_rng_state);
 
-		int sample = task.sample;
 		offset = align_up(offset, __alignof(sample));
 
-		cuda_assert(cuParamSeti(cuPathTrace, offset, task.sample))
-		offset += sizeof(task.sample);
+		cuda_assert(cuParamSeti(cuPathTrace, offset, sample))
+		offset += sizeof(sample);
 
-		cuda_assert(cuParamSeti(cuPathTrace, offset, task.x))
-		offset += sizeof(task.x);
+		cuda_assert(cuParamSeti(cuPathTrace, offset, rtile.x))
+		offset += sizeof(rtile.x);
 
-		cuda_assert(cuParamSeti(cuPathTrace, offset, task.y))
-		offset += sizeof(task.y);
+		cuda_assert(cuParamSeti(cuPathTrace, offset, rtile.y))
+		offset += sizeof(rtile.y);
 
-		cuda_assert(cuParamSeti(cuPathTrace, offset, task.w))
-		offset += sizeof(task.w);
+		cuda_assert(cuParamSeti(cuPathTrace, offset, rtile.w))
+		offset += sizeof(rtile.w);
 
-		cuda_assert(cuParamSeti(cuPathTrace, offset, task.h))
-		offset += sizeof(task.h);
+		cuda_assert(cuParamSeti(cuPathTrace, offset, rtile.h))
+		offset += sizeof(rtile.h);
 
-		cuda_assert(cuParamSeti(cuPathTrace, offset, task.offset))
-		offset += sizeof(task.offset);
+		cuda_assert(cuParamSeti(cuPathTrace, offset, rtile.offset))
+		offset += sizeof(rtile.offset);
 
-		cuda_assert(cuParamSeti(cuPathTrace, offset, task.stride))
-		offset += sizeof(task.stride);
+		cuda_assert(cuParamSeti(cuPathTrace, offset, rtile.stride))
+		offset += sizeof(rtile.stride);
 
 		cuda_assert(cuParamSetSize(cuPathTrace, offset))
 
@@ -520,23 +524,25 @@ public:
 		int xthreads = 8;
 		int ythreads = 8;
 #endif
-		int xblocks = (task.w + xthreads - 1)/xthreads;
-		int yblocks = (task.h + ythreads - 1)/ythreads;
+		int xblocks = (rtile.w + xthreads - 1)/xthreads;
+		int yblocks = (rtile.h + ythreads - 1)/ythreads;
 
 		cuda_assert(cuFuncSetCacheConfig(cuPathTrace, CU_FUNC_CACHE_PREFER_L1))
 		cuda_assert(cuFuncSetBlockShape(cuPathTrace, xthreads, ythreads, 1))
 		cuda_assert(cuLaunchGrid(cuPathTrace, xblocks, yblocks))
 
+		cuda_assert(cuCtxSynchronize())
+
 		cuda_pop_context();
 	}
 
-	void tonemap(DeviceTask& task)
+	void tonemap(DeviceTask& task, device_ptr buffer, device_ptr rgba)
 	{
 		cuda_push_context();
 
 		CUfunction cuFilmConvert;
-		CUdeviceptr d_rgba = map_pixels(task.rgba);
-		CUdeviceptr d_buffer = cuda_device_ptr(task.buffer);
+		CUdeviceptr d_rgba = map_pixels(rgba);
+		CUdeviceptr d_buffer = cuda_device_ptr(buffer);
 
 		/* get kernel function */
 		cuda_assert(cuModuleGetFunction(&cuFilmConvert, cuModule, "kernel_cuda_tonemap"))
@@ -820,27 +826,71 @@ public:
 		Device::draw_pixels(mem, y, w, h, dy, width, height, transparent);
 	}
 
-	void task_add(DeviceTask& task)
+	void thread_run(DeviceTask *task)
 	{
-		if(task.type == DeviceTask::TONEMAP)
-			tonemap(task);
-		else if(task.type == DeviceTask::PATH_TRACE)
-			path_trace(task);
-		else if(task.type == DeviceTask::SHADER)
-			shader(task);
+		if(task->type == DeviceTask::PATH_TRACE) {
+			RenderTile tile;
+			
+			/* keep rendering tiles until done */
+			while(task->acquire_tile(this, tile)) {
+				int start_sample = tile.start_sample;
+				int end_sample = tile.start_sample + tile.num_samples;
+
+				for(int sample = start_sample; sample < end_sample; sample++) {
+					if (task->get_cancel())
+						break;
+
+					path_trace(tile, sample);
+
+					tile.sample = sample + 1;
+
+					task->update_progress(tile);
+				}
+
+				task->release_tile(tile);
+			}
+		}
+		else if(task->type == DeviceTask::SHADER) {
+			shader(*task);
+
+			cuda_push_context();
+			cuda_assert(cuCtxSynchronize())
+			cuda_pop_context();
+		}
 	}
 
-	void task_wait()
+	class CUDADeviceTask : public DeviceTask {
+	public:
+		CUDADeviceTask(CUDADevice *device, DeviceTask& task)
+		: DeviceTask(task)
+		{
+			run = function_bind(&CUDADevice::thread_run, device, this);
+		}
+	};
+
+	void task_add(DeviceTask& task)
 	{
-		cuda_push_context();
+		if(task.type == DeviceTask::TONEMAP) {
+			/* must be done in main thread due to opengl access */
+			tonemap(task, task.buffer, task.rgba);
 
-		cuda_assert(cuCtxSynchronize())
+			cuda_push_context();
+			cuda_assert(cuCtxSynchronize())
+			cuda_pop_context();
+		}
+		else {
+			task_pool.push(new CUDADeviceTask(this, task));
+		}
+	}
 
-		cuda_pop_context();
+	void task_wait()
+	{
+		task_pool.wait_work();
 	}
 
 	void task_cancel()
 	{
+		task_pool.cancel();
 	}
 };
 
diff --git a/intern/cycles/device/device_multi.cpp b/intern/cycles/device/device_multi.cpp
index 83e69b98f5d..546ffe5e4b9 100644
--- a/intern/cycles/device/device_multi.cpp
+++ b/intern/cycles/device/device_multi.cpp
@@ -23,6 +23,8 @@
 #include "device_intern.h"
 #include "device_network.h"
 
+#include "buffers.h"
+
 #include "util_foreach.h"
 #include "util_list.h"
 #include "util_map.h"
@@ -255,6 +257,30 @@ public:
 		rgba.device_pointer = tmp;
 	}
 
+	void map_tile(Device *sub_device, RenderTile& tile)
+	{
+		foreach(SubDevice& sub, devices) {
+			if(sub.device == sub_device) {
+				if(tile.buffer) tile.buffer = sub.ptr_map[tile.buffer];
+				if(tile.rng_state) tile.rng_state = sub.ptr_map[tile.rng_state];
+				if(tile.rgba) tile.rgba = sub.ptr_map[tile.rgba];
+			}
+		}
+	}
+
+	int device_number(Device *sub_device)
+	{
+		int i = 0;
+
+		foreach(SubDevice& sub, devices) {
+			if(sub.device == sub_device)
+				return i;
+			i++;
+		}
+
+		return -1;
+	}
+
 	void task_add(DeviceTask& task)
 	{
 		list<DeviceTask> tasks;
@@ -266,7 +292,6 @@ public:
 				tasks.pop_front();
 
 				if(task.buffer) subtask.buffer = sub.ptr_map[task.buffer];
-				if(task.rng_state) subtask.rng_state = sub.ptr_map[task.rng_state];
 				if(task.rgba) subtask.rgba = sub.ptr_map[task.rgba];
 				if(task.shader_input) subtask.shader_input = sub.ptr_map[task.shader_input];
 				if(task.shader_output) subtask.shader_output = sub.ptr_map[task.shader_output];
diff --git a/intern/cycles/device/device_opencl.cpp b/intern/cycles/device/device_opencl.cpp
index c9ec7c75063..3c78b4895ae 100644
--- a/intern/cycles/device/device_opencl.cpp
+++ b/intern/cycles/device/device_opencl.cpp
@@ -25,6 +25,8 @@
 #include "device.h"
 #include "device_intern.h"
 
+#include "buffers.h"
+
 #include "util_foreach.h"
 #include "util_map.h"
 #include "util_math.h"
@@ -41,6 +43,7 @@ CCL_NAMESPACE_BEGIN
 class OpenCLDevice : public Device
 {
 public:
+	TaskPool task_pool;
 	cl_context cxContext;
 	cl_command_queue cqCommandQueue;
 	cl_platform_id cpPlatform;
@@ -435,6 +438,8 @@ public:
 
 	~OpenCLDevice()
 	{
+		task_pool.stop();
+
 		if(null_mem)
 			clReleaseMemObject(CL_MEM_PTR(null_mem));
 
@@ -540,19 +545,19 @@ public:
 		return global_size + ((r == 0)? 0: group_size - r);
 	}
 
-	void path_trace(DeviceTask& task)
+	void path_trace(RenderTile& rtile, int sample)
 	{
 		/* cast arguments to cl types */
 		cl_mem d_data = CL_MEM_PTR(const_mem_map["__data"]->device_pointer);
-		cl_mem d_buffer = CL_MEM_PTR(task.buffer);
-		cl_mem d_rng_state = CL_MEM_PTR(task.rng_state);
-		cl_int d_x = task.x;
-		cl_int d_y = task.y;
-		cl_int d_w = task.w;
-		cl_int d_h = task.h;
-		cl_int d_sample = task.sample;
-		cl_int d_offset = task.offset;
-		cl_int d_stride = task.stride;
+		cl_mem d_buffer = CL_MEM_PTR(rtile.buffer);
+		cl_mem d_rng_state = CL_MEM_PTR(rtile.rng_state);
+		cl_int d_x = rtile.x;
+		cl_int d_y = rtile.y;
+		cl_int d_w = rtile.w;
+		cl_int d_h = rtile.h;
+		cl_int d_sample = sample;
+		cl_int d_offset = rtile.offset;
+		cl_int d_stride = rtile.stride;
 
 		/* sample arguments */
 		int narg = 0;
@@ -613,12 +618,12 @@ public:
 		return err;
 	}
 
-	void tonemap(DeviceTask& task)
+	void tonemap(DeviceTask& task, device_ptr buffer, device_ptr rgba)
 	{
 		/* cast arguments to cl types */
 		cl_mem d_data = CL_MEM_PTR(const_mem_map["__data"]->device_pointer);
-		cl_mem d_rgba = CL_MEM_PTR(task.rgba);
-		cl_mem d_buffer = CL_MEM_PTR(task.buffer);
+		cl_mem d_rgba = CL_MEM_PTR(rgba);
+		cl_mem d_buffer = CL_MEM_PTR(buffer);
 		cl_int d_x = task.x;
 		cl_int d_y = task.y;
 		cl_int d_w = task.w;
@@ -667,30 +672,57 @@ public:
 		opencl_assert(clFinish(cqCommandQueue));
 	}
 
-	void task_add(DeviceTask& maintask)
+	void thread_run(DeviceTask *task)
 	{
-		list<DeviceTask> tasks;
+		if(task->type == DeviceTask::TONEMAP) {
+			tonemap(*task, task->buffer, task->rgba);
+		}
+		else if(task->type == DeviceTask::PATH_TRACE) {
+			RenderTile tile;
+			
+			/* keep rendering tiles until done */
+			while(task->acquire_tile(this, tile)) {
+				int start_sample = tile.start_sample;
+				int end_sample = tile.start_sample + tile.num_samples;
 
-		/* arbitrary limit to work around apple ATI opencl issue */
-		if(platform_name == "Apple")
-			maintask.split_max_size(tasks, 76800);
-		else
-			tasks.push_back(maintask);
+				for(int sample = start_sample; sample < end_sample; sample++) {
+					if (task->get_cancel())
+						break;
+
+					path_trace(tile, sample);
+
+					tile.sample = sample + 1;
 
-		foreach(DeviceTask& task, tasks) {
-			if(task.type == DeviceTask::TONEMAP)
-				tonemap(task);
-			else if(task.type == DeviceTask::PATH_TRACE)
-				path_trace(task);
+					task->update_progress(tile);
+				}
+
+				task->release_tile(tile);
+			}
 		}
 	}
 
+	class OpenCLDeviceTask : public DeviceTask {
+	public:
+		OpenCLDeviceTask(OpenCLDevice *device, DeviceTask& task)
+		: DeviceTask(task)
+		{
+			run = function_bind(&OpenCLDevice::thread_run, device, this);
+		}
+	};
+
+	void task_add(DeviceTask& task)
+	{
+		task_pool.push(new OpenCLDeviceTask(this, task));
+	}
+
 	void task_wait()
 	{
+		task_pool.wait_work();
 	}
 
 	void task_cancel()
 	{
+		task_pool.cancel();
 	}
 };
 
diff --git a/intern/cycles/device/device_task.cpp b/intern/cycles/device/device_task.cpp
new file mode 100644
index 00000000000..c85e182d629
--- /dev/null
+++ b/intern/cycles/device/device_task.cpp
@@ -0,0 +1,113 @@
+/*
+ * 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.
+ */
+
+#include <stdlib.h>
+#include <string.h>
+
+#include "device_task.h"
+
+#include "util_algorithm.h"
+#include "util_time.h"
+
+CCL_NAMESPACE_BEGIN
+
+/* Device Task */
+
+DeviceTask::DeviceTask(Type type_)
+: type(type_), x(0), y(0), w(0), h(0), rgba(0), buffer(0),
+  sample(0), num_samples(1), resolution(0),
+  shader_input(0), shader_output(0),
+  shader_eval_type(0), shader_x(0), shader_w(0)
+{
+	last_update_time = time_dt();
+}
+
+void DeviceTask::split_max_size(list<DeviceTask>& tasks, int max_size)
+{
+	int num;
+
+	if(type == SHADER) {
+		num = (shader_w + max_size - 1)/max_size;
+	}
+	else {
+		max_size = max(1, max_size/w);
+		num = (h + max_size - 1)/max_size;
+	}
+
+	split(tasks, num);
+}
+
+void DeviceTask::split(list<DeviceTask>& tasks, int num)
+{
+	if(type == SHADER) {
+		num = min(shader_w, num);
+
+		for(int i = 0; i < num; i++) {
+			int tx = shader_x + (shader_w/num)*i;
+			int tw = (i == num-1)? shader_w - i*(shader_w/num): shader_w/num;
+
+			DeviceTask task = *this;
+
+			task.shader_x = tx;
+			task.shader_w = tw;
+
+			tasks.push_back(task);
+		}
+	}
+	else if(type == PATH_TRACE) {
+		for(int i = 0; i < num; i++)
+			tasks.push_back(*this);
+	}
+	else {
+		num = min(h, num);
+
+		for(int i = 0; i < num; i++) {
+			int ty = y + (h/num)*i;
+			int th = (i == num-1)? h - i*(h/num): h/num;
+
+			DeviceTask task = *this;
+
+			task.y = ty;
+			task.h = th;
+
+			tasks.push_back(task);
+		}
+	}
+}
+
+void DeviceTask::update_progress(RenderTile &rtile)
+{
+	if (type != PATH_TRACE)
+		return;
+
+	if(update_progress_sample)
+		update_progress_sample();
+
+	if(update_tile_sample) {
+		double current_time = time_dt();
+
+		if (current_time - last_update_time >= 1.0f) {
+			update_tile_sample(rtile);
+
+			last_update_time = current_time;
+		}
+	}
+}
+
+CCL_NAMESPACE_END
+
diff --git a/intern/cycles/device/device_task.h b/intern/cycles/device/device_task.h
new file mode 100644
index 00000000000..cfb3d8d988e
--- /dev/null
+++ b/intern/cycles/device/device_task.h
@@ -0,0 +1,75 @@
+/*
+ * 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.
+ */
+
+#ifndef __DEVICE_TASK_H__
+#define __DEVICE_TASK_H__
+
+#include "device_memory.h"
+
+#include "util_function.h"
+#include "util_list.h"
+#include "util_task.h"
+
+CCL_NAMESPACE_BEGIN
+
+/* Device Task */
+
+class Device;
+class RenderBuffers;
+class RenderTile;
+class Tile;
+
+class DeviceTask : public Task {
+public:
+	typedef enum { PATH_TRACE, TONEMAP, SHADER } Type;
+	Type type;
+
+	int x, y, w, h;
+	device_ptr rgba;
+	device_ptr buffer;
+	int sample;
+	int num_samples;
+	int resolution;
+	int offset, stride;
+
+	device_ptr shader_input;
+	device_ptr shader_output;
+	int shader_eval_type;
+	int shader_x, shader_w;
+
+	DeviceTask(Type type = PATH_TRACE);
+
+	void split(list<DeviceTask>& tasks, int num);
+	void split_max_size(list<DeviceTask>& tasks, int max_size);
+
+	void update_progress(RenderTile &rtile);
+
+	boost::function<bool(Device *device, RenderTile&)> acquire_tile;
+	boost::function<void(void)> update_progress_sample;
+	boost::function<void(RenderTile&)> update_tile_sample;
+	boost::function<void(RenderTile&)> release_tile;
+	boost::function<bool(void)> get_cancel;
+
+protected:
+	double last_update_time;
+};
+
+CCL_NAMESPACE_END
+
+#endif /* __DEVICE_TASK_H__ */
+