Cycles:

* Fix missing update when editing objects with emission materials.
* Fix preview pass rendering set to 1 not showing full resolution.
* Fix CUDA runtime compiling failing due to missing cache directory.
* Use settings from first render layer for visibility and material override.

And a bunch of incomplete and still disabled code mostly related to closure
sampling.

3 files changed, 50 insertions, 13 deletions

diff --git a/intern/cycles/device/device.cpp b/intern/cycles/device/device.cpp
index e21dce45d80..ab57d7cfe86 100644
--- a/intern/cycles/device/device.cpp
+++ b/intern/cycles/device/device.cpp
@@ -84,7 +84,7 @@ void Device::pixels_alloc(device_memory& mem)
 
 void Device::pixels_copy_from(device_memory& mem, int y, int w, int h)
 {
-	mem_copy_from(mem, sizeof(uchar)*4*y*w, sizeof(uchar)*4*w*h);
+	mem_copy_from(mem, sizeof(uint8_t)*4*y*w, sizeof(uint8_t)*4*w*h);
 }
 
 void Device::pixels_free(device_memory& mem)
@@ -104,7 +104,13 @@ void Device::draw_pixels(device_memory& rgba, int y, int w, int h, int width, in
 	glPixelZoom((float)width/(float)w, (float)height/(float)h);
 	glRasterPos2f(0, y);
 
-	glDrawPixels(w, h, GL_RGBA, GL_UNSIGNED_BYTE, (void*)rgba.data_pointer);
+	uint8_t *pixels = (uint8_t*)rgba.data_pointer;
+
+	/* for multi devices, this assumes the ineffecient method that we allocate
+	   all pixels on the device even though we only render to a subset */
+	pixels += sizeof(uint8_t)*4*y*w;
+
+	glDrawPixels(w, h, GL_RGBA, GL_UNSIGNED_BYTE, pixels);
 
 	glRasterPos2f(0.0f, 0.0f);
 	glPixelZoom(1.0f, 1.0f);
diff --git a/intern/cycles/device/device_cuda.cpp b/intern/cycles/device/device_cuda.cpp
index 49ffd3d0834..7fc0afce1ac 100644
--- a/intern/cycles/device/device_cuda.cpp
+++ b/intern/cycles/device/device_cuda.cpp
@@ -232,11 +232,16 @@ public:
 		double starttime = time_dt();
 		printf("Compiling CUDA kernel ...\n");
 
+		path_create_directories(cubin);
+
 		string command = string_printf("%s -arch=sm_%d%d -m%d --cubin \"%s\" --use_fast_math "
 			"-o \"%s\" --ptxas-options=\"-v\" --maxrregcount=%d --opencc-options -OPT:Olimit=0 -I\"%s\" -DNVCC",
 			nvcc.c_str(), major, minor, machine, kernel.c_str(), cubin.c_str(), maxreg, include.c_str());
 
-		system(command.c_str());
+		if(system(command.c_str()) == -1) {
+			fprintf(stderr, "Failed to execute compilation command.\n");
+			return "";
+		}
 
 		/* verify if compilation succeeded */
 		if(!path_exists(cubin)) {
@@ -708,9 +713,13 @@ public:
 
 			cuda_push_context();
 
+			/* for multi devices, this assumes the ineffecient method that we allocate
+			   all pixels on the device even though we only render to a subset */
+			size_t offset = sizeof(uint8_t)*4*y*w;
+
 			glBindBufferARB(GL_PIXEL_UNPACK_BUFFER_ARB, pmem.cuPBO);
 			glBindTexture(GL_TEXTURE_2D, pmem.cuTexId);
-			glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, w, h, GL_RGBA, GL_UNSIGNED_BYTE, 0);
+			glTexSubImage2D(GL_TEXTURE_2D, 0, 0, 0, w, h, GL_RGBA, GL_UNSIGNED_BYTE, (void*)offset);
 			glBindBufferARB(GL_PIXEL_UNPACK_BUFFER_ARB, 0);
 			
 			glEnable(GL_TEXTURE_2D);
@@ -729,11 +738,11 @@ public:
 			
 			glTexCoord2f(0.0f, 0.0f);
 			glVertex2f(0.0f, 0.0f);
-			glTexCoord2f((float)w/(float)width, 0);
+			glTexCoord2f((float)w/(float)pmem.w, 0.0f);
 			glVertex2f((float)width, 0.0f);
-			glTexCoord2f((float)w/(float)width, (float)h/(float)height);
+			glTexCoord2f((float)w/(float)pmem.w, (float)h/(float)pmem.h);
 			glVertex2f((float)width, (float)height);
-			glTexCoord2f(0.0f, (float)h/(float)height);
+			glTexCoord2f(0.0f, (float)h/(float)pmem.h);
 			glVertex2f(0.0f, (float)height);
 
 			glEnd();
diff --git a/intern/cycles/device/device_multi.cpp b/intern/cycles/device/device_multi.cpp
index cae63596d3b..f2f6251685e 100644
--- a/intern/cycles/device/device_multi.cpp
+++ b/intern/cycles/device/device_multi.cpp
@@ -47,9 +47,6 @@ public:
 	MultiDevice(bool background_)
 	: unique_ptr(1)
 	{
-		/* enforce background for now */
-		background = true;
-
 		Device *device;
 
 		/* add CPU device */
@@ -125,6 +122,15 @@ public:
 		return desc.str();
 	}
 
+	bool load_kernels()
+	{
+		foreach(SubDevice& sub, devices)
+			if(!sub.device->load_kernels())
+				return false;
+
+		return true;
+	}
+
 	void mem_alloc(device_memory& mem, MemoryType type)
 	{
 		foreach(SubDevice& sub, devices) {
@@ -219,12 +225,26 @@ public:
 
 	void pixels_alloc(device_memory& mem)
 	{
-		Device::pixels_alloc(mem);
+		foreach(SubDevice& sub, devices) {
+			mem.device_pointer = 0;
+			sub.device->pixels_alloc(mem);
+			sub.ptr_map[unique_ptr] = mem.device_pointer;
+		}
+
+		mem.device_pointer = unique_ptr++;
 	}
 
 	void pixels_free(device_memory& mem)
 	{
-		Device::pixels_free(mem);
+		device_ptr tmp = mem.device_pointer;
+
+		foreach(SubDevice& sub, devices) {
+			mem.device_pointer = sub.ptr_map[tmp];
+			sub.device->pixels_free(mem);
+			sub.ptr_map.erase(sub.ptr_map.find(tmp));
+		}
+
+		mem.device_pointer = 0;
 	}
 
 	void pixels_copy_from(device_memory& mem, int y, int w, int h)
@@ -248,14 +268,16 @@ public:
 	{
 		device_ptr tmp = rgba.device_pointer;
 		int i = 0, sub_h = h/devices.size();
+		int sub_height = height/devices.size();
 
 		foreach(SubDevice& sub, devices) {
 			int sy = y + i*sub_h;
 			int sh = (i == (int)devices.size() - 1)? h - sub_h*i: sub_h;
+			int sheight = (i == (int)devices.size() - 1)? height - sub_height*i: sub_height;
 			/* adjust math for w/width */
 
 			rgba.device_pointer = sub.ptr_map[tmp];
-			sub.device->draw_pixels(rgba, sy, w, sh, width, height, transparent);
+			sub.device->draw_pixels(rgba, sy, w, sh, width, sheight, transparent);
 			i++;
 		}