Merge branch 'master' into dyntopo_holes

17 files changed, 56 insertions, 49 deletions

diff --git a/intern/cycles/device/device_cpu.cpp b/intern/cycles/device/device_cpu.cpp
index 4623764d210..c9b8a5b726b 100644
--- a/intern/cycles/device/device_cpu.cpp
+++ b/intern/cycles/device/device_cpu.cpp
@@ -17,6 +17,11 @@
 #include <stdlib.h>
 #include <string.h>
 
+/* So ImathMath is included before our kernel_cpu_compat. */
+#ifdef WITH_OSL
+#  include <OSL/oslexec.h>
+#endif
+
 #include "device.h"
 #include "device_intern.h"
 
diff --git a/intern/cycles/device/device_cuda.cpp b/intern/cycles/device/device_cuda.cpp
index bd265f0fdad..44be7779891 100644
--- a/intern/cycles/device/device_cuda.cpp
+++ b/intern/cycles/device/device_cuda.cpp
@@ -278,6 +278,10 @@ public:
 		if(experimental)
 			command += " -D__KERNEL_CUDA_EXPERIMENTAL__";
 
+#ifdef WITH_CYCLES_DEBUG
+		command += " -D__KERNEL_DEBUG__";
+#endif
+
 		printf("%s\n", command.c_str());
 
 		if(system(command.c_str()) == -1) {
diff --git a/intern/cycles/device/device_opencl.cpp b/intern/cycles/device/device_opencl.cpp
index d950d084cd4..58b2bcafb82 100644
--- a/intern/cycles/device/device_opencl.cpp
+++ b/intern/cycles/device/device_opencl.cpp
@@ -102,7 +102,11 @@ static string opencl_kernel_build_options(const string& platform, const string *
 
 	if(opencl_kernel_use_debug())
 		build_options += "-D__KERNEL_OPENCL_DEBUG__ ";
-	
+
+#ifdef WITH_CYCLES_DEBUG
+	build_options += "-D__KERNEL_DEBUG__ ";
+#endif
+
 	return build_options;
 }
 
diff --git a/intern/cycles/kernel/CMakeLists.txt b/intern/cycles/kernel/CMakeLists.txt
index c5ea3abc567..c521e1383a4 100644
--- a/intern/cycles/kernel/CMakeLists.txt
+++ b/intern/cycles/kernel/CMakeLists.txt
@@ -170,6 +170,12 @@ if(WITH_CYCLES_CUDA_BINARIES)
 			set(cuda_cubin kernel_${arch}.cubin)
 		endif()
 
+		if(WITH_CYCLES_DEBUG)
+			set(cuda_debug_flags "-D__KERNEL_DEBUG__")
+		else()
+			set(cuda_debug_flags "")
+		endif()
+
 		set(cuda_version_flags "-D__KERNEL_CUDA_VERSION__=${CUDA_VERSION}")
 		set(cuda_math_flags "--use_fast_math")
 
@@ -185,6 +191,7 @@ if(WITH_CYCLES_CUDA_BINARIES)
 					${cuda_version_flags}
 					${cuda_math_flags}
 					${cuda_extra_flags}
+					${cuda_debug_flags}
 					-I${CMAKE_CURRENT_SOURCE_DIR}/../util
 					-I${CMAKE_CURRENT_SOURCE_DIR}/svm
 					-DCCL_NAMESPACE_BEGIN=
@@ -197,6 +204,7 @@ if(WITH_CYCLES_CUDA_BINARIES)
 		list(APPEND cuda_cubins ${cuda_cubin})
 
 		unset(cuda_extra_flags)
+		unset(cuda_debug_flags)
 	endmacro()
 
 	foreach(arch ${CYCLES_CUDA_BINARIES_ARCH})
diff --git a/intern/cycles/kernel/SConscript b/intern/cycles/kernel/SConscript
index 5a9e57c5342..c0d969e24ae 100644
--- a/intern/cycles/kernel/SConscript
+++ b/intern/cycles/kernel/SConscript
@@ -79,6 +79,9 @@ if env['WITH_BF_CYCLES_CUDA_BINARIES']:
     nvcc_flags += " -DCCL_NAMESPACE_BEGIN= -DCCL_NAMESPACE_END= -DNVCC"
     nvcc_flags += " -I \"%s\" -I \"%s\" -I \"%s\" -I \"%s\"" % (util_dir, svm_dir, geom_dir, closure_dir)
 
+    if env['WITH_BF_CYCLES_DEBUG']:
+        nvcc_flags += " -D__KERNEL_DEBUG__"
+
     # dependencies
     dependencies = ['kernel.cu'] + kernel.Glob('*.h') + kernel.Glob('../util/*.h') + kernel.Glob('svm/*.h') + kernel.Glob('geom/*.h') + kernel.Glob('closure/*.h')
     last_cubin_file = None
diff --git a/intern/cycles/kernel/kernel_compat_cpu.h b/intern/cycles/kernel/kernel_compat_cpu.h
index c2aab93c87b..25531843993 100644
--- a/intern/cycles/kernel/kernel_compat_cpu.h
+++ b/intern/cycles/kernel/kernel_compat_cpu.h
@@ -25,6 +25,13 @@
 #include "util_half.h"
 #include "util_types.h"
 
+/* On 64bit linux single precision exponent is really slow comparing to the
+ * double precision version, even with float<->double conversion involved.
+ */
+#if !defined(__KERNEL_GPU__) && defined(__linux__) && defined(__x86_64__)
+#  define expf(x) ((float)exp((double)x))
+#endif
+
 CCL_NAMESPACE_BEGIN
 
 /* Assertions inside the kernel only work for the CPU device, so we wrap it in
diff --git a/intern/cycles/kernel/kernel_compat_opencl.h b/intern/cycles/kernel/kernel_compat_opencl.h
index 9e58ebff599..78b3fd012dd 100644
--- a/intern/cycles/kernel/kernel_compat_opencl.h
+++ b/intern/cycles/kernel/kernel_compat_opencl.h
@@ -24,14 +24,6 @@
 #define CCL_NAMESPACE_BEGIN
 #define CCL_NAMESPACE_END
 
-#ifdef __KERNEL_OPENCL_AMD__
-#define __CL_NO_FLOAT3__
-#endif
-
-#ifdef __CL_NO_FLOAT3__
-#define float3 float4
-#endif
-
 #ifdef __CL_NOINLINE__
 #define ccl_noinline __attribute__((noinline))
 #else
@@ -73,11 +65,7 @@
 #endif
 
 #define make_float2(x, y) ((float2)(x, y))
-#ifdef __CL_NO_FLOAT3__
-#define make_float3(x, y, z) ((float4)(x, y, z, 0.0f))
-#else
 #define make_float3(x, y, z) ((float3)(x, y, z))
-#endif
 #define make_float4(x, y, z, w) ((float4)(x, y, z, w))
 #define make_int2(x, y) ((int2)(x, y))
 #define make_int3(x, y, z) ((int3)(x, y, z))
diff --git a/intern/cycles/kernel/kernel_debug.h b/intern/cycles/kernel/kernel_debug.h
index 81ce1e76420..bf1bc0e9db8 100644
--- a/intern/cycles/kernel/kernel_debug.h
+++ b/intern/cycles/kernel/kernel_debug.h
@@ -27,9 +27,12 @@ ccl_device_inline void kernel_write_debug_passes(KernelGlobals *kg,
                                                  DebugData *debug_data,
                                                  int sample)
 {
-	kernel_write_pass_float(buffer + kernel_data.film.pass_bvh_traversal_steps,
-	                        sample,
-	                        debug_data->num_bvh_traversal_steps);
+	int flag = kernel_data.film.pass_flag;
+	if(flag & PASS_BVH_TRAVERSAL_STEPS) {
+		kernel_write_pass_float(buffer + kernel_data.film.pass_bvh_traversal_steps,
+		                        sample,
+		                        debug_data->num_bvh_traversal_steps);
+	}
 }
 
 CCL_NAMESPACE_END
diff --git a/intern/cycles/kernel/kernel_path.h b/intern/cycles/kernel/kernel_path.h
index 0e07d8a95f8..c03229f0a3a 100644
--- a/intern/cycles/kernel/kernel_path.h
+++ b/intern/cycles/kernel/kernel_path.h
@@ -46,7 +46,7 @@
 #include "kernel_path_volume.h"
 
 #ifdef __KERNEL_DEBUG__
-#  include "kernel_debug.h"
+#include "kernel_debug.h"
 #endif
 
 CCL_NAMESPACE_BEGIN
@@ -373,10 +373,8 @@ ccl_device void kernel_path_subsurface_update_volume_stack(KernelGlobals *kg,
 
 	Ray volume_ray = *ray;
 	Intersection isect;
-	const float3 Pend = volume_ray.P + volume_ray.D*volume_ray.t;
 
-	while(scene_intersect(kg, &volume_ray, PATH_RAY_ALL_VISIBILITY,
-	                      &isect, NULL, 0.0f, 0.0f))
+	while(scene_intersect_volume(kg, &volume_ray, &isect))
 	{
 		ShaderData sd;
 		shader_setup_from_ray(kg, &sd, &isect, &volume_ray, 0, 0);
@@ -384,15 +382,7 @@ ccl_device void kernel_path_subsurface_update_volume_stack(KernelGlobals *kg,
 
 		/* Move ray forward. */
 		volume_ray.P = ray_offset(sd.P, -sd.Ng);
-		volume_ray.D = normalize_len(Pend - volume_ray.P,
-		                             &volume_ray.t);
-
-		/* TODO(sergey): Find a faster way detecting that ray_offset moved
-		 * us pass through the end point.
-		 */
-		if(dot(ray->D, volume_ray.D) < 0.0f) {
-			break;
-		}
+		volume_ray.t -= sd.ray_length;
 	}
 }
 #endif
diff --git a/intern/cycles/kernel/kernel_types.h b/intern/cycles/kernel/kernel_types.h
index 43becf14825..e5cd23c0925 100644
--- a/intern/cycles/kernel/kernel_types.h
+++ b/intern/cycles/kernel/kernel_types.h
@@ -864,7 +864,7 @@ typedef struct KernelFilm {
 
 #ifdef __KERNEL_DEBUG__
 	int pass_bvh_traversal_steps;
-	int pad[3];
+	int pass_pad3, pass_pad4, pass_pad5;
 #endif
 } KernelFilm;
 
diff --git a/intern/cycles/kernel/kernel_volume.h b/intern/cycles/kernel/kernel_volume.h
index dd947a54530..93cb4c120ea 100644
--- a/intern/cycles/kernel/kernel_volume.h
+++ b/intern/cycles/kernel/kernel_volume.h
@@ -976,8 +976,9 @@ ccl_device void kernel_volume_stack_init(KernelGlobals *kg,
 		return;
 	}
 
-	const float3 Pend = ray->P + ray->D*ray->t;
 	Ray volume_ray = *ray;
+	volume_ray.t = FLT_MAX;
+
 	int stack_index = 0, enclosed_index = 0;
 	int enclosed_volumes[VOLUME_STACK_SIZE];
 
@@ -1019,15 +1020,6 @@ ccl_device void kernel_volume_stack_init(KernelGlobals *kg,
 
 		/* Move ray forward. */
 		volume_ray.P = ray_offset(sd.P, -sd.Ng);
-		if(volume_ray.t != FLT_MAX) {
-			volume_ray.D = normalize_len(Pend - volume_ray.P, &volume_ray.t);
-			/* TODO(sergey): Find a faster way detecting that ray_offset moved
-			 * us pass through the end point.
-			 */
-			if(dot(ray->D, volume_ray.D) < 0.0f) {
-				break;
-			}
-		}
 	}
 	/* stack_index of 0 means quick checks outside of the kernel gave false
 	 * positive, nothing to worry about, just we've wasted quite a few of
diff --git a/intern/cycles/kernel/osl/SConscript b/intern/cycles/kernel/osl/SConscript
index 4685bb7753e..d721edbaf6e 100644
--- a/intern/cycles/kernel/osl/SConscript
+++ b/intern/cycles/kernel/osl/SConscript
@@ -43,6 +43,9 @@ defs.append('CCL_NAMESPACE_BEGIN=namespace ccl {')
 defs.append('CCL_NAMESPACE_END=}')
 defs.append('WITH_OSL')
 
+if env['WITH_BF_CYCLES_DEBUG']:
+    defs.append('WITH_CYCLES_DEBUG')
+
 if env['OURPLATFORM'] in ('win32-vc', 'win64-vc'):
     cxxflags.append('-DBOOST_NO_RTTI -DBOOST_NO_TYPEID /fp:fast'.split())
     incs.append(env['BF_PTHREADS_INC'])
diff --git a/intern/cycles/kernel/osl/osl_globals.h b/intern/cycles/kernel/osl/osl_globals.h
index 5a658d8244a..9c3134e41c9 100644
--- a/intern/cycles/kernel/osl/osl_globals.h
+++ b/intern/cycles/kernel/osl/osl_globals.h
@@ -20,7 +20,6 @@
 #ifdef WITH_OSL
 
 #include <OSL/oslexec.h>
-#include <cmath>
 
 #include "util_map.h"
 #include "util_param.h"
diff --git a/intern/cycles/kernel/osl/osl_shader.cpp b/intern/cycles/kernel/osl/osl_shader.cpp
index 48498116874..ca0c2cc4415 100644
--- a/intern/cycles/kernel/osl/osl_shader.cpp
+++ b/intern/cycles/kernel/osl/osl_shader.cpp
@@ -14,6 +14,8 @@
  * limitations under the License
  */
 
+#include <OSL/oslexec.h>
+
 #include "kernel_compat_cpu.h"
 #include "kernel_montecarlo.h"
 #include "kernel_types.h"
@@ -34,7 +36,6 @@
 
 #include "attribute.h"
 
-#include <OSL/oslexec.h>
 
 CCL_NAMESPACE_BEGIN
 
diff --git a/intern/cycles/render/shader.h b/intern/cycles/render/shader.h
index 0ed6d2ddf01..368496fd188 100644
--- a/intern/cycles/render/shader.h
+++ b/intern/cycles/render/shader.h
@@ -17,6 +17,10 @@
 #ifndef __SHADER_H__
 #define __SHADER_H__
 
+#ifdef WITH_OSL
+#  include <OSL/oslexec.h>
+#endif
+
 #include "attribute.h"
 #include "kernel_types.h"
 
@@ -25,10 +29,6 @@
 #include "util_string.h"
 #include "util_types.h"
 
-#ifdef WITH_OSL
-#include <OSL/oslexec.h>
-#endif
-
 CCL_NAMESPACE_BEGIN
 
 class Device;
diff --git a/intern/cycles/render/tile.cpp b/intern/cycles/render/tile.cpp
index d6094a4fa0a..e37d8e5f8a1 100644
--- a/intern/cycles/render/tile.cpp
+++ b/intern/cycles/render/tile.cpp
@@ -200,8 +200,8 @@ list<Tile>::iterator TileManager::next_background_tile(int device, TileOrder til
 			
 			switch (tile_order) {
 				case TILE_CENTER:
-					distx = centx - (cur_tile.x + cur_tile.w);
-					disty = centy - (cur_tile.y + cur_tile.h);
+					distx = centx - (cur_tile.x + (cur_tile.w / 2));
+					disty = centy - (cur_tile.y + (cur_tile.h / 2));
 					distx = (int64_t)sqrt((double)(distx * distx + disty * disty));
 					break;
 				case TILE_RIGHT_TO_LEFT:
diff --git a/intern/ghost/intern/GHOST_SystemCocoa.mm b/intern/ghost/intern/GHOST_SystemCocoa.mm
index 7b08c46b65d..ccbac91f0b4 100644
--- a/intern/ghost/intern/GHOST_SystemCocoa.mm
+++ b/intern/ghost/intern/GHOST_SystemCocoa.mm
@@ -740,7 +740,7 @@ bool GHOST_SystemCocoa::processEvents(bool waitForEvent)
 				// For some reason NSApp is swallowing the key up events when modifier
 				// key is pressed, even if there seems to be no apparent reason to do
 				// so, as a workaround we always handle these up events.
-				if ([event type] == NSKeyUp && (([event modifierFlags] & NSCommandKeyMask) || ([event modifierFlags] & NSAlternateKeyMask)))
+				if ([event type] == NSKeyUp && ([event modifierFlags] & (NSCommandKeyMask | NSAlternateKeyMask)))
 					handleKeyEvent(event);
 
 				[NSApp sendEvent:event];