Code refactor: make Transform always affine, dropping last row.

This save a little memory and copying in the kernel by storing only a 4x3
matrix instead of a 4x4 matrix. We already did this in a few places, and
those don't need to be special exceptions anymore now.

10 files changed, 13 insertions, 40 deletions

diff --git a/intern/cycles/kernel/bvh/bvh_nodes.h b/intern/cycles/kernel/bvh/bvh_nodes.h
index 6c33dad5426..060b3934a41 100644
--- a/intern/cycles/kernel/bvh/bvh_nodes.h
+++ b/intern/cycles/kernel/bvh/bvh_nodes.h
@@ -25,7 +25,6 @@ ccl_device_forceinline Transform bvh_unaligned_node_fetch_space(KernelGlobals *k
 	space.x = kernel_tex_fetch(__bvh_nodes, child_addr+1);
 	space.y = kernel_tex_fetch(__bvh_nodes, child_addr+2);
 	space.z = kernel_tex_fetch(__bvh_nodes, child_addr+3);
-	space.w = make_float4(0.0f, 0.0f, 0.0f, 1.0f);
 	return space;
 }
 
diff --git a/intern/cycles/kernel/geom/geom_attribute.h b/intern/cycles/kernel/geom/geom_attribute.h
index 6e2ee3b62d7..42c053704d5 100644
--- a/intern/cycles/kernel/geom/geom_attribute.h
+++ b/intern/cycles/kernel/geom/geom_attribute.h
@@ -103,7 +103,6 @@ ccl_device Transform primitive_attribute_matrix(KernelGlobals *kg, const ShaderD
 	tfm.x = kernel_tex_fetch(__attributes_float3, desc.offset + 0);
 	tfm.y = kernel_tex_fetch(__attributes_float3, desc.offset + 1);
 	tfm.z = kernel_tex_fetch(__attributes_float3, desc.offset + 2);
-	tfm.w = kernel_tex_fetch(__attributes_float3, desc.offset + 3);
 
 	return tfm;
 }
diff --git a/intern/cycles/kernel/geom/geom_curve_intersect.h b/intern/cycles/kernel/geom/geom_curve_intersect.h
index faf3e3cdf2b..46c3f408f0b 100644
--- a/intern/cycles/kernel/geom/geom_curve_intersect.h
+++ b/intern/cycles/kernel/geom/geom_curve_intersect.h
@@ -170,8 +170,7 @@ ccl_device_forceinline bool cardinal_curve_intersect(
 		htfm = make_transform(
 			dir.z / d, 0, -dir.x /d, 0,
 			-dir.x * dir.y /d, d, -dir.y * dir.z /d, 0,
-			dir.x, dir.y, dir.z, 0,
-			0, 0, 0, 1);
+			dir.x, dir.y, dir.z, 0);
 
 		float4 v00 = kernel_tex_fetch(__curves, prim);
 
diff --git a/intern/cycles/kernel/geom/geom_object.h b/intern/cycles/kernel/geom/geom_object.h
index cd330308e07..62bc46dc21b 100644
--- a/intern/cycles/kernel/geom/geom_object.h
+++ b/intern/cycles/kernel/geom/geom_object.h
@@ -28,12 +28,12 @@ CCL_NAMESPACE_BEGIN
 
 enum ObjectTransform {
 	OBJECT_TRANSFORM = 0,
-	OBJECT_INVERSE_TRANSFORM = 4,
+	OBJECT_INVERSE_TRANSFORM = 1,
 };
 
 enum ObjectVectorTransform {
 	OBJECT_VECTOR_MOTION_PRE = 0,
-	OBJECT_VECTOR_MOTION_POST = 3
+	OBJECT_VECTOR_MOTION_POST = 1
 };
 
 /* Object to world space transformation */
@@ -51,8 +51,6 @@ ccl_device_inline Transform object_fetch_transform(KernelGlobals *kg, int object
 		tfm.y = kernel_tex_fetch(__objects, object).tfm.pre.y;
 		tfm.z = kernel_tex_fetch(__objects, object).tfm.pre.z;
 	}
-	tfm.w = make_float4(0.0f, 0.0f, 0.0f, 1.0f);
-
 	return tfm;
 }
 
@@ -60,20 +58,12 @@ ccl_device_inline Transform object_fetch_transform(KernelGlobals *kg, int object
 
 ccl_device_inline Transform lamp_fetch_transform(KernelGlobals *kg, int lamp, bool inverse)
 {
-	Transform tfm;
 	if(inverse) {
-		tfm.x = kernel_tex_fetch(__lights, lamp).itfm[0];
-		tfm.y = kernel_tex_fetch(__lights, lamp).itfm[1];
-		tfm.z = kernel_tex_fetch(__lights, lamp).itfm[2];
+		return kernel_tex_fetch(__lights, lamp).itfm;
 	}
 	else {
-		tfm.x = kernel_tex_fetch(__lights, lamp).tfm[0];
-		tfm.y = kernel_tex_fetch(__lights, lamp).tfm[1];
-		tfm.z = kernel_tex_fetch(__lights, lamp).tfm[2];
+		return kernel_tex_fetch(__lights, lamp).tfm;
 	}
-	tfm.w = make_float4(0.0f, 0.0f, 0.0f, 1.0f);
-
-	return tfm;
 }
 
 /* Object to world space transformation for motion vectors */
@@ -81,14 +71,7 @@ ccl_device_inline Transform lamp_fetch_transform(KernelGlobals *kg, int lamp, bo
 ccl_device_inline Transform object_fetch_vector_transform(KernelGlobals *kg, int object, enum ObjectVectorTransform type)
 {
 	int offset = object*OBJECT_VECTOR_SIZE + (int)type;
-
-	Transform tfm;
-	tfm.x = kernel_tex_fetch(__objects_vector, offset + 0);
-	tfm.y = kernel_tex_fetch(__objects_vector, offset + 1);
-	tfm.z = kernel_tex_fetch(__objects_vector, offset + 2);
-	tfm.w = make_float4(0.0f, 0.0f, 0.0f, 1.0f);
-
-	return tfm;
+	return kernel_tex_fetch(__objects_vector, offset);
 }
 
 /* Motion blurred object transformations */
diff --git a/intern/cycles/kernel/kernel_textures.h b/intern/cycles/kernel/kernel_textures.h
index 892e0c84cba..6e07d88008e 100644
--- a/intern/cycles/kernel/kernel_textures.h
+++ b/intern/cycles/kernel/kernel_textures.h
@@ -32,7 +32,7 @@ KERNEL_TEX(float2, __prim_time)
 
 /* objects */
 KERNEL_TEX(KernelObject, __objects)
-KERNEL_TEX(float4, __objects_vector)
+KERNEL_TEX(Transform, __objects_vector)
 
 /* triangles */
 KERNEL_TEX(uint, __tri_shader)
diff --git a/intern/cycles/kernel/kernel_types.h b/intern/cycles/kernel/kernel_types.h
index 5934ff53f57..567f2a04028 100644
--- a/intern/cycles/kernel/kernel_types.h
+++ b/intern/cycles/kernel/kernel_types.h
@@ -35,10 +35,10 @@
 CCL_NAMESPACE_BEGIN
 
 /* Constants */
-#define OBJECT_VECTOR_SIZE	6
+#define OBJECT_VECTOR_SIZE	2
 #define FILTER_TABLE_SIZE	1024
 #define RAMP_TABLE_SIZE		256
-#define SHUTTER_TABLE_SIZE		256
+#define SHUTTER_TABLE_SIZE	256
 
 #define BSSRDF_MIN_RADIUS			1e-8f
 #define BSSRDF_MAX_HITS				4
@@ -1483,8 +1483,8 @@ typedef struct KernelLight {
 	int samples;
 	float max_bounces;
 	float random;
-	float4 tfm[3];
-	float4 itfm[3];
+	Transform tfm;
+	Transform itfm;
 	union {
 		KernelSpotLight spot;
 		KernelAreaLight area;
diff --git a/intern/cycles/kernel/osl/osl_services.cpp b/intern/cycles/kernel/osl/osl_services.cpp
index dec56c6665b..0c5e5e30e47 100644
--- a/intern/cycles/kernel/osl/osl_services.cpp
+++ b/intern/cycles/kernel/osl/osl_services.cpp
@@ -64,8 +64,7 @@ CCL_NAMESPACE_BEGIN
 
 static void copy_matrix(OSL::Matrix44& m, const Transform& tfm)
 {
-	// TODO: remember when making affine
-	Transform t = transform_transpose(tfm);
+	ProjectionTransform t = projection_transpose(ProjectionTransform(tfm));
 	memcpy(&m, &t, sizeof(m));
 }
 
@@ -553,8 +552,7 @@ static bool set_attribute_float3_3(float3 P[3], TypeDesc type, bool derivatives,
 static bool set_attribute_matrix(const Transform& tfm, TypeDesc type, void *val)
 {
 	if(type == TypeDesc::TypeMatrix) {
-		Transform transpose = transform_transpose(tfm);
-		memcpy(val, &transpose, sizeof(Transform));
+		copy_matrix(*(OSL::Matrix44*)val, tfm);
 		return true;
 	}
 
diff --git a/intern/cycles/kernel/svm/svm_mapping.h b/intern/cycles/kernel/svm/svm_mapping.h
index 0a890545af4..42a7ae9946f 100644
--- a/intern/cycles/kernel/svm/svm_mapping.h
+++ b/intern/cycles/kernel/svm/svm_mapping.h
@@ -26,7 +26,6 @@ ccl_device void svm_node_mapping(KernelGlobals *kg, ShaderData *sd, float *stack
 	tfm.x = read_node_float(kg, offset);
 	tfm.y = read_node_float(kg, offset);
 	tfm.z = read_node_float(kg, offset);
-	tfm.w = read_node_float(kg, offset);
 
 	float3 r = transform_point(&tfm, v);
 	stack_store_float3(stack, out_offset, r);
diff --git a/intern/cycles/kernel/svm/svm_tex_coord.h b/intern/cycles/kernel/svm/svm_tex_coord.h
index c94327401f5..6ff39e5f587 100644
--- a/intern/cycles/kernel/svm/svm_tex_coord.h
+++ b/intern/cycles/kernel/svm/svm_tex_coord.h
@@ -42,7 +42,6 @@ ccl_device void svm_node_tex_coord(KernelGlobals *kg,
 				tfm.x = read_node_float(kg, offset);
 				tfm.y = read_node_float(kg, offset);
 				tfm.z = read_node_float(kg, offset);
-				tfm.w = read_node_float(kg, offset);
 				data = transform_point(&tfm, data);
 			}
 			break;
@@ -123,7 +122,6 @@ ccl_device void svm_node_tex_coord_bump_dx(KernelGlobals *kg,
 				tfm.x = read_node_float(kg, offset);
 				tfm.y = read_node_float(kg, offset);
 				tfm.z = read_node_float(kg, offset);
-				tfm.w = read_node_float(kg, offset);
 				data = transform_point(&tfm, data);
 			}
 			break;
@@ -207,7 +205,6 @@ ccl_device void svm_node_tex_coord_bump_dy(KernelGlobals *kg,
 				tfm.x = read_node_float(kg, offset);
 				tfm.y = read_node_float(kg, offset);
 				tfm.z = read_node_float(kg, offset);
-				tfm.w = read_node_float(kg, offset);
 				data = transform_point(&tfm, data);
 			}
 			break;
diff --git a/intern/cycles/kernel/svm/svm_voxel.h b/intern/cycles/kernel/svm/svm_voxel.h
index d967516a5c9..43b433683e0 100644
--- a/intern/cycles/kernel/svm/svm_voxel.h
+++ b/intern/cycles/kernel/svm/svm_voxel.h
@@ -39,7 +39,6 @@ ccl_device void svm_node_tex_voxel(KernelGlobals *kg,
 		tfm.x = read_node_float(kg, offset);
 		tfm.y = read_node_float(kg, offset);
 		tfm.z = read_node_float(kg, offset);
-		tfm.w = read_node_float(kg, offset);
 		co = transform_point(&tfm, co);
 	}