Cycles: Add support for single channel float textures on CPU.

Until now, single channel textures were packed into a float4, wasting 3 floats per pixel. Memory usage of such textures is now reduced by 3/4.
Voxel Attributes such as density, flame and heat benefit from this, but also Bumpmaps with one channel.
This commit also includes some cleanup and code deduplication for image loading.

Example Smoke render from Cosmos Laundromat: http://www.pasteall.org/pic/show.php?id=102972
Memory here went down from ~600MB to ~300MB.

Reviewers: #cycles, brecht

Differential Revision: https://developer.blender.org/D1981

3 files changed, 42 insertions, 4 deletions

diff --git a/intern/cycles/kernel/kernel_compat_cpu.h b/intern/cycles/kernel/kernel_compat_cpu.h
index ee8cee0ccfc..7fc8d2b5706 100644
--- a/intern/cycles/kernel/kernel_compat_cpu.h
+++ b/intern/cycles/kernel/kernel_compat_cpu.h
@@ -40,6 +40,7 @@
 #include "util_simd.h"
 #include "util_half.h"
 #include "util_types.h"
+#include "util_texture.h"
 
 #define ccl_addr_space
 
@@ -108,6 +109,13 @@ template<typename T> struct texture_image  {
 		return make_float4(r.x*f, r.y*f, r.z*f, r.w*f);
 	}
 
+	ccl_always_inline float4 read(float r)
+	{
+		/* TODO(dingto): Optimize this, so interpolation
+		 * happens on float instead of float4 */
+		return make_float4(r, r, r, 1.0f);
+	}
+
 	ccl_always_inline int wrap_periodic(int x, int width)
 	{
 		x %= width;
@@ -470,6 +478,7 @@ typedef texture<uint> texture_uint;
 typedef texture<int> texture_int;
 typedef texture<uint4> texture_uint4;
 typedef texture<uchar4> texture_uchar4;
+typedef texture_image<float> texture_image_float;
 typedef texture_image<float4> texture_image_float4;
 typedef texture_image<uchar4> texture_image_uchar4;
 
@@ -479,9 +488,21 @@ typedef texture_image<uchar4> texture_image_uchar4;
 #define kernel_tex_fetch_ssef(tex, index) (kg->tex.fetch_ssef(index))
 #define kernel_tex_fetch_ssei(tex, index) (kg->tex.fetch_ssei(index))
 #define kernel_tex_lookup(tex, t, offset, size) (kg->tex.lookup(t, offset, size))
-#define kernel_tex_image_interp(tex, x, y) ((tex < TEX_NUM_FLOAT4_IMAGES_CPU) ? kg->texture_float4_images[tex].interp(x, y) : kg->texture_byte4_images[tex - TEX_NUM_FLOAT4_IMAGES_CPU].interp(x, y))
-#define kernel_tex_image_interp_3d(tex, x, y, z) ((tex < TEX_NUM_FLOAT4_IMAGES_CPU) ? kg->texture_float4_images[tex].interp_3d(x, y, z) : kg->texture_byte4_images[tex - TEX_NUM_FLOAT4_IMAGES_CPU].interp_3d(x, y, z))
-#define kernel_tex_image_interp_3d_ex(tex, x, y, z, interpolation) ((tex < TEX_NUM_FLOAT4_IMAGES_CPU) ? kg->texture_float4_images[tex].interp_3d_ex(x, y, z, interpolation) : kg->texture_byte4_images[tex - TEX_NUM_FLOAT4_IMAGES_CPU].interp_3d_ex(x, y, z, interpolation))
+
+#define kernel_tex_image_interp(tex, x, y) \
+	((tex >= TEX_IMAGE_FLOAT_START_CPU) ? kg->texture_float_images[tex - TEX_IMAGE_FLOAT_START_CPU].interp(x, y) : \
+	(tex >= TEX_IMAGE_BYTE4_START_CPU) ? kg->texture_byte4_images[tex - TEX_IMAGE_BYTE4_START_CPU].interp(x, y) : \
+	kg->texture_float4_images[tex].interp(x, y))
+
+#define kernel_tex_image_interp_3d(tex, x, y, z) \
+	((tex >= TEX_IMAGE_FLOAT_START_CPU) ? kg->texture_float_images[tex - TEX_IMAGE_FLOAT_START_CPU].interp_3d(x, y, z) : \
+	(tex >= TEX_IMAGE_BYTE4_START_CPU) ? kg->texture_byte4_images[tex - TEX_IMAGE_BYTE4_START_CPU].interp_3d(x, y, z) : \
+	kg->texture_float4_images[tex].interp_3d(x, y, z))
+
+#define kernel_tex_image_interp_3d_ex(tex, x, y, z, interpolation) \
+	((tex >= TEX_IMAGE_FLOAT_START_CPU) ? kg->texture_float_images[tex - TEX_IMAGE_FLOAT_START_CPU].interp_3d_ex(x, y, z, interpolation) : \
+	(tex >= TEX_IMAGE_BYTE4_START_CPU) ? kg->texture_byte4_images[tex - TEX_IMAGE_BYTE4_START_CPU].interp_3d_ex(x, y, z, interpolation) : \
+	kg->texture_float4_images[tex].interp_3d_ex(x, y, z, interpolation))
 
 #define kernel_data (kg->__data)
 
diff --git a/intern/cycles/kernel/kernel_globals.h b/intern/cycles/kernel/kernel_globals.h
index b08d8924dc7..3af44e06179 100644
--- a/intern/cycles/kernel/kernel_globals.h
+++ b/intern/cycles/kernel/kernel_globals.h
@@ -34,6 +34,7 @@ struct OSLShadingSystem;
 typedef struct KernelGlobals {
 	texture_image_uchar4 texture_byte4_images[TEX_NUM_BYTE4_IMAGES_CPU];
 	texture_image_float4 texture_float4_images[TEX_NUM_FLOAT4_IMAGES_CPU];
+	texture_image_float texture_float_images[TEX_NUM_FLOAT_IMAGES_CPU];
 
 #  define KERNEL_TEX(type, ttype, name) ttype name;
 #  define KERNEL_IMAGE_TEX(type, ttype, name)
diff --git a/intern/cycles/kernel/kernels/cpu/kernel.cpp b/intern/cycles/kernel/kernels/cpu/kernel.cpp
index e7d0d8a0408..960012e95e3 100644
--- a/intern/cycles/kernel/kernels/cpu/kernel.cpp
+++ b/intern/cycles/kernel/kernels/cpu/kernel.cpp
@@ -106,10 +106,26 @@ void kernel_tex_copy(KernelGlobals *kg,
 			tex->extension = extension;
 		}
 	}
+	else if(strstr(name, "__tex_image_float")) {
+		texture_image_float *tex = NULL;
+		int id = atoi(name + strlen("__tex_image_float_"));
+		int array_index = id - TEX_IMAGE_FLOAT_START_CPU;
+
+		if(array_index >= 0 && array_index < TEX_NUM_FLOAT_IMAGES_CPU) {
+			tex = &kg->texture_float_images[array_index];
+		}
+
+		if(tex) {
+			tex->data = (float*)mem;
+			tex->dimensions_set(width, height, depth);
+			tex->interpolation = interpolation;
+			tex->extension = extension;
+		}
+	}
 	else if(strstr(name, "__tex_image_byte4")) {
 		texture_image_uchar4 *tex = NULL;
 		int id = atoi(name + strlen("__tex_image_byte4_"));
-		int array_index = id - TEX_NUM_FLOAT4_IMAGES_CPU;
+		int array_index = id - TEX_IMAGE_BYTE4_START_CPU;
 
 		if(array_index >= 0 && array_index < TEX_NUM_BYTE4_IMAGES_CPU) {
 			tex = &kg->texture_byte4_images[array_index];