[Bf-blender-cvs] [3c85e1c] master: Cycles: Add support for single channel byte textures.
Thomas Dinges
noreply at git.blender.org
Thu May 12 14:55:07 CEST 2016
Commit: 3c85e1ca1a916fe2ded9ab508f4cd55a2ee22549
Author: Thomas Dinges
Date: Thu May 12 14:51:42 2016 +0200
Branches: master
https://developer.blender.org/rB3c85e1ca1a916fe2ded9ab508f4cd55a2ee22549
Cycles: Add support for single channel byte textures.
This way, we also save 3/4th of memory for single channel byte textures (e.g. Bump Maps).
Note: In order for this to work, the texture *must* have 1 channel only.
In Gimp you can e.g. do that via the menu: Image -> Mode -> Grayscale
===================================================================
M intern/cycles/kernel/kernel_compat_cpu.h
M intern/cycles/kernel/kernel_globals.h
M intern/cycles/kernel/kernels/cpu/kernel.cpp
M intern/cycles/render/image.cpp
M intern/cycles/render/image.h
M intern/cycles/render/scene.h
M intern/cycles/util/util_texture.h
===================================================================
diff --git a/intern/cycles/kernel/kernel_compat_cpu.h b/intern/cycles/kernel/kernel_compat_cpu.h
index 7fc8d2b..46bef96 100644
--- a/intern/cycles/kernel/kernel_compat_cpu.h
+++ b/intern/cycles/kernel/kernel_compat_cpu.h
@@ -109,6 +109,12 @@ 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(uchar r)
+ {
+ float f = r*(1.0f/255.0f);
+ return make_float4(f, f, f, 1.0);
+ }
+
ccl_always_inline float4 read(float r)
{
/* TODO(dingto): Optimize this, so interpolation
@@ -479,6 +485,7 @@ 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<uchar> texture_image_uchar;
typedef texture_image<float4> texture_image_float4;
typedef texture_image<uchar4> texture_image_uchar4;
@@ -490,17 +497,20 @@ typedef texture_image<uchar4> texture_image_uchar4;
#define kernel_tex_lookup(tex, t, offset, size) (kg->tex.lookup(t, offset, size))
#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_BYTE_START_CPU) ? kg->texture_byte_images[tex - TEX_IMAGE_BYTE_START_CPU].interp(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_BYTE_START_CPU) ? kg->texture_byte_images[tex - TEX_IMAGE_BYTE_START_CPU].interp_3d(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_BYTE_START_CPU) ? kg->texture_byte_images[tex - TEX_IMAGE_BYTE_START_CPU].interp_3d_ex(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))
diff --git a/intern/cycles/kernel/kernel_globals.h b/intern/cycles/kernel/kernel_globals.h
index 3af44e0..c44ea1b 100644
--- a/intern/cycles/kernel/kernel_globals.h
+++ b/intern/cycles/kernel/kernel_globals.h
@@ -35,6 +35,7 @@ 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];
+ texture_image_uchar texture_byte_images[TEX_NUM_BYTE_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 960012e..365ce89 100644
--- a/intern/cycles/kernel/kernels/cpu/kernel.cpp
+++ b/intern/cycles/kernel/kernels/cpu/kernel.cpp
@@ -138,6 +138,22 @@ void kernel_tex_copy(KernelGlobals *kg,
tex->extension = extension;
}
}
+ else if(strstr(name, "__tex_image_byte")) {
+ texture_image_uchar *tex = NULL;
+ int id = atoi(name + strlen("__tex_image_byte_"));
+ int array_index = id - TEX_IMAGE_BYTE_START_CPU;
+
+ if(array_index >= 0 && array_index < TEX_NUM_BYTE_IMAGES_CPU) {
+ tex = &kg->texture_byte_images[array_index];
+ }
+
+ if(tex) {
+ tex->data = (uchar*)mem;
+ tex->dimensions_set(width, height, depth);
+ tex->interpolation = interpolation;
+ tex->extension = extension;
+ }
+ }
else
assert(0);
}
diff --git a/intern/cycles/render/image.cpp b/intern/cycles/render/image.cpp
index 0985cca..23e0708 100644
--- a/intern/cycles/render/image.cpp
+++ b/intern/cycles/render/image.cpp
@@ -43,40 +43,50 @@ ImageManager::ImageManager(const DeviceInfo& info)
tex_num_images[IMAGE_DATA_TYPE_BYTE4] = TEX_NUM_BYTE4_IMAGES_CPU;
tex_num_images[IMAGE_DATA_TYPE_FLOAT4] = TEX_NUM_FLOAT4_IMAGES_CPU;
tex_num_images[IMAGE_DATA_TYPE_FLOAT] = TEX_NUM_FLOAT_IMAGES_CPU;
+ tex_num_images[IMAGE_DATA_TYPE_BYTE] = TEX_NUM_BYTE_IMAGES_CPU;
tex_image_byte4_start = TEX_IMAGE_BYTE4_START_CPU;
tex_image_float_start = TEX_IMAGE_FLOAT_START_CPU;
+ tex_image_byte_start = TEX_IMAGE_BYTE_START_CPU;
}
/* CUDA (Fermi) */
else if((info.type == DEVICE_CUDA || info.type == DEVICE_MULTI) && !info.extended_images) {
tex_num_images[IMAGE_DATA_TYPE_BYTE4] = TEX_NUM_BYTE4_IMAGES_CUDA;
tex_num_images[IMAGE_DATA_TYPE_FLOAT4] = TEX_NUM_FLOAT4_IMAGES_CUDA;
tex_num_images[IMAGE_DATA_TYPE_FLOAT] = TEX_NUM_FLOAT_IMAGES_CUDA;
+ tex_num_images[IMAGE_DATA_TYPE_BYTE] = TEX_NUM_BYTE_IMAGES_CUDA;
tex_image_byte4_start = TEX_IMAGE_BYTE4_START_CUDA;
tex_image_float_start = TEX_IMAGE_FLOAT_START_CUDA;
+ tex_image_byte_start = TEX_IMAGE_BYTE_START_CUDA;
}
/* CUDA (Kepler and above) */
else if((info.type == DEVICE_CUDA || info.type == DEVICE_MULTI) && info.extended_images) {
tex_num_images[IMAGE_DATA_TYPE_BYTE4] = TEX_NUM_BYTE4_IMAGES_CUDA_KEPLER;
tex_num_images[IMAGE_DATA_TYPE_FLOAT4] = TEX_NUM_FLOAT4_IMAGES_CUDA_KEPLER;
tex_num_images[IMAGE_DATA_TYPE_FLOAT] = TEX_NUM_FLOAT_IMAGES_CUDA_KEPLER;
+ tex_num_images[IMAGE_DATA_TYPE_BYTE] = TEX_NUM_BYTE_IMAGES_CUDA_KEPLER;
tex_image_byte4_start = TEX_IMAGE_BYTE4_START_CUDA_KEPLER;
tex_image_float_start = TEX_IMAGE_FLOAT_START_CUDA_KEPLER;
+ tex_image_byte_start = TEX_IMAGE_BYTE_START_CUDA_KEPLER;
}
/* OpenCL */
else if(info.pack_images) {
tex_num_images[IMAGE_DATA_TYPE_BYTE4] = TEX_NUM_BYTE4_IMAGES_OPENCL;
tex_num_images[IMAGE_DATA_TYPE_FLOAT4] = TEX_NUM_FLOAT4_IMAGES_OPENCL;
tex_num_images[IMAGE_DATA_TYPE_FLOAT] = TEX_NUM_FLOAT_IMAGES_OPENCL;
+ tex_num_images[IMAGE_DATA_TYPE_BYTE] = TEX_NUM_BYTE_IMAGES_OPENCL;
tex_image_byte4_start = TEX_IMAGE_BYTE4_START_OPENCL;
tex_image_float_start = TEX_IMAGE_FLOAT_START_OPENCL;
+ tex_image_byte_start = TEX_IMAGE_BYTE_START_OPENCL;
}
/* Should never happen */
else {
tex_num_images[IMAGE_DATA_TYPE_BYTE4] = 0;
tex_num_images[IMAGE_DATA_TYPE_FLOAT4] = 0;
tex_num_images[IMAGE_DATA_TYPE_FLOAT] = 0;
+ tex_num_images[IMAGE_DATA_TYPE_BYTE] = 0;
tex_image_byte4_start = 0;
tex_image_float_start = 0;
+ tex_image_byte_start = 0;
assert(0);
}
}
@@ -137,8 +147,13 @@ ImageManager::ImageDataType ImageManager::get_image_metadata(const string& filen
else
return IMAGE_DATA_TYPE_FLOAT;
}
- else
- return IMAGE_DATA_TYPE_BYTE4;
+ else {
+ if(channels > 1)
+ return IMAGE_DATA_TYPE_BYTE4;
+ else
+ return IMAGE_DATA_TYPE_BYTE;
+ }
+
}
ImageInput *in = ImageInput::create(filename);
@@ -192,12 +207,16 @@ ImageManager::ImageDataType ImageManager::get_image_metadata(const string& filen
else
return IMAGE_DATA_TYPE_FLOAT;
}
- else
- return IMAGE_DATA_TYPE_BYTE4;
+ else {
+ if(channels > 1)
+ return IMAGE_DATA_TYPE_BYTE4;
+ else
+ return IMAGE_DATA_TYPE_BYTE;
+ }
}
/* We use a consecutive slot counting scheme on the devices, in order
- * float4, byte4, float.
+ * float4, byte4, float, byte.
* These functions convert the slot ids from ImageManager "images" ones
* to device ones and vice versa. */
int ImageManager::type_index_to_flattened_slot(int slot, ImageDataType type)
@@ -206,13 +225,20 @@ int ImageManager::type_index_to_flattened_slot(int slot, ImageDataType type)
return slot + tex_image_byte4_start;
else if(type == IMAGE_DATA_TYPE_FLOAT)
return slot + tex_image_float_start;
+ else if(type == IMAGE_DATA_TYPE_BYTE)
+ return slot + tex_image_byte_start;
else
return slot;
}
int ImageManager::flattened_slot_to_type_index(int flat_slot, ImageDataType *type)
{
- if(flat_slot >= tex_image_float_start)
+ if(flat_slot >= tex_image_byte_start)
+ {
+ *type = IMAGE_DATA_TYPE_BYTE;
+ return flat_slot - tex_image_byte_start;
+ }
+ else if(flat_slot >= tex_image_float_start)
{
*type = IMAGE_DATA_TYPE_FLOAT;
return flat_slot - tex_image_float_start;
@@ -233,6 +259,8 @@ string ImageManager::name_from_type(int type)
return "float4";
else if(type == IMAGE_DATA_TYPE_FLOAT)
return "float";
+ else if(type == IMAGE_DATA_TYPE_BYTE)
+ return "byte";
else
return "byte4";
}
@@ -268,9 +296,11 @@ int ImageManager::add_image(const string& filename,
if(type == IMAGE_DATA_TYPE_FLOAT || type == IMAGE_DATA_TYPE_FLOAT4)
is_float = true;
- /* No float textures on GPU yet */
+ /* No float and byte textures on GPU yet */
if(type == IMAGE_DATA_TYPE_FLOAT && tex_num_images[type] == 0)
type = IMAGE_DATA_TYPE_FLOAT4;
+ if(type == IMAGE_DATA_TYPE_BYTE && tex_num_images[type] == 0)
+ type = IMAGE_DATA_TYPE_BYTE4;
/* Fnd existing image. */
for(slot = 0; slot < images[type].size(); slot++) {
@@ -531,6 +561,44 @@ bool ImageManager::file_load_byte4_image(Image *img, device_vector<uchar4>& tex_
return true;
}
+bool ImageManager::file_load_byte_image(Image *img, device_vector<uchar>& tex_img)
+{
+ ImageInput *in = NULL;
+ int width, height, depth, components;
+
+ if(!file_load_image_generic(img, &in, width, height, depth, components))
+ return false;
+
+ /* read BW pixels */
+ uchar *pixels = (uchar*)tex_img.resize(width, height, depth);
+ if(pixels == NULL) {
+ return false;
+ }
+
+ if(in) {
+ if(depth <= 1) {
+ int scanlinesize = width*components*sizeof(uchar);
+
+ in->read_image(TypeDesc::UINT8,
+ (uchar*)pixels + (((size_t)height)-1)*scanlinesize,
+ AutoStrid
@@ Diff output truncated at 10240 characters. @@
More information about the Bf-blender-cvs
mailing list