Commit: 9ddbf6ab5b36f459c71d86a3a39e8489ac039601
Author: Thomas Dinges
Date: Sat May 28 01:35:20 2016 +0200
Branches: soc-2016-cycles_images
https://developer.blender.org/rB9ddbf6ab5b36f459c71d86a3a39e8489ac039601
Cycles: Add some basic code for half4 and half image textures.
Not finished yet, so noting to see here yet.
===================================================================
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 bb303b3..b7177ab 100644
--- a/intern/cycles/kernel/kernel_compat_cpu.h
+++ b/intern/cycles/kernel/kernel_compat_cpu.h
@@ -486,8 +486,10 @@ 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<half> texture_image_half;
typedef texture_image<float4> texture_image_float4;
typedef texture_image<uchar4> texture_image_uchar4;
+typedef texture_image<half4> texture_image_half4;
/* Macros to handle different memory storage on different devices */
diff --git a/intern/cycles/kernel/kernel_globals.h
b/intern/cycles/kernel/kernel_globals.h
index e06c68f..8e66a3a 100644
--- a/intern/cycles/kernel/kernel_globals.h
+++ b/intern/cycles/kernel/kernel_globals.h
@@ -37,8 +37,10 @@ struct VolumeStep;
typedef struct KernelGlobals {
texture_image_uchar4 texture_byte4_images[TEX_NUM_BYTE4_CPU];
texture_image_float4 texture_float4_images[TEX_NUM_FLOAT4_CPU];
+ texture_image_half4 texture_half4_images[TEX_NUM_HALF4_CPU];
texture_image_float texture_float_images[TEX_NUM_FLOAT_CPU];
texture_image_uchar texture_byte_images[TEX_NUM_BYTE_CPU];
+ texture_image_half texture_half_images[TEX_NUM_HALF_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 d8a83f6..f11c85d 100644
--- a/intern/cycles/kernel/kernels/cpu/kernel.cpp
+++ b/intern/cycles/kernel/kernels/cpu/kernel.cpp
@@ -154,6 +154,38 @@ void kernel_tex_copy(KernelGlobals *kg,
tex->extension = extension;
}
}
+ else if(strstr(name, "__tex_image_half4")) {
+ texture_image_half4 *tex = NULL;
+ int id = atoi(name + strlen("__tex_image_half4_"));
+ int array_index = id - TEX_START_HALF4_CPU;
+
+ if(array_index >= 0 && array_index < TEX_NUM_HALF4_CPU) {
+ tex = &kg->texture_half4_images[array_index];
+ }
+
+ if(tex) {
+ tex->data = (half4*)mem;
+ tex->dimensions_set(width, height, depth);
+ tex->interpolation = interpolation;
+ tex->extension = extension;
+ }
+ }
+ else if(strstr(name, "__tex_image_half")) {
+ texture_image_half *tex = NULL;
+ int id = atoi(name + strlen("__tex_image_half_"));
+ int array_index = id - TEX_START_HALF_CPU;
+
+ if(array_index >= 0 && array_index < TEX_NUM_HALF_CPU) {
+ tex = &kg->texture_half_images[array_index];
+ }
+
+ if(tex) {
+ tex->data = (half*)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 71dc85f..bdcb8c4 100644
--- a/intern/cycles/render/image.cpp
+++ b/intern/cycles/render/image.cpp
@@ -54,10 +54,14 @@ ImageManager::ImageManager(const DeviceInfo& info)
tex_num_images[IMAGE_DATA_TYPE_BYTE4] = TEX_NUM_BYTE4_ ## ARCH;
\
tex_num_images[IMAGE_DATA_TYPE_FLOAT] = TEX_NUM_FLOAT_ ## ARCH;
\
tex_num_images[IMAGE_DATA_TYPE_BYTE] = TEX_NUM_BYTE_ ## ARCH; \
+ tex_num_images[IMAGE_DATA_TYPE_HALF4] = TEX_NUM_HALF4_ ## ARCH;
\
+ tex_num_images[IMAGE_DATA_TYPE_HALF] = TEX_NUM_HALF_ ## ARCH; \
tex_start_images[IMAGE_DATA_TYPE_FLOAT4] = TEX_START_FLOAT4_ ##
ARCH; \
tex_start_images[IMAGE_DATA_TYPE_BYTE4] = TEX_START_BYTE4_ ##
ARCH; \
tex_start_images[IMAGE_DATA_TYPE_FLOAT] = TEX_START_FLOAT_ ##
ARCH; \
tex_start_images[IMAGE_DATA_TYPE_BYTE] = TEX_START_BYTE_ ##
ARCH; \
+ tex_start_images[IMAGE_DATA_TYPE_HALF4] = TEX_START_HALF4_ ##
ARCH; \
+ tex_start_images[IMAGE_DATA_TYPE_HALF] = TEX_START_HALF_ ##
ARCH; \
}
if(device_type == DEVICE_CPU) {
@@ -80,10 +84,14 @@ ImageManager::ImageManager(const DeviceInfo& info)
tex_num_images[IMAGE_DATA_TYPE_BYTE4] = 0;
tex_num_images[IMAGE_DATA_TYPE_FLOAT] = 0;
tex_num_images[IMAGE_DATA_TYPE_BYTE] = 0;
+ tex_num_images[IMAGE_DATA_TYPE_HALF4] = 0;
+ tex_num_images[IMAGE_DATA_TYPE_HALF] = 0;
tex_start_images[IMAGE_DATA_TYPE_FLOAT4] = 0;
tex_start_images[IMAGE_DATA_TYPE_BYTE4] = 0;
tex_start_images[IMAGE_DATA_TYPE_FLOAT] = 0;
tex_start_images[IMAGE_DATA_TYPE_BYTE] = 0;
+ tex_start_images[IMAGE_DATA_TYPE_HALF4] = 0;
+ tex_start_images[IMAGE_DATA_TYPE_HALF] = 0;
assert(0);
}
@@ -230,6 +238,10 @@ string ImageManager::name_from_type(int type)
return "float";
else if(type == IMAGE_DATA_TYPE_BYTE)
return "byte";
+ else if(type == IMAGE_DATA_TYPE_HALF4)
+ return "half4";
+ else if(type == IMAGE_DATA_TYPE_HALF)
+ return "half";
else
return "byte4";
}
@@ -265,11 +277,16 @@ int ImageManager::add_image(const string& filename,
if(type == IMAGE_DATA_TYPE_FLOAT || type == IMAGE_DATA_TYPE_FLOAT4)
is_float = true;
- /* No single channel textures on CUDA (Fermi) and OpenCL, use available
slots */
- if(type == IMAGE_DATA_TYPE_FLOAT && tex_num_images[type] == 0)
+ /* No single channel and half textures on CUDA (Fermi) and OpenCL, use
available slots */
+ if((type == IMAGE_DATA_TYPE_FLOAT ||
+ type == IMAGE_DATA_TYPE_HALF4 ||
+ type == IMAGE_DATA_TYPE_HALF) &&
+ tex_num_images[type] == 0) {
type = IMAGE_DATA_TYPE_FLOAT4;
- if(type == IMAGE_DATA_TYPE_BYTE && tex_num_images[type] == 0)
+ }
+ 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++) {
@@ -645,6 +662,118 @@ bool ImageManager::file_load_float_image(Image *img,
ImageDataType type, device_
return true;
}
+template<typename T>
+bool ImageManager::file_load_half_image(Image *img, ImageDataType type,
device_vector<T>& tex_img)
+{
+ ImageInput *in = NULL;
+ int width, height, depth, components;
+
+ if(!file_load_image_generic(img, &in, width, height, depth, components))
+ return false;
+
+ /* read RGBA pixels */
+ half *pixels = (half*)tex_img.resize(width, height, depth);
+ if(pixels == NULL) {
+ return false;
+ }
+ bool cmyk = false;
+
+ if(in) {
+ half *readpixels = pixels;
+ vector<half> tmppixels;
+
+ if(components > 4) {
+ tmppixels.resize(((size_t)width)*height*components);
+ readpixels = &tmppixels[0];
+ }
+
+ if(depth <= 1) {
+ int scanlinesize = width*components*sizeof(half);
+
+ in->read_image(TypeDesc::HALF,
+ (uchar*)readpixels +
(height-1)*scanlinesize, /*TODO(dingto): why uchar cast? */
+ AutoStride,
+ -scanlinesize,
+ AutoStride);
+ }
+ else {
+ in->read_image(TypeDesc::HALF, (uchar*)readpixels);
+ }
+
+ if(components > 4) {
+ size_t dimensions = ((size_t)width)*height;
+ for(size_t i = dimensions-1, pixel = 0; pixel <
dimensions; pixel++, i--) {
+ pixels[i*4+3] = tmppixels[i*components+3];
+ pixels[i*4+2] = tmppixels[i*components+2];
+ pixels[i*4+1] = tmppixels[i*components+1];
+ pixels[i*4+0] = tmppixels[i*components+0];
+ }
+
+ tmppixels.clear();
+ }
+
+ cmyk = strcmp(in->format_name(), "jpeg") == 0 && components ==
4; /*TODO(dingto): kick, we wont have jpeg in half format ? */
+
+ in->close();
+ delete in;
+ }
+#if 0 /* TODO(dingto): add support for half here ? */
+ else {
+ builtin_image_float_pixels_cb(img->filename, img->builtin_data,
pixels);
+ }
+#endif
+
+ /* Check if we actually have a half4 slot, in case components == 1, but
device
+ * doesn't support single channel textures. */
+ if(type == IMAGE_DATA_TYPE_FLOAT4) {
+ size_t num_pixels = ((size_t)width) * height * depth;
+ if(cmyk) {
+ /* CMYK */
+ for(size_t i = num_pixels-1, pixel = 0; pixel <
num_pixels; pixel++, i--) {
+ pixels[i*4+3] = 255;
+ pixels[i*4+2] =
(pixels[i*4+2]*pixels[i*4+3])/255;
+ pixels[i*4+1] =
(pixels[i*4+1]*pixels[i*4+3])/255;
+ pixels[i*4+0] =
(pixels[i*4+0]*pixels[i*4+3])/255;
+ }
+ }
+ else if(components == 2) {
+ /* grayscale + alpha */
+ for(size_t i = num_pixels-1, pixel = 0; pixel <
num_pixels; pixel++, i--) {
+ pixels[i*4+3] = pixels[i*2+1];
+ pixels[i*4+2] = pixels[i*2+0];
+ pixels[i*4+1] = pixels[i*2+0];
+ pixels[i*4+0] = pixels[i*2+0];
+ }
+ }
+ else if(components == 3) {
+ /* RGB */
+ for(size_t i = num_pixels-1, pixel = 0; pixel <
num_pixels; pixel++, i--) {
+ pixels[i*4+3] = 1.0f;
+ pixels[i*4+2] = pixels[i*3+2];
+ pixels[i*4+1] = pixels[i*3+1];
+ pixels[i*4+0] = pixels[i*3+0];
+ }
+ }
+ else if(components == 1) {
+ /* grayscale */
+ for(size_t i = num_pixels-1, pixel = 0; pixel <
num_pixels; pixel++, i--) {
+ pixels[i*4+3] = 1.0f;
+ pixels[i*4+2] = pixels[i];
+ pixels[i*4+1] = pixels[i];
+ pixels[i*4+0] = pixels[i];
+ }
+ }
+
+ if(img->use_alpha == false) {
+ for(size_t i = num_pixels-1, pixel = 0; pixel <
num_pixels; pixel++, i--) {
+ pixels[i*4+3] = 1.0f;
+ }
+ }
+ }
+
+ return true;
+}
+
void ImageManager::device_load_image(Device *device, DeviceScene *dscene,
ImageDataType type, int slot, Progress *progress)
{
if(progress->get_cancel())
@@ -744,7 +873,7 @@ void ImageManager::device_load_image(Device *device,
DeviceScene *dscene, ImageD
img->extension);
}
}
- else {
+ else if(type == IMAGE_DATA_TYPE_BYTE){
device_vector<uchar>& tex_img = dscene->tex_byte_image[slot];
if(tex_img.device_pointer) {
@@ -767,6 +896,57 @@ void ImageManager::device_load_image(Device *device,
DeviceScene *dscene, ImageD
img->extension);
}
}
+ else if(type == IMAGE_DATA_TYPE_HALF4){
+ device_vector<half4>& tex_img = dscene->tex_half4_image[slot];
+
+ if(tex_img.device_pointer) {
+ thread_scoped_lock device_lock(device_mutex);
+ device->tex_free(tex_img);
+ }
+
+ if(!file_load_float_image(img, type, tex_img)) {
+ /* on failure to load, we set a 1x1 pixels pink image */
+ half4 *pixels = (half4*)tex_img.resize(1, 1);
+
+#if 0 /* TODO(dingto): Fix this... */
+ pixels[0] = (half)TEX_IMAGE_MISSING_R;
+ pixels[1] = (half)TEX_IMAGE_MISSING_G;
+ pixels[2] = (half)TEX_IMAGE_MISSING_B;
+ pixels[3] = (half)TEX_IMAGE_MISSING_A;
+#endif
+ }
+
+ if(!pack_images) {
+ thread_scoped_lock device_
@@ Diff output truncated at 10240 characters. @@
_______________________________________________
Bf-blender-cvs mailing list
[email protected]
https://lists.blender.org/mailman/listinfo/bf-blender-cvs
