Page MenuHome
Differential D12370 Diff 41359 intern/cycles/kernel/device/gpu/kernel.h

Changeset View

Standalone View

View Options

intern/cycles/kernel/device/gpu/kernel.h

Show First 20 Lines • Show All 406 Lines • ▼ Show 20 Lines
*/ */
/* Common implementation for float destination. */ /* Common implementation for float destination. */
template<typename Processor> template<typename Processor>
ccl_device_inline void kernel_gpu_film_convert_common(const KernelFilmConvert *kfilm_convert, ccl_device_inline void kernel_gpu_film_convert_common(const KernelFilmConvert *kfilm_convert,
float *pixels, float *pixels,
float *render_buffer, float *render_buffer,
int num_pixels, int num_pixels,
int width,
int offset, int offset,
int stride, int stride,
int dst_offset, int dst_offset,
int dst_stride,
const Processor &processor) const Processor &processor)
{ {
const int render_pixel_index = ccl_gpu_global_id_x(); const int render_pixel_index = ccl_gpu_global_id_x();
if (render_pixel_index >= num_pixels) { if (render_pixel_index >= num_pixels) {
return; return;
} }
const uint64_t render_buffer_offset = (uint64_t)render_pixel_index * kfilm_convert->pass_stride; const uint64_t render_buffer_offset = (uint64_t)render_pixel_index * kfilm_convert->pass_stride;
ccl_global const float *buffer = render_buffer + render_buffer_offset; ccl_global const float *buffer = render_buffer + render_buffer_offset;
ccl_global float *pixel = pixels + ccl_global float *pixel = pixels +
(render_pixel_index + dst_offset) * kfilm_convert->pixel_stride; (render_pixel_index + dst_offset) * kfilm_convert->pixel_stride;
processor(kfilm_convert, buffer, pixel); processor(kfilm_convert, buffer, pixel);
} }
/* Common implementation for half4 destination and 4-channel input pass. */ /* Common implementation for half4 destination and 4-channel input pass. */
template<typename Processor> template<typename Processor>
ccl_device_inline void kernel_gpu_film_convert_half_rgba_common_rgba( ccl_device_inline void kernel_gpu_film_convert_half_rgba_common_rgba(
const KernelFilmConvert *kfilm_convert, const KernelFilmConvert *kfilm_convert,
uchar4 *rgba, uchar4 *rgba,
float *render_buffer, float *render_buffer,
int num_pixels, int num_pixels,
int width,
int offset, int offset,
int stride, int stride,
int rgba_offset, int rgba_offset,
int rgba_stride,
const Processor &processor) const Processor &processor)
{ {
const int render_pixel_index = ccl_gpu_global_id_x(); const int render_pixel_index = ccl_gpu_global_id_x();
if (render_pixel_index >= num_pixels) { if (render_pixel_index >= num_pixels) {
return; return;
} }
const uint64_t render_buffer_offset = (uint64_t)render_pixel_index * kfilm_convert->pass_stride; const uint64_t render_buffer_offset = (uint64_t)render_pixel_index * kfilm_convert->pass_stride;
ccl_global const float *buffer = render_buffer + render_buffer_offset; ccl_global const float *buffer = render_buffer + render_buffer_offset;
float pixel[4]; float pixel[4];
processor(kfilm_convert, buffer, pixel); processor(kfilm_convert, buffer, pixel);
film_apply_pass_pixel_overlays_rgba(kfilm_convert, buffer, pixel); film_apply_pass_pixel_overlays_rgba(kfilm_convert, buffer, pixel);
ccl_global half *out = (ccl_global half *)rgba + (rgba_offset + render_pixel_index) * 4; const int x = render_pixel_index % width;
float4_store_half(out, make_float4(pixel[0], pixel[1], pixel[2], pixel[3])); const int y = render_pixel_index / width;
ccl_global half4 *out = ((ccl_global half4 *)rgba) + rgba_offset + y * rgba_stride + x;
float4_store_half((ccl_global half *)out, make_float4(pixel[0], pixel[1], pixel[2], pixel[3]));
} }
/* Common implementation for half4 destination and 3-channel input pass. */ /* Common implementation for half4 destination and 3-channel input pass. */
template<typename Processor> template<typename Processor>
ccl_device_inline void kernel_gpu_film_convert_half_rgba_common_rgb( ccl_device_inline void kernel_gpu_film_convert_half_rgba_common_rgb(
const KernelFilmConvert *kfilm_convert, const KernelFilmConvert *kfilm_convert,
uchar4 *rgba, uchar4 *rgba,
float *render_buffer, float *render_buffer,
int num_pixels, int num_pixels,
int width,
int offset, int offset,
int stride, int stride,
int rgba_offset, int rgba_offset,
int rgba_stride,
const Processor &processor) const Processor &processor)
{ {
kernel_gpu_film_convert_half_rgba_common_rgba( kernel_gpu_film_convert_half_rgba_common_rgba(
kfilm_convert, kfilm_convert,
rgba, rgba,
render_buffer, render_buffer,
num_pixels, num_pixels,
width,
offset, offset,
stride, stride,
rgba_offset, rgba_offset,
rgba_stride,
[&processor](const KernelFilmConvert *kfilm_convert, [&processor](const KernelFilmConvert *kfilm_convert,
ccl_global const float *buffer, ccl_global const float *buffer,
float *pixel_rgba) { float *pixel_rgba) {
processor(kfilm_convert, buffer, pixel_rgba); processor(kfilm_convert, buffer, pixel_rgba);
pixel_rgba[3] = 1.0f; pixel_rgba[3] = 1.0f;
}); });
} }
/* Common implementation for half4 destination and single channel input pass. */ /* Common implementation for half4 destination and single channel input pass. */
template<typename Processor> template<typename Processor>
ccl_device_inline void kernel_gpu_film_convert_half_rgba_common_value( ccl_device_inline void kernel_gpu_film_convert_half_rgba_common_value(
const KernelFilmConvert *kfilm_convert, const KernelFilmConvert *kfilm_convert,
uchar4 *rgba, uchar4 *rgba,
float *render_buffer, float *render_buffer,
int num_pixels, int num_pixels,
int width,
int offset, int offset,
int stride, int stride,
int rgba_offset, int rgba_offset,
int rgba_stride,
const Processor &processor) const Processor &processor)
{ {
kernel_gpu_film_convert_half_rgba_common_rgba( kernel_gpu_film_convert_half_rgba_common_rgba(
kfilm_convert, kfilm_convert,
rgba, rgba,
render_buffer, render_buffer,
num_pixels, num_pixels,
width,
offset, offset,
stride, stride,
rgba_offset, rgba_offset,
rgba_stride,
[&processor](const KernelFilmConvert *kfilm_convert, [&processor](const KernelFilmConvert *kfilm_convert,
ccl_global const float *buffer, ccl_global const float *buffer,
float *pixel_rgba) { float *pixel_rgba) {
float value; float value;
processor(kfilm_convert, buffer, &value); processor(kfilm_convert, buffer, &value);
pixel_rgba[0] = value; pixel_rgba[0] = value;
pixel_rgba[1] = value; pixel_rgba[1] = value;
pixel_rgba[2] = value; pixel_rgba[2] = value;
pixel_rgba[3] = 1.0f; pixel_rgba[3] = 1.0f;
}); });
} }
#define KERNEL_FILM_CONVERT_PROC(name) \ #define KERNEL_FILM_CONVERT_PROC(name) \
ccl_gpu_kernel(GPU_KERNEL_BLOCK_NUM_THREADS, GPU_KERNEL_MAX_REGISTERS) name ccl_gpu_kernel(GPU_KERNEL_BLOCK_NUM_THREADS, GPU_KERNEL_MAX_REGISTERS) name
#define KERNEL_FILM_CONVERT_DEFINE(variant, channels) \ #define KERNEL_FILM_CONVERT_DEFINE(variant, channels) \
KERNEL_FILM_CONVERT_PROC(kernel_gpu_film_convert_##variant) \ KERNEL_FILM_CONVERT_PROC(kernel_gpu_film_convert_##variant) \
(const KernelFilmConvert kfilm_convert, \ (const KernelFilmConvert kfilm_convert, \
float *pixels, \ float *pixels, \
float *render_buffer, \ float *render_buffer, \
int num_pixels, \ int num_pixels, \
int width, \
int offset, \ int offset, \
int stride, \ int stride, \
int rgba_offset) \ int rgba_offset, \
int rgba_stride) \
{ \ { \
kernel_gpu_film_convert_common(&kfilm_convert, \ kernel_gpu_film_convert_common(&kfilm_convert, \
pixels, \ pixels, \
render_buffer, \ render_buffer, \
num_pixels, \ num_pixels, \
width, \
offset, \ offset, \
stride, \ stride, \
rgba_offset, \ rgba_offset, \
rgba_stride, \
film_get_pass_pixel_##variant); \ film_get_pass_pixel_##variant); \
} \ } \
KERNEL_FILM_CONVERT_PROC(kernel_gpu_film_convert_##variant##_half_rgba) \ KERNEL_FILM_CONVERT_PROC(kernel_gpu_film_convert_##variant##_half_rgba) \
(const KernelFilmConvert kfilm_convert, \ (const KernelFilmConvert kfilm_convert, \
uchar4 *rgba, \ uchar4 *rgba, \
float *render_buffer, \ float *render_buffer, \
int num_pixels, \ int num_pixels, \
int width, \
int offset, \ int offset, \
int stride, \ int stride, \
int rgba_offset) \ int rgba_offset, \
int rgba_stride) \
{ \ { \
kernel_gpu_film_convert_half_rgba_common_##channels(&kfilm_convert, \ kernel_gpu_film_convert_half_rgba_common_##channels(&kfilm_convert, \
rgba, \ rgba, \
render_buffer, \ render_buffer, \
num_pixels, \ num_pixels, \
width, \
offset, \ offset, \
stride, \ stride, \
rgba_offset, \ rgba_offset, \
rgba_stride, \
film_get_pass_pixel_##variant); \ film_get_pass_pixel_##variant); \
} }
KERNEL_FILM_CONVERT_DEFINE(depth, value) KERNEL_FILM_CONVERT_DEFINE(depth, value)
KERNEL_FILM_CONVERT_DEFINE(mist, value) KERNEL_FILM_CONVERT_DEFINE(mist, value)
KERNEL_FILM_CONVERT_DEFINE(sample_count, value) KERNEL_FILM_CONVERT_DEFINE(sample_count, value)
KERNEL_FILM_CONVERT_DEFINE(float, value) KERNEL_FILM_CONVERT_DEFINE(float, value)
▲ Show 20 Lines • Show All 252 Lines • Show Last 20 Lines