diff options
Diffstat (limited to 'intern/cycles/kernel/split/kernel_buffer_update.h')
| -rw-r--r-- | intern/cycles/kernel/split/kernel_buffer_update.h | 64 |
1 files changed, 20 insertions, 44 deletions
diff --git a/intern/cycles/kernel/split/kernel_buffer_update.h b/intern/cycles/kernel/split/kernel_buffer_update.h index 4c1fdd2d69c..3b61319e349 100644 --- a/intern/cycles/kernel/split/kernel_buffer_update.h +++ b/intern/cycles/kernel/split/kernel_buffer_update.h @@ -75,57 +75,34 @@ ccl_device void kernel_buffer_update(KernelGlobals *kg, if(ray_index != QUEUE_EMPTY_SLOT) { #endif - ccl_global uint *rng_state = kernel_split_params.rng_state; int stride = kernel_split_params.stride; ccl_global char *ray_state = kernel_split_state.ray_state; -#ifdef __KERNEL_DEBUG__ - DebugData *debug_data = &kernel_split_state.debug_data[ray_index]; -#endif ccl_global PathState *state = &kernel_split_state.path_state[ray_index]; PathRadiance *L = &kernel_split_state.path_radiance[ray_index]; ccl_global Ray *ray = &kernel_split_state.ray[ray_index]; ccl_global float3 *throughput = &kernel_split_state.throughput[ray_index]; - ccl_global float *L_transparent = &kernel_split_state.L_transparent[ray_index]; - RNG rng = kernel_split_state.rng[ray_index]; - ccl_global float *buffer = kernel_split_params.buffer; - - unsigned int work_index; - ccl_global uint *initial_rng; - - unsigned int sample; - unsigned int tile_x; - unsigned int tile_y; - unsigned int pixel_x; - unsigned int pixel_y; - - work_index = kernel_split_state.work_array[ray_index]; - sample = get_work_sample(kg, work_index, ray_index) + kernel_split_params.start_sample; - get_work_pixel_tile_position(kg, &pixel_x, &pixel_y, - &tile_x, &tile_y, - work_index, - ray_index); - initial_rng = rng_state; - - rng_state += kernel_split_params.offset + pixel_x + pixel_y*stride; - buffer += (kernel_split_params.offset + pixel_x + pixel_y*stride) * kernel_data.film.pass_stride; if(IS_STATE(ray_state, ray_index, RAY_UPDATE_BUFFER)) { -#ifdef __KERNEL_DEBUG__ - kernel_write_debug_passes(kg, buffer, state, debug_data, sample); -#endif + uint work_index = kernel_split_state.work_array[ray_index]; + uint sample = get_work_sample(kg, work_index, ray_index) + kernel_split_params.start_sample; + + uint tile_x, tile_y, pixel_x, pixel_y; + get_work_pixel_tile_position(kg, &pixel_x, &pixel_y, &tile_x, &tile_y, work_index, ray_index); + + ccl_global float *buffer = kernel_split_params.buffer; + buffer += (kernel_split_params.offset + pixel_x + pixel_y*stride) * kernel_data.film.pass_stride; /* accumulate result in output buffer */ bool is_shadow_catcher = (state->flag & PATH_RAY_SHADOW_CATCHER); - kernel_write_result(kg, buffer, sample, L, 1.0f - (*L_transparent), is_shadow_catcher); - - path_rng_end(kg, rng_state, rng); + kernel_write_result(kg, buffer, sample, L, is_shadow_catcher); ASSIGN_RAY_STATE(ray_state, ray_index, RAY_TO_REGENERATE); } if(IS_STATE(ray_state, ray_index, RAY_TO_REGENERATE)) { /* We have completed current work; So get next work */ + uint work_index; int valid_work = get_next_work(kg, &work_index, ray_index); if(!valid_work) { /* If work is invalid, this means no more work is available and the thread may exit */ @@ -135,32 +112,33 @@ ccl_device void kernel_buffer_update(KernelGlobals *kg, if(IS_STATE(ray_state, ray_index, RAY_TO_REGENERATE)) { kernel_split_state.work_array[ray_index] = work_index; /* Get the sample associated with the current work */ - sample = get_work_sample(kg, work_index, ray_index) + kernel_split_params.start_sample; + uint sample = get_work_sample(kg, work_index, ray_index) + kernel_split_params.start_sample; /* Get pixel and tile position associated with current work */ + uint tile_x, tile_y, pixel_x, pixel_y; get_work_pixel_tile_position(kg, &pixel_x, &pixel_y, &tile_x, &tile_y, work_index, ray_index); /* Remap rng_state according to the current work */ - rng_state = initial_rng + kernel_split_params.offset + pixel_x + pixel_y*stride; + ccl_global uint *rng_state = kernel_split_params.rng_state; + rng_state += kernel_split_params.offset + pixel_x + pixel_y*stride; + /* Remap buffer according to the current work */ + ccl_global float *buffer = kernel_split_params.buffer; buffer += (kernel_split_params.offset + pixel_x + pixel_y*stride) * kernel_data.film.pass_stride; /* Initialize random numbers and ray. */ - kernel_path_trace_setup(kg, rng_state, sample, pixel_x, pixel_y, &rng, ray); + uint rng_hash; + kernel_path_trace_setup(kg, rng_state, sample, pixel_x, pixel_y, &rng_hash, ray); if(ray->t != 0.0f) { - /* Initialize throughput, L_transparent, Ray, PathState; + /* Initialize throughput, path radiance, Ray, PathState; * These rays proceed with path-iteration. */ *throughput = make_float3(1.0f, 1.0f, 1.0f); - *L_transparent = 0.0f; path_radiance_init(L, kernel_data.film.use_light_pass); - path_state_init(kg, &kernel_split_state.sd_DL_shadow[ray_index], state, &rng, sample, ray); + path_state_init(kg, &kernel_split_state.sd_DL_shadow[ray_index], state, rng_hash, sample, ray); #ifdef __SUBSURFACE__ kernel_path_subsurface_init_indirect(&kernel_split_state.ss_rays[ray_index]); #endif -#ifdef __KERNEL_DEBUG__ - debug_data_init(debug_data); -#endif ASSIGN_RAY_STATE(ray_state, ray_index, RAY_REGENERATED); enqueue_flag = 1; } @@ -169,13 +147,11 @@ ccl_device void kernel_buffer_update(KernelGlobals *kg, float4 L_rad = make_float4(0.0f, 0.0f, 0.0f, 0.0f); /* Accumulate result in output buffer. */ kernel_write_pass_float4(buffer, sample, L_rad); - path_rng_end(kg, rng_state, rng); ASSIGN_RAY_STATE(ray_state, ray_index, RAY_TO_REGENERATE); } } } - kernel_split_state.rng[ray_index] = rng; #ifndef __COMPUTE_DEVICE_GPU__ } |