intern/cycles/kernel/kernels/opencl/kernel_background_buffer_update.cl

   1 /*
   2  * Copyright 2011-2015 Blender Foundation
   3  *
   4  * Licensed under the Apache License, Version 2.0 (the "License");
   5  * you may not use this file except in compliance with the License.
   6  * You may obtain a copy of the License at
   7  *
   8  * http://www.apache.org/licenses/LICENSE-2.0
   9  *
  10  * Unless required by applicable law or agreed to in writing, software
  11  * distributed under the License is distributed on an "AS IS" BASIS,
  12  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  13  * See the License for the specific language governing permissions and
  14  * limitations under the License.
  15  */
  16
  17 #include "split/kernel_background_buffer_update.h"
  18
  19 __kernel void kernel_ocl_path_trace_background_buffer_update(
  20         ccl_global char *kg,
  21         ccl_constant KernelData *data,
  22         ccl_global float *per_sample_output_buffers,
  23         ccl_global uint *rng_state,
  24         ccl_global uint *rng_coop,             /* Required for buffer Update */
  25         ccl_global float3 *throughput_coop,    /* Required for background hit processing */
  26         PathRadiance *PathRadiance_coop,       /* Required for background hit processing and buffer Update */
  27         ccl_global Ray *Ray_coop,              /* Required for background hit processing */
  28         ccl_global PathState *PathState_coop,  /* Required for background hit processing */
  29         ccl_global float *L_transparent_coop,  /* Required for background hit processing and buffer Update */
  30         ccl_global char *ray_state,            /* Stores information on the current state of a ray */
  31         int sw, int sh, int sx, int sy, int stride,
  32         int rng_state_offset_x,
  33         int rng_state_offset_y,
  34         int rng_state_stride,
  35         ccl_global unsigned int *work_array,   /* Denotes work of each ray */
  36         ccl_global int *Queue_data,            /* Queues memory */
  37         ccl_global int *Queue_index,           /* Tracks the number of elements in each queue */
  38         int queuesize,                         /* Size (capacity) of each queue */
  39         int end_sample,
  40         int start_sample,
  41 #ifdef __WORK_STEALING__
  42         ccl_global unsigned int *work_pool_wgs,
  43         unsigned int num_samples,
  44 #endif
  45 #ifdef __KERNEL_DEBUG__
  46         DebugData *debugdata_coop,
  47 #endif
  48         int parallel_samples)                  /* Number of samples to be processed in parallel */
  49 {
  50         ccl_local unsigned int local_queue_atomics;
  51         if(get_local_id(0) == 0 && get_local_id(1) == 0) {
  52                 local_queue_atomics = 0;
  53         }
  54         barrier(CLK_LOCAL_MEM_FENCE);
  55
  56         int ray_index = get_global_id(1) * get_global_size(0) + get_global_id(0);
  57         if(ray_index == 0) {
  58                 /* We will empty this queue in this kernel. */
  59                 Queue_index[QUEUE_HITBG_BUFF_UPDATE_TOREGEN_RAYS] = 0;
  60         }
  61         char enqueue_flag = 0;
  62         ray_index = get_ray_index(ray_index,
  63                                   QUEUE_HITBG_BUFF_UPDATE_TOREGEN_RAYS,
  64                                   Queue_data,
  65                                   queuesize,
  66                                   1);
  67
  68 #ifdef __COMPUTE_DEVICE_GPU__
  69         /* If we are executing on a GPU device, we exit all threads that are not
  70          * required.
  71          *
  72          * If we are executing on a CPU device, then we need to keep all threads
  73          * active since we have barrier() calls later in the kernel. CPU devices,
  74          * expect all threads to execute barrier statement.
  75          */
  76         if(ray_index == QUEUE_EMPTY_SLOT) {
  77                 return;
  78         }
  79 #endif
  80
  81 #ifndef __COMPUTE_DEVICE_GPU__
  82         if(ray_index != QUEUE_EMPTY_SLOT) {
  83 #endif
  84                 enqueue_flag =
  85                         kernel_background_buffer_update((KernelGlobals *)kg,
  86                                                         per_sample_output_buffers,
  87                                                         rng_state,
  88                                                         rng_coop,
  89                                                         throughput_coop,
  90                                                         PathRadiance_coop,
  91                                                         Ray_coop,
  92                                                         PathState_coop,
  93                                                         L_transparent_coop,
  94                                                         ray_state,
  95                                                         sw, sh, sx, sy, stride,
  96                                                         rng_state_offset_x,
  97                                                         rng_state_offset_y,
  98                                                         rng_state_stride,
  99                                                         work_array,
 100                                                         end_sample,
 101                                                         start_sample,
 102 #ifdef __WORK_STEALING__
 103                                                         work_pool_wgs,
 104                                                         num_samples,
 105 #endif
 106 #ifdef __KERNEL_DEBUG__
 107                                                         debugdata_coop,
 108 #endif
 109                                                         parallel_samples,
 110                                                         ray_index);
 111 #ifndef __COMPUTE_DEVICE_GPU__
 112         }
 113 #endif
 114
 115         /* Enqueue RAY_REGENERATED rays into QUEUE_ACTIVE_AND_REGENERATED_RAYS;
 116          * These rays will be made active during next SceneIntersectkernel.
 117          */
 118         enqueue_ray_index_local(ray_index,
 119                                 QUEUE_ACTIVE_AND_REGENERATED_RAYS,
 120                                 enqueue_flag,
 121                                 queuesize,
 122                                 &local_queue_atomics,
 123                                 Queue_data,
 124                                 Queue_index);
 125 }