Merge branch 'blender-v3.0-release'

5 files changed, 43 insertions, 13 deletions

diff --git a/intern/cycles/kernel/integrator/intersect_closest.h b/intern/cycles/kernel/integrator/intersect_closest.h
index 5522b46205b..366bfba7aca 100644
--- a/intern/cycles/kernel/integrator/intersect_closest.h
+++ b/intern/cycles/kernel/integrator/intersect_closest.h
@@ -88,7 +88,10 @@ ccl_device_forceinline bool integrator_intersect_terminate(KernelGlobals kg,
 #ifdef __SHADOW_CATCHER__
 /* Split path if a shadow catcher was hit. */
 ccl_device_forceinline void integrator_split_shadow_catcher(
-    KernelGlobals kg, IntegratorState state, ccl_private const Intersection *ccl_restrict isect)
+    KernelGlobals kg,
+    IntegratorState state,
+    ccl_private const Intersection *ccl_restrict isect,
+    ccl_global float *ccl_restrict render_buffer)
 {
   /* Test if we hit a shadow catcher object, and potentially split the path to continue tracing two
    * paths from here. */
@@ -97,6 +100,8 @@ ccl_device_forceinline void integrator_split_shadow_catcher(
     return;
   }
 
+  kernel_write_shadow_catcher_bounce_data(kg, state, render_buffer);
+
   /* Mark state as having done a shadow catcher split so that it stops contributing to
    * the shadow catcher matte pass, but keeps contributing to the combined pass. */
   INTEGRATOR_STATE_WRITE(state, path, flag) |= PATH_RAY_SHADOW_CATCHER_HIT;
@@ -191,6 +196,7 @@ ccl_device_forceinline void integrator_intersect_next_kernel(
     KernelGlobals kg,
     IntegratorState state,
     ccl_private const Intersection *ccl_restrict isect,
+    ccl_global float *ccl_restrict render_buffer,
     const bool hit)
 {
   /* Continue with volume kernel if we are inside a volume, regardless if we hit anything. */
@@ -233,7 +239,7 @@ ccl_device_forceinline void integrator_intersect_next_kernel(
 
 #ifdef __SHADOW_CATCHER__
         /* Handle shadow catcher. */
-        integrator_split_shadow_catcher(kg, state, isect);
+        integrator_split_shadow_catcher(kg, state, isect, render_buffer);
 #endif
       }
       else {
@@ -253,7 +259,10 @@ ccl_device_forceinline void integrator_intersect_next_kernel(
  * volume shading and termination testing have already been done. */
 template<uint32_t current_kernel>
 ccl_device_forceinline void integrator_intersect_next_kernel_after_volume(
-    KernelGlobals kg, IntegratorState state, ccl_private const Intersection *ccl_restrict isect)
+    KernelGlobals kg,
+    IntegratorState state,
+    ccl_private const Intersection *ccl_restrict isect,
+    ccl_global float *ccl_restrict render_buffer)
 {
   if (isect->prim != PRIM_NONE) {
     /* Hit a surface, continue with light or surface kernel. */
@@ -278,7 +287,7 @@ ccl_device_forceinline void integrator_intersect_next_kernel_after_volume(
 
 #ifdef __SHADOW_CATCHER__
       /* Handle shadow catcher. */
-      integrator_split_shadow_catcher(kg, state, isect);
+      integrator_split_shadow_catcher(kg, state, isect, render_buffer);
 #endif
       return;
     }
@@ -290,7 +299,9 @@ ccl_device_forceinline void integrator_intersect_next_kernel_after_volume(
   }
 }
 
-ccl_device void integrator_intersect_closest(KernelGlobals kg, IntegratorState state)
+ccl_device void integrator_intersect_closest(KernelGlobals kg,
+                                             IntegratorState state,
+                                             ccl_global float *ccl_restrict render_buffer)
 {
   PROFILING_INIT(kg, PROFILING_INTERSECT_CLOSEST);
 
@@ -341,7 +352,7 @@ ccl_device void integrator_intersect_closest(KernelGlobals kg, IntegratorState s
 
   /* Setup up next kernel to be executed. */
   integrator_intersect_next_kernel<DEVICE_KERNEL_INTEGRATOR_INTERSECT_CLOSEST>(
-      kg, state, &isect, hit);
+      kg, state, &isect, render_buffer, hit);
 }
 
 CCL_NAMESPACE_END
diff --git a/intern/cycles/kernel/integrator/megakernel.h b/intern/cycles/kernel/integrator/megakernel.h
index d8cc794dc7a..43313400a11 100644
--- a/intern/cycles/kernel/integrator/megakernel.h
+++ b/intern/cycles/kernel/integrator/megakernel.h
@@ -76,7 +76,7 @@ ccl_device void integrator_megakernel(KernelGlobals kg,
     if (queued_kernel) {
       switch (queued_kernel) {
         case DEVICE_KERNEL_INTEGRATOR_INTERSECT_CLOSEST:
-          integrator_intersect_closest(kg, state);
+          integrator_intersect_closest(kg, state, render_buffer);
           break;
         case DEVICE_KERNEL_INTEGRATOR_SHADE_BACKGROUND:
           integrator_shade_background(kg, state, render_buffer);
diff --git a/intern/cycles/kernel/integrator/shade_surface.h b/intern/cycles/kernel/integrator/shade_surface.h
index e6fe5a87120..3c84dcc3728 100644
--- a/intern/cycles/kernel/integrator/shade_surface.h
+++ b/intern/cycles/kernel/integrator/shade_surface.h
@@ -492,10 +492,6 @@ ccl_device bool integrate_surface(KernelGlobals kg,
     kernel_write_denoising_features_surface(kg, state, &sd, render_buffer);
 #endif
 
-#ifdef __SHADOW_CATCHER__
-    kernel_write_shadow_catcher_bounce_data(kg, state, &sd, render_buffer);
-#endif
-
     /* Direct light. */
     PROFILING_EVENT(PROFILING_SHADE_SURFACE_DIRECT_LIGHT);
     integrate_surface_direct_light(kg, state, &sd, &rng_state);
diff --git a/intern/cycles/kernel/integrator/shade_volume.h b/intern/cycles/kernel/integrator/shade_volume.h
index 86d712cdf32..b4d02b1b664 100644
--- a/intern/cycles/kernel/integrator/shade_volume.h
+++ b/intern/cycles/kernel/integrator/shade_volume.h
@@ -1024,7 +1024,7 @@ ccl_device void integrator_shade_volume(KernelGlobals kg,
   else {
     /* Continue to background, light or surface. */
     integrator_intersect_next_kernel_after_volume<DEVICE_KERNEL_INTEGRATOR_SHADE_VOLUME>(
-        kg, state, &isect);
+        kg, state, &isect, render_buffer);
     return;
   }
 #endif /* __VOLUME__ */
diff --git a/intern/cycles/kernel/integrator/shadow_catcher.h b/intern/cycles/kernel/integrator/shadow_catcher.h
index ac55678c9cb..4d72aced3fc 100644
--- a/intern/cycles/kernel/integrator/shadow_catcher.h
+++ b/intern/cycles/kernel/integrator/shadow_catcher.h
@@ -16,6 +16,7 @@
 
 #pragma once
 
+#include "kernel/film/write_passes.h"
 #include "kernel/integrator/path_state.h"
 #include "kernel/integrator/state_util.h"
 
@@ -47,7 +48,7 @@ ccl_device_inline bool kernel_shadow_catcher_is_path_split_bounce(KernelGlobals
     return false;
   }
 
-  if (path_flag & PATH_RAY_SHADOW_CATCHER_PASS) {
+  if (path_flag & PATH_RAY_SHADOW_CATCHER_HIT) {
     return false;
   }
 
@@ -88,6 +89,28 @@ ccl_device_forceinline bool kernel_shadow_catcher_is_object_pass(const uint32_t
   return path_flag & PATH_RAY_SHADOW_CATCHER_PASS;
 }
 
+/* Write shadow catcher passes on a bounce from the shadow catcher object. */
+ccl_device_forceinline void kernel_write_shadow_catcher_bounce_data(
+    KernelGlobals kg, IntegratorState state, ccl_global float *ccl_restrict render_buffer)
+{
+  kernel_assert(kernel_data.film.pass_shadow_catcher_sample_count != PASS_UNUSED);
+  kernel_assert(kernel_data.film.pass_shadow_catcher_matte != PASS_UNUSED);
+
+  const uint32_t render_pixel_index = INTEGRATOR_STATE(state, path, render_pixel_index);
+  const uint64_t render_buffer_offset = (uint64_t)render_pixel_index *
+                                        kernel_data.film.pass_stride;
+  ccl_global float *buffer = render_buffer + render_buffer_offset;
+
+  /* Count sample for the shadow catcher object. */
+  kernel_write_pass_float(buffer + kernel_data.film.pass_shadow_catcher_sample_count, 1.0f);
+
+  /* Since the split is done, the sample does not contribute to the matte, so accumulate it as
+   * transparency to the matte. */
+  const float3 throughput = INTEGRATOR_STATE(state, path, throughput);
+  kernel_write_pass_float(buffer + kernel_data.film.pass_shadow_catcher_matte + 3,
+                          average(throughput));
+}
+
 #endif /* __SHADOW_CATCHER__ */
 
 CCL_NAMESPACE_END