Cycles: simplify handling of ray differentials

* Store compact ray differentials in ShaderData and compute full differentials
  on demand. This reduces register pressure on the GPU.
* Remove BSDF differential code that was effectively doing nothing as the
  differential orientation was discarded when making it compact.

This gives a 1-5% speedup with RTX A6000 + OptiX in our benchmarks, with the
bigger speedups in simpler scenes.

Renders appear to be identical except for the Both displacement option that
does both displacement and bump.

Differential Revision: https://developer.blender.org/D15677

1 files changed, 0 insertions, 10 deletions

diff --git a/intern/cycles/kernel/closure/bsdf_principled_diffuse.h b/intern/cycles/kernel/closure/bsdf_principled_diffuse.h
index 90ef252b3b9..39cca1bd970 100644
--- a/intern/cycles/kernel/closure/bsdf_principled_diffuse.h
+++ b/intern/cycles/kernel/closure/bsdf_principled_diffuse.h
@@ -142,14 +142,10 @@ ccl_device Spectrum bsdf_principled_diffuse_eval_transmit(ccl_private const Shad
 ccl_device int bsdf_principled_diffuse_sample(ccl_private const ShaderClosure *sc,
                                               float3 Ng,
                                               float3 I,
-                                              float3 dIdx,
-                                              float3 dIdy,
                                               float randu,
                                               float randv,
                                               ccl_private Spectrum *eval,
                                               ccl_private float3 *omega_in,
-                                              ccl_private float3 *domega_in_dx,
-                                              ccl_private float3 *domega_in_dy,
                                               ccl_private float *pdf)
 {
   ccl_private const PrincipledDiffuseBsdf *bsdf = (ccl_private const PrincipledDiffuseBsdf *)sc;
@@ -160,12 +156,6 @@ ccl_device int bsdf_principled_diffuse_sample(ccl_private const ShaderClosure *s
 
   if (dot(Ng, *omega_in) > 0) {
     *eval = bsdf_principled_diffuse_compute_brdf(bsdf, N, I, *omega_in, pdf);
-
-#ifdef __RAY_DIFFERENTIALS__
-    // TODO: find a better approximation for the diffuse bounce
-    *domega_in_dx = -((2 * dot(N, dIdx)) * N - dIdx);
-    *domega_in_dy = -((2 * dot(N, dIdy)) * N - dIdy);
-#endif
   }
   else {
     *pdf = 0.0f;