[Bf-blender-cvs] [207338bb58b] master: Cycles: port curve-ray intersection from Embree for use in Cycles GPU
Brecht Van Lommel
noreply at git.blender.org
Mon Jun 22 13:28:16 CEST 2020
Commit: 207338bb58b1a44c531e6d78fad68672c6d3b2e1
Author: Brecht Van Lommel
Date: Tue Feb 18 20:54:41 2020 +0100
Branches: master
https://developer.blender.org/rB207338bb58b1a44c531e6d78fad68672c6d3b2e1
Cycles: port curve-ray intersection from Embree for use in Cycles GPU
This keeps render results compatible for combined CPU + GPU rendering.
Peformance and quality primitives is quite different than before. There
are now two options:
* Rounded Ribbon: render hair as flat ribbon with (fake) rounded normals, for
fast rendering. Hair curves are subdivided with a fixed number of user
specified subdivisions.
This gives relatively good results, especially when used with the Principled
Hair BSDF and hair viewed from a typical distance. There are artifacts when
viewed closed up, though this was also the case with all previous primitives
(but different ones).
* 3D Curve: render hair as 3D curve, for accurate results when viewing hair
close up. This automatically subdivides the curve until it is smooth.
This gives higher quality than any of the previous primitives, but does come
at a performance cost and is somewhat slower than our previous Thick curves.
The main problem here is performance. For CPU and OpenCL rendering performance
seems usually quite close or better for similar quality results.
However for CUDA and Optix, performance of 3D curve intersection is problematic,
with e.g. 1.45x longer render time in Koro (though there is no equivalent quality
and rounded ribbons seem fine for that scene). Any help or ideas to optimize this
are welcome.
Ref T73778
Depends on D8012
Maniphest Tasks: T73778
Differential Revision: https://developer.blender.org/D8013
===================================================================
M intern/cycles/blender/addon/properties.py
M intern/cycles/blender/addon/ui.py
M intern/cycles/device/device.h
M intern/cycles/device/device_optix.cpp
M intern/cycles/kernel/closure/bsdf.h
M intern/cycles/kernel/closure/bsdf_hair_principled.h
M intern/cycles/kernel/geom/geom_curve.h
M intern/cycles/kernel/geom/geom_curve_intersect.h
M intern/cycles/kernel/geom/geom_motion_curve.h
M intern/cycles/kernel/kernel_shader.h
M intern/cycles/kernel/kernels/optix/kernel_optix.cu
M intern/cycles/render/curves.cpp
M intern/cycles/render/session.cpp
===================================================================
diff --git a/intern/cycles/blender/addon/properties.py b/intern/cycles/blender/addon/properties.py
index f0f7d24002f..ceaeda6e798 100644
--- a/intern/cycles/blender/addon/properties.py
+++ b/intern/cycles/blender/addon/properties.py
@@ -73,8 +73,8 @@ enum_panorama_types = (
)
enum_curve_shape = (
- ('RIBBONS', "Ribbons", "Ignore thickness of each hair"),
- ('THICK', "Thick", "Use thickness of hair when rendering"),
+ ('RIBBONS', "Rounded Ribbons", "Render hair as flat ribbon with rounded normals, for fast rendering"),
+ ('THICK', "3D Curves", "Render hair as 3D curve, for accurate results when viewing hair close up"),
)
enum_tile_order = (
@@ -1223,7 +1223,7 @@ class CyclesCurveRenderSettings(bpy.types.PropertyGroup):
name="Shape",
description="Form of hair",
items=enum_curve_shape,
- default='THICK',
+ default='RIBBONS',
)
use_curves: BoolProperty(
name="Use Cycles Hair Rendering",
@@ -1234,7 +1234,7 @@ class CyclesCurveRenderSettings(bpy.types.PropertyGroup):
name="Subdivisions",
description="Number of subdivisions used in Cardinal curve intersection (power of 2)",
min=0, max=24,
- default=4,
+ default=2,
)
@classmethod
diff --git a/intern/cycles/blender/addon/ui.py b/intern/cycles/blender/addon/ui.py
index 0859a8a82b0..daab26d8ec0 100644
--- a/intern/cycles/blender/addon/ui.py
+++ b/intern/cycles/blender/addon/ui.py
@@ -407,7 +407,6 @@ class CYCLES_RENDER_PT_hair(CyclesButtonsPanel, Panel):
col = layout.column()
col.prop(ccscene, "shape", text="Shape")
if ccscene.shape == 'RIBBONS':
- # TODO: use for embree
col.prop(ccscene, "subdivisions", text="Curve subdivisions")
diff --git a/intern/cycles/device/device.h b/intern/cycles/device/device.h
index 67828103394..69f22aeb35c 100644
--- a/intern/cycles/device/device.h
+++ b/intern/cycles/device/device.h
@@ -132,6 +132,7 @@ class DeviceRequestedFeatures {
/* BVH/sampling kernel features. */
bool use_hair;
+ bool use_hair_thick;
bool use_object_motion;
bool use_camera_motion;
@@ -178,6 +179,7 @@ class DeviceRequestedFeatures {
max_nodes_group = 0;
nodes_features = 0;
use_hair = false;
+ use_hair_thick = false;
use_object_motion = false;
use_camera_motion = false;
use_baking = false;
@@ -200,6 +202,7 @@ class DeviceRequestedFeatures {
max_nodes_group == requested_features.max_nodes_group &&
nodes_features == requested_features.nodes_features &&
use_hair == requested_features.use_hair &&
+ use_hair_thick == requested_features.use_hair_thick &&
use_object_motion == requested_features.use_object_motion &&
use_camera_motion == requested_features.use_camera_motion &&
use_baking == requested_features.use_baking &&
diff --git a/intern/cycles/device/device_optix.cpp b/intern/cycles/device/device_optix.cpp
index bc2aeb0ae90..7aab2c96db4 100644
--- a/intern/cycles/device/device_optix.cpp
+++ b/intern/cycles/device/device_optix.cpp
@@ -428,11 +428,20 @@ class OptiXDevice : public CUDADevice {
group_descs[PG_HITS].hitgroup.entryFunctionNameAH = "__anyhit__kernel_optix_shadow_all_hit";
if (requested_features.use_hair) {
- // Add curve intersection programs
group_descs[PG_HITD].hitgroup.moduleIS = optix_module;
- group_descs[PG_HITD].hitgroup.entryFunctionNameIS = "__intersection__curve";
group_descs[PG_HITS].hitgroup.moduleIS = optix_module;
- group_descs[PG_HITS].hitgroup.entryFunctionNameIS = "__intersection__curve";
+
+ // Add curve intersection programs
+ if (requested_features.use_hair_thick) {
+ // Slower programs for thick hair since that also slows down ribbons.
+ // Ideally this should not be needed.
+ group_descs[PG_HITD].hitgroup.entryFunctionNameIS = "__intersection__curve_all";
+ group_descs[PG_HITS].hitgroup.entryFunctionNameIS = "__intersection__curve_all";
+ }
+ else {
+ group_descs[PG_HITD].hitgroup.entryFunctionNameIS = "__intersection__curve_ribbon";
+ group_descs[PG_HITS].hitgroup.entryFunctionNameIS = "__intersection__curve_ribbon";
+ }
}
if (requested_features.use_subsurface || requested_features.use_shader_raytrace) {
diff --git a/intern/cycles/kernel/closure/bsdf.h b/intern/cycles/kernel/closure/bsdf.h
index 4cc61e8ee71..6070fd983f5 100644
--- a/intern/cycles/kernel/closure/bsdf.h
+++ b/intern/cycles/kernel/closure/bsdf.h
@@ -119,13 +119,16 @@ ccl_device_inline int bsdf_sample(KernelGlobals *kg,
differential3 *domega_in,
float *pdf)
{
+ /* For curves use the smooth normal, particularly for ribbons the geometric
+ * normal gives too much darkening otherwise. */
int label;
+ const float3 Ng = (sd->type & PRIMITIVE_ALL_CURVE) ? sc->N : sd->Ng;
switch (sc->type) {
case CLOSURE_BSDF_DIFFUSE_ID:
case CLOSURE_BSDF_BSSRDF_ID:
label = bsdf_diffuse_sample(sc,
- sd->Ng,
+ Ng,
sd->I,
sd->dI.dx,
sd->dI.dy,
@@ -140,7 +143,7 @@ ccl_device_inline int bsdf_sample(KernelGlobals *kg,
#ifdef __SVM__
case CLOSURE_BSDF_OREN_NAYAR_ID:
label = bsdf_oren_nayar_sample(sc,
- sd->Ng,
+ Ng,
sd->I,
sd->dI.dx,
sd->dI.dy,
@@ -155,7 +158,7 @@ ccl_device_inline int bsdf_sample(KernelGlobals *kg,
# ifdef __OSL__
case CLOSURE_BSDF_PHONG_RAMP_ID:
label = bsdf_phong_ramp_sample(sc,
- sd->Ng,
+ Ng,
sd->I,
sd->dI.dx,
sd->dI.dy,
@@ -169,7 +172,7 @@ ccl_device_inline int bsdf_sample(KernelGlobals *kg,
break;
case CLOSURE_BSDF_DIFFUSE_RAMP_ID:
label = bsdf_diffuse_ramp_sample(sc,
- sd->Ng,
+ Ng,
sd->I,
sd->dI.dx,
sd->dI.dy,
@@ -184,7 +187,7 @@ ccl_device_inline int bsdf_sample(KernelGlobals *kg,
# endif
case CLOSURE_BSDF_TRANSLUCENT_ID:
label = bsdf_translucent_sample(sc,
- sd->Ng,
+ Ng,
sd->I,
sd->dI.dx,
sd->dI.dy,
@@ -198,7 +201,7 @@ ccl_device_inline int bsdf_sample(KernelGlobals *kg,
break;
case CLOSURE_BSDF_REFLECTION_ID:
label = bsdf_reflection_sample(sc,
- sd->Ng,
+ Ng,
sd->I,
sd->dI.dx,
sd->dI.dy,
@@ -212,7 +215,7 @@ ccl_device_inline int bsdf_sample(KernelGlobals *kg,
break;
case CLOSURE_BSDF_REFRACTION_ID:
label = bsdf_refraction_sample(sc,
- sd->Ng,
+ Ng,
sd->I,
sd->dI.dx,
sd->dI.dy,
@@ -226,7 +229,7 @@ ccl_device_inline int bsdf_sample(KernelGlobals *kg,
break;
case CLOSURE_BSDF_TRANSPARENT_ID:
label = bsdf_transparent_sample(sc,
- sd->Ng,
+ Ng,
sd->I,
sd->dI.dx,
sd->dI.dy,
@@ -244,7 +247,7 @@ ccl_device_inline int bsdf_sample(KernelGlobals *kg,
case CLOSURE_BSDF_MICROFACET_GGX_REFRACTION_ID:
label = bsdf_microfacet_ggx_sample(kg,
sc,
- sd->Ng,
+ Ng,
sd->I,
sd->dI.dx,
sd->dI.dy,
@@ -260,7 +263,7 @@ ccl_device_inline int bsdf_sample(KernelGlobals *kg,
case CLOSURE_BSDF_MICROFACET_MULTI_GGX_FRESNEL_ID:
label = bsdf_microfacet_multi_ggx_sample(kg,
sc,
- sd->Ng,
+ Ng,
sd->I,
sd->dI.dx,
sd->dI.dy,
@@ -277,7 +280,7 @@ ccl_device_inline int bsdf_sample(KernelGlobals *kg,
case CLOSURE_BSDF_MICROFACET_MULTI_GGX_GLASS_FRESNEL_ID:
label = bsdf_microfacet_multi_ggx_glass_sample(kg,
sc,
- sd->Ng,
+ Ng,
sd->I,
sd->dI.dx,
sd->dI.dy,
@@ -294,7 +297,7 @@ ccl_device_inline int bsdf_sample(KernelGlobals *kg,
case CLOSURE_BSDF_MICROFACET_BECKMANN_REFRACTION_ID:
label = bsdf_microfacet_beckmann_sample(kg,
sc,
- sd->Ng,
+ Ng,
sd->I,
sd->dI.dx,
sd->dI.dy,
@@ -308,7
@@ Diff output truncated at 10240 characters. @@
More information about the Bf-blender-cvs
mailing list