[Bf-blender-cvs] SVN commit: /data/svn/bf-blender [57282] trunk/blender/intern/cycles: Cycles: experimental correlated multi-jittered sampling pattern that can be used
Brecht Van Lommel
brechtvanlommel at pandora.be
Fri Jun 7 18:06:22 CEST 2013
Revision: 57282
http://projects.blender.org/scm/viewvc.php?view=rev&root=bf-blender&revision=57282
Author: blendix
Date: 2013-06-07 16:06:22 +0000 (Fri, 07 Jun 2013)
Log Message:
-----------
Cycles: experimental correlated multi-jittered sampling pattern that can be used
instead of sobol. So far one doesn't seem to be consistently better or worse than
the other for the same number of samples but more testing is needed.
The random number generator itself is slower than sobol for most number of samples,
except 16, 64, 256, .. because they can be computed faster. This can probably be
optimized, but we can do that when/if this actually turns out to be useful.
Paper this implementation is based on:
http://graphics.pixar.com/library/MultiJitteredSampling/
Also includes some refactoring of RNG code, fixing a Sobol correlation issue with
the first BSDF and < 16 samples, skipping some unneeded RNG calls and using a
simpler unit square to unit disk function.
Modified Paths:
--------------
trunk/blender/intern/cycles/blender/addon/properties.py
trunk/blender/intern/cycles/blender/addon/ui.py
trunk/blender/intern/cycles/blender/blender_sync.cpp
trunk/blender/intern/cycles/kernel/CMakeLists.txt
trunk/blender/intern/cycles/kernel/kernel_montecarlo.h
trunk/blender/intern/cycles/kernel/kernel_path.h
trunk/blender/intern/cycles/kernel/kernel_random.h
trunk/blender/intern/cycles/kernel/kernel_types.h
trunk/blender/intern/cycles/render/integrator.cpp
trunk/blender/intern/cycles/render/integrator.h
trunk/blender/intern/cycles/render/session.cpp
Added Paths:
-----------
trunk/blender/intern/cycles/kernel/kernel_jitter.h
Modified: trunk/blender/intern/cycles/blender/addon/properties.py
===================================================================
--- trunk/blender/intern/cycles/blender/addon/properties.py 2013-06-07 16:06:17 UTC (rev 57281)
+++ trunk/blender/intern/cycles/blender/addon/properties.py 2013-06-07 16:06:22 UTC (rev 57282)
@@ -114,7 +114,12 @@
('IGNORE', "Ignore", "Ignore per render layer number of samples"),
)
+enum_sampling_pattern = (
+ ('SOBOL', "Sobol", "Use Sobol random sampling pattern"),
+ ('CORRELATED_MUTI_JITTER', "Correlated Multi-Jitter", "Use Correlated Multi-Jitter random sampling pattern"),
+ )
+
class CyclesRenderSettings(bpy.types.PropertyGroup):
@classmethod
def register(cls):
@@ -219,6 +224,13 @@
default=1,
)
+ cls.sampling_pattern = EnumProperty(
+ name="Sampling Pattern",
+ description="Random sampling pattern used by the integrator",
+ items=enum_sampling_pattern,
+ default='SOBOL',
+ )
+
cls.use_layer_samples = EnumProperty(
name="Layer Samples",
description="How to use per render layer sample settings",
Modified: trunk/blender/intern/cycles/blender/addon/ui.py
===================================================================
--- trunk/blender/intern/cycles/blender/addon/ui.py 2013-06-07 16:06:17 UTC (rev 57281)
+++ trunk/blender/intern/cycles/blender/addon/ui.py 2013-06-07 16:06:22 UTC (rev 57282)
@@ -85,6 +85,9 @@
sub.prop(cscene, "mesh_light_samples", text="Mesh Light")
sub.prop(cscene, "subsurface_samples", text="Subsurface")
+ if cscene.feature_set == 'EXPERIMENTAL':
+ layout.row().prop(cscene, "sampling_pattern", text="Pattern")
+
for rl in scene.render.layers:
if rl.samples > 0:
layout.separator()
Modified: trunk/blender/intern/cycles/blender/blender_sync.cpp
===================================================================
--- trunk/blender/intern/cycles/blender/blender_sync.cpp 2013-06-07 16:06:17 UTC (rev 57281)
+++ trunk/blender/intern/cycles/blender/blender_sync.cpp 2013-06-07 16:06:22 UTC (rev 57282)
@@ -199,6 +199,9 @@
integrator->subsurface_samples = get_int(cscene, "subsurface_samples");
integrator->progressive = get_boolean(cscene, "progressive");
+ if(experimental)
+ integrator->sampling_pattern = (SamplingPattern)RNA_enum_get(&cscene, "sampling_pattern");
+
if(integrator->modified(previntegrator))
integrator->tag_update(scene);
}
Modified: trunk/blender/intern/cycles/kernel/CMakeLists.txt
===================================================================
--- trunk/blender/intern/cycles/kernel/CMakeLists.txt 2013-06-07 16:06:17 UTC (rev 57281)
+++ trunk/blender/intern/cycles/kernel/CMakeLists.txt 2013-06-07 16:06:22 UTC (rev 57282)
@@ -33,6 +33,7 @@
kernel_emission.h
kernel_film.h
kernel_globals.h
+ kernel_jitter.h
kernel_light.h
kernel_math.h
kernel_montecarlo.h
Added: trunk/blender/intern/cycles/kernel/kernel_jitter.h
===================================================================
--- trunk/blender/intern/cycles/kernel/kernel_jitter.h (rev 0)
+++ trunk/blender/intern/cycles/kernel/kernel_jitter.h 2013-06-07 16:06:22 UTC (rev 57282)
@@ -0,0 +1,181 @@
+/*
+ * Copyright 2013, Blender Foundation.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
+ */
+
+CCL_NAMESPACE_BEGIN
+
+/* "Correlated Multi-Jittered Sampling"
+ * Andrew Kensler, Pixar Technical Memo 13-01, 2013 */
+
+/* todo: find good value, suggested 64 gives pattern on cornell box ceiling */
+#define CMJ_RANDOM_OFFSET_LIMIT 4096
+
+__device_inline bool cmj_is_pow2(int i)
+{
+ return (i & (i - 1)) == 0;
+}
+
+__device_inline int cmj_fast_mod_pow2(int a, int b)
+{
+ return (a & (b - 1));
+}
+
+/* a must be > 0 and b must be > 1 */
+__device_inline int cmj_fast_div_pow2(int a, int b)
+{
+#ifdef __KERNEL_SSE2__
+ return a >> __builtin_ctz(b);
+#else
+ return a/b;
+#endif
+}
+
+__device_inline uint cmj_w_mask(uint w)
+{
+#ifdef __KERNEL_SSE2__
+ return ((1 << (32 - __builtin_clz(w))) - 1);
+#else
+ w |= w >> 1;
+ w |= w >> 2;
+ w |= w >> 4;
+ w |= w >> 8;
+ w |= w >> 16;
+
+ return w;
+#endif
+}
+
+__device_inline uint cmj_permute(uint i, uint l, uint p)
+{
+ uint w = l - 1;
+
+ if((l & w) == 0) {
+ /* l is a power of two (fast) */
+ i ^= p;
+ i *= 0xe170893d;
+ i ^= p >> 16;
+ i ^= (i & w) >> 4;
+ i ^= p >> 8;
+ i *= 0x0929eb3f;
+ i ^= p >> 23;
+ i ^= (i & w) >> 1;
+ i *= 1 | p >> 27;
+ i *= 0x6935fa69;
+ i ^= (i & w) >> 11;
+ i *= 0x74dcb303;
+ i ^= (i & w) >> 2;
+ i *= 0x9e501cc3;
+ i ^= (i & w) >> 2;
+ i *= 0xc860a3df;
+ i &= w;
+ i ^= i >> 5;
+
+ return (i + p) & w;
+ }
+ else {
+ /* l is not a power of two (slow) */
+ w = cmj_w_mask(w);
+
+ do {
+ i ^= p;
+ i *= 0xe170893d;
+ i ^= p >> 16;
+ i ^= (i & w) >> 4;
+ i ^= p >> 8;
+ i *= 0x0929eb3f;
+ i ^= p >> 23;
+ i ^= (i & w) >> 1;
+ i *= 1 | p >> 27;
+ i *= 0x6935fa69;
+ i ^= (i & w) >> 11;
+ i *= 0x74dcb303;
+ i ^= (i & w) >> 2;
+ i *= 0x9e501cc3;
+ i ^= (i & w) >> 2;
+ i *= 0xc860a3df;
+ i &= w;
+ i ^= i >> 5;
+ } while (i >= l);
+
+ return (i + p) % l;
+ }
+}
+
+__device_inline uint cmj_hash(uint i, uint p)
+{
+ i ^= p;
+ i ^= i >> 17;
+ i ^= i >> 10;
+ i *= 0xb36534e5;
+ i ^= i >> 12;
+ i ^= i >> 21;
+ i *= 0x93fc4795;
+ i ^= 0xdf6e307f;
+ i ^= i >> 17;
+ i *= 1 | p >> 18;
+
+ return i;
+}
+
+__device_inline float cmj_randfloat(uint i, uint p)
+{
+ return cmj_hash(i, p) * (1.0f / 4294967808.0f);
+}
+
+#ifdef __CMJ__
+__device_noinline float cmj_sample_1D(int s, int N, int p)
+{
+ uint x = cmj_permute(s, N, p * 0x68bc21eb);
+ float jx = cmj_randfloat(s, p * 0x967a889b);
+
+ float invN = 1.0f/N;
+ return (x + jx)*invN;
+}
+
+__device_noinline float2 cmj_sample_2D(int s, int N, int p)
+{
+ int m = float_to_int(sqrtf(N));
+ int n = (N + m - 1)/m;
+ float invN = 1.0f/N;
+ float invm = 1.0f/m;
+ float invn = 1.0f/n;
+
+ s = cmj_permute(s, N, p * 0x51633e2d);
+
+ int sdivm, smodm;
+
+ if(cmj_is_pow2(m)) {
+ sdivm = cmj_fast_div_pow2(s, m);
+ smodm = cmj_fast_mod_pow2(s, m);
+ }
+ else {
+ sdivm = float_to_int(s * invm);
+ smodm = s - sdivm*m;
+ }
+
+ uint sx = cmj_permute(smodm, m, p * 0x68bc21eb);
+ uint sy = cmj_permute(sdivm, n, p * 0x02e5be93);
+
+ float jx = cmj_randfloat(s, p * 0x967a889b);
+ float jy = cmj_randfloat(s, p * 0x368cc8b7);
+
+ return make_float2((sx + (sy + jx)*invn)*invm, (s + jy)*invN);
+}
+#endif
+
+CCL_NAMESPACE_END
+
Modified: trunk/blender/intern/cycles/kernel/kernel_montecarlo.h
===================================================================
--- trunk/blender/intern/cycles/kernel/kernel_montecarlo.h 2013-06-07 16:06:17 UTC (rev 57281)
+++ trunk/blender/intern/cycles/kernel/kernel_montecarlo.h 2013-06-07 16:06:22 UTC (rev 57282)
@@ -36,36 +36,12 @@
CCL_NAMESPACE_BEGIN
/// Given values x and y on [0,1], convert them in place to values on
-/// [-1,1] uniformly distributed over a unit sphere. This code is
-/// derived from Peter Shirley, "Realistic Ray Tracing", p. 103.
+/// [-1,1] uniformly distributed over a unit sphere.
__device void to_unit_disk(float *x, float *y)
{
- float r, phi;
- float a = 2.0f * (*x) - 1.0f;
- float b = 2.0f * (*y) - 1.0f;
- if(a > -b) {
- if(a > b) {
- r = a;
- phi = M_PI_4_F *(b/a);
- }
- else {
- r = b;
- phi = M_PI_4_F *(2.0f - a/b);
- }
- }
- else {
- if(a < b) {
- r = -a;
- phi = M_PI_4_F *(4.0f + b/a);
- }
- else {
- r = -b;
- if(b != 0.0f)
- phi = M_PI_4_F *(6.0f - a/b);
- else
- phi = 0.0f;
- }
- }
+ float phi = 2.0f * M_PI_F * (*x);
+ float r = sqrtf(*y);
+
*x = r * cosf(phi);
*y = r * sinf(phi);
}
Modified: trunk/blender/intern/cycles/kernel/kernel_path.h
===================================================================
--- trunk/blender/intern/cycles/kernel/kernel_path.h 2013-06-07 16:06:17 UTC (rev 57281)
+++ trunk/blender/intern/cycles/kernel/kernel_path.h 2013-06-07 16:06:22 UTC (rev 57282)
@@ -233,7 +233,7 @@
return result;
}
-__device float4 kernel_path_progressive(KernelGlobals *kg, RNG *rng, int sample, Ray ray, __global float *buffer)
+__device float4 kernel_path_progressive(KernelGlobals *kg, RNG rng, int sample, Ray ray, __global float *buffer)
{
/* initialize */
PathRadiance L;
@@ -249,6 +249,7 @@
#endif
PathState state;
int rng_offset = PRNG_BASE_NUM;
+ int num_samples = kernel_data.integrator.aa_samples;
path_state_init(&state);
@@ -270,7 +271,7 @@
}
extmax = kernel_data.curve_kernel_data.maximum_width;
- lcg_state = lcg_init(*rng + rng_offset + sample*0x51633e2d);
+ lcg_state = lcg_init(rng + rng_offset + sample*0x51633e2d);
}
bool hit = scene_intersect(kg, &ray, visibility, &isect, &lcg_state, difl, extmax);
@@ -292,7 +293,7 @@
light_ray.dP = ray.dP;
/* intersect with lamp */
- float light_t = path_rng(kg, rng, sample, rng_offset + PRNG_LIGHT);
+ float light_t = path_rng_1D(kg, rng, sample, num_samples, rng_offset + PRNG_LIGHT);
float3 emission;
if(indirect_lamp_emission(kg, &light_ray, state.flag, ray_pdf, light_t, &emission))
@@ -323,7 +324,7 @@
/* setup shading */
ShaderData sd;
shader_setup_from_ray(kg, &sd, &isect, &ray);
- float rbsdf = path_rng(kg, rng, sample, rng_offset + PRNG_BSDF);
+ float rbsdf = path_rng_1D(kg, rng, sample, num_samples, rng_offset + PRNG_BSDF);
shader_eval_surface(kg, &sd, rbsdf, state.flag, SHADER_CONTEXT_MAIN);
/* holdout */
@@ -373,13 +374,19 @@
* mainly due to the mixed in MIS that we use. gives too many unneeded
* shader evaluations, only need emission if we are going to terminate */
float probability = path_state_terminate_probability(kg, &state, throughput);
@@ Diff output truncated at 10240 characters. @@
More information about the Bf-blender-cvs
mailing list