mitsuba/src/samplers/ldsampler.cpp

252 lines
8.5 KiB
C++

/*
This file is part of Mitsuba, a physically based rendering system.
Copyright (c) 2007-2012 by Wenzel Jakob and others.
Mitsuba is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License Version 3
as published by the Free Software Foundation.
Mitsuba 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, see <http://www.gnu.org/licenses/>.
*/
#include <mitsuba/render/sampler.h>
#include <mitsuba/core/qmc.h>
MTS_NAMESPACE_BEGIN
/*!\plugin{ldsampler}{Low discrepancy sampler}
* \order{3}
* \parameters{
* \parameter{sampleCount}{\Integer}{
* Number of samples per pixel; should be a power of two
* (e.g. 1, 2, 4, 8, 16, etc.), or it will be rounded up to the next one
* \default{4}
* }
* \parameter{dimension}{\Integer}{
* Effective dimension, up to which low discrepancy samples are provided. The
* number here is to be interpreted as the number of subsequent 1D or 2D sample
* requests that can be satisfied using ``good'' samples. Higher high values
* increase both storage and computational costs.
* \default{4}
* }
* }
* \vspace{-2mm}
* \renderings{
* \unframedrendering{A projection of the first 1024 points
* onto the first two dimensions.}{sampler_ldsampler_0}
* \unframedrendering{A projection of the first 1024 points
* onto the 32 and 33th dimension, which look almost identical. However,
* note that the points have been scrambled to reduce
* correlations between dimensions.}{sampler_ldsampler_32}
* }
* This plugin implements a simple hybrid sampler that combines aspects of a Quasi-Monte
* Carlo sequence with a pseudorandom number generator based on a technique proposed
* by Kollig and Keller \cite{Kollig2002Efficient}.
* It is a good and fast general-purpose sample generator and therefore chosen as
* the default option in Mitsuba. Some of the QMC samplers in the following pages can generate
* even better distributed samples, but this comes at a higher cost in terms of performance.
*
* Roughly, the idea of this sampler is that all of the individual 2D sample dimensions are
* first filled using the same (0, 2)-sequence, which is then randomly scrambled and permuted
* using numbers generated by a Mersenne Twister pseudorandom number generator \cite{Saito2008SIMD}.
* Note that due to internal storage costs, low discrepancy samples are only provided
* up to a certain dimension, after which independent sampling takes over.
* The name of this plugin stems from the fact that (0, 2) sequences minimize the so-called
* \emph{star disrepancy}, which is a quality criterion on their spatial distribution. By
* now, the name has become slightly misleading since there are other samplers in Mitsuba
* that just as much try to minimize discrepancy, namely the \pluginref{sobol} and
* \pluginref{halton} plugins.
*
* Like the \pluginref{independent} sampler, multicore and network renderings
* will generally produce different images in subsequent runs due to the nondeterminism
* introduced by the operating system scheduler.
*/
class LowDiscrepancySampler : public Sampler {
public:
LowDiscrepancySampler() : Sampler(Properties()) { }
LowDiscrepancySampler(const Properties &props) : Sampler(props) {
/* Sample count (will be rounded up to the next power of two) */
m_sampleCount = props.getSize("sampleCount", 4);
/* Dimension, up to which which low discrepancy samples are guaranteed to be available. */
m_maxDimension = props.getInteger("dimension", 4);
if (!isPowerOfTwo(m_sampleCount)) {
m_sampleCount = roundToPowerOfTwo(m_sampleCount);
Log(EWarn, "Sample count should be a power of two -- rounding to "
SIZE_T_FMT, m_sampleCount);
}
m_samples1D = new Float*[m_maxDimension];
m_samples2D = new Point2*[m_maxDimension];
for (size_t i=0; i<m_maxDimension; i++) {
m_samples1D[i] = new Float[m_sampleCount];
m_samples2D[i] = new Point2[m_sampleCount];
}
m_random = new Random();
}
LowDiscrepancySampler(Stream *stream, InstanceManager *manager)
: Sampler(stream, manager) {
m_random = static_cast<Random *>(manager->getInstance(stream));
m_maxDimension = stream->readSize();
m_samples1D = new Float*[m_maxDimension];
m_samples2D = new Point2*[m_maxDimension];
for (size_t i=0; i<m_maxDimension; i++) {
m_samples1D[i] = new Float[(size_t) m_sampleCount];
m_samples2D[i] = new Point2[(size_t) m_sampleCount];
}
}
virtual ~LowDiscrepancySampler() {
for (size_t i=0; i<m_maxDimension; i++) {
delete[] m_samples1D[i];
delete[] m_samples2D[i];
}
delete[] m_samples1D;
delete[] m_samples2D;
}
void serialize(Stream *stream, InstanceManager *manager) const {
Sampler::serialize(stream, manager);
manager->serialize(stream, m_random.get());
stream->writeSize(m_maxDimension);
}
ref<Sampler> clone() {
ref<LowDiscrepancySampler> sampler = new LowDiscrepancySampler();
sampler->m_sampleCount = m_sampleCount;
sampler->m_maxDimension = m_maxDimension;
sampler->m_random = new Random(m_random);
sampler->m_samples1D = new Float*[m_maxDimension];
sampler->m_samples2D = new Point2*[m_maxDimension];
for (size_t i=0; i<m_maxDimension; i++) {
sampler->m_samples1D[i] = new Float[m_sampleCount];
sampler->m_samples2D[i] = new Point2[m_sampleCount];
}
for (size_t i=0; i<m_req1D.size(); ++i)
sampler->request2DArray(m_req1D[i]);
for (size_t i=0; i<m_req2D.size(); ++i)
sampler->request2DArray(m_req2D[i]);
return sampler.get();
}
inline void generate1D(Float *samples, size_t sampleCount) {
#if defined(SINGLE_PRECISION)
uint32_t scramble = m_random->nextULong() & 0xFFFFFFFF;
for (size_t i = 0; i < sampleCount; ++i)
samples[i] = radicalInverse2Single((uint32_t) i, scramble);
#else
uint64_t scramble = m_random->nextULong();
for (size_t i = 0; i < sampleCount; ++i)
samples[i] = radicalInverse2Double(i, scramble);
#endif
m_random->shuffle(samples, samples + sampleCount);
}
inline void generate2D(Point2 *samples, size_t sampleCount) {
#if defined(SINGLE_PRECISION)
union {
uint64_t qword;
uint32_t dword[2];
} scramble;
scramble.qword = m_random->nextULong();
for (size_t i = 0; i < sampleCount; ++i)
samples[i] = sample02Single((uint32_t) i, scramble.dword);
#else
uint64_t scramble[2];
scramble[0] = m_random->nextULong();
scramble[1] = m_random->nextULong();
for (size_t i = 0; i < sampleCount; ++i)
samples[i] = sample02Double(i, scramble);
#endif
m_random->shuffle(samples, samples + sampleCount);
}
void generate(const Point2i &) {
for (size_t i=0; i<m_maxDimension; ++i) {
generate1D(m_samples1D[i], m_sampleCount);
generate2D(m_samples2D[i], m_sampleCount);
}
for (size_t i=0; i<m_req1D.size(); i++)
generate1D(m_sampleArrays1D[i], m_sampleCount * m_req1D[i]);
for (size_t i=0; i<m_req2D.size(); i++)
generate2D(m_sampleArrays2D[i], m_sampleCount * m_req2D[i]);
m_sampleIndex = 0;
m_dimension1D = m_dimension2D = 0;
m_dimension1DArray = m_dimension2DArray = 0;
}
void advance() {
m_sampleIndex++;
m_dimension1D = m_dimension2D = 0;
m_dimension1DArray = m_dimension2DArray = 0;
}
void setSampleIndex(size_t sampleIndex) {
m_sampleIndex = sampleIndex;
m_dimension1D = m_dimension2D = 0;
m_dimension1DArray = m_dimension2DArray = 0;
}
Float next1D() {
Assert(m_sampleIndex < m_sampleCount);
if (m_dimension1D < m_maxDimension)
return m_samples1D[m_dimension1D++][m_sampleIndex];
else
return m_random->nextFloat();
}
Point2 next2D() {
Assert(m_sampleIndex < m_sampleCount);
if (m_dimension2D < m_maxDimension)
return m_samples2D[m_dimension2D++][m_sampleIndex];
else
return Point2(m_random->nextFloat(), m_random->nextFloat());
}
std::string toString() const {
std::ostringstream oss;
oss << "LowDiscrepancySampler[" << endl
<< " sampleCount = " << m_sampleCount << "," << endl
<< " dimension = " << m_maxDimension << endl
<< "]";
return oss.str();
}
MTS_DECLARE_CLASS()
private:
ref<Random> m_random;
size_t m_maxDimension;
size_t m_dimension1D;
size_t m_dimension2D;
Float **m_samples1D;
Point2 **m_samples2D;
};
MTS_IMPLEMENT_CLASS_S(LowDiscrepancySampler, false, Sampler)
MTS_EXPORT_PLUGIN(LowDiscrepancySampler, "Low discrepancy sampler");
MTS_NAMESPACE_END