Classes
struct	AliasTableEntry
	Alias sampling data structure (see makeAliasTable() for details) More...

struct	SSEVector4f

struct	SSEVector4i

Functions
template<typename Scalar , typename QuantizedScalar , typename Index >
float	makeAliasTable (AliasTableEntry< QuantizedScalar, Index > tbl, Scalar pmf, Index size)
	Create the lookup table needed for Walker's alias sampling method implemented in sampleAlias(). Runs in linear time. More...

template<typename Scalar , typename QuantizedScalar , typename Index >
Index	sampleAlias (const AliasTableEntry< QuantizedScalar, Index > *tbl, Index size, Scalar sample)
	Generate a sample in constant time using the alias method. More...

template<typename Scalar , typename QuantizedScalar , typename Index >
Index	sampleAliasReuse (const AliasTableEntry< QuantizedScalar, Index > *tbl, Index size, Scalar &sample)
	Generate a sample in constant time using the alias method. More...

template<int idx3, int idx2, int idx1, int idx0>
SSEVector4f	shuffle (const SSEVector4f &low, const SSEVector4f &hi)

template<int idx3, int idx2, int idx1, int idx0>
SSEVector4f	shuffle (const SSEVector4f &a)

SSEVector4i	castAsInt (const SSEVector4f &a)
	Reinterprets `as` a `SSEVector4i`. More...

SSEVector4i	toInt (const SSEVector4f &a)
	Convert `a` to integer using truncate. More...

SSEVector4i	roundToInt (const SSEVector4f &a)
	Converts `a` to integer using round. More...

SSEVector4f	castAsFloat (const SSEVector4i &a)
	Reinterprets `a` as a `SSEVector4f`. More...

SSEVector4f	toFloat (const SSEVector4i &a)
	Convert `a` to floating point. More...

FINLINE void	transpose (SSEVector4f &row0, SSEVector4f &row1, SSEVector4f &row2, SSEVector4f &row3)
	The arguments `row0`, `row1`, `row2` and `row3` are `__m128` values whose elements form the corresponding rows of a 4-by-4 matrix. The matrix transposition is returned in arguments `row0`, `row1`, `row2` and `row3` where `row0` now holds column 0 of the original matrix, `row1` now holds column 1 of the original matrix, and so on. More...

Function Documentation

SSEVector4f mitsuba::math::castAsFloat ( const SSEVector4i & a )

inline

Reinterprets a as a SSEVector4f.

SSEVector4i mitsuba::math::castAsInt ( const SSEVector4f & a )

inline

Reinterprets as a SSEVector4i.

template<typename Scalar , typename QuantizedScalar , typename Index >

float mitsuba::math::makeAliasTable	(	AliasTableEntry< QuantizedScalar, Index > *	tbl,
		Scalar *	pmf,
		Index	size
	)

Create the lookup table needed for Walker's alias sampling method implemented in sampleAlias(). Runs in linear time.

The basic idea of this method is that one can "redistribute" the probability mass of a distribution to make it uniform. This this can be done in a way such that the probability of each entry in the "flattened" PMF consists of probability mass from at most two entries in the original PMF. That then leads to an efficient O(1) sampling algorithm with a O(n) preprocessing step to set up this special decomposition.

The downside of this method is that it generally does not preserve the nice stratification properties of QMC number sequences.

Returns: The original (un-normalized) sum of all probabilities in pmf.

SSEVector4i mitsuba::math::roundToInt ( const SSEVector4f & a )

inline

Converts a to integer using round.

template<typename Scalar , typename QuantizedScalar , typename Index >

Index mitsuba::math::sampleAlias	(	const AliasTableEntry< QuantizedScalar, Index > *	tbl,
		Index	size,
		Scalar	sample
	)

Generate a sample in constant time using the alias method.

template<typename Scalar , typename QuantizedScalar , typename Index >

Index mitsuba::math::sampleAliasReuse	(	const AliasTableEntry< QuantizedScalar, Index > *	tbl,
		Index	size,
		Scalar &	sample
	)

Generate a sample in constant time using the alias method.

This variation shifts and scales the uniform random sample so that it can be reused for another sampling operation

template<int idx3, int idx2, int idx1, int idx0>

SSEVector4f mitsuba::math::shuffle	(	const SSEVector4f &	low,
		const SSEVector4f &	hi
	)

template<int idx3, int idx2, int idx1, int idx0>

SSEVector4f mitsuba::math::shuffle ( const SSEVector4f & a )

SSEVector4f mitsuba::math::toFloat ( const SSEVector4i & a )

inline

Convert a to floating point.

SSEVector4i mitsuba::math::toInt ( const SSEVector4f & a )

inline

Convert a to integer using truncate.

FINLINE void mitsuba::math::transpose	(	SSEVector4f &	row0,
		SSEVector4f &	row1,
		SSEVector4f &	row2,
		SSEVector4f &	row3
	)

The arguments row0, row1, row2 and row3 are __m128 values whose elements form the corresponding rows of a 4-by-4 matrix. The matrix transposition is returned in arguments row0, row1, row2 and row3 where row0 now holds column 0 of the original matrix, row1 now holds column 1 of the original matrix, and so on.

Author: Intel Intrinsics Guide for AVX2

Classes

Functions

Function Documentation