//----------------------------------------------------------------------------------------------- // SmoothNoise.cpp // #include "./RawNoise.hpp" #include #include #include #include #include ///////////////////////////////////////////////////////////////////////////////////////////////// // For all fractal (and Perlin) noise functions, the following internal naming conventions // are used, primarily to help me visualize 3D and 4D constructs clearly. They need not // have any actual bearing on / relationship to actual external coordinate systems. // // 1D noise: only X (+east / -west) // 2D noise: also Y (+north / -south) // 3D noise: also Z (+above / -below) // 4D noise: also T (+after / -before) ///////////////////////////////////////////////////////////////////////////////////////////////// namespace SquirrelNoise4 { namespace Easing { static float SmoothStep( float t ) { //return 2*(t*t*t) - 3*(t*t); //return t*t * (2*t-3); // optimised //return 3*(t*t) - 2*(t*t*t); return t*t * (3-2*t); // optimised } } // Easing //----------------------------------------------------------------------------------------------- float Compute1dFractalNoise( float position, float scale, uint32_t numOctaves, float octavePersistence, float octaveScale, bool renormalize, uint32_t seed ) { const float OCTAVE_OFFSET = 0.636764989593174f; // Translation/bias to add to each octave float totalNoise = 0.f; float totalAmplitude = 0.f; float currentAmplitude = 1.f; float currentPosition = position * (1.f / scale); for( uint32_t octaveNum = 0; octaveNum < numOctaves; ++ octaveNum ) { // Determine noise values at nearby integer "grid point" positions float positionFloor = ::glm::floor( currentPosition ); int32_t indexWest = (int32_t) positionFloor; int32_t indexEast = indexWest + 1; float valueWest = Get1dNoiseZeroToOne( indexWest, seed ); float valueEast = Get1dNoiseZeroToOne( indexEast, seed ); // Do a smoothed (nonlinear) weighted average of nearby grid point values float distanceFromWest = currentPosition - positionFloor; float weightEast = Easing::SmoothStep( distanceFromWest ); // Gives rounder (nonlinear) results float weightWest = 1.f - weightEast; float noiseZeroToOne = (valueWest * weightWest) + (valueEast * weightEast); float noiseThisOctave = 2.f * (noiseZeroToOne - 0.5f); // Map from [0,1] to [-1,1] // Accumulate results and prepare for next octave (if any) totalNoise += noiseThisOctave * currentAmplitude; totalAmplitude += currentAmplitude; currentAmplitude *= octavePersistence; currentPosition *= octaveScale; currentPosition += OCTAVE_OFFSET; // Add "irrational" offset to de-align octave grids ++ seed; // Eliminates octaves "echoing" each other (since each octave is uniquely seeded) } // Re-normalize total noise to within [-1,1] and fix octaves pulling us far away from limits if( renormalize && totalAmplitude > 0.f ) { totalNoise /= totalAmplitude; // Amplitude exceeds 1.0 if octaves are used! totalNoise = (totalNoise * 0.5f) + 0.5f; // Map to [0,1] totalNoise = Easing::SmoothStep( totalNoise ); // Push towards extents (octaves pull us away) totalNoise = (totalNoise * 2.0f) - 1.f; // Map back to [-1,1] } return totalNoise; } //----------------------------------------------------------------------------------------------- float Compute2dFractalNoise( float posX, float posY, float scale, uint32_t numOctaves, float octavePersistence, float octaveScale, bool renormalize, uint32_t seed ) { const float OCTAVE_OFFSET = 0.636764989593174f; // Translation/bias to add to each octave float totalNoise = 0.f; float totalAmplitude = 0.f; float currentAmplitude = 1.f; float invScale = (1.f / scale); ::glm::vec2 currentPos( posX * invScale, posY * invScale ); for( uint32_t octaveNum = 0; octaveNum < numOctaves; ++ octaveNum ) { // Determine noise values at nearby integer "grid point" positions ::glm::vec2 cellMins( ::glm::floor( currentPos.x ), ::glm::floor( currentPos.y ) ); int32_t indexWestX = (int32_t) cellMins.x; int32_t indexSouthY = (int32_t) cellMins.y; int32_t indexEastX = indexWestX + 1; int32_t indexNorthY = indexSouthY + 1; float valueSouthWest = Get2dNoiseZeroToOne( indexWestX, indexSouthY, seed ); float valueSouthEast = Get2dNoiseZeroToOne( indexEastX, indexSouthY, seed ); float valueNorthWest = Get2dNoiseZeroToOne( indexWestX, indexNorthY, seed ); float valueNorthEast = Get2dNoiseZeroToOne( indexEastX, indexNorthY, seed ); // Do a smoothed (nonlinear) weighted average of nearby grid point values ::glm::vec2 displacementFromMins = currentPos - cellMins; float weightEast = Easing::SmoothStep( displacementFromMins.x ); float weightNorth = Easing::SmoothStep( displacementFromMins.y ); float weightWest = 1.f - weightEast; float weightSouth = 1.f - weightNorth; float blendSouth = (weightEast * valueSouthEast) + (weightWest * valueSouthWest); float blendNorth = (weightEast * valueNorthEast) + (weightWest * valueNorthWest); float blendTotal = (weightSouth * blendSouth) + (weightNorth * blendNorth); float noiseThisOctave = 2.f * (blendTotal - 0.5f); // Map from [0,1] to [-1,1] // Accumulate results and prepare for next octave (if any) totalNoise += noiseThisOctave * currentAmplitude; totalAmplitude += currentAmplitude; currentAmplitude *= octavePersistence; currentPos *= octaveScale; currentPos.x += OCTAVE_OFFSET; // Add "irrational" offsets to noise position components currentPos.y += OCTAVE_OFFSET; // at each octave to break up their grid alignment ++ seed; // Eliminates octaves "echoing" each other (since each octave is uniquely seeded) } // Re-normalize total noise to within [-1,1] and fix octaves pulling us far away from limits if( renormalize && totalAmplitude > 0.f ) { totalNoise /= totalAmplitude; // Amplitude exceeds 1.0 if octaves are used totalNoise = (totalNoise * 0.5f) + 0.5f; // Map to [0,1] totalNoise = Easing::SmoothStep( totalNoise ); // Push towards extents (octaves pull us away) totalNoise = (totalNoise * 2.0f) - 1.f; // Map back to [-1,1] } return totalNoise; } //----------------------------------------------------------------------------------------------- float Compute3dFractalNoise( float posX, float posY, float posZ, float scale, uint32_t numOctaves, float octavePersistence, float octaveScale, bool renormalize, uint32_t seed ) { const float OCTAVE_OFFSET = 0.636764989593174f; // Translation/bias to add to each octave float totalNoise = 0.f; float totalAmplitude = 0.f; float currentAmplitude = 1.f; float invScale = (1.f / scale); ::glm::vec3 currentPos( posX * invScale, posY * invScale, posZ * invScale ); for( uint32_t octaveNum = 0; octaveNum < numOctaves; ++ octaveNum ) { // Determine noise values at nearby integer "grid point" positions ::glm::vec3 cellMins( ::glm::floor( currentPos.x ), ::glm::floor( currentPos.y ), ::glm::floor( currentPos.z ) ); int32_t indexWestX = (int32_t) cellMins.x; int32_t indexSouthY = (int32_t) cellMins.y; int32_t indexBelowZ = (int32_t) cellMins.z; int32_t indexEastX = indexWestX + 1; int32_t indexNorthY = indexSouthY + 1; int32_t indexAboveZ = indexBelowZ + 1; // Noise grid cell has 8 corners in 3D float aboveSouthWest = Get3dNoiseZeroToOne( indexWestX, indexSouthY, indexAboveZ, seed ); float aboveSouthEast = Get3dNoiseZeroToOne( indexEastX, indexSouthY, indexAboveZ, seed ); float aboveNorthWest = Get3dNoiseZeroToOne( indexWestX, indexNorthY, indexAboveZ, seed ); float aboveNorthEast = Get3dNoiseZeroToOne( indexEastX, indexNorthY, indexAboveZ, seed ); float belowSouthWest = Get3dNoiseZeroToOne( indexWestX, indexSouthY, indexBelowZ, seed ); float belowSouthEast = Get3dNoiseZeroToOne( indexEastX, indexSouthY, indexBelowZ, seed ); float belowNorthWest = Get3dNoiseZeroToOne( indexWestX, indexNorthY, indexBelowZ, seed ); float belowNorthEast = Get3dNoiseZeroToOne( indexEastX, indexNorthY, indexBelowZ, seed ); // Do a smoothed (nonlinear) weighted average of nearby grid point values ::glm::vec3 displacementFromMins = currentPos - cellMins; float weightEast = Easing::SmoothStep( displacementFromMins.x ); float weightNorth = Easing::SmoothStep( displacementFromMins.y ); float weightAbove = Easing::SmoothStep( displacementFromMins.z ); float weightWest = 1.f - weightEast; float weightSouth = 1.f - weightNorth; float weightBelow = 1.f - weightAbove; // 8-way blend (8 -> 4 -> 2 -> 1) float blendBelowSouth = (weightEast * belowSouthEast) + (weightWest * belowSouthWest); float blendBelowNorth = (weightEast * belowNorthEast) + (weightWest * belowNorthWest); float blendAboveSouth = (weightEast * aboveSouthEast) + (weightWest * aboveSouthWest); float blendAboveNorth = (weightEast * aboveNorthEast) + (weightWest * aboveNorthWest); float blendBelow = (weightSouth * blendBelowSouth) + (weightNorth * blendBelowNorth); float blendAbove = (weightSouth * blendAboveSouth) + (weightNorth * blendAboveNorth); float blendTotal = (weightBelow * blendBelow) + (weightAbove * blendAbove); float noiseThisOctave = 2.f * (blendTotal - 0.5f); // Map from [0,1] to [-1,1] // Accumulate results and prepare for next octave (if any) totalNoise += noiseThisOctave * currentAmplitude; totalAmplitude += currentAmplitude; currentAmplitude *= octavePersistence; currentPos *= octaveScale; currentPos.x += OCTAVE_OFFSET; // Add "irrational" offsets to noise position components currentPos.y += OCTAVE_OFFSET; // at each octave to break up their grid alignment currentPos.z += OCTAVE_OFFSET; ++ seed; // Eliminates octaves "echoing" each other (since each octave is uniquely seeded) } // Re-normalize total noise to within [-1,1] and fix octaves pulling us far away from limits if( renormalize && totalAmplitude > 0.f ) { totalNoise /= totalAmplitude; // Amplitude exceeds 1.0 if octaves are used totalNoise = (totalNoise * 0.5f) + 0.5f; // Map to [0,1] totalNoise = Easing::SmoothStep( totalNoise ); // Push towards extents (octaves pull us away) totalNoise = (totalNoise * 2.0f) - 1.f; // Map back to [-1,1] } return totalNoise; } //----------------------------------------------------------------------------------------------- float Compute4dFractalNoise( float posX, float posY, float posZ, float posT, float scale, uint32_t numOctaves, float octavePersistence, float octaveScale, bool renormalize, uint32_t seed ) { const float OCTAVE_OFFSET = 0.636764989593174f; // Translation/bias to add to each octave float totalNoise = 0.f; float totalAmplitude = 0.f; float currentAmplitude = 1.f; float invScale = (1.f / scale); ::glm::vec4 currentPos( posX * invScale, posY * invScale, posZ * invScale, posT * invScale ); for( uint32_t octaveNum = 0; octaveNum < numOctaves; ++ octaveNum ) { // Determine noise values at nearby integer "grid point" positions ::glm::vec4 cellMins( ::glm::floor( currentPos.x ), ::glm::floor( currentPos.y ), ::glm::floor( currentPos.z ), ::glm::floor( currentPos.w ) ); int32_t indexWestX = (int32_t) cellMins.x; int32_t indexSouthY = (int32_t) cellMins.y; int32_t indexBelowZ = (int32_t) cellMins.z; int32_t indexBeforeT = (int32_t) cellMins.w; int32_t indexEastX = indexWestX + 1; int32_t indexNorthY = indexSouthY + 1; int32_t indexAboveZ = indexBelowZ + 1; int32_t indexAfterT = indexBeforeT + 1; // Noise grid cell has 16 "corners" in 4D float beforeBelowSW = Get4dNoiseZeroToOne( indexWestX, indexSouthY, indexBelowZ, indexBeforeT, seed ); float beforeBelowSE = Get4dNoiseZeroToOne( indexEastX, indexSouthY, indexBelowZ, indexBeforeT, seed ); float beforeBelowNW = Get4dNoiseZeroToOne( indexWestX, indexNorthY, indexBelowZ, indexBeforeT, seed ); float beforeBelowNE = Get4dNoiseZeroToOne( indexEastX, indexNorthY, indexBelowZ, indexBeforeT, seed ); float beforeAboveSW = Get4dNoiseZeroToOne( indexWestX, indexSouthY, indexAboveZ, indexBeforeT, seed ); float beforeAboveSE = Get4dNoiseZeroToOne( indexEastX, indexSouthY, indexAboveZ, indexBeforeT, seed ); float beforeAboveNW = Get4dNoiseZeroToOne( indexWestX, indexNorthY, indexAboveZ, indexBeforeT, seed ); float beforeAboveNE = Get4dNoiseZeroToOne( indexEastX, indexNorthY, indexAboveZ, indexBeforeT, seed ); float afterBelowSW = Get4dNoiseZeroToOne( indexWestX, indexSouthY, indexBelowZ, indexAfterT, seed ); float afterBelowSE = Get4dNoiseZeroToOne( indexEastX, indexSouthY, indexBelowZ, indexAfterT, seed ); float afterBelowNW = Get4dNoiseZeroToOne( indexWestX, indexNorthY, indexBelowZ, indexAfterT, seed ); float afterBelowNE = Get4dNoiseZeroToOne( indexEastX, indexNorthY, indexBelowZ, indexAfterT, seed ); float afterAboveSW = Get4dNoiseZeroToOne( indexWestX, indexSouthY, indexAboveZ, indexAfterT, seed ); float afterAboveSE = Get4dNoiseZeroToOne( indexEastX, indexSouthY, indexAboveZ, indexAfterT, seed ); float afterAboveNW = Get4dNoiseZeroToOne( indexWestX, indexNorthY, indexAboveZ, indexAfterT, seed ); float afterAboveNE = Get4dNoiseZeroToOne( indexEastX, indexNorthY, indexAboveZ, indexAfterT, seed ); // Do a smoothed (nonlinear) weighted average of nearby grid point values ::glm::vec4 displacementFromMins = currentPos - cellMins; float weightEast = Easing::SmoothStep( displacementFromMins.x ); float weightNorth = Easing::SmoothStep( displacementFromMins.y ); float weightAbove = Easing::SmoothStep( displacementFromMins.z ); float weightAfter = Easing::SmoothStep( displacementFromMins.w ); float weightWest = 1.f - weightEast; float weightSouth = 1.f - weightNorth; float weightBelow = 1.f - weightAbove; float weightBefore = 1.f - weightAfter; // 16-way blend (16 -> 8 -> 4 -> 2 -> 1) float blendBeforeBelowSouth = (weightEast * beforeBelowSE) + (weightWest * beforeBelowSW); float blendBeforeBelowNorth = (weightEast * beforeBelowNE) + (weightWest * beforeBelowNW); float blendBeforeAboveSouth = (weightEast * beforeAboveSE) + (weightWest * beforeAboveSW); float blendBeforeAboveNorth = (weightEast * beforeAboveNE) + (weightWest * beforeAboveNW); float blendAfterBelowSouth = (weightEast * afterBelowSE) + (weightWest * afterBelowSW); float blendAfterBelowNorth = (weightEast * afterBelowNE) + (weightWest * afterBelowNW); float blendAfterAboveSouth = (weightEast * afterAboveSE) + (weightWest * afterAboveSW); float blendAfterAboveNorth = (weightEast * afterAboveNE) + (weightWest * afterAboveNW); float blendBeforeBelow = (weightSouth * blendBeforeBelowSouth) + (weightNorth * blendBeforeBelowNorth); float blendBeforeAbove = (weightSouth * blendBeforeAboveSouth) + (weightNorth * blendBeforeAboveNorth); float blendAfterBelow = (weightSouth * blendAfterBelowSouth) + (weightNorth * blendAfterBelowNorth); float blendAfterAbove = (weightSouth * blendAfterAboveSouth) + (weightNorth * blendAfterAboveNorth); float blendBefore = (weightBelow * blendBeforeBelow) + (weightAbove * blendBeforeAbove); float blendAfter = (weightBelow * blendAfterBelow) + (weightAbove * blendAfterAbove); float blendTotal = (weightBefore * blendBefore) + (weightAfter * blendAfter); float noiseThisOctave = 2.f * (blendTotal - 0.5f); // Map from [0,1] to [-1,1] // Accumulate results and prepare for next octave (if any) totalNoise += noiseThisOctave * currentAmplitude; totalAmplitude += currentAmplitude; currentAmplitude *= octavePersistence; currentPos *= octaveScale; currentPos.x += OCTAVE_OFFSET; // Add "irrational" offsets to noise position components currentPos.y += OCTAVE_OFFSET; // at each octave to break up their grid alignment currentPos.z += OCTAVE_OFFSET; currentPos.w += OCTAVE_OFFSET; ++ seed; // Eliminates octaves "echoing" each other (since each octave is uniquely seeded) } // Re-normalize total noise to within [-1,1] and fix octaves pulling us far away from limits if( renormalize && totalAmplitude > 0.f ) { totalNoise /= totalAmplitude; // Amplitude exceeds 1.0 if octaves are used totalNoise = (totalNoise * 0.5f) + 0.5f; // Map to [0,1] totalNoise = Easing::SmoothStep( totalNoise ); // Push towards extents (octaves pull us away) totalNoise = (totalNoise * 2.0f) - 1.f; // Map back to [-1,1] } return totalNoise; } //----------------------------------------------------------------------------------------------- // Perlin noise is fractal noise with "gradient vector smoothing" applied. // // In 1D, the gradients are trivial: -1.0 or 1.0, so resulting noise is boring at one octave. // float Compute1dPerlinNoise( float position, float scale, uint32_t numOctaves, float octavePersistence, float octaveScale, bool renormalize, uint32_t seed ) { const float OCTAVE_OFFSET = 0.636764989593174f; // Translation/bias to add to each octave const float gradients[2] = { -1.f, 1.f }; // 1D unit "gradient" vectors; one back, one forward float totalNoise = 0.f; float totalAmplitude = 0.f; float currentAmplitude = 1.f; float currentPosition = position * (1.f / scale); for( uint32_t octaveNum = 0; octaveNum < numOctaves; ++ octaveNum ) { // Determine random "gradient vectors" (just +1 or -1 for 1D Perlin) for surrounding corners float positionFloor = (float) ::glm::floor( currentPosition ); int32_t indexWest = (int32_t) positionFloor; int32_t indexEast = indexWest + 1; float gradientWest = gradients[ Get1dNoiseUint32( indexWest, seed ) & 0x00000001 ]; float gradientEast = gradients[ Get1dNoiseUint32( indexEast, seed ) & 0x00000001 ]; // Dot each point's gradient with displacement from point to position float displacementFromWest = currentPosition - positionFloor; // always positive float displacementFromEast = displacementFromWest - 1.f; // always negative float dotWest = gradientWest * displacementFromWest; // 1D "dot product" is... multiply float dotEast = gradientEast * displacementFromEast; // Do a smoothed (nonlinear) weighted average of dot results float weightEast = Easing::SmoothStep( displacementFromWest ); float weightWest = 1.f - weightEast; float blendTotal = (weightWest * dotWest) + (weightEast * dotEast); float noiseThisOctave = 2.f * blendTotal; // 1D Perlin is in [-.5,.5]; map to [-1,1] // Accumulate results and prepare for next octave (if any) totalNoise += noiseThisOctave * currentAmplitude; totalAmplitude += currentAmplitude; currentAmplitude *= octavePersistence; currentPosition *= octaveScale; currentPosition += OCTAVE_OFFSET; // Add "irrational" offset to de-align octave grids ++ seed; // Eliminates octaves "echoing" each other (since each octave is uniquely seeded) } // Re-normalize total noise to within [-1,1] and fix octaves pulling us far away from limits if( renormalize && totalAmplitude > 0.f ) { totalNoise /= totalAmplitude; // Amplitude exceeds 1.0 if octaves are used totalNoise = (totalNoise * 0.5f) + 0.5f; // Map to [0,1] totalNoise = Easing::SmoothStep( totalNoise ); // Push towards extents (octaves pull us away) totalNoise = (totalNoise * 2.0f) - 1.f; // Map back to [-1,1] } return totalNoise; } //----------------------------------------------------------------------------------------------- // Perlin noise is fractal noise with "gradient vector smoothing" applied. // // In 2D, gradients are unit-length vectors in various directions with even angular distribution. // float Compute2dPerlinNoise( float posX, float posY, float scale, uint32_t numOctaves, float octavePersistence, float octaveScale, bool renormalize, uint32_t seed ) { const float OCTAVE_OFFSET = 0.636764989593174f; // Translation/bias to add to each octave const ::glm::vec2 gradients[ 8 ] = // Normalized unit vectors in 8 quarter-cardinal directions { ::glm::vec2( +0.923879533f, +0.382683432f ), // 22.5 degrees (ENE) ::glm::vec2( +0.382683432f, +0.923879533f ), // 67.5 degrees (NNE) ::glm::vec2( -0.382683432f, +0.923879533f ), // 112.5 degrees (NNW) ::glm::vec2( -0.923879533f, +0.382683432f ), // 157.5 degrees (WNW) ::glm::vec2( -0.923879533f, -0.382683432f ), // 202.5 degrees (WSW) ::glm::vec2( -0.382683432f, -0.923879533f ), // 247.5 degrees (SSW) ::glm::vec2( +0.382683432f, -0.923879533f ), // 292.5 degrees (SSE) ::glm::vec2( +0.923879533f, -0.382683432f ) // 337.5 degrees (ESE) }; float totalNoise = 0.f; float totalAmplitude = 0.f; float currentAmplitude = 1.f; float invScale = (1.f / scale); ::glm::vec2 currentPos( posX * invScale, posY * invScale ); for( uint32_t octaveNum = 0; octaveNum < numOctaves; ++ octaveNum ) { // Determine random unit "gradient vectors" for surrounding corners ::glm::vec2 cellMins( ::glm::floor( currentPos.x ), ::glm::floor( currentPos.y ) ); ::glm::vec2 cellMaxs( cellMins.x + 1.f, cellMins.y + 1.f ); int32_t indexWestX = (int32_t) cellMins.x; int32_t indexSouthY = (int32_t) cellMins.y; int32_t indexEastX = indexWestX + 1; int32_t indexNorthY = indexSouthY + 1; uint32_t noiseSW = Get2dNoiseUint32( indexWestX, indexSouthY, seed ); uint32_t noiseSE = Get2dNoiseUint32( indexEastX, indexSouthY, seed ); uint32_t noiseNW = Get2dNoiseUint32( indexWestX, indexNorthY, seed ); uint32_t noiseNE = Get2dNoiseUint32( indexEastX, indexNorthY, seed ); const ::glm::vec2& gradientSW = gradients[ noiseSW & 0x00000007 ]; const ::glm::vec2& gradientSE = gradients[ noiseSE & 0x00000007 ]; const ::glm::vec2& gradientNW = gradients[ noiseNW & 0x00000007 ]; const ::glm::vec2& gradientNE = gradients[ noiseNE & 0x00000007 ]; // Dot each corner's gradient with displacement from corner to position ::glm::vec2 displacementFromSW( currentPos.x - cellMins.x, currentPos.y - cellMins.y ); ::glm::vec2 displacementFromSE( currentPos.x - cellMaxs.x, currentPos.y - cellMins.y ); ::glm::vec2 displacementFromNW( currentPos.x - cellMins.x, currentPos.y - cellMaxs.y ); ::glm::vec2 displacementFromNE( currentPos.x - cellMaxs.x, currentPos.y - cellMaxs.y ); float dotSouthWest = ::glm::dot( gradientSW, displacementFromSW ); float dotSouthEast = ::glm::dot( gradientSE, displacementFromSE ); float dotNorthWest = ::glm::dot( gradientNW, displacementFromNW ); float dotNorthEast = ::glm::dot( gradientNE, displacementFromNE ); // Do a smoothed (nonlinear) weighted average of dot results float weightEast = Easing::SmoothStep( displacementFromSW.x ); float weightNorth = Easing::SmoothStep( displacementFromSW.y ); float weightWest = 1.f - weightEast; float weightSouth = 1.f - weightNorth; float blendSouth = (weightEast * dotSouthEast) + (weightWest * dotSouthWest); float blendNorth = (weightEast * dotNorthEast) + (weightWest * dotNorthWest); float blendTotal = (weightSouth * blendSouth) + (weightNorth * blendNorth); float noiseThisOctave = blendTotal * (1.f / 0.662578106f); // 2D Perlin is in [-.662578106,.662578106]; map to ~[-1,1] // Accumulate results and prepare for next octave (if any) totalNoise += noiseThisOctave * currentAmplitude; totalAmplitude += currentAmplitude; currentAmplitude *= octavePersistence; currentPos *= octaveScale; currentPos.x += OCTAVE_OFFSET; // Add "irrational" offset to de-align octave grids currentPos.y += OCTAVE_OFFSET; // Add "irrational" offset to de-align octave grids ++ seed; // Eliminates octaves "echoing" each other (since each octave is uniquely seeded) } // Re-normalize total noise to within [-1,1] and fix octaves pulling us far away from limits if( renormalize && totalAmplitude > 0.f ) { totalNoise /= totalAmplitude; // Amplitude exceeds 1.0 if octaves are used totalNoise = (totalNoise * 0.5f) + 0.5f; // Map to [0,1] totalNoise = Easing::SmoothStep( totalNoise ); // Push towards extents (octaves pull us away) totalNoise = (totalNoise * 2.0f) - 1.f; // Map back to [-1,1] } return totalNoise; } //----------------------------------------------------------------------------------------------- // Perlin noise is fractal noise with "gradient vector smoothing" applied. // // In 3D, gradients are unit-length vectors in random (3D) directions. // float Compute3dPerlinNoise( float posX, float posY, float posZ, float scale, uint32_t numOctaves, float octavePersistence, float octaveScale, bool renormalize, uint32_t seed ) { const float OCTAVE_OFFSET = 0.636764989593174f; // Translation/bias to add to each octave const float fSQRT_3_OVER_3 = 0.577350269189f; const ::glm::vec3 gradients[ 8 ] = // Traditional "12 edges" requires modulus and isn't any better. { ::glm::vec3( +fSQRT_3_OVER_3, +fSQRT_3_OVER_3, +fSQRT_3_OVER_3 ), // Normalized unit 3D vectors ::glm::vec3( -fSQRT_3_OVER_3, +fSQRT_3_OVER_3, +fSQRT_3_OVER_3 ), // pointing toward cube ::glm::vec3( +fSQRT_3_OVER_3, -fSQRT_3_OVER_3, +fSQRT_3_OVER_3 ), // corners, so components ::glm::vec3( -fSQRT_3_OVER_3, -fSQRT_3_OVER_3, +fSQRT_3_OVER_3 ), // are all sqrt(3)/3, i.e. ::glm::vec3( +fSQRT_3_OVER_3, +fSQRT_3_OVER_3, -fSQRT_3_OVER_3 ), // 0.5773502691896257645091f. ::glm::vec3( -fSQRT_3_OVER_3, +fSQRT_3_OVER_3, -fSQRT_3_OVER_3 ), // These are slightly better ::glm::vec3( +fSQRT_3_OVER_3, -fSQRT_3_OVER_3, -fSQRT_3_OVER_3 ), // than axes (1,0,0) and much ::glm::vec3( -fSQRT_3_OVER_3, -fSQRT_3_OVER_3, -fSQRT_3_OVER_3 ) // faster than edges (1,1,0). }; float totalNoise = 0.f; float totalAmplitude = 0.f; float currentAmplitude = 1.f; float invScale = (1.f / scale); ::glm::vec3 currentPos( posX * invScale, posY * invScale, posZ * invScale ); for( uint32_t octaveNum = 0; octaveNum < numOctaves; ++ octaveNum ) { // Determine random unit "gradient vectors" for surrounding corners ::glm::vec3 cellMins( ::glm::floor( currentPos.x ), ::glm::floor( currentPos.y ), ::glm::floor( currentPos.z ) ); ::glm::vec3 cellMaxs( cellMins.x + 1.f, cellMins.y + 1.f, cellMins.z + 1.f ); int32_t indexWestX = (int32_t) cellMins.x; int32_t indexSouthY = (int32_t) cellMins.y; int32_t indexBelowZ = (int32_t) cellMins.z; int32_t indexEastX = indexWestX + 1; int32_t indexNorthY = indexSouthY + 1; int32_t indexAboveZ = indexBelowZ + 1; uint32_t noiseBelowSW = Get3dNoiseUint32( indexWestX, indexSouthY, indexBelowZ, seed ); uint32_t noiseBelowSE = Get3dNoiseUint32( indexEastX, indexSouthY, indexBelowZ, seed ); uint32_t noiseBelowNW = Get3dNoiseUint32( indexWestX, indexNorthY, indexBelowZ, seed ); uint32_t noiseBelowNE = Get3dNoiseUint32( indexEastX, indexNorthY, indexBelowZ, seed ); uint32_t noiseAboveSW = Get3dNoiseUint32( indexWestX, indexSouthY, indexAboveZ, seed ); uint32_t noiseAboveSE = Get3dNoiseUint32( indexEastX, indexSouthY, indexAboveZ, seed ); uint32_t noiseAboveNW = Get3dNoiseUint32( indexWestX, indexNorthY, indexAboveZ, seed ); uint32_t noiseAboveNE = Get3dNoiseUint32( indexEastX, indexNorthY, indexAboveZ, seed ); ::glm::vec3 gradientBelowSW = gradients[ noiseBelowSW & 0x00000007 ]; ::glm::vec3 gradientBelowSE = gradients[ noiseBelowSE & 0x00000007 ]; ::glm::vec3 gradientBelowNW = gradients[ noiseBelowNW & 0x00000007 ]; ::glm::vec3 gradientBelowNE = gradients[ noiseBelowNE & 0x00000007 ]; ::glm::vec3 gradientAboveSW = gradients[ noiseAboveSW & 0x00000007 ]; ::glm::vec3 gradientAboveSE = gradients[ noiseAboveSE & 0x00000007 ]; ::glm::vec3 gradientAboveNW = gradients[ noiseAboveNW & 0x00000007 ]; ::glm::vec3 gradientAboveNE = gradients[ noiseAboveNE & 0x00000007 ]; // Dot each corner's gradient with displacement from corner to position ::glm::vec3 displacementFromBelowSW( currentPos.x - cellMins.x, currentPos.y - cellMins.y, currentPos.z - cellMins.z ); ::glm::vec3 displacementFromBelowSE( currentPos.x - cellMaxs.x, currentPos.y - cellMins.y, currentPos.z - cellMins.z ); ::glm::vec3 displacementFromBelowNW( currentPos.x - cellMins.x, currentPos.y - cellMaxs.y, currentPos.z - cellMins.z ); ::glm::vec3 displacementFromBelowNE( currentPos.x - cellMaxs.x, currentPos.y - cellMaxs.y, currentPos.z - cellMins.z ); ::glm::vec3 displacementFromAboveSW( currentPos.x - cellMins.x, currentPos.y - cellMins.y, currentPos.z - cellMaxs.z ); ::glm::vec3 displacementFromAboveSE( currentPos.x - cellMaxs.x, currentPos.y - cellMins.y, currentPos.z - cellMaxs.z ); ::glm::vec3 displacementFromAboveNW( currentPos.x - cellMins.x, currentPos.y - cellMaxs.y, currentPos.z - cellMaxs.z ); ::glm::vec3 displacementFromAboveNE( currentPos.x - cellMaxs.x, currentPos.y - cellMaxs.y, currentPos.z - cellMaxs.z ); float dotBelowSW = ::glm::dot( gradientBelowSW, displacementFromBelowSW ); float dotBelowSE = ::glm::dot( gradientBelowSE, displacementFromBelowSE ); float dotBelowNW = ::glm::dot( gradientBelowNW, displacementFromBelowNW ); float dotBelowNE = ::glm::dot( gradientBelowNE, displacementFromBelowNE ); float dotAboveSW = ::glm::dot( gradientAboveSW, displacementFromAboveSW ); float dotAboveSE = ::glm::dot( gradientAboveSE, displacementFromAboveSE ); float dotAboveNW = ::glm::dot( gradientAboveNW, displacementFromAboveNW ); float dotAboveNE = ::glm::dot( gradientAboveNE, displacementFromAboveNE ); // Do a smoothed (nonlinear) weighted average of dot results float weightEast = Easing::SmoothStep( displacementFromBelowSW.x ); float weightNorth = Easing::SmoothStep( displacementFromBelowSW.y ); float weightAbove = Easing::SmoothStep( displacementFromBelowSW.z ); float weightWest = 1.f - weightEast; float weightSouth = 1.f - weightNorth; float weightBelow = 1.f - weightAbove; // 8-way blend (8 -> 4 -> 2 -> 1) float blendBelowSouth = (weightEast * dotBelowSE) + (weightWest * dotBelowSW); float blendBelowNorth = (weightEast * dotBelowNE) + (weightWest * dotBelowNW); float blendAboveSouth = (weightEast * dotAboveSE) + (weightWest * dotAboveSW); float blendAboveNorth = (weightEast * dotAboveNE) + (weightWest * dotAboveNW); float blendBelow = (weightSouth * blendBelowSouth) + (weightNorth * blendBelowNorth); float blendAbove = (weightSouth * blendAboveSouth) + (weightNorth * blendAboveNorth); float blendTotal = (weightBelow * blendBelow) + (weightAbove * blendAbove); float noiseThisOctave = blendTotal * (1.f / 0.793856621f); // 3D Perlin is in [-.793856621,.793856621]; map to ~[-1,1] // Accumulate results and prepare for next octave (if any) totalNoise += noiseThisOctave * currentAmplitude; totalAmplitude += currentAmplitude; currentAmplitude *= octavePersistence; currentPos *= octaveScale; currentPos.x += OCTAVE_OFFSET; // Add "irrational" offset to de-align octave grids currentPos.y += OCTAVE_OFFSET; // Add "irrational" offset to de-align octave grids currentPos.z += OCTAVE_OFFSET; // Add "irrational" offset to de-align octave grids ++ seed; // Eliminates octaves "echoing" each other (since each octave is uniquely seeded) } // Re-normalize total noise to within [-1,1] and fix octaves pulling us far away from limits if( renormalize && totalAmplitude > 0.f ) { totalNoise /= totalAmplitude; // Amplitude exceeds 1.0 if octaves are used totalNoise = (totalNoise * 0.5f) + 0.5f; // Map to [0,1] totalNoise = Easing::SmoothStep( totalNoise ); // Push towards extents (octaves pull us away) totalNoise = (totalNoise * 2.0f) - 1.f; // Map back to [-1,1] } return totalNoise; } //----------------------------------------------------------------------------------------------- // Perlin noise is fractal noise with "gradient vector smoothing" applied. // // In 4D, gradients are unit-length hyper-vectors in random (4D) directions. // float Compute4dPerlinNoise( float posX, float posY, float posZ, float posT, float scale, uint32_t numOctaves, float octavePersistence, float octaveScale, bool renormalize, uint32_t seed ) { const float OCTAVE_OFFSET = 0.636764989593174f; // Translation/bias to add to each octave const ::glm::vec4 gradients[ 16 ] = // Hard to tell if this is any better in 4D than just having 8 { ::glm::vec4( +0.5f, +0.5f, +0.5f, +0.5f ), // Normalized unit 4D vectors pointing toward each ::glm::vec4( -0.5f, +0.5f, +0.5f, +0.5f ), // of the 16 hypercube corners, so components are ::glm::vec4( +0.5f, -0.5f, +0.5f, +0.5f ), // all sqrt(4)/4, i.e. one-half. ::glm::vec4( -0.5f, -0.5f, +0.5f, +0.5f ), // ::glm::vec4( +0.5f, +0.5f, -0.5f, +0.5f ), // It's hard to tell whether these are any better ::glm::vec4( -0.5f, +0.5f, -0.5f, +0.5f ), // or worse than vectors facing axes (1,0,0,0) or ::glm::vec4( +0.5f, -0.5f, -0.5f, +0.5f ), // 3D edges (.7,.7,0,0) or 4D edges (.57,.57,.57,0) ::glm::vec4( -0.5f, -0.5f, -0.5f, +0.5f ), // but less-axial gradients looked a little better ::glm::vec4( +0.5f, +0.5f, +0.5f, -0.5f ), // with 2D and 3D noise so I'm assuming this is as ::glm::vec4( -0.5f, +0.5f, +0.5f, -0.5f ), // good or better as any other gradient-selection ::glm::vec4( +0.5f, -0.5f, +0.5f, -0.5f ), // scheme (and is crazy-fast). *shrug* ::glm::vec4( -0.5f, -0.5f, +0.5f, -0.5f ), // ::glm::vec4( +0.5f, +0.5f, -0.5f, -0.5f ), // Plus, we want a power-of-two number of evenly- ::glm::vec4( -0.5f, +0.5f, -0.5f, -0.5f ), // distributed gradients, so we can cheaply select ::glm::vec4( +0.5f, -0.5f, -0.5f, -0.5f ), // one from bit-noise (use bit-mask, not modulus). ::glm::vec4( -0.5f, -0.5f, -0.5f, -0.5f ) // }; float totalNoise = 0.f; float totalAmplitude = 0.f; float currentAmplitude = 1.f; float invScale = (1.f / scale); ::glm::vec4 currentPos( posX * invScale, posY * invScale, posZ * invScale, posT * invScale ); for( uint32_t octaveNum = 0; octaveNum < numOctaves; ++ octaveNum ) { // Determine random unit "gradient vectors" for 16 surrounding 4D (hypercube) cell corners ::glm::vec4 cellMins( ::glm::floor( currentPos.x ), ::glm::floor( currentPos.y ), ::glm::floor( currentPos.z ), ::glm::floor( currentPos.w ) ); ::glm::vec4 cellMaxs( cellMins.x + 1.f, cellMins.y + 1.f, cellMins.z + 1.f, cellMins.w + 1.f ); int32_t indexWestX = (int32_t) cellMins.x; int32_t indexSouthY = (int32_t) cellMins.y; int32_t indexBelowZ = (int32_t) cellMins.z; int32_t indexBeforeT = (int32_t) cellMins.w; int32_t indexEastX = indexWestX + 1; int32_t indexNorthY = indexSouthY + 1; int32_t indexAboveZ = indexBelowZ + 1; int32_t indexAfterT = indexBeforeT + 1; // "BeforeBSW" stands for "BeforeBelowSouthWest" below (i.e. 4D hypercube mins), etc. uint32_t noiseBeforeBSW = Get4dNoiseUint32( indexWestX, indexSouthY, indexBelowZ, indexBeforeT, seed ); uint32_t noiseBeforeBSE = Get4dNoiseUint32( indexEastX, indexSouthY, indexBelowZ, indexBeforeT, seed ); uint32_t noiseBeforeBNW = Get4dNoiseUint32( indexWestX, indexNorthY, indexBelowZ, indexBeforeT, seed ); uint32_t noiseBeforeBNE = Get4dNoiseUint32( indexEastX, indexNorthY, indexBelowZ, indexBeforeT, seed ); uint32_t noiseBeforeASW = Get4dNoiseUint32( indexWestX, indexSouthY, indexAboveZ, indexBeforeT, seed ); uint32_t noiseBeforeASE = Get4dNoiseUint32( indexEastX, indexSouthY, indexAboveZ, indexBeforeT, seed ); uint32_t noiseBeforeANW = Get4dNoiseUint32( indexWestX, indexNorthY, indexAboveZ, indexBeforeT, seed ); uint32_t noiseBeforeANE = Get4dNoiseUint32( indexEastX, indexNorthY, indexAboveZ, indexBeforeT, seed ); uint32_t noiseAfterBSW = Get4dNoiseUint32( indexWestX, indexSouthY, indexBelowZ, indexAfterT, seed ); uint32_t noiseAfterBSE = Get4dNoiseUint32( indexEastX, indexSouthY, indexBelowZ, indexAfterT, seed ); uint32_t noiseAfterBNW = Get4dNoiseUint32( indexWestX, indexNorthY, indexBelowZ, indexAfterT, seed ); uint32_t noiseAfterBNE = Get4dNoiseUint32( indexEastX, indexNorthY, indexBelowZ, indexAfterT, seed ); uint32_t noiseAfterASW = Get4dNoiseUint32( indexWestX, indexSouthY, indexAboveZ, indexAfterT, seed ); uint32_t noiseAfterASE = Get4dNoiseUint32( indexEastX, indexSouthY, indexAboveZ, indexAfterT, seed ); uint32_t noiseAfterANW = Get4dNoiseUint32( indexWestX, indexNorthY, indexAboveZ, indexAfterT, seed ); uint32_t noiseAfterANE = Get4dNoiseUint32( indexEastX, indexNorthY, indexAboveZ, indexAfterT, seed ); // Mask with 15 (mod 16) to look up in gradients table ::glm::vec4 gradientBeforeBSW = gradients[ noiseBeforeBSW & 0x0000000F ]; ::glm::vec4 gradientBeforeBSE = gradients[ noiseBeforeBSE & 0x0000000F ]; ::glm::vec4 gradientBeforeBNW = gradients[ noiseBeforeBNW & 0x0000000F ]; ::glm::vec4 gradientBeforeBNE = gradients[ noiseBeforeBNE & 0x0000000F ]; ::glm::vec4 gradientBeforeASW = gradients[ noiseBeforeASW & 0x0000000F ]; ::glm::vec4 gradientBeforeASE = gradients[ noiseBeforeASE & 0x0000000F ]; ::glm::vec4 gradientBeforeANW = gradients[ noiseBeforeANW & 0x0000000F ]; ::glm::vec4 gradientBeforeANE = gradients[ noiseBeforeANE & 0x0000000F ]; ::glm::vec4 gradientAfterBSW = gradients[ noiseAfterBSW & 0x0000000F ]; ::glm::vec4 gradientAfterBSE = gradients[ noiseAfterBSE & 0x0000000F ]; ::glm::vec4 gradientAfterBNW = gradients[ noiseAfterBNW & 0x0000000F ]; ::glm::vec4 gradientAfterBNE = gradients[ noiseAfterBNE & 0x0000000F ]; ::glm::vec4 gradientAfterASW = gradients[ noiseAfterASW & 0x0000000F ]; ::glm::vec4 gradientAfterASE = gradients[ noiseAfterASE & 0x0000000F ]; ::glm::vec4 gradientAfterANW = gradients[ noiseAfterANW & 0x0000000F ]; ::glm::vec4 gradientAfterANE = gradients[ noiseAfterANE & 0x0000000F ]; // Dot each corner's gradient with displacement from corner to position ::glm::vec4 displacementFromBeforeBSW( currentPos.x - cellMins.x, currentPos.y - cellMins.y, currentPos.z - cellMins.z, currentPos.w - cellMins.w ); ::glm::vec4 displacementFromBeforeBSE( currentPos.x - cellMaxs.x, currentPos.y - cellMins.y, currentPos.z - cellMins.z, currentPos.w - cellMins.w ); ::glm::vec4 displacementFromBeforeBNW( currentPos.x - cellMins.x, currentPos.y - cellMaxs.y, currentPos.z - cellMins.z, currentPos.w - cellMins.w ); ::glm::vec4 displacementFromBeforeBNE( currentPos.x - cellMaxs.x, currentPos.y - cellMaxs.y, currentPos.z - cellMins.z, currentPos.w - cellMins.w ); ::glm::vec4 displacementFromBeforeASW( currentPos.x - cellMins.x, currentPos.y - cellMins.y, currentPos.z - cellMaxs.z, currentPos.w - cellMins.w ); ::glm::vec4 displacementFromBeforeASE( currentPos.x - cellMaxs.x, currentPos.y - cellMins.y, currentPos.z - cellMaxs.z, currentPos.w - cellMins.w ); ::glm::vec4 displacementFromBeforeANW( currentPos.x - cellMins.x, currentPos.y - cellMaxs.y, currentPos.z - cellMaxs.z, currentPos.w - cellMins.w ); ::glm::vec4 displacementFromBeforeANE( currentPos.x - cellMaxs.x, currentPos.y - cellMaxs.y, currentPos.z - cellMaxs.z, currentPos.w - cellMins.w ); ::glm::vec4 displacementFromAfterBSW( currentPos.x - cellMins.x, currentPos.y - cellMins.y, currentPos.z - cellMins.z, currentPos.w - cellMaxs.w ); ::glm::vec4 displacementFromAfterBSE( currentPos.x - cellMaxs.x, currentPos.y - cellMins.y, currentPos.z - cellMins.z, currentPos.w - cellMaxs.w ); ::glm::vec4 displacementFromAfterBNW( currentPos.x - cellMins.x, currentPos.y - cellMaxs.y, currentPos.z - cellMins.z, currentPos.w - cellMaxs.w ); ::glm::vec4 displacementFromAfterBNE( currentPos.x - cellMaxs.x, currentPos.y - cellMaxs.y, currentPos.z - cellMins.z, currentPos.w - cellMaxs.w ); ::glm::vec4 displacementFromAfterASW( currentPos.x - cellMins.x, currentPos.y - cellMins.y, currentPos.z - cellMaxs.z, currentPos.w - cellMaxs.w ); ::glm::vec4 displacementFromAfterASE( currentPos.x - cellMaxs.x, currentPos.y - cellMins.y, currentPos.z - cellMaxs.z, currentPos.w - cellMaxs.w ); ::glm::vec4 displacementFromAfterANW( currentPos.x - cellMins.x, currentPos.y - cellMaxs.y, currentPos.z - cellMaxs.z, currentPos.w - cellMaxs.w ); ::glm::vec4 displacementFromAfterANE( currentPos.x - cellMaxs.x, currentPos.y - cellMaxs.y, currentPos.z - cellMaxs.z, currentPos.w - cellMaxs.w ); float dotBeforeBSW = ::glm::dot( gradientBeforeBSW, displacementFromBeforeBSW ); float dotBeforeBSE = ::glm::dot( gradientBeforeBSE, displacementFromBeforeBSE ); float dotBeforeBNW = ::glm::dot( gradientBeforeBNW, displacementFromBeforeBNW ); float dotBeforeBNE = ::glm::dot( gradientBeforeBNE, displacementFromBeforeBNE ); float dotBeforeASW = ::glm::dot( gradientBeforeASW, displacementFromBeforeASW ); float dotBeforeASE = ::glm::dot( gradientBeforeASE, displacementFromBeforeASE ); float dotBeforeANW = ::glm::dot( gradientBeforeANW, displacementFromBeforeANW ); float dotBeforeANE = ::glm::dot( gradientBeforeANE, displacementFromBeforeANE ); float dotAfterBSW = ::glm::dot( gradientAfterBSW, displacementFromAfterBSW ); float dotAfterBSE = ::glm::dot( gradientAfterBSE, displacementFromAfterBSE ); float dotAfterBNW = ::glm::dot( gradientAfterBNW, displacementFromAfterBNW ); float dotAfterBNE = ::glm::dot( gradientAfterBNE, displacementFromAfterBNE ); float dotAfterASW = ::glm::dot( gradientAfterASW, displacementFromAfterASW ); float dotAfterASE = ::glm::dot( gradientAfterASE, displacementFromAfterASE ); float dotAfterANW = ::glm::dot( gradientAfterANW, displacementFromAfterANW ); float dotAfterANE = ::glm::dot( gradientAfterANE, displacementFromAfterANE ); // Do a smoothed (nonlinear) weighted average of dot results float weightEast = Easing::SmoothStep( displacementFromBeforeBSW.x ); float weightNorth = Easing::SmoothStep( displacementFromBeforeBSW.y ); float weightAbove = Easing::SmoothStep( displacementFromBeforeBSW.z ); float weightAfter = Easing::SmoothStep( displacementFromBeforeBSW.w ); float weightWest = 1.f - weightEast; float weightSouth = 1.f - weightNorth; float weightBelow = 1.f - weightAbove; float weightBefore = 1.f - weightAfter; // 16-way blend (16 -> 8 -> 4 -> 2 -> 1) float blendBeforeBelowSouth = (weightEast * dotBeforeBSE) + (weightWest * dotBeforeBSW); float blendBeforeBelowNorth = (weightEast * dotBeforeBNE) + (weightWest * dotBeforeBNW); float blendBeforeAboveSouth = (weightEast * dotBeforeASE) + (weightWest * dotBeforeASW); float blendBeforeAboveNorth = (weightEast * dotBeforeANE) + (weightWest * dotBeforeANW); float blendAfterBelowSouth = (weightEast * dotAfterBSE) + (weightWest * dotAfterBSW); float blendAfterBelowNorth = (weightEast * dotAfterBNE) + (weightWest * dotAfterBNW); float blendAfterAboveSouth = (weightEast * dotAfterASE) + (weightWest * dotAfterASW); float blendAfterAboveNorth = (weightEast * dotAfterANE) + (weightWest * dotAfterANW); float blendBeforeBelow = (weightSouth * blendBeforeBelowSouth) + (weightNorth * blendBeforeBelowNorth); float blendBeforeAbove = (weightSouth * blendBeforeAboveSouth) + (weightNorth * blendBeforeAboveNorth); float blendAfterBelow = (weightSouth * blendAfterBelowSouth) + (weightNorth * blendAfterBelowNorth); float blendAfterAbove = (weightSouth * blendAfterAboveSouth) + (weightNorth * blendAfterAboveNorth); float blendBefore = (weightBelow * blendBeforeBelow) + (weightAbove * blendBeforeAbove); float blendAfter = (weightBelow * blendAfterBelow) + (weightAbove * blendAfterAbove); float blendTotal = (weightBefore * blendBefore) + (weightAfter * blendAfter); float noiseThisOctave = blendTotal * (1.f / 0.6875f); // 4D Perlin is in [-.6875,.6875]; map to ~[-1,1] // Accumulate results and prepare for next octave (if any) totalNoise += noiseThisOctave * currentAmplitude; totalAmplitude += currentAmplitude; currentAmplitude *= octavePersistence; currentPos *= octaveScale; currentPos.x += OCTAVE_OFFSET; // Add "irrational" offset to de-align octave grids currentPos.y += OCTAVE_OFFSET; // Add "irrational" offset to de-align octave grids currentPos.z += OCTAVE_OFFSET; // Add "irrational" offset to de-align octave grids currentPos.w += OCTAVE_OFFSET; // Add "irrational" offset to de-align octave grids ++ seed; // Eliminates octaves "echoing" each other (since each octave is uniquely seeded) } // Re-normalize total noise to within [-1,1] and fix octaves pulling us far away from limits if( renormalize && totalAmplitude > 0.f ) { totalNoise /= totalAmplitude; // Amplitude exceeds 1.0 if octaves are used totalNoise = (totalNoise * 0.5f) + 0.5f; // Map to [0,1] totalNoise = Easing::SmoothStep( totalNoise ); // Push towards extents (octaves pull us away) totalNoise = (totalNoise * 2.0f) - 1.f; // Map back to [-1,1] } return totalNoise; } } // SquirrelNoise4