qt/qtquick3dphysics/src/3rdparty/PhysX/include/foundation/unix/neon/PxUnixNeonInlineAoS.h

Revision control

++ // Copyright (c) 2001-2004 NovodeX AG. All rights reserved.
++ // Copyright (c) 2004-2008 AGEIA Technologies, Inc. All rights reserved.
++ // SPDX-FileCopyrightText: Copyright (c) 2008-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
++ // SPDX-License-Identifier: Apache-2.0
++
++ #ifndef PXFOUNDATION_PXUNIXNEONINLINEAOS_H
++ #define PXFOUNDATION_PXUNIXNEONINLINEAOS_H
++
++ #if !COMPILE_VECTOR_INTRINSICS
++ #error Vector intrinsics should not be included when using scalar implementation.
++ #endif
++
++ namespace physx
++ {
++ namespace aos
++ {
++
++ // improved estimates
++ #define VRECIPEQ recipq_newton<1>
++ #define VRECIPE recip_newton<1>
++ #define VRECIPSQRTEQ rsqrtq_newton<1>
++ #define VRECIPSQRTE rsqrt_newton<1>
++
++ // "exact"
++ #define VRECIPQ recipq_newton<4>
++ #if PX_SWITCH
++ // StabilizationTests.AveragePoint needs more precision to succeed.
++ #define VRECIP recip_newton<5>
++ #else
++ #define VRECIP recip_newton<4>
++ #endif
++ #define VRECIPSQRTQ rsqrtq_newton<4>
++ #define VRECIPSQRT rsqrt_newton<4>
++
++ #define VECMATH_AOS_EPSILON (1e-3f)
++
++ //////////////////////////////////////////////////////////////////////
++ //Test that Vec3V and FloatV are legal
++ //////////////////////////////////
++
++ #define FLOAT_COMPONENTS_EQUAL_THRESHOLD 0.01f
++ PX_FORCE_INLINE bool isValidFloatV(const FloatV a)
++ {
++ /*
++ PX_ALIGN(16, PxF32) data[4];
++ vst1_f32(reinterpret_cast<float32_t*>(data), a);
++ return
++ PxU32* intData = reinterpret_cast<PxU32*>(data);
++ return intData[0] == intData[1];
++ */
++ PX_ALIGN(16, PxF32) data[4];
++ vst1_f32(reinterpret_cast<float32_t*>(data), a);
++ const float32_t x = data[0];
++ const float32_t y = data[1];
++
++ return (x == y);
++
++ /*if (PxAbs(x - y) < FLOAT_COMPONENTS_EQUAL_THRESHOLD)
++ {
++ return true;
++ }
++
++ if (PxAbs((x - y) / x) < FLOAT_COMPONENTS_EQUAL_THRESHOLD)
++ {
++ return true;
++ }
++
++ return false;*/
++ }
++
++ PX_FORCE_INLINE bool isValidVec3V(const Vec3V a)
++ {
++ const float32_t w = vgetq_lane_f32(a, 3);
++ return (0.0f == w);
++ //const PxU32* intData = reinterpret_cast<const PxU32*>(&w);
++ //return *intData == 0;
++ }
++
++ PX_FORCE_INLINE bool isAligned16(const void* a)
++ {
++ return(0 == (size_t(a) & 0x0f));
++ }
++
++ #if PX_DEBUG
++ #define ASSERT_ISVALIDVEC3V(a) PX_ASSERT(isValidVec3V(a))
++ #define ASSERT_ISVALIDFLOATV(a) PX_ASSERT(isValidFloatV(a))
++ #define ASSERT_ISALIGNED16(a) PX_ASSERT(isAligned16(static_cast<const void*>(a)))
++ #else
++ #define ASSERT_ISVALIDVEC3V(a)
++ #define ASSERT_ISVALIDFLOATV(a)
++ #define ASSERT_ISALIGNED16(a)
++ #endif
++
++ namespace internalUnitNeonSimd
++ {
++ PX_FORCE_INLINE PxU32 BAllTrue4_R(const BoolV a)
++ {
++ const uint16x4_t dHigh = vget_high_u16(vreinterpretq_u16_u32(a));
++ const uint16x4_t dLow = vmovn_u32(a);
++ const uint16x8_t combined = vcombine_u16(dLow, dHigh);
++ const uint32x2_t finalReduce = vreinterpret_u32_u8(vmovn_u16(combined));
++ return PxU32(vget_lane_u32(finalReduce, 0) == 0xffffFFFF);
++ }
++
++ PX_FORCE_INLINE PxU32 BAllTrue3_R(const BoolV a)
++ {
++ const uint16x4_t dHigh = vget_high_u16(vreinterpretq_u16_u32(a));
++ const uint16x4_t dLow = vmovn_u32(a);
++ const uint16x8_t combined = vcombine_u16(dLow, dHigh);
++ const uint32x2_t finalReduce = vreinterpret_u32_u8(vmovn_u16(combined));
++ return PxU32((vget_lane_u32(finalReduce, 0) & 0xffFFff) == 0xffFFff);
++ }
++
++ PX_FORCE_INLINE PxU32 BAnyTrue4_R(const BoolV a)
++ {
++ const uint16x4_t dHigh = vget_high_u16(vreinterpretq_u16_u32(a));
++ const uint16x4_t dLow = vmovn_u32(a);
++ const uint16x8_t combined = vcombine_u16(dLow, dHigh);
++ const uint32x2_t finalReduce = vreinterpret_u32_u8(vmovn_u16(combined));
++ return PxU32(vget_lane_u32(finalReduce, 0) != 0x0);
++ }
++
++ PX_FORCE_INLINE PxU32 BAnyTrue3_R(const BoolV a)
++ {
++ const uint16x4_t dHigh = vget_high_u16(vreinterpretq_u16_u32(a));
++ const uint16x4_t dLow = vmovn_u32(a);
++ const uint16x8_t combined = vcombine_u16(dLow, dHigh);
++ const uint32x2_t finalReduce = vreinterpret_u32_u8(vmovn_u16(combined));
++ return PxU32((vget_lane_u32(finalReduce, 0) & 0xffFFff) != 0);
++ }
++ }
++
++ namespace vecMathTests
++ {
++ // PT: this function returns an invalid Vec3V (W!=0.0f) just for unit-testing 'isValidVec3V'
++ PX_FORCE_INLINE Vec3V getInvalidVec3V()
++ {
++ PX_ALIGN(16, PxF32) data[4] = { 1.0f, 1.0f, 1.0f, 1.0f };
++ return V4LoadA(data);
++ }
++
++ PX_FORCE_INLINE bool allElementsEqualFloatV(const FloatV a, const FloatV b)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(b);
++ return vget_lane_u32(vceq_f32(a, b), 0) != 0;
++ }
++
++ PX_FORCE_INLINE bool allElementsEqualVec3V(const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ return V3AllEq(a, b) != 0;
++ }
++
++ PX_FORCE_INLINE bool allElementsEqualVec4V(const Vec4V a, const Vec4V b)
++ {
++ return V4AllEq(a, b) != 0;
++ }
++
++ PX_FORCE_INLINE bool allElementsEqualBoolV(const BoolV a, const BoolV b)
++ {
++ return internalUnitNeonSimd::BAllTrue4_R(vceqq_u32(a, b)) != 0;
++ }
++
++ PX_FORCE_INLINE PxU32 V4U32AllEq(const VecU32V a, const VecU32V b)
++ {
++ return internalUnitNeonSimd::BAllTrue4_R(V4IsEqU32(a, b));
++ }
++
++ PX_FORCE_INLINE bool allElementsEqualVecU32V(const VecU32V a, const VecU32V b)
++ {
++ return V4U32AllEq(a, b) != 0;
++ }
++
++ PX_FORCE_INLINE BoolV V4IsEqI32(const VecI32V a, const VecI32V b)
++ {
++ return vceqq_s32(a, b);
++ }
++
++ PX_FORCE_INLINE PxU32 V4I32AllEq(const VecI32V a, const VecI32V b)
++ {
++ return internalUnitNeonSimd::BAllTrue4_R(V4IsEqI32(a, b));
++ }
++
++ PX_FORCE_INLINE bool allElementsEqualVecI32V(const VecI32V a, const VecI32V b)
++ {
++ return V4I32AllEq(a, b) != 0;
++ }
++
++ PX_FORCE_INLINE bool allElementsNearEqualFloatV(const FloatV a, const FloatV b)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(b);
++
++ const float32x2_t c = vsub_f32(a, b);
++ const float32x2_t error = vdup_n_f32(VECMATH_AOS_EPSILON);
++ // absolute compare abs(error) > abs(c)
++ const uint32x2_t greater = vcagt_f32(error, c);
++ const uint32x2_t min = vpmin_u32(greater, greater);
++ return vget_lane_u32(min, 0) != 0x0;
++ }
++
++ PX_FORCE_INLINE bool allElementsNearEqualVec3V(const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ const float32x4_t c = vsubq_f32(a, b);
++ const float32x4_t error = vdupq_n_f32(VECMATH_AOS_EPSILON);
++ // absolute compare abs(error) > abs(c)
++ const uint32x4_t greater = vcagtq_f32(error, c);
++ return internalUnitNeonSimd::BAllTrue3_R(greater) != 0;
++ }
++
++ PX_FORCE_INLINE bool allElementsNearEqualVec4V(const Vec4V a, const Vec4V b)
++ {
++ const float32x4_t c = vsubq_f32(a, b);
++ const float32x4_t error = vdupq_n_f32(VECMATH_AOS_EPSILON);
++ // absolute compare abs(error) > abs(c)
++ const uint32x4_t greater = vcagtq_f32(error, c);
++ return internalUnitNeonSimd::BAllTrue4_R(greater) != 0x0;
++ }
++ }
++
++ #if 0 // debugging printfs
++ #include <stdio.h>
++ PX_FORCE_INLINE void printVec(const float32x4_t& v, const char* name)
++ {
++ PX_ALIGN(16, float32_t) data[4];
++ vst1q_f32(data, v);
++ printf("%s: (%f, %f, %f, %f)\n", name, data[0], data[1], data[2], data[3]);
++ }
++
++ PX_FORCE_INLINE void printVec(const float32x2_t& v, const char* name)
++ {
++ PX_ALIGN(16, float32_t) data[2];
++ vst1_f32(data, v);
++ printf("%s: (%f, %f)\n", name, data[0], data[1]);
++ }
++
++ PX_FORCE_INLINE void printVec(const uint32x4_t& v, const char* name)
++ {
++ PX_ALIGN(16, uint32_t) data[4];
++ vst1q_u32(data, v);
++ printf("%s: (0x%x, 0x%x, 0x%x, 0x%x)\n", name, data[0], data[1], data[2], data[3]);
++ }
++
++ PX_FORCE_INLINE void printVec(const uint16x8_t& v, const char* name)
++ {
++ PX_ALIGN(16, uint16_t) data[8];
++ vst1q_u16(data, v);
++ printf("%s: (0x%x, 0x%x, 0x%x, 0x%x, 0x%x, 0x%x, 0x%x, 0x%x)\n", name, data[0], data[1], data[2], data[3],
++ data[4], data[5], data[6], data[7]);
++ }
++
++ PX_FORCE_INLINE void printVec(const int32x4_t& v, const char* name)
++ {
++ PX_ALIGN(16, int32_t) data[4];
++ vst1q_s32(data, v);
++ printf("%s: (0x%x, 0x%x, 0x%x, 0x%x)\n", name, data[0], data[1], data[2], data[3]);
++ }
++
++ PX_FORCE_INLINE void printVec(const int16x8_t& v, const char* name)
++ {
++ PX_ALIGN(16, int16_t) data[8];
++ vst1q_s16(data, v);
++ printf("%s: (0x%x, 0x%x, 0x%x, 0x%x, 0x%x, 0x%x, 0x%x, 0x%x)\n", name, data[0], data[1], data[2], data[3],
++ data[4], data[5], data[6], data[7]);
++ }
++
++ PX_FORCE_INLINE void printVec(const uint16x4_t& v, const char* name)
++ {
++ PX_ALIGN(16, uint16_t) data[4];
++ vst1_u16(data, v);
++ printf("%s: (0x%x, 0x%x, 0x%x, 0x%x)\n", name, data[0], data[1], data[2], data[3]);
++ }
++
++ PX_FORCE_INLINE void printVec(const uint32x2_t& v, const char* name)
++ {
++ PX_ALIGN(16, uint32_t) data[2];
++ vst1_u32(data, v);
++ printf("%s: (0x%x, 0x%x)\n", name, data[0], data[1]);
++ }
++
++ PX_FORCE_INLINE void printVar(const PxU32 v, const char* name)
++ {
++ printf("%s: 0x%x\n", name, v);
++ }
++
++ PX_FORCE_INLINE void printVar(const PxF32 v, const char* name)
++ {
++ printf("%s: %f\n", name, v);
++ }
++
++ #define PRINT_VAR(X) printVar((X), #X)
++ #define PRINT_VEC(X) printVec((X), #X)
++ #define PRINT_VEC_TITLE(TITLE, X) printVec((X), TITLE #X)
++ #endif // debugging printf
++
++ /////////////////////////////////////////////////////////////////////
++ ////FUNCTIONS USED ONLY FOR ASSERTS IN VECTORISED IMPLEMENTATIONS
++ /////////////////////////////////////////////////////////////////////
++
++ PX_FORCE_INLINE bool isFiniteFloatV(const FloatV a)
++ {
++ PX_ALIGN(16, PxF32) data[4];
++ vst1_f32(reinterpret_cast<float32_t*>(data), a);
++ return PxIsFinite(data[0]) && PxIsFinite(data[1]);
++ }
++
++ PX_FORCE_INLINE bool isFiniteVec3V(const Vec3V a)
++ {
++ PX_ALIGN(16, PxF32) data[4];
++ vst1q_f32(reinterpret_cast<float32_t*>(data), a);
++ return PxIsFinite(data[0]) && PxIsFinite(data[1]) && PxIsFinite(data[2]);
++ }
++
++ PX_FORCE_INLINE bool isFiniteVec4V(const Vec4V a)
++ {
++ PX_ALIGN(16, PxF32) data[4];
++ vst1q_f32(reinterpret_cast<float32_t*>(data), a);
++ return PxIsFinite(data[0]) && PxIsFinite(data[1]) && PxIsFinite(data[2]) && PxIsFinite(data[3]);
++ }
++
++ PX_FORCE_INLINE bool hasZeroElementinFloatV(const FloatV a)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ return vget_lane_u32(vreinterpret_u32_f32(a), 0) == 0;
++ }
++
++ PX_FORCE_INLINE bool hasZeroElementInVec3V(const Vec3V a)
++ {
++ const uint32x2_t dLow = vget_low_u32(vreinterpretq_u32_f32(a));
++ const uint32x2_t dMin = vpmin_u32(dLow, dLow);
++
++ return vget_lane_u32(dMin, 0) == 0 || vgetq_lane_u32(vreinterpretq_u32_f32(a), 2) == 0;
++ }
++
++ PX_FORCE_INLINE bool hasZeroElementInVec4V(const Vec4V a)
++ {
++ const uint32x2_t dHigh = vget_high_u32(vreinterpretq_u32_f32(a));
++ const uint32x2_t dLow = vget_low_u32(vreinterpretq_u32_f32(a));
++
++ const uint32x2_t dMin = vmin_u32(dHigh, dLow);
++ const uint32x2_t pairMin = vpmin_u32(dMin, dMin);
++ return vget_lane_u32(pairMin, 0) == 0;
++ }
++
++ /////////////////////////////////////////////////////////////////////
++ ////VECTORISED FUNCTION IMPLEMENTATIONS
++ /////////////////////////////////////////////////////////////////////
++
++ PX_FORCE_INLINE FloatV FLoad(const PxF32 f)
++ {
++ return vdup_n_f32(reinterpret_cast<const float32_t&>(f));
++ }
++
++ PX_FORCE_INLINE FloatV FLoadA(const PxF32* const f)
++ {
++ ASSERT_ISALIGNED16(f);
++ return vld1_f32(reinterpret_cast<const float32_t*>(f));
++ }
++
++ PX_FORCE_INLINE Vec3V V3Load(const PxF32 f)
++ {
++ PX_ALIGN(16, PxF32) data[4] = { f, f, f, 0.0f };
++ return V4LoadA(data);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Load(const PxF32 f)
++ {
++ return vdupq_n_f32(reinterpret_cast<const float32_t&>(f));
++ }
++
++ PX_FORCE_INLINE BoolV BLoad(const bool f)
++ {
++ const PxU32 i = static_cast<PxU32>(-(static_cast<PxI32>(f)));
++ return vdupq_n_u32(i);
++ }
++
++ PX_FORCE_INLINE Vec3V V3LoadA(const PxVec3& f)
++ {
++ ASSERT_ISALIGNED16(&f);
++ PX_ALIGN(16, PxF32) data[4] = { f.x, f.y, f.z, 0.0f };
++ return V4LoadA(data);
++ }
++
++ PX_FORCE_INLINE Vec3V V3LoadU(const PxVec3& f)
++ {
++ PX_ALIGN(16, PxF32) data[4] = { f.x, f.y, f.z, 0.0f };
++ return V4LoadA(data);
++ }
++
++ PX_FORCE_INLINE Vec3V V3LoadUnsafeA(const PxVec3& f)
++ {
++ ASSERT_ISALIGNED16(&f);
++ PX_ALIGN(16, PxF32) data[4] = { f.x, f.y, f.z, 0.0f };
++ return V4LoadA(data);
++ }
++
++ PX_FORCE_INLINE Vec3V V3LoadA(const PxF32* f)
++ {
++ ASSERT_ISALIGNED16(f);
++ PX_ALIGN(16, PxF32) data[4] = { f[0], f[1], f[2], 0.0f };
++ return V4LoadA(data);
++ }
++
++ PX_FORCE_INLINE Vec3V V3LoadU(const PxF32* f)
++ {
++ PX_ALIGN(16, PxF32) data[4] = { f[0], f[1], f[2], 0.0f };
++ return V4LoadA(data);
++ }
++
++ PX_FORCE_INLINE Vec3V Vec3V_From_Vec4V(Vec4V v)
++ {
++ return vsetq_lane_f32(0.0f, v, 3);
++ }
++
++ PX_FORCE_INLINE Vec3V Vec3V_From_Vec4V_WUndefined(Vec4V v)
++ {
++ return v;
++ }
++
++ PX_FORCE_INLINE Vec4V Vec4V_From_Vec3V(Vec3V f)
++ {
++ return f; // ok if it is implemented as the same type.
++ }
++
++ PX_FORCE_INLINE Vec4V Vec4V_From_FloatV(FloatV f)
++ {
++ return vcombine_f32(f, f);
++ }
++
++ PX_FORCE_INLINE Vec3V Vec3V_From_FloatV(FloatV f)
++ {
++ return Vec3V_From_Vec4V(Vec4V_From_FloatV(f));
++ }
++
++ PX_FORCE_INLINE Vec3V Vec3V_From_FloatV_WUndefined(FloatV f)
++ {
++ return Vec3V_From_Vec4V_WUndefined(Vec4V_From_FloatV(f));
++ }
++
++ PX_FORCE_INLINE Vec4V Vec4V_From_PxVec3_WUndefined(const PxVec3& f)
++ {
++ PX_ALIGN(16, PxF32) data[4] = { f.x, f.y, f.z, 0.0f };
++ return V4LoadA(data);
++ }
++
++ PX_FORCE_INLINE Mat33V Mat33V_From_PxMat33(const PxMat33& m)
++ {
++ return Mat33V(V3LoadU(m.column0), V3LoadU(m.column1), V3LoadU(m.column2));
++ }
++
++ PX_FORCE_INLINE void PxMat33_From_Mat33V(const Mat33V& m, PxMat33& out)
++ {
++ V3StoreU(m.col0, out.column0);
++ V3StoreU(m.col1, out.column1);
++ V3StoreU(m.col2, out.column2);
++ }
++
++ PX_FORCE_INLINE Vec4V V4LoadA(const PxF32* const f)
++ {
++ ASSERT_ISALIGNED16(f);
++ return vld1q_f32(reinterpret_cast<const float32_t*>(f));
++ }
++
++ PX_FORCE_INLINE void V4StoreA(Vec4V a, PxF32* f)
++ {
++ ASSERT_ISALIGNED16(f);
++ vst1q_f32(reinterpret_cast<float32_t*>(f), a);
++ }
++
++ PX_FORCE_INLINE void V4StoreU(const Vec4V a, PxF32* f)
++ {
++ PX_ALIGN(16, PxF32) f2[4];
++ vst1q_f32(reinterpret_cast<float32_t*>(f2), a);
++ f[0] = f2[0];
++ f[1] = f2[1];
++ f[2] = f2[2];
++ f[3] = f2[3];
++ }
++
++ PX_FORCE_INLINE void BStoreA(const BoolV a, PxU32* u)
++ {
++ ASSERT_ISALIGNED16(u);
++ vst1q_u32(reinterpret_cast<uint32_t*>(u), a);
++ }
++
++ PX_FORCE_INLINE void U4StoreA(const VecU32V uv, PxU32* u)
++ {
++ ASSERT_ISALIGNED16(u);
++ vst1q_u32(reinterpret_cast<uint32_t*>(u), uv);
++ }
++
++ PX_FORCE_INLINE void I4StoreA(const VecI32V iv, PxI32* i)
++ {
++ ASSERT_ISALIGNED16(i);
++ vst1q_s32(reinterpret_cast<int32_t*>(i), iv);
++ }
++
++ PX_FORCE_INLINE Vec4V V4LoadU(const PxF32* const f)
++ {
++ return vld1q_f32(reinterpret_cast<const float32_t*>(f));
++ }
++
++ PX_FORCE_INLINE BoolV BLoad(const bool* const f)
++ {
++ const PX_ALIGN(16, PxU32) b[4] = { static_cast<PxU32>(-static_cast<PxI32>(f[0])),
++ static_cast<PxU32>(-static_cast<PxI32>(f[1])),
++ static_cast<PxU32>(-static_cast<PxI32>(f[2])),
++ static_cast<PxU32>(-static_cast<PxI32>(f[3])) };
++ return vld1q_u32(b);
++ }
++
++ PX_FORCE_INLINE void FStore(const FloatV a, PxF32* PX_RESTRICT f)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ // vst1q_lane_f32(f, a, 0); // causes vst1 alignment bug
++ *f = vget_lane_f32(a, 0);
++ }
++
++ PX_FORCE_INLINE void Store_From_BoolV(const BoolV a, PxU32* PX_RESTRICT f)
++ {
++ *f = vget_lane_u32(vget_low_u32(a), 0);
++ }
++
++ PX_FORCE_INLINE void V3StoreA(const Vec3V a, PxVec3& f)
++ {
++ ASSERT_ISALIGNED16(&f);
++ PX_ALIGN(16, PxF32) f2[4];
++ vst1q_f32(reinterpret_cast<float32_t*>(f2), a);
++ f = PxVec3(f2[0], f2[1], f2[2]);
++ }
++
++ PX_FORCE_INLINE void V3StoreU(const Vec3V a, PxVec3& f)
++ {
++ PX_ALIGN(16, PxF32) f2[4];
++ vst1q_f32(reinterpret_cast<float32_t*>(f2), a);
++ f = PxVec3(f2[0], f2[1], f2[2]);
++ }
++
++ //////////////////////////////////
++ // FLOATV
++ //////////////////////////////////
++
++ PX_FORCE_INLINE FloatV FZero()
++ {
++ return FLoad(0.0f);
++ }
++
++ PX_FORCE_INLINE FloatV FOne()
++ {
++ return FLoad(1.0f);
++ }
++
++ PX_FORCE_INLINE FloatV FHalf()
++ {
++ return FLoad(0.5f);
++ }
++
++ PX_FORCE_INLINE FloatV FEps()
++ {
++ return FLoad(PX_EPS_REAL);
++ }
++
++ PX_FORCE_INLINE FloatV FEps6()
++ {
++ return FLoad(1e-6f);
++ }
++
++ PX_FORCE_INLINE FloatV FMax()
++ {
++ return FLoad(PX_MAX_REAL);
++ }
++
++ PX_FORCE_INLINE FloatV FNegMax()
++ {
++ return FLoad(-PX_MAX_REAL);
++ }
++
++ PX_FORCE_INLINE FloatV IZero()
++ {
++ return vreinterpret_f32_u32(vdup_n_u32(0));
++ }
++
++ PX_FORCE_INLINE FloatV IOne()
++ {
++ return vreinterpret_f32_u32(vdup_n_u32(1));
++ }
++
++ PX_FORCE_INLINE FloatV ITwo()
++ {
++ return vreinterpret_f32_u32(vdup_n_u32(2));
++ }
++
++ PX_FORCE_INLINE FloatV IThree()
++ {
++ return vreinterpret_f32_u32(vdup_n_u32(3));
++ }
++
++ PX_FORCE_INLINE FloatV IFour()
++ {
++ return vreinterpret_f32_u32(vdup_n_u32(4));
++ }
++
++ PX_FORCE_INLINE FloatV FNeg(const FloatV f)
++ {
++ ASSERT_ISVALIDFLOATV(f);
++ return vneg_f32(f);
++ }
++
++ PX_FORCE_INLINE FloatV FAdd(const FloatV a, const FloatV b)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(b);
++ return vadd_f32(a, b);
++ }
++
++ PX_FORCE_INLINE FloatV FSub(const FloatV a, const FloatV b)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(b);
++ return vsub_f32(a, b);
++ }
++
++ PX_FORCE_INLINE FloatV FMul(const FloatV a, const FloatV b)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(b);
++ return vmul_f32(a, b);
++ }
++
++ template <int n>
++ PX_FORCE_INLINE float32x2_t recip_newton(const float32x2_t& in)
++ {
++ float32x2_t recip = vrecpe_f32(in);
++ for(int i = 0; i < n; ++i)
++ recip = vmul_f32(recip, vrecps_f32(in, recip));
++ return recip;
++ }
++
++ template <int n>
++ PX_FORCE_INLINE float32x4_t recipq_newton(const float32x4_t& in)
++ {
++ float32x4_t recip = vrecpeq_f32(in);
++ for(int i = 0; i < n; ++i)
++ recip = vmulq_f32(recip, vrecpsq_f32(recip, in));
++ return recip;
++ }
++
++ template <int n>
++ PX_FORCE_INLINE float32x2_t rsqrt_newton(const float32x2_t& in)
++ {
++ float32x2_t rsqrt = vrsqrte_f32(in);
++ for(int i = 0; i < n; ++i)
++ rsqrt = vmul_f32(rsqrt, vrsqrts_f32(vmul_f32(rsqrt, rsqrt), in));
++ return rsqrt;
++ }
++
++ template <int n>
++ PX_FORCE_INLINE float32x4_t rsqrtq_newton(const float32x4_t& in)
++ {
++ float32x4_t rsqrt = vrsqrteq_f32(in);
++ for(int i = 0; i < n; ++i)
++ rsqrt = vmulq_f32(rsqrt, vrsqrtsq_f32(vmulq_f32(rsqrt, rsqrt), in));
++ return rsqrt;
++ }
++
++ PX_FORCE_INLINE FloatV FDiv(const FloatV a, const FloatV b)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(b);
++ return vmul_f32(a, VRECIP(b));
++ }
++
++ PX_FORCE_INLINE FloatV FDivFast(const FloatV a, const FloatV b)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(b);
++ return vmul_f32(a, VRECIPE(b));
++ }
++
++ PX_FORCE_INLINE FloatV FRecip(const FloatV a)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ return VRECIP(a);
++ }
++
++ PX_FORCE_INLINE FloatV FRecipFast(const FloatV a)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ return VRECIPE(a);
++ }
++
++ PX_FORCE_INLINE FloatV FRsqrt(const FloatV a)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ return VRECIPSQRT(a);
++ }
++
++ PX_FORCE_INLINE FloatV FSqrt(const FloatV a)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ return FSel(FIsEq(a, FZero()), a, vmul_f32(a, VRECIPSQRT(a)));
++ }
++
++ PX_FORCE_INLINE FloatV FRsqrtFast(const FloatV a)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ return VRECIPSQRTE(a);
++ }
++
++ PX_FORCE_INLINE FloatV FScaleAdd(const FloatV a, const FloatV b, const FloatV c)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(b);
++ ASSERT_ISVALIDFLOATV(c);
++ return vmla_f32(c, a, b);
++ }
++
++ PX_FORCE_INLINE FloatV FNegScaleSub(const FloatV a, const FloatV b, const FloatV c)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(b);
++ ASSERT_ISVALIDFLOATV(c);
++ return vmls_f32(c, a, b);
++ }
++
++ PX_FORCE_INLINE FloatV FAbs(const FloatV a)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ return vabs_f32(a);
++ }
++
++ PX_FORCE_INLINE FloatV FSel(const BoolV c, const FloatV a, const FloatV b)
++ {
++ PX_ASSERT( vecMathTests::allElementsEqualBoolV(c, BTTTT()) ||
++ vecMathTests::allElementsEqualBoolV(c, BFFFF()));
++ ASSERT_ISVALIDFLOATV(vbsl_f32(vget_low_u32(c), a, b));
++ return vbsl_f32(vget_low_u32(c), a, b);
++ }
++
++ PX_FORCE_INLINE BoolV FIsGrtr(const FloatV a, const FloatV b)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(b);
++ return vdupq_lane_u32(vcgt_f32(a, b), 0);
++ }
++
++ PX_FORCE_INLINE BoolV FIsGrtrOrEq(const FloatV a, const FloatV b)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(b);
++ return vdupq_lane_u32(vcge_f32(a, b), 0);
++ }
++
++ PX_FORCE_INLINE BoolV FIsEq(const FloatV a, const FloatV b)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(b);
++ return vdupq_lane_u32(vceq_f32(a, b), 0);
++ }
++
++ PX_FORCE_INLINE FloatV FMax(const FloatV a, const FloatV b)
++ {
++ //ASSERT_ISVALIDFLOATV(a);
++ //ASSERT_ISVALIDFLOATV(b);
++ return vmax_f32(a, b);
++ }
++
++ PX_FORCE_INLINE FloatV FMin(const FloatV a, const FloatV b)
++ {
++ //ASSERT_ISVALIDFLOATV(a);
++ //ASSERT_ISVALIDFLOATV(b);
++ return vmin_f32(a, b);
++ }
++
++ PX_FORCE_INLINE FloatV FClamp(const FloatV a, const FloatV minV, const FloatV maxV)
++ {
++ ASSERT_ISVALIDFLOATV(minV);
++ ASSERT_ISVALIDFLOATV(maxV);
++ return vmax_f32(vmin_f32(a, maxV), minV);
++ }
++
++ PX_FORCE_INLINE PxU32 FAllGrtr(const FloatV a, const FloatV b)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(b);
++ return vget_lane_u32(vcgt_f32(a, b), 0);
++ }
++
++ PX_FORCE_INLINE PxU32 FAllGrtrOrEq(const FloatV a, const FloatV b)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(b);
++ return vget_lane_u32(vcge_f32(a, b), 0);
++ }
++
++ PX_FORCE_INLINE PxU32 FAllEq(const FloatV a, const FloatV b)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(b);
++ return vget_lane_u32(vceq_f32(a, b), 0);
++ }
++
++ PX_FORCE_INLINE FloatV FRound(const FloatV a)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++
++ // truncate(a + (0.5f - sign(a)))
++ const float32x2_t half = vdup_n_f32(0.5f);
++ const float32x2_t sign = vcvt_f32_u32((vshr_n_u32(vreinterpret_u32_f32(a), 31)));
++ const float32x2_t aPlusHalf = vadd_f32(a, half);
++ const float32x2_t aRound = vsub_f32(aPlusHalf, sign);
++ int32x2_t tmp = vcvt_s32_f32(aRound);
++ return vcvt_f32_s32(tmp);
++ }
++
++ PX_FORCE_INLINE FloatV FSin(const FloatV a)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++
++ // Modulo the range of the given angles such that -XM_2PI <= Angles < XM_2PI
++ const FloatV recipTwoPi = FLoadA(g_PXReciprocalTwoPi.f);
++ const FloatV twoPi = FLoadA(g_PXTwoPi.f);
++ const FloatV tmp = FMul(a, recipTwoPi);
++ const FloatV b = FRound(tmp);
++ const FloatV V1 = FNegScaleSub(twoPi, b, a);
++
++ // sin(V) ~= V - V^3 / 3! + V^5 / 5! - V^7 / 7! + V^9 / 9! - V^11 / 11! + V^13 / 13! -
++ // V^15 / 15! + V^17 / 17! - V^19 / 19! + V^21 / 21! - V^23 / 23! (for -PI <= V < PI)
++ const FloatV V2 = FMul(V1, V1);
++ const FloatV V3 = FMul(V2, V1);
++ const FloatV V5 = FMul(V3, V2);
++ const FloatV V7 = FMul(V5, V2);
++ const FloatV V9 = FMul(V7, V2);
++ const FloatV V11 = FMul(V9, V2);
++ const FloatV V13 = FMul(V11, V2);
++ const FloatV V15 = FMul(V13, V2);
++ const FloatV V17 = FMul(V15, V2);
++ const FloatV V19 = FMul(V17, V2);
++ const FloatV V21 = FMul(V19, V2);
++ const FloatV V23 = FMul(V21, V2);
++
++ const Vec4V sinCoefficients0 = V4LoadA(g_PXSinCoefficients0.f);
++ const Vec4V sinCoefficients1 = V4LoadA(g_PXSinCoefficients1.f);
++ const Vec4V sinCoefficients2 = V4LoadA(g_PXSinCoefficients2.f);
++
++ const FloatV S1 = V4GetY(sinCoefficients0);
++ const FloatV S2 = V4GetZ(sinCoefficients0);
++ const FloatV S3 = V4GetW(sinCoefficients0);
++ const FloatV S4 = V4GetX(sinCoefficients1);
++ const FloatV S5 = V4GetY(sinCoefficients1);
++ const FloatV S6 = V4GetZ(sinCoefficients1);
++ const FloatV S7 = V4GetW(sinCoefficients1);
++ const FloatV S8 = V4GetX(sinCoefficients2);
++ const FloatV S9 = V4GetY(sinCoefficients2);
++ const FloatV S10 = V4GetZ(sinCoefficients2);
++ const FloatV S11 = V4GetW(sinCoefficients2);
++
++ FloatV Result;
++ Result = FScaleAdd(S1, V3, V1);
++ Result = FScaleAdd(S2, V5, Result);
++ Result = FScaleAdd(S3, V7, Result);
++ Result = FScaleAdd(S4, V9, Result);
++ Result = FScaleAdd(S5, V11, Result);
++ Result = FScaleAdd(S6, V13, Result);
++ Result = FScaleAdd(S7, V15, Result);
++ Result = FScaleAdd(S8, V17, Result);
++ Result = FScaleAdd(S9, V19, Result);
++ Result = FScaleAdd(S10, V21, Result);
++ Result = FScaleAdd(S11, V23, Result);
++
++ return Result;
++ }
++
++ PX_FORCE_INLINE FloatV FCos(const FloatV a)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++
++ // Modulo the range of the given angles such that -XM_2PI <= Angles < XM_2PI
++ const FloatV recipTwoPi = FLoadA(g_PXReciprocalTwoPi.f);
++ const FloatV twoPi = FLoadA(g_PXTwoPi.f);
++ const FloatV tmp = FMul(a, recipTwoPi);
++ const FloatV b = FRound(tmp);
++ const FloatV V1 = FNegScaleSub(twoPi, b, a);
++
++ // cos(V) ~= 1 - V^2 / 2! + V^4 / 4! - V^6 / 6! + V^8 / 8! - V^10 / 10! + V^12 / 12! -
++ // V^14 / 14! + V^16 / 16! - V^18 / 18! + V^20 / 20! - V^22 / 22! (for -PI <= V < PI)
++ const FloatV V2 = FMul(V1, V1);
++ const FloatV V4 = FMul(V2, V2);
++ const FloatV V6 = FMul(V4, V2);
++ const FloatV V8 = FMul(V4, V4);
++ const FloatV V10 = FMul(V6, V4);
++ const FloatV V12 = FMul(V6, V6);
++ const FloatV V14 = FMul(V8, V6);
++ const FloatV V16 = FMul(V8, V8);
++ const FloatV V18 = FMul(V10, V8);
++ const FloatV V20 = FMul(V10, V10);
++ const FloatV V22 = FMul(V12, V10);
++
++ const Vec4V cosCoefficients0 = V4LoadA(g_PXCosCoefficients0.f);
++ const Vec4V cosCoefficients1 = V4LoadA(g_PXCosCoefficients1.f);
++ const Vec4V cosCoefficients2 = V4LoadA(g_PXCosCoefficients2.f);
++
++ const FloatV C1 = V4GetY(cosCoefficients0);
++ const FloatV C2 = V4GetZ(cosCoefficients0);
++ const FloatV C3 = V4GetW(cosCoefficients0);
++ const FloatV C4 = V4GetX(cosCoefficients1);
++ const FloatV C5 = V4GetY(cosCoefficients1);
++ const FloatV C6 = V4GetZ(cosCoefficients1);
++ const FloatV C7 = V4GetW(cosCoefficients1);
++ const FloatV C8 = V4GetX(cosCoefficients2);
++ const FloatV C9 = V4GetY(cosCoefficients2);
++ const FloatV C10 = V4GetZ(cosCoefficients2);
++ const FloatV C11 = V4GetW(cosCoefficients2);
++
++ FloatV Result;
++ Result = FScaleAdd(C1, V2, FOne());
++ Result = FScaleAdd(C2, V4, Result);
++ Result = FScaleAdd(C3, V6, Result);
++ Result = FScaleAdd(C4, V8, Result);
++ Result = FScaleAdd(C5, V10, Result);
++ Result = FScaleAdd(C6, V12, Result);
++ Result = FScaleAdd(C7, V14, Result);
++ Result = FScaleAdd(C8, V16, Result);
++ Result = FScaleAdd(C9, V18, Result);
++ Result = FScaleAdd(C10, V20, Result);
++ Result = FScaleAdd(C11, V22, Result);
++
++ return Result;
++ }
++
++ PX_FORCE_INLINE PxU32 FOutOfBounds(const FloatV a, const FloatV min, const FloatV max)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(min);
++ ASSERT_ISVALIDFLOATV(max);
++
++ const BoolV c = BOr(FIsGrtr(a, max), FIsGrtr(min, a));
++ return PxU32(!BAllEqFFFF(c));
++ }
++
++ PX_FORCE_INLINE PxU32 FInBounds(const FloatV a, const FloatV min, const FloatV max)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(min);
++ ASSERT_ISVALIDFLOATV(max);
++
++ const BoolV c = BAnd(FIsGrtrOrEq(a, min), FIsGrtrOrEq(max, a));
++ return PxU32(BAllEqTTTT(c));
++ }
++
++ PX_FORCE_INLINE PxU32 FOutOfBounds(const FloatV a, const FloatV bounds)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(bounds);
++ const uint32x2_t greater = vcagt_f32(a, bounds);
++ return vget_lane_u32(greater, 0);
++ }
++
++ PX_FORCE_INLINE PxU32 FInBounds(const FloatV a, const FloatV bounds)
++ {
++ ASSERT_ISVALIDFLOATV(a);
++ ASSERT_ISVALIDFLOATV(bounds);
++ const uint32x2_t geq = vcage_f32(bounds, a);
++ return vget_lane_u32(geq, 0);
++ }
++
++ //////////////////////////////////
++ // VEC3V
++ //////////////////////////////////
++
++ PX_FORCE_INLINE Vec3V V3Splat(const FloatV f)
++ {
++ ASSERT_ISVALIDFLOATV(f);
++
++ const uint32x2_t mask = { 0xffffFFFF, 0x0 };
++ const uint32x2_t uHigh = vreinterpret_u32_f32(f);
++ const float32x2_t dHigh = vreinterpret_f32_u32(vand_u32(uHigh, mask));
++
++ return vcombine_f32(f, dHigh);
++ }
++
++ PX_FORCE_INLINE Vec3V V3Merge(const FloatVArg x, const FloatVArg y, const FloatVArg z)
++ {
++ ASSERT_ISVALIDFLOATV(x);
++ ASSERT_ISVALIDFLOATV(y);
++ ASSERT_ISVALIDFLOATV(z);
++
++ const uint32x2_t mask = { 0xffffFFFF, 0x0 };
++ const uint32x2_t dHigh = vand_u32(vreinterpret_u32_f32(z), mask);
++ const uint32x2_t dLow = vext_u32(vreinterpret_u32_f32(x), vreinterpret_u32_f32(y), 1);
++ return vreinterpretq_f32_u32(vcombine_u32(dLow, dHigh));
++ }
++
++ PX_FORCE_INLINE Vec3V V3UnitX()
++ {
++ const float32x4_t x = { 1.0f, 0.0f, 0.0f, 0.0f };
++ return x;
++ }
++
++ PX_FORCE_INLINE Vec3V V3UnitY()
++ {
++ const float32x4_t y = { 0, 1.0f, 0, 0 };
++ return y;
++ }
++
++ PX_FORCE_INLINE Vec3V V3UnitZ()
++ {
++ const float32x4_t z = { 0, 0, 1.0f, 0 };
++ return z;
++ }
++
++ PX_FORCE_INLINE FloatV V3GetX(const Vec3V f)
++ {
++ ASSERT_ISVALIDVEC3V(f);
++ const float32x2_t fLow = vget_low_f32(f);
++ return vdup_lane_f32(fLow, 0);
++ }
++
++ PX_FORCE_INLINE FloatV V3GetY(const Vec3V f)
++ {
++ ASSERT_ISVALIDVEC3V(f);
++ const float32x2_t fLow = vget_low_f32(f);
++ return vdup_lane_f32(fLow, 1);
++ }
++
++ PX_FORCE_INLINE FloatV V3GetZ(const Vec3V f)
++ {
++ ASSERT_ISVALIDVEC3V(f);
++ const float32x2_t fhigh = vget_high_f32(f);
++ return vdup_lane_f32(fhigh, 0);
++ }
++
++ PX_FORCE_INLINE Vec3V V3SetX(const Vec3V v, const FloatV f)
++ {
++ ASSERT_ISVALIDVEC3V(v);
++ ASSERT_ISVALIDFLOATV(f);
++ return V4Sel(BFTTT(), v, vcombine_f32(f, f));
++ }
++
++ PX_FORCE_INLINE Vec3V V3SetY(const Vec3V v, const FloatV f)
++ {
++ ASSERT_ISVALIDVEC3V(v);
++ ASSERT_ISVALIDFLOATV(f);
++ return V4Sel(BTFTT(), v, vcombine_f32(f, f));
++ }
++
++ PX_FORCE_INLINE Vec3V V3SetZ(const Vec3V v, const FloatV f)
++ {
++ ASSERT_ISVALIDVEC3V(v);
++ ASSERT_ISVALIDFLOATV(f);
++ return V4Sel(BTTFT(), v, vcombine_f32(f, f));
++ }
++
++ PX_FORCE_INLINE Vec3V V3ColX(const Vec3V a, const Vec3V b, const Vec3V c)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ ASSERT_ISVALIDVEC3V(c);
++
++ const float32x2_t aLow = vget_low_f32(a);
++ const float32x2_t bLow = vget_low_f32(b);
++ const float32x2_t cLow = vget_low_f32(c);
++ const float32x2_t zero = vdup_n_f32(0.0f);
++
++ const float32x2x2_t zipL = vzip_f32(aLow, bLow);
++ const float32x2x2_t zipH = vzip_f32(cLow, zero);
++
++ return vcombine_f32(zipL.val[0], zipH.val[0]);
++ }
++
++ PX_FORCE_INLINE Vec3V V3ColY(const Vec3V a, const Vec3V b, const Vec3V c)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ ASSERT_ISVALIDVEC3V(c);
++
++ const float32x2_t aLow = vget_low_f32(a);
++ const float32x2_t bLow = vget_low_f32(b);
++ const float32x2_t cLow = vget_low_f32(c);
++ const float32x2_t zero = vdup_n_f32(0.0f);
++
++ const float32x2x2_t zipL = vzip_f32(aLow, bLow);
++ const float32x2x2_t zipH = vzip_f32(cLow, zero);
++
++ return vcombine_f32(zipL.val[1], zipH.val[1]);
++ }
++
++ PX_FORCE_INLINE Vec3V V3ColZ(const Vec3V a, const Vec3V b, const Vec3V c)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ ASSERT_ISVALIDVEC3V(c);
++
++ const float32x2_t aHi = vget_high_f32(a);
++ const float32x2_t bHi = vget_high_f32(b);
++ const float32x2_t cHi = vget_high_f32(c);
++
++ const float32x2x2_t zipL = vzip_f32(aHi, bHi);
++
++ return vcombine_f32(zipL.val[0], cHi);
++ }
++
++ PX_FORCE_INLINE Vec3V V3Zero()
++ {
++ return vdupq_n_f32(0.0f);
++ }
++
++ PX_FORCE_INLINE Vec3V V3Eps()
++ {
++ return V3Load(PX_EPS_REAL);
++ }
++
++ PX_FORCE_INLINE Vec3V V3One()
++ {
++ return V3Load(1.0f);
++ }
++
++ PX_FORCE_INLINE Vec3V V3Neg(const Vec3V f)
++ {
++ ASSERT_ISVALIDVEC3V(f);
++ const float32x4_t tmp = vnegq_f32(f);
++ return vsetq_lane_f32(0.0f, tmp, 3);
++ }
++
++ PX_FORCE_INLINE Vec3V V3Add(const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ return vaddq_f32(a, b);
++ }
++
++ PX_FORCE_INLINE Vec3V V3Add(const Vec3V a, const FloatV b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDFLOATV(b);
++ return vaddq_f32(a, Vec3V_From_FloatV(b));
++ }
++
++ PX_FORCE_INLINE Vec3V V3Sub(const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ return vsubq_f32(a, b);
++ }
++
++ PX_FORCE_INLINE Vec3V V3Sub(const Vec3V a, const FloatV b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDFLOATV(b);
++ return vsubq_f32(a, Vec3V_From_FloatV(b));
++ }
++
++ PX_FORCE_INLINE Vec3V V3Scale(const Vec3V a, const FloatV b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDFLOATV(b);
++ const float32x4_t tmp = vmulq_lane_f32(a, b, 0);
++ return vsetq_lane_f32(0.0f, tmp, 3);
++ }
++
++ PX_FORCE_INLINE Vec3V V3Mul(const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ return vmulq_f32(a, b);
++ }
++
++ PX_FORCE_INLINE Vec3V V3ScaleInv(const Vec3V a, const FloatV b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDFLOATV(b);
++ const float32x2_t invB = VRECIP(b);
++ const float32x4_t tmp = vmulq_lane_f32(a, invB, 0);
++ return vsetq_lane_f32(0.0f, tmp, 3);
++ }
++
++ PX_FORCE_INLINE Vec3V V3Div(const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ float32x4_t invB = VRECIPQ(b);
++ invB = vsetq_lane_f32(0.0f, invB, 3);
++ return vmulq_f32(a, invB);
++ }
++
++ PX_FORCE_INLINE Vec3V V3ScaleInvFast(const Vec3V a, const FloatV b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDFLOATV(b);
++ const float32x2_t invB = VRECIPE(b);
++ const float32x4_t tmp = vmulq_lane_f32(a, invB, 0);
++ return vsetq_lane_f32(0.0f, tmp, 3);
++ }
++
++ PX_FORCE_INLINE Vec3V V3DivFast(const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ float32x4_t invB = VRECIPEQ(b);
++ invB = vsetq_lane_f32(0.0f, invB, 3);
++ return vmulq_f32(a, invB);
++ }
++
++ PX_FORCE_INLINE Vec3V V3Recip(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ const float32x4_t recipA = VRECIPQ(a);
++ return vsetq_lane_f32(0.0f, recipA, 3);
++ }
++
++ PX_FORCE_INLINE Vec3V V3RecipFast(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ const float32x4_t recipA = VRECIPEQ(a);
++ return vsetq_lane_f32(0.0f, recipA, 3);
++ }
++
++ PX_FORCE_INLINE Vec3V V3Rsqrt(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ const float32x4_t rSqrA = VRECIPSQRTQ(a);
++ return vsetq_lane_f32(0.0f, rSqrA, 3);
++ }
++
++ PX_FORCE_INLINE Vec3V V3RsqrtFast(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ const float32x4_t rSqrA = VRECIPSQRTEQ(a);
++ return vsetq_lane_f32(0.0f, rSqrA, 3);
++ }
++
++ PX_FORCE_INLINE Vec3V V3ScaleAdd(const Vec3V a, const FloatV b, const Vec3V c)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDFLOATV(b);
++ ASSERT_ISVALIDVEC3V(c);
++
++ float32x4_t tmp = vmlaq_lane_f32(c, a, b, 0);
++ // using vsetq_lane_f32 resulted in failures,
++ // probably related to a compiler bug on
++ // ndk r9d-win32, gcc 4.8, cardhu/shield
++
++ // code with issue
++ // return vsetq_lane_f32(0.0f, tmp, 3);
++
++ // workaround
++ float32x2_t w_z = vget_high_f32(tmp);
++ float32x2_t y_x = vget_low_f32(tmp);
++ w_z = vset_lane_f32(0.0f, w_z, 1);
++ return vcombine_f32(y_x, w_z);
++ }
++
++ PX_FORCE_INLINE Vec3V V3NegScaleSub(const Vec3V a, const FloatV b, const Vec3V c)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDFLOATV(b);
++ ASSERT_ISVALIDVEC3V(c);
++
++ float32x4_t tmp = vmlsq_lane_f32(c, a, b, 0);
++ // using vsetq_lane_f32 resulted in failures,
++ // probably related to a compiler bug on
++ // ndk r9d-win32, gcc 4.8, cardhu/shield
++
++ // code with issue
++ // return vsetq_lane_f32(0.0f, tmp, 3);
++
++ // workaround
++ float32x2_t w_z = vget_high_f32(tmp);
++ float32x2_t y_x = vget_low_f32(tmp);
++ w_z = vset_lane_f32(0.0f, w_z, 1);
++ return vcombine_f32(y_x, w_z);
++ }
++
++ PX_FORCE_INLINE Vec3V V3MulAdd(const Vec3V a, const Vec3V b, const Vec3V c)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ ASSERT_ISVALIDVEC3V(c);
++ return vmlaq_f32(c, a, b);
++ }
++
++ PX_FORCE_INLINE Vec3V V3NegMulSub(const Vec3V a, const Vec3V b, const Vec3V c)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ ASSERT_ISVALIDVEC3V(c);
++ return vmlsq_f32(c, a, b);
++ }
++
++ PX_FORCE_INLINE Vec3V V3Abs(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ return vabsq_f32(a);
++ }
++
++ PX_FORCE_INLINE FloatV V3Dot(const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++
++ // const uint32x2_t mask = {0xffffFFFF, 0x0};
++ const float32x4_t tmp = vmulq_f32(a, b);
++
++ const float32x2_t low = vget_low_f32(tmp);
++ const float32x2_t high = vget_high_f32(tmp);
++ // const float32x2_t high = vreinterpret_f32_u32(vand_u32(vreinterpret_u32_f32(high_), mask));
++
++ const float32x2_t sumTmp = vpadd_f32(low, high); // = {0+z, x+y}
++ const float32x2_t sum0ZYX = vpadd_f32(sumTmp, sumTmp); // = {x+y+z, x+y+z}
++
++ return sum0ZYX;
++ }
++
++ PX_FORCE_INLINE Vec3V V3Cross(const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++
++ const uint32x2_t TF = { 0xffffFFFF, 0x0 };
++ const float32x2_t ay_ax = vget_low_f32(a); // d2
++ const float32x2_t aw_az = vget_high_f32(a); // d3
++ const float32x2_t by_bx = vget_low_f32(b); // d4
++ const float32x2_t bw_bz = vget_high_f32(b); // d5
++ // Hi, Lo
++ const float32x2_t bz_by = vext_f32(by_bx, bw_bz, 1); // bz, by
++ const float32x2_t az_ay = vext_f32(ay_ax, aw_az, 1); // az, ay
++
++ const float32x2_t azbx = vmul_f32(aw_az, by_bx); // 0, az*bx
++ const float32x2_t aybz_axby = vmul_f32(ay_ax, bz_by); // ay*bz, ax*by
++
++ const float32x2_t azbxSUBaxbz = vmls_f32(azbx, bw_bz, ay_ax); // 0, az*bx-ax*bz
++ const float32x2_t aybzSUBazby_axbySUBaybx = vmls_f32(aybz_axby, by_bx, az_ay); // ay*bz-az*by, ax*by-ay*bx
++
++ const float32x2_t retLow = vext_f32(aybzSUBazby_axbySUBaybx, azbxSUBaxbz, 1); // az*bx-ax*bz, ay*bz-az*by
++ const uint32x2_t retHigh = vand_u32(TF, vreinterpret_u32_f32(aybzSUBazby_axbySUBaybx)); // 0, ax*by-ay*bx
++
++ return vcombine_f32(retLow, vreinterpret_f32_u32(retHigh));
++ }
++
++ PX_FORCE_INLINE VecCrossV V3PrepareCross(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ return a;
++ }
++
++ PX_FORCE_INLINE FloatV V3Length(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++
++ // const uint32x2_t mask = {0xffffFFFF, 0x0};
++
++ const float32x4_t tmp = vmulq_f32(a, a);
++ const float32x2_t low = vget_low_f32(tmp);
++ const float32x2_t high = vget_high_f32(tmp);
++ // const float32x2_t high = vreinterpret_f32_u32(vand_u32(vreinterpret_u32_f32(high_), mask));
++
++ const float32x2_t sumTmp = vpadd_f32(low, high); // = {0+z, x+y}
++ const float32x2_t sum0ZYX = vpadd_f32(sumTmp, sumTmp); // = {x+y+z, x+y+z}
++
++ return FSqrt(sum0ZYX);
++ }
++
++ PX_FORCE_INLINE FloatV V3LengthSq(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ return V3Dot(a, a);
++ }
++
++ PX_FORCE_INLINE Vec3V V3Normalize(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ //PX_ASSERT(!FAllEq(V4LengthSq(a), FZero()));
++ return V3ScaleInv(a, V3Length(a));
++ }
++
++ PX_FORCE_INLINE Vec3V V3NormalizeFast(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ //PX_ASSERT(!FAllEq(V4LengthSq(a), FZero()));
++ return V3Scale(a, VRECIPSQRTE(V3Dot(a, a)));
++ }
++
++ PX_FORCE_INLINE Vec3V V3NormalizeSafe(const Vec3V a, const Vec3V unsafeReturnValue)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ const FloatV zero = vdup_n_f32(0.0f);
++ const FloatV length = V3Length(a);
++ const uint32x4_t isGreaterThanZero = FIsGrtr(length, zero);
++ return V3Sel(isGreaterThanZero, V3ScaleInv(a, length), unsafeReturnValue);
++ }
++
++ PX_FORCE_INLINE Vec3V V3Sel(const BoolV c, const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V( vbslq_f32(c, a, b));
++ return vbslq_f32(c, a, b);
++ }
++
++ PX_FORCE_INLINE BoolV V3IsGrtr(const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ return vcgtq_f32(a, b);
++ }
++
++ PX_FORCE_INLINE BoolV V3IsGrtrOrEq(const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ return vcgeq_f32(a, b);
++ }
++
++ PX_FORCE_INLINE BoolV V3IsEq(const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ return vceqq_f32(a, b);
++ }
++
++ PX_FORCE_INLINE Vec3V V3Max(const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ return vmaxq_f32(a, b);
++ }
++
++ PX_FORCE_INLINE Vec3V V3Min(const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ return vminq_f32(a, b);
++ }
++
++ PX_FORCE_INLINE FloatV V3ExtractMax(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++
++ const float32x2_t low = vget_low_f32(a);
++ const float32x2_t high = vget_high_f32(a);
++
++ const float32x2_t zz = vdup_lane_f32(high, 0);
++ const float32x2_t max0 = vpmax_f32(zz, low);
++ const float32x2_t max1 = vpmax_f32(max0, max0);
++
++ return max1;
++ }
++
++ PX_FORCE_INLINE FloatV V3ExtractMin(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++
++ const float32x2_t low = vget_low_f32(a);
++ const float32x2_t high = vget_high_f32(a);
++
++ const float32x2_t zz = vdup_lane_f32(high, 0);
++ const float32x2_t min0 = vpmin_f32(zz, low);
++ const float32x2_t min1 = vpmin_f32(min0, min0);
++
++ return min1;
++ }
++
++ // return (a >= 0.0f) ? 1.0f : -1.0f;
++ PX_FORCE_INLINE Vec3V V3Sign(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ const Vec3V zero = V3Zero();
++ const Vec3V one = V3One();
++ const Vec3V none = V3Neg(one);
++ return V3Sel(V3IsGrtrOrEq(a, zero), one, none);
++ }
++
++ PX_FORCE_INLINE Vec3V V3Clamp(const Vec3V a, const Vec3V minV, const Vec3V maxV)
++ {
++ ASSERT_ISVALIDVEC3V(minV);
++ ASSERT_ISVALIDVEC3V(maxV);
++ return V3Max(V3Min(a, maxV), minV);
++ }
++
++ PX_FORCE_INLINE PxU32 V3AllGrtr(const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ return internalUnitNeonSimd::BAllTrue3_R(V4IsGrtr(a, b));
++ }
++
++ PX_FORCE_INLINE PxU32 V3AllGrtrOrEq(const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ return internalUnitNeonSimd::BAllTrue3_R(V4IsGrtrOrEq(a, b));
++ }
++
++ PX_FORCE_INLINE PxU32 V3AllEq(const Vec3V a, const Vec3V b)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(b);
++ return internalUnitNeonSimd::BAllTrue3_R(V4IsEq(a, b));
++ }
++
++ PX_FORCE_INLINE Vec3V V3Round(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ // truncate(a + (0.5f - sign(a)))
++ const Vec3V half = V3Load(0.5f);
++ const float32x4_t sign = vcvtq_f32_u32((vshrq_n_u32(vreinterpretq_u32_f32(a), 31)));
++ const Vec3V aPlusHalf = V3Add(a, half);
++ const Vec3V aRound = V3Sub(aPlusHalf, sign);
++ return vcvtq_f32_s32(vcvtq_s32_f32(aRound));
++ }
++
++ PX_FORCE_INLINE Vec3V V3Sin(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++
++ // Modulo the range of the given angles such that -XM_2PI <= Angles < XM_2PI
++ const Vec4V recipTwoPi = V4LoadA(g_PXReciprocalTwoPi.f);
++ const Vec4V twoPi = V4LoadA(g_PXTwoPi.f);
++ const Vec3V tmp = V4Mul(a, recipTwoPi);
++ const Vec3V b = V3Round(tmp);
++ const Vec3V V1 = V4NegMulSub(twoPi, b, a);
++
++ // sin(V) ~= V - V^3 / 3! + V^5 / 5! - V^7 / 7! + V^9 / 9! - V^11 / 11! + V^13 / 13! -
++ // V^15 / 15! + V^17 / 17! - V^19 / 19! + V^21 / 21! - V^23 / 23! (for -PI <= V < PI)
++ const Vec3V V2 = V3Mul(V1, V1);
++ const Vec3V V3 = V3Mul(V2, V1);
++ const Vec3V V5 = V3Mul(V3, V2);
++ const Vec3V V7 = V3Mul(V5, V2);
++ const Vec3V V9 = V3Mul(V7, V2);
++ const Vec3V V11 = V3Mul(V9, V2);
++ const Vec3V V13 = V3Mul(V11, V2);
++ const Vec3V V15 = V3Mul(V13, V2);
++ const Vec3V V17 = V3Mul(V15, V2);
++ const Vec3V V19 = V3Mul(V17, V2);
++ const Vec3V V21 = V3Mul(V19, V2);
++ const Vec3V V23 = V3Mul(V21, V2);
++
++ const Vec4V sinCoefficients0 = V4LoadA(g_PXSinCoefficients0.f);
++ const Vec4V sinCoefficients1 = V4LoadA(g_PXSinCoefficients1.f);
++ const Vec4V sinCoefficients2 = V4LoadA(g_PXSinCoefficients2.f);
++
++ const FloatV S1 = V4GetY(sinCoefficients0);
++ const FloatV S2 = V4GetZ(sinCoefficients0);
++ const FloatV S3 = V4GetW(sinCoefficients0);
++ const FloatV S4 = V4GetX(sinCoefficients1);
++ const FloatV S5 = V4GetY(sinCoefficients1);
++ const FloatV S6 = V4GetZ(sinCoefficients1);
++ const FloatV S7 = V4GetW(sinCoefficients1);
++ const FloatV S8 = V4GetX(sinCoefficients2);
++ const FloatV S9 = V4GetY(sinCoefficients2);
++ const FloatV S10 = V4GetZ(sinCoefficients2);
++ const FloatV S11 = V4GetW(sinCoefficients2);
++
++ Vec3V Result;
++ Result = V4ScaleAdd(V3, S1, V1);
++ Result = V4ScaleAdd(V5, S2, Result);
++ Result = V4ScaleAdd(V7, S3, Result);
++ Result = V4ScaleAdd(V9, S4, Result);
++ Result = V4ScaleAdd(V11, S5, Result);
++ Result = V4ScaleAdd(V13, S6, Result);
++ Result = V4ScaleAdd(V15, S7, Result);
++ Result = V4ScaleAdd(V17, S8, Result);
++ Result = V4ScaleAdd(V19, S9, Result);
++ Result = V4ScaleAdd(V21, S10, Result);
++ Result = V4ScaleAdd(V23, S11, Result);
++
++ return Result;
++ }
++
++ PX_FORCE_INLINE Vec3V V3Cos(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++
++ // Modulo the range of the given angles such that -XM_2PI <= Angles < XM_2PI
++ const Vec4V recipTwoPi = V4LoadA(g_PXReciprocalTwoPi.f);
++ const Vec4V twoPi = V4LoadA(g_PXTwoPi.f);
++ const Vec3V tmp = V4Mul(a, recipTwoPi);
++ const Vec3V b = V3Round(tmp);
++ const Vec3V V1 = V4NegMulSub(twoPi, b, a);
++
++ // cos(V) ~= 1 - V^2 / 2! + V^4 / 4! - V^6 / 6! + V^8 / 8! - V^10 / 10! + V^12 / 12! -
++ // V^14 / 14! + V^16 / 16! - V^18 / 18! + V^20 / 20! - V^22 / 22! (for -PI <= V < PI)
++ const Vec3V V2 = V3Mul(V1, V1);
++ const Vec3V V4 = V3Mul(V2, V2);
++ const Vec3V V6 = V3Mul(V4, V2);
++ const Vec3V V8 = V3Mul(V4, V4);
++ const Vec3V V10 = V3Mul(V6, V4);
++ const Vec3V V12 = V3Mul(V6, V6);
++ const Vec3V V14 = V3Mul(V8, V6);
++ const Vec3V V16 = V3Mul(V8, V8);
++ const Vec3V V18 = V3Mul(V10, V8);
++ const Vec3V V20 = V3Mul(V10, V10);
++ const Vec3V V22 = V3Mul(V12, V10);
++
++ const Vec4V cosCoefficients0 = V4LoadA(g_PXCosCoefficients0.f);
++ const Vec4V cosCoefficients1 = V4LoadA(g_PXCosCoefficients1.f);
++ const Vec4V cosCoefficients2 = V4LoadA(g_PXCosCoefficients2.f);
++
++ const FloatV C1 = V4GetY(cosCoefficients0);
++ const FloatV C2 = V4GetZ(cosCoefficients0);
++ const FloatV C3 = V4GetW(cosCoefficients0);
++ const FloatV C4 = V4GetX(cosCoefficients1);
++ const FloatV C5 = V4GetY(cosCoefficients1);
++ const FloatV C6 = V4GetZ(cosCoefficients1);
++ const FloatV C7 = V4GetW(cosCoefficients1);
++ const FloatV C8 = V4GetX(cosCoefficients2);
++ const FloatV C9 = V4GetY(cosCoefficients2);
++ const FloatV C10 = V4GetZ(cosCoefficients2);
++ const FloatV C11 = V4GetW(cosCoefficients2);
++
++ Vec3V Result;
++ Result = V4ScaleAdd(V2, C1, V4One());
++ Result = V4ScaleAdd(V4, C2, Result);
++ Result = V4ScaleAdd(V6, C3, Result);
++ Result = V4ScaleAdd(V8, C4, Result);
++ Result = V4ScaleAdd(V10, C5, Result);
++ Result = V4ScaleAdd(V12, C6, Result);
++ Result = V4ScaleAdd(V14, C7, Result);
++ Result = V4ScaleAdd(V16, C8, Result);
++ Result = V4ScaleAdd(V18, C9, Result);
++ Result = V4ScaleAdd(V20, C10, Result);
++ Result = V4ScaleAdd(V22, C11, Result);
++
++ return V4ClearW(Result);
++ }
++
++ PX_FORCE_INLINE Vec3V V3PermYZZ(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ const float32x2_t xy = vget_low_f32(a);
++ const float32x2_t zw = vget_high_f32(a);
++ const float32x2_t yz = vext_f32(xy, zw, 1);
++ return vcombine_f32(yz, zw);
++ }
++
++ PX_FORCE_INLINE Vec3V V3PermXYX(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++
++ const uint32x2_t mask = { 0xffffFFFF, 0x0 };
++ const uint32x2_t xy = vget_low_u32(vreinterpretq_u32_f32(a));
++ const uint32x2_t xw = vand_u32(xy, mask);
++ return vreinterpretq_f32_u32(vcombine_u32(xy, xw));
++ }
++
++ PX_FORCE_INLINE Vec3V V3PermYZX(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++
++ const uint32x2_t mask = { 0xffffFFFF, 0x0 };
++ const uint32x2_t xy = vget_low_u32(vreinterpretq_u32_f32(a));
++ const uint32x2_t zw = vget_high_u32(vreinterpretq_u32_f32(a));
++ const uint32x2_t yz = vext_u32(xy, zw, 1);
++ const uint32x2_t xw = vand_u32(xy, mask);
++ return vreinterpretq_f32_u32(vcombine_u32(yz, xw));
++ }
++
++ PX_FORCE_INLINE Vec3V V3PermZXY(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++
++ const uint32x2_t xy = vget_low_u32(vreinterpretq_u32_f32(a));
++ const uint32x2_t zw = vget_high_u32(vreinterpretq_u32_f32(a));
++ const uint32x2_t wz = vrev64_u32(zw);
++
++ const uint32x2_t zx = vext_u32(wz, xy, 1);
++ const uint32x2_t yw = vext_u32(xy, wz, 1);
++
++ return vreinterpretq_f32_u32(vcombine_u32(zx, yw));
++ }
++
++ PX_FORCE_INLINE Vec3V V3PermZZY(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++
++ const uint32x2_t xy = vget_low_u32(vreinterpretq_u32_f32(a));
++ const uint32x2_t zw = vget_high_u32(vreinterpretq_u32_f32(a));
++
++ const uint32x2_t wz = vrev64_u32(zw);
++ const uint32x2_t yw = vext_u32(xy, wz, 1);
++ const uint32x2_t zz = vdup_lane_u32(wz, 1);
++
++ return vreinterpretq_f32_u32(vcombine_u32(zz, yw));
++ }
++
++ PX_FORCE_INLINE Vec3V V3PermYXX(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++
++ const uint32x2_t mask = { 0xffffFFFF, 0x0 };
++ const uint32x2_t xy = vget_low_u32(vreinterpretq_u32_f32(a));
++ const uint32x2_t yx = vrev64_u32(xy);
++ const uint32x2_t xw = vand_u32(xy, mask);
++ return vreinterpretq_f32_u32(vcombine_u32(yx, xw));
++ }
++
++ PX_FORCE_INLINE Vec3V V3Perm_Zero_1Z_0Y(const Vec3V v0, const Vec3V v1)
++ {
++ ASSERT_ISVALIDVEC3V(v0);
++ ASSERT_ISVALIDVEC3V(v1);
++
++ const uint32x2_t xy = vget_low_u32(vreinterpretq_u32_f32(v0));
++ const uint32x2_t zw = vget_high_u32(vreinterpretq_u32_f32(v1));
++ const uint32x2_t wz = vrev64_u32(zw);
++ const uint32x2_t yw = vext_u32(xy, wz, 1);
++
++ return vreinterpretq_f32_u32(vcombine_u32(wz, yw));
++ }
++
++ PX_FORCE_INLINE Vec3V V3Perm_0Z_Zero_1X(const Vec3V v0, const Vec3V v1)
++ {
++ ASSERT_ISVALIDVEC3V(v0);
++ ASSERT_ISVALIDVEC3V(v1);
++
++ const uint32x2_t mask = { 0xffffFFFF, 0x0 };
++ const uint32x2_t zw = vget_high_u32(vreinterpretq_u32_f32(v0));
++ const uint32x2_t xy = vget_low_u32(vreinterpretq_u32_f32(v1));
++ const uint32x2_t xw = vand_u32(xy, mask);
++
++ return vreinterpretq_f32_u32(vcombine_u32(zw, xw));
++ }
++
++ PX_FORCE_INLINE Vec3V V3Perm_1Y_0X_Zero(const Vec3V v0, const Vec3V v1)
++ {
++ ASSERT_ISVALIDVEC3V(v0);
++ ASSERT_ISVALIDVEC3V(v1);
++
++ const uint32x2_t axy = vget_low_u32(vreinterpretq_u32_f32(v0));
++ const uint32x2_t bxy = vget_low_u32(vreinterpretq_u32_f32(v1));
++ const uint32x2_t byax = vext_u32(bxy, axy, 1);
++ const uint32x2_t ww = vdup_n_u32(0);
++
++ return vreinterpretq_f32_u32(vcombine_u32(byax, ww));
++ }
++
++ PX_FORCE_INLINE FloatV V3SumElems(const Vec3V a)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++
++ // const uint32x2_t mask = {0xffffFFFF, 0x0};
++
++ const float32x2_t low = vget_low_f32(a);
++ const float32x2_t high = vget_high_f32(a);
++ // const float32x2_t high = vreinterpret_f32_u32(vand_u32(vreinterpret_u32_f32(high_), mask));
++
++ const float32x2_t sumTmp = vpadd_f32(low, high); // = {0+z, x+y}
++ const float32x2_t sum0ZYX = vpadd_f32(sumTmp, sumTmp); // = {x+y+z, x+y+z}
++
++ return sum0ZYX;
++ }
++
++ PX_FORCE_INLINE PxU32 V3OutOfBounds(const Vec3V a, const Vec3V min, const Vec3V max)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(min);
++ ASSERT_ISVALIDVEC3V(max);
++
++ const BoolV c = BOr(V3IsGrtr(a, max), V3IsGrtr(min, a));
++ return internalUnitNeonSimd::BAnyTrue3_R(c);
++ }
++
++ PX_FORCE_INLINE PxU32 V3InBounds(const Vec3V a, const Vec3V min, const Vec3V max)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(min);
++ ASSERT_ISVALIDVEC3V(max);
++
++ const BoolV c = BAnd(V3IsGrtrOrEq(a, min), V3IsGrtrOrEq(max, a));
++ return internalUnitNeonSimd::BAllTrue4_R(c);
++ }
++
++ PX_FORCE_INLINE PxU32 V3OutOfBounds(const Vec3V a, const Vec3V bounds)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(bounds);
++
++ const BoolV greater = V3IsGrtr(V3Abs(a), bounds);
++ return internalUnitNeonSimd::BAnyTrue3_R(greater);
++ }
++
++ PX_FORCE_INLINE PxU32 V3InBounds(const Vec3V a, const Vec3V bounds)
++ {
++ ASSERT_ISVALIDVEC3V(a);
++ ASSERT_ISVALIDVEC3V(bounds);
++
++ const BoolV greaterOrEq = V3IsGrtrOrEq(bounds, V3Abs(a));
++ return internalUnitNeonSimd::BAllTrue4_R(greaterOrEq);
++ }
++
++ PX_FORCE_INLINE void V3Transpose(Vec3V& col0, Vec3V& col1, Vec3V& col2)
++ {
++ ASSERT_ISVALIDVEC3V(col0);
++ ASSERT_ISVALIDVEC3V(col1);
++ ASSERT_ISVALIDVEC3V(col2);
++
++ Vec3V col3 = V3Zero();
++ const float32x4x2_t v0v1 = vzipq_f32(col0, col2);
++ const float32x4x2_t v2v3 = vzipq_f32(col1, col3);
++ const float32x4x2_t zip0 = vzipq_f32(v0v1.val[0], v2v3.val[0]);
++ const float32x4x2_t zip1 = vzipq_f32(v0v1.val[1], v2v3.val[1]);
++ col0 = zip0.val[0];
++ col1 = zip0.val[1];
++ col2 = zip1.val[0];
++ // col3 = zip1.val[1];
++ }
++
++ //////////////////////////////////
++ // VEC4V
++ //////////////////////////////////
++
++ PX_FORCE_INLINE Vec4V V4Splat(const FloatV f)
++ {
++ ASSERT_ISVALIDFLOATV(f);
++ return vcombine_f32(f, f);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Merge(const FloatV* const floatVArray)
++ {
++ ASSERT_ISVALIDFLOATV(floatVArray[0]);
++ ASSERT_ISVALIDFLOATV(floatVArray[1]);
++ ASSERT_ISVALIDFLOATV(floatVArray[2]);
++ ASSERT_ISVALIDFLOATV(floatVArray[3]);
++
++ const uint32x2_t xLow = vreinterpret_u32_f32(floatVArray[0]);
++ const uint32x2_t yLow = vreinterpret_u32_f32(floatVArray[1]);
++ const uint32x2_t zLow = vreinterpret_u32_f32(floatVArray[2]);
++ const uint32x2_t wLow = vreinterpret_u32_f32(floatVArray[3]);
++
++ const uint32x2_t dLow = vext_u32(xLow, yLow, 1);
++ const uint32x2_t dHigh = vext_u32(zLow, wLow, 1);
++
++ return vreinterpretq_f32_u32(vcombine_u32(dLow, dHigh));
++ }
++
++ PX_FORCE_INLINE Vec4V V4Merge(const FloatVArg x, const FloatVArg y, const FloatVArg z, const FloatVArg w)
++ {
++ ASSERT_ISVALIDFLOATV(x);
++ ASSERT_ISVALIDFLOATV(y);
++ ASSERT_ISVALIDFLOATV(z);
++ ASSERT_ISVALIDFLOATV(w);
++
++ const uint32x2_t xLow = vreinterpret_u32_f32(x);
++ const uint32x2_t yLow = vreinterpret_u32_f32(y);
++ const uint32x2_t zLow = vreinterpret_u32_f32(z);
++ const uint32x2_t wLow = vreinterpret_u32_f32(w);
++
++ const uint32x2_t dLow = vext_u32(xLow, yLow, 1);
++ const uint32x2_t dHigh = vext_u32(zLow, wLow, 1);
++
++ return vreinterpretq_f32_u32(vcombine_u32(dLow, dHigh));
++ }
++
++ PX_FORCE_INLINE Vec4V V4MergeW(const Vec4VArg x, const Vec4VArg y, const Vec4VArg z, const Vec4VArg w)
++ {
++ const float32x2_t xx = vget_high_f32(x);
++ const float32x2_t yy = vget_high_f32(y);
++ const float32x2_t zz = vget_high_f32(z);
++ const float32x2_t ww = vget_high_f32(w);
++
++ const float32x2x2_t zipL = vzip_f32(xx, yy);
++ const float32x2x2_t zipH = vzip_f32(zz, ww);
++
++ return vcombine_f32(zipL.val[1], zipH.val[1]);
++ }
++
++ PX_FORCE_INLINE Vec4V V4MergeZ(const Vec4VArg x, const Vec4VArg y, const Vec4VArg z, const Vec4VArg w)
++ {
++ const float32x2_t xx = vget_high_f32(x);
++ const float32x2_t yy = vget_high_f32(y);
++ const float32x2_t zz = vget_high_f32(z);
++ const float32x2_t ww = vget_high_f32(w);
++
++ const float32x2x2_t zipL = vzip_f32(xx, yy);
++ const float32x2x2_t zipH = vzip_f32(zz, ww);
++
++ return vcombine_f32(zipL.val[0], zipH.val[0]);
++ }
++
++ PX_FORCE_INLINE Vec4V V4MergeY(const Vec4VArg x, const Vec4VArg y, const Vec4VArg z, const Vec4VArg w)
++ {
++ const float32x2_t xx = vget_low_f32(x);
++ const float32x2_t yy = vget_low_f32(y);
++ const float32x2_t zz = vget_low_f32(z);
++ const float32x2_t ww = vget_low_f32(w);
++
++ const float32x2x2_t zipL = vzip_f32(xx, yy);
++ const float32x2x2_t zipH = vzip_f32(zz, ww);
++
++ return vcombine_f32(zipL.val[1], zipH.val[1]);
++ }
++
++ PX_FORCE_INLINE Vec4V V4MergeX(const Vec4VArg x, const Vec4VArg y, const Vec4VArg z, const Vec4VArg w)
++ {
++ const float32x2_t xx = vget_low_f32(x);
++ const float32x2_t yy = vget_low_f32(y);
++ const float32x2_t zz = vget_low_f32(z);
++ const float32x2_t ww = vget_low_f32(w);
++
++ const float32x2x2_t zipL = vzip_f32(xx, yy);
++ const float32x2x2_t zipH = vzip_f32(zz, ww);
++
++ return vcombine_f32(zipL.val[0], zipH.val[0]);
++ }
++
++ PX_FORCE_INLINE Vec4V V4UnpackXY(const Vec4VArg a, const Vec4VArg b)
++ {
++ return vzipq_f32(a, b).val[0];
++ }
++
++ PX_FORCE_INLINE Vec4V V4UnpackZW(const Vec4VArg a, const Vec4VArg b)
++ {
++ return vzipq_f32(a, b).val[1];
++ }
++
++ PX_FORCE_INLINE Vec4V V4UnitW()
++ {
++ const float32x2_t zeros = vreinterpret_f32_u32(vmov_n_u32(0));
++ const float32x2_t ones = vmov_n_f32(1.0f);
++ const float32x2_t zo = vext_f32(zeros, ones, 1);
++ return vcombine_f32(zeros, zo);
++ }
++
++ PX_FORCE_INLINE Vec4V V4UnitX()
++ {
++ const float32x2_t zeros = vreinterpret_f32_u32(vmov_n_u32(0));
++ const float32x2_t ones = vmov_n_f32(1.0f);
++ const float32x2_t oz = vext_f32(ones, zeros, 1);
++ return vcombine_f32(oz, zeros);
++ }
++
++ PX_FORCE_INLINE Vec4V V4UnitY()
++ {
++ const float32x2_t zeros = vreinterpret_f32_u32(vmov_n_u32(0));
++ const float32x2_t ones = vmov_n_f32(1.0f);
++ const float32x2_t zo = vext_f32(zeros, ones, 1);
++ return vcombine_f32(zo, zeros);
++ }
++
++ PX_FORCE_INLINE Vec4V V4UnitZ()
++ {
++ const float32x2_t zeros = vreinterpret_f32_u32(vmov_n_u32(0));
++ const float32x2_t ones = vmov_n_f32(1.0f);
++ const float32x2_t oz = vext_f32(ones, zeros, 1);
++ return vcombine_f32(zeros, oz);
++ }
++
++ PX_FORCE_INLINE FloatV V4GetW(const Vec4V f)
++ {
++ const float32x2_t fhigh = vget_high_f32(f);
++ return vdup_lane_f32(fhigh, 1);
++ }
++
++ PX_FORCE_INLINE FloatV V4GetX(const Vec4V f)
++ {
++ const float32x2_t fLow = vget_low_f32(f);
++ return vdup_lane_f32(fLow, 0);
++ }
++
++ PX_FORCE_INLINE FloatV V4GetY(const Vec4V f)
++ {
++ const float32x2_t fLow = vget_low_f32(f);
++ return vdup_lane_f32(fLow, 1);
++ }
++
++ PX_FORCE_INLINE FloatV V4GetZ(const Vec4V f)
++ {
++ const float32x2_t fhigh = vget_high_f32(f);
++ return vdup_lane_f32(fhigh, 0);
++ }
++
++ PX_FORCE_INLINE Vec4V V4SetW(const Vec4V v, const FloatV f)
++ {
++ ASSERT_ISVALIDFLOATV(f);
++ return V4Sel(BTTTF(), v, vcombine_f32(f, f));
++ }
++
++ PX_FORCE_INLINE Vec4V V4SetX(const Vec4V v, const FloatV f)
++ {
++ ASSERT_ISVALIDFLOATV(f);
++ return V4Sel(BFTTT(), v, vcombine_f32(f, f));
++ }
++
++ PX_FORCE_INLINE Vec4V V4SetY(const Vec4V v, const FloatV f)
++ {
++ ASSERT_ISVALIDFLOATV(f);
++ return V4Sel(BTFTT(), v, vcombine_f32(f, f));
++ }
++
++ PX_FORCE_INLINE Vec4V V4SetZ(const Vec4V v, const FloatV f)
++ {
++ ASSERT_ISVALIDFLOATV(f);
++ return V4Sel(BTTFT(), v, vcombine_f32(f, f));
++ }
++
++ PX_FORCE_INLINE Vec4V V4ClearW(const Vec4V v)
++ {
++ return V4Sel(BTTTF(), v, V4Zero());
++ }
++
++ PX_FORCE_INLINE Vec4V V4PermYXWZ(const Vec4V a)
++ {
++ const float32x2_t xy = vget_low_f32(a);
++ const float32x2_t zw = vget_high_f32(a);
++ const float32x2_t yx = vext_f32(xy, xy, 1);
++ const float32x2_t wz = vext_f32(zw, zw, 1);
++ return vcombine_f32(yx, wz);
++ }
++
++ PX_FORCE_INLINE Vec4V V4PermXZXZ(const Vec4V a)
++ {
++ const float32x2_t xy = vget_low_f32(a);
++ const float32x2_t zw = vget_high_f32(a);
++ const float32x2x2_t xzyw = vzip_f32(xy, zw);
++ return vcombine_f32(xzyw.val[0], xzyw.val[0]);
++ }
++
++ PX_FORCE_INLINE Vec4V V4PermYWYW(const Vec4V a)
++ {
++ const float32x2_t xy = vget_low_f32(a);
++ const float32x2_t zw = vget_high_f32(a);
++ const float32x2x2_t xzyw = vzip_f32(xy, zw);
++ return vcombine_f32(xzyw.val[1], xzyw.val[1]);
++ }
++
++ PX_FORCE_INLINE Vec4V V4PermYZXW(const Vec4V a)
++ {
++ const uint32x2_t xy = vget_low_u32(vreinterpretq_u32_f32(a));
++ const uint32x2_t zw = vget_high_u32(vreinterpretq_u32_f32(a));
++ const uint32x2_t yz = vext_u32(xy, zw, 1);
++ const uint32x2_t xw = vrev64_u32(vext_u32(zw, xy, 1));
++ return vreinterpretq_f32_u32(vcombine_u32(yz, xw));
++ }
++
++ PX_FORCE_INLINE Vec4V V4PermZWXY(const Vec4V a)
++ {
++ const float32x2_t low = vget_low_f32(a);
++ const float32x2_t high = vget_high_f32(a);
++ return vcombine_f32(high, low);
++ }
++
++ template <PxU8 E0, PxU8 E1, PxU8 E2, PxU8 E3>
++ PX_FORCE_INLINE Vec4V V4Perm(const Vec4V V)
++ {
++ static const uint32_t ControlElement[4] =
++ {
++ #if 1
++ 0x03020100, // XM_SWIZZLE_X
++ 0x07060504, // XM_SWIZZLE_Y
++ 0x0B0A0908, // XM_SWIZZLE_Z
++ 0x0F0E0D0C, // XM_SWIZZLE_W
++ #else
++ 0x00010203, // XM_SWIZZLE_X
++ 0x04050607, // XM_SWIZZLE_Y
++ 0x08090A0B, // XM_SWIZZLE_Z
++ 0x0C0D0E0F, // XM_SWIZZLE_W
++ #endif
++ };
++
++ uint8x8x2_t tbl;
++ tbl.val[0] = vreinterpret_u8_f32(vget_low_f32(V));
++ tbl.val[1] = vreinterpret_u8_f32(vget_high_f32(V));
++
++ uint8x8_t idx =
++ vcreate_u8(static_cast<uint64_t>(ControlElement[E0]) | (static_cast<uint64_t>(ControlElement[E1]) << 32));
++ const uint8x8_t rL = vtbl2_u8(tbl, idx);
++ idx = vcreate_u8(static_cast<uint64_t>(ControlElement[E2]) | (static_cast<uint64_t>(ControlElement[E3]) << 32));
++ const uint8x8_t rH = vtbl2_u8(tbl, idx);
++ return vreinterpretq_f32_u8(vcombine_u8(rL, rH));
++ }
++
++ // PT: this seems measurably slower than the hardcoded version
++ /*PX_FORCE_INLINE Vec4V V4PermYZXW(const Vec4V a)
++ {
++ return V4Perm<1, 2, 0, 3>(a);
++ }*/
++
++ PX_FORCE_INLINE Vec4V V4Zero()
++ {
++ return vreinterpretq_f32_u32(vmovq_n_u32(0));
++ // return vmovq_n_f32(0.0f);
++ }
++
++ PX_FORCE_INLINE Vec4V V4One()
++ {
++ return vmovq_n_f32(1.0f);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Eps()
++ {
++ // return vmovq_n_f32(PX_EPS_REAL);
++ return V4Load(PX_EPS_REAL);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Neg(const Vec4V f)
++ {
++ return vnegq_f32(f);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Add(const Vec4V a, const Vec4V b)
++ {
++ return vaddq_f32(a, b);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Sub(const Vec4V a, const Vec4V b)
++ {
++ return vsubq_f32(a, b);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Scale(const Vec4V a, const FloatV b)
++ {
++ return vmulq_lane_f32(a, b, 0);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Mul(const Vec4V a, const Vec4V b)
++ {
++ return vmulq_f32(a, b);
++ }
++
++ PX_FORCE_INLINE Vec4V V4ScaleInv(const Vec4V a, const FloatV b)
++ {
++ ASSERT_ISVALIDFLOATV(b);
++ const float32x2_t invB = VRECIP(b);
++ return vmulq_lane_f32(a, invB, 0);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Div(const Vec4V a, const Vec4V b)
++ {
++ const float32x4_t invB = VRECIPQ(b);
++ return vmulq_f32(a, invB);
++ }
++
++ PX_FORCE_INLINE Vec4V V4ScaleInvFast(const Vec4V a, const FloatV b)
++ {
++ ASSERT_ISVALIDFLOATV(b);
++ const float32x2_t invB = VRECIPE(b);
++ return vmulq_lane_f32(a, invB, 0);
++ }
++
++ PX_FORCE_INLINE Vec4V V4DivFast(const Vec4V a, const Vec4V b)
++ {
++ const float32x4_t invB = VRECIPEQ(b);
++ return vmulq_f32(a, invB);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Recip(const Vec4V a)
++ {
++ return VRECIPQ(a);
++ }
++
++ PX_FORCE_INLINE Vec4V V4RecipFast(const Vec4V a)
++ {
++ return VRECIPEQ(a);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Rsqrt(const Vec4V a)
++ {
++ return VRECIPSQRTQ(a);
++ }
++
++ PX_FORCE_INLINE Vec4V V4RsqrtFast(const Vec4V a)
++ {
++ return VRECIPSQRTEQ(a);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Sqrt(const Vec4V a)
++ {
++ return V4Sel(V4IsEq(a, V4Zero()), a, V4Mul(a, VRECIPSQRTQ(a)));
++ }
++
++ PX_FORCE_INLINE Vec4V V4ScaleAdd(const Vec4V a, const FloatV b, const Vec4V c)
++ {
++ ASSERT_ISVALIDFLOATV(b);
++ return vmlaq_lane_f32(c, a, b, 0);
++ }
++
++ PX_FORCE_INLINE Vec4V V4NegScaleSub(const Vec4V a, const FloatV b, const Vec4V c)
++ {
++ ASSERT_ISVALIDFLOATV(b);
++ return vmlsq_lane_f32(c, a, b, 0);
++ }
++
++ PX_FORCE_INLINE Vec4V V4MulAdd(const Vec4V a, const Vec4V b, const Vec4V c)
++ {
++ return vmlaq_f32(c, a, b);
++ }
++
++ PX_FORCE_INLINE Vec4V V4NegMulSub(const Vec4V a, const Vec4V b, const Vec4V c)
++ {
++ return vmlsq_f32(c, a, b);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Abs(const Vec4V a)
++ {
++ return vabsq_f32(a);
++ }
++
++ PX_FORCE_INLINE FloatV V4SumElements(const Vec4V a)
++ {
++ const Vec4V xy = V4UnpackXY(a, a); // x,x,y,y
++ const Vec4V zw = V4UnpackZW(a, a); // z,z,w,w
++ const Vec4V xz_yw = V4Add(xy, zw); // x+z,x+z,y+w,y+w
++ const FloatV xz = V4GetX(xz_yw); // x+z
++ const FloatV yw = V4GetZ(xz_yw); // y+w
++ return FAdd(xz, yw); // sum
++ }
++
++ PX_FORCE_INLINE FloatV V4Dot(const Vec4V a, const Vec4V b)
++ {
++ const float32x4_t tmp = vmulq_f32(a, b);
++ const float32x2_t low = vget_low_f32(tmp);
++ const float32x2_t high = vget_high_f32(tmp);
++
++ const float32x2_t sumTmp = vpadd_f32(low, high); // = {z+w, x+y}
++ const float32x2_t sumWZYX = vpadd_f32(sumTmp, sumTmp); // = {x+y+z+w, x+y+z+w}
++ return sumWZYX;
++ }
++
++ PX_FORCE_INLINE FloatV V4Dot3(const Vec4V aa, const Vec4V bb)
++ {
++ // PT: the V3Dot code relies on the fact that W=0 so we can't reuse it as-is, we need to clear W first.
++ // TODO: find a better implementation that does not need to clear W.
++ const Vec4V a = V4ClearW(aa);
++ const Vec4V b = V4ClearW(bb);
++
++ const float32x4_t tmp = vmulq_f32(a, b);
++ const float32x2_t low = vget_low_f32(tmp);
++ const float32x2_t high = vget_high_f32(tmp);
++
++ const float32x2_t sumTmp = vpadd_f32(low, high); // = {0+z, x+y}
++ const float32x2_t sum0ZYX = vpadd_f32(sumTmp, sumTmp); // = {x+y+z, x+y+z}
++ return sum0ZYX;
++ }
++
++ PX_FORCE_INLINE Vec4V V4Cross(const Vec4V a, const Vec4V b)
++ {
++ const uint32x2_t TF = { 0xffffFFFF, 0x0 };
++ const float32x2_t ay_ax = vget_low_f32(a); // d2
++ const float32x2_t aw_az = vget_high_f32(a); // d3
++ const float32x2_t by_bx = vget_low_f32(b); // d4
++ const float32x2_t bw_bz = vget_high_f32(b); // d5
++ // Hi, Lo
++ const float32x2_t bz_by = vext_f32(by_bx, bw_bz, 1); // bz, by
++ const float32x2_t az_ay = vext_f32(ay_ax, aw_az, 1); // az, ay
++
++ const float32x2_t azbx = vmul_f32(aw_az, by_bx); // 0, az*bx
++ const float32x2_t aybz_axby = vmul_f32(ay_ax, bz_by); // ay*bz, ax*by
++
++ const float32x2_t azbxSUBaxbz = vmls_f32(azbx, bw_bz, ay_ax); // 0, az*bx-ax*bz
++ const float32x2_t aybzSUBazby_axbySUBaybx = vmls_f32(aybz_axby, by_bx, az_ay); // ay*bz-az*by, ax*by-ay*bx
++
++ const float32x2_t retLow = vext_f32(aybzSUBazby_axbySUBaybx, azbxSUBaxbz, 1); // az*bx-ax*bz, ay*bz-az*by
++ const uint32x2_t retHigh = vand_u32(TF, vreinterpret_u32_f32(aybzSUBazby_axbySUBaybx)); // 0, ax*by-ay*bx
++
++ return vcombine_f32(retLow, vreinterpret_f32_u32(retHigh));
++ }
++
++ PX_FORCE_INLINE FloatV V4Length(const Vec4V a)
++ {
++ const float32x4_t tmp = vmulq_f32(a, a);
++ const float32x2_t low = vget_low_f32(tmp);
++ const float32x2_t high = vget_high_f32(tmp);
++
++ const float32x2_t sumTmp = vpadd_f32(low, high); // = {0+z, x+y}
++ const float32x2_t sumWZYX = vpadd_f32(sumTmp, sumTmp); // = {x+y+z, x+y+z}
++ return FSqrt(sumWZYX);
++ }
++
++ PX_FORCE_INLINE FloatV V4LengthSq(const Vec4V a)
++ {
++ return V4Dot(a, a);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Normalize(const Vec4V a)
++ {
++ //PX_ASSERT(!FAllEq(V4LengthSq(a), FZero()));
++ return V4ScaleInv(a, V4Length(a));
++ }
++
++ PX_FORCE_INLINE Vec4V V4NormalizeFast(const Vec4V a)
++ {
++ //PX_ASSERT(!FAllEq(V4LengthSq(a), FZero()));
++ return V4Scale(a, FRsqrtFast(V4Dot(a, a)));
++ }
++
++ PX_FORCE_INLINE Vec4V V4NormalizeSafe(const Vec4V a, const Vec4V unsafeReturnValue)
++ {
++ const FloatV zero = FZero();
++ const FloatV length = V4Length(a);
++ const uint32x4_t isGreaterThanZero = FIsGrtr(length, zero);
++ return V4Sel(isGreaterThanZero, V4ScaleInv(a, length), unsafeReturnValue);
++ }
++
++ PX_FORCE_INLINE BoolV V4IsEqU32(const VecU32V a, const VecU32V b)
++ {
++ return vceqq_u32(a, b);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Sel(const BoolV c, const Vec4V a, const Vec4V b)
++ {
++ return vbslq_f32(c, a, b);
++ }
++
++ PX_FORCE_INLINE BoolV V4IsGrtr(const Vec4V a, const Vec4V b)
++ {
++ return vcgtq_f32(a, b);
++ }
++
++ PX_FORCE_INLINE BoolV V4IsGrtrOrEq(const Vec4V a, const Vec4V b)
++ {
++ return vcgeq_f32(a, b);
++ }
++
++ PX_FORCE_INLINE BoolV V4IsEq(const Vec4V a, const Vec4V b)
++ {
++ return vceqq_f32(a, b);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Max(const Vec4V a, const Vec4V b)
++ {
++ return vmaxq_f32(a, b);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Min(const Vec4V a, const Vec4V b)
++ {
++ return vminq_f32(a, b);
++ }
++
++ PX_FORCE_INLINE FloatV V4ExtractMax(const Vec4V a)
++ {
++ const float32x2_t low = vget_low_f32(a);
++ const float32x2_t high = vget_high_f32(a);
++
++ const float32x2_t max0 = vpmax_f32(high, low);
++ const float32x2_t max1 = vpmax_f32(max0, max0);
++
++ return max1;
++ }
++
++ PX_FORCE_INLINE FloatV V4ExtractMin(const Vec4V a)
++ {
++ const float32x2_t low = vget_low_f32(a);
++ const float32x2_t high = vget_high_f32(a);
++
++ const float32x2_t min0 = vpmin_f32(high, low);
++ const float32x2_t min1 = vpmin_f32(min0, min0);
++
++ return min1;
++ }
++
++ PX_FORCE_INLINE Vec4V V4Clamp(const Vec4V a, const Vec4V minV, const Vec4V maxV)
++ {
++ return V4Max(V4Min(a, maxV), minV);
++ }
++
++ PX_FORCE_INLINE PxU32 V4AllGrtr(const Vec4V a, const Vec4V b)
++ {
++ return internalUnitNeonSimd::BAllTrue4_R(V4IsGrtr(a, b));
++ }
++
++ PX_FORCE_INLINE PxU32 V4AllGrtrOrEq(const Vec4V a, const Vec4V b)
++ {
++ return internalUnitNeonSimd::BAllTrue4_R(V4IsGrtrOrEq(a, b));
++ }
++
++ PX_FORCE_INLINE PxU32 V4AllGrtrOrEq3(const Vec4V a, const Vec4V b)
++ {
++ return internalUnitNeonSimd::BAllTrue3_R(V4IsGrtrOrEq(a, b));
++ }
++
++ PX_FORCE_INLINE PxU32 V4AllEq(const Vec4V a, const Vec4V b)
++ {
++ return internalUnitNeonSimd::BAllTrue4_R(V4IsEq(a, b));
++ }
++
++ PX_FORCE_INLINE PxU32 V4AnyGrtr3(const Vec4V a, const Vec4V b)
++ {
++ return internalUnitNeonSimd::BAnyTrue3_R(V4IsGrtr(a, b));
++ }
++
++ PX_FORCE_INLINE Vec4V V4Round(const Vec4V a)
++ {
++ // truncate(a + (0.5f - sign(a)))
++ const Vec4V half = V4Load(0.5f);
++ const float32x4_t sign = vcvtq_f32_u32((vshrq_n_u32(vreinterpretq_u32_f32(a), 31)));
++ const Vec4V aPlusHalf = V4Add(a, half);
++ const Vec4V aRound = V4Sub(aPlusHalf, sign);
++ return vcvtq_f32_s32(vcvtq_s32_f32(aRound));
++ }
++
++ PX_FORCE_INLINE Vec4V V4Sin(const Vec4V a)
++ {
++ const Vec4V recipTwoPi = V4LoadA(g_PXReciprocalTwoPi.f);
++ const Vec4V twoPi = V4LoadA(g_PXTwoPi.f);
++ const Vec4V tmp = V4Mul(a, recipTwoPi);
++ const Vec4V b = V4Round(tmp);
++ const Vec4V V1 = V4NegMulSub(twoPi, b, a);
++
++ // sin(V) ~= V - V^3 / 3! + V^5 / 5! - V^7 / 7! + V^9 / 9! - V^11 / 11! + V^13 / 13! -
++ // V^15 / 15! + V^17 / 17! - V^19 / 19! + V^21 / 21! - V^23 / 23! (for -PI <= V < PI)
++ const Vec4V V2 = V4Mul(V1, V1);
++ const Vec4V V3 = V4Mul(V2, V1);
++ const Vec4V V5 = V4Mul(V3, V2);
++ const Vec4V V7 = V4Mul(V5, V2);
++ const Vec4V V9 = V4Mul(V7, V2);
++ const Vec4V V11 = V4Mul(V9, V2);
++ const Vec4V V13 = V4Mul(V11, V2);
++ const Vec4V V15 = V4Mul(V13, V2);
++ const Vec4V V17 = V4Mul(V15, V2);
++ const Vec4V V19 = V4Mul(V17, V2);
++ const Vec4V V21 = V4Mul(V19, V2);
++ const Vec4V V23 = V4Mul(V21, V2);
++
++ const Vec4V sinCoefficients0 = V4LoadA(g_PXSinCoefficients0.f);
++ const Vec4V sinCoefficients1 = V4LoadA(g_PXSinCoefficients1.f);
++ const Vec4V sinCoefficients2 = V4LoadA(g_PXSinCoefficients2.f);
++
++ const FloatV S1 = V4GetY(sinCoefficients0);
++ const FloatV S2 = V4GetZ(sinCoefficients0);
++ const FloatV S3 = V4GetW(sinCoefficients0);
++ const FloatV S4 = V4GetX(sinCoefficients1);
++ const FloatV S5 = V4GetY(sinCoefficients1);
++ const FloatV S6 = V4GetZ(sinCoefficients1);
++ const FloatV S7 = V4GetW(sinCoefficients1);
++ const FloatV S8 = V4GetX(sinCoefficients2);
++ const FloatV S9 = V4GetY(sinCoefficients2);
++ const FloatV S10 = V4GetZ(sinCoefficients2);
++ const FloatV S11 = V4GetW(sinCoefficients2);
++
++ Vec4V Result;
++ Result = V4ScaleAdd(V3, S1, V1);
++ Result = V4ScaleAdd(V5, S2, Result);
++ Result = V4ScaleAdd(V7, S3, Result);
++ Result = V4ScaleAdd(V9, S4, Result);
++ Result = V4ScaleAdd(V11, S5, Result);
++ Result = V4ScaleAdd(V13, S6, Result);
++ Result = V4ScaleAdd(V15, S7, Result);
++ Result = V4ScaleAdd(V17, S8, Result);
++ Result = V4ScaleAdd(V19, S9, Result);
++ Result = V4ScaleAdd(V21, S10, Result);
++ Result = V4ScaleAdd(V23, S11, Result);
++
++ return Result;
++ }
++
++ PX_FORCE_INLINE Vec4V V4Cos(const Vec4V a)
++ {
++ const Vec4V recipTwoPi = V4LoadA(g_PXReciprocalTwoPi.f);
++ const Vec4V twoPi = V4LoadA(g_PXTwoPi.f);
++ const Vec4V tmp = V4Mul(a, recipTwoPi);
++ const Vec4V b = V4Round(tmp);
++ const Vec4V V1 = V4NegMulSub(twoPi, b, a);
++
++ // cos(V) ~= 1 - V^2 / 2! + V^4 / 4! - V^6 / 6! + V^8 / 8! - V^10 / 10! + V^12 / 12! -
++ // V^14 / 14! + V^16 / 16! - V^18 / 18! + V^20 / 20! - V^22 / 22! (for -PI <= V < PI)
++ const Vec4V V2 = V4Mul(V1, V1);
++ const Vec4V V4 = V4Mul(V2, V2);
++ const Vec4V V6 = V4Mul(V4, V2);
++ const Vec4V V8 = V4Mul(V4, V4);
++ const Vec4V V10 = V4Mul(V6, V4);
++ const Vec4V V12 = V4Mul(V6, V6);
++ const Vec4V V14 = V4Mul(V8, V6);
++ const Vec4V V16 = V4Mul(V8, V8);
++ const Vec4V V18 = V4Mul(V10, V8);
++ const Vec4V V20 = V4Mul(V10, V10);
++ const Vec4V V22 = V4Mul(V12, V10);
++
++ const Vec4V cosCoefficients0 = V4LoadA(g_PXCosCoefficients0.f);
++ const Vec4V cosCoefficients1 = V4LoadA(g_PXCosCoefficients1.f);
++ const Vec4V cosCoefficients2 = V4LoadA(g_PXCosCoefficients2.f);
++
++ const FloatV C1 = V4GetY(cosCoefficients0);
++ const FloatV C2 = V4GetZ(cosCoefficients0);
++ const FloatV C3 = V4GetW(cosCoefficients0);
++ const FloatV C4 = V4GetX(cosCoefficients1);
++ const FloatV C5 = V4GetY(cosCoefficients1);
++ const FloatV C6 = V4GetZ(cosCoefficients1);
++ const FloatV C7 = V4GetW(cosCoefficients1);
++ const FloatV C8 = V4GetX(cosCoefficients2);
++ const FloatV C9 = V4GetY(cosCoefficients2);
++ const FloatV C10 = V4GetZ(cosCoefficients2);
++ const FloatV C11 = V4GetW(cosCoefficients2);
++
++ Vec4V Result;
++ Result = V4ScaleAdd(V2, C1, V4One());
++ Result = V4ScaleAdd(V4, C2, Result);
++ Result = V4ScaleAdd(V6, C3, Result);
++ Result = V4ScaleAdd(V8, C4, Result);
++ Result = V4ScaleAdd(V10, C5, Result);
++ Result = V4ScaleAdd(V12, C6, Result);
++ Result = V4ScaleAdd(V14, C7, Result);
++ Result = V4ScaleAdd(V16, C8, Result);
++ Result = V4ScaleAdd(V18, C9, Result);
++ Result = V4ScaleAdd(V20, C10, Result);
++ Result = V4ScaleAdd(V22, C11, Result);
++
++ return Result;
++ }
++
++ PX_FORCE_INLINE void V4Transpose(Vec4V& col0, Vec4V& col1, Vec4V& col2, Vec4V& col3)
++ {
++ const float32x4x2_t v0v1 = vzipq_f32(col0, col2);
++ const float32x4x2_t v2v3 = vzipq_f32(col1, col3);
++ const float32x4x2_t zip0 = vzipq_f32(v0v1.val[0], v2v3.val[0]);
++ const float32x4x2_t zip1 = vzipq_f32(v0v1.val[1], v2v3.val[1]);
++ col0 = zip0.val[0];
++ col1 = zip0.val[1];
++ col2 = zip1.val[0];
++ col3 = zip1.val[1];
++ }
++
++ //////////////////////////////////
++ // VEC4V
++ //////////////////////////////////
++
++ PX_FORCE_INLINE BoolV BFFFF()
++ {
++ return vmovq_n_u32(0);
++ }
++
++ PX_FORCE_INLINE BoolV BFFFT()
++ {
++ const uint32x2_t zeros = vmov_n_u32(0);
++ const uint32x2_t ones = vmov_n_u32(0xffffFFFF);
++ const uint32x2_t zo = vext_u32(zeros, ones, 1);
++ return vcombine_u32(zeros, zo);
++ }
++
++ PX_FORCE_INLINE BoolV BFFTF()
++ {
++ const uint32x2_t zeros = vmov_n_u32(0);
++ const uint32x2_t ones = vmov_n_u32(0xffffFFFF);
++ const uint32x2_t oz = vext_u32(ones, zeros, 1);
++ return vcombine_u32(zeros, oz);
++ }
++
++ PX_FORCE_INLINE BoolV BFFTT()
++ {
++ const uint32x2_t zeros = vmov_n_u32(0);
++ const uint32x2_t ones = vmov_n_u32(0xffffFFFF);
++ return vcombine_u32(zeros, ones);
++ }
++
++ PX_FORCE_INLINE BoolV BFTFF()
++ {
++ const uint32x2_t zeros = vmov_n_u32(0);
++ const uint32x2_t ones = vmov_n_u32(0xffffFFFF);
++ const uint32x2_t zo = vext_u32(zeros, ones, 1);
++ return vcombine_u32(zo, zeros);
++ }
++
++ PX_FORCE_INLINE BoolV BFTFT()
++ {
++ const uint32x2_t zeros = vmov_n_u32(0);
++ const uint32x2_t ones = vmov_n_u32(0xffffFFFF);
++ const uint32x2_t zo = vext_u32(zeros, ones, 1);
++ return vcombine_u32(zo, zo);
++ }
++
++ PX_FORCE_INLINE BoolV BFTTF()
++ {
++ const uint32x2_t zeros = vmov_n_u32(0);
++ const uint32x2_t ones = vmov_n_u32(0xffffFFFF);
++ const uint32x2_t zo = vext_u32(zeros, ones, 1);
++ const uint32x2_t oz = vext_u32(ones, zeros, 1);
++ return vcombine_u32(zo, oz);
++ }
++
++ PX_FORCE_INLINE BoolV BFTTT()
++ {
++ const uint32x2_t zeros = vmov_n_u32(0);
++ const uint32x2_t ones = vmov_n_u32(0xffffFFFF);
++ const uint32x2_t zo = vext_u32(zeros, ones, 1);
++ return vcombine_u32(zo, ones);
++ }
++
++ PX_FORCE_INLINE BoolV BTFFF()
++ {
++ const uint32x2_t zeros = vmov_n_u32(0);
++ const uint32x2_t ones = vmov_n_u32(0xffffFFFF);
++ // const uint32x2_t zo = vext_u32(zeros, ones, 1);
++ const uint32x2_t oz = vext_u32(ones, zeros, 1);
++ return vcombine_u32(oz, zeros);
++ }
++
++ PX_FORCE_INLINE BoolV BTFFT()
++ {
++ const uint32x2_t zeros = vmov_n_u32(0);
++ const uint32x2_t ones = vmov_n_u32(0xffffFFFF);
++ const uint32x2_t zo = vext_u32(zeros, ones, 1);
++ const uint32x2_t oz = vext_u32(ones, zeros, 1);
++ return vcombine_u32(oz, zo);
++ }
++
++ PX_FORCE_INLINE BoolV BTFTF()
++ {
++ const uint32x2_t zeros = vmov_n_u32(0);
++ const uint32x2_t ones = vmov_n_u32(0xffffFFFF);
++ const uint32x2_t oz = vext_u32(ones, zeros, 1);
++ return vcombine_u32(oz, oz);
++ }
++
++ PX_FORCE_INLINE BoolV BTFTT()
++ {
++ const uint32x2_t zeros = vmov_n_u32(0);
++ const uint32x2_t ones = vmov_n_u32(0xffffFFFF);
++ const uint32x2_t oz = vext_u32(ones, zeros, 1);
++ return vcombine_u32(oz, ones);
++ }
++
++ PX_FORCE_INLINE BoolV BTTFF()
++ {
++ const uint32x2_t zeros = vmov_n_u32(0);
++ const uint32x2_t ones = vmov_n_u32(0xffffFFFF);
++ return vcombine_u32(ones, zeros);
++ }
++
++ PX_FORCE_INLINE BoolV BTTFT()
++ {
++ const uint32x2_t zeros = vmov_n_u32(0);
++ const uint32x2_t ones = vmov_n_u32(0xffffFFFF);
++ const uint32x2_t zo = vext_u32(zeros, ones, 1);
++ return vcombine_u32(ones, zo);
++ }
++
++ PX_FORCE_INLINE BoolV BTTTF()
++ {
++ const uint32x2_t zeros = vmov_n_u32(0);
++ const uint32x2_t ones = vmov_n_u32(0xffffFFFF);
++ const uint32x2_t oz = vext_u32(ones, zeros, 1);
++ return vcombine_u32(ones, oz);
++ }
++
++ PX_FORCE_INLINE BoolV BTTTT()
++ {
++ return vmovq_n_u32(0xffffFFFF);
++ }
++
++ PX_FORCE_INLINE BoolV BXMask()
++ {
++ return BTFFF();
++ }
++
++ PX_FORCE_INLINE BoolV BYMask()
++ {
++ return BFTFF();
++ }
++
++ PX_FORCE_INLINE BoolV BZMask()
++ {
++ return BFFTF();
++ }
++
++ PX_FORCE_INLINE BoolV BWMask()
++ {
++ return BFFFT();
++ }
++
++ PX_FORCE_INLINE BoolV BGetX(const BoolV f)
++ {
++ const uint32x2_t fLow = vget_low_u32(f);
++ return vdupq_lane_u32(fLow, 0);
++ }
++
++ PX_FORCE_INLINE BoolV BGetY(const BoolV f)
++ {
++ const uint32x2_t fLow = vget_low_u32(f);
++ return vdupq_lane_u32(fLow, 1);
++ }
++
++ PX_FORCE_INLINE BoolV BGetZ(const BoolV f)
++ {
++ const uint32x2_t fHigh = vget_high_u32(f);
++ return vdupq_lane_u32(fHigh, 0);
++ }
++
++ PX_FORCE_INLINE BoolV BGetW(const BoolV f)
++ {
++ const uint32x2_t fHigh = vget_high_u32(f);
++ return vdupq_lane_u32(fHigh, 1);
++ }
++
++ PX_FORCE_INLINE BoolV BSetX(const BoolV v, const BoolV f)
++ {
++ return vbslq_u32(BFTTT(), v, f);
++ }
++
++ PX_FORCE_INLINE BoolV BSetY(const BoolV v, const BoolV f)
++ {
++ return vbslq_u32(BTFTT(), v, f);
++ }
++
++ PX_FORCE_INLINE BoolV BSetZ(const BoolV v, const BoolV f)
++ {
++ return vbslq_u32(BTTFT(), v, f);
++ }
++
++ PX_FORCE_INLINE BoolV BSetW(const BoolV v, const BoolV f)
++ {
++ return vbslq_u32(BTTTF(), v, f);
++ }
++
++ PX_FORCE_INLINE BoolV BAnd(const BoolV a, const BoolV b)
++ {
++ return vandq_u32(a, b);
++ }
++
++ PX_FORCE_INLINE BoolV BNot(const BoolV a)
++ {
++ return vmvnq_u32(a);
++ }
++
++ PX_FORCE_INLINE BoolV BAndNot(const BoolV a, const BoolV b)
++ {
++ // return vbicq_u32(a, b);
++ return vandq_u32(a, vmvnq_u32(b));
++ }
++
++ PX_FORCE_INLINE BoolV BOr(const BoolV a, const BoolV b)
++ {
++ return vorrq_u32(a, b);
++ }
++
++ PX_FORCE_INLINE BoolV BAllTrue4(const BoolV a)
++ {
++ const uint32x2_t allTrue = vmov_n_u32(0xffffFFFF);
++ const uint16x4_t dHigh = vget_high_u16(vreinterpretq_u16_u32(a));
++ const uint16x4_t dLow = vmovn_u32(a);
++ uint16x8_t combined = vcombine_u16(dLow, dHigh);
++ const uint32x2_t finalReduce = vreinterpret_u32_u8(vmovn_u16(combined));
++ const uint32x2_t result = vceq_u32(finalReduce, allTrue);
++ return vdupq_lane_u32(result, 0);
++ }
++
++ PX_FORCE_INLINE BoolV BAnyTrue4(const BoolV a)
++ {
++ const uint32x2_t allTrue = vmov_n_u32(0xffffFFFF);
++ const uint16x4_t dHigh = vget_high_u16(vreinterpretq_u16_u32(a));
++ const uint16x4_t dLow = vmovn_u32(a);
++ uint16x8_t combined = vcombine_u16(dLow, dHigh);
++ const uint32x2_t finalReduce = vreinterpret_u32_u8(vmovn_u16(combined));
++ const uint32x2_t result = vtst_u32(finalReduce, allTrue);
++ return vdupq_lane_u32(result, 0);
++ }
++
++ PX_FORCE_INLINE BoolV BAllTrue3(const BoolV a)
++ {
++ const uint32x2_t allTrue3 = vmov_n_u32(0x00ffFFFF);
++ const uint16x4_t dHigh = vget_high_u16(vreinterpretq_u16_u32(a));
++ const uint16x4_t dLow = vmovn_u32(a);
++ uint16x8_t combined = vcombine_u16(dLow, dHigh);
++ const uint32x2_t finalReduce = vreinterpret_u32_u8(vmovn_u16(combined));
++ const uint32x2_t result = vceq_u32(vand_u32(finalReduce, allTrue3), allTrue3);
++ return vdupq_lane_u32(result, 0);
++ }
++
++ PX_FORCE_INLINE BoolV BAnyTrue3(const BoolV a)
++ {
++ const uint32x2_t allTrue3 = vmov_n_u32(0x00ffFFFF);
++ const uint16x4_t dHigh = vget_high_u16(vreinterpretq_u16_u32(a));
++ const uint16x4_t dLow = vmovn_u32(a);
++ uint16x8_t combined = vcombine_u16(dLow, dHigh);
++ const uint32x2_t finalReduce = vreinterpret_u32_u8(vmovn_u16(combined));
++ const uint32x2_t result = vtst_u32(vand_u32(finalReduce, allTrue3), allTrue3);
++ return vdupq_lane_u32(result, 0);
++ }
++
++ PX_FORCE_INLINE PxU32 BAllEq(const BoolV a, const BoolV b)
++ {
++ const BoolV bTest = vceqq_u32(a, b);
++ return internalUnitNeonSimd::BAllTrue4_R(bTest);
++ }
++
++ PX_FORCE_INLINE PxU32 BAllEqTTTT(const BoolV a)
++ {
++ return BAllEq(a, BTTTT());
++ }
++
++ PX_FORCE_INLINE PxU32 BAllEqFFFF(const BoolV a)
++ {
++ return BAllEq(a, BFFFF());
++ }
++
++ PX_FORCE_INLINE PxU32 BGetBitMask(const BoolV a)
++ {
++ static PX_ALIGN(16, const PxU32) bitMaskData[4] = { 1, 2, 4, 8 };
++ const uint32x4_t bitMask = *(reinterpret_cast<const uint32x4_t*>(bitMaskData));
++ const uint32x4_t t0 = vandq_u32(a, bitMask);
++ const uint32x2_t t1 = vpadd_u32(vget_low_u32(t0), vget_high_u32(t0)); // Pairwise add (0 + 1), (2 + 3)
++ return PxU32(vget_lane_u32(vpadd_u32(t1, t1), 0));
++ }
++
++ //////////////////////////////////
++ // MAT33V
++ //////////////////////////////////
++
++ PX_FORCE_INLINE Vec3V M33MulV3(const Mat33V& a, const Vec3V b)
++ {
++ const FloatV x = V3GetX(b);
++ const FloatV y = V3GetY(b);
++ const FloatV z = V3GetZ(b);
++ const Vec3V v0 = V3Scale(a.col0, x);
++ const Vec3V v1 = V3Scale(a.col1, y);
++ const Vec3V v2 = V3Scale(a.col2, z);
++ const Vec3V v0PlusV1 = V3Add(v0, v1);
++ return V3Add(v0PlusV1, v2);
++ }
++
++ PX_FORCE_INLINE Vec3V M33TrnspsMulV3(const Mat33V& a, const Vec3V b)
++ {
++ const FloatV x = V3Dot(a.col0, b);
++ const FloatV y = V3Dot(a.col1, b);
++ const FloatV z = V3Dot(a.col2, b);
++ return V3Merge(x, y, z);
++ }
++
++ PX_FORCE_INLINE Vec3V M33MulV3AddV3(const Mat33V& A, const Vec3V b, const Vec3V c)
++ {
++ const FloatV x = V3GetX(b);
++ const FloatV y = V3GetY(b);
++ const FloatV z = V3GetZ(b);
++ Vec3V result = V3ScaleAdd(A.col0, x, c);
++ result = V3ScaleAdd(A.col1, y, result);
++ return V3ScaleAdd(A.col2, z, result);
++ }
++
++ PX_FORCE_INLINE Mat33V M33MulM33(const Mat33V& a, const Mat33V& b)
++ {
++ return Mat33V(M33MulV3(a, b.col0), M33MulV3(a, b.col1), M33MulV3(a, b.col2));
++ }
++
++ PX_FORCE_INLINE Mat33V M33Add(const Mat33V& a, const Mat33V& b)
++ {
++ return Mat33V(V3Add(a.col0, b.col0), V3Add(a.col1, b.col1), V3Add(a.col2, b.col2));
++ }
++
++ PX_FORCE_INLINE Mat33V M33Scale(const Mat33V& a, const FloatV& b)
++ {
++ return Mat33V(V3Scale(a.col0, b), V3Scale(a.col1, b), V3Scale(a.col2, b));
++ }
++
++ PX_FORCE_INLINE Mat33V M33Inverse(const Mat33V& a)
++ {
++ const float32x2_t zeros = vreinterpret_f32_u32(vmov_n_u32(0));
++ const BoolV btttf = BTTTF();
++
++ const Vec3V cross01 = V3Cross(a.col0, a.col1);
++ const Vec3V cross12 = V3Cross(a.col1, a.col2);
++ const Vec3V cross20 = V3Cross(a.col2, a.col0);
++ const FloatV dot = V3Dot(cross01, a.col2);
++ const FloatV invDet = FRecipFast(dot);
++
++ const float32x4x2_t merge = vzipq_f32(cross12, cross01);
++ const float32x4_t mergeh = merge.val[0];
++ const float32x4_t mergel = merge.val[1];
++
++ // const Vec3V colInv0 = XMVectorPermute(mergeh,cross20,PxPermuteControl(0,4,1,7));
++ const float32x4_t colInv0_xxyy = vzipq_f32(mergeh, cross20).val[0];
++ const float32x4_t colInv0 = vreinterpretq_f32_u32(vandq_u32(vreinterpretq_u32_f32(colInv0_xxyy), btttf));
++
++ // const Vec3V colInv1 = XMVectorPermute(mergeh,cross20,PxPermuteControl(2,5,3,7));
++ const float32x2_t zw0 = vget_high_f32(mergeh);
++ const float32x2_t xy1 = vget_low_f32(cross20);
++ const float32x2_t yzero1 = vext_f32(xy1, zeros, 1);
++ const float32x2x2_t merge1 = vzip_f32(zw0, yzero1);
++ const float32x4_t colInv1 = vcombine_f32(merge1.val[0], merge1.val[1]);
++
++ // const Vec3V colInv2 = XMVectorPermute(mergel,cross20,PxPermuteControl(0,6,1,7));
++ const float32x2_t x0y0 = vget_low_f32(mergel);
++ const float32x2_t z1w1 = vget_high_f32(cross20);
++ const float32x2x2_t merge2 = vzip_f32(x0y0, z1w1);
++ const float32x4_t colInv2 = vcombine_f32(merge2.val[0], merge2.val[1]);
++
++ return Mat33V(vmulq_lane_f32(colInv0, invDet, 0), vmulq_lane_f32(colInv1, invDet, 0),
++ vmulq_lane_f32(colInv2, invDet, 0));
++ }
++
++ PX_FORCE_INLINE Mat33V M33Trnsps(const Mat33V& a)
++ {
++ return Mat33V(V3Merge(V3GetX(a.col0), V3GetX(a.col1), V3GetX(a.col2)),
++ V3Merge(V3GetY(a.col0), V3GetY(a.col1), V3GetY(a.col2)),
++ V3Merge(V3GetZ(a.col0), V3GetZ(a.col1), V3GetZ(a.col2)));
++ }
++
++ PX_FORCE_INLINE Mat33V M33Identity()
++ {
++ return Mat33V(V3UnitX(), V3UnitY(), V3UnitZ());
++ }
++
++ PX_FORCE_INLINE Mat33V M33Sub(const Mat33V& a, const Mat33V& b)
++ {
++ return Mat33V(V3Sub(a.col0, b.col0), V3Sub(a.col1, b.col1), V3Sub(a.col2, b.col2));
++ }
++
++ PX_FORCE_INLINE Mat33V M33Neg(const Mat33V& a)
++ {
++ return Mat33V(V3Neg(a.col0), V3Neg(a.col1), V3Neg(a.col2));
++ }
++
++ PX_FORCE_INLINE Mat33V M33Abs(const Mat33V& a)
++ {
++ return Mat33V(V3Abs(a.col0), V3Abs(a.col1), V3Abs(a.col2));
++ }
++
++ PX_FORCE_INLINE Mat33V PromoteVec3V(const Vec3V v)
++ {
++ const BoolV bTFFF = BTFFF();
++ const BoolV bFTFF = BFTFF();
++ const BoolV bFFTF = BTFTF();
++
++ const Vec3V zero = V3Zero();
++
++ return Mat33V(V3Sel(bTFFF, v, zero), V3Sel(bFTFF, v, zero), V3Sel(bFFTF, v, zero));
++ }
++
++ PX_FORCE_INLINE Mat33V M33Diagonal(const Vec3VArg d)
++ {
++ const Vec3V x = V3Mul(V3UnitX(), d);
++ const Vec3V y = V3Mul(V3UnitY(), d);
++ const Vec3V z = V3Mul(V3UnitZ(), d);
++ return Mat33V(x, y, z);
++ }
++
++ //////////////////////////////////
++ // MAT34V
++ //////////////////////////////////
++
++ PX_FORCE_INLINE Vec3V M34MulV3(const Mat34V& a, const Vec3V b)
++ {
++ const FloatV x = V3GetX(b);
++ const FloatV y = V3GetY(b);
++ const FloatV z = V3GetZ(b);
++ const Vec3V v0 = V3Scale(a.col0, x);
++ const Vec3V v1 = V3Scale(a.col1, y);
++ const Vec3V v2 = V3Scale(a.col2, z);
++ const Vec3V v0PlusV1 = V3Add(v0, v1);
++ const Vec3V v0PlusV1Plusv2 = V3Add(v0PlusV1, v2);
++ return V3Add(v0PlusV1Plusv2, a.col3);
++ }
++
++ PX_FORCE_INLINE Vec3V M34Mul33V3(const Mat34V& a, const Vec3V b)
++ {
++ const FloatV x = V3GetX(b);
++ const FloatV y = V3GetY(b);
++ const FloatV z = V3GetZ(b);
++ const Vec3V v0 = V3Scale(a.col0, x);
++ const Vec3V v1 = V3Scale(a.col1, y);
++ const Vec3V v2 = V3Scale(a.col2, z);
++ const Vec3V v0PlusV1 = V3Add(v0, v1);
++ return V3Add(v0PlusV1, v2);
++ }
++
++ PX_FORCE_INLINE Vec3V M34TrnspsMul33V3(const Mat34V& a, const Vec3V b)
++ {
++ const FloatV x = V3Dot(a.col0, b);
++ const FloatV y = V3Dot(a.col1, b);
++ const FloatV z = V3Dot(a.col2, b);
++ return V3Merge(x, y, z);
++ }
++
++ PX_FORCE_INLINE Mat34V M34MulM34(const Mat34V& a, const Mat34V& b)
++ {
++ return Mat34V(M34Mul33V3(a, b.col0), M34Mul33V3(a, b.col1), M34Mul33V3(a, b.col2), M34MulV3(a, b.col3));
++ }
++
++ PX_FORCE_INLINE Mat33V M34MulM33(const Mat34V& a, const Mat33V& b)
++ {
++ return Mat33V(M34Mul33V3(a, b.col0), M34Mul33V3(a, b.col1), M34Mul33V3(a, b.col2));
++ }
++
++ PX_FORCE_INLINE Mat33V M34Mul33MM34(const Mat34V& a, const Mat34V& b)
++ {
++ return Mat33V(M34Mul33V3(a, b.col0), M34Mul33V3(a, b.col1), M34Mul33V3(a, b.col2));
++ }
++
++ PX_FORCE_INLINE Mat34V M34Add(const Mat34V& a, const Mat34V& b)
++ {
++ return Mat34V(V3Add(a.col0, b.col0), V3Add(a.col1, b.col1), V3Add(a.col2, b.col2), V3Add(a.col3, b.col3));
++ }
++
++ PX_FORCE_INLINE Mat33V M34Trnsps33(const Mat34V& a)
++ {
++ return Mat33V(V3Merge(V3GetX(a.col0), V3GetX(a.col1), V3GetX(a.col2)),
++ V3Merge(V3GetY(a.col0), V3GetY(a.col1), V3GetY(a.col2)),
++ V3Merge(V3GetZ(a.col0), V3GetZ(a.col1), V3GetZ(a.col2)));
++ }
++
++ //////////////////////////////////
++ // MAT44V
++ //////////////////////////////////
++
++ PX_FORCE_INLINE Vec4V M44MulV4(const Mat44V& a, const Vec4V b)
++ {
++ const FloatV x = V4GetX(b);
++ const FloatV y = V4GetY(b);
++ const FloatV z = V4GetZ(b);
++ const FloatV w = V4GetW(b);
++
++ const Vec4V v0 = V4Scale(a.col0, x);
++ const Vec4V v1 = V4Scale(a.col1, y);
++ const Vec4V v2 = V4Scale(a.col2, z);
++ const Vec4V v3 = V4Scale(a.col3, w);
++ const Vec4V v0PlusV1 = V4Add(v0, v1);
++ const Vec4V v0PlusV1Plusv2 = V4Add(v0PlusV1, v2);
++ return V4Add(v0PlusV1Plusv2, v3);
++ }
++
++ PX_FORCE_INLINE Vec4V M44TrnspsMulV4(const Mat44V& a, const Vec4V b)
++ {
++ return V4Merge(V4Dot(a.col0, b), V4Dot(a.col1, b), V4Dot(a.col2, b), V4Dot(a.col3, b));
++ }
++
++ PX_FORCE_INLINE Mat44V M44MulM44(const Mat44V& a, const Mat44V& b)
++ {
++ return Mat44V(M44MulV4(a, b.col0), M44MulV4(a, b.col1), M44MulV4(a, b.col2), M44MulV4(a, b.col3));
++ }
++
++ PX_FORCE_INLINE Mat44V M44Add(const Mat44V& a, const Mat44V& b)
++ {
++ return Mat44V(V4Add(a.col0, b.col0), V4Add(a.col1, b.col1), V4Add(a.col2, b.col2), V4Add(a.col3, b.col3));
++ }
++
++ PX_FORCE_INLINE Mat44V M44Trnsps(const Mat44V& a)
++ {
++ // asm volatile(
++ // "vzip.f32 %q0, %q2 \n\t"
++ // "vzip.f32 %q1, %q3 \n\t"
++ // "vzip.f32 %q0, %q1 \n\t"
++ // "vzip.f32 %q2, %q3 \n\t"
++ // : "+w" (a.col0), "+w" (a.col1), "+w" (a.col2), "+w" a.col3));
++
++ const float32x4x2_t v0v1 = vzipq_f32(a.col0, a.col2);
++ const float32x4x2_t v2v3 = vzipq_f32(a.col1, a.col3);
++ const float32x4x2_t zip0 = vzipq_f32(v0v1.val[0], v2v3.val[0]);
++ const float32x4x2_t zip1 = vzipq_f32(v0v1.val[1], v2v3.val[1]);
++
++ return Mat44V(zip0.val[0], zip0.val[1], zip1.val[0], zip1.val[1]);
++ }
++
++ PX_FORCE_INLINE Mat44V M44Inverse(const Mat44V& a)
++ {
++ float32x4_t minor0, minor1, minor2, minor3;
++ float32x4_t row0, row1, row2, row3;
++ float32x4_t det, tmp1;
++
++ tmp1 = vmovq_n_f32(0.0f);
++ row1 = vmovq_n_f32(0.0f);
++ row3 = vmovq_n_f32(0.0f);
++
++ row0 = a.col0;
++ row1 = vextq_f32(a.col1, a.col1, 2);
++ row2 = a.col2;
++ row3 = vextq_f32(a.col3, a.col3, 2);
++
++ tmp1 = vmulq_f32(row2, row3);
++ tmp1 = vrev64q_f32(tmp1);
++ minor0 = vmulq_f32(row1, tmp1);
++ minor1 = vmulq_f32(row0, tmp1);
++ tmp1 = vextq_f32(tmp1, tmp1, 2);
++ minor0 = vsubq_f32(vmulq_f32(row1, tmp1), minor0);
++ minor1 = vsubq_f32(vmulq_f32(row0, tmp1), minor1);
++ minor1 = vextq_f32(minor1, minor1, 2);
++
++ tmp1 = vmulq_f32(row1, row2);
++ tmp1 = vrev64q_f32(tmp1);
++ minor0 = vaddq_f32(vmulq_f32(row3, tmp1), minor0);
++ minor3 = vmulq_f32(row0, tmp1);
++ tmp1 = vextq_f32(tmp1, tmp1, 2);
++ minor0 = vsubq_f32(minor0, vmulq_f32(row3, tmp1));
++ minor3 = vsubq_f32(vmulq_f32(row0, tmp1), minor3);
++ minor3 = vextq_f32(minor3, minor3, 2);
++
++ tmp1 = vmulq_f32(vextq_f32(row1, row1, 2), row3);
++ tmp1 = vrev64q_f32(tmp1);
++ row2 = vextq_f32(row2, row2, 2);
++ minor0 = vaddq_f32(vmulq_f32(row2, tmp1), minor0);
++ minor2 = vmulq_f32(row0, tmp1);
++ tmp1 = vextq_f32(tmp1, tmp1, 2);
++ minor0 = vsubq_f32(minor0, vmulq_f32(row2, tmp1));
++ minor2 = vsubq_f32(vmulq_f32(row0, tmp1), minor2);
++ minor2 = vextq_f32(minor2, minor2, 2);
++
++ tmp1 = vmulq_f32(row0, row1);
++ tmp1 = vrev64q_f32(tmp1);
++ minor2 = vaddq_f32(vmulq_f32(row3, tmp1), minor2);
++ minor3 = vsubq_f32(vmulq_f32(row2, tmp1), minor3);
++ tmp1 = vextq_f32(tmp1, tmp1, 2);
++ minor2 = vsubq_f32(vmulq_f32(row3, tmp1), minor2);
++ minor3 = vsubq_f32(minor3, vmulq_f32(row2, tmp1));
++
++ tmp1 = vmulq_f32(row0, row3);
++ tmp1 = vrev64q_f32(tmp1);
++ minor1 = vsubq_f32(minor1, vmulq_f32(row2, tmp1));
++ minor2 = vaddq_f32(vmulq_f32(row1, tmp1), minor2);
++ tmp1 = vextq_f32(tmp1, tmp1, 2);
++ minor1 = vaddq_f32(vmulq_f32(row2, tmp1), minor1);
++ minor2 = vsubq_f32(minor2, vmulq_f32(row1, tmp1));
++
++ tmp1 = vmulq_f32(row0, row2);
++ tmp1 = vrev64q_f32(tmp1);
++ minor1 = vaddq_f32(vmulq_f32(row3, tmp1), minor1);
++ minor3 = vsubq_f32(minor3, vmulq_f32(row1, tmp1));
++ tmp1 = vextq_f32(tmp1, tmp1, 2);
++ minor1 = vsubq_f32(minor1, vmulq_f32(row3, tmp1));
++ minor3 = vaddq_f32(vmulq_f32(row1, tmp1), minor3);
++
++ det = vmulq_f32(row0, minor0);
++ det = vaddq_f32(vextq_f32(det, det, 2), det);
++ det = vaddq_f32(vrev64q_f32(det), det);
++ det = vdupq_lane_f32(VRECIPE(vget_low_f32(det)), 0);
++
++ minor0 = vmulq_f32(det, minor0);
++ minor1 = vmulq_f32(det, minor1);
++ minor2 = vmulq_f32(det, minor2);
++ minor3 = vmulq_f32(det, minor3);
++ Mat44V invTrans(minor0, minor1, minor2, minor3);
++ return M44Trnsps(invTrans);
++ }
++
++ PX_FORCE_INLINE Vec4V V4LoadXYZW(PxF32 x, PxF32 y, PxF32 z, PxF32 w)
++ {
++ const float32x4_t ret = { x, y, z, w };
++ return ret;
++ }
++
++ /*
++ PX_FORCE_INLINE VecU16V V4U32PK(VecU32V a, VecU32V b)
++ {
++ return vcombine_u16(vqmovn_u32(a), vqmovn_u32(b));
++ }
++ */
++
++ PX_FORCE_INLINE VecU32V V4U32Sel(const BoolV c, const VecU32V a, const VecU32V b)
++ {
++ return vbslq_u32(c, a, b);
++ }
++
++ PX_FORCE_INLINE VecU32V V4U32or(VecU32V a, VecU32V b)
++ {
++ return vorrq_u32(a, b);
++ }
++
++ PX_FORCE_INLINE VecU32V V4U32xor(VecU32V a, VecU32V b)
++ {
++ return veorq_u32(a, b);
++ }
++
++ PX_FORCE_INLINE VecU32V V4U32and(VecU32V a, VecU32V b)
++ {
++ return vandq_u32(a, b);
++ }
++
++ PX_FORCE_INLINE VecU32V V4U32Andc(VecU32V a, VecU32V b)
++ {
++ // return vbicq_u32(a, b); // creates gcc compiler bug in RTreeQueries.cpp
++ return vandq_u32(a, vmvnq_u32(b));
++ }
++
++ /*
++ PX_FORCE_INLINE VecU16V V4U16Or(VecU16V a, VecU16V b)
++ {
++ return vorrq_u16(a, b);
++ }
++ */
++
++ /*
++ PX_FORCE_INLINE VecU16V V4U16And(VecU16V a, VecU16V b)
++ {
++ return vandq_u16(a, b);
++ }
++ */
++ /*
++ PX_FORCE_INLINE VecU16V V4U16Andc(VecU16V a, VecU16V b)
++ {
++ return vbicq_u16(a, b);
++ }
++ */
++
++ PX_FORCE_INLINE VecI32V I4LoadXYZW(const PxI32& x, const PxI32& y, const PxI32& z, const PxI32& w)
++ {
++ const int32x4_t ret = { x, y, z, w };
++ return ret;
++ }
++
++ PX_FORCE_INLINE VecI32V I4Load(const PxI32 i)
++ {
++ return vdupq_n_s32(i);
++ }
++
++ PX_FORCE_INLINE VecI32V I4LoadU(const PxI32* i)
++ {
++ return vld1q_s32(i);
++ }
++
++ PX_FORCE_INLINE VecI32V I4LoadA(const PxI32* i)
++ {
++ return vld1q_s32(i);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_Add(const VecI32VArg a, const VecI32VArg b)
++ {
++ return vaddq_s32(a, b);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_Sub(const VecI32VArg a, const VecI32VArg b)
++ {
++ return vsubq_s32(a, b);
++ }
++
++ PX_FORCE_INLINE BoolV VecI32V_IsGrtr(const VecI32VArg a, const VecI32VArg b)
++ {
++ return vcgtq_s32(a, b);
++ }
++
++ PX_FORCE_INLINE BoolV VecI32V_IsEq(const VecI32VArg a, const VecI32VArg b)
++ {
++ return vceqq_s32(a, b);
++ }
++
++ PX_FORCE_INLINE VecI32V V4I32Sel(const BoolV c, const VecI32V a, const VecI32V b)
++ {
++ return vbslq_s32(c, a, b);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_Zero()
++ {
++ return vdupq_n_s32(0);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_One()
++ {
++ return vdupq_n_s32(1);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_Two()
++ {
++ return vdupq_n_s32(2);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_MinusOne()
++ {
++ return vdupq_n_s32(-1);
++ }
++
++ PX_FORCE_INLINE VecU32V U4Zero()
++ {
++ return U4Load(0);
++ }
++
++ PX_FORCE_INLINE VecU32V U4One()
++ {
++ return U4Load(1);
++ }
++
++ PX_FORCE_INLINE VecU32V U4Two()
++ {
++ return U4Load(2);
++ }
++
++ PX_FORCE_INLINE VecShiftV VecI32V_PrepareShift(const VecI32VArg shift)
++ {
++ return shift;
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_LeftShift(const VecI32VArg a, const VecShiftVArg count)
++ {
++ return vshlq_s32(a, count);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_RightShift(const VecI32VArg a, const VecShiftVArg count)
++ {
++ return vshlq_s32(a, VecI32V_Sub(I4Load(0), count));
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_LeftShift(const VecI32VArg a, const PxU32 count)
++ {
++ const int32x4_t shiftCount = { (PxI32)count, (PxI32)count, (PxI32)count, (PxI32)count };
++ return vshlq_s32(a, shiftCount);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_RightShift(const VecI32VArg a, const PxU32 count)
++ {
++ const int32x4_t shiftCount = { -(PxI32)count, -(PxI32)count, -(PxI32)count, -(PxI32)count };
++ return vshlq_s32(a, shiftCount);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_And(const VecI32VArg a, const VecI32VArg b)
++ {
++ return vandq_s32(a, b);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_Or(const VecI32VArg a, const VecI32VArg b)
++ {
++ return vorrq_s32(a, b);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_GetX(const VecI32VArg f)
++ {
++ const int32x2_t fLow = vget_low_s32(f);
++ return vdupq_lane_s32(fLow, 0);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_GetY(const VecI32VArg f)
++ {
++ const int32x2_t fLow = vget_low_s32(f);
++ return vdupq_lane_s32(fLow, 1);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_GetZ(const VecI32VArg f)
++ {
++ const int32x2_t fHigh = vget_high_s32(f);
++ return vdupq_lane_s32(fHigh, 0);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_GetW(const VecI32VArg f)
++ {
++ const int32x2_t fHigh = vget_high_s32(f);
++ return vdupq_lane_s32(fHigh, 1);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_Sel(const BoolV c, const VecI32VArg a, const VecI32VArg b)
++ {
++ return vbslq_s32(c, a, b);
++ }
++
++ PX_FORCE_INLINE void PxI32_From_VecI32V(const VecI32VArg a, PxI32* i)
++ {
++ *i = vgetq_lane_s32(a, 0);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_Merge(const VecI32VArg a, const VecI32VArg b, const VecI32VArg c, const VecI32VArg d)
++ {
++ const int32x2_t aLow = vget_low_s32(a);
++ const int32x2_t bLow = vget_low_s32(b);
++ const int32x2_t cLow = vget_low_s32(c);
++ const int32x2_t dLow = vget_low_s32(d);
++
++ const int32x2_t low = vext_s32(aLow, bLow, 1);
++ const int32x2_t high = vext_s32(cLow, dLow, 1);
++
++ return vcombine_s32(low, high);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_From_BoolV(const BoolVArg a)
++ {
++ return vreinterpretq_s32_u32(a);
++ }
++
++ PX_FORCE_INLINE VecU32V VecU32V_From_BoolV(const BoolVArg a)
++ {
++ return a;
++ }
++
++ /*
++ template<int a> PX_FORCE_INLINE VecI32V V4ISplat()
++ {
++ return vdupq_n_s32(a);
++ }
++
++ template<PxU32 a> PX_FORCE_INLINE VecU32V V4USplat()
++ {
++ return vdupq_n_u32(a);
++ }
++ */
++
++ /*
++ PX_FORCE_INLINE void V4U16StoreAligned(VecU16V val, VecU16V* address)
++ {
++ vst1q_u16((uint16_t*)address, val);
++ }
++ */
++
++ PX_FORCE_INLINE void V4U32StoreAligned(VecU32V val, VecU32V* address)
++ {
++ vst1q_u32(reinterpret_cast<uint32_t*>(address), val);
++ }
++
++ PX_FORCE_INLINE Vec4V V4LoadAligned(Vec4V* addr)
++ {
++ return vld1q_f32(reinterpret_cast<float32_t*>(addr));
++ }
++
++ PX_FORCE_INLINE Vec4V V4LoadUnaligned(Vec4V* addr)
++ {
++ return vld1q_f32(reinterpret_cast<float32_t*>(addr));
++ }
++
++ PX_FORCE_INLINE Vec4V V4Andc(const Vec4V a, const VecU32V b)
++ {
++ return vreinterpretq_f32_u32(V4U32Andc(vreinterpretq_u32_f32(a), b));
++ }
++
++ PX_FORCE_INLINE VecU32V V4IsGrtrV32u(const Vec4V a, const Vec4V b)
++ {
++ return V4IsGrtr(a, b);
++ }
++
++ PX_FORCE_INLINE VecU16V V4U16LoadAligned(VecU16V* addr)
++ {
++ return vld1q_u16(reinterpret_cast<uint16_t*>(addr));
++ }
++
++ PX_FORCE_INLINE VecU16V V4U16LoadUnaligned(VecU16V* addr)
++ {
++ return vld1q_u16(reinterpret_cast<uint16_t*>(addr));
++ }
++
++ PX_FORCE_INLINE VecU16V V4U16CompareGt(VecU16V a, VecU16V b)
++ {
++ return vcgtq_u16(a, b);
++ }
++
++ PX_FORCE_INLINE VecU16V V4I16CompareGt(VecI16V a, VecI16V b)
++ {
++ return vcgtq_s16(a, b);
++ }
++
++ PX_FORCE_INLINE Vec4V Vec4V_From_VecU32V(VecU32V a)
++ {
++ return vcvtq_f32_u32(a);
++ }
++
++ PX_FORCE_INLINE Vec4V Vec4V_From_VecI32V(VecI32V a)
++ {
++ return vcvtq_f32_s32(a);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_From_Vec4V(Vec4V a)
++ {
++ return vcvtq_s32_f32(a);
++ }
++
++ PX_FORCE_INLINE Vec4V Vec4V_ReinterpretFrom_VecU32V(VecU32V a)
++ {
++ return vreinterpretq_f32_u32(a);
++ }
++
++ PX_FORCE_INLINE Vec4V Vec4V_ReinterpretFrom_VecI32V(VecI32V a)
++ {
++ return vreinterpretq_f32_s32(a);
++ }
++
++ PX_FORCE_INLINE VecU32V VecU32V_ReinterpretFrom_Vec4V(Vec4V a)
++ {
++ return vreinterpretq_u32_f32(a);
++ }
++
++ PX_FORCE_INLINE VecI32V VecI32V_ReinterpretFrom_Vec4V(Vec4V a)
++ {
++ return vreinterpretq_s32_f32(a);
++ }
++
++ #if !PX_SWITCH
++ template <int index>
++ PX_FORCE_INLINE BoolV BSplatElement(BoolV a)
++ {
++ if(index < 2)
++ {
++ return vdupq_lane_u32(vget_low_u32(a), index);
++ }
++ else if(index == 2)
++ {
++ return vdupq_lane_u32(vget_high_u32(a), 0);
++ }
++ else if(index == 3)
++ {
++ return vdupq_lane_u32(vget_high_u32(a), 1);
++ }
++ }
++ #else
++ //workaround for template compile issue
++ template <int index> PX_FORCE_INLINE BoolV BSplatElement(BoolV a);
++ template<> PX_FORCE_INLINE BoolV BSplatElement<0>(BoolV a) { return vdupq_lane_u32(vget_low_u32(a), 0); }
++ template<> PX_FORCE_INLINE BoolV BSplatElement<1>(BoolV a) { return vdupq_lane_u32(vget_low_u32(a), 1); }
++ template<> PX_FORCE_INLINE BoolV BSplatElement<2>(BoolV a) { return vdupq_lane_u32(vget_high_u32(a), 0); }
++ template<> PX_FORCE_INLINE BoolV BSplatElement<3>(BoolV a) { return vdupq_lane_u32(vget_high_u32(a), 1); }
++ #endif
++
++ #if !PX_SWITCH
++ template <int index>
++ PX_FORCE_INLINE VecU32V V4U32SplatElement(VecU32V a)
++ {
++ if(index < 2)
++ {
++ return vdupq_lane_u32(vget_low_u32(a), index);
++ }
++ else if(index == 2)
++ {
++ return vdupq_lane_u32(vget_high_u32(a), 0);
++ }
++ else if(index == 3)
++ {
++ return vdupq_lane_u32(vget_high_u32(a), 1);
++ }
++ }
++ #else
++ //workaround for template compile issue
++ template <int index> PX_FORCE_INLINE VecU32V V4U32SplatElement(VecU32V a);
++ template <> PX_FORCE_INLINE VecU32V V4U32SplatElement<0>(VecU32V a) { return vdupq_lane_u32(vget_low_u32(a), 0); }
++ template <> PX_FORCE_INLINE VecU32V V4U32SplatElement<1>(VecU32V a) { return vdupq_lane_u32(vget_low_u32(a), 1); }
++ template <> PX_FORCE_INLINE VecU32V V4U32SplatElement<2>(VecU32V a) { return vdupq_lane_u32(vget_high_u32(a), 0); }
++ template <> PX_FORCE_INLINE VecU32V V4U32SplatElement<3>(VecU32V a) { return vdupq_lane_u32(vget_high_u32(a), 1); }
++ #endif
++
++ #if !PX_SWITCH
++ template <int index>
++ PX_FORCE_INLINE Vec4V V4SplatElement(Vec4V a)
++ {
++ if(index < 2)
++ {
++ return vdupq_lane_f32(vget_low_f32(a), index);
++ }
++ else if(index == 2)
++ {
++ return vdupq_lane_f32(vget_high_f32(a), 0);
++ }
++ else if(index == 3)
++ {
++ return vdupq_lane_f32(vget_high_f32(a), 1);
++ }
++ }
++ #else
++ //workaround for template compile issue
++ template <int index> PX_FORCE_INLINE Vec4V V4SplatElement(Vec4V a);
++ template <> PX_FORCE_INLINE Vec4V V4SplatElement<0>(Vec4V a) { return vdupq_lane_f32(vget_low_f32(a), 0); }
++ template <> PX_FORCE_INLINE Vec4V V4SplatElement<1>(Vec4V a) { return vdupq_lane_f32(vget_low_f32(a), 1); }
++ template <> PX_FORCE_INLINE Vec4V V4SplatElement<2>(Vec4V a) { return vdupq_lane_f32(vget_high_f32(a), 0); }
++ template <> PX_FORCE_INLINE Vec4V V4SplatElement<3>(Vec4V a) { return vdupq_lane_f32(vget_high_f32(a), 1); }
++ #endif
++
++ PX_FORCE_INLINE VecU32V U4LoadXYZW(PxU32 x, PxU32 y, PxU32 z, PxU32 w)
++ {
++ const uint32x4_t ret = { x, y, z, w };
++ return ret;
++ }
++
++ PX_FORCE_INLINE VecU32V U4Load(const PxU32 i)
++ {
++ return vdupq_n_u32(i);
++ }
++
++ PX_FORCE_INLINE VecU32V U4LoadU(const PxU32* i)
++ {
++ return vld1q_u32(i);
++ }
++
++ PX_FORCE_INLINE VecU32V U4LoadA(const PxU32* i)
++ {
++ return vld1q_u32(i);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Ceil(const Vec4V in)
++ {
++ const float32x4_t ones = vdupq_n_f32(1.0f);
++ const float32x4_t rdToZero = vcvtq_f32_s32(vcvtq_s32_f32(in));
++ const float32x4_t rdToZeroPlusOne = vaddq_f32(rdToZero, ones);
++ const uint32x4_t gt = vcgtq_f32(in, rdToZero);
++ return vbslq_f32(gt, rdToZeroPlusOne, rdToZero);
++ }
++
++ PX_FORCE_INLINE Vec4V V4Floor(const Vec4V in)
++ {
++ const float32x4_t ones = vdupq_n_f32(1.0f);
++ const float32x4_t rdToZero = vcvtq_f32_s32(vcvtq_s32_f32(in));
++ const float32x4_t rdToZeroMinusOne = vsubq_f32(rdToZero, ones);
++ const uint32x4_t lt = vcltq_f32(in, rdToZero);
++ return vbslq_f32(lt, rdToZeroMinusOne, rdToZero);
++ }
++
++ PX_FORCE_INLINE VecU32V V4ConvertToU32VSaturate(const Vec4V in, PxU32 power)
++ {
++ PX_ASSERT(power == 0 && "Non-zero power not supported in convertToU32VSaturate");
++ PX_UNUSED(power); // prevent warning in release builds
++
++ return vcvtq_u32_f32(in);
++ }
++
++ PX_FORCE_INLINE void QuatGetMat33V(const QuatVArg q, Vec3V& column0, Vec3V& column1, Vec3V& column2)
++ {
++ const FloatV one = FOne();
++ const FloatV x = V4GetX(q);
++ const FloatV y = V4GetY(q);
++ const FloatV z = V4GetZ(q);
++ const FloatV w = V4GetW(q);
++
++ const FloatV x2 = FAdd(x, x);
++ const FloatV y2 = FAdd(y, y);
++ const FloatV z2 = FAdd(z, z);
++
++ const FloatV xx = FMul(x2, x);
++ const FloatV yy = FMul(y2, y);
++ const FloatV zz = FMul(z2, z);
++
++ const FloatV xy = FMul(x2, y);
++ const FloatV xz = FMul(x2, z);
++ const FloatV xw = FMul(x2, w);
++
++ const FloatV yz = FMul(y2, z);
++ const FloatV yw = FMul(y2, w);
++ const FloatV zw = FMul(z2, w);
++
++ const FloatV v = FSub(one, xx);
++
++ column0 = V3Merge(FSub(FSub(one, yy), zz), FAdd(xy, zw), FSub(xz, yw));
++ column1 = V3Merge(FSub(xy, zw), FSub(v, zz), FAdd(yz, xw));
++ column2 = V3Merge(FAdd(xz, yw), FSub(yz, xw), FSub(v, yy));
++ }
++
++ } // namespace aos
++ } // namespace physx
++
++ #endif // PXFOUNDATION_PXUNIXNEONINLINEAOS_H