MCPcopy Create free account

hub / github.com/TinyMPC/TinyMPC / functions

Functions7,313 in github.com/TinyMPC/TinyMPC

⨍ Functions 7,313◇ Types & classes 2,787
FunctionCwiseBinaryNotEqualReturnType<OtherDerived> cwiseNotEqual
include/Eigen/Eigen/src/plugins/MatrixCwiseBinaryOps.h:68
MethodCwiseBinaryOp
Required for Visual Studio or the Copy constructor will probably not get inlined!
include/Eigen/Eigen/src/Core/CwiseBinaryOp.h:110
FunctionCwiseInverseReturnType cwiseInverse
include/Eigen/Eigen/src/plugins/MatrixCwiseUnaryOps.h:83
FunctionCwisePowReturnType<ScalarExponent> cwisePow
include/Eigen/Eigen/src/plugins/MatrixCwiseUnaryOps.h:108
FunctionCwiseScalarEqualReturnType cwiseEqual
\returns an expression of the coefficient-wise == operator of \c *this and a scalar \a s * * \warning this performs an exact comparison, which is
include/Eigen/Eigen/src/plugins/MatrixCwiseBinaryOps.h:197
FunctionCwiseScalarGreaterOrEqualReturnType cwiseGreaterOrEqual
\returns an expression of the coefficient-wise >= operator of \c *this and a scalar \a s */
include/Eigen/Eigen/src/plugins/MatrixCwiseBinaryOps.h:246
FunctionCwiseScalarGreaterReturnType cwiseGreater
\returns an expression of the coefficient-wise > operator of \c *this and a scalar \a s */
include/Eigen/Eigen/src/plugins/MatrixCwiseBinaryOps.h:230
FunctionCwiseScalarLessOrEqualReturnType cwiseLessOrEqual
\returns an expression of the coefficient-wise <= operator of \c *this and a scalar \a s */
include/Eigen/Eigen/src/plugins/MatrixCwiseBinaryOps.h:238
FunctionCwiseScalarLessReturnType cwiseLess
\returns an expression of the coefficient-wise < operator of \c *this and a scalar \a s */
include/Eigen/Eigen/src/plugins/MatrixCwiseBinaryOps.h:222
FunctionCwiseScalarNotEqualReturnType cwiseNotEqual
\returns an expression of the coefficient-wise == operator of \c *this and a scalar \a s * * \warning this performs an exact comparison, which is
include/Eigen/Eigen/src/plugins/MatrixCwiseBinaryOps.h:214
FunctionCwiseSqrtReturnType cwiseSqrt
include/Eigen/Eigen/src/plugins/MatrixCwiseUnaryOps.h:58
MethodCwiseTernaryOp
include/Eigen/Eigen/src/Core/CwiseTernaryOp.h:123
MethodCwiseUnaryOp
include/Eigen/Eigen/src/Core/CwiseUnaryOp.h:67
FunctionCwiseUnaryOp<CustomUnaryOp, const Derived> unaryExpr
include/Eigen/Eigen/src/plugins/CommonCwiseUnaryOps.h:133
MethodCwiseUnaryView
include/Eigen/Eigen/src/Core/CwiseUnaryView.h:74
FunctionCwiseUnaryView<CustomViewOp, const Derived> unaryViewExpr
include/Eigen/Eigen/src/plugins/CommonCwiseUnaryOps.h:154
MethodData
include/Eigen/Eigen/src/Core/CoreEvaluators.h:614
MethodData
include/Eigen/Eigen/src/Core/CoreEvaluators.h:909
MethodData
include/Eigen/Eigen/src/Core/CoreEvaluators.h:1003
MethodData
include/Eigen/Eigen/src/Core/CoreEvaluators.h:1069
MethodDecodePartition
include/Eigen/Eigen/src/ThreadPool/NonBlockingThreadPool.h:181
FunctionDenseBase
include/Eigen/Eigen/src/Core/DenseBase.h:683
MethodDenseStorage
include/Eigen/Eigen/src/Core/DenseStorage.h:178
MethodDenseStorage
include/Eigen/Eigen/src/Core/DenseStorage.h:218
MethodDenseStorage
include/Eigen/Eigen/src/Core/DenseStorage.h:239
MethodDenseStorage
include/Eigen/Eigen/src/Core/DenseStorage.h:273
MethodDenseStorage
include/Eigen/Eigen/src/Core/DenseStorage.h:304
MethodDenseStorage
include/Eigen/Eigen/src/Core/DenseStorage.h:340
MethodDenseStorage
include/Eigen/Eigen/src/Core/DenseStorage.h:384
MethodDenseStorage
include/Eigen/Eigen/src/Core/DenseStorage.h:422
MethodDenseStorage
include/Eigen/Eigen/src/Core/DenseStorage.h:460
MethodDenseStorage
include/Eigen/Eigen/src/Core/DenseStorage.h:544
MethodDenseStorage
include/Eigen/Eigen/src/Core/DenseStorage.h:619
MethodDiagonal
include/Eigen/Eigen/src/Core/Diagonal.h:75
MethodDiagonalDifferenceReturnType<OtherDerived> operator-
include/Eigen/Eigen/src/Core/DiagonalMatrix.h:151
MethodDiagonalMatrix
Default constructor without initialization */
include/Eigen/Eigen/src/Core/DiagonalMatrix.h:209
MethodDiagonalPreconditioner
include/Eigen/Eigen/src/IterativeLinearSolvers/BasicPreconditioners.h:49
MethodDiagonalProductReturnType<OtherDerived> operator*
include/Eigen/Eigen/src/Core/DiagonalMatrix.h:104
MethodDiagonalSumReturnType<OtherDerived> operator+
include/Eigen/Eigen/src/Core/DiagonalMatrix.h:140
MethodDiagonalWrapper
include/Eigen/Eigen/src/Core/DiagonalMatrix.h:363
FunctionDigammaReturnType digamma
\returns an expression of the coefficient-wise digamma (psi, derivative of lgamma). * * \specialfunctions_module * * \note This function suppo
include/Eigen/Eigen/src/plugins/ArrayCwiseUnaryOps.h:642
MethodDstPacketType packet
include/Eigen/Eigen/src/Core/CoreEvaluators.h:715
FunctionEIGEN_ALWAYS_INLINE bool (isfinite)
include/Eigen/Eigen/src/Core/arch/Default/BFloat16.h:789
FunctionEIGEN_ALWAYS_INLINE bool (isinf)
include/Eigen/Eigen/src/Core/arch/Default/BFloat16.h:783
FunctionEIGEN_ALWAYS_INLINE bool (isnan)
include/Eigen/Eigen/src/Core/arch/Default/BFloat16.h:777
FunctionEIGEN_ALWAYS_INLINE Packet16c pgather<signed char, Packet16c>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:840
FunctionEIGEN_ALWAYS_INLINE Packet16c pgather_partial<signed char, Packet16c>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:875
FunctionEIGEN_ALWAYS_INLINE Packet16uc pgather<unsigned char, Packet16uc>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:845
FunctionEIGEN_ALWAYS_INLINE Packet16uc pgather_partial<unsigned char, Packet16uc>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:880
FunctionEIGEN_ALWAYS_INLINE Packet2d pgather<double, Packet2d>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:2871
FunctionEIGEN_ALWAYS_INLINE Packet2d pgather_partial<double, Packet2d>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:2875
FunctionEIGEN_ALWAYS_INLINE Packet4f pgather<float, Packet4f>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:815
FunctionEIGEN_ALWAYS_INLINE Packet4f pgather_partial<float, Packet4f>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:850
FunctionEIGEN_ALWAYS_INLINE Packet4i pgather<int, Packet4i>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:820
FunctionEIGEN_ALWAYS_INLINE Packet4i pgather_partial<int, Packet4i>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:855
FunctionEIGEN_ALWAYS_INLINE Packet8bf pgather<bfloat16, Packet8bf>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:835
FunctionEIGEN_ALWAYS_INLINE Packet8bf pgather_partial<bfloat16, Packet8bf>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:870
FunctionEIGEN_ALWAYS_INLINE Packet8s pgather<short int, Packet8s>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:825
FunctionEIGEN_ALWAYS_INLINE Packet8s pgather_partial<short int, Packet8s>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:860
FunctionEIGEN_ALWAYS_INLINE Packet8us pgather<unsigned short int, Packet8us>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:830
FunctionEIGEN_ALWAYS_INLINE Packet8us pgather_partial<unsigned short int, Packet8us>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:865
FunctionEIGEN_ALWAYS_INLINE bool(isfinite)
include/Eigen/Eigen/src/Core/arch/Default/Half.h:918
FunctionEIGEN_ALWAYS_INLINE bool(isinf)
include/Eigen/Eigen/src/Core/arch/Default/Half.h:913
FunctionEIGEN_ALWAYS_INLINE bool(isnan)
include/Eigen/Eigen/src/Core/arch/Default/Half.h:908
FunctionEIGEN_ALWAYS_INLINE double2 ploadt_ro<double2, Aligned>
include/Eigen/Eigen/src/Core/arch/GPU/PacketMath.h:386
FunctionEIGEN_ALWAYS_INLINE double2 ploadt_ro<double2, Unaligned>
include/Eigen/Eigen/src/Core/arch/GPU/PacketMath.h:403
FunctionEIGEN_ALWAYS_INLINE float4 ploadt_ro<float4, Aligned>
include/Eigen/Eigen/src/Core/arch/GPU/PacketMath.h:378
FunctionEIGEN_ALWAYS_INLINE float4 ploadt_ro<float4, Unaligned>
include/Eigen/Eigen/src/Core/arch/GPU/PacketMath.h:395
FunctionEIGEN_ALWAYS_INLINE void pscatter<bfloat16, Packet8bf>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:924
FunctionEIGEN_ALWAYS_INLINE void pscatter<double, Packet2d>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:2879
FunctionEIGEN_ALWAYS_INLINE void pscatter<float, Packet4f>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:904
FunctionEIGEN_ALWAYS_INLINE void pscatter<int, Packet4i>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:909
FunctionEIGEN_ALWAYS_INLINE void pscatter<short int, Packet8s>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:914
FunctionEIGEN_ALWAYS_INLINE void pscatter<signed char, Packet16c>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:929
FunctionEIGEN_ALWAYS_INLINE void pscatter<unsigned char, Packet16uc>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:934
FunctionEIGEN_ALWAYS_INLINE void pscatter<unsigned short int, Packet8us>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:919
FunctionEIGEN_ALWAYS_INLINE void pscatter_partial<bfloat16, Packet8bf>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:959
FunctionEIGEN_ALWAYS_INLINE void pscatter_partial<double, Packet2d>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:2883
FunctionEIGEN_ALWAYS_INLINE void pscatter_partial<float, Packet4f>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:939
FunctionEIGEN_ALWAYS_INLINE void pscatter_partial<int, Packet4i>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:944
FunctionEIGEN_ALWAYS_INLINE void pscatter_partial<short int, Packet8s>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:949
FunctionEIGEN_ALWAYS_INLINE void pscatter_partial<signed char, Packet16c>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:964
FunctionEIGEN_ALWAYS_INLINE void pscatter_partial<unsigned char, Packet16uc>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:969
FunctionEIGEN_ALWAYS_INLINE void pscatter_partial<unsigned short int, Packet8us>
include/Eigen/Eigen/src/Core/arch/AltiVec/PacketMath.h:954
FunctionEIGEN_DECLARE_CONST_FAST_Packet4i
These constants are endian-agnostic
include/Eigen/Eigen/src/Core/arch/ZVector/PacketMath.h:88
FunctionEIGEN_DECLARE_CONST_FAST_Packet4ui
include/Eigen/Eigen/src/Core/arch/AltiVec/TypeCasting.h:76
FunctionEIGEN_DECLARE_CONST_Packet4f(cephes_SQRTHF, 0.707106781186547524f); static EIGEN_DECLARE_CONST_Packet4f(cephes_log_p0, 7.0376836292e-2f); static EIGEN_DECLARE_CONST_Packet4f(cephes_log_p1, -1.1514610310e-1f); static EIGEN_DECLARE_CONST_Packet4f(cephes_log_p2, 1.1676998740e-1f); static EIGEN_DECLARE_CONST_Packet4f(cephes_log_p3, -1.2420140846e-1f); static EIGEN_DECLARE_CONST_Packet4f(cephes_log_p4, +1.4249322787e-1f); static EIGEN_DECLARE_CONST_Packet4f(cephes_log_p5, -1.6668057665e-1f); static EIGEN_DECLARE_CONST_Packet4f(cephes_log_p6, +2.0000714765e-1f); static EIGEN_DECLARE_CONST_Packet4f(cephes_log_p7, -2.4999993993e-1f); static EIGEN_DECLARE_CONST_Packet4f(cephes_log_p8, +3.3333331174e-1f); static EIGEN_DECLARE_CONST_Packet4f(cephes_log_q1, -2.12194440e-4f); static EIGEN_DECLARE_CONST_Packet4f(cephes_log_q2, 0.693359375f); static EIGEN_DECLARE_CONST_Packet4f(half, 0.5f); static EIGEN_DECLARE_CONST_Packet4f(1, 1.0f); // Convert negative argument into NAN (quiet negative, to be specific). Packet4f zero = (Packet4f)__builtin_msa_ldi_w(0); Packet4i neg_mask = __builtin_msa_fclt_w(_x, zero); Packet4i zero_mask = __builtin_msa_fceq_w(_x, zero); Packet4f non_neg_x_or_nan = padd(_x, (Packet4f)neg_mask); // Add 0.0 or NAN. Packet4f x = non_neg_x_or_nan; // Extract exponent from x = mantissa * 2**exponent, where 1.0 <= mantissa < 2.0. // N.B. the exponent is one less of what frexpf() would return. Packet4i e_int = __builtin_msa_ftint_s_w(__builtin_msa_flog2_w(x)); // Multiply x by 2**(-exponent-1) to get 0.5 <= x < 1.0 as from frexpf(). x = __builtin_msa_fexp2_w(x, (Packet4i)__builtin_msa_nori_b((v16u8)e_int, 0)); /* if (x < SQRTHF) { x = x + x - 1.0; } else { e += 1; x = x - 1.0; } */ Packet4f xx = padd(x, x); Packet4i ge_mask = __builtin_msa_fcle_w(p4f_cephes_SQRTHF, x); e_int = psub(e_int, ge_mask); x = (Packet4f)__builtin_msa_bsel_v((v16u8)ge_mask, (v16u8)xx, (v16u8)x); x = psub(x, p4f_1); Packet4f e = __builtin_msa_ffint_s_w(e_int); Packet4f x2 = pmul(x, x); Packet4f x3 = pmul(x2, x); Packet4f y, y1, y2; y = pmadd(p4f_cephes_log_p0, x, p4f_cephes_log_p1); y1 = pmadd(p4f_cephes_log_p3, x, p4f_cephes_log_p4); y2 = pmadd(p4f_cephes_log_p6, x, p4f_cephes_log_p7); y = pmadd(y, x, p4f_cephes_log_p2); y1 = pmadd(y1, x, p4f_cephes_log_p5); y2 = pmadd(y2, x, p4f_cephes_log_p8); y = pmadd(y, x3, y1); y = pmadd(y, x3, y2); y = pmul(y, x3); y = pmadd(e, p4f_cephes_log_q1, y); x = __builtin_msa_fmsub_w(x, x2, p4f_half); x = padd(x, y); x = pmadd(e, p4f_cephes_log_q2, x); // x is now the logarithm result candidate. We still need to handle the // extreme arguments of zero and positive infinity, though. // N.B. if the argument is +INFINITY, x is NAN because the polynomial terms // contain infinities of both signs (see the coefficients and code above). // INFINITY - INFINITY is NAN. // If the argument is +INFINITY, make it the new result candidate. // To achieve that we choose the smaller of the result candidate and the // argument. // This is correct for all finite pairs of values (the logarithm is smaller // than the argument). // This is also correct in the special case when the argument is +INFINITY // and the result candidate is NAN. This is because the fmin.df instruction // prefers non-NANs to NANs. x = __builtin_msa_fmin_w(x, non_neg_x_or_nan); // If the argument is zero (including -0.0), the result becomes -INFINITY. Packet4i neg_infs = __builtin_msa_slli_w(zero_mask, 23); x = (Packet4f)__builtin_msa_bsel_v((v16u8)zero_mask, (v16u8)x, (v16u8)neg_infs); return x; } template <> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS Packet4f pexp<Packet4f>(const Packet4f& _x) { // Limiting single-precision pexp's argument to [-128, +128] lets pexp // reach 0 and INFINITY naturally. static EIGEN_DECLARE_CONST_Packet4f(exp_lo, -128.0f); static EIGEN_DECLARE_CONST_Packet4f(exp_hi, +128.0f); static EIGEN_DECLARE_CONST_Packet4f(cephes_LOG2EF, 1.44269504088896341f); static EIGEN_DECLARE_CONST_Packet4f(cephes_exp_C1, 0.693359375f); static EIGEN_DECLARE_CONST_Packet4f(cephes_exp_C2, -2.12194440e-4f); static EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p0, 1.9875691500e-4f); static EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p1, 1.3981999507e-3f); static EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p2, 8.3334519073e-3f); static EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p3, 4.1665795894e-2f); static EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p4, 1.6666665459e-1f); static EIGEN_DECLARE_CONST_Packet4f(cephes_exp_p5, 5.0000001201e-1f); static EIGEN_DECLARE_CONST_Packet4f(half, 0.5f); static EIGEN_DECLARE_CONST_Packet4f(1, 1.0f); Packet4f x = _x; // Clamp x. x = (Packet4f)__builtin_msa_bsel_v((v16u8)__builtin_msa_fclt_w(x, p4f_exp_lo), (v16u8)x, (v16u8)p4f_exp_lo); x = (Packet4f)__builtin_msa_bsel_v((v16u8)__builtin_msa_fclt_w(p4f_exp_hi, x), (v16u8)x, (v16u8)p4f_exp_hi); // Round to nearest integer by adding 0.5 (with x's sign) and truncating. Packet4f x2_add = (Packet4f)__builtin_msa_binsli_w((v4u32)p4f_half, (v4u32)x, 0); Packet4f x2 = pmadd(x, p4f_cephes_LOG2EF, x2_add); Packet4i x2_int = __builtin_msa_ftrunc_s_w(x2); Packet4f x2_int_f = __builtin_msa_ffint_s_w(x2_int); x = __builtin_msa_fmsub_w(x, x2_int_f, p4f_cephes_exp_C1); x = __builtin_msa_fmsub_w(x, x2_int_f, p4f_cephes_exp_C2); Packet4f z = pmul(x, x); Packet4f y = p4f_cephes_exp_p0; y = pmadd(y, x, p4f_cephes_exp_p1); y = pmadd(y, x, p4f_cephes_exp_p2); y = pmadd(y, x, p4f_cephes_exp_p3); y = pmadd(y, x, p4f_cephes_exp_p4); y = pmadd(y, x, p4f_cephes_exp_p5); y = pmadd(y, z, x); y = padd(y, p4f_1); // y *= 2**exponent. y = __builtin_msa_fexp2_w(y, x2_int); return y; } template <> EIGEN_DEFINE_FUNCTION_ALLOWING_MULTIPLE_DEFINITIONS Packet4f ptanh<Packet4f>(const Packet4f& _x) { static EIGEN_DECLARE_CONST_Packet4f(tanh_tiny, 1e-4f); static EIGEN_DECLARE_CONST_Packet4f(tanh_hi, 9.0f); // The monomial coefficients of the numerator polynomial (odd). static EIGEN_DECLARE_CONST_Packet4f(alpha_1, 4.89352455891786e-3f); static EIGEN_DECLARE_CONST_Packet4f(alpha_3, 6.37261928875436e-4f); static EIGEN_DECLARE_CONST_Packet4f(alpha_5, 1.48572235717979e-5f); static EIGEN_DECLARE_CONST_Packet4f(alpha_7, 5.12229709037114e-8f); static EIGEN_DECLARE_CONST_Packet4f(alpha_9, -8.60467152213735e-11f); static EIGEN_DECLARE_CONST_Packet4f(alpha_11, 2.00018790482477e-13f); static EIGEN_DECLARE_CONST_Packet4f(alpha_13, -2.76076847742355e-16f); // The monomial coefficients of the denominator polynomial (even). static EIGEN_DECLARE_CONST_Packet4f(beta_0, 4.89352518554385e-3f); static EIGEN_DECLARE_CONST_Packet4f(beta_2, 2.26843463243900e-3f); static EIGEN_DECLARE_CONST_Packet4f(beta_4, 1.18534705686654e-4f); static EIGEN_DECLARE_CONST_Packet4f(beta_6, 1.19825839466702e-6f); Packet4f x = pabs(_x); Packet4i tiny_mask = __builtin_msa_fclt_w(x, p4f_tanh_tiny); // Clamp the inputs to the range [-9, 9] since anything outside // this range is -/+1.0f in single-precision. x = (Packet4f)__builtin_msa_bsel_v((v16u8)__builtin_msa_fclt_w(p4f_tanh_hi, x), (v16u8)x, (v16u8)p4f_tanh_hi); // Since the polynomials are odd/even, we need x**2. Packet4f x2 = pmul(x, x); // Evaluate the numerator polynomial p. Packet4f p = pmadd(x2, p4f_alpha_13, p4f_alpha_11); p = pmadd(x2, p, p4f_alpha_9); p = pmadd(x2, p, p4f_alpha_7); p = pmadd(x2, p, p4f_alpha_5); p = pmadd(x2, p, p4f_alpha_3); p = pmadd(x2, p, p4f_alpha_1); p = pmul(x, p); // Evaluate the denominator polynomial q. Packet4f q = pmadd(x2, p4f_beta_6, p4f_beta_4); q = pmadd(x2, q, p4f_beta_2); q = pmadd(x2, q, p4f_beta_0); // Divide the numerator by the denominator. p = pdiv(p, q); // Reinstate the sign. p = (Packet4f)__builtin_msa_binsli_w((v4u32)p, (v4u32)_x, 0); // When the argument is very small in magnitude it's more accurate to just return it. p = (Packet4f)__builtin_msa_bsel_v((v16u8)tiny_mask, (v16u8)p, (v16u8)_x); return p; } template <bool sine> Packet4f psincos_inner_msa_float(const Packet4f& _x) { static EIGEN_DECLARE_CONST_Packet4f(sincos_max_arg, 13176795.0f); // Approx. (2**24) / (4/Pi). static EIGEN_DECLARE_CONST_Packet4f(minus_cephes_DP1, -0.78515625f); static EIGEN_DECLARE_CONST_Packet4f(minus_cephes_DP2, -2.4187564849853515625e-4f); static EIGEN_DECLARE_CONST_Packet4f(minus_cephes_DP3, -3.77489497744594108e-8f); static EIGEN_DECLARE_CONST_Packet4f(sincof_p0, -1.9515295891e-4f); static EIGEN_DECLARE_CONST_Packet4f(sincof_p1, 8.3321608736e-3f); static EIGEN_DECLARE_CONST_Packet4f(sincof_p2, -1.6666654611e-1f); static EIGEN_DECLARE_CONST_Packet4f(coscof_p0, 2.443315711809948e-5f); static EIGEN_DECLARE_CONST_Packet4f(coscof_p1, -1.388731625493765e-3f); static EIGEN_DECLARE_CONST_Packet4f(coscof_p2, 4.166664568298827e-2f); static EIGEN_DECLARE_CONST_Packet4f(cephes_FOPI, 1.27323954473516f); // 4/Pi. static EIGEN_DECLARE_CONST_Packet4f(half, 0.5f); static EIGEN_DECLARE_CONST_Packet4f(1, 1.0f); Packet4f x = pabs(_x); // Translate infinite arguments into NANs. Packet4f zero_or_nan_if_inf = psub(_x, _x); x = padd(x, zero_or_nan_if_inf); // Prevent sin/cos from generating values larger than 1.0 in magnitude // for very large arguments by setting x to 0.0. Packet4i small_or_nan_mask = __builtin_msa_fcult_w(x, p4f_sincos_max_arg); x = pand(x, (Packet4f)small_or_nan_mask); // Scale x by 4/Pi to find x's octant. Packet4f y = pmul(x, p4f_cephes_FOPI); // Get the octant. We'll reduce x by this number of octants or by one more than it. Packet4i y_int = __builtin_msa_ftrunc_s_w(y); // x's from even-numbered octants will translate to octant 0: [0, +Pi/4]. // x's from odd-numbered octants will translate to octant -1: [-Pi/4, 0]. // Adjustment for odd-numbered octants: octant = (octant + 1) & (~1). Packet4i y_int1 = __builtin_msa_addvi_w(y_int, 1); Packet4i y_int2 = (Packet4i)__builtin_msa_bclri_w((Packet4ui)y_int1, 0); // bclri = bit-clear y = __builtin_msa_ffint_s_w(y_int2); // Compute the sign to apply to the polynomial. Packet4i sign_mask = sine ? pxor(__builtin_msa_slli_w(y_int1, 29), (Packet4i)_x) : __builtin_msa_slli_w(__builtin_msa_addvi_w(y_int, 3), 29); // Get the polynomial selection mask. // We'll calculate both (sin and cos) polynomials and then select from the two. Packet4i poly_mask = __builtin_msa_ceqi_w(__builtin_msa_slli_w(y_int2, 30), 0); // Reduce x by y octants to get: -Pi/4 <= x <= +Pi/4. // The magic pass: "Extended precision modular arithmetic" // x = ((x - y * DP1) - y * DP2) - y * DP3 Packet4f tmp1 = pmul(y, p4f_minus_cephes_DP1); Packet4f tmp2 = pmul(y, p4f_minus_cephes_DP2); Packet4f tmp3 = pmul(y, p4f_minus_cephes_DP3); x = padd(x, tmp1); x = padd(x, tmp2); x = padd(x, tmp3); // Evaluate the cos(x) polynomial. y = p4f_coscof_p0; Packet4f z = pmul(x, x); y = pmadd(y, z, p4f_coscof_p1); y = pmadd(y, z, p4f_coscof_p2); y = pmul(y, z); y = pmul(y, z); y = __builtin_msa_fmsub_w(y, z, p4f_half); y = padd(y, p4f_1); // Evaluate the sin(x) polynomial. Packet4f y2 = p4f_sincof_p0; y2 = pmadd(y2, z, p4f_sincof_p1); y2 = pmadd(y2, z, p4f_sincof_p2); y2 = pmul(y2, z); y2 = pmadd(y2, x, x); // Select the correct result from the two polynomials. y = sine ? (Packet4f)__builtin_msa_bsel_v((v16u8)poly_mask, (v16u8)y, (v16u8)y2) : (Packet4f)__builtin_msa_bsel_v((v16u8)poly_mask, (v16u8)y2, (v16u8)y); // Update the sign. sign_mask = pxor(sign_mask, (Packet4i)y); y = (Packet4f)
include/Eigen/Eigen/src/Core/arch/MSA/MathFunctions.h:38
MethodEIGEN_DEVICE_FUNC replicate
include/Eigen/Eigen/src/Core/VectorwiseOp.h:567
FunctionEIGEN_DEVICE_FUNC ~aligned_stack_memory_handler
include/Eigen/Eigen/src/Core/util/Memory.h:685
FunctionEIGEN_DEVICE_FUNC ~local_nested_eval_wrapper
include/Eigen/Eigen/src/Core/util/Memory.h:736
FunctionEIGEN_DEVICE_FUNC Packet4hf ptanh<Packet4hf>
include/Eigen/Eigen/src/Core/arch/NEON/MathFunctions.h:22
FunctionEIGEN_DEVICE_FUNC Packet8hf ptanh<Packet8hf>
include/Eigen/Eigen/src/Core/arch/NEON/MathFunctions.h:29
FunctionEIGEN_DEVICE_FUNC bool sqrt<bool>
include/Eigen/Eigen/src/Core/MathFunctions.h:1376
MethodEIGEN_DEVICE_FUNC Inverse
include/Eigen/Eigen/src/Core/Inverse.h:55
MethodEIGEN_DEVICE_FUNC bfloat16
include/Eigen/Eigen/src/Core/arch/Default/BFloat16.h:115
FunctionEIGEN_DEVICE_FUNC bfloat16 (max)
include/Eigen/Eigen/src/Core/arch/Default/BFloat16.h:690
FunctionEIGEN_DEVICE_FUNC bfloat16 (min)
include/Eigen/Eigen/src/Core/arch/Default/BFloat16.h:684
FunctionEIGEN_DEVICE_FUNC bool (isfinite)
include/Eigen/Eigen/src/Core/arch/Default/Half.h:708
FunctionEIGEN_DEVICE_FUNC bool (isfinite)
include/Eigen/Eigen/src/Core/arch/Default/BFloat16.h:597
FunctionEIGEN_DEVICE_FUNC bool (isinf)
include/Eigen/Eigen/src/Core/arch/Default/Half.h:691
← previousnext →1,201–1,300 of 7,313, ranked by callers