sequenzo 0.1.31__cp310-cp310-macosx_10_9_x86_64.whl

sequenzo/dissimilarity_measures/src/xsimd/benchmark/xsimd_benchmark.hpp

@@ -0,0 +1,565 @@
+/***************************************************************************
+ * Copyright (c) Johan Mabille, Sylvain Corlay, Wolf Vollprecht and         *
+ * Martin Renou                                                             *
+ * Copyright (c) QuantStack                                                 *
+ * Copyright (c) Serge Guelton                                              *
+ *                                                                          *
+ * Distributed under the terms of the BSD 3-Clause License.                 *
+ *                                                                          *
+ * The full license is in the file LICENSE, distributed with this software. *
+ ****************************************************************************/
+
+#ifndef XSIMD_BENCHMARK_HPP
+#define XSIMD_BENCHMARK_HPP
+
+#include "xsimd/arch/xsimd_scalar.hpp"
+#include "xsimd/xsimd.hpp"
+#include <chrono>
+#include <iostream>
+#include <string>
+#include <vector>
+
+namespace xsimd
+{
+    using duration_type = std::chrono::duration<double, std::milli>;
+
+    template <class T>
+    using bench_vector = std::vector<T, xsimd::aligned_allocator<T>>;
+
+    template <class T>
+    void init_benchmark(bench_vector<T>& lhs, bench_vector<T>& rhs, bench_vector<T>& res, size_t size)
+    {
+        lhs.resize(size);
+        rhs.resize(size);
+        res.resize(size);
+        for (size_t i = 0; i < size; ++i)
+        {
+            lhs[i] = T(0.5) + std::sqrt(T(i)) * T(9.) / T(size);
+            rhs[i] = T(10.2) / T(i + 2) + T(0.25);
+        }
+    }
+
+    template <class T>
+    void init_benchmark(bench_vector<T>& op0, bench_vector<T>& op1, bench_vector<T>& op2, bench_vector<T>& res, size_t size)
+    {
+        op0.resize(size);
+        op1.resize(size);
+        op2.resize(size);
+        res.resize(size);
+        for (size_t i = 0; i < size; ++i)
+        {
+            op0[i] = T(0.5) + std::sqrt(T(i)) * T(9.) / T(size);
+            op1[i] = T(10.2) / T(i + 3) + T(0.25);
+            op2[i] = T(20.1) / T(i + 2) + T(0.65);
+        }
+    }
+
+    template <class T>
+    void init_benchmark_arctrigo(bench_vector<T>& lhs, bench_vector<T>& rhs, bench_vector<T>& res, size_t size)
+    {
+        lhs.resize(size);
+        rhs.resize(size);
+        res.resize(size);
+        for (size_t i = 0; i < size; ++i)
+        {
+            lhs[i] = T(-1.) + T(2.) * T(i) / T(size);
+            rhs[i] = T(i) / T(i + 2) + T(0.25);
+        }
+    }
+
+    enum class init_method
+    {
+        classic,
+        arctrigo
+    };
+
+    template <class F, class V>
+    duration_type benchmark_scalar(F f, V& lhs, V& res, std::size_t number)
+    {
+        size_t s = lhs.size();
+        duration_type t_res = duration_type::max();
+        for (std::size_t count = 0; count < number; ++count)
+        {
+            auto start = std::chrono::steady_clock::now();
+            for (size_t i = 0; i < s; ++i)
+            {
+                res[i] = f(lhs[i]);
+            }
+            auto end = std::chrono::steady_clock::now();
+            auto tmp = end - start;
+            t_res = tmp < t_res ? tmp : t_res;
+        }
+        return t_res;
+    }
+
+    template <class F, class V>
+    duration_type benchmark_scalar(F f, V& lhs, V& rhs, V& res, std::size_t number)
+    {
+        size_t s = lhs.size();
+        duration_type t_res = duration_type::max();
+        for (std::size_t count = 0; count < number; ++count)
+        {
+            auto start = std::chrono::steady_clock::now();
+            for (size_t i = 0; i < s; ++i)
+            {
+                res[i] = f(lhs[i], rhs[i]);
+            }
+            auto end = std::chrono::steady_clock::now();
+            auto tmp = end - start;
+            t_res = tmp < t_res ? tmp : t_res;
+        }
+        return t_res;
+    }
+
+    template <class F, class V>
+    duration_type benchmark_scalar(F f, V& op0, V& op1, V& op2, V& res, std::size_t number)
+    {
+        size_t s = op0.size();
+        duration_type t_res = duration_type::max();
+        for (std::size_t count = 0; count < number; ++count)
+        {
+            auto start = std::chrono::steady_clock::now();
+            for (size_t i = 0; i < s; ++i)
+            {
+                res[i] = f(op0[i], op1[i], op2[i]);
+            }
+            auto end = std::chrono::steady_clock::now();
+            auto tmp = end - start;
+            t_res = tmp < t_res ? tmp : t_res;
+        }
+        return t_res;
+    }
+
+    template <class B, class F, class V>
+    duration_type benchmark_simd(F f, V& lhs, V& res, std::size_t number)
+    {
+        std::size_t s = lhs.size();
+        duration_type t_res = duration_type::max();
+        for (std::size_t count = 0; count < number; ++count)
+        {
+            auto start = std::chrono::steady_clock::now();
+            for (std::size_t i = 0; i <= (s - B::size); i += B::size)
+            {
+                B blhs = B::load_aligned(&lhs[i]);
+                B bres = f(blhs);
+                bres.store_aligned(&res[i]);
+            }
+            auto end = std::chrono::steady_clock::now();
+            auto tmp = end - start;
+            t_res = tmp < t_res ? tmp : t_res;
+        }
+        return t_res;
+    }
+
+    template <class B, class F, class V>
+    duration_type benchmark_simd_unrolled(F f, V& lhs, V& res, std::size_t number)
+    {
+        std::size_t s = lhs.size();
+        std::size_t inc = 4 * B::size;
+        duration_type t_res = duration_type::max();
+        for (std::size_t count = 0; count < number; ++count)
+        {
+            auto start = std::chrono::steady_clock::now();
+            for (std::size_t i = 0; i <= (s - inc); i += inc)
+            {
+                size_t j = i + B::size;
+                size_t k = j + B::size;
+                size_t l = k + B::size;
+                B blhs = B::load_aligned(&lhs[i]),
+                  blhs2 = B::load_aligned(&lhs[j]),
+                  blhs3 = B::load_aligned(&lhs[k]),
+                  blhs4 = B::load_aligned(&lhs[l]);
+                B bres = f(blhs);
+                B bres2 = f(blhs2);
+                B bres3 = f(blhs3);
+                B bres4 = f(blhs4);
+                bres.store_aligned(&res[i]);
+                bres2.store_aligned(&res[j]);
+                bres3.store_aligned(&res[k]);
+                bres4.store_aligned(&res[l]);
+            }
+            auto end = std::chrono::steady_clock::now();
+            auto tmp = end - start;
+            t_res = tmp < t_res ? tmp : t_res;
+        }
+        return t_res;
+    }
+
+    template <class B, class F, class V>
+    duration_type benchmark_simd(F f, V& lhs, V& rhs, V& res, std::size_t number)
+    {
+        std::size_t s = lhs.size();
+        duration_type t_res = duration_type::max();
+        for (std::size_t count = 0; count < number; ++count)
+        {
+            auto start = std::chrono::steady_clock::now();
+            for (std::size_t i = 0; i <= (s - B::size); i += B::size)
+            {
+                B blhs = B::load_aligned(&lhs[i]),
+                  brhs = B::load_aligned(&rhs[i]);
+                B bres = f(blhs, brhs);
+                bres.store_aligned(&res[i]);
+            }
+            auto end = std::chrono::steady_clock::now();
+            auto tmp = end - start;
+            t_res = tmp < t_res ? tmp : t_res;
+        }
+        return t_res;
+    }
+
+    template <class B, class F, class V>
+    duration_type benchmark_simd_unrolled(F f, V& lhs, V& rhs, V& res, std::size_t number)
+    {
+        std::size_t s = lhs.size();
+        std::size_t inc = 4 * B::size;
+        duration_type t_res = duration_type::max();
+        for (std::size_t count = 0; count < number; ++count)
+        {
+            auto start = std::chrono::steady_clock::now();
+            for (std::size_t i = 0; i <= (s - inc); i += inc)
+            {
+                size_t j = i + B::size;
+                size_t k = j + B::size;
+                size_t l = k + B::size;
+                B blhs = B::load_aligned(&lhs[i]),
+                  brhs = B::load_aligned(&rhs[i]),
+                  blhs2 = B::load_aligned(&lhs[j]),
+                  brhs2 = B::load_aligned(&rhs[j]);
+                B blhs3 = B::load_aligned(&lhs[k]),
+                  brhs3 = B::load_aligned(&rhs[k]),
+                  blhs4 = B::load_aligned(&lhs[l]),
+                  brhs4 = B::load_aligned(&rhs[l]);
+                B bres = f(blhs, brhs);
+                B bres2 = f(blhs2, brhs2);
+                B bres3 = f(blhs3, brhs3);
+                B bres4 = f(blhs4, brhs4);
+                bres.store_aligned(&res[i]);
+                bres2.store_aligned(&res[j]);
+                bres3.store_aligned(&res[k]);
+                bres4.store_aligned(&res[l]);
+            }
+            auto end = std::chrono::steady_clock::now();
+            auto tmp = end - start;
+            t_res = tmp < t_res ? tmp : t_res;
+        }
+        return t_res;
+    }
+
+    template <class B, class F, class V>
+    duration_type benchmark_simd(F f, V& op0, V& op1, V& op2, V& res, std::size_t number)
+    {
+        std::size_t s = op0.size();
+        duration_type t_res = duration_type::max();
+        for (std::size_t count = 0; count < number; ++count)
+        {
+            auto start = std::chrono::steady_clock::now();
+            for (std::size_t i = 0; i <= (s - B::size); i += B::size)
+            {
+                B bop0 = B::load_aligned(&op0[i]),
+                  bop1 = B::load_aligned(&op1[i]),
+                  bop2 = B::load_aligned(&op2[i]);
+                B bres = f(bop0, bop1, bop2);
+                bres.store_aligned(&res[i]);
+            }
+            auto end = std::chrono::steady_clock::now();
+            auto tmp = end - start;
+            t_res = tmp < t_res ? tmp : t_res;
+        }
+        return t_res;
+    }
+
+    template <class B, class F, class V>
+    duration_type benchmark_simd_unrolled(F f, V& op0, V& op1, V& op2, V& res, std::size_t number)
+    {
+        std::size_t s = op0.size();
+        std::size_t inc = 4 * B::size;
+        duration_type t_res = duration_type::max();
+        for (std::size_t count = 0; count < number; ++count)
+        {
+            auto start = std::chrono::steady_clock::now();
+            for (std::size_t i = 0; i <= (s - inc); i += inc)
+            {
+                size_t j = i + B::size;
+                size_t k = j + B::size;
+                size_t l = k + B::size;
+                B bop0_i = B::load_aligned(&op0[i]),
+                  bop1_i = B::load_aligned(&op1[i]),
+                  bop2_i = B::load_aligned(&op2[i]);
+                B bop0_j = B::load_aligned(&op0[j]),
+                  bop1_j = B::load_aligned(&op1[j]),
+                  bop2_j = B::load_aligned(&op2[j]);
+                B bop0_k = B::load_aligned(&op0[k]),
+                  bop1_k = B::load_aligned(&op1[k]),
+                  bop2_k = B::load_aligned(&op2[k]);
+                B bop0_l = B::load_aligned(&op0[l]),
+                  bop1_l = B::load_aligned(&op1[l]),
+                  bop2_l = B::load_aligned(&op2[l]);
+                B bres_i = f(bop0_i, bop1_i, bop2_i);
+                B bres_j = f(bop0_j, bop1_j, bop2_j);
+                B bres_k = f(bop0_k, bop1_k, bop2_k);
+                B bres_l = f(bop0_l, bop1_l, bop2_l);
+                bres_i.store_aligned(&res[i]);
+                bres_j.store_aligned(&res[j]);
+                bres_k.store_aligned(&res[k]);
+                bres_l.store_aligned(&res[l]);
+            }
+            auto end = std::chrono::steady_clock::now();
+            auto tmp = end - start;
+            t_res = tmp < t_res ? tmp : t_res;
+        }
+        return t_res;
+    }
+
+    template <class F, class OS>
+    void run_benchmark_1op(F f, OS& out, std::size_t size, std::size_t iter, init_method init = init_method::classic)
+    {
+        bench_vector<float> f_lhs, f_rhs, f_res;
+        bench_vector<double> d_lhs, d_rhs, d_res;
+
+        switch (init)
+        {
+        case init_method::classic:
+            init_benchmark(f_lhs, f_rhs, f_res, size);
+            init_benchmark(d_lhs, d_rhs, d_res, size);
+            break;
+        case init_method::arctrigo:
+            init_benchmark_arctrigo(f_lhs, f_rhs, f_res, size);
+            init_benchmark_arctrigo(d_lhs, d_rhs, d_res, size);
+            break;
+        default:
+            init_benchmark(f_lhs, f_rhs, f_res, size);
+            init_benchmark(d_lhs, d_rhs, d_res, size);
+            break;
+        }
+
+#ifndef XSIMD_POLY_BENCHMARKS
+        duration_type t_float_scalar = benchmark_scalar(f, f_lhs, f_res, iter);
+        duration_type t_double_scalar = benchmark_scalar(f, d_lhs, d_res, iter);
+#endif
+
+        duration_type t_float_vector = benchmark_simd<batch<float>>(f, f_lhs, f_res, iter);
+        duration_type t_float_vector_u = benchmark_simd_unrolled<batch<float>>(f, f_lhs, f_res, iter);
+#if !XSIMD_WITH_NEON || XSIMD_WITH_NEON64
+        duration_type t_double_vector = benchmark_simd<batch<double>>(f, d_lhs, d_res, iter);
+        duration_type t_double_vector_u = benchmark_simd_unrolled<batch<double>>(f, d_lhs, d_res, iter);
+#endif
+
+        out << "============================" << std::endl;
+        out << f.name() << std::endl;
+#ifndef XSIMD_POLY_BENCHMARKS
+        out << "scalar float      : " << t_float_scalar.count() << "ms" << std::endl;
+#endif
+        out << "vector float      : " << t_float_vector.count() << "ms" << std::endl;
+        out << "vector float unr  : " << t_float_vector_u.count() << "ms" << std::endl;
+
+#ifndef XSIMD_POLY_BENCHMARKS
+        out << "scalar double     : " << t_double_scalar.count() << "ms" << std::endl;
+#endif
+#if !XSIMD_WITH_NEON || XSIMD_WITH_NEON64
+        out << "vector double     : " << t_double_vector.count() << "ms" << std::endl;
+        out << "vector double unr : " << t_double_vector_u.count() << "ms" << std::endl;
+#endif
+        out << "============================" << std::endl;
+    }
+
+    template <class F, class OS>
+    void run_benchmark_2op(F f, OS& out, std::size_t size, std::size_t iter)
+    {
+        bench_vector<float> f_lhs, f_rhs, f_res;
+        bench_vector<double> d_lhs, d_rhs, d_res;
+
+        init_benchmark(f_lhs, f_rhs, f_res, size);
+        init_benchmark(d_lhs, d_rhs, d_res, size);
+
+        duration_type t_float_scalar = benchmark_scalar(f, f_lhs, f_rhs, f_res, iter);
+        duration_type t_float_vector = benchmark_simd<batch<float>>(f, f_lhs, f_rhs, f_res, iter);
+        duration_type t_float_vector_u = benchmark_simd_unrolled<batch<float>>(f, f_lhs, f_rhs, f_res, iter);
+
+        duration_type t_double_scalar = benchmark_scalar(f, d_lhs, d_rhs, d_res, iter);
+#if !XSIMD_WITH_NEON || XSIMD_WITH_NEON64
+        duration_type t_double_vector = benchmark_simd<batch<double>>(f, d_lhs, d_rhs, d_res, iter);
+        duration_type t_double_vector_u = benchmark_simd_unrolled<batch<double>>(f, d_lhs, d_rhs, d_res, iter);
+#endif
+
+        out << "============================" << std::endl;
+        out << f.name() << std::endl;
+        out << "scalar float      : " << t_float_scalar.count() << "ms" << std::endl;
+        out << "vector float      : " << t_float_vector.count() << "ms" << std::endl;
+        out << "vector float unr  : " << t_float_vector_u.count() << "ms" << std::endl;
+        out << "scalar double     : " << t_double_scalar.count() << "ms" << std::endl;
+#if !XSIMD_WITH_NEON || XSIMD_WITH_NEON64
+        out << "vector double     : " << t_double_vector.count() << "ms" << std::endl;
+        out << "vector double unr : " << t_double_vector_u.count() << "ms" << std::endl;
+#endif
+        out << "============================" << std::endl;
+    }
+
+    template <class F, class OS>
+    void run_benchmark_3op(F f, OS& out, std::size_t size, std::size_t iter)
+    {
+        bench_vector<float> f_op0, f_op1, f_op2, f_res;
+        bench_vector<double> d_op0, d_op1, d_op2, d_res;
+
+        init_benchmark(f_op0, f_op1, f_op2, f_res, size);
+        init_benchmark(d_op0, d_op1, d_op2, d_res, size);
+
+        duration_type t_float_scalar = benchmark_scalar(f, f_op0, f_op1, f_op2, f_res, iter);
+        duration_type t_float_vector = benchmark_simd<batch<float>>(f, f_op0, f_op1, f_op2, f_res, iter);
+        duration_type t_float_vector_u = benchmark_simd_unrolled<batch<float>>(f, f_op0, f_op1, f_op2, f_res, iter);
+        duration_type t_double_scalar = benchmark_scalar(f, d_op0, d_op1, d_op2, d_res, iter);
+#if !XSIMD_WITH_NEON || XSIMD_WITH_NEON64
+        duration_type t_double_vector = benchmark_simd<batch<double>>(f, d_op0, d_op1, d_op2, d_res, iter);
+        duration_type t_double_vector_u = benchmark_simd_unrolled<batch<double>>(f, d_op0, d_op1, d_op2, d_res, iter);
+#endif
+
+        out << "============================" << std::endl;
+        out << f.name() << std::endl;
+        out << "scalar float      : " << t_float_scalar.count() << "ms" << std::endl;
+        out << "vector float      : " << t_float_vector.count() << "ms" << std::endl;
+        out << "vector float unr  : " << t_float_vector_u.count() << "ms" << std::endl;
+        out << "scalar double     : " << t_double_scalar.count() << "ms" << std::endl;
+#if !XSIMD_WITH_NEON || XSIMD_WITH_NEON64
+        out << "vector double     : " << t_double_vector.count() << "ms" << std::endl;
+        out << "vector double unr : " << t_double_vector_u.count() << "ms" << std::endl;
+#endif
+        out << "============================" << std::endl;
+    }
+
+#define DEFINE_OP_FUNCTOR_2OP(OP, NAME)                       \
+    struct NAME##_fn                                          \
+    {                                                         \
+        template <class T>                                    \
+        inline T operator()(const T& lhs, const T& rhs) const \
+        {                                                     \
+            return lhs OP rhs;                                \
+        }                                                     \
+        inline std::string name() const                       \
+        {                                                     \
+            return #NAME;                                     \
+        }                                                     \
+    }
+
+#define DEFINE_FUNCTOR_1OP(FN)                \
+    struct FN##_fn                            \
+    {                                         \
+        template <class T>                    \
+        inline T operator()(const T& x) const \
+        {                                     \
+            using xsimd::FN;                  \
+            return FN(x);                     \
+        }                                     \
+        inline std::string name() const       \
+        {                                     \
+            return #FN;                       \
+        }                                     \
+    }
+
+#define DEFINE_FUNCTOR_1OP_TEMPLATE(NAME, FN, N, ...) \
+    struct NAME##_##N##_fn                            \
+    {                                                 \
+        template <class T>                            \
+        inline T operator()(const T& x) const         \
+        {                                             \
+            using xsimd::FN;                          \
+            return FN<T, __VA_ARGS__>(x);             \
+        }                                             \
+        inline std::string name() const               \
+        {                                             \
+            return #FN " " #N;                        \
+        }                                             \
+    }
+
+#define DEFINE_FUNCTOR_2OP(FN)                                \
+    struct FN##_fn                                            \
+    {                                                         \
+        template <class T>                                    \
+        inline T operator()(const T& lhs, const T& rhs) const \
+        {                                                     \
+            using xsimd::FN;                                  \
+            return FN(lhs, rhs);                              \
+        }                                                     \
+        inline std::string name() const                       \
+        {                                                     \
+            return #FN;                                       \
+        }                                                     \
+    }
+
+#define DEFINE_FUNCTOR_3OP(FN)                                              \
+    struct FN##_fn                                                          \
+    {                                                                       \
+        template <class T>                                                  \
+        inline T operator()(const T& op0, const T& op1, const T& op2) const \
+        {                                                                   \
+            using xsimd::FN;                                                \
+            return FN(op0, op1, op2);                                       \
+        }                                                                   \
+        inline std::string name() const                                     \
+        {                                                                   \
+            return #FN;                                                     \
+        }                                                                   \
+    }
+
+    DEFINE_OP_FUNCTOR_2OP(+, add);
+    DEFINE_OP_FUNCTOR_2OP(-, sub);
+    DEFINE_OP_FUNCTOR_2OP(*, mul);
+    DEFINE_OP_FUNCTOR_2OP(/, div);
+
+    DEFINE_FUNCTOR_1OP(exp);
+    DEFINE_FUNCTOR_1OP(exp2);
+    DEFINE_FUNCTOR_1OP(expm1);
+    DEFINE_FUNCTOR_1OP(log);
+    DEFINE_FUNCTOR_1OP(log10);
+    DEFINE_FUNCTOR_1OP(log2);
+    DEFINE_FUNCTOR_1OP(log1p);
+
+    DEFINE_FUNCTOR_1OP(sin);
+    DEFINE_FUNCTOR_1OP(cos);
+    DEFINE_FUNCTOR_1OP(tan);
+    DEFINE_FUNCTOR_1OP(asin);
+    DEFINE_FUNCTOR_1OP(acos);
+    DEFINE_FUNCTOR_1OP(atan);
+
+    DEFINE_FUNCTOR_1OP(sinh);
+    DEFINE_FUNCTOR_1OP(cosh);
+    DEFINE_FUNCTOR_1OP(tanh);
+    DEFINE_FUNCTOR_1OP(asinh);
+    DEFINE_FUNCTOR_1OP(acosh);
+    DEFINE_FUNCTOR_1OP(atanh);
+
+    DEFINE_FUNCTOR_2OP(pow);
+    DEFINE_FUNCTOR_1OP(sqrt);
+    DEFINE_FUNCTOR_1OP(cbrt);
+    DEFINE_FUNCTOR_2OP(hypot);
+
+    DEFINE_FUNCTOR_1OP(ceil);
+    DEFINE_FUNCTOR_1OP(floor);
+    DEFINE_FUNCTOR_1OP(trunc);
+    DEFINE_FUNCTOR_1OP(round);
+    DEFINE_FUNCTOR_1OP(nearbyint);
+    DEFINE_FUNCTOR_1OP(rint);
+
+    DEFINE_FUNCTOR_2OP(fmod);
+    DEFINE_FUNCTOR_2OP(remainder);
+    DEFINE_FUNCTOR_2OP(fdim);
+    DEFINE_FUNCTOR_3OP(clip);
+#if 0
+DEFINE_FUNCTOR_1OP(isfinite);
+DEFINE_FUNCTOR_1OP(isinf);
+DEFINE_FUNCTOR_1OP(is_flint);
+DEFINE_FUNCTOR_1OP(is_odd);
+DEFINE_FUNCTOR_1OP(is_even);
+#endif
+
+    DEFINE_FUNCTOR_1OP_TEMPLATE(horner, kernel::horner, 5, 1, 2, 3, 4, 5);
+    DEFINE_FUNCTOR_1OP_TEMPLATE(estrin, kernel::estrin, 5, 1, 2, 3, 4, 5);
+    DEFINE_FUNCTOR_1OP_TEMPLATE(horner, kernel::horner, 10, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10);
+    DEFINE_FUNCTOR_1OP_TEMPLATE(estrin, kernel::estrin, 10, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10);
+    DEFINE_FUNCTOR_1OP_TEMPLATE(horner, kernel::horner, 12, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12);
+    DEFINE_FUNCTOR_1OP_TEMPLATE(estrin, kernel::estrin, 12, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12);
+    DEFINE_FUNCTOR_1OP_TEMPLATE(horner, kernel::horner, 14, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14);
+    DEFINE_FUNCTOR_1OP_TEMPLATE(estrin, kernel::estrin, 14, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14);
+    DEFINE_FUNCTOR_1OP_TEMPLATE(horner, kernel::horner, 16, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16);
+    DEFINE_FUNCTOR_1OP_TEMPLATE(estrin, kernel::estrin, 16, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16);
+
+}
+#endif

sequenzo/dissimilarity_measures/src/xsimd/docs/source/conf.py

@@ -0,0 +1,37 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+
+import os
+import subprocess
+
+on_rtd = os.environ.get('READTHEDOCS', None) == 'True'
+
+if on_rtd:
+    subprocess.check_call('cd ..; doxygen', shell=True)
+
+import sphinx_rtd_theme
+
+html_theme = "sphinx_rtd_theme"
+html_theme_path = [sphinx_rtd_theme.get_html_theme_path()]
+
+def setup(app):
+    app.add_css_file("main_stylesheet.css")
+
+extensions = ['breathe', 'sphinx_rtd_theme']
+breathe_projects = { 'xsimd': '../xml' }
+templates_path = ['_templates']
+html_static_path = ['_static']
+source_suffix = '.rst'
+master_doc = 'index'
+project = 'xsimd'
+copyright = '2016, Johan Mabille and Sylvain Corlay'
+author = 'Johan Mabille and Sylvain Corlay'
+
+html_logo = 'quantstack-white.svg'
+
+exclude_patterns = []
+highlight_language = 'c++'
+pygments_style = 'sphinx'
+todo_include_todos = False
+htmlhelp_basename = 'xsimddoc'
+