love2d.megasource/libs/openal-soft/common/phase_shifter.h

#ifndef PHASE_SHIFTER_H
#define PHASE_SHIFTER_H

#include "config_simd.h"

#if HAVE_SSE_INTRINSICS
#include <xmmintrin.h>
#elif HAVE_NEON
#include <arm_neon.h>
#endif

#include <array>
#include <cmath>
#include <cstddef>

#include "alnumbers.h"
#include "alspan.h"
#include "opthelpers.h"


/* Implements a wide-band +90 degree phase-shift. Note that this should be
 * given one sample less of a delay (FilterSize/2 - 1) compared to the direct
 * signal delay (FilterSize/2) to properly align.
 */
template<std::size_t FilterSize>
struct SIMDALIGN PhaseShifterT {
    static_assert(FilterSize >= 16, "FilterSize needs to be at least 16");
    static_assert((FilterSize&(FilterSize-1)) == 0, "FilterSize needs to be power-of-two");

    alignas(16) std::array<float,FilterSize/2> mCoeffs{};

    PhaseShifterT()
    {
        /* Every other coefficient is 0, so we only need to calculate and store
         * the non-0 terms and double-step over the input to apply it. The
         * calculated coefficients are in reverse to make applying in the time-
         * domain more efficient.
         */
        for(std::size_t i{0};i < FilterSize/2;++i)
        {
            const auto k = static_cast<int>(i*2 + 1) - int{FilterSize/2};

            /* Calculate the Blackman window value for this coefficient. */
            const auto w = 2.0*al::numbers::pi/double{FilterSize} * static_cast<double>(i*2 + 1);
            const auto window = 0.3635819 - 0.4891775*std::cos(w) + 0.1365995*std::cos(2.0*w)
                - 0.0106411*std::cos(3.0*w);

            const auto pk = al::numbers::pi * static_cast<double>(k);
            mCoeffs[i] = static_cast<float>(window * (1.0-std::cos(pk)) / pk);
        }
    }

    void process(const al::span<float> dst, const al::span<const float> src) const;

private:
#if HAVE_NEON
    static auto load4(float32_t a, float32_t b, float32_t c, float32_t d)
    {
        float32x4_t ret{vmovq_n_f32(a)};
        ret = vsetq_lane_f32(b, ret, 1);
        ret = vsetq_lane_f32(c, ret, 2);
        ret = vsetq_lane_f32(d, ret, 3);
        return ret;
    }
    static void vtranspose4(float32x4_t &x0, float32x4_t &x1, float32x4_t &x2, float32x4_t &x3)
    {
        float32x4x2_t t0_{vzipq_f32(x0, x2)};
        float32x4x2_t t1_{vzipq_f32(x1, x3)};
        float32x4x2_t u0_{vzipq_f32(t0_.val[0], t1_.val[0])};
        float32x4x2_t u1_{vzipq_f32(t0_.val[1], t1_.val[1])};
        x0 = u0_.val[0];
        x1 = u0_.val[1];
        x2 = u1_.val[0];
        x3 = u1_.val[1];
    }
#endif
};

template<std::size_t S>
NOINLINE inline
void PhaseShifterT<S>::process(const al::span<float> dst, const al::span<const float> src) const
{
    auto in = src.begin();
#if HAVE_SSE_INTRINSICS
    if(const std::size_t todo{dst.size()>>2})
    {
        auto out = al::span{reinterpret_cast<__m128*>(dst.data()), todo};
        std::generate(out.begin(), out.end(), [&in,this]
        {
            __m128 r0{_mm_setzero_ps()};
            __m128 r1{_mm_setzero_ps()};
            __m128 r2{_mm_setzero_ps()};
            __m128 r3{_mm_setzero_ps()};
            for(std::size_t j{0};j < mCoeffs.size();j+=4)
            {
                const __m128 coeffs{_mm_load_ps(&mCoeffs[j])};
                const __m128 s0{_mm_loadu_ps(&in[j*2])};
                const __m128 s1{_mm_loadu_ps(&in[j*2 + 4])};
                const __m128 s2{_mm_movehl_ps(_mm_movelh_ps(s1, s1), s0)};
                const __m128 s3{_mm_loadh_pi(_mm_movehl_ps(s1, s1),
                    reinterpret_cast<const __m64*>(&in[j*2 + 8]))};

                __m128 s{_mm_shuffle_ps(s0, s1, _MM_SHUFFLE(2, 0, 2, 0))};
                r0 = _mm_add_ps(r0, _mm_mul_ps(s, coeffs));

                s = _mm_shuffle_ps(s0, s1, _MM_SHUFFLE(3, 1, 3, 1));
                r1 = _mm_add_ps(r1, _mm_mul_ps(s, coeffs));

                s = _mm_shuffle_ps(s2, s3, _MM_SHUFFLE(2, 0, 2, 0));
                r2 = _mm_add_ps(r2, _mm_mul_ps(s, coeffs));

                s = _mm_shuffle_ps(s2, s3, _MM_SHUFFLE(3, 1, 3, 1));
                r3 = _mm_add_ps(r3, _mm_mul_ps(s, coeffs));
            }
            in += 4;

            _MM_TRANSPOSE4_PS(r0, r1, r2, r3);
            return _mm_add_ps(_mm_add_ps(r0, r1), _mm_add_ps(r2, r3));
        });
    }
    if(const std::size_t todo{dst.size()&3})
    {
        auto out = dst.last(todo);
        std::generate(out.begin(), out.end(), [&in,this]
        {
            __m128 r4{_mm_setzero_ps()};
            for(std::size_t j{0};j < mCoeffs.size();j+=4)
            {
                const __m128 coeffs{_mm_load_ps(&mCoeffs[j])};
                const __m128 s{_mm_setr_ps(in[j*2], in[j*2 + 2], in[j*2 + 4], in[j*2 + 6])};
                r4 = _mm_add_ps(r4, _mm_mul_ps(s, coeffs));
            }
            ++in;
            r4 = _mm_add_ps(r4, _mm_shuffle_ps(r4, r4, _MM_SHUFFLE(0, 1, 2, 3)));
            r4 = _mm_add_ps(r4, _mm_movehl_ps(r4, r4));
            return _mm_cvtss_f32(r4);
        });
    }

#elif HAVE_NEON

    if(const std::size_t todo{dst.size()>>2})
    {
        auto out = al::span{reinterpret_cast<float32x4_t*>(dst.data()), todo};
        std::generate(out.begin(), out.end(), [&in,this]
        {
            float32x4_t r0{vdupq_n_f32(0.0f)};
            float32x4_t r1{vdupq_n_f32(0.0f)};
            float32x4_t r2{vdupq_n_f32(0.0f)};
            float32x4_t r3{vdupq_n_f32(0.0f)};
            for(std::size_t j{0};j < mCoeffs.size();j+=4)
            {
                const float32x4_t coeffs{vld1q_f32(&mCoeffs[j])};
                const float32x4_t s0{vld1q_f32(&in[j*2])};
                const float32x4_t s1{vld1q_f32(&in[j*2 + 4])};
                const float32x4_t s2{vcombine_f32(vget_high_f32(s0), vget_low_f32(s1))};
                const float32x4_t s3{vcombine_f32(vget_high_f32(s1), vld1_f32(&in[j*2 + 8]))};
                const float32x4x2_t values0{vuzpq_f32(s0, s1)};
                const float32x4x2_t values1{vuzpq_f32(s2, s3)};

                r0 = vmlaq_f32(r0, values0.val[0], coeffs);
                r1 = vmlaq_f32(r1, values0.val[1], coeffs);
                r2 = vmlaq_f32(r2, values1.val[0], coeffs);
                r3 = vmlaq_f32(r3, values1.val[1], coeffs);
            }
            in += 4;

            vtranspose4(r0, r1, r2, r3);
            return vaddq_f32(vaddq_f32(r0, r1), vaddq_f32(r2, r3));
        });
    }
    if(const std::size_t todo{dst.size()&3})
    {
        auto out = dst.last(todo);
        std::generate(out.begin(), out.end(), [&in,this]
        {
            float32x4_t r4{vdupq_n_f32(0.0f)};
            for(std::size_t j{0};j < mCoeffs.size();j+=4)
            {
                const float32x4_t coeffs{vld1q_f32(&mCoeffs[j])};
                const float32x4_t s{load4(in[j*2], in[j*2 + 2], in[j*2 + 4], in[j*2 + 6])};
                r4 = vmlaq_f32(r4, s, coeffs);
            }
            ++in;
            r4 = vaddq_f32(r4, vrev64q_f32(r4));
            return vget_lane_f32(vadd_f32(vget_low_f32(r4), vget_high_f32(r4)), 0);
        });
    }

#else

    std::generate(dst.begin(), dst.end(), [&in,this]
    {
        float ret{0.0f};
        for(std::size_t j{0};j < mCoeffs.size();++j)
            ret += in[j*2] * mCoeffs[j];
        ++in;
        return ret;
    });
#endif
}

#endif /* PHASE_SHIFTER_H */