superbuild/modules/openal-soft/Alc/mixer/mixer_neon.cpp

#include "config.h"

#include <arm_neon.h>

#include <limits>

#include "AL/al.h"
#include "AL/alc.h"
#include "alMain.h"
#include "alu.h"
#include "hrtf.h"
#include "defs.h"
#include "hrtfbase.h"



template<>
const ALfloat *Resample_<LerpTag,NEONTag>(const InterpState* UNUSED(state),
  const ALfloat *RESTRICT src, ALsizei frac, ALint increment,
  ALfloat *RESTRICT dst, ALsizei dstlen)
{
    const int32x4_t increment4 = vdupq_n_s32(increment*4);
    const float32x4_t fracOne4 = vdupq_n_f32(1.0f/FRACTIONONE);
    const int32x4_t fracMask4 = vdupq_n_s32(FRACTIONMASK);
    alignas(16) ALsizei pos_[4], frac_[4];
    int32x4_t pos4, frac4;
    ALsizei todo, pos, i;

    ASSUME(frac >= 0);
    ASSUME(increment > 0);
    ASSUME(dstlen > 0);

    InitiatePositionArrays(frac, increment, frac_, pos_, 4);
    frac4 = vld1q_s32(frac_);
    pos4 = vld1q_s32(pos_);

    todo = dstlen & ~3;
    for(i = 0;i < todo;i += 4)
    {
        const int pos0 = vgetq_lane_s32(pos4, 0);
        const int pos1 = vgetq_lane_s32(pos4, 1);
        const int pos2 = vgetq_lane_s32(pos4, 2);
        const int pos3 = vgetq_lane_s32(pos4, 3);
        const float32x4_t val1 = (float32x4_t){src[pos0], src[pos1], src[pos2], src[pos3]};
        const float32x4_t val2 = (float32x4_t){src[pos0+1], src[pos1+1], src[pos2+1], src[pos3+1]};

        /* val1 + (val2-val1)*mu */
        const float32x4_t r0 = vsubq_f32(val2, val1);
        const float32x4_t mu = vmulq_f32(vcvtq_f32_s32(frac4), fracOne4);
        const float32x4_t out = vmlaq_f32(val1, mu, r0);

        vst1q_f32(&dst[i], out);

        frac4 = vaddq_s32(frac4, increment4);
        pos4 = vaddq_s32(pos4, vshrq_n_s32(frac4, FRACTIONBITS));
        frac4 = vandq_s32(frac4, fracMask4);
    }

    /* NOTE: These four elements represent the position *after* the last four
     * samples, so the lowest element is the next position to resample.
     */
    pos = vgetq_lane_s32(pos4, 0);
    frac = vgetq_lane_s32(frac4, 0);

    for(;i < dstlen;++i)
    {
        dst[i] = lerp(src[pos], src[pos+1], frac * (1.0f/FRACTIONONE));

        frac += increment;
        pos  += frac>>FRACTIONBITS;
        frac &= FRACTIONMASK;
    }
    return dst;
}

template<>
const ALfloat *Resample_<BSincTag,NEONTag>(const InterpState *state, const ALfloat *RESTRICT src,
    ALsizei frac, ALint increment, ALfloat *RESTRICT dst, ALsizei dstlen)
{
    const ALfloat *const filter = state->bsinc.filter;
    const float32x4_t sf4 = vdupq_n_f32(state->bsinc.sf);
    const ALsizei m = state->bsinc.m;
    const float32x4_t *fil, *scd, *phd, *spd;
    ALsizei pi, i, j, offset;
    float32x4_t r4;
    ALfloat pf;

    ASSUME(m > 0);
    ASSUME(dstlen > 0);
    ASSUME(increment > 0);
    ASSUME(frac >= 0);

    src -= state->bsinc.l;
    for(i = 0;i < dstlen;i++)
    {
        // Calculate the phase index and factor.
#define FRAC_PHASE_BITDIFF (FRACTIONBITS-BSINC_PHASE_BITS)
        pi = frac >> FRAC_PHASE_BITDIFF;
        pf = (frac & ((1<<FRAC_PHASE_BITDIFF)-1)) * (1.0f/(1<<FRAC_PHASE_BITDIFF));
#undef FRAC_PHASE_BITDIFF

        offset = m*pi*4;
        fil = (const float32x4_t*)(filter + offset); offset += m;
        scd = (const float32x4_t*)(filter + offset); offset += m;
        phd = (const float32x4_t*)(filter + offset); offset += m;
        spd = (const float32x4_t*)(filter + offset);

        // Apply the scale and phase interpolated filter.
        r4 = vdupq_n_f32(0.0f);
        {
            const ALsizei count = m >> 2;
            const float32x4_t pf4 = vdupq_n_f32(pf);

            ASSUME(count > 0);

            for(j = 0;j < count;j++)
            {
                /* f = ((fil + sf*scd) + pf*(phd + sf*spd)) */
                const float32x4_t f4 = vmlaq_f32(
                    vmlaq_f32(fil[j], sf4, scd[j]),
                    pf4, vmlaq_f32(phd[j], sf4, spd[j])
                );
                /* r += f*src */
                r4 = vmlaq_f32(r4, f4, vld1q_f32(&src[j*4]));
            }
        }
        r4 = vaddq_f32(r4, vcombine_f32(vrev64_f32(vget_high_f32(r4)),
                                        vrev64_f32(vget_low_f32(r4))));
        dst[i] = vget_lane_f32(vadd_f32(vget_low_f32(r4), vget_high_f32(r4)), 0);

        frac += increment;
        src  += frac>>FRACTIONBITS;
        frac &= FRACTIONMASK;
    }
    return dst;
}


static inline void ApplyCoeffs(ALsizei /*Offset*/, float2 *RESTRICT Values, const ALsizei IrSize,
    const HrirArray<ALfloat> &Coeffs, const ALfloat left, const ALfloat right)
{
    ASSUME(IrSize >= 2);

    float32x4_t leftright4;
    {
        float32x2_t leftright2 = vdup_n_f32(0.0);
        leftright2 = vset_lane_f32(left, leftright2, 0);
        leftright2 = vset_lane_f32(right, leftright2, 1);
        leftright4 = vcombine_f32(leftright2, leftright2);
    }

    for(ALsizei c{0};c < IrSize;c += 2)
    {
        float32x4_t vals = vcombine_f32(vld1_f32((float32_t*)&Values[c  ][0]),
                                        vld1_f32((float32_t*)&Values[c+1][0]));
        float32x4_t coefs = vld1q_f32((float32_t*)&Coeffs[c][0]);

        vals = vmlaq_f32(vals, coefs, leftright4);

        vst1_f32((float32_t*)&Values[c  ][0], vget_low_f32(vals));
        vst1_f32((float32_t*)&Values[c+1][0], vget_high_f32(vals));
    }
}

template<>
void MixHrtf_<NEONTag>(ALfloat *RESTRICT LeftOut, ALfloat *RESTRICT RightOut, const ALfloat *data,
    float2 *RESTRICT AccumSamples, const ALsizei OutPos, const ALsizei IrSize,
    MixHrtfParams *hrtfparams, const ALsizei BufferSize)
{
    MixHrtfBase<ApplyCoeffs>(LeftOut, RightOut, data, AccumSamples, OutPos, IrSize, hrtfparams,
        BufferSize);
}

template<>
void MixHrtfBlend_<NEONTag>(ALfloat *RESTRICT LeftOut, ALfloat *RESTRICT RightOut,
    const ALfloat *data, float2 *RESTRICT AccumSamples, const ALsizei OutPos, const ALsizei IrSize,
    const HrtfParams *oldparams, MixHrtfParams *newparams, const ALsizei BufferSize)
{
    MixHrtfBlendBase<ApplyCoeffs>(LeftOut, RightOut, data, AccumSamples, OutPos, IrSize, oldparams,
        newparams, BufferSize);
}

template<>
void MixDirectHrtf_<NEONTag>(ALfloat *RESTRICT LeftOut, ALfloat *RESTRICT RightOut,
    const ALfloat (*data)[BUFFERSIZE], float2 *RESTRICT AccumSamples, DirectHrtfState *State,
    const ALsizei NumChans, const ALsizei BufferSize)
{
    MixDirectHrtfBase<ApplyCoeffs>(LeftOut, RightOut, data, AccumSamples, State, NumChans,
        BufferSize);
}


template<>
void Mix_<NEONTag>(const ALfloat *data, const ALsizei OutChans, ALfloat (*OutBuffer)[BUFFERSIZE],
    ALfloat *CurrentGains, const ALfloat *TargetGains, const ALsizei Counter, const ALsizei OutPos,
    const ALsizei BufferSize)
{
    ASSUME(OutChans > 0);
    ASSUME(BufferSize > 0);

    const ALfloat delta{(Counter > 0) ? 1.0f/(ALfloat)Counter : 0.0f};
    for(ALsizei c{0};c < OutChans;c++)
    {
        ALfloat *RESTRICT dst{al::assume_aligned<16>(&OutBuffer[c][OutPos])};
        ALsizei pos{0};
        ALfloat gain{CurrentGains[c]};
        const ALfloat diff{TargetGains[c] - gain};

        if(std::fabs(diff) > std::numeric_limits<float>::epsilon())
        {
            ALsizei minsize{mini(BufferSize, Counter)};
            const ALfloat step{diff * delta};
            ALfloat step_count{0.0f};
            /* Mix with applying gain steps in aligned multiples of 4. */
            if(LIKELY(minsize > 3))
            {
                const float32x4_t four4{vdupq_n_f32(4.0f)};
                const float32x4_t step4{vdupq_n_f32(step)};
                const float32x4_t gain4{vdupq_n_f32(gain)};
                float32x4_t step_count4{vsetq_lane_f32(0.0f,
                    vsetq_lane_f32(1.0f,
                    vsetq_lane_f32(2.0f,
                    vsetq_lane_f32(3.0f, vdupq_n_f32(0.0f), 3),
                    2), 1), 0
                )};
                ALsizei todo{minsize >> 2};

                do {
                    const float32x4_t val4 = vld1q_f32(&data[pos]);
                    float32x4_t dry4 = vld1q_f32(&dst[pos]);
                    dry4 = vmlaq_f32(dry4, val4, vmlaq_f32(gain4, step4, step_count4));
                    step_count4 = vaddq_f32(step_count4, four4);
                    vst1q_f32(&dst[pos], dry4);
                    pos += 4;
                } while(--todo);
                /* NOTE: step_count4 now represents the next four counts after
                 * the last four mixed samples, so the lowest element
                 * represents the next step count to apply.
                 */
                step_count = vgetq_lane_f32(step_count4, 0);
            }
            /* Mix with applying left over gain steps that aren't aligned multiples of 4. */
            for(;pos < minsize;pos++)
            {
                dst[pos] += data[pos]*(gain + step*step_count);
                step_count += 1.0f;
            }
            if(pos == Counter)
                gain = TargetGains[c];
            else
                gain += step*step_count;
            CurrentGains[c] = gain;

            /* Mix until pos is aligned with 4 or the mix is done. */
            minsize = mini(BufferSize, (pos+3)&~3);
            for(;pos < minsize;pos++)
                dst[pos] += data[pos]*gain;
        }

        if(!(std::fabs(gain) > GAIN_SILENCE_THRESHOLD))
            continue;
        if(LIKELY(BufferSize-pos > 3))
        {
            ALsizei todo{(BufferSize-pos) >> 2};
            const float32x4_t gain4 = vdupq_n_f32(gain);
            do {
                const float32x4_t val4 = vld1q_f32(&data[pos]);
                float32x4_t dry4 = vld1q_f32(&dst[pos]);
                dry4 = vmlaq_f32(dry4, val4, gain4);
                vst1q_f32(&dst[pos], dry4);
                pos += 4;
            } while(--todo);
        }
        for(;pos < BufferSize;pos++)
            dst[pos] += data[pos]*gain;
    }
}

template<>
void MixRow_<NEONTag>(ALfloat *OutBuffer, const ALfloat *Gains, const ALfloat (*data)[BUFFERSIZE],
    const ALsizei InChans, const ALsizei InPos, const ALsizei BufferSize)
{
    ASSUME(InChans > 0);
    ASSUME(BufferSize > 0);

    for(ALsizei c{0};c < InChans;c++)
    {
        const ALfloat *RESTRICT src{al::assume_aligned<16>(&data[c][InPos])};
        ALsizei pos{0};
        const ALfloat gain{Gains[c]};
        if(!(std::fabs(gain) > GAIN_SILENCE_THRESHOLD))
            continue;

        if(LIKELY(BufferSize > 3))
        {
            ALsizei todo{BufferSize >> 2};
            float32x4_t gain4{vdupq_n_f32(gain)};
            do {
                const float32x4_t val4 = vld1q_f32(&src[pos]);
                float32x4_t dry4 = vld1q_f32(&OutBuffer[pos]);
                dry4 = vmlaq_f32(dry4, val4, gain4);
                vst1q_f32(&OutBuffer[pos], dry4);
                pos += 4;
            } while(--todo);
        }
        for(;pos < BufferSize;pos++)
            OutBuffer[pos] += src[pos]*gain;
    }
}