path: root/src/libultra/audio/reverb.c

#include "libaudio.h"
#include "synthInternals.h"
#include "stdbool.h"
#include "stddef.h"
#include "stdint.h"
#include "ultra64/convert.h"

#define RANGE 2.0

extern ALGlobals* alGlobals;

Acmd* _loadOutputBuffer(ALFx* r, ALDelay* d, s32 buff, s32 incount, Acmd* p);
Acmd* _loadBuffer(ALFx* r, s16* curr_ptr, s32 buff, s32 count, Acmd* p);
Acmd* _saveBuffer(ALFx* r, s16* curr_ptr, s32 buff, s32 count, Acmd* p);
Acmd* _filterBuffer(ALLowPass* lp, s32 buff, s32 count, Acmd* p);
f32 _doModFunc(ALDelay* d, s32 count);

Acmd* alFxPull(void* filter, s16* outp, s32 outCount, s32 sampleOffset, Acmd* p) {
    Acmd* ptr = p;
    ALFx* r = (ALFx*)filter;
    ALFilter* source = r->filter.source;
    s16 i;
    s16 buff1;
    s16 buff2;
    s16 input;
    s16 output;
    s16* in_ptr;
    s16* out_ptr;
    s16 gain;
    s16* prev_out_ptr = NULL;
    ALDelay* d;
    ALDelay* pd;

    // pull channels going into this effect first
    ptr = (*source->handler)(source, outp, outCount, sampleOffset, p);

    input = AL_AUX_L_OUT;
    output = AL_AUX_R_OUT;
    buff1 = AL_TEMP_0;
    buff2 = AL_TEMP_1;

    aSetBuffer(ptr++, 0, 0x000, 0x000, outCount << 1); // set the buffer size
    aMix(ptr++, 0, 0xDA83, AL_AUX_L_OUT, input);       // .707L = L - .293L
    aMix(ptr++, 0, 0x5A82, AL_AUX_R_OUT, input);       // mix the AuxL and AuxR into the AuxL
    // and write the mixed value to the delay line at r->input
    ptr = _saveBuffer(r, r->input, input, outCount, ptr);

    // clear the AL_AUX_R_OUT
    aClearBuffer(ptr++, output, outCount << 1);

    for (i = 0; i < r->section_count; i++) {
        d = &r->delay[i]; // get the ALDelay structure
        in_ptr = &r->input[-d->input];
        out_ptr = &r->input[-d->output];

        if (in_ptr == prev_out_ptr) {
            s16 t = buff2;
            buff2 = buff1;
            buff1 = t;
        } else {
            // load data at in_ptr into buff1
            ptr = _loadBuffer(r, in_ptr, buff1, outCount, ptr);
        }
        ptr = _loadOutputBuffer(r, d, buff2, outCount, ptr);

        if (d->ffcoef != 0) {
            aMix(ptr++, 0, (u16)d->ffcoef, buff1, buff2);
            if (d->rs == NULL && d->lp == NULL) {
                ptr = _saveBuffer(r, out_ptr, buff2, outCount, ptr);
            }
        }

        if (d->fbcoef != 0) {
            aMix(ptr++, 0, (u16)d->fbcoef, buff2, buff1);
            ptr = _saveBuffer(r, in_ptr, buff1, outCount, ptr);
        }

        if (d->lp != NULL) {
            ptr = _filterBuffer(d->lp, buff2, outCount, ptr);
        }

        if (d->rs == NULL) {
            ptr = _saveBuffer(r, out_ptr, buff2, outCount, ptr);
        }

        if (d->gain != 0) {
            aMix(ptr++, 0, (u16)d->gain, buff2, output);
        }

        prev_out_ptr = &r->input[d->output];
    }

    // bump the master delay line input pointer modulo the length
    r->input += outCount;
    if (r->input > &r->base[r->length]) {
        r->input -= r->length;
    }

    // output already in AL_AUX_R_OUT, just copy to AL_AUX_L_OUT
    aDMEMMove(ptr++, output, AL_AUX_L_OUT, outCount << 1);
    return ptr;
}

s32 alFxParam(void* filter, s32 paramID, void* param) {
    if (paramID == AL_FILTER_SET_SOURCE) {
        ALFilter* f = (ALFilter*)filter;
        f->source = (ALFilter*)param;
    }
    return 0;
}

/**
 * This routine gets called by alSynSetFXParam. No checking takes place to
 * verify the validity of the paramID or the param value. input and output
 * values must be 8 byte aligned, so round down any param passed.
 */
s32 alFxParamHdl(void* filter, s32 paramID, void* param) {
    ALFx* f = (ALFx*)filter;
    s32 p = (paramID - 2) % 8;
    s32 s = (paramID - 2) / 8;
    s32 val = *(s32*)param;

#define INPUT_PARAM 0
#define OUTPUT_PARAM 1
#define FBCOEF_PARAM 2
#define FFCOEF_PARAM 3
#define GAIN_PARAM 4
#define CHORUSRATE_PARAM 5
#define CHORUSDEPTH_PARAM 6
#define LPFILT_PARAM 7

    switch (p) {
        case INPUT_PARAM:
            f->delay[s].input = (u32)val & 0xFFFFFFF8;
            break;

        case OUTPUT_PARAM:
            f->delay[s].output = (u32)val & 0xFFFFFFF8;
            break;

        case FFCOEF_PARAM:
            f->delay[s].ffcoef = (s16)val;
            break;

        case FBCOEF_PARAM:
            f->delay[s].fbcoef = (s16)val;
            break;

        case GAIN_PARAM:
            f->delay[s].gain = (s16)val;
            break;

        case CHORUSRATE_PARAM:
            f->delay[s].rsinc = (((f32)val / 1000) * RANGE) / alGlobals->drvr.outputRate;
            break;

            // the following constant is derived from:
            //
            //      ratio = 2^(cents/1200)
            //
            // and therefore for hundredths of a cent
            //                         x
            //      ln(ratio) = ---------------
            //                  (120,000)/ln(2)
            // where
            //      120,000/ln(2) = 173123.40...
#define CONVERT 173123.404906676

        case CHORUSDEPTH_PARAM:
            f->delay[s].rsgain = ((f32)val / CONVERT) * (f->delay[s].output - f->delay[s].input);
            break;

        case LPFILT_PARAM:
            if (f->delay[s].lp != NULL) {
                f->delay[s].lp->fc = (s16)val;
                _init_lpfilter(f->delay[s].lp);
            }
            break;
    }
    return 0;
}

Acmd* _loadOutputBuffer(ALFx* r, ALDelay* d, s32 buff, s32 incount, Acmd* p) {
    Acmd* ptr = p;
    s32 ratio;
    s32 count;
    s32 rbuff = AL_TEMP_2;
    s16* out_ptr;
    f32 fincount;
    f32 fratio;
    f32 delta;
    s32 ramalign = 0;
    s32 length;

    // The following section implements the chorus resampling. Modulate where you pull
    // the samples from, since you need varying amounts of samples.
    if (d->rs != NULL) {
        length = d->output - d->input;
        // get the number of samples to modulate by
        delta = _doModFunc(d, incount);
        // find ratio of delta to delay length and quantize to same resolution as resampler
        delta /= length;                    // convert delta from number of samples to a pitch ratio
        delta = (s32)(delta * UNITY_PITCH); // quantize to value microcode will use
        delta = delta / UNITY_PITCH;
        // pitch ratio needs to be centered around 1, not zero
        fratio = 1.0 - delta;

        // d->rs->delta is the difference between the fractional and integer value of the samples needed.
        // fratio * incount + rs->delta gives the number of samples needed for this frame.
        fincount = d->rs->delta + (fratio * (f32)incount);
        count = (s32)fincount;                // quantize to s32
        d->rs->delta = fincount - (f32)count; // calculate the round off and store

        // d->rsdelta is amount the out_ptr has deviated from its starting position.
        // You calc the out_ptr by taking d->output - d->rsdelta, and then using the
        // negative of that as an index into the delay buffer. loadBuffer that uses this
        // value then bumps it up if it is below the  delay buffer.
        out_ptr = &r->input[-(d->output - d->rsdelta)];
        // calculate the number of samples needed to align the buffer
        ramalign = ((intptr_t)out_ptr & 7) >> 1;

        // load the rbuff with samples, note that there will be ramalign worth of samples at the
        // begining which you don't care about.
        ptr = _loadBuffer(r, out_ptr - ramalign, rbuff, count + ramalign, ptr);

        // convert fratio to 16 bit fraction for microcode use
        ratio = (s32)(fratio * UNITY_PITCH);
        // set the buffers, and do the resample
        aSetBuffer(ptr++, 0, rbuff + (ramalign << 1), buff, incount << 1);
        aResample(ptr++, d->rs->first, ratio, osVirtualToPhysical(d->rs->state));

        d->rs->first = false;          // turn off first time flag
        d->rsdelta += count - incount; // add the number of samples to d->rsdelta
    } else {
        out_ptr = &r->input[-d->output];
        ptr = _loadBuffer(r, out_ptr, buff, incount, ptr);
    }
    return ptr;
}

/**
 * This routine is for loading data from the delay line buff. If the
 * address of curr_ptr < r->base, it will force it to be within r->base
 * space, If the load goes past the end of r->base it will wrap around.
 * Cause count bytes of data at curr_ptr (within the delay line) to be
 * loaded into buff. (Buff is a dmem buffer)
 */
Acmd* _loadBuffer(ALFx* r, s16* curr_ptr, s32 buff, s32 count, Acmd* p) {
    Acmd* ptr = p;
    s32 after_end;
    s32 before_end;
    s16* updated_ptr;
    s16* delay_end;

    delay_end = &r->base[r->length];

    if (curr_ptr < r->base) {
        curr_ptr += r->length;
    }
    updated_ptr = curr_ptr + count;

    if (updated_ptr > delay_end) {
        after_end = updated_ptr - delay_end;
        before_end = delay_end - curr_ptr;

        aSetBuffer(ptr++, 0, buff, 0x000, before_end << 1);
        aLoadBuffer(ptr++, osVirtualToPhysical(curr_ptr));
        aSetBuffer(ptr++, 0, buff + (before_end << 1), 0, after_end << 1);
        aLoadBuffer(ptr++, osVirtualToPhysical(r->base));
    } else {
        aSetBuffer(ptr++, 0, buff, 0x000, count << 1);
        aLoadBuffer(ptr++, osVirtualToPhysical(curr_ptr));
    }
    aSetBuffer(ptr++, 0, 0x000, 0x000, count << 1);
    return ptr;
}

/**
 * This routine is for writing data to the delay line buff. If the
 * address of curr_ptr < r->base, it will force it to be within r->base
 * space. If the write goes past the end of r->base, it will wrap around
 * Cause count bytes of data at buff to be written to delay line, curr_ptr.
 */
Acmd* _saveBuffer(ALFx* r, s16* curr_ptr, s32 buff, s32 count, Acmd* p) {
    Acmd* ptr = p;
    s32 after_end;
    s32 before_end;
    s16* updated_ptr;
    s16* delay_end;

    delay_end = &r->base[r->length];
    if (curr_ptr < r->base) {
        // probably just security, shouldn't occur
        curr_ptr += r->length;
    }
    updated_ptr = curr_ptr + count;

    if (updated_ptr > delay_end) { // if the data wraps past end of r->base
        after_end = updated_ptr - delay_end;
        before_end = delay_end - curr_ptr;

        aSetBuffer(ptr++, 0, 0x000, buff, before_end << 1);
        aSaveBuffer(ptr++, osVirtualToPhysical(curr_ptr));
        aSetBuffer(ptr++, 0, 0x000, buff + (before_end << 1), after_end << 1);
        aSaveBuffer(ptr++, osVirtualToPhysical(r->base));
        aSetBuffer(ptr++, 0, 0x000, 0x000, count << 1);
    } else {
        aSetBuffer(ptr++, 0, 0x000, buff, count << 1);
        aSaveBuffer(ptr++, osVirtualToPhysical(curr_ptr));
    }
    return ptr;
}

Acmd* _filterBuffer(ALLowPass* lp, s32 buff, s32 count, Acmd* p) {
    Acmd* ptr = p;

    aSetBuffer(ptr++, 0, buff, buff, count << 1);
    aLoadADPCM(ptr++, 32, osVirtualToPhysical(lp->fcvec.fccoef));
    aPoleFilter(ptr++, lp->first, lp->fgain, osVirtualToPhysical(lp->fstate));
    lp->first = false;

    return ptr;
}

/**
 * Generate a triangle wave from -1 to 1, and find the current position
 * in the wave. (Rate of the wave is controlled by d->rsinc, which is chorus
 * rate) Multiply the current triangle wave value by d->rsgain, (chorus depth)
 * which is expressed in number of samples back from output pointer the chorus
 * should go at it's full chorus. In otherwords, this function returns a number
 * of samples the output pointer should modulate backwards.
 */
f32 _doModFunc(ALDelay* d, s32 count) {
    f32 val;

    // generate bipolar sawtooth from -RANGE to +RANGE
    d->rsval += d->rsinc * count;
    d->rsval = (d->rsval > RANGE) ? (d->rsval - RANGE * 2) : d->rsval;

    // convert to monopolar triangle from 0 to RANGE
    val = d->rsval;
    val = (val < 0) ? -val : val;

    // convert to bipolar triangle from -1 to 1
    val -= RANGE / 2;

    return d->rsgain * val;
}