static void ghosts(NodeGlare* ndg, CompBuf* dst, CompBuf* src)
{
// colormodulation and scale factors (cm & scalef) for 16 passes max: 64
int x, y, n, p, np;
fRGB c, tc, cm[64];
float sc, isc, u, v, sm, s, t, ofs, scalef[64];
CompBuf *tbuf1, *tbuf2, *gbuf;
const float cmo = 1.f - ndg->colmod;
const int qt = 1 << ndg->quality;
const float s1 = 4.f/(float)qt, s2 = 2.f*s1;
gbuf = BTP(src, ndg->threshold, qt);
tbuf1 = dupalloc_compbuf(gbuf);
IIR_gauss(tbuf1, s1, 0, 3);
IIR_gauss(tbuf1, s1, 1, 3);
IIR_gauss(tbuf1, s1, 2, 3);
tbuf2 = dupalloc_compbuf(tbuf1);
IIR_gauss(tbuf2, s2, 0, 3);
IIR_gauss(tbuf2, s2, 1, 3);
IIR_gauss(tbuf2, s2, 2, 3);
if (ndg->iter & 1) ofs = 0.5f; else ofs = 0.f;
for (x=0; x<(ndg->iter*4); x++) {
y = x & 3;
cm[x][0] = cm[x][1] = cm[x][2] = 1;
if (y==1) fRGB_rgbmult(cm[x], 1.f, cmo, cmo);
if (y==2) fRGB_rgbmult(cm[x], cmo, cmo, 1.f);
if (y==3) fRGB_rgbmult(cm[x], cmo, 1.f, cmo);
scalef[x] = 2.1f*(1.f-(x+ofs)/(float)(ndg->iter*4));
if (x & 1) scalef[x] = -0.99f/scalef[x];
}
sc = 2.13;
isc = -0.97;
for (y=0; y<gbuf->y; y++) {
v = (float)(y+0.5f) / (float)gbuf->y;
for (x=0; x<gbuf->x; x++) {
u = (float)(x+0.5f) / (float)gbuf->x;
s = (u-0.5f)*sc + 0.5f, t = (v-0.5f)*sc + 0.5f;
qd_getPixelLerp(tbuf1, s*gbuf->x, t*gbuf->y, c);
sm = smoothMask(s, t);
fRGB_mult(c, sm);
s = (u-0.5f)*isc + 0.5f, t = (v-0.5f)*isc + 0.5f;
qd_getPixelLerp(tbuf2, s*gbuf->x - 0.5f, t*gbuf->y - 0.5f, tc);
sm = smoothMask(s, t);
fRGB_madd(c, tc, sm);
qd_setPixel(gbuf, x, y, c);
}
}
memset(tbuf1->rect, 0, tbuf1->x*tbuf1->y*tbuf1->type*sizeof(float));
for (n=1; n<ndg->iter; n++) {
for (y=0; y<gbuf->y; y++) {
v = (float)(y+0.5f) / (float)gbuf->y;
for (x=0; x<gbuf->x; x++) {
u = (float)(x+0.5f) / (float)gbuf->x;
tc[0] = tc[1] = tc[2] = 0.f;
for (p=0;p<4;p++) {
np = (n<<2) + p;
s = (u-0.5f)*scalef[np] + 0.5f;
t = (v-0.5f)*scalef[np] + 0.5f;
qd_getPixelLerp(gbuf, s*gbuf->x - 0.5f, t*gbuf->y - 0.5f, c);
fRGB_colormult(c, cm[np]);
sm = smoothMask(s, t)*0.25f;
fRGB_madd(tc, c, sm);
}
p = (x + y*tbuf1->x)*tbuf1->type;
tbuf1->rect[p] += tc[0];
tbuf1->rect[p+1] += tc[1];
tbuf1->rect[p+2] += tc[2];
}
}
memcpy(gbuf->rect, tbuf1->rect, tbuf1->x*tbuf1->y*tbuf1->type*sizeof(float));
}
free_compbuf(tbuf1);
free_compbuf(tbuf2);
mixImages(dst, gbuf, 0.5f + 0.5f*ndg->mix);
free_compbuf(gbuf);
}
void GlareGhostOperation::generateGlare(float *data, MemoryBuffer *inputTile, NodeGlare *settings)
{
const int qt = 1 << settings->quality;
const float s1 = 4.0f / (float)qt, s2 = 2.0f * s1;
int x, y, n, p, np;
fRGB c, tc, cm[64];
float sc, isc, u, v, sm, s, t, ofs, scalef[64];
const float cmo = 1.0f - settings->colmod;
MemoryBuffer *gbuf = inputTile->duplicate();
MemoryBuffer *tbuf1 = inputTile->duplicate();
bool breaked = false;
FastGaussianBlurOperation::IIR_gauss(tbuf1, s1, 0, 3);
if (!breaked) FastGaussianBlurOperation::IIR_gauss(tbuf1, s1, 1, 3);
if (isBreaked()) breaked = true;
if (!breaked) FastGaussianBlurOperation::IIR_gauss(tbuf1, s1, 2, 3);
MemoryBuffer *tbuf2 = tbuf1->duplicate();
if (isBreaked()) breaked = true;
if (!breaked) FastGaussianBlurOperation::IIR_gauss(tbuf2, s2, 0, 3);
if (isBreaked()) breaked = true;
if (!breaked) FastGaussianBlurOperation::IIR_gauss(tbuf2, s2, 1, 3);
if (isBreaked()) breaked = true;
if (!breaked) FastGaussianBlurOperation::IIR_gauss(tbuf2, s2, 2, 3);
ofs = (settings->iter & 1) ? 0.5f : 0.0f;
for (x = 0; x < (settings->iter * 4); x++) {
y = x & 3;
cm[x][0] = cm[x][1] = cm[x][2] = 1;
if (y == 1) fRGB_rgbmult(cm[x], 1.0f, cmo, cmo);
if (y == 2) fRGB_rgbmult(cm[x], cmo, cmo, 1.0f);
if (y == 3) fRGB_rgbmult(cm[x], cmo, 1.0f, cmo);
scalef[x] = 2.1f * (1.0f - (x + ofs) / (float)(settings->iter * 4));
if (x & 1) scalef[x] = -0.99f / scalef[x];
}
sc = 2.13;
isc = -0.97;
for (y = 0; y < gbuf->getHeight() && (!breaked); y++) {
v = ((float)y + 0.5f) / (float)gbuf->getHeight();
for (x = 0; x < gbuf->getWidth(); x++) {
u = ((float)x + 0.5f) / (float)gbuf->getWidth();
s = (u - 0.5f) * sc + 0.5f, t = (v - 0.5f) * sc + 0.5f;
tbuf1->readBilinear(c, s * gbuf->getWidth(), t * gbuf->getHeight());
sm = smoothMask(s, t);
mul_v3_fl(c, sm);
s = (u - 0.5f) * isc + 0.5f, t = (v - 0.5f) * isc + 0.5f;
tbuf2->readBilinear(tc, s * gbuf->getWidth() - 0.5f, t * gbuf->getHeight() - 0.5f);
sm = smoothMask(s, t);
madd_v3_v3fl(c, tc, sm);
gbuf->writePixel(x, y, c);
}
if (isBreaked()) breaked = true;
}
memset(tbuf1->getBuffer(), 0, tbuf1->getWidth() * tbuf1->getHeight() * COM_NUM_CHANNELS_COLOR * sizeof(float));
for (n = 1; n < settings->iter && (!breaked); n++) {
for (y = 0; y < gbuf->getHeight() && (!breaked); y++) {
v = ((float)y + 0.5f) / (float)gbuf->getHeight();
for (x = 0; x < gbuf->getWidth(); x++) {
u = ((float)x + 0.5f) / (float)gbuf->getWidth();
tc[0] = tc[1] = tc[2] = 0.0f;
for (p = 0; p < 4; p++) {
np = (n << 2) + p;
s = (u - 0.5f) * scalef[np] + 0.5f;
t = (v - 0.5f) * scalef[np] + 0.5f;
gbuf->readBilinear(c, s * gbuf->getWidth() - 0.5f, t * gbuf->getHeight() - 0.5f);
mul_v3_v3(c, cm[np]);
sm = smoothMask(s, t) * 0.25f;
madd_v3_v3fl(tc, c, sm);
}
tbuf1->addPixel(x, y, tc);
}
if (isBreaked()) breaked = true;
}
memcpy(gbuf->getBuffer(), tbuf1->getBuffer(), tbuf1->getWidth() * tbuf1->getHeight() * COM_NUM_CHANNELS_COLOR * sizeof(float));
}
memcpy(data, gbuf->getBuffer(), gbuf->getWidth() * gbuf->getHeight() * COM_NUM_CHANNELS_COLOR * sizeof(float));
delete gbuf;
delete tbuf1;
delete tbuf2;
}
