void GaussianXBlurOperation::executePixel(float output[4], int x, int y, void *data)
{
float color_accum[4] = {0.0f, 0.0f, 0.0f, 0.0f};
float multiplier_accum = 0.0f;
MemoryBuffer *inputBuffer = (MemoryBuffer *)data;
float *buffer = inputBuffer->getBuffer();
int bufferwidth = inputBuffer->getWidth();
int bufferstartx = inputBuffer->getRect()->xmin;
int bufferstarty = inputBuffer->getRect()->ymin;
int miny = y;
int minx = x - this->m_rad;
int maxx = x + this->m_rad;
miny = max(miny, inputBuffer->getRect()->ymin);
minx = max(minx, inputBuffer->getRect()->xmin);
maxx = min(maxx, inputBuffer->getRect()->xmax - 1);
int step = getStep();
int offsetadd = getOffsetAdd();
int bufferindex = ((minx - bufferstartx) * 4) + ((miny - bufferstarty) * 4 * bufferwidth);
for (int nx = minx, index = (minx - x) + this->m_rad; nx <= maxx; nx += step, index += step) {
const float multiplier = this->m_gausstab[index];
madd_v4_v4fl(color_accum, &buffer[bufferindex], multiplier);
multiplier_accum += multiplier;
bufferindex += offsetadd;
}
mul_v4_v4fl(output, color_accum, 1.0f / multiplier_accum);
}
void GaussianAlphaXBlurOperation::executePixel(float output[4], int x, int y, void *data)
{
const bool do_invert = this->m_do_subtract;
MemoryBuffer *inputBuffer = (MemoryBuffer *)data;
float *buffer = inputBuffer->getBuffer();
int bufferwidth = inputBuffer->getWidth();
int bufferstartx = inputBuffer->getRect()->xmin;
int bufferstarty = inputBuffer->getRect()->ymin;
int miny = y;
// int maxy = y; // UNUSED
int minx = x - this->m_rad;
int maxx = x + this->m_rad; // UNUSED
miny = max(miny, inputBuffer->getRect()->ymin);
minx = max(minx, inputBuffer->getRect()->xmin);
// maxy = min(maxy, inputBuffer->getRect()->ymax);
maxx = min(maxx, inputBuffer->getRect()->xmax);
/* *** this is the main part which is different to 'GaussianXBlurOperation' *** */
int step = getStep();
int offsetadd = getOffsetAdd();
int bufferindex = ((minx - bufferstartx) * 4) + ((miny - bufferstarty) * 4 * bufferwidth);
/* gauss */
float alpha_accum = 0.0f;
float multiplier_accum = 0.0f;
/* dilate */
float value_max = finv_test(buffer[(x * 4) + (y * 4 * bufferwidth)], do_invert); /* init with the current color to avoid unneeded lookups */
float distfacinv_max = 1.0f; /* 0 to 1 */
for (int nx = minx; nx <= maxx; nx += step) {
const int index = (nx - x) + this->m_rad;
float value = finv_test(buffer[bufferindex], do_invert);
float multiplier;
/* gauss */
{
multiplier = this->m_gausstab[index];
alpha_accum += value * multiplier;
multiplier_accum += multiplier;
}
/* dilate - find most extreme color */
if (value > value_max) {
multiplier = this->m_distbuf_inv[index];
value *= multiplier;
if (value > value_max) {
value_max = value;
distfacinv_max = multiplier;
}
}
bufferindex += offsetadd;
}
/* blend between the max value and gauss blue - gives nice feather */
const float value_blur = alpha_accum / multiplier_accum;
const float value_final = (value_max * distfacinv_max) + (value_blur * (1.0f - distfacinv_max));
output[0] = finv_test(value_final, do_invert);
}
void BokehBlurOperation::executePixel(float output[4], int x, int y, void *data)
{
float color_accum[4];
float tempBoundingBox[4];
float bokeh[4];
this->m_inputBoundingBoxReader->readSampled(tempBoundingBox, x, y, COM_PS_NEAREST);
if (tempBoundingBox[0] > 0.0f) {
float multiplier_accum[4] = {0.0f, 0.0f, 0.0f, 0.0f};
MemoryBuffer *inputBuffer = (MemoryBuffer *)data;
float *buffer = inputBuffer->getBuffer();
int bufferwidth = inputBuffer->getWidth();
int bufferstartx = inputBuffer->getRect()->xmin;
int bufferstarty = inputBuffer->getRect()->ymin;
const float max_dim = max(this->getWidth(), this->getHeight());
int pixelSize = this->m_size * max_dim / 100.0f;
zero_v4(color_accum);
if (pixelSize < 2) {
this->m_inputProgram->readSampled(color_accum, x, y, COM_PS_NEAREST);
multiplier_accum[0] = 1.0f;
multiplier_accum[1] = 1.0f;
multiplier_accum[2] = 1.0f;
multiplier_accum[3] = 1.0f;
}
int miny = y - pixelSize;
int maxy = y + pixelSize;
int minx = x - pixelSize;
int maxx = x + pixelSize;
miny = max(miny, inputBuffer->getRect()->ymin);
minx = max(minx, inputBuffer->getRect()->xmin);
maxy = min(maxy, inputBuffer->getRect()->ymax);
maxx = min(maxx, inputBuffer->getRect()->xmax);
int step = getStep();
int offsetadd = getOffsetAdd() * COM_NUM_CHANNELS_COLOR;
float m = this->m_bokehDimension / pixelSize;
for (int ny = miny; ny < maxy; ny += step) {
int bufferindex = ((minx - bufferstartx) * COM_NUM_CHANNELS_COLOR) +
((ny - bufferstarty) * COM_NUM_CHANNELS_COLOR * bufferwidth);
for (int nx = minx; nx < maxx; nx += step) {
float u = this->m_bokehMidX - (nx - x) * m;
float v = this->m_bokehMidY - (ny - y) * m;
this->m_inputBokehProgram->readSampled(bokeh, u, v, COM_PS_NEAREST);
madd_v4_v4v4(color_accum, bokeh, &buffer[bufferindex]);
add_v4_v4(multiplier_accum, bokeh);
bufferindex += offsetadd;
}
}
output[0] = color_accum[0] * (1.0f / multiplier_accum[0]);
output[1] = color_accum[1] * (1.0f / multiplier_accum[1]);
output[2] = color_accum[2] * (1.0f / multiplier_accum[2]);
output[3] = color_accum[3] * (1.0f / multiplier_accum[3]);
}
else {
this->m_inputProgram->readSampled(output, x, y, COM_PS_NEAREST);
}
}
void GaussianXBlurOperation::executePixel(float output[4], int x, int y, void *data)
{
float color_accum[4] = {0.0f, 0.0f, 0.0f, 0.0f};
float multiplier_accum = 0.0f;
MemoryBuffer *inputBuffer = (MemoryBuffer *)data;
float *buffer = inputBuffer->getBuffer();
int bufferwidth = inputBuffer->getWidth();
int bufferstartx = inputBuffer->getRect()->xmin;
int bufferstarty = inputBuffer->getRect()->ymin;
rcti &rect = *inputBuffer->getRect();
int xmin = max_ii(x - m_filtersize, rect.xmin);
int xmax = min_ii(x + m_filtersize + 1, rect.xmax);
int ymin = max_ii(y, rect.ymin);
int step = getStep();
int offsetadd = getOffsetAdd();
int bufferindex = ((xmin - bufferstartx) * 4) + ((ymin - bufferstarty) * 4 * bufferwidth);
#ifdef __SSE2__
__m128 accum_r = _mm_load_ps(color_accum);
for (int nx = xmin, index = (xmin - x) + this->m_filtersize; nx < xmax; nx += step, index += step) {
__m128 reg_a = _mm_load_ps(&buffer[bufferindex]);
reg_a = _mm_mul_ps(reg_a, this->m_gausstab_sse[index]);
accum_r = _mm_add_ps(accum_r, reg_a);
multiplier_accum += this->m_gausstab[index];
bufferindex += offsetadd;
}
_mm_store_ps(color_accum, accum_r);
#else
for (int nx = xmin, index = (xmin - x) + this->m_filtersize; nx < xmax; nx += step, index += step) {
const float multiplier = this->m_gausstab[index];
madd_v4_v4fl(color_accum, &buffer[bufferindex], multiplier);
multiplier_accum += multiplier;
bufferindex += offsetadd;
}
#endif
mul_v4_v4fl(output, color_accum, 1.0f / multiplier_accum);
}
