liz/rawTherapee
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Apply HaldCLUT::getRGB() per tile line

Browse Source 
`getRGB()` now takes a whole tile line instead of a single pixel.
This commit is contained in:
Flössie
2016-04-26 21:57:58 +02:00
parent 78c08e9e5c
commit bf499055e1
3 changed files with 75 additions and 61 deletions
Download Patch File Download Diff File Expand all files Collapse all files

106
rtengine/clutstore.cc
View File

@@ -164,92 +164,94 @@ Glib::ustring rtengine::HaldCLUT::getProfile() const
return clut_profile;
}
void rtengine::HaldCLUT::getRGB(float r, float g, float b, float out_rgbx[4]) const
void rtengine::HaldCLUT::getRGB(std::size_t line_size, const float* r, const float* g, const float* b, float* out_rgbx) const
{
const unsigned int level = clut_level; // This is important
const unsigned int red = std::min(flevel_minus_two, r * flevel_minus_one);
const unsigned int green = std::min(flevel_minus_two, g * flevel_minus_one);
const unsigned int blue = std::min(flevel_minus_two, b * flevel_minus_one);
const unsigned int level_square = level * level;
const unsigned int color = red + green * level + blue * level_square;
for (std::size_t column = 0; column < line_size; ++column, ++r, ++g, ++b, out_rgbx += 4) {
const unsigned int red = std::min(flevel_minus_two, *r * flevel_minus_one);
const unsigned int green = std::min(flevel_minus_two, *g * flevel_minus_one);
const unsigned int blue = std::min(flevel_minus_two, *b * flevel_minus_one);
const unsigned int color = red + green * level + blue * level_square;
#ifndef __SSE2__
r = r * flevel_minus_one - red;
g = g * flevel_minus_one - green;
b = b * flevel_minus_one - blue;
const float re = *r * flevel_minus_one - red;
const float gr = *g * flevel_minus_one - green;
const float bl = *b * flevel_minus_one - blue;
size_t index = color * 4;
size_t index = color * 4;
float tmp1[4] ALIGNED16;
tmp1[0] = intp<float>(r, clut_image.data[index + 4], clut_image.data[index]);
tmp1[1] = intp<float>(r, clut_image.data[index + 5], clut_image.data[index + 1]);
tmp1[2] = intp<float>(r, clut_image.data[index + 6], clut_image.data[index + 2]);
float tmp1[4] ALIGNED16;
tmp1[0] = intp<float>(re, clut_image.data[index + 4], clut_image.data[index]);
tmp1[1] = intp<float>(re, clut_image.data[index + 5], clut_image.data[index + 1]);
tmp1[2] = intp<float>(re, clut_image.data[index + 6], clut_image.data[index + 2]);
index = (color + level) * 4;
index = (color + level) * 4;
float tmp2[4] ALIGNED16;
tmp2[0] = intp<float>(r, clut_image.data[index + 4], clut_image.data[index]);
tmp2[1] = intp<float>(r, clut_image.data[index + 5], clut_image.data[index + 1]);
tmp2[2] = intp<float>(r, clut_image.data[index + 6], clut_image.data[index + 2]);
float tmp2[4] ALIGNED16;
tmp2[0] = intp<float>(re, clut_image.data[index + 4], clut_image.data[index]);
tmp2[1] = intp<float>(re, clut_image.data[index + 5], clut_image.data[index + 1]);
tmp2[2] = intp<float>(re, clut_image.data[index + 6], clut_image.data[index + 2]);
out_rgbx[0] = intp<float>(g, tmp2[0], tmp1[0]);
out_rgbx[1] = intp<float>(g, tmp2[1], tmp1[1]);
out_rgbx[2] = intp<float>(g, tmp2[2], tmp1[2]);
out_rgbx[0] = intp<float>(gr, tmp2[0], tmp1[0]);
out_rgbx[1] = intp<float>(gr, tmp2[1], tmp1[1]);
out_rgbx[2] = intp<float>(gr, tmp2[2], tmp1[2]);
index = (color + level_square) * 4;
index = (color + level_square) * 4;
tmp1[0] = intp<float>(r, clut_image.data[index + 4], clut_image.data[index]);
tmp1[1] = intp<float>(r, clut_image.data[index + 5], clut_image.data[index + 1]);
tmp1[2] = intp<float>(r, clut_image.data[index + 6], clut_image.data[index + 2]);
tmp1[0] = intp<float>(re, clut_image.data[index + 4], clut_image.data[index]);
tmp1[1] = intp<float>(re, clut_image.data[index + 5], clut_image.data[index + 1]);
tmp1[2] = intp<float>(re, clut_image.data[index + 6], clut_image.data[index + 2]);
index = (color + level + level_square) * 4;
index = (color + level + level_square) * 4;
tmp2[0] = intp<float>(r, clut_image.data[index + 4], clut_image.data[index]);
tmp2[1] = intp<float>(r, clut_image.data[index + 5], clut_image.data[index + 1]);
tmp2[2] = intp<float>(r, clut_image.data[index + 6], clut_image.data[index + 2]);
tmp2[0] = intp<float>(re, clut_image.data[index + 4], clut_image.data[index]);
tmp2[1] = intp<float>(re, clut_image.data[index + 5], clut_image.data[index + 1]);
tmp2[2] = intp<float>(re, clut_image.data[index + 6], clut_image.data[index + 2]);
tmp1[0] = intp<float>(g, tmp2[0], tmp1[0]);
tmp1[1] = intp<float>(g, tmp2[1], tmp1[1]);
tmp1[2] = intp<float>(g, tmp2[2], tmp1[2]);
tmp1[0] = intp<float>(gr, tmp2[0], tmp1[0]);
tmp1[1] = intp<float>(gr, tmp2[1], tmp1[1]);
tmp1[2] = intp<float>(gr, tmp2[2], tmp1[2]);
out_rgbx[0] = intp<float>(b, tmp1[0], out_rgbx[0]);
out_rgbx[1] = intp<float>(b, tmp1[1], out_rgbx[1]);
out_rgbx[2] = intp<float>(b, tmp1[2], out_rgbx[2]);
out_rgbx[0] = intp<float>(bl, tmp1[0], out_rgbx[0]);
out_rgbx[1] = intp<float>(bl, tmp1[1], out_rgbx[1]);
out_rgbx[2] = intp<float>(bl, tmp1[2], out_rgbx[2]);
#else
const vfloat v_tmp = _mm_set_ps(0.0f, b, g, r) * _mm_load_ps1(&flevel_minus_one);
const vfloat v_rgb = v_tmp - _mm_cvtepi32_ps(_mm_cvttps_epi32(_mm_min_ps(_mm_load_ps1(&flevel_minus_two), v_tmp)));
const vfloat v_tmp = _mm_set_ps(0.0f, *b, *g, *r) * _mm_load_ps1(&flevel_minus_one);
const vfloat v_rgb = v_tmp - _mm_cvtepi32_ps(_mm_cvttps_epi32(_mm_min_ps(_mm_load_ps1(&flevel_minus_two), v_tmp)));
size_t index = color * 4;
size_t index = color * 4;
const vfloat v_r = PERMUTEPS(v_rgb, 0x00);
const vfloat v_r = PERMUTEPS(v_rgb, 0x00);
vfloat v_tmp1 = vintpf(v_r, getClutValue(clut_image, index + 4), getClutValue(clut_image, index));
vfloat v_tmp1 = vintpf(v_r, getClutValue(clut_image, index + 4), getClutValue(clut_image, index));
index = (color + level) * 4;
index = (color + level) * 4;
vfloat v_tmp2 = vintpf(v_r, getClutValue(clut_image, index + 4), getClutValue(clut_image, index));
vfloat v_tmp2 = vintpf(v_r, getClutValue(clut_image, index + 4), getClutValue(clut_image, index));
const vfloat v_g = PERMUTEPS(v_rgb, 0x55);
const vfloat v_g = PERMUTEPS(v_rgb, 0x55);
vfloat v_out = vintpf(v_g, v_tmp2, v_tmp1);
vfloat v_out = vintpf(v_g, v_tmp2, v_tmp1);
index = (color + level_square) * 4;
index = (color + level_square) * 4;
v_tmp1 = vintpf(v_r, getClutValue(clut_image, index + 4), getClutValue(clut_image, index));
v_tmp1 = vintpf(v_r, getClutValue(clut_image, index + 4), getClutValue(clut_image, index));
index = (color + level + level_square) * 4;
index = (color + level + level_square) * 4;
v_tmp2 = vintpf(v_r, getClutValue(clut_image, index + 4), getClutValue(clut_image, index));
v_tmp2 = vintpf(v_r, getClutValue(clut_image, index + 4), getClutValue(clut_image, index));
v_tmp1 = vintpf(v_g, v_tmp2, v_tmp1);
v_tmp1 = vintpf(v_g, v_tmp2, v_tmp1);
const vfloat v_b = PERMUTEPS(v_rgb, 0xAA);
const vfloat v_b = PERMUTEPS(v_rgb, 0xAA);
_mm_store_ps(out_rgbx, vintpf(v_b, v_tmp1, v_out));
_mm_store_ps(out_rgbx, vintpf(v_b, v_tmp1, v_out));
#endif
}
}
rtengine::CLUTStore& rtengine::CLUTStore::getInstance()

4
rtengine/clutstore.h
View File

@@ -24,7 +24,7 @@ public:
virtual Glib::ustring getFilename() const = 0;
virtual Glib::ustring getProfile() const = 0;
virtual void getRGB(float r, float g, float b, float out_rgbx[4]) const = 0;
virtual void getRGB(std::size_t line_size, const float* r, const float* g, const float* b, float* out_rgbx) const = 0;
static void splitClutFilename(
const Glib::ustring& filename,
@@ -48,7 +48,7 @@ public:
Glib::ustring getFilename() const override;
Glib::ustring getProfile() const override;
void getRGB(float r, float g, float b, float out_rgbx[4]) const override;
void getRGB(std::size_t line_size, const float* r, const float* g, const float* b, float* out_rgbx) const override;
private:
AlignedBuffer<std::uint16_t> clut_image;

26
rtengine/improcfun.cc
View File

@@ -16,6 +16,7 @@
* You should have received a copy of the GNU General Public License
* along with RawTherapee. If not, see <http://www.gnu.org/licenses/>.
*/
#include <memory>
#include <cmath>
#include <glib.h>
#include <glibmm.h>
@@ -3224,8 +3225,8 @@ void ImProcFunctions::rgbProc (Imagefloat* working, LabImage* lab, PipetteBuffer
}
}
double filmSimCorrectedStrength = double(params->filmSimulation.strength) / 100.;
double filmSimSourceStrength = double(100 - params->filmSimulation.strength) / 100.;
float filmSimCorrectedStrength = static_cast<float>(params->filmSimulation.strength) / 100.0f;
float filmSimSourceStrength = 1.0f - filmSimCorrectedStrength;
const float exp_scale = pow (2.0, expcomp);
const float comp = (max(0.0, expcomp) + 1.0) * hlcompr / 100.0;
@@ -4354,13 +4355,24 @@ void ImProcFunctions::rgbProc (Imagefloat* working, LabImage* lab, PipetteBuffer
sourceR = CLIP<float>( Color::gamma_srgb( sourceR ) );
sourceG = CLIP<float>( Color::gamma_srgb( sourceG ) );
sourceB = CLIP<float>( Color::gamma_srgb( sourceB ) );
}
const std::size_t line_size = std::min(TS, tW - jstart);
std::size_t out_rgbx_size = 4 * (line_size + 16);
std::unique_ptr<float> out_rgbx_buf(new float[out_rgbx_size]);
void* out_rgbx_ptr = out_rgbx_buf.get();
float* const out_rgbx = reinterpret_cast<float*>(std::align(16, 4 * line_size, out_rgbx_ptr, out_rgbx_size));
colorLUT->getRGB(line_size, rtemp + ti * TS, gtemp + ti * TS, btemp + ti * TS, out_rgbx);
for (int j = jstart, tj = 0; j < tW; j++, tj++) {
float &sourceR = rtemp[ti * TS + tj];
float &sourceG = gtemp[ti * TS + tj];
float &sourceB = btemp[ti * TS + tj];
float out_rgbx[4] ALIGNED16;
colorLUT->getRGB( sourceR, sourceG, sourceB, out_rgbx );
// apply strength
sourceR = out_rgbx[0] * filmSimCorrectedStrength + sourceR * filmSimSourceStrength;
sourceG = out_rgbx[1] * filmSimCorrectedStrength + sourceG * filmSimSourceStrength;
sourceB = out_rgbx[2] * filmSimCorrectedStrength + sourceB * filmSimSourceStrength;
sourceR = out_rgbx[tj * 4 + 0] * filmSimCorrectedStrength + sourceR * filmSimSourceStrength;
sourceG = out_rgbx[tj * 4 + 1] * filmSimCorrectedStrength + sourceG * filmSimSourceStrength;
sourceB = out_rgbx[tj * 4 + 2] * filmSimCorrectedStrength + sourceB * filmSimSourceStrength;
// apply inverse gamma sRGB
sourceR = Color::igamma_srgb( sourceR );
sourceG = Color::igamma_srgb( sourceG );