lab2rgb performance optimizations

This commit is contained in:
Oliver Duis
2011-04-09 13:11:31 +02:00
parent 60a528d845
commit e972e992ba

View File

@@ -21,7 +21,7 @@
#include <glibmm.h>
#include <iccstore.h>
#include <iccmatrices.h>
#include <mytime.h>
//#include <sRGBgamutbdy.h>
#ifdef _OPENMP
@@ -56,60 +56,68 @@ const char* wprofnames[] = {"sRGB", "Adobe RGB", "ProPhoto", "WideGamut", "Bruce
const int numprof = 7;
void ImProcFunctions::lab2rgb (LabImage* lab, Image8* image) {
//MyTime tBeg,tEnd;
// tBeg.set();
//gamutmap(lab);
if (monitorTransform) {
int ix = 0;
float g;
float* buffer = new float [3*lab->W];
// cmsDoTransform is relatively expensive
// Causes problems on some machines
//#pragma omp parallel for if (multiThread)
#pragma omp parallel for if (multiThread)
for (int i=0; i<lab->H; i++) {
float buffer[3*lab->W];
const int ix = i * 3 * lab->W;
int iy = 0;
float* rL = lab->L[i];
float* ra = lab->a[i];
float* rb = lab->b[i];
int iy = 0;
float R,G,B;
float fy,fx,fz,x_,y_,z_;
for (int j=0; j<lab->W; j++) {
float fy = (0.00862069 * rL[j])/327.68 + 0.137932; // (L+16)/116
float fx = (0.002 * ra[j])/327.68 + fy;
float fz = fy - (0.005 * rb[j])/327.68;
fy = (0.00862069 * rL[j]) / 327.68 + 0.137932; // (L+16)/116
fx = (0.002 * ra[j]) / 327.68 + fy;
fz = fy - (0.005 * rb[j]) / 327.68;
float x_ = Lab2xyz(fx)*D50x;//should this be 32767??? buffer is short int !!!
float y_ = Lab2xyz(fy);
float z_ = Lab2xyz(fz)*D50z;
x_ = Lab2xyz(fx)*D50x;//should this be 32767??? buffer is short int !!!
y_ = Lab2xyz(fy);
z_ = Lab2xyz(fz)*D50z;
buffer[iy++] = CLIP01(x_);
buffer[iy++] = CLIP01(y_);
buffer[iy++] = CLIP01(z_);
}
cmsDoTransform (monitorTransform, buffer, image->data + ix, lab->W);
ix += 3*lab->W;
}
delete [] buffer;
} else {
#pragma omp parallel for if (multiThread)
for (int i=0; i<lab->H; i++) {
float* rL = lab->L[i];
float* ra = lab->a[i];
float* rb = lab->b[i];
int ix = 3*i*lab->W;
for (int j=0; j<lab->W; j++) {
float g;
int ix = i * 3 * lab->W;
float R,G,B;
float fy,fx,fz,x_,y_,z_;
for (int j=0; j<lab->W; j++) {
//float L1=rL[j],a1=ra[j],b1=rb[j];//for testing
float fy = (0.00862069 * rL[j])/327.68 + 0.137932; // (L+16)/116
float fx = (0.002 * ra[j])/327.68 + fy;
float fz = fy - (0.005 * rb[j])/327.68;
fy = (0.00862069 * rL[j]) / 327.68 + 0.137932; // (L+16)/116
fx = (0.002 * ra[j]) / 327.68 + fy;
fz = fy - (0.005 * rb[j]) / 327.68;
float x_ = 65535*Lab2xyz(fx)*D50x;
float y_ = 65535*Lab2xyz(fy);
float z_ = 65535*Lab2xyz(fz)*D50z;
x_ = 65535*Lab2xyz(fx)*D50x;
y_ = 65535*Lab2xyz(fy);
z_ = 65535*Lab2xyz(fz)*D50z;
xyz2srgb(x_,y_,z_,R,G,B);
@@ -120,6 +128,9 @@ void ImProcFunctions::lab2rgb (LabImage* lab, Image8* image) {
}
}
}
//tEnd.set();
//printf("lab2rgb %i %d\n", lab->W, tEnd.etime(tBeg));
}
Image8* ImProcFunctions::lab2rgb (LabImage* lab, int cx, int cy, int cw, int ch, Glib::ustring profile) {
@@ -141,18 +152,19 @@ Image8* ImProcFunctions::lab2rgb (LabImage* lab, int cx, int cy, int cw, int ch,
cmsHTRANSFORM hTransform = cmsCreateTransform (iprof, TYPE_RGB_16, oprof, TYPE_RGB_8, settings->colorimetricIntent,
cmsFLAGS_NOOPTIMIZE | cmsFLAGS_NOCACHE ); // NOCACHE is important for thread safety
lcmsMutex->unlock ();
int ix = 0;
float g;
short* buffer = new short [3*cw];
// cmsDoTransform is relatively expensive
// Causes problems on some machines
//#pragma omp parallel for if (multiThread)
#pragma omp parallel for if (multiThread)
for (int i=cy; i<cy+ch; i++) {
short buffer [3*cw];
const int ix = i * 3 * cw;
int iy = 0;
float* rL = lab->L[i];
float* ra = lab->a[i];
float* rb = lab->b[i];
int iy = 0;
for (int j=cx; j<cx+cw; j++) {
float fy = (0.00862069 * rL[j])/327.68 + 0.137932; // (L+16)/116
@@ -167,10 +179,10 @@ Image8* ImProcFunctions::lab2rgb (LabImage* lab, int cx, int cy, int cw, int ch,
buffer[iy++] = CLIP((int)y_);
buffer[iy++] = CLIP((int)z_);
}
cmsDoTransform (hTransform, buffer, image->data + ix, cw);
ix += 3*cw;
}
delete [] buffer;
cmsDeleteTransform(hTransform);
} else {