Skip to content

O
opencv
Commit a7db950b authored by k-shinotsuka's avatar k-shinotsuka
Browse files
Options

add SSE code for Lab2RGB_f.

parent 7fe0fb6b
Hide whitespace changes
Inline Side-by-side
Showing with 166 additions and 5 deletions
modules/imgproc/src/color.cpp
View file @ a7db950b
......@@ -144,7 +144,7 @@ template<typename _Tp> static inline _Tp splineInterpolate(_Tp x, const _Tp* tab
#if CV_SSE2
template<typename _Tp> static inline void splineInterpolate(__m128& v_x, const _Tp* tab, int n)
{
__m128i v_ix = _mm_cvtps_epi32(_mm_min_ps(_mm_max_ps(v_x, _mm_setzero_ps()), _mm_set1_ps(float(n - 1))));
__m128i v_ix = _mm_cvttps_epi32(_mm_min_ps(_mm_max_ps(v_x, _mm_setzero_ps()), _mm_set1_ps(float(n - 1))));
v_x = _mm_sub_ps(v_x, _mm_cvtepi32_ps(v_ix));
v_ix = _mm_slli_epi32(v_ix, 2);
......@@ -5474,11 +5474,106 @@ struct Lab2RGB_f
coeffs[i+3] = _coeffs[i+3]*_whitept[i];
coeffs[i+blueIdx*3] = _coeffs[i+6]*_whitept[i];
}
lThresh = 0.008856f * 903.3f;
fThresh = 7.787f * 0.008856f + 16.0f / 116.0f;
#if CV_SSE2
haveSIMD = checkHardwareSupport(CV_CPU_SSE2);
#endif
}
#if CV_SSE2
void process(__m128& v_li0, __m128& v_li1, __m128& v_ai0,
__m128& v_ai1, __m128& v_bi0, __m128& v_bi1) const
{
__m128 v_y00 = _mm_mul_ps(v_li0, _mm_set1_ps(1.0f/903.3f));
__m128 v_y01 = _mm_mul_ps(v_li1, _mm_set1_ps(1.0f/903.3f));
__m128 v_fy00 = _mm_add_ps(_mm_mul_ps(_mm_set1_ps(7.787f), v_y00), _mm_set1_ps(16.0f/116.0f));
__m128 v_fy01 = _mm_add_ps(_mm_mul_ps(_mm_set1_ps(7.787f), v_y01), _mm_set1_ps(16.0f/116.0f));
__m128 v_fy10 = _mm_mul_ps(_mm_add_ps(v_li0, _mm_set1_ps(16.0f)), _mm_set1_ps(1.0f/116.0f));
__m128 v_fy11 = _mm_mul_ps(_mm_add_ps(v_li1, _mm_set1_ps(16.0f)), _mm_set1_ps(1.0f/116.0f));
__m128 v_y10 = _mm_mul_ps(_mm_mul_ps(v_fy10, v_fy10), v_fy10);
__m128 v_y11 = _mm_mul_ps(_mm_mul_ps(v_fy11, v_fy11), v_fy11);
__m128 v_cmpli0 = _mm_cmple_ps(v_li0, _mm_set1_ps(lThresh));
__m128 v_cmpli1 = _mm_cmple_ps(v_li1, _mm_set1_ps(lThresh));
v_y00 = _mm_and_ps(v_cmpli0, v_y00);
v_y01 = _mm_and_ps(v_cmpli1, v_y01);
v_fy00 = _mm_and_ps(v_cmpli0, v_fy00);
v_fy01 = _mm_and_ps(v_cmpli1, v_fy01);
v_y10 = _mm_andnot_ps(v_cmpli0, v_y10);
v_y11 = _mm_andnot_ps(v_cmpli1, v_y11);
v_fy10 = _mm_andnot_ps(v_cmpli0, v_fy10);
v_fy11 = _mm_andnot_ps(v_cmpli1, v_fy11);
__m128 v_y0 = _mm_or_ps(v_y00, v_y10);
__m128 v_y1 = _mm_or_ps(v_y01, v_y11);
__m128 v_fy0 = _mm_or_ps(v_fy00, v_fy10);
__m128 v_fy1 = _mm_or_ps(v_fy01, v_fy11);
__m128 v_fxz00 = _mm_add_ps(v_fy0, _mm_mul_ps(v_ai0, _mm_set1_ps(0.002f)));
__m128 v_fxz01 = _mm_add_ps(v_fy1, _mm_mul_ps(v_ai1, _mm_set1_ps(0.002f)));
__m128 v_fxz10 = _mm_sub_ps(v_fy0, _mm_mul_ps(v_bi0, _mm_set1_ps(0.005f)));
__m128 v_fxz11 = _mm_sub_ps(v_fy1, _mm_mul_ps(v_bi1, _mm_set1_ps(0.005f)));
__m128 v_fxz000 = _mm_mul_ps(_mm_sub_ps(v_fxz00, _mm_set1_ps(16.0f/116.0f)), _mm_set1_ps(1.0f/7.787f));
__m128 v_fxz001 = _mm_mul_ps(_mm_sub_ps(v_fxz01, _mm_set1_ps(16.0f/116.0f)), _mm_set1_ps(1.0f/7.787f));
__m128 v_fxz010 = _mm_mul_ps(_mm_sub_ps(v_fxz10, _mm_set1_ps(16.0f/116.0f)), _mm_set1_ps(1.0f/7.787f));
__m128 v_fxz011 = _mm_mul_ps(_mm_sub_ps(v_fxz11, _mm_set1_ps(16.0f/116.0f)), _mm_set1_ps(1.0f/7.787f));
__m128 v_fxz100 = _mm_mul_ps(_mm_mul_ps(v_fxz00, v_fxz00), v_fxz00);
__m128 v_fxz101 = _mm_mul_ps(_mm_mul_ps(v_fxz01, v_fxz01), v_fxz01);
__m128 v_fxz110 = _mm_mul_ps(_mm_mul_ps(v_fxz10, v_fxz10), v_fxz10);
__m128 v_fxz111 = _mm_mul_ps(_mm_mul_ps(v_fxz11, v_fxz11), v_fxz11);
__m128 v_cmpfxz00 = _mm_cmple_ps(v_fxz00, _mm_set1_ps(fThresh));
__m128 v_cmpfxz01 = _mm_cmple_ps(v_fxz01, _mm_set1_ps(fThresh));
__m128 v_cmpfxz10 = _mm_cmple_ps(v_fxz10, _mm_set1_ps(fThresh));
__m128 v_cmpfxz11 = _mm_cmple_ps(v_fxz11, _mm_set1_ps(fThresh));
v_fxz000 = _mm_and_ps(v_cmpfxz00, v_fxz000);
v_fxz001 = _mm_and_ps(v_cmpfxz01, v_fxz001);
v_fxz010 = _mm_and_ps(v_cmpfxz10, v_fxz010);
v_fxz011 = _mm_and_ps(v_cmpfxz11, v_fxz011);
v_fxz100 = _mm_andnot_ps(v_cmpfxz00, v_fxz100);
v_fxz101 = _mm_andnot_ps(v_cmpfxz01, v_fxz101);
v_fxz110 = _mm_andnot_ps(v_cmpfxz10, v_fxz110);
v_fxz111 = _mm_andnot_ps(v_cmpfxz11, v_fxz111);
__m128 v_x0 = _mm_or_ps(v_fxz000, v_fxz100);
__m128 v_x1 = _mm_or_ps(v_fxz001, v_fxz101);
__m128 v_z0 = _mm_or_ps(v_fxz010, v_fxz110);
__m128 v_z1 = _mm_or_ps(v_fxz011, v_fxz111);
__m128 v_ro0 = _mm_mul_ps(_mm_set1_ps(coeffs[0]), v_x0);
__m128 v_ro1 = _mm_mul_ps(_mm_set1_ps(coeffs[0]), v_x1);
__m128 v_go0 = _mm_mul_ps(_mm_set1_ps(coeffs[3]), v_x0);
__m128 v_go1 = _mm_mul_ps(_mm_set1_ps(coeffs[3]), v_x1);
__m128 v_bo0 = _mm_mul_ps(_mm_set1_ps(coeffs[6]), v_x0);
__m128 v_bo1 = _mm_mul_ps(_mm_set1_ps(coeffs[6]), v_x1);
v_ro0 = _mm_add_ps(v_ro0, _mm_mul_ps(_mm_set1_ps(coeffs[1]), v_y0));
v_ro1 = _mm_add_ps(v_ro1, _mm_mul_ps(_mm_set1_ps(coeffs[1]), v_y1));
v_go0 = _mm_add_ps(v_go0, _mm_mul_ps(_mm_set1_ps(coeffs[4]), v_y0));
v_go1 = _mm_add_ps(v_go1, _mm_mul_ps(_mm_set1_ps(coeffs[4]), v_y1));
v_bo0 = _mm_add_ps(v_bo0, _mm_mul_ps(_mm_set1_ps(coeffs[7]), v_y0));
v_bo1 = _mm_add_ps(v_bo1, _mm_mul_ps(_mm_set1_ps(coeffs[7]), v_y1));
v_ro0 = _mm_add_ps(v_ro0, _mm_mul_ps(_mm_set1_ps(coeffs[2]), v_z0));
v_ro1 = _mm_add_ps(v_ro1, _mm_mul_ps(_mm_set1_ps(coeffs[2]), v_z1));
v_go0 = _mm_add_ps(v_go0, _mm_mul_ps(_mm_set1_ps(coeffs[5]), v_z0));
v_go1 = _mm_add_ps(v_go1, _mm_mul_ps(_mm_set1_ps(coeffs[5]), v_z1));
v_bo0 = _mm_add_ps(v_bo0, _mm_mul_ps(_mm_set1_ps(coeffs[8]), v_z0));
v_bo1 = _mm_add_ps(v_bo1, _mm_mul_ps(_mm_set1_ps(coeffs[8]), v_z1));
v_li0 = _mm_min_ps(_mm_max_ps(v_ro0, _mm_setzero_ps()), _mm_set1_ps(1.0f));
v_li1 = _mm_min_ps(_mm_max_ps(v_ro1, _mm_setzero_ps()), _mm_set1_ps(1.0f));
v_ai0 = _mm_min_ps(_mm_max_ps(v_go0, _mm_setzero_ps()), _mm_set1_ps(1.0f));
v_ai1 = _mm_min_ps(_mm_max_ps(v_go1, _mm_setzero_ps()), _mm_set1_ps(1.0f));
v_bi0 = _mm_min_ps(_mm_max_ps(v_bo0, _mm_setzero_ps()), _mm_set1_ps(1.0f));
v_bi1 = _mm_min_ps(_mm_max_ps(v_bo1, _mm_setzero_ps()), _mm_set1_ps(1.0f));
}
#endif
void operator()(const float* src, float* dst, int n) const
{
int i, dcn = dstcn;
int i = 0, dcn = dstcn;
const float* gammaTab = srgb ? sRGBInvGammaTab : 0;
float gscale = GammaTabScale;
float C0 = coeffs[0], C1 = coeffs[1], C2 = coeffs[2],
......@@ -5487,9 +5582,70 @@ struct Lab2RGB_f
float alpha = ColorChannel<float>::max();
n *= 3;
static const float lThresh = 0.008856f * 903.3f;
static const float fThresh = 7.787f * 0.008856f + 16.0f / 116.0f;
for (i = 0; i < n; i += 3, dst += dcn)
#if CV_SSE2
if (haveSIMD)
{
for (; i <= n - 24; i += 24, dst += dcn * 8)
{
__m128 v_li0 = _mm_loadu_ps(src + i + 0);
__m128 v_li1 = _mm_loadu_ps(src + i + 4);
__m128 v_ai0 = _mm_loadu_ps(src + i + 8);
__m128 v_ai1 = _mm_loadu_ps(src + i + 12);
__m128 v_bi0 = _mm_loadu_ps(src + i + 16);
__m128 v_bi1 = _mm_loadu_ps(src + i + 20);
_mm_deinterleave_ps(v_li0, v_li1, v_ai0, v_ai1, v_bi0, v_bi1);
process(v_li0, v_li1, v_ai0, v_ai1, v_bi0, v_bi1);
if (gammaTab)
{
__m128 v_gscale = _mm_set1_ps(gscale);
v_li0 = _mm_mul_ps(v_li0, v_gscale);
v_li1 = _mm_mul_ps(v_li1, v_gscale);
v_ai0 = _mm_mul_ps(v_ai0, v_gscale);
v_ai1 = _mm_mul_ps(v_ai1, v_gscale);
v_bi0 = _mm_mul_ps(v_bi0, v_gscale);
v_bi1 = _mm_mul_ps(v_bi1, v_gscale);
splineInterpolate(v_li0, gammaTab, GAMMA_TAB_SIZE);
splineInterpolate(v_li1, gammaTab, GAMMA_TAB_SIZE);
splineInterpolate(v_ai0, gammaTab, GAMMA_TAB_SIZE);
splineInterpolate(v_ai1, gammaTab, GAMMA_TAB_SIZE);
splineInterpolate(v_bi0, gammaTab, GAMMA_TAB_SIZE);
splineInterpolate(v_bi1, gammaTab, GAMMA_TAB_SIZE);
}
if( dcn == 4 )
{
__m128 v_a0 = _mm_set1_ps(alpha);
__m128 v_a1 = _mm_set1_ps(alpha);
_mm_interleave_ps(v_li0, v_li1, v_ai0, v_ai1, v_bi0, v_bi1, v_a0, v_a1);
_mm_storeu_ps(dst + 0, v_li0);
_mm_storeu_ps(dst + 4, v_li1);
_mm_storeu_ps(dst + 8, v_ai0);
_mm_storeu_ps(dst + 12, v_ai1);
_mm_storeu_ps(dst + 16, v_bi0);
_mm_storeu_ps(dst + 20, v_bi1);
_mm_storeu_ps(dst + 24, v_a0);
_mm_storeu_ps(dst + 28, v_a1);
}
else
{
_mm_interleave_ps(v_li0, v_li1, v_ai0, v_ai1, v_bi0, v_bi1);
_mm_storeu_ps(dst + 0, v_li0);
_mm_storeu_ps(dst + 4, v_li1);
_mm_storeu_ps(dst + 8, v_ai0);
_mm_storeu_ps(dst + 12, v_ai1);
_mm_storeu_ps(dst + 16, v_bi0);
_mm_storeu_ps(dst + 20, v_bi1);
}
}
}
#endif
for (; i < n; i += 3, dst += dcn)
{
float li = src[i];
float ai = src[i + 1];
......@@ -5540,6 +5696,11 @@ struct Lab2RGB_f
int dstcn;
float coeffs[9];
bool srgb;
float lThresh;
float fThresh;
#if CV_SSE2
bool haveSIMD;
#endif
};
#undef clip
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment