From c1c7c7f83e7a07151e838b9aab20896355eba129 Mon Sep 17 00:00:00 2001 From: k-shinotsuka Date: Sun, 30 Oct 2016 12:02:39 +0900 Subject: [PATCH] add SSE code for Luv2RGB_f. --- modules/imgproc/src/color.cpp | 133 +++++++++++++++++++++++++++++++++- 1 file changed, 131 insertions(+), 2 deletions(-) diff --git a/modules/imgproc/src/color.cpp b/modules/imgproc/src/color.cpp index 77af633af3..7207cf9de3 100644 --- a/modules/imgproc/src/color.cpp +++ b/modules/imgproc/src/color.cpp @@ -5995,13 +5995,77 @@ struct Luv2RGB_f float d = 1.f/(whitept[0] + whitept[1]*15 + whitept[2]*3); un = 4*whitept[0]*d; vn = 9*whitept[1]*d; + #if CV_SSE2 + haveSIMD = checkHardwareSupport(CV_CPU_SSE2); + #endif CV_Assert(whitept[1] == 1.f); } + #if CV_SSE2 + void process(__m128& v_l0, __m128& v_l1, __m128& v_u0, + __m128& v_u1, __m128& v_v0, __m128& v_v1) const + { + __m128 v_y0 = _mm_mul_ps(_mm_add_ps(v_l0, _mm_set1_ps(16.0f)), _mm_set1_ps(1.f/116.f)); + __m128 v_y1 = _mm_mul_ps(_mm_add_ps(v_l1, _mm_set1_ps(16.0f)), _mm_set1_ps(1.f/116.f)); + v_y0 = _mm_mul_ps(_mm_mul_ps(v_y0, v_y0), v_y0); + v_y1 = _mm_mul_ps(_mm_mul_ps(v_y1, v_y1), v_y1); + __m128 v_d0 = _mm_div_ps(_mm_set1_ps(1.f/13.f), v_l0); + __m128 v_d1 = _mm_div_ps(_mm_set1_ps(1.f/13.f), v_l1); + v_u0 = _mm_mul_ps(_mm_add_ps(_mm_mul_ps(v_u0, v_d0), _mm_set1_ps(un)), _mm_set1_ps(3.f)); + v_u1 = _mm_mul_ps(_mm_add_ps(_mm_mul_ps(v_u1, v_d1), _mm_set1_ps(un)), _mm_set1_ps(3.f)); + v_v0 = _mm_add_ps(_mm_mul_ps(v_v0, v_d0), _mm_set1_ps(vn)); + v_v1 = _mm_add_ps(_mm_mul_ps(v_v1, v_d1), _mm_set1_ps(vn)); + __m128 v_iv0 = _mm_div_ps(_mm_set1_ps(0.25f), v_v0); + __m128 v_iv1 = _mm_div_ps(_mm_set1_ps(0.25f), v_v1); + __m128 v_x0 = _mm_mul_ps(_mm_mul_ps(_mm_set1_ps(3.f), v_u0), v_iv0); + __m128 v_x1 = _mm_mul_ps(_mm_mul_ps(_mm_set1_ps(3.f), v_u1), v_iv1); + __m128 v_z0 = _mm_mul_ps(_mm_sub_ps(_mm_sub_ps(_mm_set1_ps(12.f), v_u0), _mm_mul_ps(_mm_set1_ps(20.f), v_v0)), v_iv0); + __m128 v_z1 = _mm_mul_ps(_mm_sub_ps(_mm_sub_ps(_mm_set1_ps(12.f), v_u1), _mm_mul_ps(_mm_set1_ps(20.f), v_v1)), v_iv1); + + v_l0 = _mm_mul_ps(v_x0, _mm_set1_ps(coeffs[0])); + v_l1 = _mm_mul_ps(v_x1, _mm_set1_ps(coeffs[0])); + v_u0 = _mm_mul_ps(v_x0, _mm_set1_ps(coeffs[3])); + v_u1 = _mm_mul_ps(v_x1, _mm_set1_ps(coeffs[3])); + v_v0 = _mm_mul_ps(v_x0, _mm_set1_ps(coeffs[6])); + v_v1 = _mm_mul_ps(v_x1, _mm_set1_ps(coeffs[6])); + v_l0 = _mm_add_ps(v_l0, _mm_set1_ps(coeffs[1])); + v_l1 = _mm_add_ps(v_l1, _mm_set1_ps(coeffs[1])); + v_u0 = _mm_add_ps(v_u0, _mm_set1_ps(coeffs[4])); + v_u1 = _mm_add_ps(v_u1, _mm_set1_ps(coeffs[4])); + v_v0 = _mm_add_ps(v_v0, _mm_set1_ps(coeffs[7])); + v_v1 = _mm_add_ps(v_v1, _mm_set1_ps(coeffs[7])); + v_l0 = _mm_add_ps(v_l0, _mm_mul_ps(v_z0, _mm_set1_ps(coeffs[2]))); + v_l1 = _mm_add_ps(v_l1, _mm_mul_ps(v_z1, _mm_set1_ps(coeffs[2]))); + v_u0 = _mm_add_ps(v_u0, _mm_mul_ps(v_z0, _mm_set1_ps(coeffs[5]))); + v_u1 = _mm_add_ps(v_u1, _mm_mul_ps(v_z1, _mm_set1_ps(coeffs[5]))); + v_v0 = _mm_add_ps(v_v0, _mm_mul_ps(v_z0, _mm_set1_ps(coeffs[8]))); + v_v1 = _mm_add_ps(v_v1, _mm_mul_ps(v_z1, _mm_set1_ps(coeffs[8]))); + v_l0 = _mm_mul_ps(v_l0, v_y0); + v_l1 = _mm_mul_ps(v_l1, v_y1); + v_u0 = _mm_mul_ps(v_u0, v_y0); + v_u1 = _mm_mul_ps(v_u1, v_y1); + v_v0 = _mm_mul_ps(v_v0, v_y0); + v_v1 = _mm_mul_ps(v_v1, v_y1); + + v_l0 = _mm_max_ps(v_l0, _mm_setzero_ps()); + v_l1 = _mm_max_ps(v_l1, _mm_setzero_ps()); + v_u0 = _mm_max_ps(v_u0, _mm_setzero_ps()); + v_u1 = _mm_max_ps(v_u1, _mm_setzero_ps()); + v_v0 = _mm_max_ps(v_v0, _mm_setzero_ps()); + v_v1 = _mm_max_ps(v_v1, _mm_setzero_ps()); + v_l0 = _mm_min_ps(v_l0, _mm_set1_ps(1.f)); + v_l1 = _mm_min_ps(v_l1, _mm_set1_ps(1.f)); + v_u0 = _mm_min_ps(v_u0, _mm_set1_ps(1.f)); + v_u1 = _mm_min_ps(v_u1, _mm_set1_ps(1.f)); + v_v0 = _mm_min_ps(v_v0, _mm_set1_ps(1.f)); + v_v1 = _mm_min_ps(v_v1, _mm_set1_ps(1.f)); + } + #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], @@ -6011,7 +6075,69 @@ struct Luv2RGB_f float _un = un, _vn = vn; n *= 3; - 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_l0 = _mm_loadu_ps(src + i + 0); + __m128 v_l1 = _mm_loadu_ps(src + i + 4); + __m128 v_u0 = _mm_loadu_ps(src + i + 8); + __m128 v_u1 = _mm_loadu_ps(src + i + 12); + __m128 v_v0 = _mm_loadu_ps(src + i + 16); + __m128 v_v1 = _mm_loadu_ps(src + i + 20); + + _mm_deinterleave_ps(v_l0, v_l1, v_u0, v_u1, v_v0, v_v1); + + process(v_l0, v_l1, v_u0, v_u1, v_v0, v_v1); + + if( gammaTab ) + { + __m128 v_gscale = _mm_set1_ps(gscale); + v_l0 = _mm_mul_ps(v_l0, v_gscale); + v_l1 = _mm_mul_ps(v_l1, v_gscale); + v_u0 = _mm_mul_ps(v_u0, v_gscale); + v_u1 = _mm_mul_ps(v_u1, v_gscale); + v_v0 = _mm_mul_ps(v_v0, v_gscale); + v_v1 = _mm_mul_ps(v_v1, v_gscale); + splineInterpolate(v_l0, gammaTab, GAMMA_TAB_SIZE); + splineInterpolate(v_l1, gammaTab, GAMMA_TAB_SIZE); + splineInterpolate(v_u0, gammaTab, GAMMA_TAB_SIZE); + splineInterpolate(v_u1, gammaTab, GAMMA_TAB_SIZE); + splineInterpolate(v_v0, gammaTab, GAMMA_TAB_SIZE); + splineInterpolate(v_v1, 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_l0, v_l1, v_u0, v_u1, v_v0, v_v1, v_a0, v_a1); + + _mm_storeu_ps(dst + 0, v_l0); + _mm_storeu_ps(dst + 4, v_l1); + _mm_storeu_ps(dst + 8, v_u0); + _mm_storeu_ps(dst + 12, v_u1); + _mm_storeu_ps(dst + 16, v_v0); + _mm_storeu_ps(dst + 20, v_v1); + _mm_storeu_ps(dst + 24, v_a0); + _mm_storeu_ps(dst + 28, v_a1); + } + else + { + _mm_interleave_ps(v_l0, v_l1, v_u0, v_u1, v_v0, v_v1); + + _mm_storeu_ps(dst + 0, v_l0); + _mm_storeu_ps(dst + 4, v_l1); + _mm_storeu_ps(dst + 8, v_u0); + _mm_storeu_ps(dst + 12, v_u1); + _mm_storeu_ps(dst + 16, v_v0); + _mm_storeu_ps(dst + 20, v_v1); + } + } + } + #endif + for( ; i < n; i += 3, dst += dcn ) { float L = src[i], u = src[i+1], v = src[i+2], d, X, Y, Z; Y = (L + 16.f) * (1.f/116.f); @@ -6047,6 +6173,9 @@ struct Luv2RGB_f int dstcn; float coeffs[9], un, vn; bool srgb; + #if CV_SSE2 + bool haveSIMD; + #endif };