generic-library/opencv
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

cv::phase; cv::cartToPolar

Browse Source
This commit is contained in:
Ilya Lavrenov 2014-10-12 03:04:03 -07:00
parent 7e6fb668ed
commit fb97273b3c
2 changed files with 41 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
modules/core/include/opencv2/core/base.hpp
View File

@ -605,6 +605,22 @@ inline uint32x4_t cv_vrndq_u32_f32(float32x4_t v)
return vcvtq_u32_f32(vaddq_f32(v, v_05));
}
inline float32x4_t cv_vrecpq_f32(float32x4_t val)
{
float32x4_t reciprocal = vrecpeq_f32(val);
reciprocal = vmulq_f32(vrecpsq_f32(val, reciprocal), reciprocal);
reciprocal = vmulq_f32(vrecpsq_f32(val, reciprocal), reciprocal);
return reciprocal;
}
inline float32x2_t cv_vrecp_f32(float32x2_t val)
{
float32x2_t reciprocal = vrecpe_f32(val);
reciprocal = vmul_f32(vrecps_f32(val, reciprocal), reciprocal);
reciprocal = vmul_f32(vrecps_f32(val, reciprocal), reciprocal);
return reciprocal;
}
#endif
} // cv

25
modules/core/src/mathfuncs.cpp
View File

@ -168,6 +168,31 @@ static void FastAtan2_32f(const float *Y, const float *X, float *angle, int len,
_mm_storeu_ps(angle + i, a);
}
}
#elif CV_NEON
float32x4_t eps = vdupq_n_f32((float)DBL_EPSILON);
float32x4_t _90 = vdupq_n_f32(90.f), _180 = vdupq_n_f32(180.f), _360 = vdupq_n_f32(360.f);
float32x4_t z = vdupq_n_f32(0.0f), scale4 = vdupq_n_f32(scale);
float32x4_t p1 = vdupq_n_f32(atan2_p1), p3 = vdupq_n_f32(atan2_p3);
float32x4_t p5 = vdupq_n_f32(atan2_p5), p7 = vdupq_n_f32(atan2_p7);
for( ; i <= len - 4; i += 4 )
{
float32x4_t x = vld1q_f32(X + i), y = vld1q_f32(Y + i);
float32x4_t ax = vabsq_f32(x), ay = vabsq_f32(y);
float32x4_t tmin = vminq_f32(ax, ay), tmax = vmaxq_f32(ax, ay);
float32x4_t c = vmulq_f32(tmin, cv_vrecpq_f32(vaddq_f32(tmax, eps)));
float32x4_t c2 = vmulq_f32(c, c);
float32x4_t a = vmulq_f32(c2, p7);
a = vmulq_f32(vaddq_f32(a, p5), c2);
a = vmulq_f32(vaddq_f32(a, p3), c2);
a = vmulq_f32(vaddq_f32(a, p1), c);
a = vbslq_f32(vcgeq_f32(ax, ay), a, vsubq_f32(_90, a));
a = vbslq_f32(vcltq_f32(x, z), vsubq_f32(_180, a), a);
a = vbslq_f32(vcltq_f32(y, z), vsubq_f32(_360, a), a);
vst1q_f32(angle + i, vmulq_f32(a, scale4));
}
#endif
for( ; i < len; i++ )