mirror/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2025-06-11 11:45:30 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

imgproc: fix warp optimizations

Browse Source
This commit is contained in:
Alexander Alekhin 2017-07-17 15:12:41 +03:00
parent b4716b1d92
commit 4bb4a349c9
2 changed files with 22 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
modules/imgproc/src/imgwarp.avx2.cpp
View File

@ -125,6 +125,7 @@ public:
}
}
}
_mm256_zeroupper();
}
private:
@ -230,6 +231,7 @@ public:
}
}
}
_mm256_zeroupper();
}
private:

41
modules/imgproc/src/imgwarp.sse4_1.cpp
View File

@ -387,19 +387,19 @@ class WarpPerspectiveLine_SSE4_Impl: public WarpPerspectiveLine_SSE4
public:
WarpPerspectiveLine_SSE4_Impl(const double *M)
{
v_M0 = _mm_set1_pd(M[0]);
v_M3 = _mm_set1_pd(M[3]);
v_M6 = _mm_set1_pd(M[6]);
v_intmax = _mm_set1_pd((double)INT_MAX);
v_intmin = _mm_set1_pd((double)INT_MIN);
v_2 = _mm_set1_pd(2);
v_zero = _mm_setzero_pd();
v_1 = _mm_set1_pd(1);
v_its = _mm_set1_pd(INTER_TAB_SIZE);
v_itsi1 = _mm_set1_epi32(INTER_TAB_SIZE - 1);
CV_UNUSED(M);
}
virtual void processNN(const double *M, short* xy, double X0, double Y0, double W0, int bw)
{
const __m128d v_M0 = _mm_set1_pd(M[0]);
const __m128d v_M3 = _mm_set1_pd(M[3]);
const __m128d v_M6 = _mm_set1_pd(M[6]);
const __m128d v_intmax = _mm_set1_pd((double)INT_MAX);
const __m128d v_intmin = _mm_set1_pd((double)INT_MIN);
const __m128d v_2 = _mm_set1_pd(2);
const __m128d v_zero = _mm_setzero_pd();
const __m128d v_1 = _mm_set1_pd(1);
int x1 = 0;
__m128d v_X0d = _mm_set1_pd(X0);
__m128d v_Y0d = _mm_set1_pd(Y0);
@ -521,6 +521,16 @@ public:
}
virtual void process(const double *M, short* xy, short* alpha, double X0, double Y0, double W0, int bw)
{
const __m128d v_M0 = _mm_set1_pd(M[0]);
const __m128d v_M3 = _mm_set1_pd(M[3]);
const __m128d v_M6 = _mm_set1_pd(M[6]);
const __m128d v_intmax = _mm_set1_pd((double)INT_MAX);
const __m128d v_intmin = _mm_set1_pd((double)INT_MIN);
const __m128d v_2 = _mm_set1_pd(2);
const __m128d v_zero = _mm_setzero_pd();
const __m128d v_its = _mm_set1_pd(INTER_TAB_SIZE);
const __m128i v_itsi1 = _mm_set1_epi32(INTER_TAB_SIZE - 1);
int x1 = 0;
__m128d v_X0d = _mm_set1_pd(X0);
@ -656,17 +666,6 @@ public:
}
}
virtual ~WarpPerspectiveLine_SSE4_Impl() {};
private:
__m128d v_M0;
__m128d v_M3;
__m128d v_M6;
__m128d v_intmax;
__m128d v_intmin;
__m128d v_2,
v_zero,
v_1,
v_its;
__m128i v_itsi1;
};
Ptr<WarpPerspectiveLine_SSE4> WarpPerspectiveLine_SSE4::getImpl(const double *M)