potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-29 15:23:05 +04:00
Code Packages Releases Wiki Activity

Clean up the Universal Intrinsic API.

Browse Source
This commit is contained in:
Liutong HAN
2023-10-13 19:23:30 +08:00
parent 0507043a55
commit a287605c3e
40 changed files with 1611 additions and 1531 deletions
Download Patch File Download Diff File Expand all files Collapse all files
samples/cpp/simd_basic.cpp
+2 -2
View File
@@ -38,8 +38,8 @@ int main(int /*argc*/, char** /*argv*/)
printf("================== arithm check =================\n");
v_uint8 a = vx_setall_u8(10);
v_uint8 c = a + vx_setall_u8(45);
printf("(vx_setall_u8(10) + vx_setall_u8(45)).get0() => %d\n", (int)c.get0());
v_uint8 c = v_add(a, vx_setall_u8(45));
printf("v_get0(vx_setall_u8(10) + vx_setall_u8(45)) => %d\n", (int)v_get0(c));
#else
printf("\nSIMD intrinsics are not available. Check compilation target and passed build options.\n");
#endif
samples/cpp/tutorial_code/core/univ_intrin/univ_intrin.cpp
+4 -4
View File
@@ -85,7 +85,7 @@ void conv1dsimd(Mat src, Mat kernel, float *ans, int row = 0, int rowk = 0, int
//! [convolution-1D-main]
//! [convolution-1D-main-h1]
int step = v_float32().nlanes;
int step = VTraits<v_float32x4>::vlanes();
float *sptr = src_32.ptr<float>(row), *kptr = kernel.ptr<float>(rowk);
for (int k = 0; k < ksize; k++)
{
@@ -96,7 +96,7 @@ void conv1dsimd(Mat src, Mat kernel, float *ans, int row = 0, int rowk = 0, int
for (i = 0; i + step < len; i += step)
{
v_float32 window = vx_load(sptr + i + k);
v_float32 sum = vx_load(ans + i) + kernel_wide * window;
v_float32 sum = v_add(vx_load(ans + i), v_mul(kernel_wide, window));
v_store(ans + i, sum);
}
//! [convolution-1D-main-h2]
@@ -122,7 +122,7 @@ void convolute_simd(Mat src, Mat &dst, Mat kernel)
copyMakeBorder(src, src, sz, sz, 0, 0, BORDER_REPLICATE);
int step = v_float32().nlanes;
int step = VTraits<v_float32x4>::vlanes();
//! [convolution-2D-init]
//! [convolution-2D-main]
@@ -135,7 +135,7 @@ void convolute_simd(Mat src, Mat &dst, Mat kernel)
int j;
for (j = 0; j + step < cols; j += step)
{
v_float32 sum = vx_load(&dst.ptr<float>(i)[j]) + vx_load(&ans[j]);
v_float32 sum = v_add(vx_load(&dst.ptr<float>(i)[j]), vx_load(&ans[j]));
v_store(&dst.ptr<float>(i)[j], sum);
}