opencv/modules/core/perf/perf_arithm.cpp
Daniil Anufriev b5f5540e8a
Merge pull request #26886 from sk1er52:feature/exp64f
Enable SIMD_SCALABLE for exp and sqrt #26886

### Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [x] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
      Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake
```
CPU - Banana Pi k1, compiler - clang 18.1.4
```
```
Geometric mean (ms)

              Name of Test               baseline  hal     ui      hal         ui    
                                                                    vs         vs
                                                                 baseline   baseline
                                                                (x-factor) (x-factor)
Exp::ExpFixture::(127x61, 32FC1)          0.358     --   0.033      --       10.70   
Exp::ExpFixture::(640x480, 32FC1)         14.304    --   1.167      --       12.26   
Exp::ExpFixture::(1280x720, 32FC1)        42.785    --   3.538      --       12.09
Exp::ExpFixture::(1920x1080, 32FC1)       96.206    --   7.927      --       12.14   
Exp::ExpFixture::(127x61, 64FC1)          0.433   0.050  0.098     8.59       4.40   
Exp::ExpFixture::(640x480, 64FC1)         17.315  1.935  3.813     8.95       4.54   
Exp::ExpFixture::(1280x720, 64FC1)        52.181  5.877  11.519    8.88       4.53   
Exp::ExpFixture::(1920x1080, 64FC1)      117.082  13.157 25.854    8.90       4.53
```
Additionally, this PR brings Sqrt optimization with UI:
```
Geometric mean (ms)

              Name of Test                     baseline    ui       ui    
                                                                    vs
                                                                 baseline
                                                                (x-factor)
Sqrt::SqrtFixture::(127x61, 5, false)            0.111   0.027     4.11   
Sqrt::SqrtFixture::(127x61, 6, false)            0.149   0.053     2.82   
Sqrt::SqrtFixture::(640x480, 5, false)           4.374   0.967     4.52   
Sqrt::SqrtFixture::(640x480, 6, false)           5.885   2.046     2.88   
Sqrt::SqrtFixture::(1280x720, 5, false)          12.960  2.915     4.45   
Sqrt::SqrtFixture::(1280x720, 6, false)          17.648  6.107     2.89   
Sqrt::SqrtFixture::(1920x1080, 5, false)         29.178  6.524     4.47   
Sqrt::SqrtFixture::(1920x1080, 6, false)         39.709  13.670    2.90   
```

Reference
Muller, J.-M. Elementary Functions: Algorithms and Implementation. 2nd ed. Boston: Birkhäuser, 2006.
https://www.springer.com/gp/book/9780817643720
2025-02-21 17:36:54 +03:00

849 lines
20 KiB
C++

#include "perf_precomp.hpp"
#include <numeric>
#include "opencv2/core/softfloat.hpp"
namespace opencv_test
{
using namespace perf;
using BroadcastTest = perf::TestBaseWithParam<std::tuple<std::vector<int>, perf::MatType, std::vector<int>>>;
typedef Size_MatType BinaryOpTest;
PERF_TEST_P_(BroadcastTest, basic)
{
std::vector<int> shape_src = get<0>(GetParam());
int dt_type = get<1>(GetParam());
std::vector<int> shape_dst = get<2>(GetParam());
cv::Mat src(static_cast<int>(shape_src.size()), shape_src.data(), dt_type);
cv::Mat dst(static_cast<int>(shape_dst.size()), shape_dst.data(), dt_type);
cv::randu(src, -1.f, 1.f);
TEST_CYCLE() cv::broadcast(src, shape_dst, dst);
SANITY_CHECK_NOTHING();
}
INSTANTIATE_TEST_CASE_P(/*nothing*/ , BroadcastTest,
testing::Combine(
testing::Values(std::vector<int>{1, 100, 800},
std::vector<int>{10, 1, 800},
std::vector<int>{10, 100, 1}),
testing::Values(CV_32FC1),
testing::Values(std::vector<int>{10, 100, 800})
)
);
PERF_TEST_P_(BinaryOpTest, min)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a = Mat(sz, type);
cv::Mat b = Mat(sz, type);
cv::Mat c = Mat(sz, type);
declare.in(a, b, WARMUP_RNG).out(c);
TEST_CYCLE() cv::min(a, b, c);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, minScalarDouble)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a = Mat(sz, type);
cv::Scalar b;
cv::Mat c = Mat(sz, type);
declare.in(a, b, WARMUP_RNG).out(c);
TEST_CYCLE() cv::min(a, b, c);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, minScalarSameType)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a = Mat(sz, type);
cv::Scalar b;
cv::Mat c = Mat(sz, type);
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(type) < CV_32S)
{
b = Scalar(1, 0, 3, 4); // don't pass non-integer values for 8U/8S/16U/16S processing
}
else if (CV_MAT_DEPTH(type) == CV_32S)
{
b = Scalar(1, 0, -3, 4); // don't pass non-integer values for 32S processing
}
TEST_CYCLE() cv::min(a, b, c);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, max)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a = Mat(sz, type);
cv::Mat b = Mat(sz, type);
cv::Mat c = Mat(sz, type);
declare.in(a, b, WARMUP_RNG).out(c);
TEST_CYCLE() cv::max(a, b, c);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, maxScalarDouble)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a = Mat(sz, type);
cv::Scalar b;
cv::Mat c = Mat(sz, type);
declare.in(a, b, WARMUP_RNG).out(c);
TEST_CYCLE() cv::max(a, b, c);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, maxScalarSameType)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a = Mat(sz, type);
cv::Scalar b;
cv::Mat c = Mat(sz, type);
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(type) < CV_32S)
{
b = Scalar(1, 0, 3, 4); // don't pass non-integer values for 8U/8S/16U/16S processing
}
else if (CV_MAT_DEPTH(type) == CV_32S)
{
b = Scalar(1, 0, -3, 4); // don't pass non-integer values for 32S processing
}
TEST_CYCLE() cv::max(a, b, c);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, absdiff)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a = Mat(sz, type);
cv::Mat b = Mat(sz, type);
cv::Mat c = Mat(sz, type);
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(type) == CV_32S)
{
//see ticket 1529: absdiff can be without saturation on 32S
a /= 2;
b /= 2;
}
TEST_CYCLE() cv::absdiff(a, b, c);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, absdiffScalarDouble)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a = Mat(sz, type);
cv::Scalar b;
cv::Mat c = Mat(sz, type);
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(type) == CV_32S)
{
//see ticket 1529: absdiff can be without saturation on 32S
a /= 2;
b /= 2;
}
TEST_CYCLE() cv::absdiff(a, b, c);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, absdiffScalarSameType)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a = Mat(sz, type);
cv::Scalar b;
cv::Mat c = Mat(sz, type);
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(type) < CV_32S)
{
b = Scalar(1, 0, 3, 4); // don't pass non-integer values for 8U/8S/16U/16S processing
}
else if (CV_MAT_DEPTH(type) == CV_32S)
{
//see ticket 1529: absdiff can be without saturation on 32S
a /= 2;
b = Scalar(1, 0, -3, 4); // don't pass non-integer values for 32S processing
}
TEST_CYCLE() cv::absdiff(a, b, c);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, add)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a = Mat(sz, type);
cv::Mat b = Mat(sz, type);
cv::Mat c = Mat(sz, type);
declare.in(a, b, WARMUP_RNG).out(c);
declare.time(50);
if (CV_MAT_DEPTH(type) == CV_32S)
{
//see ticket 1529: add can be without saturation on 32S
a /= 2;
b /= 2;
}
TEST_CYCLE() cv::add(a, b, c);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, addScalarDouble)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a = Mat(sz, type);
cv::Scalar b;
cv::Mat c = Mat(sz, type);
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(type) == CV_32S)
{
//see ticket 1529: add can be without saturation on 32S
a /= 2;
b /= 2;
}
TEST_CYCLE() cv::add(a, b, c);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, addScalarSameType)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a = Mat(sz, type);
cv::Scalar b;
cv::Mat c = Mat(sz, type);
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(type) < CV_32S)
{
b = Scalar(1, 0, 3, 4); // don't pass non-integer values for 8U/8S/16U/16S processing
}
else if (CV_MAT_DEPTH(type) == CV_32S)
{
//see ticket 1529: add can be without saturation on 32S
a /= 2;
b = Scalar(1, 0, -3, 4); // don't pass non-integer values for 32S processing
}
TEST_CYCLE() cv::add(a, b, c, noArray(), type);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, subtract)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a = Mat(sz, type);
cv::Mat b = Mat(sz, type);
cv::Mat c = Mat(sz, type);
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(type) == CV_32S)
{
//see ticket 1529: subtract can be without saturation on 32S
a /= 2;
b /= 2;
}
TEST_CYCLE() cv::subtract(a, b, c);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, subtractScalarDouble)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a = Mat(sz, type);
cv::Scalar b;
cv::Mat c = Mat(sz, type);
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(type) == CV_32S)
{
//see ticket 1529: subtract can be without saturation on 32S
a /= 2;
b /= 2;
}
TEST_CYCLE() cv::subtract(a, b, c);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, subtractScalarSameType)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a = Mat(sz, type);
cv::Scalar b;
cv::Mat c = Mat(sz, type);
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(type) < CV_32S)
{
b = Scalar(1, 0, 3, 4); // don't pass non-integer values for 8U/8S/16U/16S processing
}
else if (CV_MAT_DEPTH(type) == CV_32S)
{
//see ticket 1529: subtract can be without saturation on 32S
a /= 2;
b = Scalar(1, 0, -3, 4); // don't pass non-integer values for 32S processing
}
TEST_CYCLE() cv::subtract(a, b, c, noArray(), type);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, multiply)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a(sz, type), b(sz, type), c(sz, type);
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(type) == CV_32S)
{
//According to docs, saturation is not applied when result is 32bit integer
a /= (2 << 16);
b /= (2 << 16);
}
TEST_CYCLE() cv::multiply(a, b, c);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, multiplyScale)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a(sz, type), b(sz, type), c(sz, type);
double scale = 0.5;
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(type) == CV_32S)
{
//According to docs, saturation is not applied when result is 32bit integer
a /= (2 << 16);
b /= (2 << 16);
}
TEST_CYCLE() cv::multiply(a, b, c, scale);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, divide)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a(sz, type), b(sz, type), c(sz, type);
double scale = 0.5;
declare.in(a, b, WARMUP_RNG).out(c);
TEST_CYCLE() cv::divide(a, b, c, scale);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, reciprocal)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat b(sz, type), c(sz, type);
double scale = 0.5;
declare.in(b, WARMUP_RNG).out(c);
TEST_CYCLE() cv::divide(scale, b, c);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(BinaryOpTest, transposeND)
{
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
cv::Mat a = Mat(sz, type).reshape(1);
std::vector<int> order(a.dims);
std::iota(order.begin(), order.end(), 0);
std::reverse(order.begin(), order.end());
std::vector<int> new_sz(a.dims);
std::copy(a.size.p, a.size.p + a.dims, new_sz.begin());
std::reverse(new_sz.begin(), new_sz.end());
cv::Mat b = Mat(new_sz, type);
declare.in(a,WARMUP_RNG).out(b);
TEST_CYCLE() cv::transposeND(a, order, b);
SANITY_CHECK_NOTHING();
}
INSTANTIATE_TEST_CASE_P(/*nothing*/ , BinaryOpTest,
testing::Combine(
testing::Values(szVGA, sz720p, sz1080p),
testing::Values(CV_8UC1, CV_8UC3, CV_8UC4, CV_8SC1, CV_16SC1, CV_16SC2, CV_16SC3, CV_16SC4, CV_32SC1, CV_32FC1)
)
);
///////////// Mixed type arithmetics ////////
typedef perf::TestBaseWithParam<std::tuple<cv::Size, std::tuple<perf::MatType, perf::MatType>>> ArithmMixedTest;
PERF_TEST_P_(ArithmMixedTest, add)
{
auto p = GetParam();
Size sz = get<0>(p);
int srcType = get<0>(get<1>(p));
int dstType = get<1>(get<1>(p));
cv::Mat a = Mat(sz, srcType);
cv::Mat b = Mat(sz, srcType);
cv::Mat c = Mat(sz, dstType);
declare.in(a, b, WARMUP_RNG).out(c);
declare.time(50);
if (CV_MAT_DEPTH(dstType) == CV_32S)
{
//see ticket 1529: add can be without saturation on 32S
a /= 2;
b /= 2;
}
TEST_CYCLE() cv::add(a, b, c, /* mask */ noArray(), dstType);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(ArithmMixedTest, addScalarDouble)
{
auto p = GetParam();
Size sz = get<0>(p);
int srcType = get<0>(get<1>(p));
int dstType = get<1>(get<1>(p));
cv::Mat a = Mat(sz, srcType);
cv::Scalar b;
cv::Mat c = Mat(sz, dstType);
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(dstType) == CV_32S)
{
//see ticket 1529: add can be without saturation on 32S
a /= 2;
b /= 2;
}
TEST_CYCLE() cv::add(a, b, c, /* mask */ noArray(), dstType);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(ArithmMixedTest, addScalarSameType)
{
auto p = GetParam();
Size sz = get<0>(p);
int srcType = get<0>(get<1>(p));
int dstType = get<1>(get<1>(p));
cv::Mat a = Mat(sz, srcType);
cv::Scalar b;
cv::Mat c = Mat(sz, dstType);
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(dstType) < CV_32S)
{
b = Scalar(1, 0, 3, 4); // don't pass non-integer values for 8U/8S/16U/16S processing
}
else if (CV_MAT_DEPTH(dstType) == CV_32S)
{
//see ticket 1529: add can be without saturation on 32S
a /= 2;
b = Scalar(1, 0, -3, 4); // don't pass non-integer values for 32S processing
}
TEST_CYCLE() cv::add(a, b, c, /* mask */ noArray(), dstType);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(ArithmMixedTest, subtract)
{
auto p = GetParam();
Size sz = get<0>(p);
int srcType = get<0>(get<1>(p));
int dstType = get<1>(get<1>(p));
cv::Mat a = Mat(sz, srcType);
cv::Mat b = Mat(sz, srcType);
cv::Mat c = Mat(sz, dstType);
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(dstType) == CV_32S)
{
//see ticket 1529: subtract can be without saturation on 32S
a /= 2;
b /= 2;
}
TEST_CYCLE() cv::subtract(a, b, c, /* mask */ noArray(), dstType);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(ArithmMixedTest, subtractScalarDouble)
{
auto p = GetParam();
Size sz = get<0>(p);
int srcType = get<0>(get<1>(p));
int dstType = get<1>(get<1>(p));
cv::Mat a = Mat(sz, srcType);
cv::Scalar b;
cv::Mat c = Mat(sz, dstType);
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(dstType) == CV_32S)
{
//see ticket 1529: subtract can be without saturation on 32S
a /= 2;
b /= 2;
}
TEST_CYCLE() cv::subtract(a, b, c, /* mask */ noArray(), dstType);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(ArithmMixedTest, subtractScalarSameType)
{
auto p = GetParam();
Size sz = get<0>(p);
int srcType = get<0>(get<1>(p));
int dstType = get<1>(get<1>(p));
cv::Mat a = Mat(sz, srcType);
cv::Scalar b;
cv::Mat c = Mat(sz, dstType);
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(dstType) < CV_32S)
{
b = Scalar(1, 0, 3, 4); // don't pass non-integer values for 8U/8S/16U/16S processing
}
else if (CV_MAT_DEPTH(dstType) == CV_32S)
{
//see ticket 1529: subtract can be without saturation on 32S
a /= 2;
b = Scalar(1, 0, -3, 4); // don't pass non-integer values for 32S processing
}
TEST_CYCLE() cv::subtract(a, b, c, /* mask */ noArray(), dstType);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(ArithmMixedTest, multiply)
{
auto p = GetParam();
Size sz = get<0>(p);
int srcType = get<0>(get<1>(p));
int dstType = get<1>(get<1>(p));
cv::Mat a(sz, srcType), b(sz, srcType), c(sz, dstType);
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(dstType) == CV_32S)
{
//According to docs, saturation is not applied when result is 32bit integer
a /= (2 << 16);
b /= (2 << 16);
}
TEST_CYCLE() cv::multiply(a, b, c, /* scale */ 1.0, dstType);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(ArithmMixedTest, multiplyScale)
{
auto p = GetParam();
Size sz = get<0>(p);
int srcType = get<0>(get<1>(p));
int dstType = get<1>(get<1>(p));
cv::Mat a(sz, srcType), b(sz, srcType), c(sz, dstType);
double scale = 0.5;
declare.in(a, b, WARMUP_RNG).out(c);
if (CV_MAT_DEPTH(dstType) == CV_32S)
{
//According to docs, saturation is not applied when result is 32bit integer
a /= (2 << 16);
b /= (2 << 16);
}
TEST_CYCLE() cv::multiply(a, b, c, scale, dstType);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(ArithmMixedTest, divide)
{
auto p = GetParam();
Size sz = get<0>(p);
int srcType = get<0>(get<1>(p));
int dstType = get<1>(get<1>(p));
cv::Mat a(sz, srcType), b(sz, srcType), c(sz, dstType);
double scale = 0.5;
declare.in(a, b, WARMUP_RNG).out(c);
TEST_CYCLE() cv::divide(a, b, c, scale, dstType);
SANITY_CHECK_NOTHING();
}
PERF_TEST_P_(ArithmMixedTest, reciprocal)
{
auto p = GetParam();
Size sz = get<0>(p);
int srcType = get<0>(get<1>(p));
int dstType = get<1>(get<1>(p));
cv::Mat b(sz, srcType), c(sz, dstType);
double scale = 0.5;
declare.in(b, WARMUP_RNG).out(c);
TEST_CYCLE() cv::divide(scale, b, c, dstType);
SANITY_CHECK_NOTHING();
}
INSTANTIATE_TEST_CASE_P(/*nothing*/ , ArithmMixedTest,
testing::Combine(
testing::Values(szVGA, sz720p, sz1080p),
testing::Values(std::tuple<perf::MatType, perf::MatType>{CV_8U, CV_16U},
std::tuple<perf::MatType, perf::MatType>{CV_8S, CV_16S},
std::tuple<perf::MatType, perf::MatType>{CV_8U, CV_32F},
std::tuple<perf::MatType, perf::MatType>{CV_8S, CV_32F}
)
)
);
typedef perf::TestBaseWithParam<std::tuple<cv::Size, int, bool>> SqrtFixture;
PERF_TEST_P_(SqrtFixture, Sqrt) {
Size sz = get<0>(GetParam());
int type = get<1>(GetParam());
bool inverse = get<2>(GetParam());
Mat src(sz, type), dst(sz, type);
randu(src, FLT_EPSILON, 1000);
declare.in(src).out(dst);
TEST_CYCLE() cv::pow(src, inverse ? -0.5 : 0.5, dst);
SANITY_CHECK_NOTHING();
}
INSTANTIATE_TEST_CASE_P(/*nothing*/ , SqrtFixture,
testing::Combine(
testing::Values(TYPICAL_MAT_SIZES),
testing::Values(CV_32FC1, CV_64FC1),
testing::Bool()
)
);
///////////// Rotate ////////////////////////
typedef perf::TestBaseWithParam<std::tuple<cv::Size, int, perf::MatType>> RotateTest;
PERF_TEST_P_(RotateTest, rotate)
{
Size sz = get<0>(GetParam());
int rotatecode = get<1>(GetParam());
int type = get<2>(GetParam());
cv::Mat a(sz, type), b(sz, type);
declare.in(a, WARMUP_RNG).out(b);
TEST_CYCLE() cv::rotate(a, b, rotatecode);
SANITY_CHECK_NOTHING();
}
INSTANTIATE_TEST_CASE_P(/*nothing*/ , RotateTest,
testing::Combine(
testing::Values(szVGA, sz720p, sz1080p),
testing::Values(ROTATE_180, ROTATE_90_CLOCKWISE, ROTATE_90_COUNTERCLOCKWISE),
testing::Values(CV_8UC1, CV_8UC2, CV_8UC3, CV_8UC4, CV_8SC1, CV_16SC1, CV_16SC2, CV_16SC3, CV_16SC4, CV_32SC1, CV_32FC1)
)
);
///////////// PatchNaNs ////////////////////////
template<typename _Tp>
_Tp randomNan(RNG& rng);
template<>
float randomNan(RNG& rng)
{
uint32_t r = rng.next();
Cv32suf v;
v.u = r;
// exp & set a bit to avoid zero mantissa
v.u = v.u | 0x7f800001;
return v.f;
}
template<>
double randomNan(RNG& rng)
{
uint32_t r0 = rng.next();
uint32_t r1 = rng.next();
Cv64suf v;
v.u = (uint64_t(r0) << 32) | uint64_t(r1);
// exp &set a bit to avoid zero mantissa
v.u = v.u | 0x7ff0000000000001;
return v.f;
}
typedef Size_MatType PatchNaNsFixture;
PERF_TEST_P_(PatchNaNsFixture, PatchNaNs)
{
const Size_MatType_t params = GetParam();
Size srcSize = get<0>(params);
const int type = get<1>(params), cn = CV_MAT_CN(type);
Mat src(srcSize, type);
declare.in(src, WARMUP_RNG).out(src);
// generating NaNs
{
srcSize.width *= cn;
RNG& rng = theRNG();
for (int y = 0; y < srcSize.height; ++y)
{
float *const ptrf = src.ptr<float>(y);
for (int x = 0; x < srcSize.width; ++x)
{
ptrf[x] = (x + y) % 2 == 0 ? randomNan<float >(rng) : ptrf[x];
}
}
}
TEST_CYCLE() cv::patchNaNs(src, 17.7);
SANITY_CHECK(src);
}
INSTANTIATE_TEST_CASE_P(/*nothing*/ , PatchNaNsFixture,
testing::Combine(
testing::Values(szVGA, sz720p, sz1080p, sz2160p),
testing::Values(CV_32FC1, CV_32FC2, CV_32FC3, CV_32FC4)
)
);
//////////////EXP////////////
typedef Size_MatType ExpFixture;
PERF_TEST_P(ExpFixture, Exp,
testing::Combine(testing::Values(TYPICAL_MAT_SIZES), testing::Values(CV_32F, CV_64F)))
{
cv::Size size = std::get<0>(GetParam());
int type = std::get<1>(GetParam());
cv::Mat src(size, type);
cv::Mat dst(size, type);
declare.in(src).out(dst);
cv::randu(src, -5.0, 5.0);
TEST_CYCLE()
{
cv::exp(src, dst);
}
SANITY_CHECK_NOTHING();
}
} // namespace