| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | #include "gutyansky_a_img_contrast_incr/all/include/ops_all.hpp" | ||
| 2 | |||
| 3 | #include <mpi.h> | ||
| 4 | |||
| 5 | #include <algorithm> | ||
| 6 | #include <array> | ||
| 7 | #include <cstddef> | ||
| 8 | #include <cstdint> | ||
| 9 | #include <limits> | ||
| 10 | #include <utility> | ||
| 11 | #include <vector> | ||
| 12 | |||
| 13 | #include "gutyansky_a_img_contrast_incr/common/include/common.hpp" | ||
| 14 | |||
| 15 | namespace gutyansky_a_img_contrast_incr { | ||
| 16 | |||
| 17 |
1/2✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
|
20 | GutyanskyAImgContrastIncrALL::GutyanskyAImgContrastIncrALL(const InType &in) { |
| 18 | SetTypeOfTask(GetStaticTypeOfTask()); | ||
| 19 |
1/2✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
|
20 | GetInput() = in; |
| 20 | GetOutput() = {}; | ||
| 21 | 20 | } | |
| 22 | |||
| 23 | 20 | bool GutyanskyAImgContrastIncrALL::ValidationImpl() { | |
| 24 | 20 | return !GetInput().empty(); | |
| 25 | } | ||
| 26 | |||
| 27 | 20 | bool GutyanskyAImgContrastIncrALL::PreProcessingImpl() { | |
| 28 | 20 | return true; | |
| 29 | } | ||
| 30 | |||
| 31 | 20 | bool GutyanskyAImgContrastIncrALL::RunImpl() { | |
| 32 | 20 | int rank = -1; | |
| 33 | 20 | int p_count = -1; | |
| 34 | |||
| 35 | 20 | MPI_Comm_rank(MPI_COMM_WORLD, &rank); | |
| 36 | 20 | MPI_Comm_size(MPI_COMM_WORLD, &p_count); | |
| 37 | |||
| 38 | 20 | size_t sz = GetInput().size(); | |
| 39 | |||
| 40 | 20 | MPI_Bcast(&sz, 1, MPI_UINT64_T, 0, MPI_COMM_WORLD); | |
| 41 | |||
| 42 | 20 | size_t chunk_size = sz / static_cast<size_t>(p_count); | |
| 43 | 20 | size_t remainder_size = sz % static_cast<size_t>(p_count); | |
| 44 | |||
| 45 |
1/2✓ Branch 0 taken 20 times.
✗ Branch 1 not taken.
|
20 | size_t local_count = chunk_size + (std::cmp_less(rank, remainder_size) ? 1 : 0); |
| 46 | |||
| 47 | 20 | std::vector<uint8_t> local_data(local_count); | |
| 48 | |||
| 49 |
1/4✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
20 | std::vector<int> send_counts(p_count); |
| 50 |
1/4✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
20 | std::vector<int> locs(p_count); |
| 51 | |||
| 52 |
2/2✓ Branch 0 taken 40 times.
✓ Branch 1 taken 20 times.
|
60 | for (int i = 0; i < p_count; i++) { |
| 53 | 40 | size_t elems_for_proc = chunk_size + (std::cmp_less(i, remainder_size) ? 1 : 0); | |
| 54 | 40 | send_counts[i] = static_cast<int>(elems_for_proc); | |
| 55 | } | ||
| 56 | |||
| 57 |
2/2✓ Branch 0 taken 20 times.
✓ Branch 1 taken 20 times.
|
40 | for (int i = 1; i < p_count; i++) { |
| 58 | 20 | locs[i] = locs[i - 1] + send_counts[i - 1]; | |
| 59 | } | ||
| 60 | |||
| 61 |
1/2✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
|
20 | MPI_Scatterv(GetInput().data(), send_counts.data(), locs.data(), MPI_UINT8_T, local_data.data(), |
| 62 | static_cast<int>(local_count), MPI_UINT8_T, 0, MPI_COMM_WORLD); | ||
| 63 | |||
| 64 | 20 | uint8_t local_lower_bound = std::numeric_limits<uint8_t>::max(); | |
| 65 | 20 | uint8_t local_upper_bound = std::numeric_limits<uint8_t>::min(); | |
| 66 | |||
| 67 | 20 | #pragma omp parallel for default(none) shared(local_data, local_count) reduction(min : local_lower_bound) \ | |
| 68 | reduction(max : local_upper_bound) | ||
| 69 | for (size_t i = 0; i < local_count; i++) { | ||
| 70 | uint8_t val = local_data[i]; | ||
| 71 | local_lower_bound = std::min(local_lower_bound, val); | ||
| 72 | local_upper_bound = std::max(local_upper_bound, val); | ||
| 73 | } | ||
| 74 | |||
| 75 | 20 | uint8_t lower_bound = std::numeric_limits<uint8_t>::max(); | |
| 76 | 20 | uint8_t upper_bound = std::numeric_limits<uint8_t>::min(); | |
| 77 | |||
| 78 |
1/2✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
|
20 | MPI_Allreduce(&local_lower_bound, &lower_bound, 1, MPI_UINT8_T, MPI_MIN, MPI_COMM_WORLD); |
| 79 |
1/2✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
|
20 | MPI_Allreduce(&local_upper_bound, &upper_bound, 1, MPI_UINT8_T, MPI_MAX, MPI_COMM_WORLD); |
| 80 | |||
| 81 | 20 | uint8_t delta = upper_bound - lower_bound; | |
| 82 | |||
| 83 |
2/2✓ Branch 0 taken 12 times.
✓ Branch 1 taken 8 times.
|
20 | if (delta != 0) { |
| 84 | 12 | std::array<uint8_t, 256> lut{}; | |
| 85 |
2/2✓ Branch 0 taken 3072 times.
✓ Branch 1 taken 12 times.
|
3084 | for (int pix = 0; pix < 256; pix++) { |
| 86 | 3072 | int shifted = pix - lower_bound; | |
| 87 | 3072 | lut.at(pix) = static_cast<uint8_t>(255 * shifted / delta); | |
| 88 | } | ||
| 89 | |||
| 90 | 12 | #pragma omp parallel for default(none) shared(local_data, local_count, lut) | |
| 91 | for (size_t i = 0; i < local_count; i++) { | ||
| 92 | local_data[i] = lut.at(local_data[i]); | ||
| 93 | } | ||
| 94 | } | ||
| 95 | |||
| 96 |
1/2✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
|
20 | GetOutput().resize(sz); |
| 97 |
1/2✓ Branch 1 taken 20 times.
✗ Branch 2 not taken.
|
20 | MPI_Allgatherv(local_data.data(), static_cast<int>(local_count), MPI_UINT8_T, GetOutput().data(), send_counts.data(), |
| 98 | locs.data(), MPI_UINT8_T, MPI_COMM_WORLD); | ||
| 99 | |||
| 100 | 20 | return true; | |
| 101 | } | ||
| 102 | |||
| 103 | 20 | bool GutyanskyAImgContrastIncrALL::PostProcessingImpl() { | |
| 104 | 20 | return true; | |
| 105 | } | ||
| 106 | |||
| 107 | } // namespace gutyansky_a_img_contrast_incr | ||
| 108 |