| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | #include "vector_scalar_product/mpi/include/ops_mpi.hpp" | ||
| 2 | |||
| 3 | #include <mpi.h> | ||
| 4 | |||
| 5 | #include <cstddef> | ||
| 6 | #include <numeric> | ||
| 7 | #include <vector> | ||
| 8 | |||
| 9 | #include "vector_scalar_product/common/include/common.hpp" | ||
| 10 | |||
| 11 | namespace vector_scalar_product { | ||
| 12 | namespace { | ||
| 13 | 6 | std::vector<int> BuildCounts(int total, int parts) { | |
| 14 | 6 | std::vector<int> counts(parts, 0); | |
| 15 | 6 | const int base = total / parts; | |
| 16 | 6 | int remainder = total % parts; | |
| 17 |
2/2✓ Branch 0 taken 12 times.
✓ Branch 1 taken 6 times.
|
18 | for (int i = 0; i < parts; ++i) { |
| 18 |
4/4✓ Branch 0 taken 8 times.
✓ Branch 1 taken 4 times.
✓ Branch 2 taken 4 times.
✓ Branch 3 taken 8 times.
|
20 | counts[i] = base + (remainder > 0 ? 1 : 0); |
| 19 |
2/2✓ Branch 0 taken 4 times.
✓ Branch 1 taken 8 times.
|
12 | if (remainder > 0) { |
| 20 | 4 | --remainder; | |
| 21 | } | ||
| 22 | } | ||
| 23 | 6 | return counts; | |
| 24 | } | ||
| 25 | |||
| 26 | 6 | std::vector<int> BuildDisplacements(const std::vector<int> &counts) { | |
| 27 | 6 | std::vector<int> displs(counts.size(), 0); | |
| 28 |
2/2✓ Branch 0 taken 6 times.
✓ Branch 1 taken 6 times.
|
12 | for (std::size_t i = 1; i < counts.size(); ++i) { |
| 29 | 6 | displs[i] = displs[i - 1] + counts[i - 1]; | |
| 30 | } | ||
| 31 | 6 | return displs; | |
| 32 | } | ||
| 33 | } // namespace | ||
| 34 | |||
| 35 |
1/2✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
|
6 | VectorScalarProductMpi::VectorScalarProductMpi(const InType &in) { |
| 36 | SetTypeOfTask(GetStaticTypeOfTask()); | ||
| 37 | GetInput() = in; | ||
| 38 | 6 | GetOutput() = 0.0; | |
| 39 | 6 | } | |
| 40 | |||
| 41 |
1/2✓ Branch 0 taken 6 times.
✗ Branch 1 not taken.
|
6 | bool VectorScalarProductMpi::ValidationImpl() { |
| 42 | const auto &lhs = GetInput().lhs; | ||
| 43 | const auto &rhs = GetInput().rhs; | ||
| 44 |
2/4✓ Branch 0 taken 6 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 6 times.
|
6 | return !lhs.empty() && lhs.size() == rhs.size(); |
| 45 | } | ||
| 46 | |||
| 47 | 6 | bool VectorScalarProductMpi::PreProcessingImpl() { | |
| 48 | 6 | MPI_Comm_rank(MPI_COMM_WORLD, &rank_); | |
| 49 | 6 | MPI_Comm_size(MPI_COMM_WORLD, &world_size_); | |
| 50 | |||
| 51 | const auto &lhs = GetInput().lhs; | ||
| 52 | const auto &rhs = GetInput().rhs; | ||
| 53 | 6 | int global_size = 0; | |
| 54 |
2/2✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
|
6 | if (rank_ == 0) { |
| 55 | 3 | global_size = static_cast<int>(lhs.size()); | |
| 56 | } | ||
| 57 | 6 | MPI_Bcast(&global_size, 1, MPI_INT, 0, MPI_COMM_WORLD); | |
| 58 | |||
| 59 | // Prepare scattering meta information | ||
| 60 | 6 | auto counts = BuildCounts(global_size, world_size_); | |
| 61 |
1/2✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
|
6 | auto displs = BuildDisplacements(counts); |
| 62 |
1/2✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
|
6 | const int local_count = counts[rank_]; |
| 63 | |||
| 64 |
1/2✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
|
6 | local_lhs_.assign(static_cast<std::size_t>(local_count), 0.0); |
| 65 |
1/4✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
✗ Branch 4 not taken.
|
6 | local_rhs_.assign(static_cast<std::size_t>(local_count), 0.0); |
| 66 | |||
| 67 |
2/2✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
|
6 | const double *lhs_ptr = rank_ == 0 ? lhs.data() : nullptr; |
| 68 |
2/2✓ Branch 0 taken 3 times.
✓ Branch 1 taken 3 times.
|
6 | const double *rhs_ptr = rank_ == 0 ? rhs.data() : nullptr; |
| 69 | |||
| 70 |
1/2✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
|
6 | MPI_Scatterv(lhs_ptr, counts.data(), displs.data(), MPI_DOUBLE, local_lhs_.data(), local_count, MPI_DOUBLE, 0, |
| 71 | MPI_COMM_WORLD); | ||
| 72 |
1/2✓ Branch 1 taken 6 times.
✗ Branch 2 not taken.
|
6 | MPI_Scatterv(rhs_ptr, counts.data(), displs.data(), MPI_DOUBLE, local_rhs_.data(), local_count, MPI_DOUBLE, 0, |
| 73 | MPI_COMM_WORLD); | ||
| 74 | |||
| 75 | 6 | local_sum_ = 0.0; | |
| 76 |
1/2✓ Branch 0 taken 6 times.
✗ Branch 1 not taken.
|
6 | result_ = 0.0; |
| 77 | 6 | return true; | |
| 78 | } | ||
| 79 | |||
| 80 | 6 | bool VectorScalarProductMpi::RunImpl() { | |
| 81 | 6 | local_sum_ = std::inner_product(local_lhs_.begin(), local_lhs_.end(), local_rhs_.begin(), 0.0); | |
| 82 | |||
| 83 | 6 | return MPI_Reduce(&local_sum_, &result_, 1, MPI_DOUBLE, MPI_SUM, 0, MPI_COMM_WORLD) == MPI_SUCCESS; | |
| 84 | } | ||
| 85 | |||
| 86 | 6 | bool VectorScalarProductMpi::PostProcessingImpl() { | |
| 87 | 6 | MPI_Bcast(&result_, 1, MPI_DOUBLE, 0, MPI_COMM_WORLD); | |
| 88 | 6 | GetOutput() = result_; | |
| 89 | 6 | return true; | |
| 90 | } | ||
| 91 | |||
| 92 | } // namespace vector_scalar_product | ||
| 93 |