Bausager/Flux-openbuild
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Sync public subset from Flux

Browse Source
This commit is contained in:
Gitea CI
2025-10-08 16:08:04 +00:00
parent 35023cb7e1
commit 1c93024ff2
10 changed files with 260 additions and 133 deletions
Download Patch File Download Diff File Expand all files Collapse all files

160
include/numerics/matadd.h
View File

@@ -7,6 +7,39 @@
namespace numerics{
template <typename T>
void inplace_matadd_mat(utils::Matrix<T>& A, const utils::Matrix<T>& B) {
const uint64_t rows = A.rows();
const uint64_t cols = A.cols();
if (rows != B.rows() || cols != B.cols()) {
throw std::runtime_error("inplace_matadd: dimension mismatch");
}
for (uint64_t i = 0; i < cols; ++i) {
for (uint64_t j = 0; j < rows; ++j) {
A(j, i) += B(j, i);
}
}
}
template <typename T>
utils::Matrix<T> matadd_mat(const utils::Matrix<T>& A, const utils::Matrix<T>& B) {
const uint64_t rows = A.rows();
const uint64_t cols = A.cols();
utils::Matrix<T> C = A;
if (rows != B.rows() || cols != B.cols()) {
throw std::runtime_error("inplace_matadd: dimension mismatch");
}
inplace_matadd_mat(C, B);
return C;
}
template <typename T>
void inplace_matadd_colvec(utils::Matrix<T>& A, const utils::Vector<T>& x) {
@@ -94,133 +127,6 @@ namespace numerics{
}
}
/*
// -------------- Collapse(2) OpenMP ----------------
template <typename T>
utils::Vector<T> matvec_omp(const utils::Matrix<T>& A, const utils::Vector<T>& x) {
if (A.cols() != x.size()) {
throw std::runtime_error("matvec: dimension mismatch");
}
const uint64_t m = A.rows();
const uint64_t n = A.cols();
utils::Vector<T> y(m, T{0}); // <-- y has length m (rows)
const T* xptr = x.data();
const T* Aptr = A.data(); // row-major: A(i,j) == Aptr[i*n + j]
// Each row i is an independent dot product: y[i] = dot(A[i,*], x)
#pragma omp parallel for schedule(static)
for (uint64_t i = 0; i < m; ++i) {
const T* row = Aptr + i * n; // contiguous row i
T acc = T{0};
#pragma omp simd reduction(+:acc)
for (uint64_t j = 0; j < n; ++j) {
acc += row[j] * xptr[j];
}
y[i] = acc;
}
return y;
}
// -------------- Auto OpenMP ----------------
template <typename T>
utils::Vector<T> matvec_auto(const utils::Matrix<T>& A,
const utils::Vector<T>& x) {
uint64_t work = A.rows() * A.cols();
bool can_parallel = omp_config::omp_parallel_allowed();
#ifdef _OPENMP
int threads = omp_get_max_threads();
#else
int threads = 1;
#endif
if (can_parallel || work > static_cast<uint64_t>(threads) * 4ull) {
return matvec_omp(A,x);
}
else{
// Safe fallback
return matvec(A,x);
}
}
// =================================================
// y = x * A (Vector–Matrix product)
// =================================================
template <typename T>
utils::Vector<T> vecmat(const utils::Vector<T>& x, const utils::Matrix<T>& A) {
if (x.size() != A.rows()) {
throw std::runtime_error("vecmat: dimension mismatch");
}
const uint64_t m = A.rows();
const uint64_t n = A.cols();
utils::Vector<T> y(n, T{0});
for (uint64_t j = 0; j < n; ++j) {
for (uint64_t i = 0; i < m; ++i) {
y[j] += x[i] * A(i, j);
}
}
return y;
}
// -------------- Collapse(2) OpenMP ----------------
template <typename T>
utils::Vector<T> vecmat_omp(const utils::Vector<T>& x, const utils::Matrix<T>& A) {
if (x.size() != A.rows()) {
throw std::runtime_error("vecmat: dimension mismatch");
}
const uint64_t m = A.rows();
const uint64_t n = A.cols();
utils::Vector<T> y(n, T{0});
#pragma omp parallel for schedule(static)
for (uint64_t j = 0; j < n; ++j) {
T acc = T{0};
for (uint64_t i = 0; i < m; ++i) {
acc += x[i] * A(i, j);
}
y[j] = acc;
}
return y;
}
// -------------- Auto OpenMP ----------------
template <typename T>
utils::Vector<T> vecmat_auto(const utils::Vector<T>& x,
const utils::Matrix<T>& A) {
uint64_t work = A.rows() * A.cols();
bool can_parallel = omp_config::omp_parallel_allowed();
#ifdef _OPENMP
int threads = omp_get_max_threads();
#else
int threads = 1;
#endif
if (can_parallel || work > static_cast<uint64_t>(threads) * 4ull) {
return vecmat_omp(x,A);
}
else{
// Safe fallback
return vecmat(x,A);
}
}
*/
} // namespace numerics
#endif // _matadd_n_