Bausager/Flux
1
0
Fork 0
Code Issues 36 Pull Requests Actions Packages Projects 1 Releases Wiki Activity

Finishing up and starting lu decomp

Browse Source
This commit is contained in:
Bausager
2025-09-13 21:44:20 +02:00
parent 320436ce98
commit 88087ea6a6
24 changed files with 1502 additions and 699 deletions
Download Patch File Download Diff File Expand all files Collapse all files
include/numerics/matvec.h
+97 -3
View File
@@ -3,11 +3,13 @@
#include "./utils/matrix.h"
#include "./core/omp_config.h"
namespace numerics{
// y = A * x, where A is (m×n) and x is length n and y is length m
// =================================================
// y = A * x (Matrix–Vector product)
// =================================================
template <typename T>
utils::Vector<T> matvec(const utils::Matrix<T>& A, const utils::Vector<T>& x) {
if (A.cols() != x.size()) {
@@ -18,6 +20,27 @@ namespace numerics{
const uint64_t n = A.cols();
utils::Vector<T> y(m, T{0});
for (uint64_t i = 0; i < m; ++i) {
for (uint64_t j = 0; j < n; ++j) {
y[i] += A(i, j) * x[j];
}
}
return y;
}
// -------------- 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});
#pragma omp parallel for schedule(static)
for (uint64_t i = 0; i < m; ++i) {
T acc = T{0};
for (uint64_t j = 0; j < n; ++j) {
@@ -28,7 +51,34 @@ namespace numerics{
return y;
}
// y = x * A, where x is length m and A is (m×n) -> y is length n
// -------------- 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()) {
@@ -38,6 +88,26 @@ namespace numerics{
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) {
@@ -48,6 +118,30 @@ namespace numerics{
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