jiri/LA_library
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

NRVec: constructor with GPU location option, otimes2vec

Browse Source
This commit is contained in:
Jiri Pittner
2024-05-17 16:17:43 +02:00
parent 0b91e88dca
commit ea2b494abb
2 changed files with 88 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

33
vec.h
View File

@@ -130,7 +130,8 @@ public:
};
//! inlined constructor creating vector of given size filled with prescribed value
inline NRVec(const T &a, const int n);
//inline NRVec(const T &a, const int n);
inline NRVec(const T &a, const int n, const GPUID loc = undefined);
//! inlined constructor creating vector froman array
template<int SIZE> inline NRVec(const T (&a)[SIZE]);
@@ -356,9 +357,13 @@ public:
//! compute the outer product of two vectors
const NRMat<T> otimes(const NRVec<T> &rhs, const bool conjugate = false, const T &scale = 1) const;
//! opeartor for outer product computation
inline const NRMat<T> operator|(const NRVec<T> &rhs) const { return otimes(rhs,true); };
//! compute the outer product of two vectors, result interpreted as a vector
const NRVec otimes2vec(const NRVec<T> &rhs, const bool conjugate = false, const T &scale = 1) const;
//! compute the sum of the vector elements
inline const T sum() const {
T sum(v[0]);
@@ -678,6 +683,8 @@ std::istream & operator>>(std::istream &s, NRVec<T> &x) {
* @param[in] a value to be assigned to all vector elements
* @param[in] n required vector size
******************************************************************************/
/* replaced by the one with optional GPUID
template <typename T>
inline NRVec<T>::NRVec(const T& a, const int n): nn(n), count(new int) {
*count = 1;
@@ -699,6 +706,30 @@ inline NRVec<T>::NRVec(const T& a, const int n): nn(n), count(new int) {
}
#endif
}
*/
template <typename T>
inline NRVec<T>::NRVec(const T& a, const int n, const GPUID loc): nn(n), count(new int) {
*count = 1;
#ifdef CUDALA
location = (loc==undefined?DEFAULT_LOC:loc);
if(location == cpu){
#endif
v = new T[n];
if(!LA_traits<T>::is_plaindata() || a != (T)0){
for(register int i=0; i<n; i++) v[i] = a;
}else{
memset(v, 0, nn*sizeof(T));
}
#ifdef CUDALA
}else{
if(sizeof(T)%sizeof(float) != 0)laerror("memory alignment error");
v = (T*) gpualloc(n*sizeof(T));
if(!LA_traits<T>::is_plaindata()) laerror("only implemented for plain data");
smart_gpu_set(n, a, v);
}
#endif
}
/***************************************************************************//**
* inline constructor creating vector from an array