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

added tests for plain data in constructors of vec mat smat

Browse Source
This commit is contained in:
Jiri Pittner
2021-10-28 20:17:32 +02:00
parent b50f9b36b1
commit d96531f340
5 changed files with 52 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
mat.h
View File

@@ -478,7 +478,7 @@ NRMat<T>::NRMat(const T &a, const int n, const int m, const GPUID loc) : nn(n),
#else
p = v = new T[nm];
#endif
if (a != (T)0){
if (!LA_traits<T>::is_plaindata() || a != (T)0){
for (register int i=0; i<nm; i++) *p++ = a;
}else{
memset(p, 0, nm*sizeof(T));
@@ -488,6 +488,7 @@ NRMat<T>::NRMat(const T &a, const int n, const int m, const GPUID loc) : nn(n),
if(sizeof(T)%sizeof(float) != 0)laerror("memory alignment error");
v = (T*)gpualloc(nm*sizeof(T));
if(!LA_traits<T>::is_plaindata()) laerror("only implemented for plain data");
smart_gpu_set(nm, a, v);
}
#endif
@@ -522,7 +523,7 @@ inline NRMat<T>::NRMat(const T (&a)[R][C]) : count(new int) {
#endif
#ifdef CUDALA
}else{
if!LA_traits<T>::is_plaindata()) laerror("only implemented for plain data");
if(!LA_traits<T>::is_plaindata()) laerror("only implemented for plain data");
v = (T*) gpualloc(nn*mm*sizeof(T));
cublasSetVector(nm, sizeof(T), a, 1, v, 1);
}
@@ -554,13 +555,14 @@ NRMat<T>::NRMat(const T &a, const int n, const int m) : nn(n), mm(m), count(new
#else
p = v = new T[nm];
#endif
if (a != (T)0)
if (!LA_traits<T>::is_plaindata() || a != (T)0)
for (register int i=0; i<nm; i++) *p++ = a;
else
memset(p, 0, nm*sizeof(T));
#ifdef CUDALA
}else{
v = (T*)gpualloc(nm*sizeof(T));
if(!LA_traits<T>::is_plaindata()) laerror("only implemented for plain data");
smart_gpu_set(nm, a, v);
}
#endif
@@ -588,14 +590,17 @@ NRMat<T>::NRMat(const T *a, const int n, const int m) : nn(n), mm(m), count(new
v = new T*[n];
v[0] = new T[nm];
for (register int i=1; i<n; i++) v[i] = v[i-1] + m;
memcpy(v[0], a, nm*sizeof(T));
if(LA_traits<T>::is_plaindata()) memcpy(v[0], a, nm*sizeof(T));
else for(int i=0; i<nm; ++i) v[0][i] = a[i];
#else
v = new T[nm];
memcpy(v, a, nm*sizeof(T));
if(LA_traits<T>::is_plaindata()) memcpy(v, a, nm*sizeof(T));
else for(int i=0; i<nm; ++i) v[i] = a[i];
#endif
#ifdef CUDALA
}else{
v = (T*) gpualloc(nm*sizeof(T));
if(!LA_traits<T>::is_plaindata()) laerror("only implemented for plain data");
cublasSetVector(nm, sizeof(T), a, 1, v, 1);
}
#endif