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

*** empty log message ***

Browse Source
This commit is contained in:
jiri
2010-06-25 15:28:19 +00:00
parent eb0aaf9adf
commit 074c943862
13 changed files with 1938 additions and 464 deletions
Download Patch File Download Diff File Expand all files Collapse all files

46
vec.cc
View File

@@ -84,6 +84,16 @@ NRVec<T>::NRVec(const NRMat<T> &rhs)
template <typename T>
void NRVec<T>::put(int fd, bool dim, bool transp) const
{
#ifdef CUDALA
if(location!=cpu)
{
NRVec<T> tmp= *this;
tmp.moveto(cpu);
tmp.put(fd,dim,transp);
return;
}
#endif
errno=0;
int pad=1; //align at least 8-byte
if(dim)
@@ -94,9 +104,21 @@ if(sizeof(int) != write(fd,&pad,sizeof(int))) laerror("cannot write");
LA_traits<T>::multiput(nn,fd,v,dim);
}
template <typename T>
void NRVec<T>::get(int fd, bool dim, bool transp)
{
#ifdef CUDALA
if(location!=cpu)
{
NRVec<T> tmp;
tmp.moveto(cpu);
tmp.get(fd,dim,transp);
tmp.moveto(location);
*this = tmp;
return;
}
#endif
int nn0[2]; //align at least 8-byte
errno=0;
if(dim)
@@ -319,9 +341,16 @@ void NRVec<double>::gemv(const double beta, const NRMat<double> &A,
if ((trans == 'n'?A.ncols():A.nrows()) != x.size())
laerror("incompatible sizes in gemv A*x");
#endif
SAME_LOC3(*this,x,A);
copyonwrite();
cblas_dgemv(CblasRowMajor, (trans=='n' ? CblasNoTrans:CblasTrans),
A.nrows(), A.ncols(), alpha, A, A.ncols(), x.v, 1, beta, v, 1);
#ifdef CUDALA
if(location==cpu)
#endif
cblas_dgemv(CblasRowMajor, (trans=='n' ? CblasNoTrans:CblasTrans), A.nrows(), A.ncols(), alpha, A, A.ncols(), x.v, 1, beta, v, 1);
#ifdef CUDALA
else
cublasDgemv((trans=='n' ?'T':'N'),A.ncols(), A.nrows(),alpha, A, A.ncols(), x.v, 1, beta, v, 1);
#endif
}
@@ -362,8 +391,16 @@ void NRVec<double>::gemv(const double beta, const NRSMat<double> &A,
#ifdef DEBUG
if (A.ncols()!=x.size()) laerror("incompatible dimension in gemv A*x");
#endif
SAME_LOC3(*this,A,x);
copyonwrite();
#ifdef CUDALA
if(location==cpu)
#endif
cblas_dspmv(CblasRowMajor, CblasLower, A.ncols(), alpha, A, x.v, 1, beta, v, 1);
#ifdef CUDALA
else
cublasDspmv('U',A.ncols(), alpha, A, x.v, 1, beta, v, 1);
#endif
}
template<>
@@ -420,6 +457,7 @@ NRVec<complex<double> >::otimes(const NRVec< complex<double> > &b, const bool co
template<typename T>
int NRVec<T>::sort(int direction, int from, int to, int *perm)
{
NOT_GPU(*this);
copyonwrite();
if(to == -1) to=nn-1;
if(direction) return memqsort<1,NRVec<T>,int,int>(*this,perm,from,to);
@@ -427,6 +465,10 @@ else return memqsort<0,NRVec<T>,int,int>(*this,perm,from,to);
}
//////////////////////////////////////////////////////////////////////////////
//// forced instantization in the corespoding object file