phynexis-deps/armadillo-code
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

optionally use openmp

Browse Source
This commit is contained in:
conrad
2022-09-15 14:39:33 +10:00
parent 67d500d878
commit 52076bef60
1 changed files with 81 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files
include/armadillo_bits/glue_powext_meat.hpp
+81 -20
View File
@@ -229,8 +229,6 @@ glue_powext_cx::apply(Mat<typename T1::elem_type>& out, const mtGlue<typename T1
arma_debug_assert_same_size(A, B, "element-wise pow()");
// TODO: investigate use of openmp
if(UA.is_alias(out) && (UA.has_subview))
{
Mat<eT> tmp;
@@ -264,9 +262,26 @@ glue_powext_cx::apply(Mat< std::complex<T> >& out, const Mat< std::complex<T> >&
const eT* A_mem = A.memptr();
const T* B_mem = B.memptr();
for(uword i=0; i<N; ++i)
if( arma_config::openmp && mp_gate<eT>::eval(N) )
{
out_mem[i] = std::pow(A_mem[i], B_mem[i]);
#if defined(ARMA_USE_OPENMP)
{
const int n_threads = mp_thread_limit::get();
#pragma omp parallel for schedule(static) num_threads(n_threads)
for(uword i=0; i<N; ++i)
{
out_mem[i] = std::pow(A_mem[i], B_mem[i]);
}
}
#endif
}
else
{
for(uword i=0; i<N; ++i)
{
out_mem[i] = std::pow(A_mem[i], B_mem[i]);
}
}
}
@@ -298,36 +313,82 @@ glue_powext_cx::apply
X.check_size(B);
// TODO: investigate use of openmp
const T* B_mem = B.memptr();
if(mode == 0) // each column
{
for(uword i=0; i < A_n_cols; ++i)
if( arma_config::openmp && mp_gate<eT>::eval(A.n_elem) )
{
const eT* A_mem = A.colptr(i);
eT* out_mem = out.colptr(i);
for(uword row=0; row < A_n_rows; ++row)
#if defined(ARMA_USE_OPENMP)
{
out_mem[row] = std::pow(A_mem[row], B_mem[row]);
const int n_threads = int( (std::min)(uword(mp_thread_limit::get()), A_n_cols) );
#pragma omp parallel for schedule(static) num_threads(n_threads)
for(uword i=0; i < A_n_cols; ++i)
{
const eT* A_mem = A.colptr(i);
eT* out_mem = out.colptr(i);
for(uword row=0; row < A_n_rows; ++row)
{
out_mem[row] = std::pow(A_mem[row], B_mem[row]);
}
}
}
#endif
}
else
{
for(uword i=0; i < A_n_cols; ++i)
{
const eT* A_mem = A.colptr(i);
eT* out_mem = out.colptr(i);
for(uword row=0; row < A_n_rows; ++row)
{
out_mem[row] = std::pow(A_mem[row], B_mem[row]);
}
}
}
}
if(mode == 1) // each row
{
for(uword i=0; i < A_n_cols; ++i)
if( arma_config::openmp && mp_gate<eT>::eval(A.n_elem) )
{
const eT* A_mem = A.colptr(i);
eT* out_mem = out.colptr(i);
const eT B_val = B_mem[i];
for(uword row=0; row < A_n_rows; ++row)
#if defined(ARMA_USE_OPENMP)
{
out_mem[row] = std::pow(A_mem[row], B_val);
const int n_threads = int( (std::min)(uword(mp_thread_limit::get()), A_n_cols) );
#pragma omp parallel for schedule(static) num_threads(n_threads)
for(uword i=0; i < A_n_cols; ++i)
{
const eT* A_mem = A.colptr(i);
eT* out_mem = out.colptr(i);
const eT B_val = B_mem[i];
for(uword row=0; row < A_n_rows; ++row)
{
out_mem[row] = std::pow(A_mem[row], B_val);
}
}
}
#endif
}
else
{
for(uword i=0; i < A_n_cols; ++i)
{
const eT* A_mem = A.colptr(i);
eT* out_mem = out.colptr(i);
const eT B_val = B_mem[i];
for(uword row=0; row < A_n_rows; ++row)
{
out_mem[row] = std::pow(A_mem[row], B_val);
}
}
}
}