Add nstream to OpenACC, but it looks like NVHPC is ignoring the += in the kernel so results are wrong

This commit is contained in:
Tom Deakin 2021-02-03 16:58:47 +00:00
parent c53b635a3c
commit f90d911551
2 changed files with 17 additions and 0 deletions

View File

@ -120,6 +120,22 @@ void ACCStream<T>::triad()
} }
} }
template <class T>
void ACCStream<T>::nstream()
{
const T scalar = startScalar;
int array_size = this->array_size;
T * restrict a = this->a;
T * restrict b = this->b;
T * restrict c = this->c;
#pragma acc parallel loop present(a[0:array_size], b[0:array_size], c[0:array_size]) wait
for (int i = 0; i < array_size; i++)
{
a[i] += b[i] + scalar * c[i];
}
}
template <class T> template <class T>
T ACCStream<T>::dot() T ACCStream<T>::dot()
{ {

View File

@ -35,6 +35,7 @@ class ACCStream : public Stream<T>
virtual void add() override; virtual void add() override;
virtual void mul() override; virtual void mul() override;
virtual void triad() override; virtual void triad() override;
virtual void nstream() override;
virtual T dot() override; virtual T dot() override;
virtual void init_arrays(T initA, T initB, T initC) override; virtual void init_arrays(T initA, T initB, T initC) override;