Issue #
Only a part of an array was transferred to the GPU and it does not cover the used range.
Relevance #
Minimising data transfers is one of the main optimization points when offloading computations to the GPU. An opportunity for such optimization occurs whenever only part of an array is required in a computation. In such cases, only a part of the array may be transferred to or from the GPU. However, the developer must ensure that the copied array range includes those array positions accessed in the code. Otherwise, the result is undefined behavior, most likely causing invalid memory accesses and crashes.
Actions #
Update the copied array range to match the actual array usage in the code.
Code example #
The following code performs the sum of two arrays:
void foo() {
int A[100], B[100], sum[100];
#pragma omp target map(to: A[0:50], B[0:50]) map(from: sum[0:50])
#pragma omp parallel for
for (int i = 0; i < 100; i++) {
sum[i] = A[i] + B[i];
}
}
However, only half of the array elements have been copied to the GPU. This must be fixed by matching the transferred range with array range actually used in the loop:
Related resources #
- PWD005 examples at GitHub
- PWR015: Avoid copying unnecessary array elements to the GPU
- OpenMP 4.5 Complete Specifications, November 2015 [last checked May 2019]
