I thought the code like below should compile with Clang:
int *p;
p = new int[1024];
for (int i=0;i<1024;i++)
p[i] = new int[1024];
#pragma omp target parallel for map(p[0:1024][0:1024]) reduction(+: sum)
… but I get a build-time error. Is this expected even in the recent trunk?