I have a mutli-GPU code which I am compiling using pgc++.
Below is a section of the code:
#pragma omp parallel for
for(int dev=0;dev<NGPUS;dev++)
{
acc_set_device_num(dev,acc_device_nvidia);
#pragma acc update device(ta_64[0:N],tb_64[0:N],tc_64[0:N])
** int s=dev*(N/NGPUS);**
#pragma acc data present(ta_64[0:N],tb_64[0:N],tc_64[0:N],wa[0:N],wb[0:N],wc[0:N],KA[0:4*N],KB[0:4*N],KC[0:4*N],this)
{.....................
upon compiling with -Minfo=accel, there are a few output lines which look like this:
…
288, Generating update device(_T25_2687->tb_64[:_T25_2687->N],_T25_2687->tc_64[:_T25_2687->N],_T25_2687->ta_64[:_T25_2687->N])
…
what do these _T25_2687… mean? I have no variables by that name in my code. The lines in bold denote the corresponding line number 287-288