How to store the data in std::map when looping over the FArrayBoxes in MultiFabs?

[mafiami]

Hello, I'm trying to output part of the flow field data in specific format, so I'm trying to store the data from MultiFab to std::map data, and I use MFIter to do so. But it turns out only part of the data that I need is stored after looping over the FArrayBoxes in the MultiFab. Here's the code.

    for (amrex::MFIter mfi(vlmdata, amrex::TilingIfNotGPU());
         mfi.isValid(); ++mfi) {
      const auto& bx = mfi.tilebox();
      const int state_idx = m_dmConvert[mfi.index()];
      auto const& plane = vlmdata.array(mfi);
      auto const& state = a_state[lev]->const_array(state_idx, 0);
      auto* idx_d_p = m_fieldIndices_d.dataPtr();

      amrex::ParallelFor(
        bx, m_fieldNames.size(),
        [=] AMREX_GPU_DEVICE(int i, int j, int k, int n) noexcept {
          int stIdx = idx_d_p[n];
          if (j == p0) {
            if (k <= m_k0hi[2] && k >= m_k0low[2]) {
              facedata.insert(std::make_pair((i - m_k0low[0]) + leni * (k - m_k0low[2]), state(i,j,k,stIdx)));
            }
          }
        });
    }

As the whole domain is divided into a few parts, the result shows that only 1 part is considered during the loop, far from what I expected. How can I get the data from all the threads? Looking forward to your reply.
Sincerely,
Jessie

[WeiqunZhang]

Maybe you can copy the data you need to a single process first. For example,

BoxArray ba(Box(IntVect(m_k0low[0],p0,m_k0low[2]),IntVect(m_k0hi[0],p0,m_k0hi[2]),v_lmdata.ixType());
DistributionMapping dm({0});
MultiFab single_box_mf(ba,dm,ncomp, 0);
single_box_mf.ParallelCopy(vlmdata, ....)

[mafiami]

Thanks for your reply. It works!