diff --git a/coreneuron/CMakeLists.txt b/coreneuron/CMakeLists.txt
index 810bc57f8..f65e7e7da 100644
--- a/coreneuron/CMakeLists.txt
+++ b/coreneuron/CMakeLists.txt
@@ -107,6 +107,7 @@ if(CORENRN_ENABLE_GPU)
       ${CMAKE_CURRENT_SOURCE_DIR}/utils/randoms/nrnran123.cpp
       ${CMAKE_CURRENT_SOURCE_DIR}/io/nrn_setup.cpp
       ${CMAKE_CURRENT_SOURCE_DIR}/io/setup_fornetcon.cpp
+      ${CMAKE_CURRENT_SOURCE_DIR}/io/corenrn_data_return.cpp
       ${CMAKE_CURRENT_SOURCE_DIR}/io/global_vars.cpp)
 
   set_source_files_properties(${DIMPLIC_CODE_FILE} ${NMODL_INBUILT_MOD_OUTPUTS} PROPERTIES
diff --git a/coreneuron/apps/main1.cpp b/coreneuron/apps/main1.cpp
index 9075af336..9037b036c 100644
--- a/coreneuron/apps/main1.cpp
+++ b/coreneuron/apps/main1.cpp
@@ -62,6 +62,7 @@ THE POSSIBILITY OF SUCH DAMAGE.
 #include "coreneuron/network/multisend.hpp"
 #include "coreneuron/io/file_utils.hpp"
 #include "coreneuron/io/nrn2core_direct.h"
+#include "coreneuron/io/core2nrn_data_return.hpp"
 
 extern "C" {
 const char* corenrn_version() {
@@ -609,6 +610,8 @@ extern "C" int run_solve_core(int argc, char** argv) {
         (*nrn2core_all_weights_return_)(weights);
     }
 
+    core2nrn_data_return();
+
     {
         Instrumentor::phase p("checkpoint");
         write_checkpoint(nrn_threads, nrn_nthread, corenrn_param.checkpointpath.c_str());
diff --git a/coreneuron/io/core2nrn_data_return.cpp b/coreneuron/io/core2nrn_data_return.cpp
new file mode 100644
index 000000000..a380cf6cc
--- /dev/null
+++ b/coreneuron/io/core2nrn_data_return.cpp
@@ -0,0 +1,174 @@
+/*
+Copyright (c) 2020, Blue Brain Project
+All rights reserved.
+     
+Redistribution and use in source and binary forms, with or without modification,
+are permitted provided that the following conditions are met:
+1. Redistributions of source code must retain the above copyright notice,
+   this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright notice,
+   this list of conditions and the following disclaimer in the documentation
+   and/or other materials provided with the distribution.
+3. Neither the name of the copyright holder nor the names of its contributors
+   may be used to endorse or promote products derived from this software
+   without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
+THE POSSIBILITY OF SUCH DAMAGE.
+*/
+
+#include "coreneuron/coreneuron.hpp"
+#include "coreneuron/io/nrn2core_direct.h"
+#include "coreneuron/sim/multicore.hpp"
+#include "coreneuron/nrniv/nrniv_decl.h"
+#include "coreneuron/io/core2nrn_data_return.hpp"
+
+/** @brief, Information from NEURON to help with copying data to NEURON.
+ *  Info for copying voltage, i_membrane_, and mechanism data.
+ *  See implementaton in
+ *  nrn/src/nrniv/nrnbbcore_write.cpp:nrnthreads_type_return.
+ *  Return is size of either the returned data pointer or the number
+ *  of pointers in mdata. tid is the thread index.
+ */
+size_t (*nrn2core_type_return_)(int type, int tid,
+  double*& data, double**& mdata);
+
+namespace coreneuron {
+
+/** @brief permuted array copied to unpermuted array
+ *  If permute is NULL then just a copy
+ */
+static void inverse_permute_copy(size_t n, double* permuted_src, double* dest,
+                                 int* permute)
+{
+  if (permute) {
+    for (size_t i = 0; i < n; ++i) {
+      dest[i] = permuted_src[permute[i]];
+    }
+  }else{
+    std::copy(permuted_src, permuted_src + n, dest);
+  }
+}
+
+/** @brief SoA permuted mechanism data copied to unpermuted AoS data.
+ *  dest is an array of n pointers to the beginning of each sz length array.
+ *  src is a contiguous array of sz segments of size stride. The stride
+ *  may be slightly greater than n for purposes of alignment.
+ *  Each of the sz segments of src are permuted.
+ */
+static void soa2aos_inverse_permute_copy(size_t n, int sz, int stride,
+  double* src, double** dest, int* permute)
+{
+  // src is soa and permuted. dest is n pointers to sz doubles (aos).
+  for (size_t instance = 0; instance < n; ++instance) {
+    double* d = dest[instance];
+    double* s = src + permute[instance];
+    for (int i = 0; i < sz; ++i) {
+      d[i] = s[i*stride];
+    }
+  }
+}
+
+/** @brief SoA unpermuted mechanism data copied to unpermuted AoS data.
+ *  dest is an array of n pointers to the beginning of each sz length array.
+ *  src is a contiguous array of sz segments of size stride. The stride
+ *  may be slightly greater than n for purposes of alignment.
+ *  Each of the sz segments of src have the same order as the n pointers
+ *  of dest.
+ */
+static void soa2aos_unpermuted_copy(size_t n, int sz, int stride,
+  double* src, double** dest)
+{
+  // src is soa and permuted. dest is n pointers to sz doubles (aos).
+  for (size_t instance = 0; instance < n; ++instance) {
+    double* d = dest[instance];
+    double* s = src + instance;
+    for (int i = 0; i < sz; ++i) {
+      d[i] = s[i*stride];
+    }
+  }
+}
+
+/** @brief AoS mechanism data copied to AoS data.
+ *  dest is an array of n pointers to the beginning of each sz length array.
+ *  src is a contiguous array of n segments of size sz.
+ */
+static void aos2aos_copy(size_t n, int sz, double* src, double** dest) {
+  for (size_t instance = 0; instance < n; ++instance) {
+    double* d = dest[instance];
+    double* s = src + (instance*sz);
+    std::copy(s, s + sz, d);
+  }
+}
+      
+/** @brief copy data back to NEURON.
+ *  Copies t, voltage, i_membrane_ if it used, and mechanism param data.
+ */
+void core2nrn_data_return() {
+  if (!nrn2core_type_return_) {
+    return;
+  }
+  for (int tid = 0; tid < nrn_nthread; ++tid) {
+    size_t n = 0;
+    double* data = nullptr;
+    double** mdata = nullptr;
+    NrnThread& nt = nrn_threads[tid];
+
+    n = (*nrn2core_type_return_)(0, tid, data, mdata); // 0 means time
+    if (n) { // not the empty thread
+      data[0] = nt._t;
+    }
+
+    if (nt.end) { // transfer voltage and possibly i_membrane_
+      n = (*nrn2core_type_return_)(voltage, tid, data, mdata);
+      assert(n == size_t(nt.end) && data);
+      inverse_permute_copy(n, nt._actual_v, data, nt._permute);
+
+      if (nt.nrn_fast_imem) {
+        n = (*nrn2core_type_return_)(i_membrane_, tid, data, mdata);
+        assert(n == size_t(nt.end) && data);
+        inverse_permute_copy(n, nt.nrn_fast_imem->nrn_sav_rhs, data, nt._permute);
+      }
+    }
+
+    for (NrnThreadMembList* tml = nt.tml; tml; tml = tml->next) {
+      int mtype = tml->index;
+      Memb_list* ml = tml->ml;
+      n = (*nrn2core_type_return_)(mtype, tid, data, mdata);
+      assert(n == size_t(ml->nodecount) && mdata);
+      if (n == 0) {
+        continue;
+      }
+      // NEURON is AoS, CoreNEURON may be SoA and may be permuted.
+      // On the NEURON side, the data is actually contiguous because of
+      // cache_efficient, but that may not be the case for ARTIFICIAL_CELL.
+      // For initial implementation simplicity, use the mdata info which gives
+      // a double* for each param_size mech instance.
+      int* permute = ml->_permute;
+      double* cndat = ml->data;
+      int layout = corenrn.get_mech_data_layout()[mtype];
+      int sz = corenrn.get_prop_param_size()[mtype];
+      if (layout == 0) { /* SoA */
+        int stride = ml->_nodecount_padded;
+        if (permute) {
+          soa2aos_inverse_permute_copy(n, sz, stride, cndat, mdata, permute);
+        } else {
+          soa2aos_unpermuted_copy(n, sz, stride, cndat, mdata);
+        }
+      } else { /* AoS */
+        aos2aos_copy(n, sz, cndat, mdata);
+      }
+    }
+  }
+}
+
+}  // namespace coreneuron
diff --git a/coreneuron/io/core2nrn_data_return.hpp b/coreneuron/io/core2nrn_data_return.hpp
new file mode 100644
index 000000000..8f87d23f8
--- /dev/null
+++ b/coreneuron/io/core2nrn_data_return.hpp
@@ -0,0 +1,40 @@
+/*
+Copyright (c) 2016, Blue Brain Project
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without modification,
+are permitted provided that the following conditions are met:
+1. Redistributions of source code must retain the above copyright notice,
+   this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright notice,
+   this list of conditions and the following disclaimer in the documentation
+   and/or other materials provided with the distribution.
+3. Neither the name of the copyright holder nor the names of its contributors
+   may be used to endorse or promote products derived from this software
+   without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
+THE POSSIBILITY OF SUCH DAMAGE.
+*/
+
+#ifndef _H_CORENRNDATARETURN_
+#define _H_CORENRNDATARETURN_
+
+namespace coreneuron {
+
+/** @brief Copies back to NEURON the voltage, i_membrane_, and mechanism data.
+ */
+extern void core2nrn_data_return();
+
+}  // namespace coreneuron
+#endif // _H_CORENRNDATARETURN_
+
diff --git a/coreneuron/io/nrn2core_direct.h b/coreneuron/io/nrn2core_direct.h
index 0df66608f..3c4990a05 100644
--- a/coreneuron/io/nrn2core_direct.h
+++ b/coreneuron/io/nrn2core_direct.h
@@ -113,6 +113,9 @@ extern int (*nrn2core_all_spike_vectors_return_)(std::vector<double>& spikevec,
 /* send all weights to NEURON */
 extern void (*nrn2core_all_weights_return_)(std::vector<double*>& weights);
 
+/* get data array pointer from NEURON to copy into. */
+extern size_t (*nrn2core_type_return_)(int type, int tid, double*& data,
+                                       double**& mdata);
 } // extern "C"
 
 #endif /* nrn2core_direct_h */
diff --git a/coreneuron/mechanism/mech/enginemech.cpp b/coreneuron/mechanism/mech/enginemech.cpp
index 346f69cd7..b500116e8 100644
--- a/coreneuron/mechanism/mech/enginemech.cpp
+++ b/coreneuron/mechanism/mech/enginemech.cpp
@@ -95,8 +95,11 @@ int corenrn_embedded_run(int nthread,
     mk_mech_init(argc, argv);
 
     // initialize extra arguments built into special-core
-    coreneuron::modl_reg();
-
+    static bool modl_reg_called = false;
+    if (!modl_reg_called) {
+        coreneuron::modl_reg();
+        modl_reg_called = true;
+    }
     // run simulation
     run_solve_core(argc, argv);