Merge branch 'master' into 1uc/optimize-diam-area

.github/workflows/docs.yml

@@ -78,11 +78,3 @@ jobs:
       - name: live debug session on failure (manual steps required, check `.github/docs.yml`)
         if: failure() && contains(github.event.pull_request.title, 'live-debug-docs')
         uses: mxschmitt/action-tmate@v3
-
-      - name: Deploy 🚀
-        uses: JamesIves/github-pages-deploy-action@v4
-        if: env.OK_TO_DEPLOY_DOCS == 'true' && github.ref == 'refs/heads/master'
-        with:
-          branch: gh-pages # The branch the action should deploy to.
-          folder: ${{runner.workspace}}/nrn/docs/_build # The folder the action should deploy.
-          single-commit: true #have a single commit on the deployment branch instead of maintaining the full history

docs/courses/mpi_parallelization.rst

@@ -43,6 +43,6 @@ Unfortunately MPI can't be a part of the binary installation because I don't kno
 Going Further 
 ----------
 
-The ring model from the `above ModelDB entry <https://modeldb.science/96444>`_ is a good next step. See also the documentation for the `ParallelContext <https://nrn.readthedocs.io/en/latest/hoc/modelspec/programmatic/network/parcon.html?highlight=parallelcontext>`_ class, especialy the subset of methods gathered under the `ParallelNetwork <https://nrn.readthedocs.io/en/latest/hoc/modelspec/programmatic/network/parcon.html?highlight=parallelcontext>`_ heading. A large portion of the `ParallelNetManager <https://nrn.readthedocs.io/en/latest/hoc/modelspec/programmatic/network/parnet.html?highlight=parallelnetmanager>`_ wrapper is better off done directly from the underlying ParallelContext though it can be mined for interesting pieces. A good place to find the most recent idioms is the NEURON implementation of the Vogels and Abbott model found in the `Brette et al. ModelDB entry <https://senselab.med.yale.edu/ModelDB/ShowModel?model=83319#tabs-1>`_. However, to run in parallel, the NetCon delay between cells needs to be set greater than zero.
+The ring model from the `above ModelDB entry <https://modeldb.science/96444>`_ is a good next step. See also the documentation for the `ParallelContext <https://nrn.readthedocs.io/en/latest/hoc/modelspec/programmatic/network/parcon.html?highlight=parallelcontext>`_ class, especialy the subset of methods gathered under the `ParallelNetwork <https://nrn.readthedocs.io/en/latest/hoc/modelspec/programmatic/network/parcon.html?highlight=parallelcontext>`_ heading. A large portion of the `ParallelNetManager <https://nrn.readthedocs.io/en/latest/hoc/modelspec/programmatic/network/parnet.html?highlight=parallelnetmanager>`_ wrapper is better off done directly from the underlying ParallelContext though it can be mined for interesting pieces. A good place to find the most recent idioms is the NEURON implementation of the Vogels and Abbott model found in the `Brette et al. ModelDB entry <https://modeldb.science/83319>`_. However, to run in parallel, the NetCon delay between cells needs to be set greater than zero.
 
 

docs/guide/how_to_get_started_with_neuron.rst

@@ -41,7 +41,7 @@ To learn how to use NMODL to add new mechanisms to NEURON:
     NEURON comes with a bunch of mod files that can serve as starting points for "programming by example." Under MSWin the default mechanisms (hh, pas, expsyn etc.) are in ``c:\nrn\src\nrnoc`` (on my Linux box this is ``/usr/local/src/nrn-x.x/src/nrnoc``). A large collection of mod files is in ``c:\nrn\examples\nrniv\nmodl`` (Linux ``/usr/local/src/nrn-x.x/share/examples/nrniv/nmodl``).
 
 4.
-    You may also find useful examples in `ModelDB <https://smodeldb.science>`_.
+    You may also find useful examples in `ModelDB <https://modeldb.science>`_.
 
 For courses about NEURON, see the :ref:`Course Exercises <exercises2018>` page and the :ref:`Training Videos <training_videos>` page.
 

packaging/python/test_wheels.sh

@@ -260,7 +260,7 @@ $python_exe -m pip install --upgrade pip
 # install numpy, pytest and neuron
 # we install setuptools because since python 3.12 it is no more installed
 # by default
-$python_exe -m pip install numpy pytest setuptools
+$python_exe -m pip install "numpy<2" pytest setuptools
 $python_exe -m pip install $python_wheel
 $python_exe -m pip show neuron || $python_exe -m pip show neuron-nightly
 

src/neuron/cache/mechanism_range.hpp

@@ -215,7 +215,7 @@ struct MechanismInstance: MechanismRange<NumFloatingPointFields, NumDatumFields>
     MechanismInstance(Prop* prop)
         : base_type{_nrn_mechanism_get_type(prop), mechanism::_get::_current_row(prop), 0} {
         if (!prop) {
-            // grrr...see cagkftab test where setdata is not called(?) and extcall_prop is null(?)
+            // see cagkftab test where setdata is not called(?) and extcall_prop is null(?)
             return;
         }
         indices_to_cache(_nrn_mechanism_get_type(prop), [this, prop](auto field) {

src/nmodl/deriv.cpp

@@ -161,7 +161,7 @@ void solv_diffeq(Item* qsol,
         // derivimplicit_thread
         Sprintf(buf,
                 "%s %s%s_thread(%d, _slist%d, _dlist%d, neuron::scopmath::row_view{_ml, _iml}, %s, "
-                "_ml, _iml, _ppvar, _thread, _nt);\n%s",
+                "_ml, _iml, _ppvar, _thread, _globals, _nt);\n%s",
                 deriv1_advance,
                 ssprefix,
                 method->name,

src/nmodl/noccout.cpp

@@ -536,6 +536,8 @@ void c_out_vectorize() {
     P("_ni = _ml_arg->_nodeindices;\n");
     P("_cntml = _ml_arg->_nodecount;\n");
     P("_thread = _ml_arg->_thread;\n");
+    P("double* _globals = nullptr;\n");
+    P("if (gind != 0 && _thread != nullptr) { _globals = _thread[_gth].get<double*>(); }\n");
     /*check_tables();*/
     P("for (_iml = 0; _iml < _cntml; ++_iml) {\n");
     P(" _ppvar = _ml_arg->_pdata[_iml];\n");
@@ -598,6 +600,8 @@ void c_out_vectorize() {
         P("_ni = _ml_arg->_nodeindices;\n");
         P("_cntml = _ml_arg->_nodecount;\n");
         P("_thread = _ml_arg->_thread;\n");
+        P("double* _globals = nullptr;\n");
+        P("if (gind != 0 && _thread != nullptr) { _globals = _thread[_gth].get<double*>(); }\n");
         P("for (_iml = 0; _iml < _cntml; ++_iml) {\n");
         P(" _ppvar = _ml_arg->_pdata[_iml];\n");
         ext_vdef();
@@ -664,6 +668,8 @@ void c_out_vectorize() {
         P("_ni = _ml_arg->_nodeindices;\n");
         P("_cntml = _ml_arg->_nodecount;\n");
         P("_thread = _ml_arg->_thread;\n");
+        P("double* _globals = nullptr;\n");
+        P("if (gind != 0 && _thread != nullptr) { _globals = _thread[_gth].get<double*>(); }\n");
         P("for (_iml = 0; _iml < _cntml; ++_iml) {\n");
         if (electrode_current) {
             P(" _nd = _ml_arg->_nodelist[_iml];\n");
@@ -700,6 +706,8 @@ void c_out_vectorize() {
         P("_ni = _ml_arg->_nodeindices;\n");
         P("size_t _cntml = _ml_arg->_nodecount;\n");
         P("_thread = _ml_arg->_thread;\n");
+        P("double* _globals = nullptr;\n");
+        P("if (gind != 0 && _thread != nullptr) { _globals = _thread[_gth].get<double*>(); }\n");
         P("for (size_t _iml = 0; _iml < _cntml; ++_iml) {\n");
         P(" _ppvar = _ml_arg->_pdata[_iml];\n");
         P(" _nd = _ml_arg->_nodelist[_iml];\n");

src/nmodl/nocpout.cpp

@@ -296,12 +296,12 @@ void parout() {
     if (vectorize) {
         Lappendstr(defs_list,
                    "\n\
-#define _threadargscomma_ _ml, _iml, _ppvar, _thread, _nt,\n\
-#define _threadargsprotocomma_ Memb_list* _ml, size_t _iml, Datum* _ppvar, Datum* _thread, NrnThread* _nt,\n\
-#define _internalthreadargsprotocomma_ _nrn_mechanism_cache_range* _ml, size_t _iml, Datum* _ppvar, Datum* _thread, NrnThread* _nt,\n\
-#define _threadargs_ _ml, _iml, _ppvar, _thread, _nt\n\
-#define _threadargsproto_ Memb_list* _ml, size_t _iml, Datum* _ppvar, Datum* _thread, NrnThread* _nt\n\
-#define _internalthreadargsproto_ _nrn_mechanism_cache_range* _ml, size_t _iml, Datum* _ppvar, Datum* _thread, NrnThread* _nt\n\
+#define _threadargscomma_ _ml, _iml, _ppvar, _thread, _globals, _nt,\n\
+#define _threadargsprotocomma_ Memb_list* _ml, size_t _iml, Datum* _ppvar, Datum* _thread, double* _globals, NrnThread* _nt,\n\
+#define _internalthreadargsprotocomma_ _nrn_mechanism_cache_range* _ml, size_t _iml, Datum* _ppvar, Datum* _thread, double* _globals, NrnThread* _nt,\n\
+#define _threadargs_ _ml, _iml, _ppvar, _thread, _globals, _nt\n\
+#define _threadargsproto_ Memb_list* _ml, size_t _iml, Datum* _ppvar, Datum* _thread, double* _globals, NrnThread* _nt\n\
+#define _internalthreadargsproto_ _nrn_mechanism_cache_range* _ml, size_t _iml, Datum* _ppvar, Datum* _thread, double* _globals, NrnThread* _nt\n\
 ");
     } else {
         Lappendstr(defs_list,
@@ -544,11 +544,6 @@ extern void nrn_promote(Prop*, int, int);\n\
         }
     }
 
-    emit_check_table_thread = 0;
-    if (vectorize && check_tables_threads(defs_list)) {
-        emit_check_table_thread = 1;
-    }
-
     /* per thread top LOCAL */
     /* except those that are marked assigned_to_ == 2 stay static double */
     if (vectorize && toplocal_) {
@@ -601,7 +596,7 @@ extern void nrn_promote(Prop*, int, int);\n\
     }
     /* per thread global data */
     gind = 0;
-    if (vectorize)
+    if (vectorize) {
         SYMLISTITER {
             s = SYM(q);
             if (s->nrntype & (NRNGLOBAL) && s->assigned_to_ == 1) {
@@ -612,8 +607,15 @@ extern void nrn_promote(Prop*, int, int);\n\
                 }
             }
         }
+    }
     /* double scalars declared internally */
     Lappendstr(defs_list, "/* declare global and static user variables */\n");
+    Sprintf(buf, "#define gind %d\n", gind);
+    Lappendstr(defs_list, buf);
+    if (!gind) {
+        Sprintf(buf, "#define _gth 0\n");
+        Lappendstr(defs_list, buf);
+    }
     if (gind) {
         Sprintf(buf,
                 "static int _thread1data_inuse = 0;\nstatic double _thread1data[%d];\n#define _gth "
@@ -644,7 +646,7 @@ extern void nrn_promote(Prop*, int, int);\n\
                 if (s->subtype & ARRAY) {
                     Sprintf(buf,
                             "#define %s%s (_thread1data + %d)\n\
-#define %s (_thread[_gth].get<double*>() + %d)\n",
+#define %s (_globals + %d)\n",
                             s->name,
                             suffix,
                             gind,
@@ -653,7 +655,7 @@ extern void nrn_promote(Prop*, int, int);\n\
                 } else {
                     Sprintf(buf,
                             "#define %s%s _thread1data[%d]\n\
-#define %s _thread[_gth].get<double*>()[%d]\n",
+#define %s _globals[%d]\n",
                             s->name,
                             suffix,
                             gind,
@@ -684,6 +686,11 @@ extern void nrn_promote(Prop*, int, int);\n\
         }
     }
 
+    emit_check_table_thread = 0;
+    if (vectorize && check_tables_threads(defs_list)) {
+        emit_check_table_thread = 1;
+    }
+
     Lappendstr(defs_list, "/* some parameters have upper and lower limits */\n");
     Lappendstr(defs_list, "static HocParmLimits _hoc_parm_limits[] = {\n");
     SYMLISTITER {
@@ -739,17 +746,13 @@ extern void nrn_promote(Prop*, int, int);\n\
     Lappendstr(defs_list, "{0, 0, 0}\n};\n");
     Lappendstr(defs_list, "static double _sav_indep;\n");
     if (ba_index_ > 0) {
-        Lappendstr(defs_list,
-                   "static void _ba1(Node*_nd, Datum* _ppd, Datum* _thread, NrnThread* _nt, "
-                   "Memb_list* _ml, size_t _iml, _nrn_model_sorted_token const&)");
-        for (i = 2; i <= ba_index_; ++i) {
+        for (int i = 1; i <= ba_index_; ++i) {
             Sprintf(buf,
-                    ", _ba%d(Node*_nd, Datum* _ppd, Datum* _thread, NrnThread* _nt, Memb_list* "
-                    "_ml, size_t _iml, _nrn_model_sorted_token const&)",
+                    "static void _ba%d(Node*_nd, Datum* _ppd, Datum* _thread, NrnThread* _nt, "
+                    "Memb_list* _ml, size_t _iml, _nrn_model_sorted_token const&);\n",
                     i);
             Lappendstr(defs_list, buf);
         }
-        Lappendstr(defs_list, ";\n");
     }
 
     /******** what normally goes into cabvars.h structures */
@@ -1579,6 +1582,8 @@ void ldifusreg() {
             "_nrn_model_sorted_token const& _sorted_token) {\n"
             "  _nrn_mechanism_cache_range _lmr{_sorted_token, *_nt, *_ml_arg, _ml_arg->_type()};\n"
             "  auto* const _ml = &_lmr;\n"
+            "  double* _globals = nullptr;\n"
+            "  if (gind != 0 && _thread != nullptr) { _globals = _thread[_gth].get<double*>(); }\n"
             "  *_pdvol = ",
             n,
             n);
@@ -1933,10 +1938,13 @@ void bablk(int ba, int type, Item* q1, Item* q2) {
     insertstr(q1, buf);
     q = q1->next;
     vectorize_substitute(insertstr(q, ""), "Datum* _ppvar;");
-    qv = insertstr(q,
-                   "_nrn_mechanism_cache_range _lmr{_sorted_token, *_nt, *_ml_arg, "
-                   "_ml_arg->_type()}; auto* const "
-                   "_ml = &_lmr;\n");
+    qv = insertstr(
+        q,
+        "_nrn_mechanism_cache_range _lmr{_sorted_token, *_nt, *_ml_arg, "
+        "_ml_arg->_type()}; auto* const "
+        "_ml = &_lmr;\n"
+        "double* _globals = nullptr;\n"
+        "if (gind != 0 && _thread != nullptr) { _globals = _thread[_gth].get<double*>(); }\n");
     qv = insertstr(q, "_ppvar = _ppd;\n");
     movelist(qb, q2, procfunc);
 
@@ -2773,18 +2781,21 @@ void out_nt_ml_frag(List* p) {
     vectorize_substitute(lappendstr(p, ""), "  Datum* _ppvar;\n");
     vectorize_substitute(lappendstr(p, ""), "  size_t _iml;");
     vectorize_substitute(lappendstr(p, ""), "  _nrn_mechanism_cache_range* _ml;");
-    Lappendstr(p,
-               "  Node* _nd{};\n"
-               "  double _v{};\n"
-               "  int _cntml;\n"
-               "  _nrn_mechanism_cache_range _lmr{_sorted_token, *_nt, *_ml_arg, _type};\n"
-               "  _ml = &_lmr;\n"
-               "  _cntml = _ml_arg->_nodecount;\n"
-               "  Datum *_thread{_ml_arg->_thread};\n"
-               "  for (_iml = 0; _iml < _cntml; ++_iml) {\n"
-               "    _ppvar = _ml_arg->_pdata[_iml];\n"
-               "    _nd = _ml_arg->_nodelist[_iml];\n"
-               "    v = NODEV(_nd);\n");
+    Lappendstr(
+        p,
+        "  Node* _nd{};\n"
+        "  double _v{};\n"
+        "  int _cntml;\n"
+        "  _nrn_mechanism_cache_range _lmr{_sorted_token, *_nt, *_ml_arg, _type};\n"
+        "  _ml = &_lmr;\n"
+        "  _cntml = _ml_arg->_nodecount;\n"
+        "  Datum *_thread{_ml_arg->_thread};\n"
+        "  double* _globals = nullptr;\n"
+        "  if (gind != 0 && _thread != nullptr) { _globals = _thread[_gth].get<double*>(); }\n"
+        "  for (_iml = 0; _iml < _cntml; ++_iml) {\n"
+        "    _ppvar = _ml_arg->_pdata[_iml];\n"
+        "    _nd = _ml_arg->_nodelist[_iml];\n"
+        "    v = NODEV(_nd);\n");
 }
 
 void cvode_emit_interface() {
@@ -3046,7 +3057,9 @@ void net_receive(Item* qarg, Item* qp1, Item* qp2, Item* qstmt, Item* qend) {
         "  auto* const _ml = &_ml_real;\n"
         "  size_t const _iml{};\n");
     q = insertstr(qstmt, "  _ppvar = _nrn_mechanism_access_dparam(_pnt->_prop);\n");
-    vectorize_substitute(insertstr(q, ""), "  _thread = nullptr; _nt = (NrnThread*)_pnt->_vnt;");
+    vectorize_substitute(
+        insertstr(q, ""),
+        "  _thread = nullptr; double* _globals = nullptr; _nt = (NrnThread*)_pnt->_vnt;");
     if (debugging_) {
         if (0) {
             insertstr(qstmt, " assert(_tsav <= t); _tsav = t;");
@@ -3137,7 +3150,8 @@ void net_init(Item* qinit, Item* qp2) {
                          "  auto* const _ml = &_ml_real;\n"
                          "  size_t const _iml{};\n"
                          "  Datum* _ppvar = _nrn_mechanism_access_dparam(_pnt->_prop);\n"
-                         "  Datum* _thread = (Datum*)0;\n"
+                         "  Datum* _thread = nullptr;\n"
+                         "  double* _globals = nullptr;\n"
                          "  NrnThread* _nt = (NrnThread*)_pnt->_vnt;\n");
     if (net_init_q1_) {
         diag("NET_RECEIVE block can contain only one INITIAL block", (char*) 0);

src/nmodl/parsact.cpp

@@ -374,13 +374,15 @@ int check_tables_threads(List* p) {
             Sprintf(buf, "\nstatic void %s(_internalthreadargsproto_);", STR(q));
             lappendstr(p, buf);
         }
-        lappendstr(p,
-                   "\n"
-                   "static void _check_table_thread(_threadargsprotocomma_ int _type, "
-                   "_nrn_model_sorted_token const& _sorted_token) {\n"
-                   "  _nrn_mechanism_cache_range _lmr{_sorted_token, *_nt, *_ml, _type};\n"
-                   "  {\n"
-                   "    auto* const _ml = &_lmr;\n");
+        lappendstr(
+            p,
+            "\n"
+            "static void _check_table_thread(_threadargsprotocomma_ int _type, "
+            "_nrn_model_sorted_token const& _sorted_token) {\n"
+            "  if (gind != 0 && _thread != nullptr) { _globals = _thread[_gth].get<double*>(); } \n"
+            "  _nrn_mechanism_cache_range _lmr{_sorted_token, *_nt, *_ml, _type};\n"
+            "  {\n"
+            "    auto* const _ml = &_lmr;\n");
         ITERATE(q, check_table_thread_list) {
             Sprintf(buf, "  %s(_threadargs_);\n", STR(q));
             lappendstr(p, buf);
@@ -760,13 +762,16 @@ static void funchack(Symbol* n, bool ishoc, int hack) {
                    "    hoc_execerror(\"POINT_PROCESS data instance not valid\", NULL);\n"
                    "  }\n");
         q = lappendstr(procfunc, "  _setdata(_p);\n");
-        vectorize_substitute(q,
-                             "  _nrn_mechanism_cache_instance _ml_real{_p};\n"
-                             "  auto* const _ml = &_ml_real;\n"
-                             "  size_t const _iml{};\n"
-                             "  _ppvar = _nrn_mechanism_access_dparam(_p);\n"
-                             "  _thread = _extcall_thread.data();\n"
-                             "  _nt = static_cast<NrnThread*>(_pnt->_vnt);\n");
+        vectorize_substitute(
+            q,
+            "  _nrn_mechanism_cache_instance _ml_real{_p};\n"
+            "  auto* const _ml = &_ml_real;\n"
+            "  size_t const _iml{};\n"
+            "  _ppvar = _nrn_mechanism_access_dparam(_p);\n"
+            "  _thread = _extcall_thread.data();\n"
+            "  double* _globals = nullptr;\n"
+            "  if (gind != 0 && _thread != nullptr) { _globals = _thread[_gth].get<double*>(); }\n"
+            "  _nt = static_cast<NrnThread*>(_pnt->_vnt);\n");
     } else if (ishoc) {
         hocfunc_setdata_item(n, lappendstr(procfunc, ""));
         vectorize_substitute(
@@ -776,18 +781,23 @@ static void funchack(Symbol* n, bool ishoc, int hack) {
             "size_t const _iml{};\n"
             "_ppvar = _local_prop ? _nrn_mechanism_access_dparam(_local_prop) : nullptr;\n"
             "_thread = _extcall_thread.data();\n"
+            "double* _globals = nullptr;\n"
+            "if (gind != 0 && _thread != nullptr) { _globals = _thread[_gth].get<double*>(); }\n"
             "_nt = nrn_threads;\n");
     } else {  // _npy_...
         q = lappendstr(procfunc,
                        "  neuron::legacy::set_globals_from_prop(_prop, _ml_real, _ml, _iml);\n"
                        "  _ppvar = _nrn_mechanism_access_dparam(_prop);\n");
-        vectorize_substitute(q,
-                             "_nrn_mechanism_cache_instance _ml_real{_prop};\n"
-                             "auto* const _ml = &_ml_real;\n"
-                             "size_t const _iml{};\n"
-                             "_ppvar = _nrn_mechanism_access_dparam(_prop);\n"
-                             "_thread = _extcall_thread.data();\n"
-                             "_nt = nrn_threads;\n");
+        vectorize_substitute(
+            q,
+            "_nrn_mechanism_cache_instance _ml_real{_prop};\n"
+            "auto* const _ml = &_ml_real;\n"
+            "size_t const _iml{};\n"
+            "_ppvar = _nrn_mechanism_access_dparam(_prop);\n"
+            "_thread = _extcall_thread.data();\n"
+            "double* _globals = nullptr;\n"
+            "if (gind != 0 && _thread != nullptr) { _globals = _thread[_gth].get<double*>(); }\n"
+            "_nt = nrn_threads;\n");
     }
     if (n == last_func_using_table) {
         qp = lappendstr(procfunc, "");