Add default loop patterns back into the global par_loops

CHANGELOG.md

@@ -21,6 +21,7 @@
 - [[PR 1031]](https://github.com/parthenon-hpc-lab/parthenon/pull/1031) Fix bug in non-cell centered AMR
 
 ### Infrastructure (changes irrelevant to downstream codes)
+- [[PR 1066]](https://github.com/parthenon-hpc-lab/parthenon/pull/1066) Re-introduce default loop patterns and exec spaces
 - [[PR 1064]](https://github.com/parthenon-hpc-lab/parthenon/pull/1064) Forbid erroneous edge case when adding MeshData on a partition
 - [[PR 1035]](https://github.com/parthenon-hpc-lab/parthenon/pull/1035) Fix multigrid infrastructure to work with forest
 - [[PR 1048]](https://github.com/parthenon-hpc-lab/parthenon/pull/1048) Tiny fixes to custom coords logic

doc/sphinx/src/nested_par_for.rst

@@ -210,7 +210,7 @@ An ``IndexSplit`` object is typically used as:
 
   // Par for
   par_for_outer(
-	  DEFAULT_OUTER_LOOP_PATTERN, "KernalOuter", DevExecSpace(), scratch_size,
+	  "KernelOuter", scratch_size,
 	  scratch_level, 0, nblocks - 1, 0, idx_sp.outer_size() - 1,
 	  KOKKOS_LAMBDA(team_mbr_t member, const int b, const int outer_idx) {
 	    ScratchPad1D<Real> scratch(member.team_scratch(scratch_level), Nmax);
@@ -231,7 +231,7 @@ An ``IndexSplit`` object is typically used as:
 	      Real *var = &pack(b, ivar, k, jrange.s, flattened_inner_ijrange.s);
 
 	      // Do something with the pointer in the inner loop.
-	      par_for_inner(DEFAULT_INNER_LOOP_PATTERN, member, 0, flattened_inner_size,
+	      par_for_inner(member, 0, flattened_inner_size,
 	        [&](const int i) {
 		  foo(var[i]);
 		});

example/kokkos_pi/kokkos_pi.cpp

@@ -1,5 +1,5 @@
 //========================================================================================
-// (C) (or copyright) 2020-2021. Triad National Security, LLC. All rights reserved.
+// (C) (or copyright) 2020-2024. Triad National Security, LLC. All rights reserved.
 //
 // This program was produced under U.S. Government contract 89233218CNA000001 for Los
 // Alamos National Laboratory (LANL), which is operated by Triad National Security, LLC
@@ -272,9 +272,9 @@ result_t naiveParFor(int n_block, int n_mesh, int n_iter, double radius) {
       auto inOrOut = base->PackVariables({Metadata::Independent});
       // iops = 0  fops = 11
       par_for(
-          DEFAULT_LOOP_PATTERN, PARTHENON_AUTO_LABEL, DevExecSpace(), 0,
-          inOrOut.GetDim(4) - 1, nghost, inOrOut.GetDim(3) - nghost - 1, nghost,
-          inOrOut.GetDim(2) - nghost - 1, nghost, inOrOut.GetDim(1) - nghost - 1,
+          PARTHENON_AUTO_LABEL, 0, inOrOut.GetDim(4) - 1, nghost,
+          inOrOut.GetDim(3) - nghost - 1, nghost, inOrOut.GetDim(2) - nghost - 1, nghost,
+          inOrOut.GetDim(1) - nghost - 1,
           KOKKOS_LAMBDA(const int l, const int k_grid, const int j_grid,
                         const int i_grid) {
             const Real x =

example/poisson/poisson_package.cpp

@@ -1,5 +1,5 @@
 //========================================================================================
-// (C) (or copyright) 2021-2023. Triad National Security, LLC. All rights reserved.
+// (C) (or copyright) 2021-2024. Triad National Security, LLC. All rights reserved.
 //
 // This program was produced under U.S. Government contract 89233218CNA000001 for Los
 // Alamos National Laboratory (LANL), which is operated by Triad National Security, LLC
@@ -129,8 +129,7 @@ TaskStatus SetMatrixElements(T *u) {
   const int ndim = v.GetNdim();
   const Real w0 = -2.0 * ndim;
   parthenon::par_for(
-      DEFAULT_LOOP_PATTERN, PARTHENON_AUTO_LABEL, DevExecSpace(), 0, v.GetDim(5) - 1,
-      kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
+      PARTHENON_AUTO_LABEL, 0, v.GetDim(5) - 1, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
       KOKKOS_LAMBDA(const int b, const int k, const int j, const int i) {
         for (int n = isp_lo; n <= isp_hi; n++) {
           v(b, n, k, j, i) = 1;
@@ -246,8 +245,7 @@ TaskStatus UpdatePhi(T *u, T *du) {
   if (isp_hi < 0) { // there is no sparse matrix, so we must be using the stencil
     const auto &stencil = pkg->Param<Stencil_t>("stencil");
     parthenon::par_for(
-        DEFAULT_LOOP_PATTERN, PARTHENON_AUTO_LABEL, DevExecSpace(), 0, v.GetDim(5) - 1,
-        kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
+        PARTHENON_AUTO_LABEL, 0, v.GetDim(5) - 1, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
         KOKKOS_LAMBDA(const int b, const int k, const int j, const int i) {
           const Real rhs = dV * v(b, irho, k, j, i);
           const Real phi_new = stencil.Jacobi(v, iphi, b, k, j, i, rhs);
@@ -258,8 +256,7 @@ TaskStatus UpdatePhi(T *u, T *du) {
     const auto &sp_accessor =
         pkg->Param<parthenon::solvers::SparseMatrixAccessor>("sparse_accessor");
     parthenon::par_for(
-        DEFAULT_LOOP_PATTERN, PARTHENON_AUTO_LABEL, DevExecSpace(), 0, v.GetDim(5) - 1,
-        kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
+        PARTHENON_AUTO_LABEL, 0, v.GetDim(5) - 1, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
         KOKKOS_LAMBDA(const int b, const int k, const int j, const int i) {
           const Real rhs = dV * v(b, irho, k, j, i);
           const Real phi_new =
@@ -269,8 +266,7 @@ TaskStatus UpdatePhi(T *u, T *du) {
   }
 
   parthenon::par_for(
-      DEFAULT_LOOP_PATTERN, PARTHENON_AUTO_LABEL, DevExecSpace(), 0, dv.GetDim(5) - 1,
-      kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
+      PARTHENON_AUTO_LABEL, 0, dv.GetDim(5) - 1, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
       KOKKOS_LAMBDA(const int b, const int k, const int j, const int i) {
         v(b, iphi, k, j, i) += dv(b, idphi, k, j, i);
       });

example/poisson_gmg/poisson_equation.hpp

@@ -1,5 +1,5 @@
 //========================================================================================
-// (C) (or copyright) 2023. Triad National Security, LLC. All rights reserved.
+// (C) (or copyright) 2023-2024. Triad National Security, LLC. All rights reserved.
 //
 // This program was produced under U.S. Government contract 89233218CNA000001 for Los
 // Alamos National Laboratory (LANL), which is operated by Triad National Security, LLC
@@ -75,8 +75,7 @@ class PoissonEquation {
     auto desc = parthenon::MakePackDescriptor<diag_t, D>(md.get());
     auto pack = desc.GetPack(md.get(), include_block);
     parthenon::par_for(
-        DEFAULT_LOOP_PATTERN, "StoreDiagonal", DevExecSpace(), 0, pack.GetNBlocks() - 1,
-        kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
+        "StoreDiagonal", 0, pack.GetNBlocks() - 1, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
         KOKKOS_LAMBDA(const int b, const int k, const int j, const int i) {
           const auto &coords = pack.GetCoordinates(b);
           // Build the unigrid diagonal of the matrix
@@ -122,8 +121,7 @@ class PoissonEquation {
         parthenon::MakePackDescriptor<var_t, D>(md.get(), {}, {PDOpt::WithFluxes});
     auto pack = desc.GetPack(md.get(), include_block);
     parthenon::par_for(
-        DEFAULT_LOOP_PATTERN, "CaclulateFluxes", DevExecSpace(), 0, pack.GetNBlocks() - 1,
-        kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
+        "CaclulateFluxes", 0, pack.GetNBlocks() - 1, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
         KOKKOS_LAMBDA(const int b, const int k, const int j, const int i) {
           const auto &coords = pack.GetCoordinates(b);
           Real dx1 = coords.template Dxc<X1DIR>(k, j, i);
@@ -185,9 +183,8 @@ class PoissonEquation {
         parthenon::MakePackDescriptor<in_t, out_t>(md.get(), {}, {PDOpt::WithFluxes});
     auto pack = desc.GetPack(md.get(), include_block);
     parthenon::par_for(
-        DEFAULT_LOOP_PATTERN, "FluxMultiplyMatrix", DevExecSpace(), 0,
-        pack.GetNBlocks() - 1, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
-        KOKKOS_LAMBDA(const int b, const int k, const int j, const int i) {
+        "FluxMultiplyMatrix", 0, pack.GetNBlocks() - 1, kb.s, kb.e, jb.s, jb.e, ib.s,
+        ib.e, KOKKOS_LAMBDA(const int b, const int k, const int j, const int i) {
           const auto &coords = pack.GetCoordinates(b);
           Real dx1 = coords.template Dxc<X1DIR>(k, j, i);
           pack(b, te, out_t(), k, j, i) = -alpha * pack(b, te, in_t(), k, j, i);

src/interface/update.cpp

@@ -1,5 +1,5 @@
 //========================================================================================
-// (C) (or copyright) 2020-2023. Triad National Security, LLC. All rights reserved.
+// (C) (or copyright) 2020-2024. Triad National Security, LLC. All rights reserved.
 //
 // This program was produced under U.S. Government contract 89233218CNA000001 for Los
 // Alamos National Laboratory (LANL), which is operated by Triad National Security, LLC

src/interface/update.hpp

@@ -1,5 +1,5 @@
 //========================================================================================
-// (C) (or copyright) 2020-2023. Triad National Security, LLC. All rights reserved.
+// (C) (or copyright) 2020-2024. Triad National Security, LLC. All rights reserved.
 //
 // This program was produced under U.S. Government contract 89233218CNA000001 for Los
 // Alamos National Laboratory (LANL), which is operated by Triad National Security, LLC
@@ -75,8 +75,8 @@ TaskStatus WeightedSumData(const F &flags, T *in1, T *in2, const Real w1, const
   const auto &y = in2->PackVariables(flags);
   const auto &z = out->PackVariables(flags);
   parthenon::par_for(
-      DEFAULT_LOOP_PATTERN, PARTHENON_AUTO_LABEL, DevExecSpace(), 0, x.GetDim(5) - 1, 0,
-      x.GetDim(4) - 1, 0, x.GetDim(3) - 1, 0, x.GetDim(2) - 1, 0, x.GetDim(1) - 1,
+      PARTHENON_AUTO_LABEL, 0, x.GetDim(5) - 1, 0, x.GetDim(4) - 1, 0, x.GetDim(3) - 1, 0,
+      x.GetDim(2) - 1, 0, x.GetDim(1) - 1,
       KOKKOS_LAMBDA(const int b, const int l, const int k, const int j, const int i) {
         // TOOD(someone) This is potentially dangerous and/or not intended behavior
         // as we still may want to update (or populate) z if any of those vars are
@@ -98,8 +98,8 @@ TaskStatus SetDataToConstant(const F &flags, T *data, const Real val) {
   PARTHENON_INSTRUMENT
   const auto &x = data->PackVariables(flags);
   parthenon::par_for(
-      DEFAULT_LOOP_PATTERN, PARTHENON_AUTO_LABEL, DevExecSpace(), 0, x.GetDim(5) - 1, 0,
-      x.GetDim(4) - 1, 0, x.GetDim(3) - 1, 0, x.GetDim(2) - 1, 0, x.GetDim(1) - 1,
+      PARTHENON_AUTO_LABEL, 0, x.GetDim(5) - 1, 0, x.GetDim(4) - 1, 0, x.GetDim(3) - 1, 0,
+      x.GetDim(2) - 1, 0, x.GetDim(1) - 1,
       KOKKOS_LAMBDA(const int b, const int l, const int k, const int j, const int i) {
         if (x.IsAllocated(b, l)) {
           x(b, l, k, j, i) = val;
@@ -161,8 +161,8 @@ TaskStatus Update2S(const F &flags, T *s0_data, T *s1_data, T *rhs_data,
   Real gam0 = pint->gam0[stage - 1];
   Real gam1 = pint->gam1[stage - 1];
   parthenon::par_for(
-      DEFAULT_LOOP_PATTERN, PARTHENON_AUTO_LABEL, DevExecSpace(), 0, s0.GetDim(5) - 1, 0,
-      s0.GetDim(4) - 1, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
+      PARTHENON_AUTO_LABEL, 0, s0.GetDim(5) - 1, 0, s0.GetDim(4) - 1, kb.s, kb.e, jb.s,
+      jb.e, ib.s, ib.e,
       KOKKOS_LAMBDA(const int b, const int l, const int k, const int j, const int i) {
         if (s0.IsAllocated(b, l) && s1.IsAllocated(b, l) && rhs.IsAllocated(b, l)) {
           if (update_s1) {
@@ -199,8 +199,8 @@ TaskStatus SumButcher(const F &flags, std::shared_ptr<T> base_data,
   const IndexRange jb = out_data->GetBoundsJ(interior);
   const IndexRange kb = out_data->GetBoundsK(interior);
   parthenon::par_for(
-      DEFAULT_LOOP_PATTERN, PARTHENON_AUTO_LABEL, DevExecSpace(), 0, out.GetDim(5) - 1, 0,
-      out.GetDim(4) - 1, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
+      PARTHENON_AUTO_LABEL, 0, out.GetDim(5) - 1, 0, out.GetDim(4) - 1, kb.s, kb.e, jb.s,
+      jb.e, ib.s, ib.e,
       KOKKOS_LAMBDA(const int b, const int l, const int k, const int j, const int i) {
         if (out.IsAllocated(b, l) && in.IsAllocated(b, l)) {
           out(b, l, k, j, i) = in(b, l, k, j, i);
@@ -210,8 +210,8 @@ TaskStatus SumButcher(const F &flags, std::shared_ptr<T> base_data,
     Real a = pint->a[stage - 1][prev];
     const auto &in = stage_data[stage]->PackVariables(flags);
     parthenon::par_for(
-        DEFAULT_LOOP_PATTERN, PARTHENON_AUTO_LABEL, DevExecSpace(), 0, out.GetDim(5) - 1,
-        0, out.GetDim(4) - 1, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
+        PARTHENON_AUTO_LABEL, 0, out.GetDim(5) - 1, 0, out.GetDim(4) - 1, kb.s, kb.e,
+        jb.s, jb.e, ib.s, ib.e,
         KOKKOS_LAMBDA(const int b, const int l, const int k, const int j, const int i) {
           if (out.IsAllocated(b, l) && in.IsAllocated(b, l)) {
             out(b, l, k, j, i) += dt * a * in(b, l, k, j, i);
@@ -247,8 +247,8 @@ TaskStatus UpdateButcher(const F &flags, std::vector<std::shared_ptr<T>> stage_d
     const Real butcher_b = pint->b[stage];
     const auto &in = stage_data[stage]->PackVariables(flags);
     parthenon::par_for(
-        DEFAULT_LOOP_PATTERN, PARTHENON_AUTO_LABEL, DevExecSpace(), 0, out.GetDim(5) - 1,
-        0, out.GetDim(4) - 1, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
+        PARTHENON_AUTO_LABEL, 0, out.GetDim(5) - 1, 0, out.GetDim(4) - 1, kb.s, kb.e,
+        jb.s, jb.e, ib.s, ib.e,
         KOKKOS_LAMBDA(const int b, const int l, const int k, const int j, const int i) {
           if (out.IsAllocated(b, l) && in.IsAllocated(b, l)) {
             out(b, l, k, j, i) += dt * b * in(b, l, k, j, i);

src/kokkos_abstraction.hpp

@@ -3,7 +3,7 @@
 // Copyright(C) 2020-2023 The Parthenon collaboration
 // Licensed under the 3-clause BSD License, see LICENSE file for details
 //========================================================================================
-// (C) (or copyright) 2020-2023. Triad National Security, LLC. All rights reserved.
+// (C) (or copyright) 2020-2024. Triad National Security, LLC. All rights reserved.
 //
 // This program was produced under U.S. Government contract 89233218CNA000001
 // for Los Alamos National Laboratory (LANL), which is operated by Triad
@@ -28,6 +28,7 @@
 #include <Kokkos_Core.hpp>
 
 #include "basic_types.hpp"
+#include "config.hpp"
 #include "parthenon_array_generic.hpp"
 #include "utils/error_checking.hpp"
 #include "utils/instrument.hpp"
@@ -631,6 +632,12 @@ inline void par_dispatch(LoopPatternSimdFor, const std::string &name,
               function(l, m, n, k, j, i);
 }
 
+template <typename Tag, typename... Args>
+inline void par_dispatch(const std::string &name, Args &&...args) {
+  par_dispatch<Tag>(DEFAULT_LOOP_PATTERN, name, DevExecSpace(),
+                    std::forward<Args>(args)...);
+}
+
 template <class... Args>
 inline void par_for(Args &&...args) {
   par_dispatch<dispatch_impl::ParallelForDispatch>(std::forward<Args>(args)...);
@@ -715,6 +722,12 @@ inline void par_for_outer(OuterLoopPatternTeams, const std::string &name,
       });
 }
 
+template <typename... Args>
+inline void par_for_outer(const std::string &name, Args &&...args) {
+  par_for_outer(DEFAULT_OUTER_LOOP_PATTERN, name, DevExecSpace(),
+                std::forward<Args>(args)...);
+}
+
 // Inner parallel loop using TeamThreadRange
 template <typename Function>
 KOKKOS_FORCEINLINE_FUNCTION void
@@ -903,6 +916,11 @@ KOKKOS_FORCEINLINE_FUNCTION void par_for_inner(InnerLoopPatternSimdFor,
   }
 }
 
+template <typename... Args>
+KOKKOS_FORCEINLINE_FUNCTION void par_for_inner(team_mbr_t team_member, Args &&...args) {
+  par_for_inner(DEFAULT_INNER_LOOP_PATTERN, team_member, std::forward<Args>(args)...);
+}
+
 // reused from kokoks/core/perf_test/PerfTest_ExecSpacePartitioning.cpp
 // commit a0d011fb30022362c61b3bb000ae3de6906cb6a7
 template <class ExecSpace>

src/mesh/mesh-amr_loadbalance.cpp

@@ -124,8 +124,7 @@ bool TryRecvCoarseToFine(int lid_recv, int send_rank, const LogicalLocation &fin
         const int is = (ox1 == 0) ? 0 : (ib_int.e - ib_int.s + 1) / 2;
         const int idx_te = static_cast<int>(te) % 3;
         parthenon::par_for(
-            DEFAULT_LOOP_PATTERN, PARTHENON_AUTO_LABEL, DevExecSpace(), 0, nt, 0, nu, 0,
-            nv, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
+            PARTHENON_AUTO_LABEL, 0, nt, 0, nu, 0, nv, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
             KOKKOS_LAMBDA(const int t, const int u, const int v, const int k, const int j,
                           const int i) {
               cb(idx_te, t, u, v, k, j, i) = fb(idx_te, t, u, v, k + ks, j + js, i + is);
@@ -217,8 +216,7 @@ bool TryRecvFineToCoarse(int lid_recv, int send_rank, const LogicalLocation &fin
         const int is = (ox1 == 0) ? 0 : (ib.e - ib.s + 1 - TopologicalOffsetI(te));
         const int idx_te = static_cast<int>(te) % 3;
         parthenon::par_for(
-            DEFAULT_LOOP_PATTERN, PARTHENON_AUTO_LABEL, DevExecSpace(), 0, nt, 0, nu, 0,
-            nv, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
+            PARTHENON_AUTO_LABEL, 0, nt, 0, nu, 0, nv, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
             KOKKOS_LAMBDA(const int t, const int u, const int v, const int k, const int j,
                           const int i) {
               fb(idx_te, t, u, v, k + ks, j + js, i + is) = cb(idx_te, t, u, v, k, j, i);

src/outputs/histogram.cpp

@@ -3,7 +3,7 @@
 // Copyright(C) 2023 The Parthenon collaboration
 // Licensed under the 3-clause BSD License, see LICENSE file for details
 //========================================================================================
-// (C) (or copyright) 2023. Triad National Security, LLC. All rights reserved.
+// (C) (or copyright) 2023-2024. Triad National Security, LLC. All rights reserved.
 //
 // This program was produced under U.S. Government contract 89233218CNA000001 for Los
 // Alamos National Laboratory (LANL), which is operated by Triad National Security, LLC
@@ -287,8 +287,7 @@ void Histogram::CalcHist(Mesh *pm) {
     const auto kb = md->GetBoundsK(IndexDomain::interior);
 
     parthenon::par_for(
-        DEFAULT_LOOP_PATTERN, "CalcHist", DevExecSpace(), 0, md->NumBlocks() - 1, kb.s,
-        kb.e, jb.s, jb.e, ib.s, ib.e,
+        "CalcHist", 0, md->NumBlocks() - 1, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
         KOKKOS_LAMBDA(const int b, const int k, const int j, const int i) {
           auto &coords = x_var.GetCoords(b);
           auto x_val = std::numeric_limits<Real>::quiet_NaN();

src/prolong_restrict/pr_loops.hpp

@@ -171,8 +171,8 @@ InnerHostProlongationRestrictionLoop(std::size_t buf, const ProResInfoArrHost_t
   auto coarse = info(buf).coarse;
   auto fine = info(buf).fine;
   par_for(
-      DEFAULT_LOOP_PATTERN, PARTHENON_AUTO_LABEL, DevExecSpace(), 0, 0, 0, 0, 0,
-      idxer.size() - 1, KOKKOS_LAMBDA(const int, const int, const int ii) {
+      PARTHENON_AUTO_LABEL, 0, 0, 0, 0, 0, idxer.size() - 1,
+      KOKKOS_LAMBDA(const int, const int, const int ii) {
         const auto [t, u, v, k, j, i] = idxer(ii);
         if (idxer.IsActive(k, j, i)) {
           Stencil::template Do<DIM, FEL, CEL>(t, u, v, k, j, i, ckb, cjb, cib, kb, jb, ib,

src/solvers/mg_solver.hpp

@@ -195,8 +195,7 @@ class MGSolver {
     auto pack = desc.GetPack(md.get(), include_block);
     if (params_.two_by_two_diagonal) {
       parthenon::par_for(
-          DEFAULT_LOOP_PATTERN, "CaclulateFluxes", DevExecSpace(), 0,
-          pack.GetNBlocks() - 1, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
+          "CaclulateFluxes", 0, pack.GetNBlocks() - 1, kb.s, kb.e, jb.s, jb.e, ib.s, ib.e,
           KOKKOS_LAMBDA(const int b, const int k, const int j, const int i) {
             const auto &coords = pack.GetCoordinates(b);