From 3a9361dc368018d6c2e2911dd2b8781d9eb0b205 Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante5.lvt.dkrz.de>
Date: Mon, 24 Feb 2025 14:21:48 +0100
Subject: [PATCH 01/33] Added bindings for rbf
---
.../mo_lib_intp_rbf_bindings.cpp | 32 +++++++++++++++++++
src/interpolation/mo_lib_intp_rbf_bindings.h | 22 +++++++++++++
2 files changed, 54 insertions(+)
create mode 100644 src/interpolation/mo_lib_intp_rbf_bindings.cpp
create mode 100644 src/interpolation/mo_lib_intp_rbf_bindings.h
diff --git a/src/interpolation/mo_lib_intp_rbf_bindings.cpp b/src/interpolation/mo_lib_intp_rbf_bindings.cpp
new file mode 100644
index 0000000..281a89f
--- /dev/null
+++ b/src/interpolation/mo_lib_intp_rbf_bindings.cpp
@@ -0,0 +1,32 @@
+// ICON
+//
+// ---------------------------------------------------------------
+// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
+// Contact information: icon-model.org
+//
+// See AUTHORS.TXT for a list of authors
+// See LICENSES/ for license information
+// SPDX-License-Identifier: BSD-3-Clause
+// ---------------------------------------------------------------
+
+#include mo_lib_intp_rbf_bindings.h
+#include mo_lib_intp_rbf.hpp
+
+
+void rbf_interpol_c2grad_lib(const wp* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk
+ const wp* rbf_c2grad_coeff, const wp* grad_x, const wp* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, bool lacc){
+
+
+rbf_interpol_c2grad_lib<wp>(p_cell_in, rbf_c2grad_idx,
+ rbf_c2grad_blk, rbf_c2grad_coeff,
+ grad_x, grad_y, i_startblk, i_endblk,
+ i_startidx_in, i_endidx_in, slev, elev,
+ nproma, lacc);
+
+
+}
+
+
+
diff --git a/src/interpolation/mo_lib_intp_rbf_bindings.h b/src/interpolation/mo_lib_intp_rbf_bindings.h
new file mode 100644
index 0000000..826fb7e
--- /dev/null
+++ b/src/interpolation/mo_lib_intp_rbf_bindings.h
@@ -0,0 +1,22 @@
+// ICON
+//
+// ---------------------------------------------------------------
+// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
+// Contact information: icon-model.org
+//
+// See AUTHORS.TXT for a list of authors
+// See LICENSES/ for license information
+// SPDX-License-Identifier: BSD-3-Clause
+// ---------------------------------------------------------------
+#pragma once
+
+#ifdef __SINGLE_PRECISION
+ using wp = single;
+#else
+ using wp = double;
+#endif
+
+void rbf_interpol_c2grad_lib(const wp* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const wp* rbf_c2grad_coeff, const wp* grad_x, const wp* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, bool lacc);
--
GitLab
From 63e4efcf558af3b5d08ecc7254b258c099036d8f Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante5.lvt.dkrz.de>
Date: Mon, 24 Feb 2025 14:42:23 +0100
Subject: [PATCH 02/33] added Views for c2grad_lib
---
src/interpolation/mo_lib_intp_rbf.cpp | 31 +++++++++++++++++++++++++++
src/interpolation/mo_lib_intp_rbf.hpp | 21 ++++++++++++++++++
2 files changed, 52 insertions(+)
create mode 100644 src/interpolation/mo_lib_intp_rbf.cpp
create mode 100644 src/interpolation/mo_lib_intp_rbf.hpp
diff --git a/src/interpolation/mo_lib_intp_rbf.cpp b/src/interpolation/mo_lib_intp_rbf.cpp
new file mode 100644
index 0000000..c74503e
--- /dev/null
+++ b/src/interpolation/mo_lib_intp_rbf.cpp
@@ -0,0 +1,31 @@
+// ICON
+//
+// ---------------------------------------------------------------
+// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
+// Contact information: icon-model.org
+//
+// See AUTHORS.TXT for a list of authors
+// See LICENSES/ for license information
+// SPDX-License-Identifier: BSD-3-Clause
+// ---------------------------------------------------------------
+
+#inlcude mo_lib_intp_rbf.hpp
+
+
+void rbf_interpol_c2grad_lib(const wp* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2gra
+ const wp* rbf_c2grad_coeff, const wp* grad_x, const wp* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, bool lacc){
+
+ //aliases for unmanaged Kokkos views
+ typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
+ typedef Kokkos::View<T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedT3D;
+ typedef Kokkos::View<const int***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstInt3D;
+
+ //to avoid memory ownership issues
+ UnmanagedConstT3D p_cell_in_view(p_cell_in, nproma, nlev, nblks_c);
+ UnmanagedConstT3D grad_x_view(grad_x, nproma, nlev, nblks_c);
+ UnmanagedConstT3D grad_y_view(grad_y, nproma, nlev, nblks_c);
+ UnmanagedConstInt3D rbf_c2grad_idx_view(rbf_c2grad_idx, rbf_c2grad_dim, nproma, nblks_c);
+ UnmanagedConstInt3D rbf_c2grad_blk_view(rbf_c2grad_blk, rbf_c2grad_dim, nproma, nblks_c);
+}
diff --git a/src/interpolation/mo_lib_intp_rbf.hpp b/src/interpolation/mo_lib_intp_rbf.hpp
new file mode 100644
index 0000000..5fbf68b
--- /dev/null
+++ b/src/interpolation/mo_lib_intp_rbf.hpp
@@ -0,0 +1,21 @@
+// ICON
+//
+// ---------------------------------------------------------------
+// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
+// Contact information: icon-model.org
+//
+// See AUTHORS.TXT for a list of authors
+// See LICENSES/ for license information
+// SPDX-License-Identifier: BSD-3-Clause
+// ---------------------------------------------------------------
+#pragma once
+
+#include "mo_lib_loopindices.hpp"
+#include <Kokkos_Core.hpp>
+#include <vector>
+
+template <typename T>
+void rbf_interpol_c2grad_lib(const wp* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2gra
+ const wp* rbf_c2grad_coeff, const wp* grad_x, const wp* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, bool lacc);
--
GitLab
From c1a0ccffe73b60020a51ad83502783d03a437a43 Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante5.lvt.dkrz.de>
Date: Mon, 24 Feb 2025 14:44:57 +0100
Subject: [PATCH 03/33] fixed typo!
---
src/interpolation/mo_lib_intp_rbf.cpp | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/interpolation/mo_lib_intp_rbf.cpp b/src/interpolation/mo_lib_intp_rbf.cpp
index c74503e..795d343 100644
--- a/src/interpolation/mo_lib_intp_rbf.cpp
+++ b/src/interpolation/mo_lib_intp_rbf.cpp
@@ -9,7 +9,7 @@
// SPDX-License-Identifier: BSD-3-Clause
// ---------------------------------------------------------------
-#inlcude mo_lib_intp_rbf.hpp
+#include mo_lib_intp_rbf.hpp
void rbf_interpol_c2grad_lib(const wp* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2gra
--
GitLab
From 016db45178c908b4e57f27e3950231ac0cf97143 Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante3.lvt.dkrz.de>
Date: Mon, 24 Feb 2025 15:22:30 +0100
Subject: [PATCH 04/33] added binding file into cmake
---
src/interpolation/CMakeLists.txt | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/interpolation/CMakeLists.txt b/src/interpolation/CMakeLists.txt
index 1051516..f1dc7b9 100644
--- a/src/interpolation/CMakeLists.txt
+++ b/src/interpolation/CMakeLists.txt
@@ -19,7 +19,7 @@ add_library(
mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib.cpp
mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib_bindings.cpp
interpolation_bindings.cpp
-)
+ mo_lib_intp_rbf_bindings.cpp)
add_library(${PROJECT_NAME}::interpolation ALIAS iconmath-interpolation)
--
GitLab
From 4c640a3a852321ca55f18c9f3d2bd9f882e05b30 Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante5.lvt.dkrz.de>
Date: Mon, 24 Feb 2025 15:32:30 +0100
Subject: [PATCH 05/33] Fixed typo
---
src/interpolation/mo_lib_intp_rbf_bindings.cpp | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/src/interpolation/mo_lib_intp_rbf_bindings.cpp b/src/interpolation/mo_lib_intp_rbf_bindings.cpp
index 281a89f..8bb0feb 100644
--- a/src/interpolation/mo_lib_intp_rbf_bindings.cpp
+++ b/src/interpolation/mo_lib_intp_rbf_bindings.cpp
@@ -9,8 +9,8 @@
// SPDX-License-Identifier: BSD-3-Clause
// ---------------------------------------------------------------
-#include mo_lib_intp_rbf_bindings.h
-#include mo_lib_intp_rbf.hpp
+#include "mo_lib_intp_rbf_bindings.h"
+#include "mo_lib_intp_rbf.hpp"
void rbf_interpol_c2grad_lib(const wp* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk
--
GitLab
From 3fe938837b825e1d47d5c203c3131cb9d0c1b6a0 Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante5.lvt.dkrz.de>
Date: Mon, 24 Feb 2025 23:00:08 +0100
Subject: [PATCH 06/33] Ported rbf_interpol_c2grad_lib
---
src/interpolation/CMakeLists.txt | 3 +-
src/interpolation/mo_lib_intp_rbf.cpp | 66 ++++++++++++++++---
src/interpolation/mo_lib_intp_rbf.hpp | 6 +-
.../mo_lib_intp_rbf_bindings.cpp | 25 +++++--
src/interpolation/mo_lib_intp_rbf_bindings.h | 18 ++---
5 files changed, 91 insertions(+), 27 deletions(-)
diff --git a/src/interpolation/CMakeLists.txt b/src/interpolation/CMakeLists.txt
index f1dc7b9..d5af767 100644
--- a/src/interpolation/CMakeLists.txt
+++ b/src/interpolation/CMakeLists.txt
@@ -19,7 +19,8 @@ add_library(
mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib.cpp
mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib_bindings.cpp
interpolation_bindings.cpp
- mo_lib_intp_rbf_bindings.cpp)
+ mo_lib_intp_rbf_bindings.cpp
+ mo_lib_intp_rbf.cpp)
add_library(${PROJECT_NAME}::interpolation ALIAS iconmath-interpolation)
diff --git a/src/interpolation/mo_lib_intp_rbf.cpp b/src/interpolation/mo_lib_intp_rbf.cpp
index 795d343..666e02c 100644
--- a/src/interpolation/mo_lib_intp_rbf.cpp
+++ b/src/interpolation/mo_lib_intp_rbf.cpp
@@ -9,23 +9,73 @@
// SPDX-License-Identifier: BSD-3-Clause
// ---------------------------------------------------------------
-#include mo_lib_intp_rbf.hpp
+#include "mo_lib_intp_rbf.hpp"
+#include <Kokkos_Core.hpp>
-
-void rbf_interpol_c2grad_lib(const wp* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2gra
- const wp* rbf_c2grad_coeff, const wp* grad_x, const wp* grad_y,
+template <typename T>
+void rbf_interpol_c2grad_lib(const T* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const T* rbf_c2grad_coeff, T* grad_x, T* grad_y,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, bool lacc){
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblks_c, bool lacc){
//aliases for unmanaged Kokkos views
typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
typedef Kokkos::View<T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedT3D;
typedef Kokkos::View<const int***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstInt3D;
+ typedef Kokkos::View<const T****, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT4D;
//to avoid memory ownership issues
UnmanagedConstT3D p_cell_in_view(p_cell_in, nproma, nlev, nblks_c);
- UnmanagedConstT3D grad_x_view(grad_x, nproma, nlev, nblks_c);
- UnmanagedConstT3D grad_y_view(grad_y, nproma, nlev, nblks_c);
+ UnmanagedT3D grad_x_view(grad_x, nproma, nlev, nblks_c);
+ UnmanagedT3D grad_y_view(grad_y, nproma, nlev, nblks_c);
UnmanagedConstInt3D rbf_c2grad_idx_view(rbf_c2grad_idx, rbf_c2grad_dim, nproma, nblks_c);
UnmanagedConstInt3D rbf_c2grad_blk_view(rbf_c2grad_blk, rbf_c2grad_dim, nproma, nblks_c);
-}
+ UnmanagedConstT4D rbf_c2grad_coeff_view(rbf_c2grad_coeff, rbf_c2grad_dim, 2, nproma, nblks_c);
+
+ for (int jb = i_startblk; jb < i_endblk; ++jb) {
+
+ int i_startidx, i_endidx;
+ get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk, i_startidx, i_endidx);
+
+ Kokkos::parallel_for("rbf_interpol_c2grad",
+ Kokkos::MDRangePolicy<Kokkos::Rank<2>>({slev, i_startidx}, {elev, i_endidx}),
+ KOKKOS_LAMBDA(const int jk, const int jc){
+
+ grad_x_view(jc, jk, jb) =
+ rbf_c2grad_coeff_view(0, 1, jc, jb)*
+ p_cell_in_view(jc, jk, jb) +
+ rbf_c2grad_coeff_view(1, 1, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(1, jc, jb), jk, rbf_c2grad_blk_view(1, jc, jb)) +
+ rbf_c2grad_coeff_view(2, 1, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(2, jc, jb), jk, rbf_c2grad_blk_view(2, jc, jb)) +
+ rbf_c2grad_coeff_view(3, 1, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(3, jc, jb), jk, rbf_c2grad_blk_view(3, jc, jb)) +
+ rbf_c2grad_coeff_view(4, 1, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(4, jc, jb), jk, rbf_c2grad_blk_view(4, jc, jb)) +
+ rbf_c2grad_coeff_view(5, 1, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(5, jc, jb), jk, rbf_c2grad_blk_view(5, jc, jb)) +
+ rbf_c2grad_coeff_view(6, 1, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(6, jc, jb), jk, rbf_c2grad_blk_view(6, jc, jb)) +
+ rbf_c2grad_coeff_view(7, 1, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(7, jc, jb), jk, rbf_c2grad_blk_view(7, jc, jb)) +
+ rbf_c2grad_coeff_view(8, 1, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(8, jc, jb), jk, rbf_c2grad_blk_view(8, jc, jb)) +
+ rbf_c2grad_coeff_view(9, 1, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(9, jc, jb), jk, rbf_c2grad_blk_view(9, jc, jb));
+
+ });
+
+ }//for
+}//void
+
+template
+void rbf_interpol_c2grad_lib<double>(const double* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const double* rbf_c2grad_coeff, double* grad_x, double* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
+
+template
+void rbf_interpol_c2grad_lib<float>(const float* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const float* rbf_c2grad_coeff, float* grad_x, float* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
diff --git a/src/interpolation/mo_lib_intp_rbf.hpp b/src/interpolation/mo_lib_intp_rbf.hpp
index 5fbf68b..8006a43 100644
--- a/src/interpolation/mo_lib_intp_rbf.hpp
+++ b/src/interpolation/mo_lib_intp_rbf.hpp
@@ -15,7 +15,7 @@
#include <vector>
template <typename T>
-void rbf_interpol_c2grad_lib(const wp* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2gra
- const wp* rbf_c2grad_coeff, const wp* grad_x, const wp* grad_y,
+void rbf_interpol_c2grad_lib(const T* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const T* rbf_c2grad_coeff, T* grad_x, T* grad_y,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, bool lacc);
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
diff --git a/src/interpolation/mo_lib_intp_rbf_bindings.cpp b/src/interpolation/mo_lib_intp_rbf_bindings.cpp
index 8bb0feb..a44a3da 100644
--- a/src/interpolation/mo_lib_intp_rbf_bindings.cpp
+++ b/src/interpolation/mo_lib_intp_rbf_bindings.cpp
@@ -9,24 +9,35 @@
// SPDX-License-Identifier: BSD-3-Clause
// ---------------------------------------------------------------
+
#include "mo_lib_intp_rbf_bindings.h"
#include "mo_lib_intp_rbf.hpp"
-void rbf_interpol_c2grad_lib(const wp* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk
- const wp* rbf_c2grad_coeff, const wp* grad_x, const wp* grad_y,
+void rbf_interpol_c2grad_lib_sp(const float* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const float* rbf_c2grad_coeff, float* grad_x, float* grad_y,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, bool lacc){
+ int slev, int elev, int nproma, int nlev, int nblk_c, int rbf_c2grad_dim, bool lacc){
-rbf_interpol_c2grad_lib<wp>(p_cell_in, rbf_c2grad_idx,
- rbf_c2grad_blk, rbf_c2grad_coeff,
+rbf_interpol_c2grad_lib<float>(p_cell_in, rbf_c2grad_idx, rbf_c2grad_blk,
+ rbf_c2grad_coeff,
grad_x, grad_y, i_startblk, i_endblk,
i_startidx_in, i_endidx_in, slev, elev,
- nproma, lacc);
+ nproma, nlev, nblk_c, rbf_c2grad_dim, lacc);
+}
-}
+void rbf_interpol_c2grad_lib_dp(const double* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const double* rbf_c2grad_coeff, double* grad_x, double* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int rbf_c2grad_dim, bool lacc){
+rbf_interpol_c2grad_lib<double>(p_cell_in, rbf_c2grad_idx, rbf_c2grad_blk,
+ rbf_c2grad_coeff,
+ grad_x, grad_y, i_startblk, i_endblk,
+ i_startidx_in, i_endidx_in, slev, elev,
+ nproma, nlev, nblk_c, rbf_c2grad_dim, lacc);
+}
diff --git a/src/interpolation/mo_lib_intp_rbf_bindings.h b/src/interpolation/mo_lib_intp_rbf_bindings.h
index 826fb7e..e7ef8f3 100644
--- a/src/interpolation/mo_lib_intp_rbf_bindings.h
+++ b/src/interpolation/mo_lib_intp_rbf_bindings.h
@@ -10,13 +10,15 @@
// ---------------------------------------------------------------
#pragma once
-#ifdef __SINGLE_PRECISION
- using wp = single;
-#else
- using wp = double;
-#endif
+extern "C" {
-void rbf_interpol_c2grad_lib(const wp* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const wp* rbf_c2grad_coeff, const wp* grad_x, const wp* grad_y,
+void rbf_interpol_c2grad_lib_sp(const float* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const float* rbf_c2grad_coeff, float* grad_x, const float* grad_y,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, bool lacc);
+ int slev, int elev, int nproma, int nlev, int nblk_c, int rbf_c2grad_dim, bool lacc);
+
+void rbf_interpol_c2grad_lib_dp(const double* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const double* rbf_c2grad_coeff, double* grad_x, const double* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int rbf_c2grad_dim, bool lacc);
+}
--
GitLab
From 19ab5976de9b83a3ac37333701826e0203851f8a Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante4.lvt.dkrz.de>
Date: Mon, 24 Feb 2025 23:24:44 +0100
Subject: [PATCH 07/33] Added grad_y
---
src/interpolation/mo_lib_intp_rbf.cpp | 25 +++++++++++++++++++++++--
1 file changed, 23 insertions(+), 2 deletions(-)
diff --git a/src/interpolation/mo_lib_intp_rbf.cpp b/src/interpolation/mo_lib_intp_rbf.cpp
index 666e02c..0b0e213 100644
--- a/src/interpolation/mo_lib_intp_rbf.cpp
+++ b/src/interpolation/mo_lib_intp_rbf.cpp
@@ -62,8 +62,29 @@ void rbf_interpol_c2grad_lib(const T* p_cell_in, const int* rbf_c2grad_idx, cons
p_cell_in_view(rbf_c2grad_idx_view(8, jc, jb), jk, rbf_c2grad_blk_view(8, jc, jb)) +
rbf_c2grad_coeff_view(9, 1, jc, jb)*
p_cell_in_view(rbf_c2grad_idx_view(9, jc, jb), jk, rbf_c2grad_blk_view(9, jc, jb));
-
- });
+
+ grad_y_view(jc, jk, jb) =
+ rbf_c2grad_coeff_view(0, 2, jc, jb)*
+ p_cell_in_view(jc, jk, jb) +
+ rbf_c2grad_coeff_view(1, 2, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(1, jc, jb), jk, rbf_c2grad_blk_view(1, jc, jb)) +
+ rbf_c2grad_coeff_view(2, 2, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(2, jc, jb), jk, rbf_c2grad_blk_view(2, jc, jb)) +
+ rbf_c2grad_coeff_view(3, 2, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(3, jc, jb), jk, rbf_c2grad_blk_view(3, jc, jb)) +
+ rbf_c2grad_coeff_view(4, 2, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(4, jc, jb), jk, rbf_c2grad_blk_view(4, jc, jb)) +
+ rbf_c2grad_coeff_view(5, 2, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(5, jc, jb), jk, rbf_c2grad_blk_view(5, jc, jb)) +
+ rbf_c2grad_coeff_view(6, 2, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(6, jc, jb), jk, rbf_c2grad_blk_view(6, jc, jb)) +
+ rbf_c2grad_coeff_view(7, 2, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(7, jc, jb), jk, rbf_c2grad_blk_view(7, jc, jb)) +
+ rbf_c2grad_coeff_view(8, 2, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(8, jc, jb), jk, rbf_c2grad_blk_view(8, jc, jb)) +
+ rbf_c2grad_coeff_view(9, 2, jc, jb)*
+ p_cell_in_view(rbf_c2grad_idx_view(9, jc, jb), jk, rbf_c2grad_blk_view(9, jc, jb));
+ });
}//for
}//void
--
GitLab
From 54ac88915fd7476b3e3f81713b6d97cd586e144b Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante0.lvt.dkrz.de>
Date: Tue, 25 Feb 2025 10:30:28 +0100
Subject: [PATCH 08/33] Ported rbf_vec_interpol_cell_lib
---
src/interpolation/mo_lib_intp_rbf.cpp | 86 +++++++++++++++++++
src/interpolation/mo_lib_intp_rbf.hpp | 7 ++
.../mo_lib_intp_rbf_bindings.cpp | 45 +++++++---
src/interpolation/mo_lib_intp_rbf_bindings.h | 16 +++-
4 files changed, 139 insertions(+), 15 deletions(-)
diff --git a/src/interpolation/mo_lib_intp_rbf.cpp b/src/interpolation/mo_lib_intp_rbf.cpp
index 0b0e213..fa9f372 100644
--- a/src/interpolation/mo_lib_intp_rbf.cpp
+++ b/src/interpolation/mo_lib_intp_rbf.cpp
@@ -100,3 +100,89 @@ void rbf_interpol_c2grad_lib<float>(const float* p_cell_in, const int* rbf_c2gra
const float* rbf_c2grad_coeff, float* grad_x, float* grad_y,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
+
+
+//------------------------------------------rbf_vec_interpol_cell_lib---------------------------------------------
+
+template <typename T>
+void rbf_vec_interpol_cell_lib(const T* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+ const T* rbf_vec_coeff_c, T* p_u_out, T* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e, int rbf_vec_dim_c, bool lacc, bool acc_async){
+
+ typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
+ typedef Kokkos::View<T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedT3D;
+ typedef Kokkos::View<const int***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstInt3D;
+ typedef Kokkos::View<const T****, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT4D;
+
+ UnmanagedConstT3D p_vn_in_view(p_vn_in, nproma, nlev, nblks_e);
+ UnmanagedConstInt3D rbf_vec_idx_c_view(rbf_vec_idx_c, rbf_vec_dim_c, nproma, nblk_c);
+ UnmanagedConstInt3D rbf_vec_blk_c_view(rbf_vec_blk_c, rbf_vec_dim_c, nproma, nblk_c);
+ UnmanagedConstT4D rbf_vec_coeff_c_view(rbf_vec_coeff_c, nproma, nblk_c); //TODO
+ UnmanagedT3D p_u_out_view(p_u_out, nproma, nlev, nblk_c);
+ UnmanagedT3D p_v_out_view(p_u_out, nproma, nlev, nblk_c);
+
+
+ for (int jb = i_startblk; jb < i_endblk; ++jb) {
+
+ int i_startidx, i_endidx;
+ get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk, i_startidx, i_endidx);
+
+ Kokkos::parallel_for("rbf_vec_interpol_cell_lib",
+ Kokkos::MDRangePolicy<Kokkos::Rank<2>>({slev, i_startidx}, {elev, i_endidx}),
+ KOKKOS_LAMBDA(const int jk, const int jc){
+
+ p_u_out_view(jc, jk, jb) =
+ rbf_vec_coeff_c_view(0, 1, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(0, jc, jb), jk, rbf_vec_blk_c_view(0, jc, jb)) +
+ rbf_vec_coeff_c_view(1, 1, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(1, jc, jb), jk, rbf_vec_blk_c_view(1, jc, jb)) +
+ rbf_vec_coeff_c_view(2, 1, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(2, jc, jb), jk, rbf_vec_blk_c_view(2, jc, jb)) +
+ rbf_vec_coeff_c_view(3, 1, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(3, jc, jb), jk, rbf_vec_blk_c_view(3, jc, jb)) +
+ rbf_vec_coeff_c_view(4, 1, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(4, jc, jb), jk, rbf_vec_blk_c_view(4, jc, jb)) +
+ rbf_vec_coeff_c_view(5, 1, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(5, jc, jb), jk, rbf_vec_blk_c_view(5, jc, jb)) +
+ rbf_vec_coeff_c_view(6, 1, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(6, jc, jb), jk, rbf_vec_blk_c_view(6, jc, jb)) +
+ rbf_vec_coeff_c_view(7, 1, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(7, jc, jb), jk, rbf_vec_blk_c_view(7, jc, jb)) +
+ rbf_vec_coeff_c_view(8, 1, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(8, jc, jb), jk, rbf_vec_blk_c_view(8, jc, jb));
+
+ p_v_out_view(jc, jk, jb) =
+ rbf_vec_coeff_c_view(0, 2, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(0, jc, jb), jk, rbf_vec_blk_c_view(0, jc, jb)) +
+ rbf_vec_coeff_c_view(1, 2, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(1, jc, jb), jk, rbf_vec_blk_c_view(1, jc, jb)) +
+ rbf_vec_coeff_c_view(2, 2, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(2, jc, jb), jk, rbf_vec_blk_c_view(2, jc, jb)) +
+ rbf_vec_coeff_c_view(3, 2, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(3, jc, jb), jk, rbf_vec_blk_c_view(3, jc, jb)) +
+ rbf_vec_coeff_c_view(4, 2, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(4, jc, jb), jk, rbf_vec_blk_c_view(4, jc, jb)) +
+ rbf_vec_coeff_c_view(5, 2, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(5, jc, jb), jk, rbf_vec_blk_c_view(5, jc, jb)) +
+ rbf_vec_coeff_c_view(6, 2, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(6, jc, jb), jk, rbf_vec_blk_c_view(6, jc, jb)) +
+ rbf_vec_coeff_c_view(7, 2, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(7, jc, jb), jk, rbf_vec_blk_c_view(7, jc, jb)) +
+ rbf_vec_coeff_c_view(8, 2, jc, jb)*
+ p_vn_in_view(rbf_vec_idx_c_view(8, jc, jb), jk, rbf_vec_blk_c_view(8, jc, jb));
+ });
+ Kokkos::fence();
+ }//for
+}//void
+
+template
+void rbf_vec_interpol_cell_lib(const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+ const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e, int rbf_vec_dim_c, bool lacc, bool acc_async);
+template
+void rbf_vec_interpol_cell_lib(const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+ const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e, int rbf_vec_dim_c, bool lacc, bool acc_async);
diff --git a/src/interpolation/mo_lib_intp_rbf.hpp b/src/interpolation/mo_lib_intp_rbf.hpp
index 8006a43..520e9c6 100644
--- a/src/interpolation/mo_lib_intp_rbf.hpp
+++ b/src/interpolation/mo_lib_intp_rbf.hpp
@@ -19,3 +19,10 @@ void rbf_interpol_c2grad_lib(const T* p_cell_in, const int* rbf_c2grad_idx, cons
const T* rbf_c2grad_coeff, T* grad_x, T* grad_y,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
+
+
+template <typename T>
+void rbf_vec_interpol_cell_lib(const T* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+ const T* rbf_vec_coeff_c, T* p_u_out, T* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e, int rbf_vec_dim_c, bool lacc, bool acc_async);
diff --git a/src/interpolation/mo_lib_intp_rbf_bindings.cpp b/src/interpolation/mo_lib_intp_rbf_bindings.cpp
index a44a3da..b19b28c 100644
--- a/src/interpolation/mo_lib_intp_rbf_bindings.cpp
+++ b/src/interpolation/mo_lib_intp_rbf_bindings.cpp
@@ -17,27 +17,46 @@
void rbf_interpol_c2grad_lib_sp(const float* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
const float* rbf_c2grad_coeff, float* grad_x, float* grad_y,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int rbf_c2grad_dim, bool lacc){
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc){
-rbf_interpol_c2grad_lib<float>(p_cell_in, rbf_c2grad_idx, rbf_c2grad_blk,
- rbf_c2grad_coeff,
- grad_x, grad_y, i_startblk, i_endblk,
- i_startidx_in, i_endidx_in, slev, elev,
- nproma, nlev, nblk_c, rbf_c2grad_dim, lacc);
+ rbf_interpol_c2grad_lib<float>(p_cell_in, rbf_c2grad_idx, rbf_c2grad_blk,
+ rbf_c2grad_coeff,
+ grad_x, grad_y, i_startblk, i_endblk,
+ i_startidx_in, i_endidx_in, slev, elev,
+ nproma, rbf_c2grad_dim, nlev, nblk_c, lacc);
}
-
void rbf_interpol_c2grad_lib_dp(const double* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
const double* rbf_c2grad_coeff, double* grad_x, double* grad_y,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int rbf_c2grad_dim, bool lacc){
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc){
+
+
+ rbf_interpol_c2grad_lib<double>(p_cell_in, rbf_c2grad_idx, rbf_c2grad_blk,
+ rbf_c2grad_coeff,
+ grad_x, grad_y, i_startblk, i_endblk,
+ i_startidx_in, i_endidx_in, slev, elev,
+ nproma, rbf_c2grad_dim, nlev, nblk_c, lacc);
+}
+void rbf_vec_interpol_cell_lib_sp(const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+ const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int nblk_e, int rbf_vec_dim_c, bool lacc, bool acc_async){
-rbf_interpol_c2grad_lib<double>(p_cell_in, rbf_c2grad_idx, rbf_c2grad_blk,
- rbf_c2grad_coeff,
- grad_x, grad_y, i_startblk, i_endblk,
- i_startidx_in, i_endidx_in, slev, elev,
- nproma, nlev, nblk_c, rbf_c2grad_dim, lacc);
+ rbf_vec_interpol_cell_lib<float>(p_vn_in, rbf_vec_idx_c, rbf_vec_blk_c,
+ rbf_vec_coeff_c, p_u_out, p_v_out,
+ i_startblk, i_endblk, i_startidx_in, i_endidx_in,
+ slev, elev, nproma, nlev, nblk_c, nblk_e, rbf_vec_dim_c, lacc, acc_async);
}
+void rbf_vec_interpol_cell_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+ const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int nblk_e, int rbf_vec_dim_c, bool lacc, bool acc_async){
+
+ rbf_vec_interpol_cell_lib<double>(p_vn_in, rbf_vec_idx_c, rbf_vec_blk_c, rbf_vec_coeff_c,
+ p_u_out, p_v_out, i_startblk, i_endblk, i_startidx_in, i_endidx_in,
+ slev, elev, nproma, nlev, nblk_c, nblk_e, rbf_vec_dim_c, lacc, acc_async);
+}
diff --git a/src/interpolation/mo_lib_intp_rbf_bindings.h b/src/interpolation/mo_lib_intp_rbf_bindings.h
index e7ef8f3..a9775fa 100644
--- a/src/interpolation/mo_lib_intp_rbf_bindings.h
+++ b/src/interpolation/mo_lib_intp_rbf_bindings.h
@@ -15,10 +15,22 @@ extern "C" {
void rbf_interpol_c2grad_lib_sp(const float* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
const float* rbf_c2grad_coeff, float* grad_x, const float* grad_y,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int rbf_c2grad_dim, bool lacc);
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
void rbf_interpol_c2grad_lib_dp(const double* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
const double* rbf_c2grad_coeff, double* grad_x, const double* grad_y,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int rbf_c2grad_dim, bool lacc);
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
+
+void rbf_vec_interpol_cell_lib_sp(const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+ const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int nblk_e, int rbf_vec_dim_c, bool lacc, bool acc_async);
+
+void rbf_vec_interpol_cell_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+ const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int nblk_e, int rbf_vec_dim_c, bool lacc, bool acc_async);
+
+
}
--
GitLab
From eac9e262eb66b25397b283b7e5e3efa5682d3a1c Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante0.lvt.dkrz.de>
Date: Tue, 25 Feb 2025 11:32:42 +0100
Subject: [PATCH 09/33] Ported rbf_vec_interpol_edge_lib
---
src/interpolation/mo_lib_intp_rbf.cpp | 59 ++++++++++++++++++-
src/interpolation/mo_lib_intp_rbf.hpp | 7 +++
.../mo_lib_intp_rbf_bindings.cpp | 25 ++++++++
src/interpolation/mo_lib_intp_rbf_bindings.h | 12 ++++
4 files changed, 101 insertions(+), 2 deletions(-)
diff --git a/src/interpolation/mo_lib_intp_rbf.cpp b/src/interpolation/mo_lib_intp_rbf.cpp
index fa9f372..f125f1f 100644
--- a/src/interpolation/mo_lib_intp_rbf.cpp
+++ b/src/interpolation/mo_lib_intp_rbf.cpp
@@ -177,12 +177,67 @@ void rbf_vec_interpol_cell_lib(const T* p_vn_in, const int* rbf_vec_idx_c, const
}//void
template
-void rbf_vec_interpol_cell_lib(const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+void rbf_vec_interpol_cell_lib<double>(const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e, int rbf_vec_dim_c, bool lacc, bool acc_async);
template
-void rbf_vec_interpol_cell_lib(const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+void rbf_vec_interpol_cell_lib<float>(const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e, int rbf_vec_dim_c, bool lacc, bool acc_async);
+
+//------------------------------------------rbf_vec_interpol_edge_lib---------------------------------------------
+
+template <typename T>
+void rbf_vec_interpol_edge_lib(const T* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
+ const T* rbf_vec_coeff_e, T* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async){
+
+ typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
+ typedef Kokkos::View<T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedT3D;
+ typedef Kokkos::View<const int***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstInt3D;
+
+ UnmanagedConstT3D p_vn_in_view(p_vn_in, nproma,nlev,nblks_e);
+ UnmanagedConstInt3D rbf_vec_idx_e_view(rbf_vec_idx_e, rbf_vec_dim_e,nproma,nblks_e);
+ UnmanagedConstInt3D rbf_vec_blk_e_view(rbf_vec_blk_e, rbf_vec_dim_e,nproma,nblks_e);
+ UnmanagedConstT3D rbf_vec_coeff_e_view(rbf_vec_coeff_e, rbf_vec_dim_e,nproma,nblks_e);
+ UnmanagedT3D p_vt_out_view(p_vt_out, nproma,nlev,nblks_e);
+
+
+ for (int jb = i_startblk; jb < i_endblk; ++jb) {
+
+ int i_startidx, i_endidx;
+ get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk, i_startidx, i_endidx);
+
+ Kokkos::parallel_for("rbf_vec_interpol_edge_lib",
+ Kokkos::MDRangePolicy<Kokkos::Rank<2>>({slev, i_startidx}, {elev, i_endidx}),
+ KOKKOS_LAMBDA(const int jk, const int je){
+
+ p_vt_out_view(je, jk, jb) =
+ rbf_vec_coeff_e_view(1, je, jb) * p_vn_in_view(rbf_vec_idx_e_view(1, je, jb), jk, rbf_vec_blk_e_view(1, je, jb)) +
+ rbf_vec_coeff_e_view(2, je, jb) * p_vn_in_view(rbf_vec_idx_e_view(2, je, jb), jk, rbf_vec_blk_e_view(2, je, jb)) +
+ rbf_vec_coeff_e_view(3, je, jb) * p_vn_in_view(rbf_vec_idx_e_view(3, je, jb), jk, rbf_vec_blk_e_view(3, je, jb)) +
+ rbf_vec_coeff_e_view(4, je, jb) * p_vn_in_view(rbf_vec_idx_e_view(4, je, jb), jk, rbf_vec_blk_e_view(4, je, jb));
+
+ });
+ }
+}
+
+template
+void rbf_vec_interpol_edge_lib<double>(const double* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
+ const double* rbf_vec_coeff_e, double* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async);
+
+
+template
+void rbf_vec_interpol_edge_lib<float>(const float* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
+ const float* rbf_vec_coeff_e, float* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async);
+
diff --git a/src/interpolation/mo_lib_intp_rbf.hpp b/src/interpolation/mo_lib_intp_rbf.hpp
index 520e9c6..e52989d 100644
--- a/src/interpolation/mo_lib_intp_rbf.hpp
+++ b/src/interpolation/mo_lib_intp_rbf.hpp
@@ -26,3 +26,10 @@ void rbf_vec_interpol_cell_lib(const T* p_vn_in, const int* rbf_vec_idx_c, const
const T* rbf_vec_coeff_c, T* p_u_out, T* p_v_out,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e, int rbf_vec_dim_c, bool lacc, bool acc_async);
+
+template <typename T>
+void rbf_vec_interpol_edge_lib(const T* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
+ const T* rbf_vec_coeff_e, T* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_asynci);
diff --git a/src/interpolation/mo_lib_intp_rbf_bindings.cpp b/src/interpolation/mo_lib_intp_rbf_bindings.cpp
index b19b28c..5c7943b 100644
--- a/src/interpolation/mo_lib_intp_rbf_bindings.cpp
+++ b/src/interpolation/mo_lib_intp_rbf_bindings.cpp
@@ -60,3 +60,28 @@ void rbf_vec_interpol_cell_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_
p_u_out, p_v_out, i_startblk, i_endblk, i_startidx_in, i_endidx_in,
slev, elev, nproma, nlev, nblk_c, nblk_e, rbf_vec_dim_c, lacc, acc_async);
}
+
+void rbf_vec_interpol_edge_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
+ const double* rbf_vec_coeff_e, double* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async){
+
+
+ rbf_vec_interpol_edge_lib<double>(p_vn_in, rbf_vec_idx_e, rbf_vec_blk_e,
+ rbf_vec_coeff_e, p_vt_out,
+ i_startblk, i_endblk, i_startidx_in,i_endidx_in,
+ slev, elev, nlev, nproma, rbf_vec_dim_e, nblks_e, lacc, acc_async);
+}
+
+void rbf_vec_interpol_edge_lib_sp(const float* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
+ const float* rbf_vec_coeff_e, float* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async){
+
+ rbf_vec_interpol_edge_lib<float>(p_vn_in, rbf_vec_idx_e, rbf_vec_blk_e,
+ rbf_vec_coeff_e, p_vt_out,
+ i_startblk, i_endblk, i_startidx_in, i_endidx_in,
+ slev, elev, nlev, nproma, rbf_vec_dim_e, nblks_e,lacc, acc_async);
+}
diff --git a/src/interpolation/mo_lib_intp_rbf_bindings.h b/src/interpolation/mo_lib_intp_rbf_bindings.h
index a9775fa..9620614 100644
--- a/src/interpolation/mo_lib_intp_rbf_bindings.h
+++ b/src/interpolation/mo_lib_intp_rbf_bindings.h
@@ -32,5 +32,17 @@ void rbf_vec_interpol_cell_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
int slev, int elev, int nproma, int nlev, int nblk_c, int nblk_e, int rbf_vec_dim_c, bool lacc, bool acc_async);
+void rbf_vec_interpol_edge_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
+ const double* rbf_vec_coeff_e, double* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async);
+
+void rbf_vec_interpol_edge_lib_sp(const float* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
+ const float* rbf_vec_coeff_e, float* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async);
+
}
--
GitLab
From 04a3601466fe73b9d396e60bdbd990d8c252e04a Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante0.lvt.dkrz.de>
Date: Tue, 25 Feb 2025 11:56:15 +0100
Subject: [PATCH 10/33] Fixed index
---
src/interpolation/mo_lib_intp_rbf.cpp | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/src/interpolation/mo_lib_intp_rbf.cpp b/src/interpolation/mo_lib_intp_rbf.cpp
index f125f1f..a7d9cbd 100644
--- a/src/interpolation/mo_lib_intp_rbf.cpp
+++ b/src/interpolation/mo_lib_intp_rbf.cpp
@@ -217,10 +217,10 @@ void rbf_vec_interpol_edge_lib(const T* p_vn_in, const int* rbf_vec_idx_e, const
KOKKOS_LAMBDA(const int jk, const int je){
p_vt_out_view(je, jk, jb) =
+ rbf_vec_coeff_e_view(0, je, jb) * p_vn_in_view(rbf_vec_idx_e_view(0, je, jb), jk, rbf_vec_blk_e_view(0, je, jb)) +
rbf_vec_coeff_e_view(1, je, jb) * p_vn_in_view(rbf_vec_idx_e_view(1, je, jb), jk, rbf_vec_blk_e_view(1, je, jb)) +
rbf_vec_coeff_e_view(2, je, jb) * p_vn_in_view(rbf_vec_idx_e_view(2, je, jb), jk, rbf_vec_blk_e_view(2, je, jb)) +
- rbf_vec_coeff_e_view(3, je, jb) * p_vn_in_view(rbf_vec_idx_e_view(3, je, jb), jk, rbf_vec_blk_e_view(3, je, jb)) +
- rbf_vec_coeff_e_view(4, je, jb) * p_vn_in_view(rbf_vec_idx_e_view(4, je, jb), jk, rbf_vec_blk_e_view(4, je, jb));
+ rbf_vec_coeff_e_view(3, je, jb) * p_vn_in_view(rbf_vec_idx_e_view(3, je, jb), jk, rbf_vec_blk_e_view(3, je, jb));
});
}
--
GitLab
From f54eaa77aed000133b4c8c689dc20b49c37a6291 Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante0.lvt.dkrz.de>
Date: Tue, 25 Feb 2025 16:03:59 +0100
Subject: [PATCH 11/33] ported mo_lib_laplace
---
src/horizontal/CMakeLists.txt | 4 +-
src/horizontal/mo_lib_laplace.cpp | 104 +++++++++++++++++++++
src/horizontal/mo_lib_laplace.hpp | 24 +++++
src/horizontal/mo_lib_laplace_bindings.cpp | 50 ++++++++++
src/horizontal/mo_lib_laplace_bindings.h | 32 +++++++
5 files changed, 213 insertions(+), 1 deletion(-)
create mode 100644 src/horizontal/mo_lib_laplace.cpp
create mode 100644 src/horizontal/mo_lib_laplace.hpp
create mode 100644 src/horizontal/mo_lib_laplace_bindings.cpp
create mode 100644 src/horizontal/mo_lib_laplace_bindings.h
diff --git a/src/horizontal/CMakeLists.txt b/src/horizontal/CMakeLists.txt
index 078a14d..44f9e44 100644
--- a/src/horizontal/CMakeLists.txt
+++ b/src/horizontal/CMakeLists.txt
@@ -13,7 +13,9 @@ add_library(
iconmath-horizontal
mo_lib_divrot.F90
mo_lib_laplace.F90
- mo_lib_gradients.F90)
+ mo_lib_gradients.F90
+ mo_lib_laplace_bindings.cpp
+ mo_lib_laplace.cpp)
add_library(${PROJECT_NAME}::horizontal ALIAS iconmath-horizontal)
diff --git a/src/horizontal/mo_lib_laplace.cpp b/src/horizontal/mo_lib_laplace.cpp
new file mode 100644
index 0000000..48c4479
--- /dev/null
+++ b/src/horizontal/mo_lib_laplace.cpp
@@ -0,0 +1,104 @@
+// ICON
+//
+// ---------------------------------------------------------------
+// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
+// Contact information: icon-model.org
+//
+// See AUTHORS.TXT for a list of authors
+// See LICENSES/ for license information
+// SPDX-License-Identifier: BSD-3-Clause
+// ---------------------------------------------------------------
+
+#include "mo_lib_laplace.hpp"
+//#include "mo_lib_gradients.hpp"
+#include <Kokkos_Core.hpp>
+#include <iostream>
+
+template<typename T>
+void nabla2_scalar_lib(const T* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
+ const int* edge_cell_idx, const int* edge_cell_blk, const T* inv_dual_edge_length,
+ const int* cell_edge_idx, const int* cell_edge_blk,
+ const T* geofac_n2s, const T* geofac_div, T* nabla2_psi_c,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
+ int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc){
+
+ typedef Kokkos::View<const T**, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT2D;
+ typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
+ typedef Kokkos::View<T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedT3D;
+ typedef Kokkos::View<const int***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstInt3D;
+
+
+ UnmanagedConstT3D psi_c_view(psi_c, nproma,nlev,nblks_c);
+ UnmanagedConstInt3D cell_neighbor_idx_view(cell_neighbor_idx, nproma,nblks_c,3);
+ UnmanagedConstInt3D cell_neighbor_blk_view(cell_neighbor_blk, nproma,nblks_c,3);
+ UnmanagedConstInt3D edge_cell_idx_view(edge_cell_idx, nproma,nblks_e,2);
+ UnmanagedConstInt3D edge_cell_blk_view(edge_cell_blk, nproma,nblks_e,2);
+ UnmanagedConstT2D inv_dual_edge_length_view(inv_dual_edge_length, nproma,nblks_e);
+ UnmanagedConstInt3D cell_edge_idx_view(cell_edge_idx, nproma,nblks_c,3);
+ UnmanagedConstInt3D cell_edge_blk_view(cell_edge_blk, nproma,nblks_c,3);
+ UnmanagedConstT3D geofac_n2s_view(geofac_n2s, nproma,cell_type+1,nblks_c);
+ UnmanagedConstT3D geofac_div_view(geofac_div, nproma,cell_type,nblks_c);
+ UnmanagedT3D nabla2_psi_c_view(nabla2_psi_c, nproma,nlev,nblks_c);
+
+
+ switch (cell_type){
+
+ case 3:
+ for (int jb = i_startblk; jb < i_endblk; ++jb) {
+
+ int i_startidx, i_endidx;
+ get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk, i_startidx, i_endidx);
+
+
+ Kokkos::parallel_for("rbf_interpol_c2grad",
+ Kokkos::MDRangePolicy<Kokkos::Rank<2>>({slev, i_startidx}, {elev, i_endidx}),
+ KOKKOS_LAMBDA(const int jk, const int jc){
+
+ nabla2_psi_c_view(jc, jk, jb) =
+ psi_c_view(jc, jk, jb)*geofac_n2s_view(jc, 1, jb) +
+ psi_c_view(cell_neighbor_idx_view(jc, jb, 1), jk, cell_neighbor_blk_view(jc, jb, 1)) *
+ geofac_n2s_view(jc, 2, jb) +
+ psi_c_view(cell_neighbor_idx_view(jc, jb, 2), jk, cell_neighbor_blk_view(jc, jb, 2)) *
+ geofac_n2s_view(jc, 3, jb) +
+ psi_c_view(cell_neighbor_idx_view(jc, jb, 3), jk, cell_neighbor_blk_view(jc, jb, 3)) *
+ geofac_n2s_view(jc, 4, jb);
+ });
+ }
+ break;
+
+
+ case 6:
+/* TODO
+ grad_fd_norm_lib(psi_c, edge_cell_idx, edge_cell_blk,
+ inv_dual_edge_length, z_grad_fd_norm_e,&
+ i_startblk_e, i_endblk_e, i_startidx_e, i_endidx_e,
+ slev, elev, nproma);
+
+ div_lib(z_grad_fd_norm_e, cell_edge_idx, cell_edge_blk,
+ geofac_div, nabla2_psi_c, i_startblk, i_endblk,
+ i_startidx_in, i_endidx_in, slev, elev, nproma);
+*/
+ default:
+ std::cout << "Unknown value for cell_type\n";
+
+ }//switch
+}//void
+
+template
+void nabla2_scalar_lib<double>(const double* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
+ const int* edge_cell_idx, const int* edge_cell_blk, const double* inv_dual_edge_length,
+ const int* cell_edge_idx, const int* cell_edge_blk,
+ const double* geofac_n2s, const double* geofac_div, double* nabla2_psi_c,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
+ int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc);
+
+template
+void nabla2_scalar_lib<float>(const float* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
+ const int* edge_cell_idx, const int* edge_cell_blk, const float* inv_dual_edge_length,
+ const int* cell_edge_idx, const int* cell_edge_blk,
+ const float* geofac_n2s, const float* geofac_div, float* nabla2_psi_c,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
+ int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc);
diff --git a/src/horizontal/mo_lib_laplace.hpp b/src/horizontal/mo_lib_laplace.hpp
new file mode 100644
index 0000000..2a6663c
--- /dev/null
+++ b/src/horizontal/mo_lib_laplace.hpp
@@ -0,0 +1,24 @@
+// ICON
+//
+// ---------------------------------------------------------------
+// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
+// Contact information: icon-model.org
+//
+// See AUTHORS.TXT for a list of authors
+// See LICENSES/ for license information
+// SPDX-License-Identifier: BSD-3-Clause
+// ---------------------------------------------------------------
+#pragma once
+
+#include "mo_lib_loopindices.hpp"
+#include <Kokkos_Core.hpp>
+#include <vector>
+
+template<typename T>
+void nabla2_scalar_lib(const T* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
+ const int* dge_cell_idx, const int* edge_cell_blk, const T* inv_dual_edge_length,
+ const int* cell_edge_idx, const int* cell_edge_blk,
+ const T* geofac_n2s, const T* geofac_div, const T* nabla2_psi_c,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
+ int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc);
diff --git a/src/horizontal/mo_lib_laplace_bindings.cpp b/src/horizontal/mo_lib_laplace_bindings.cpp
new file mode 100644
index 0000000..43a19b3
--- /dev/null
+++ b/src/horizontal/mo_lib_laplace_bindings.cpp
@@ -0,0 +1,50 @@
+// ICON
+//
+// ---------------------------------------------------------------
+// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
+// Contact information: icon-model.org
+//
+// See AUTHORS.TXT for a list of authors
+// See LICENSES/ for license information
+// SPDX-License-Identifier: BSD-3-Clause
+// ---------------------------------------------------------------
+
+
+#include "mo_lib_laplace_bindings.h"
+#include "mo_lib_laplace.hpp"
+
+
+void nabla2_scalar_lib_dp(const double* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
+ const int* edge_cell_idx, const int* edge_cell_blk, const double* inv_dual_edge_length,
+ const int* cell_edge_idx, const int* cell_edge_blk,
+ const double* geofac_n2s, const double* geofac_div, const double* nabla2_psi_c,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
+ int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc){
+
+ nabla2_scalar_lib<double>(psi_c, cell_neighbor_idx, cell_neighbor_blk,
+ edge_cell_idx, edge_cell_blk, inv_dual_edge_length,
+ cell_edge_idx, cell_edge_blk,
+ geofac_n2s, geofac_div, nabla2_psi_c,
+ i_startblk, i_endblk, i_startidx_in, i_endidx_in,
+ i_startblk_e, i_endblk_e, i_startidx_e, i_endidx_e,
+ nlev, slev, elev, nproma, nblks_e, nblks_c, cell_type, lacc);
+}
+
+
+void nabla2_scalar_lib_sp(const float* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
+ const int* edge_cell_idx, const int* edge_cell_blk, const float* inv_dual_edge_length,
+ const int* cell_edge_idx, const int* cell_edge_blk,
+ const float* geofac_n2s, const float* geofac_div, const float* nabla2_psi_c,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
+ int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc){
+
+ nabla2_scalar_lib<float>(psi_c, cell_neighbor_idx, cell_neighbor_blk,
+ edge_cell_idx, edge_cell_blk, inv_dual_edge_length,
+ cell_edge_idx, cell_edge_blk,
+ geofac_n2s, geofac_div, nabla2_psi_c,
+ i_startblk, i_endblk, i_startidx_in, i_endidx_in,
+ i_startblk_e, i_endblk_e, i_startidx_e, i_endidx_e,
+ nlev, slev, elev, nproma, nblks_e, nblks_c, cell_type, lacc);
+}
diff --git a/src/horizontal/mo_lib_laplace_bindings.h b/src/horizontal/mo_lib_laplace_bindings.h
new file mode 100644
index 0000000..96bdc64
--- /dev/null
+++ b/src/horizontal/mo_lib_laplace_bindings.h
@@ -0,0 +1,32 @@
+// ICON
+//
+// ---------------------------------------------------------------
+// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
+// Contact information: icon-model.org
+//
+// See AUTHORS.TXT for a list of authors
+// See LICENSES/ for license information
+// SPDX-License-Identifier: BSD-3-Clause
+// ---------------------------------------------------------------
+#pragma once
+
+
+
+extern "C"{
+
+void nabla2_scalar_lib_dp(const double* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
+ const int* edge_cell_idx, const int* edge_cell_blk, const double* inv_dual_edge_length,
+ const int* cell_edge_idx, const int* cell_edge_blk,
+ const double* geofac_n2s, const double* geofac_div, const double* nabla2_psi_c,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
+ int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc);
+
+void nabla2_scalar_lib_sp(const float* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
+ const int* edge_cell_idx, const int* edge_cell_blk, const float* inv_dual_edge_length,
+ const int* cell_edge_idx, const int* cell_edge_blk,
+ const float* geofac_n2s, const float* geofac_div, const float* nabla2_psi_c,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
+ int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc);
+}
--
GitLab
From bd8a950b6fa2be57b0068ddd6744d28d6fabcaa0 Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante0.lvt.dkrz.de>
Date: Tue, 25 Feb 2025 16:28:59 +0100
Subject: [PATCH 12/33] Fixed index for nabla2_scalar_lib
---
src/horizontal/mo_lib_laplace.cpp | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/src/horizontal/mo_lib_laplace.cpp b/src/horizontal/mo_lib_laplace.cpp
index 48c4479..203b3b7 100644
--- a/src/horizontal/mo_lib_laplace.cpp
+++ b/src/horizontal/mo_lib_laplace.cpp
@@ -56,13 +56,13 @@ void nabla2_scalar_lib(const T* psi_c, const int* cell_neighbor_idx, const int*
KOKKOS_LAMBDA(const int jk, const int jc){
nabla2_psi_c_view(jc, jk, jb) =
- psi_c_view(jc, jk, jb)*geofac_n2s_view(jc, 1, jb) +
+ psi_c_view(jc, jk, jb)*geofac_n2s_view(jc, 0, jb) +
+ psi_c_view(cell_neighbor_idx_view(jc, jb, 0), jk, cell_neighbor_blk_view(jc, jb, 0)) *
+ geofac_n2s_view(jc, 1, jb) +
psi_c_view(cell_neighbor_idx_view(jc, jb, 1), jk, cell_neighbor_blk_view(jc, jb, 1)) *
geofac_n2s_view(jc, 2, jb) +
psi_c_view(cell_neighbor_idx_view(jc, jb, 2), jk, cell_neighbor_blk_view(jc, jb, 2)) *
- geofac_n2s_view(jc, 3, jb) +
- psi_c_view(cell_neighbor_idx_view(jc, jb, 3), jk, cell_neighbor_blk_view(jc, jb, 3)) *
- geofac_n2s_view(jc, 4, jb);
+ geofac_n2s_view(jc, 3, jb);
});
}
break;
--
GitLab
From 18b653dbc456a0061436df5e7a091b20a64592cd Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante4.lvt.dkrz.de>
Date: Wed, 26 Feb 2025 10:04:06 +0100
Subject: [PATCH 13/33] WIP: MR seperation.
---
src/horizontal/mo_lib_laplace.cpp | 118 ++++++++++++++++++++++++++++++
src/horizontal/mo_lib_laplace.hpp | 7 ++
2 files changed, 125 insertions(+)
diff --git a/src/horizontal/mo_lib_laplace.cpp b/src/horizontal/mo_lib_laplace.cpp
index 203b3b7..57ad292 100644
--- a/src/horizontal/mo_lib_laplace.cpp
+++ b/src/horizontal/mo_lib_laplace.cpp
@@ -11,8 +11,10 @@
#include "mo_lib_laplace.hpp"
//#include "mo_lib_gradients.hpp"
+//#inlcude "mo_fortran_tools.hpp"
#include <Kokkos_Core.hpp>
#include <iostream>
+#include <utility>
template<typename T>
void nabla2_scalar_lib(const T* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
@@ -102,3 +104,119 @@ void nabla2_scalar_lib<float>(const float* psi_c, const int* cell_neighbor_idx,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc);
+
+
+//--------------------------------nabla2_scalar_avg_lib-------------------------------------
+
+
+template<typename T>
+void nabla2_scalar_avg_lib(const T* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
+ const T* geofac_n2s, const T* avg_coeff, T* nabla2_psi_c,
+ int i_startblk_in, int i_endblk_in, int i_startidx_in, int i_endidx_in,
+ int nblks_c, int cell_type, int patch_id,
+ int nlev, int slev, int elev, int nproma, bool l_limited_area, bool lacc){
+
+ typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
+ typedef Kokkos::View<T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedT3D;
+ typedef Kokkos::View<const int***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstInt3D;
+
+ UnmanagedConstT3D psi_c_view(psi_c, nproma,nlev,nblks_c);
+ UnmanagedConstInt3D cell_neighbor_idx_view(cell_neighbor_idx, nproma,nblks_c,3);
+ UnmanagedConstInt3D cell_neighbor_blk_view(cell_neighbor_blk, nproma,nblks_c,3);
+ UnmanagedConstT3D geofac_n2s_view(geofac_n2s, nproma,cell_type+1,nblks_c);
+ UnmanagedConstT3D avg_coeff_view(avg_coeff, nproma,nlev,nblks_c);
+ UnmanagedT3D nabla2_psi_c_view(nabla2_psi_c, nproma,nlev,nblks_c);
+
+ int aux_c
+ UnmanagedT3D aux_c_view(aux_c, nproma, nlev, nblks_c); //local
+
+
+
+ switch (cell_type){
+
+ case 3:
+
+ if(slev == elev){
+
+ jk = slev;
+ i_startblk = i_startblk_in[1];
+ i_endblk = i_endblk_in[1];
+
+ for(int jb = i_startblk; jb < i_endblk; ++jb) {
+
+ int i_startidx, i_endidx;
+ get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk, i_startidx, i_endidx);
+
+ Kokkos::parallel_for("aux_c",
+ Kokkos::RangePolicy<int>(i_startidx, i_endidx),
+ KOKKOS_LAMBDA(const int jc) {
+
+ aux_c_view(jc, jk, jb) =
+ psi_c_view(jc, jk, jb) * geofac_n2s_view(jc, 0, jb) +
+ psi_c_view(cell_neighbor_idx_view(jc, jb, 0), jk, cell_neighbor_blk_view(jc, jb, 0)) *
+ geofac_n2s_view(jc, 1, jb) +
+ psi_c_view(cell_neighbor_idx_view(jc, jb, 1), jk, cell_neighbor_blk_view(jc, jb, 1)) *
+ geofac_n2s_view(jc, 2, jb) +
+ psi_c_view(cell_neighbor_idx_view(jc, jb, 2), jk, cell_neighbor_blk_view(jc, jb, 2)) *
+ geofac_n2s_view(jc, 3, jb);
+
+ });
+ }
+
+ if (l_limited_area || patch_id > 1){
+
+ i_startblk = i_startblk_in[2];
+ i_endblk = i_endblk_in[2];
+
+ /*TODO
+ gradient( Kokkos::subview(aux_c_view, Kokkos::ALL(), jk, std::make_pair(i_startblk, i_endblk + 1)),
+ Kokkos::subview(nabla2_psi_c_view, Kokkos::ALL(), jk, std::make_pair(i_startblk, i_endblk + 1)), lzacc);
+ */
+ }
+
+ i_startblk = i_startblk_in[3];
+ i_endblk = i_endblk_in[3];
+
+ for(int jb = i_startblk; jb < i_endblk; ++jb) {
+
+ int i_startidx, i_endidx;
+ get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk, i_startidx, i_endidx);
+
+ Kokkos::parallel_for("DivGrad",
+ Kokkos::RangePolicy<int>(i_startidx, i_endidx),
+ KOKKOS_LAMBDA(const int jc) {
+
+ nabla2_psi_c_view(jc, jk, jb) =
+ aux_c_view(jc, jk, jb) * avg_coeff_view(jc, 0, jb) +
+ aux_c_view(cell_neighbor_idx_view(jc, jb, 0), jk, cell_neighbor_blk_view(jc, jb, 0)) *
+ avg_coeff_view(jc, 1, jb) +
+ aux_c_view(cell_neighbor_idx_view(jc, jb, 1), jk, cell_neighbor_blk_view(jc, jb, 1)) *
+ avg_coeff_view(jc, 2, jb) +
+ aux_c_view(cell_neighbor_idx_view(jc, jb, 2), jk, cell_neighbor_blk_view(jc, jb, 2)) *
+ avg_coeff_view(jc, 3, jb);
+
+ });
+ }
+ }//if
+ break;
+
+ default:
+ std::cout << "Default case.\n";
+
+}//switch
+}//void
+
+
+template
+void nabla2_scalar_avg_lib<double>(const double* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
+ const double* geofac_n2s, const double* avg_coeff, double* nabla2_psi_c,
+ int i_startblk_in, int i_endblk_in, int i_startidx_in, int i_endidx_in,
+ int nblks_c, int cell_type, int patch_id,
+ int nlev, int slev, int elev, int nproma, bool l_limited_area, bool lacc);
+
+template
+void nabla2_scalar_avg_lib<float>(const float* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
+ const float* geofac_n2s, const float* avg_coeff, float* nabla2_psi_c,
+ int i_startblk_in, int i_endblk_in, int i_startidx_in, int i_endidx_in,
+ int nblks_c, int cell_type, int patch_id,
+ int nlev, int slev, int elev, int nproma, bool l_limited_area, bool lacc);
diff --git a/src/horizontal/mo_lib_laplace.hpp b/src/horizontal/mo_lib_laplace.hpp
index 2a6663c..0a67f85 100644
--- a/src/horizontal/mo_lib_laplace.hpp
+++ b/src/horizontal/mo_lib_laplace.hpp
@@ -22,3 +22,10 @@ void nabla2_scalar_lib(const T* psi_c, const int* cell_neighbor_idx, const int*
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc);
+template<typename T>
+void nabla2_scalar_avg_lib(const T* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
+ const T* geofac_n2s, const T* avg_coeff, T* nabla2_psi_c,
+ int i_startblk_in, int i_endblk_in, int i_startidx_in, int i_endidx_in,
+ int nblks_c, int cell_type, int patch_id,
+ int nlev, int slev, int elev, int nproma, bool l_limited_area, bool lacc);
+
--
GitLab
From 4f0ea43895255e35edb7483f9187f78089403674 Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante4.lvt.dkrz.de>
Date: Wed, 26 Feb 2025 10:10:02 +0100
Subject: [PATCH 14/33] Revert "WIP: MR seperation."
This reverts commit fed348a36d7907ea9b24e154906ddf8b47ff489a.
---
src/horizontal/mo_lib_laplace.cpp | 118 ------------------------------
src/horizontal/mo_lib_laplace.hpp | 7 --
2 files changed, 125 deletions(-)
diff --git a/src/horizontal/mo_lib_laplace.cpp b/src/horizontal/mo_lib_laplace.cpp
index 57ad292..203b3b7 100644
--- a/src/horizontal/mo_lib_laplace.cpp
+++ b/src/horizontal/mo_lib_laplace.cpp
@@ -11,10 +11,8 @@
#include "mo_lib_laplace.hpp"
//#include "mo_lib_gradients.hpp"
-//#inlcude "mo_fortran_tools.hpp"
#include <Kokkos_Core.hpp>
#include <iostream>
-#include <utility>
template<typename T>
void nabla2_scalar_lib(const T* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
@@ -104,119 +102,3 @@ void nabla2_scalar_lib<float>(const float* psi_c, const int* cell_neighbor_idx,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc);
-
-
-//--------------------------------nabla2_scalar_avg_lib-------------------------------------
-
-
-template<typename T>
-void nabla2_scalar_avg_lib(const T* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
- const T* geofac_n2s, const T* avg_coeff, T* nabla2_psi_c,
- int i_startblk_in, int i_endblk_in, int i_startidx_in, int i_endidx_in,
- int nblks_c, int cell_type, int patch_id,
- int nlev, int slev, int elev, int nproma, bool l_limited_area, bool lacc){
-
- typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
- typedef Kokkos::View<T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedT3D;
- typedef Kokkos::View<const int***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstInt3D;
-
- UnmanagedConstT3D psi_c_view(psi_c, nproma,nlev,nblks_c);
- UnmanagedConstInt3D cell_neighbor_idx_view(cell_neighbor_idx, nproma,nblks_c,3);
- UnmanagedConstInt3D cell_neighbor_blk_view(cell_neighbor_blk, nproma,nblks_c,3);
- UnmanagedConstT3D geofac_n2s_view(geofac_n2s, nproma,cell_type+1,nblks_c);
- UnmanagedConstT3D avg_coeff_view(avg_coeff, nproma,nlev,nblks_c);
- UnmanagedT3D nabla2_psi_c_view(nabla2_psi_c, nproma,nlev,nblks_c);
-
- int aux_c
- UnmanagedT3D aux_c_view(aux_c, nproma, nlev, nblks_c); //local
-
-
-
- switch (cell_type){
-
- case 3:
-
- if(slev == elev){
-
- jk = slev;
- i_startblk = i_startblk_in[1];
- i_endblk = i_endblk_in[1];
-
- for(int jb = i_startblk; jb < i_endblk; ++jb) {
-
- int i_startidx, i_endidx;
- get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk, i_startidx, i_endidx);
-
- Kokkos::parallel_for("aux_c",
- Kokkos::RangePolicy<int>(i_startidx, i_endidx),
- KOKKOS_LAMBDA(const int jc) {
-
- aux_c_view(jc, jk, jb) =
- psi_c_view(jc, jk, jb) * geofac_n2s_view(jc, 0, jb) +
- psi_c_view(cell_neighbor_idx_view(jc, jb, 0), jk, cell_neighbor_blk_view(jc, jb, 0)) *
- geofac_n2s_view(jc, 1, jb) +
- psi_c_view(cell_neighbor_idx_view(jc, jb, 1), jk, cell_neighbor_blk_view(jc, jb, 1)) *
- geofac_n2s_view(jc, 2, jb) +
- psi_c_view(cell_neighbor_idx_view(jc, jb, 2), jk, cell_neighbor_blk_view(jc, jb, 2)) *
- geofac_n2s_view(jc, 3, jb);
-
- });
- }
-
- if (l_limited_area || patch_id > 1){
-
- i_startblk = i_startblk_in[2];
- i_endblk = i_endblk_in[2];
-
- /*TODO
- gradient( Kokkos::subview(aux_c_view, Kokkos::ALL(), jk, std::make_pair(i_startblk, i_endblk + 1)),
- Kokkos::subview(nabla2_psi_c_view, Kokkos::ALL(), jk, std::make_pair(i_startblk, i_endblk + 1)), lzacc);
- */
- }
-
- i_startblk = i_startblk_in[3];
- i_endblk = i_endblk_in[3];
-
- for(int jb = i_startblk; jb < i_endblk; ++jb) {
-
- int i_startidx, i_endidx;
- get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk, i_startidx, i_endidx);
-
- Kokkos::parallel_for("DivGrad",
- Kokkos::RangePolicy<int>(i_startidx, i_endidx),
- KOKKOS_LAMBDA(const int jc) {
-
- nabla2_psi_c_view(jc, jk, jb) =
- aux_c_view(jc, jk, jb) * avg_coeff_view(jc, 0, jb) +
- aux_c_view(cell_neighbor_idx_view(jc, jb, 0), jk, cell_neighbor_blk_view(jc, jb, 0)) *
- avg_coeff_view(jc, 1, jb) +
- aux_c_view(cell_neighbor_idx_view(jc, jb, 1), jk, cell_neighbor_blk_view(jc, jb, 1)) *
- avg_coeff_view(jc, 2, jb) +
- aux_c_view(cell_neighbor_idx_view(jc, jb, 2), jk, cell_neighbor_blk_view(jc, jb, 2)) *
- avg_coeff_view(jc, 3, jb);
-
- });
- }
- }//if
- break;
-
- default:
- std::cout << "Default case.\n";
-
-}//switch
-}//void
-
-
-template
-void nabla2_scalar_avg_lib<double>(const double* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
- const double* geofac_n2s, const double* avg_coeff, double* nabla2_psi_c,
- int i_startblk_in, int i_endblk_in, int i_startidx_in, int i_endidx_in,
- int nblks_c, int cell_type, int patch_id,
- int nlev, int slev, int elev, int nproma, bool l_limited_area, bool lacc);
-
-template
-void nabla2_scalar_avg_lib<float>(const float* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
- const float* geofac_n2s, const float* avg_coeff, float* nabla2_psi_c,
- int i_startblk_in, int i_endblk_in, int i_startidx_in, int i_endidx_in,
- int nblks_c, int cell_type, int patch_id,
- int nlev, int slev, int elev, int nproma, bool l_limited_area, bool lacc);
diff --git a/src/horizontal/mo_lib_laplace.hpp b/src/horizontal/mo_lib_laplace.hpp
index 0a67f85..2a6663c 100644
--- a/src/horizontal/mo_lib_laplace.hpp
+++ b/src/horizontal/mo_lib_laplace.hpp
@@ -22,10 +22,3 @@ void nabla2_scalar_lib(const T* psi_c, const int* cell_neighbor_idx, const int*
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc);
-template<typename T>
-void nabla2_scalar_avg_lib(const T* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
- const T* geofac_n2s, const T* avg_coeff, T* nabla2_psi_c,
- int i_startblk_in, int i_endblk_in, int i_startidx_in, int i_endidx_in,
- int nblks_c, int cell_type, int patch_id,
- int nlev, int slev, int elev, int nproma, bool l_limited_area, bool lacc);
-
--
GitLab
From 732769bcec238a625ee72b573854559442e12845 Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante4.lvt.dkrz.de>
Date: Wed, 26 Feb 2025 10:10:15 +0100
Subject: [PATCH 15/33] Revert "Fixed index for nabla2_scalar_lib"
This reverts commit 4122a81a9a283e42a5c8df8476e61820800bd0b0.
---
src/horizontal/mo_lib_laplace.cpp | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/src/horizontal/mo_lib_laplace.cpp b/src/horizontal/mo_lib_laplace.cpp
index 203b3b7..48c4479 100644
--- a/src/horizontal/mo_lib_laplace.cpp
+++ b/src/horizontal/mo_lib_laplace.cpp
@@ -56,13 +56,13 @@ void nabla2_scalar_lib(const T* psi_c, const int* cell_neighbor_idx, const int*
KOKKOS_LAMBDA(const int jk, const int jc){
nabla2_psi_c_view(jc, jk, jb) =
- psi_c_view(jc, jk, jb)*geofac_n2s_view(jc, 0, jb) +
- psi_c_view(cell_neighbor_idx_view(jc, jb, 0), jk, cell_neighbor_blk_view(jc, jb, 0)) *
- geofac_n2s_view(jc, 1, jb) +
+ psi_c_view(jc, jk, jb)*geofac_n2s_view(jc, 1, jb) +
psi_c_view(cell_neighbor_idx_view(jc, jb, 1), jk, cell_neighbor_blk_view(jc, jb, 1)) *
geofac_n2s_view(jc, 2, jb) +
psi_c_view(cell_neighbor_idx_view(jc, jb, 2), jk, cell_neighbor_blk_view(jc, jb, 2)) *
- geofac_n2s_view(jc, 3, jb);
+ geofac_n2s_view(jc, 3, jb) +
+ psi_c_view(cell_neighbor_idx_view(jc, jb, 3), jk, cell_neighbor_blk_view(jc, jb, 3)) *
+ geofac_n2s_view(jc, 4, jb);
});
}
break;
--
GitLab
From c115d9d0d3d52c1a49915544760b8d32d92293ce Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante4.lvt.dkrz.de>
Date: Wed, 26 Feb 2025 10:10:24 +0100
Subject: [PATCH 16/33] Revert "ported mo_lib_laplace"
This reverts commit 768b5176d0179e3cde4a76e894cc26a1d9513151.
---
src/horizontal/CMakeLists.txt | 4 +-
src/horizontal/mo_lib_laplace.cpp | 104 ---------------------
src/horizontal/mo_lib_laplace.hpp | 24 -----
src/horizontal/mo_lib_laplace_bindings.cpp | 50 ----------
src/horizontal/mo_lib_laplace_bindings.h | 32 -------
5 files changed, 1 insertion(+), 213 deletions(-)
delete mode 100644 src/horizontal/mo_lib_laplace.cpp
delete mode 100644 src/horizontal/mo_lib_laplace.hpp
delete mode 100644 src/horizontal/mo_lib_laplace_bindings.cpp
delete mode 100644 src/horizontal/mo_lib_laplace_bindings.h
diff --git a/src/horizontal/CMakeLists.txt b/src/horizontal/CMakeLists.txt
index 44f9e44..078a14d 100644
--- a/src/horizontal/CMakeLists.txt
+++ b/src/horizontal/CMakeLists.txt
@@ -13,9 +13,7 @@ add_library(
iconmath-horizontal
mo_lib_divrot.F90
mo_lib_laplace.F90
- mo_lib_gradients.F90
- mo_lib_laplace_bindings.cpp
- mo_lib_laplace.cpp)
+ mo_lib_gradients.F90)
add_library(${PROJECT_NAME}::horizontal ALIAS iconmath-horizontal)
diff --git a/src/horizontal/mo_lib_laplace.cpp b/src/horizontal/mo_lib_laplace.cpp
deleted file mode 100644
index 48c4479..0000000
--- a/src/horizontal/mo_lib_laplace.cpp
+++ /dev/null
@@ -1,104 +0,0 @@
-// ICON
-//
-// ---------------------------------------------------------------
-// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
-// Contact information: icon-model.org
-//
-// See AUTHORS.TXT for a list of authors
-// See LICENSES/ for license information
-// SPDX-License-Identifier: BSD-3-Clause
-// ---------------------------------------------------------------
-
-#include "mo_lib_laplace.hpp"
-//#include "mo_lib_gradients.hpp"
-#include <Kokkos_Core.hpp>
-#include <iostream>
-
-template<typename T>
-void nabla2_scalar_lib(const T* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
- const int* edge_cell_idx, const int* edge_cell_blk, const T* inv_dual_edge_length,
- const int* cell_edge_idx, const int* cell_edge_blk,
- const T* geofac_n2s, const T* geofac_div, T* nabla2_psi_c,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
- int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc){
-
- typedef Kokkos::View<const T**, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT2D;
- typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
- typedef Kokkos::View<T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedT3D;
- typedef Kokkos::View<const int***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstInt3D;
-
-
- UnmanagedConstT3D psi_c_view(psi_c, nproma,nlev,nblks_c);
- UnmanagedConstInt3D cell_neighbor_idx_view(cell_neighbor_idx, nproma,nblks_c,3);
- UnmanagedConstInt3D cell_neighbor_blk_view(cell_neighbor_blk, nproma,nblks_c,3);
- UnmanagedConstInt3D edge_cell_idx_view(edge_cell_idx, nproma,nblks_e,2);
- UnmanagedConstInt3D edge_cell_blk_view(edge_cell_blk, nproma,nblks_e,2);
- UnmanagedConstT2D inv_dual_edge_length_view(inv_dual_edge_length, nproma,nblks_e);
- UnmanagedConstInt3D cell_edge_idx_view(cell_edge_idx, nproma,nblks_c,3);
- UnmanagedConstInt3D cell_edge_blk_view(cell_edge_blk, nproma,nblks_c,3);
- UnmanagedConstT3D geofac_n2s_view(geofac_n2s, nproma,cell_type+1,nblks_c);
- UnmanagedConstT3D geofac_div_view(geofac_div, nproma,cell_type,nblks_c);
- UnmanagedT3D nabla2_psi_c_view(nabla2_psi_c, nproma,nlev,nblks_c);
-
-
- switch (cell_type){
-
- case 3:
- for (int jb = i_startblk; jb < i_endblk; ++jb) {
-
- int i_startidx, i_endidx;
- get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk, i_startidx, i_endidx);
-
-
- Kokkos::parallel_for("rbf_interpol_c2grad",
- Kokkos::MDRangePolicy<Kokkos::Rank<2>>({slev, i_startidx}, {elev, i_endidx}),
- KOKKOS_LAMBDA(const int jk, const int jc){
-
- nabla2_psi_c_view(jc, jk, jb) =
- psi_c_view(jc, jk, jb)*geofac_n2s_view(jc, 1, jb) +
- psi_c_view(cell_neighbor_idx_view(jc, jb, 1), jk, cell_neighbor_blk_view(jc, jb, 1)) *
- geofac_n2s_view(jc, 2, jb) +
- psi_c_view(cell_neighbor_idx_view(jc, jb, 2), jk, cell_neighbor_blk_view(jc, jb, 2)) *
- geofac_n2s_view(jc, 3, jb) +
- psi_c_view(cell_neighbor_idx_view(jc, jb, 3), jk, cell_neighbor_blk_view(jc, jb, 3)) *
- geofac_n2s_view(jc, 4, jb);
- });
- }
- break;
-
-
- case 6:
-/* TODO
- grad_fd_norm_lib(psi_c, edge_cell_idx, edge_cell_blk,
- inv_dual_edge_length, z_grad_fd_norm_e,&
- i_startblk_e, i_endblk_e, i_startidx_e, i_endidx_e,
- slev, elev, nproma);
-
- div_lib(z_grad_fd_norm_e, cell_edge_idx, cell_edge_blk,
- geofac_div, nabla2_psi_c, i_startblk, i_endblk,
- i_startidx_in, i_endidx_in, slev, elev, nproma);
-*/
- default:
- std::cout << "Unknown value for cell_type\n";
-
- }//switch
-}//void
-
-template
-void nabla2_scalar_lib<double>(const double* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
- const int* edge_cell_idx, const int* edge_cell_blk, const double* inv_dual_edge_length,
- const int* cell_edge_idx, const int* cell_edge_blk,
- const double* geofac_n2s, const double* geofac_div, double* nabla2_psi_c,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
- int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc);
-
-template
-void nabla2_scalar_lib<float>(const float* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
- const int* edge_cell_idx, const int* edge_cell_blk, const float* inv_dual_edge_length,
- const int* cell_edge_idx, const int* cell_edge_blk,
- const float* geofac_n2s, const float* geofac_div, float* nabla2_psi_c,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
- int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc);
diff --git a/src/horizontal/mo_lib_laplace.hpp b/src/horizontal/mo_lib_laplace.hpp
deleted file mode 100644
index 2a6663c..0000000
--- a/src/horizontal/mo_lib_laplace.hpp
+++ /dev/null
@@ -1,24 +0,0 @@
-// ICON
-//
-// ---------------------------------------------------------------
-// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
-// Contact information: icon-model.org
-//
-// See AUTHORS.TXT for a list of authors
-// See LICENSES/ for license information
-// SPDX-License-Identifier: BSD-3-Clause
-// ---------------------------------------------------------------
-#pragma once
-
-#include "mo_lib_loopindices.hpp"
-#include <Kokkos_Core.hpp>
-#include <vector>
-
-template<typename T>
-void nabla2_scalar_lib(const T* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
- const int* dge_cell_idx, const int* edge_cell_blk, const T* inv_dual_edge_length,
- const int* cell_edge_idx, const int* cell_edge_blk,
- const T* geofac_n2s, const T* geofac_div, const T* nabla2_psi_c,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
- int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc);
diff --git a/src/horizontal/mo_lib_laplace_bindings.cpp b/src/horizontal/mo_lib_laplace_bindings.cpp
deleted file mode 100644
index 43a19b3..0000000
--- a/src/horizontal/mo_lib_laplace_bindings.cpp
+++ /dev/null
@@ -1,50 +0,0 @@
-// ICON
-//
-// ---------------------------------------------------------------
-// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
-// Contact information: icon-model.org
-//
-// See AUTHORS.TXT for a list of authors
-// See LICENSES/ for license information
-// SPDX-License-Identifier: BSD-3-Clause
-// ---------------------------------------------------------------
-
-
-#include "mo_lib_laplace_bindings.h"
-#include "mo_lib_laplace.hpp"
-
-
-void nabla2_scalar_lib_dp(const double* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
- const int* edge_cell_idx, const int* edge_cell_blk, const double* inv_dual_edge_length,
- const int* cell_edge_idx, const int* cell_edge_blk,
- const double* geofac_n2s, const double* geofac_div, const double* nabla2_psi_c,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
- int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc){
-
- nabla2_scalar_lib<double>(psi_c, cell_neighbor_idx, cell_neighbor_blk,
- edge_cell_idx, edge_cell_blk, inv_dual_edge_length,
- cell_edge_idx, cell_edge_blk,
- geofac_n2s, geofac_div, nabla2_psi_c,
- i_startblk, i_endblk, i_startidx_in, i_endidx_in,
- i_startblk_e, i_endblk_e, i_startidx_e, i_endidx_e,
- nlev, slev, elev, nproma, nblks_e, nblks_c, cell_type, lacc);
-}
-
-
-void nabla2_scalar_lib_sp(const float* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
- const int* edge_cell_idx, const int* edge_cell_blk, const float* inv_dual_edge_length,
- const int* cell_edge_idx, const int* cell_edge_blk,
- const float* geofac_n2s, const float* geofac_div, const float* nabla2_psi_c,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
- int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc){
-
- nabla2_scalar_lib<float>(psi_c, cell_neighbor_idx, cell_neighbor_blk,
- edge_cell_idx, edge_cell_blk, inv_dual_edge_length,
- cell_edge_idx, cell_edge_blk,
- geofac_n2s, geofac_div, nabla2_psi_c,
- i_startblk, i_endblk, i_startidx_in, i_endidx_in,
- i_startblk_e, i_endblk_e, i_startidx_e, i_endidx_e,
- nlev, slev, elev, nproma, nblks_e, nblks_c, cell_type, lacc);
-}
diff --git a/src/horizontal/mo_lib_laplace_bindings.h b/src/horizontal/mo_lib_laplace_bindings.h
deleted file mode 100644
index 96bdc64..0000000
--- a/src/horizontal/mo_lib_laplace_bindings.h
+++ /dev/null
@@ -1,32 +0,0 @@
-// ICON
-//
-// ---------------------------------------------------------------
-// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
-// Contact information: icon-model.org
-//
-// See AUTHORS.TXT for a list of authors
-// See LICENSES/ for license information
-// SPDX-License-Identifier: BSD-3-Clause
-// ---------------------------------------------------------------
-#pragma once
-
-
-
-extern "C"{
-
-void nabla2_scalar_lib_dp(const double* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
- const int* edge_cell_idx, const int* edge_cell_blk, const double* inv_dual_edge_length,
- const int* cell_edge_idx, const int* cell_edge_blk,
- const double* geofac_n2s, const double* geofac_div, const double* nabla2_psi_c,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
- int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc);
-
-void nabla2_scalar_lib_sp(const float* psi_c, const int* cell_neighbor_idx, const int* cell_neighbor_blk,
- const int* edge_cell_idx, const int* edge_cell_blk, const float* inv_dual_edge_length,
- const int* cell_edge_idx, const int* cell_edge_blk,
- const float* geofac_n2s, const float* geofac_div, const float* nabla2_psi_c,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int i_startblk_e, int i_endblk_e, int i_startidx_e, int i_endidx_e,
- int nlev, int slev, int elev, int nproma, int nblks_e, int nblks_c, int cell_type, bool lacc);
-}
--
GitLab
From 6a1c7103cef418f5908d02507e0d40b30f1a923b Mon Sep 17 00:00:00 2001
From: Pradipta Samanta <samanta@dkrz.de>
Date: Wed, 26 Feb 2025 11:15:47 +0100
Subject: [PATCH 17/33] modified few things in the existing test for intp_rbf
---
test/c/test_intp_rbf.cpp | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/test/c/test_intp_rbf.cpp b/test/c/test_intp_rbf.cpp
index 0aa4f9b..4dc2d92 100644
--- a/test/c/test_intp_rbf.cpp
+++ b/test/c/test_intp_rbf.cpp
@@ -39,7 +39,7 @@ typedef ::testing::Types< MixedPrecision<double, double>,
// Define a typed test fixture.
template <typename TypePair>
-class RbfVecInterpolVertexMixedTestFixture : public ::testing::Test {
+class RbfVecInterpolMixedTestFixture : public ::testing::Test {
public:
using InType = typename TypePair::in_type;
using OutType = typename TypePair::out_type;
@@ -69,7 +69,7 @@ public:
std::vector<OutType> p_u_out; // Dimensions: (nproma, nlev, nblks_v)
std::vector<OutType> p_v_out; // Dimensions: (nproma, nlev, nblks_v)
- RbfVecInterpolVertexMixedTestFixture() {
+ RbfVecInterpolMixedTestFixture() {
// Allocate and initialize inputs.
p_e_in.resize(num_elements_3d<InType>(nproma, nlev, nblks_e), static_cast<InType>(1));
rbf_vec_idx_v.resize(num_elements_3d<int>(rbf_vec_dim, nproma, nblks_v), 1);
@@ -82,9 +82,9 @@ public:
}
};
-TYPED_TEST_SUITE(RbfVecInterpolVertexMixedTestFixture, MixedTypes);
+TYPED_TEST_SUITE(RbfVecInterpolMixedTestFixture, MixedTypes);
-TYPED_TEST(RbfVecInterpolVertexMixedTestFixture, BasicTest) {
+TYPED_TEST(RbfVecInterpolMixedTestFixture, BasicTest) {
using InType = typename TestFixture::InType;
using OutType = typename TestFixture::OutType;
@@ -106,8 +106,8 @@ TYPED_TEST(RbfVecInterpolVertexMixedTestFixture, BasicTest) {
this->lacc,
this->acc_async,
this->nlev,
- RbfVecInterpolVertexMixedTestFixture< TypeParam >::nblks_e,
- RbfVecInterpolVertexMixedTestFixture< TypeParam >::nblks_v);
+ this->nblks_e,
+ this->nblks_v);
// Check the outputs only for blocks in the range [i_startblk, i_endblk].
for (int block = this->i_startblk; block <= this->i_endblk; ++block) {
--
GitLab
From 8dadd736b6f11f68a9f86f89a5e77986358008b2 Mon Sep 17 00:00:00 2001
From: Pradipta Samanta <samanta@dkrz.de>
Date: Wed, 26 Feb 2025 11:16:28 +0100
Subject: [PATCH 18/33] fixed few bugs in one of the routine of intp_rbf
---
src/interpolation/mo_lib_intp_rbf.cpp | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/src/interpolation/mo_lib_intp_rbf.cpp b/src/interpolation/mo_lib_intp_rbf.cpp
index a7d9cbd..e87a3e5 100644
--- a/src/interpolation/mo_lib_intp_rbf.cpp
+++ b/src/interpolation/mo_lib_intp_rbf.cpp
@@ -32,13 +32,13 @@ void rbf_interpol_c2grad_lib(const T* p_cell_in, const int* rbf_c2grad_idx, cons
UnmanagedConstInt3D rbf_c2grad_blk_view(rbf_c2grad_blk, rbf_c2grad_dim, nproma, nblks_c);
UnmanagedConstT4D rbf_c2grad_coeff_view(rbf_c2grad_coeff, rbf_c2grad_dim, 2, nproma, nblks_c);
- for (int jb = i_startblk; jb < i_endblk; ++jb) {
+ for (int jb = i_startblk; jb <= i_endblk; ++jb) {
int i_startidx, i_endidx;
get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk, i_startidx, i_endidx);
Kokkos::parallel_for("rbf_interpol_c2grad",
- Kokkos::MDRangePolicy<Kokkos::Rank<2>>({slev, i_startidx}, {elev, i_endidx}),
+ Kokkos::MDRangePolicy<Kokkos::Rank<2>>({slev, i_startidx}, {elev+1, i_endidx+1}),
KOKKOS_LAMBDA(const int jk, const int jc){
grad_x_view(jc, jk, jb) =
--
GitLab
From 06cf4132c0920ba9615d9b9121a5e3cb7f2b048d Mon Sep 17 00:00:00 2001
From: Pradipta Samanta <samanta@dkrz.de>
Date: Wed, 26 Feb 2025 11:19:12 +0100
Subject: [PATCH 19/33] added a unit-test for rbf_interpol_c2grad_lib
---
test/c/test_intp_rbf.cpp | 84 +++++++++++++++++++++++++++++++++++++++-
1 file changed, 83 insertions(+), 1 deletion(-)
diff --git a/test/c/test_intp_rbf.cpp b/test/c/test_intp_rbf.cpp
index 4dc2d92..a6aa11a 100644
--- a/test/c/test_intp_rbf.cpp
+++ b/test/c/test_intp_rbf.cpp
@@ -13,6 +13,7 @@
#include <Kokkos_Core.hpp>
#include <vector>
#include "mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib.hpp"
+#include "mo_lib_intp_rbf.hpp"
// Free-function helpers for 3D and 4D array sizes (assumed column-major)
template<typename T>
@@ -37,7 +38,88 @@ typedef ::testing::Types< MixedPrecision<double, double>,
MixedPrecision<double, float>,
MixedPrecision<float, float> > MixedTypes;
-// Define a typed test fixture.
+// Define a typed test fixture for the functions which have the same input and output types
+template <typename T>
+class RbfInterpolTypedTestFixture : public ::testing::Test {
+public:
+
+ static constexpr int nproma = 3; // inner loop length
+ static constexpr int nlev = 4; // vertical levels
+ static constexpr int nblks_c = 2; // number of blocks for the cell arrays
+ static constexpr int rbf_c2grad_dim = 10; // fixed dimension for the RBF c2 gradient
+
+ // Parameters for the index ranges.
+ int i_startblk = 0;
+ int i_endblk = 1; // Note: loop over jb uses jb < i_endblk, so blocks 0 and 1
+ int i_startidx_in = 0;
+ int i_endidx_in = nproma-1; // [0, nproma) i.e. 0 .. nproma-1
+ int slev = 0;
+ int elev = nlev-1;
+ bool lacc = false; // Not using any ACC-specific behavior.
+
+ // Data arrays.
+ std::vector<T> p_cell_in; // size: nproma * nlev * nblks_c
+ std::vector<int> rbf_c2grad_idx; // size: rbf_c2grad_dim * nproma * nblks_c
+ std::vector<int> rbf_c2grad_blk; // size: rbf_c2grad_dim * nproma * nblks_c
+ std::vector<T> rbf_c2grad_coeff; // size: rbf_c2grad_dim * 2 * nproma * nblks_c
+ std::vector<T> grad_x; // size: nproma * nlev * nblks_c
+ std::vector<T> grad_y; // size: nproma * nlev * nblks_c
+
+ RbfInterpolTypedTestFixture() {
+ size_t size3d = static_cast<size_t>(nproma) * nlev * nblks_c;
+ size_t size3d_idx = static_cast<size_t>(rbf_c2grad_dim) * nproma * nblks_c;
+ size_t size4d = static_cast<size_t>(rbf_c2grad_dim) * 2 * nproma * nblks_c;
+ p_cell_in.resize(size3d, static_cast<T>(1));
+ rbf_c2grad_idx.resize(size3d_idx, 1);
+ rbf_c2grad_blk.resize(size3d_idx, 0); // Set block indices to 0 for testing.
+ rbf_c2grad_coeff.resize(size4d, static_cast<T>(1));
+ grad_x.resize(size3d, static_cast<T>(0));
+ grad_y.resize(size3d, static_cast<T>(0));
+ }
+};
+
+typedef ::testing::Types<float, double> MyTypes;
+
+TYPED_TEST_SUITE(RbfInterpolTypedTestFixture, MyTypes);
+
+TYPED_TEST(RbfInterpolTypedTestFixture, BasicTest) {
+
+ rbf_interpol_c2grad_lib<TypeParam>(
+ this->p_cell_in.data(),
+ this->rbf_c2grad_idx.data(),
+ this->rbf_c2grad_blk.data(),
+ this->rbf_c2grad_coeff.data(),
+ this->grad_x.data(),
+ this->grad_y.data(),
+ this->i_startblk,
+ this->i_endblk,
+ this->i_startidx_in,
+ this->i_endidx_in,
+ this->slev,
+ this->elev,
+ this->nproma,
+ this->rbf_c2grad_dim,
+ this->nlev,
+ this->nblks_c,
+ this->lacc);
+
+ // For each block from i_startblk to i_endblk-1, and for each (i, level)
+ // the kernel sums rbf_c2grad_dim contributions, each equal to 1.
+ // Therefore, we expect grad_x and grad_y to equal rbf_c2grad_dim (i.e., 10).
+ for (int jb = this->i_startblk; jb < this->i_endblk; ++jb) {
+ for (int jk = 0; jk < this->nlev; ++jk) {
+ for (int i = 0; i < this->nproma; ++i) {
+ size_t idx = i + static_cast<size_t>(jk) * this->nproma + static_cast<size_t>(jb) * this->nproma * this->nlev;
+ EXPECT_NEAR(this->grad_x[idx], static_cast<TypeParam>(this->rbf_c2grad_dim), static_cast<TypeParam>(1e-5))
+ << "grad_x failure at block " << jb << ", level " << jk << ", index " << i;
+ EXPECT_NEAR(this->grad_y[idx], static_cast<TypeParam>(this->rbf_c2grad_dim), static_cast<TypeParam>(1e-5))
+ << "grad_y failure at block " << jb << ", level " << jk << ", index " << i;
+ }
+ }
+ }
+}
+
+// Define a typed test fixture for the functions which have different input and output types
template <typename TypePair>
class RbfVecInterpolMixedTestFixture : public ::testing::Test {
public:
--
GitLab
From 6db33cd540dcab1d944cbf25fcf0b117ad8b8c52 Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante4.lvt.dkrz.de>
Date: Thu, 27 Feb 2025 11:15:29 +0100
Subject: [PATCH 20/33] Rbf clean up
---
src/interpolation/mo_lib_intp_rbf.cpp | 58 +++++++--------
src/interpolation/mo_lib_intp_rbf.hpp | 21 +++---
.../mo_lib_intp_rbf_bindings.cpp | 47 ++++++------
src/interpolation/mo_lib_intp_rbf_bindings.h | 72 ++++++++++---------
4 files changed, 103 insertions(+), 95 deletions(-)
diff --git a/src/interpolation/mo_lib_intp_rbf.cpp b/src/interpolation/mo_lib_intp_rbf.cpp
index e87a3e5..b22e737 100644
--- a/src/interpolation/mo_lib_intp_rbf.cpp
+++ b/src/interpolation/mo_lib_intp_rbf.cpp
@@ -91,24 +91,25 @@ void rbf_interpol_c2grad_lib(const T* p_cell_in, const int* rbf_c2grad_idx, cons
template
void rbf_interpol_c2grad_lib<double>(const double* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const double* rbf_c2grad_coeff, double* grad_x, double* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
+ const double* rbf_c2grad_coeff, double* grad_x, double* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
template
void rbf_interpol_c2grad_lib<float>(const float* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const float* rbf_c2grad_coeff, float* grad_x, float* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
+ const float* rbf_c2grad_coeff, float* grad_x, float* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
//------------------------------------------rbf_vec_interpol_cell_lib---------------------------------------------
template <typename T>
void rbf_vec_interpol_cell_lib(const T* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const T* rbf_vec_coeff_c, T* p_u_out, T* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e, int rbf_vec_dim_c, bool lacc, bool acc_async){
+ const T* rbf_vec_coeff_c, T* p_u_out, T* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e, i
+ int rbf_vec_dim_c, bool lacc, bool acc_async){
typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
typedef Kokkos::View<T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedT3D;
@@ -178,23 +179,25 @@ void rbf_vec_interpol_cell_lib(const T* p_vn_in, const int* rbf_vec_idx_c, const
template
void rbf_vec_interpol_cell_lib<double>(const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e, int rbf_vec_dim_c, bool lacc, bool acc_async);
+ const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e,
+ int rbf_vec_dim_c, bool lacc, bool acc_async);
template
void rbf_vec_interpol_cell_lib<float>(const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e, int rbf_vec_dim_c, bool lacc, bool acc_async);
+ const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e,
+ int rbf_vec_dim_c, bool lacc, bool acc_async);
//------------------------------------------rbf_vec_interpol_edge_lib---------------------------------------------
template <typename T>
void rbf_vec_interpol_edge_lib(const T* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const T* rbf_vec_coeff_e, T* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async){
+ const T* rbf_vec_coeff_e, T* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async){
typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
typedef Kokkos::View<T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedT3D;
@@ -228,16 +231,15 @@ void rbf_vec_interpol_edge_lib(const T* p_vn_in, const int* rbf_vec_idx_e, const
template
void rbf_vec_interpol_edge_lib<double>(const double* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const double* rbf_vec_coeff_e, double* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async);
+ const double* rbf_vec_coeff_e, double* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async);
template
void rbf_vec_interpol_edge_lib<float>(const float* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const float* rbf_vec_coeff_e, float* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async);
-
+ const float* rbf_vec_coeff_e, float* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async);
diff --git a/src/interpolation/mo_lib_intp_rbf.hpp b/src/interpolation/mo_lib_intp_rbf.hpp
index e52989d..1e95737 100644
--- a/src/interpolation/mo_lib_intp_rbf.hpp
+++ b/src/interpolation/mo_lib_intp_rbf.hpp
@@ -16,20 +16,21 @@
template <typename T>
void rbf_interpol_c2grad_lib(const T* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const T* rbf_c2grad_coeff, T* grad_x, T* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
+ const T* rbf_c2grad_coeff, T* grad_x, T* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
template <typename T>
void rbf_vec_interpol_cell_lib(const T* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const T* rbf_vec_coeff_c, T* p_u_out, T* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e, int rbf_vec_dim_c, bool lacc, bool acc_async);
+ const T* rbf_vec_coeff_c, T* p_u_out, T* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e,
+ int rbf_vec_dim_c, bool lacc, bool acc_async);
template <typename T>
void rbf_vec_interpol_edge_lib(const T* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const T* rbf_vec_coeff_e, T* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_asynci);
+ const T* rbf_vec_coeff_e, T* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async);
diff --git a/src/interpolation/mo_lib_intp_rbf_bindings.cpp b/src/interpolation/mo_lib_intp_rbf_bindings.cpp
index 5c7943b..2932f24 100644
--- a/src/interpolation/mo_lib_intp_rbf_bindings.cpp
+++ b/src/interpolation/mo_lib_intp_rbf_bindings.cpp
@@ -15,10 +15,9 @@
void rbf_interpol_c2grad_lib_sp(const float* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const float* rbf_c2grad_coeff, float* grad_x, float* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc){
-
+ const float* rbf_c2grad_coeff, float* grad_x, float* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc){
rbf_interpol_c2grad_lib<float>(p_cell_in, rbf_c2grad_idx, rbf_c2grad_blk,
rbf_c2grad_coeff,
@@ -27,11 +26,10 @@ void rbf_interpol_c2grad_lib_sp(const float* p_cell_in, const int*rbf_c2grad_idx
nproma, rbf_c2grad_dim, nlev, nblk_c, lacc);
}
-void rbf_interpol_c2grad_lib_dp(const double* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const double* rbf_c2grad_coeff, double* grad_x, double* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc){
-
+void rbf_interpol_c2grad_lib_dp(const double* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const double* rbf_c2grad_coeff, double* grad_x, double* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc){
rbf_interpol_c2grad_lib<double>(p_cell_in, rbf_c2grad_idx, rbf_c2grad_blk,
rbf_c2grad_coeff,
@@ -41,9 +39,10 @@ void rbf_interpol_c2grad_lib_dp(const double* p_cell_in, const int*rbf_c2grad_id
}
void rbf_vec_interpol_cell_lib_sp(const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int nblk_e, int rbf_vec_dim_c, bool lacc, bool acc_async){
+ const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int nblk_e,
+ int rbf_vec_dim_c, bool lacc, bool acc_async){
rbf_vec_interpol_cell_lib<float>(p_vn_in, rbf_vec_idx_c, rbf_vec_blk_c,
rbf_vec_coeff_c, p_u_out, p_v_out,
@@ -52,9 +51,10 @@ void rbf_vec_interpol_cell_lib_sp(const float* p_vn_in, const int* rbf_vec_idx_c
}
void rbf_vec_interpol_cell_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int nblk_e, int rbf_vec_dim_c, bool lacc, bool acc_async){
+ const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int nblk_e,
+ int rbf_vec_dim_c, bool lacc, bool acc_async){
rbf_vec_interpol_cell_lib<double>(p_vn_in, rbf_vec_idx_c, rbf_vec_blk_c, rbf_vec_coeff_c,
p_u_out, p_v_out, i_startblk, i_endblk, i_startidx_in, i_endidx_in,
@@ -62,11 +62,10 @@ void rbf_vec_interpol_cell_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_
}
void rbf_vec_interpol_edge_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const double* rbf_vec_coeff_e, double* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async){
-
+ const double* rbf_vec_coeff_e, double* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async){
rbf_vec_interpol_edge_lib<double>(p_vn_in, rbf_vec_idx_e, rbf_vec_blk_e,
rbf_vec_coeff_e, p_vt_out,
@@ -75,10 +74,10 @@ void rbf_vec_interpol_edge_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_
}
void rbf_vec_interpol_edge_lib_sp(const float* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const float* rbf_vec_coeff_e, float* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async){
+ const float* rbf_vec_coeff_e, float* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async){
rbf_vec_interpol_edge_lib<float>(p_vn_in, rbf_vec_idx_e, rbf_vec_blk_e,
rbf_vec_coeff_e, p_vt_out,
diff --git a/src/interpolation/mo_lib_intp_rbf_bindings.h b/src/interpolation/mo_lib_intp_rbf_bindings.h
index 9620614..9473aef 100644
--- a/src/interpolation/mo_lib_intp_rbf_bindings.h
+++ b/src/interpolation/mo_lib_intp_rbf_bindings.h
@@ -12,37 +12,43 @@
extern "C" {
-void rbf_interpol_c2grad_lib_sp(const float* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const float* rbf_c2grad_coeff, float* grad_x, const float* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
-
-void rbf_interpol_c2grad_lib_dp(const double* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const double* rbf_c2grad_coeff, double* grad_x, const double* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
-
-void rbf_vec_interpol_cell_lib_sp(const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int nblk_e, int rbf_vec_dim_c, bool lacc, bool acc_async);
-
-void rbf_vec_interpol_cell_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int nblk_e, int rbf_vec_dim_c, bool lacc, bool acc_async);
-
-void rbf_vec_interpol_edge_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const double* rbf_vec_coeff_e, double* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async);
-
-void rbf_vec_interpol_edge_lib_sp(const float* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const float* rbf_vec_coeff_e, float* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async);
-
-
+void rbf_interpol_c2grad_lib_sp(
+ const float* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const float* rbf_c2grad_coeff, float* grad_x, const float* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
+
+void rbf_interpol_c2grad_lib_dp(
+ const double* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const double* rbf_c2grad_coeff, double* grad_x, const double* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
+
+void rbf_vec_interpol_cell_lib_sp(
+ const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+ const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int nblk_e,
+ int rbf_vec_dim_c, bool lacc, bool acc_async);
+
+void rbf_vec_interpol_cell_lib_dp(
+ const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+ const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblk_c, int nblk_e,
+ int rbf_vec_dim_c, bool lacc, bool acc_async);
+
+void rbf_vec_interpol_edge_lib_dp(
+ const double* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
+ const double* rbf_vec_coeff_e, double* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async);
+
+void rbf_vec_interpol_edge_lib_sp(
+ const float* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
+ const float* rbf_vec_coeff_e, float* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async);
}
--
GitLab
From 1be0c3ab1cec1bcbcf3ab6b446e16ccd4a58c2a5 Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante4.lvt.dkrz.de>
Date: Thu, 27 Feb 2025 11:22:56 +0100
Subject: [PATCH 21/33] Fix
---
src/interpolation/mo_lib_intp_rbf.cpp | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/interpolation/mo_lib_intp_rbf.cpp b/src/interpolation/mo_lib_intp_rbf.cpp
index b22e737..1cce6d8 100644
--- a/src/interpolation/mo_lib_intp_rbf.cpp
+++ b/src/interpolation/mo_lib_intp_rbf.cpp
@@ -108,7 +108,7 @@ template <typename T>
void rbf_vec_interpol_cell_lib(const T* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
const T* rbf_vec_coeff_c, T* p_u_out, T* p_v_out,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e, i
+ int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e,
int rbf_vec_dim_c, bool lacc, bool acc_async){
typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
--
GitLab
From 0f5a6e7543abd5e54edf9b043b95bf3d4bca51c8 Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante0.lvt.dkrz.de>
Date: Tue, 4 Mar 2025 16:46:31 +0100
Subject: [PATCH 22/33] Added test for cell_lib
---
src/interpolation/mo_lib_intp_rbf.cpp | 42 ++++++------
src/interpolation/mo_lib_intp_rbf.hpp | 4 +-
.../mo_lib_intp_rbf_bindings.cpp | 8 +--
src/interpolation/mo_lib_intp_rbf_bindings.h | 4 +-
test/c/test_intp_rbf.cpp | 64 ++++++++++++++++++-
5 files changed, 92 insertions(+), 30 deletions(-)
diff --git a/src/interpolation/mo_lib_intp_rbf.cpp b/src/interpolation/mo_lib_intp_rbf.cpp
index 1cce6d8..bdbfbe5 100644
--- a/src/interpolation/mo_lib_intp_rbf.cpp
+++ b/src/interpolation/mo_lib_intp_rbf.cpp
@@ -93,13 +93,13 @@ template
void rbf_interpol_c2grad_lib<double>(const double* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
const double* rbf_c2grad_coeff, double* grad_x, double* grad_y,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblks_c, bool lacc);
template
void rbf_interpol_c2grad_lib<float>(const float* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
const float* rbf_c2grad_coeff, float* grad_x, float* grad_y,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblks_c, bool lacc);
//------------------------------------------rbf_vec_interpol_cell_lib---------------------------------------------
@@ -108,7 +108,7 @@ template <typename T>
void rbf_vec_interpol_cell_lib(const T* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
const T* rbf_vec_coeff_c, T* p_u_out, T* p_v_out,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e,
+ int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
int rbf_vec_dim_c, bool lacc, bool acc_async){
typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
@@ -117,20 +117,20 @@ void rbf_vec_interpol_cell_lib(const T* p_vn_in, const int* rbf_vec_idx_c, const
typedef Kokkos::View<const T****, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT4D;
UnmanagedConstT3D p_vn_in_view(p_vn_in, nproma, nlev, nblks_e);
- UnmanagedConstInt3D rbf_vec_idx_c_view(rbf_vec_idx_c, rbf_vec_dim_c, nproma, nblk_c);
- UnmanagedConstInt3D rbf_vec_blk_c_view(rbf_vec_blk_c, rbf_vec_dim_c, nproma, nblk_c);
- UnmanagedConstT4D rbf_vec_coeff_c_view(rbf_vec_coeff_c, nproma, nblk_c); //TODO
- UnmanagedT3D p_u_out_view(p_u_out, nproma, nlev, nblk_c);
- UnmanagedT3D p_v_out_view(p_u_out, nproma, nlev, nblk_c);
+ UnmanagedConstInt3D rbf_vec_idx_c_view(rbf_vec_idx_c, rbf_vec_dim_c, nproma, nblks_c);
+ UnmanagedConstInt3D rbf_vec_blk_c_view(rbf_vec_blk_c, rbf_vec_dim_c, nproma, nblks_c);
+ UnmanagedConstT4D rbf_vec_coeff_c_view(rbf_vec_coeff_c, nproma, nblks_c); //TODO
+ UnmanagedT3D p_u_out_view(p_u_out, nproma, nlev, nblks_c);
+ UnmanagedT3D p_v_out_view(p_u_out, nproma, nlev, nblks_c);
- for (int jb = i_startblk; jb < i_endblk; ++jb) {
+ for (int jb = i_startblk; jb <= i_endblk; ++jb) {
int i_startidx, i_endidx;
get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk, i_startidx, i_endidx);
Kokkos::parallel_for("rbf_vec_interpol_cell_lib",
- Kokkos::MDRangePolicy<Kokkos::Rank<2>>({slev, i_startidx}, {elev, i_endidx}),
+ Kokkos::MDRangePolicy<Kokkos::Rank<2>>({slev, i_startidx}, {elev+1, i_endidx+1}),
KOKKOS_LAMBDA(const int jk, const int jc){
p_u_out_view(jc, jk, jb) =
@@ -154,23 +154,23 @@ void rbf_vec_interpol_cell_lib(const T* p_vn_in, const int* rbf_vec_idx_c, const
p_vn_in_view(rbf_vec_idx_c_view(8, jc, jb), jk, rbf_vec_blk_c_view(8, jc, jb));
p_v_out_view(jc, jk, jb) =
- rbf_vec_coeff_c_view(0, 2, jc, jb)*
+ rbf_vec_coeff_c_view(0, 1, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(0, jc, jb), jk, rbf_vec_blk_c_view(0, jc, jb)) +
- rbf_vec_coeff_c_view(1, 2, jc, jb)*
+ rbf_vec_coeff_c_view(1, 1, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(1, jc, jb), jk, rbf_vec_blk_c_view(1, jc, jb)) +
- rbf_vec_coeff_c_view(2, 2, jc, jb)*
+ rbf_vec_coeff_c_view(2, 1, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(2, jc, jb), jk, rbf_vec_blk_c_view(2, jc, jb)) +
- rbf_vec_coeff_c_view(3, 2, jc, jb)*
+ rbf_vec_coeff_c_view(3, 1, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(3, jc, jb), jk, rbf_vec_blk_c_view(3, jc, jb)) +
- rbf_vec_coeff_c_view(4, 2, jc, jb)*
+ rbf_vec_coeff_c_view(4, 1, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(4, jc, jb), jk, rbf_vec_blk_c_view(4, jc, jb)) +
- rbf_vec_coeff_c_view(5, 2, jc, jb)*
+ rbf_vec_coeff_c_view(5, 1, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(5, jc, jb), jk, rbf_vec_blk_c_view(5, jc, jb)) +
- rbf_vec_coeff_c_view(6, 2, jc, jb)*
+ rbf_vec_coeff_c_view(6, 1, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(6, jc, jb), jk, rbf_vec_blk_c_view(6, jc, jb)) +
- rbf_vec_coeff_c_view(7, 2, jc, jb)*
+ rbf_vec_coeff_c_view(7, 1, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(7, jc, jb), jk, rbf_vec_blk_c_view(7, jc, jb)) +
- rbf_vec_coeff_c_view(8, 2, jc, jb)*
+ rbf_vec_coeff_c_view(8, 1, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(8, jc, jb), jk, rbf_vec_blk_c_view(8, jc, jb));
});
Kokkos::fence();
@@ -181,13 +181,13 @@ template
void rbf_vec_interpol_cell_lib<double>(const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e,
+ int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
int rbf_vec_dim_c, bool lacc, bool acc_async);
template
void rbf_vec_interpol_cell_lib<float>(const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e,
+ int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
int rbf_vec_dim_c, bool lacc, bool acc_async);
//------------------------------------------rbf_vec_interpol_edge_lib---------------------------------------------
diff --git a/src/interpolation/mo_lib_intp_rbf.hpp b/src/interpolation/mo_lib_intp_rbf.hpp
index 1e95737..5b9d5e1 100644
--- a/src/interpolation/mo_lib_intp_rbf.hpp
+++ b/src/interpolation/mo_lib_intp_rbf.hpp
@@ -18,14 +18,14 @@ template <typename T>
void rbf_interpol_c2grad_lib(const T* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
const T* rbf_c2grad_coeff, T* grad_x, T* grad_y,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblks_c, bool lacc);
template <typename T>
void rbf_vec_interpol_cell_lib(const T* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
const T* rbf_vec_coeff_c, T* p_u_out, T* p_v_out,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int nblks_e,
+ int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
int rbf_vec_dim_c, bool lacc, bool acc_async);
template <typename T>
diff --git a/src/interpolation/mo_lib_intp_rbf_bindings.cpp b/src/interpolation/mo_lib_intp_rbf_bindings.cpp
index 2932f24..b679619 100644
--- a/src/interpolation/mo_lib_intp_rbf_bindings.cpp
+++ b/src/interpolation/mo_lib_intp_rbf_bindings.cpp
@@ -41,24 +41,24 @@ void rbf_interpol_c2grad_lib_dp(const double* p_cell_in, const int* rbf_c2grad_i
void rbf_vec_interpol_cell_lib_sp(const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int nblk_e,
+ int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
int rbf_vec_dim_c, bool lacc, bool acc_async){
rbf_vec_interpol_cell_lib<float>(p_vn_in, rbf_vec_idx_c, rbf_vec_blk_c,
rbf_vec_coeff_c, p_u_out, p_v_out,
i_startblk, i_endblk, i_startidx_in, i_endidx_in,
- slev, elev, nproma, nlev, nblk_c, nblk_e, rbf_vec_dim_c, lacc, acc_async);
+ slev, elev, nproma, nlev, nblks_c, nblks_e, rbf_vec_dim_c, lacc, acc_async);
}
void rbf_vec_interpol_cell_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int nblk_e,
+ int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
int rbf_vec_dim_c, bool lacc, bool acc_async){
rbf_vec_interpol_cell_lib<double>(p_vn_in, rbf_vec_idx_c, rbf_vec_blk_c, rbf_vec_coeff_c,
p_u_out, p_v_out, i_startblk, i_endblk, i_startidx_in, i_endidx_in,
- slev, elev, nproma, nlev, nblk_c, nblk_e, rbf_vec_dim_c, lacc, acc_async);
+ slev, elev, nproma, nlev, nblks_c, nblks_e, rbf_vec_dim_c, lacc, acc_async);
}
void rbf_vec_interpol_edge_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
diff --git a/src/interpolation/mo_lib_intp_rbf_bindings.h b/src/interpolation/mo_lib_intp_rbf_bindings.h
index 9473aef..cc1dd5c 100644
--- a/src/interpolation/mo_lib_intp_rbf_bindings.h
+++ b/src/interpolation/mo_lib_intp_rbf_bindings.h
@@ -28,14 +28,14 @@ void rbf_vec_interpol_cell_lib_sp(
const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int nblk_e,
+ int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
int rbf_vec_dim_c, bool lacc, bool acc_async);
void rbf_vec_interpol_cell_lib_dp(
const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblk_c, int nblk_e,
+ int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
int rbf_vec_dim_c, bool lacc, bool acc_async);
void rbf_vec_interpol_edge_lib_dp(
diff --git a/test/c/test_intp_rbf.cpp b/test/c/test_intp_rbf.cpp
index a6aa11a..9414061 100644
--- a/test/c/test_intp_rbf.cpp
+++ b/test/c/test_intp_rbf.cpp
@@ -12,6 +12,8 @@
#include <gtest/gtest.h>
#include <Kokkos_Core.hpp>
#include <vector>
+#include <algorithm>
+#include <numeric>
#include "mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib.hpp"
#include "mo_lib_intp_rbf.hpp"
@@ -46,7 +48,9 @@ public:
static constexpr int nproma = 3; // inner loop length
static constexpr int nlev = 4; // vertical levels
static constexpr int nblks_c = 2; // number of blocks for the cell arrays
+ static constexpr int nblks_e = 2;
static constexpr int rbf_c2grad_dim = 10; // fixed dimension for the RBF c2 gradient
+ static constexpr int rbf_vec_dim_c = 9;
// Parameters for the index ranges.
int i_startblk = 0;
@@ -56,23 +60,43 @@ public:
int slev = 0;
int elev = nlev-1;
bool lacc = false; // Not using any ACC-specific behavior.
+ bool acc_async = false;
// Data arrays.
std::vector<T> p_cell_in; // size: nproma * nlev * nblks_c
std::vector<int> rbf_c2grad_idx; // size: rbf_c2grad_dim * nproma * nblks_c
std::vector<int> rbf_c2grad_blk; // size: rbf_c2grad_dim * nproma * nblks_c
+ std::vector<int> rbf_vec_idx_c; //size: rbf_vec_dim_c * nproma * nblks_c
+ std::vector<int> rbf_vec_blk_c; // size: rbf_vec_dim_c * nproma * nblks_c
std::vector<T> rbf_c2grad_coeff; // size: rbf_c2grad_dim * 2 * nproma * nblks_c
std::vector<T> grad_x; // size: nproma * nlev * nblks_c
std::vector<T> grad_y; // size: nproma * nlev * nblks_c
+ std::vector<T> p_vn_in;
+ std::vector<T> rbf_vec_coeff_c;
+ std::vector<T> p_u_out;
+ std::vector<T> p_v_out;
RbfInterpolTypedTestFixture() {
size_t size3d = static_cast<size_t>(nproma) * nlev * nblks_c;
size_t size3d_idx = static_cast<size_t>(rbf_c2grad_dim) * nproma * nblks_c;
size_t size4d = static_cast<size_t>(rbf_c2grad_dim) * 2 * nproma * nblks_c;
+
+ size_t size3d_vec_dim = static_cast<size_t>(rbf_vec_dim_c) * nproma * nblks_c;
+ size_t size_4d_vec_dim = static_cast<size_t>(rbf_vec_dim_c) * 2 * nproma * nblks_c;
+
p_cell_in.resize(size3d, static_cast<T>(1));
+ p_vn_in.resize(size3d, static_cast<T>(1));
+
+ rbf_vec_idx_c.resize(size3d_vec_dim, 1);
+ rbf_vec_blk_c.resize(size3d_vec_dim, 0);
rbf_c2grad_idx.resize(size3d_idx, 1);
rbf_c2grad_blk.resize(size3d_idx, 0); // Set block indices to 0 for testing.
+
+ rbf_vec_coeff_c.resize(size_4d_vec_dim, static_cast<T>(1));
rbf_c2grad_coeff.resize(size4d, static_cast<T>(1));
+
+ p_u_out.resize(size3d_vec_dim, static_cast<T>(0));
+ p_v_out.resize(size3d_vec_dim, static_cast<T>(0));
grad_x.resize(size3d, static_cast<T>(0));
grad_y.resize(size3d, static_cast<T>(0));
}
@@ -83,7 +107,7 @@ typedef ::testing::Types<float, double> MyTypes;
TYPED_TEST_SUITE(RbfInterpolTypedTestFixture, MyTypes);
TYPED_TEST(RbfInterpolTypedTestFixture, BasicTest) {
-
+ using T = TypeParam;
rbf_interpol_c2grad_lib<TypeParam>(
this->p_cell_in.data(),
this->rbf_c2grad_idx.data(),
@@ -119,6 +143,44 @@ TYPED_TEST(RbfInterpolTypedTestFixture, BasicTest) {
}
}
+
+TYPED_TEST(RbfInterpolTypedTestFixture, cell_lib) {
+ using T = TypeParam;
+
+ rbf_vec_interpol_cell_lib<T>(
+ this->p_vn_in.data(),
+ this->rbf_vec_idx_c.data(),
+ this->rbf_vec_blk_c.data(),
+ this->rbf_vec_coeff_c.data(),
+ this->p_u_out.data(),
+ this->p_v_out.data(),
+ this->i_startblk,
+ this->i_endblk,
+ this->i_startidx_in,
+ this->i_endidx_in,
+ this->slev,
+ this->elev,
+ this->nproma,
+ this->rbf_c2grad_dim,
+ this->nlev,
+ this->nblks_c,
+ this->nblks_e,
+ this->lacc,
+ this->acc_async);
+
+ for (int jb = this->i_startblk; jb < this->i_endblk; ++jb) {
+ for (int jk = 0; jk < this->nlev; ++jk) {
+ for (int i = 0; i < this->nproma; ++i) {
+ size_t idx = i + static_cast<size_t>(jk) * this->nproma + static_cast<size_t>(jb) * this->nproma * this->nlev;
+ EXPECT_NEAR(this->p_u_out[idx], static_cast<T>(this->rbf_vec_dim_c), static_cast<T>(1e-5))
+ << "p_u_out failure at block " << jb << ", level " << jk << ", index " << i;
+ // EXPECT_NEAR(this->p_v_out[idx], static_cast<T>(this->rbf_vec_dim_c), static_cast<T>(1e-5))
+ // << "p_v_out failure at block " << jb << ", level " << jk << ", index " << i;
+ }
+ }
+ }
+ }
+
// Define a typed test fixture for the functions which have different input and output types
template <typename TypePair>
class RbfVecInterpolMixedTestFixture : public ::testing::Test {
--
GitLab
From 8175944d3f080ec145d4ce59e2e6fdddb8d6153d Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante0.lvt.dkrz.de>
Date: Tue, 4 Mar 2025 16:57:03 +0100
Subject: [PATCH 23/33] Fixed strange typo
---
src/horizontal/mo_lib_divrot.F90 | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/src/horizontal/mo_lib_divrot.F90 b/src/horizontal/mo_lib_divrot.F90
index a2f2ad9..c1360f6 100644
--- a/src/horizontal/mo_lib_divrot.F90
+++ b/src/horizontal/mo_lib_divrot.F90
@@ -561,7 +561,7 @@ CONTAINS
!!
!! !LITERATURE
!! Ollivier-Gooch et al (2002): A High-Order-Accurate Unstructured Mesh
-!! Finite-Volume Scheme for the Advection-Diffusion Equation, J. Comput. Phys.,
+!! Finite-Volume Scheme for the Advection-Diffusion Equation, J. Compute. Phys.,
!! 181, 729-752
!!
SUBROUTINE recon_lsq_cell_q_lib(p_cc, lsq_idx_c, lsq_blk_c, &
@@ -800,7 +800,7 @@ CONTAINS
!!
!! !LITERATURE
!! Ollivier-Gooch et al (2002): A High-Order-Accurate Unstructured Mesh
-!! Finite-Volume Scheme for the Advection-Diffusion Equation, J. Comput. Phys.,
+!! Finite-Volume Scheme for the Advection-Diffusion Equation, J. Compute. Phys.,
!! 181, 729-752
!!
SUBROUTINE recon_lsq_cell_q_svd_lib(p_cc, lsq_idx_c, lsq_blk_c, &
@@ -1008,7 +1008,7 @@ CONTAINS
!!
!! !LITERATURE
!! Ollivier-Gooch et al (2002): A High-Order-Accurate Unstructured Mesh
-!! Finite-Volume Scheme for the Advection-Diffusion Equation, J. Comput. Phys.,
+!! Finite-Volume Scheme for the Advection-Diffusion Equation, J. Compute. Phys.,
!! 181, 729-752
!!
SUBROUTINE recon_lsq_cell_c_lib(p_cc, lsq_idx_c, lsq_blk_c, &
@@ -1295,7 +1295,7 @@ CONTAINS
!!
!! !LITERATURE
!! Ollivier-Gooch et al (2002): A High-Order-Accurate Unstructured Mesh
-!! Finite-Volume Scheme for the Advection-Diffusion Equation, J. Comput. Phys.,
+!! Finite-Volume Scheme for the Advection-Diffusion Equation, J. Compute. Phys.,
!! 181, 729-752
!!
SUBROUTINE recon_lsq_cell_c_svd_lib(p_cc, lsq_idx_c, lsq_blk_c, &
--
GitLab
From 3afe4ffe25a1621a662d055f21b04aaf8f9e3fbb Mon Sep 17 00:00:00 2001
From: Ali Sedighi <k202194@levante0.lvt.dkrz.de>
Date: Wed, 5 Mar 2025 14:57:12 +0100
Subject: [PATCH 24/33] Added test for edge_lib
---
src/interpolation/mo_lib_intp_rbf.cpp | 22 ++++++------
test/c/test_intp_rbf.cpp | 52 +++++++++++++++++++++++++++
2 files changed, 63 insertions(+), 11 deletions(-)
diff --git a/src/interpolation/mo_lib_intp_rbf.cpp b/src/interpolation/mo_lib_intp_rbf.cpp
index bdbfbe5..76e4784 100644
--- a/src/interpolation/mo_lib_intp_rbf.cpp
+++ b/src/interpolation/mo_lib_intp_rbf.cpp
@@ -154,23 +154,23 @@ void rbf_vec_interpol_cell_lib(const T* p_vn_in, const int* rbf_vec_idx_c, const
p_vn_in_view(rbf_vec_idx_c_view(8, jc, jb), jk, rbf_vec_blk_c_view(8, jc, jb));
p_v_out_view(jc, jk, jb) =
- rbf_vec_coeff_c_view(0, 1, jc, jb)*
+ rbf_vec_coeff_c_view(0, 2, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(0, jc, jb), jk, rbf_vec_blk_c_view(0, jc, jb)) +
- rbf_vec_coeff_c_view(1, 1, jc, jb)*
+ rbf_vec_coeff_c_view(1, 2, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(1, jc, jb), jk, rbf_vec_blk_c_view(1, jc, jb)) +
- rbf_vec_coeff_c_view(2, 1, jc, jb)*
+ rbf_vec_coeff_c_view(2, 2, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(2, jc, jb), jk, rbf_vec_blk_c_view(2, jc, jb)) +
- rbf_vec_coeff_c_view(3, 1, jc, jb)*
+ rbf_vec_coeff_c_view(3, 2, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(3, jc, jb), jk, rbf_vec_blk_c_view(3, jc, jb)) +
- rbf_vec_coeff_c_view(4, 1, jc, jb)*
+ rbf_vec_coeff_c_view(4, 2, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(4, jc, jb), jk, rbf_vec_blk_c_view(4, jc, jb)) +
- rbf_vec_coeff_c_view(5, 1, jc, jb)*
+ rbf_vec_coeff_c_view(5, 2, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(5, jc, jb), jk, rbf_vec_blk_c_view(5, jc, jb)) +
- rbf_vec_coeff_c_view(6, 1, jc, jb)*
+ rbf_vec_coeff_c_view(6, 2, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(6, jc, jb), jk, rbf_vec_blk_c_view(6, jc, jb)) +
- rbf_vec_coeff_c_view(7, 1, jc, jb)*
+ rbf_vec_coeff_c_view(7, 2, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(7, jc, jb), jk, rbf_vec_blk_c_view(7, jc, jb)) +
- rbf_vec_coeff_c_view(8, 1, jc, jb)*
+ rbf_vec_coeff_c_view(8, 2, jc, jb)*
p_vn_in_view(rbf_vec_idx_c_view(8, jc, jb), jk, rbf_vec_blk_c_view(8, jc, jb));
});
Kokkos::fence();
@@ -210,13 +210,13 @@ void rbf_vec_interpol_edge_lib(const T* p_vn_in, const int* rbf_vec_idx_e, const
UnmanagedT3D p_vt_out_view(p_vt_out, nproma,nlev,nblks_e);
- for (int jb = i_startblk; jb < i_endblk; ++jb) {
+ for (int jb = i_startblk; jb <= i_endblk; ++jb) {
int i_startidx, i_endidx;
get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk, i_startidx, i_endidx);
Kokkos::parallel_for("rbf_vec_interpol_edge_lib",
- Kokkos::MDRangePolicy<Kokkos::Rank<2>>({slev, i_startidx}, {elev, i_endidx}),
+ Kokkos::MDRangePolicy<Kokkos::Rank<2>>({slev, i_startidx}, {elev+1, i_endidx+1}),
KOKKOS_LAMBDA(const int jk, const int je){
p_vt_out_view(je, jk, jb) =
diff --git a/test/c/test_intp_rbf.cpp b/test/c/test_intp_rbf.cpp
index 9414061..426a86b 100644
--- a/test/c/test_intp_rbf.cpp
+++ b/test/c/test_intp_rbf.cpp
@@ -51,6 +51,7 @@ public:
static constexpr int nblks_e = 2;
static constexpr int rbf_c2grad_dim = 10; // fixed dimension for the RBF c2 gradient
static constexpr int rbf_vec_dim_c = 9;
+ static constexpr int rbf_vec_dim_e = 4;
// Parameters for the index ranges.
int i_startblk = 0;
@@ -76,6 +77,12 @@ public:
std::vector<T> p_u_out;
std::vector<T> p_v_out;
+ std::vector<int> rbf_vec_idx_e;
+ std::vector<int> rbf_vec_blk_e;
+ std::vector<T> rbf_vec_coeff_e;
+ std::vector<T> p_vt_out;
+
+
RbfInterpolTypedTestFixture() {
size_t size3d = static_cast<size_t>(nproma) * nlev * nblks_c;
size_t size3d_idx = static_cast<size_t>(rbf_c2grad_dim) * nproma * nblks_c;
@@ -84,6 +91,9 @@ public:
size_t size3d_vec_dim = static_cast<size_t>(rbf_vec_dim_c) * nproma * nblks_c;
size_t size_4d_vec_dim = static_cast<size_t>(rbf_vec_dim_c) * 2 * nproma * nblks_c;
+ size_t size3d_edge_lib = static_cast<size_t>(rbf_vec_dim_e) * nproma * nblks_c;
+ size_t size_4d_edge_lib = static_cast<size_t>(rbf_vec_dim_e) * 2 * nproma * nblks_c;
+
p_cell_in.resize(size3d, static_cast<T>(1));
p_vn_in.resize(size3d, static_cast<T>(1));
@@ -91,12 +101,17 @@ public:
rbf_vec_blk_c.resize(size3d_vec_dim, 0);
rbf_c2grad_idx.resize(size3d_idx, 1);
rbf_c2grad_blk.resize(size3d_idx, 0); // Set block indices to 0 for testing.
+ rbf_vec_idx_e.resize(size3d_vec_dim, 1);
+ rbf_vec_blk_e.resize(size3d_vec_dim, 0);
rbf_vec_coeff_c.resize(size_4d_vec_dim, static_cast<T>(1));
rbf_c2grad_coeff.resize(size4d, static_cast<T>(1));
+ rbf_vec_coeff_e.resize(size_4d_edge_lib, static_cast<T>(1));
p_u_out.resize(size3d_vec_dim, static_cast<T>(0));
p_v_out.resize(size3d_vec_dim, static_cast<T>(0));
+ p_vt_out.resize(size3d_edge_lib, static_cast<T>(0));
+
grad_x.resize(size3d, static_cast<T>(0));
grad_y.resize(size3d, static_cast<T>(0));
}
@@ -181,6 +196,43 @@ TYPED_TEST(RbfInterpolTypedTestFixture, cell_lib) {
}
}
+
+TYPED_TEST(RbfInterpolTypedTestFixture, edge_lib){
+ using T = TypeParam;
+
+
+ rbf_vec_interpol_edge_lib<T>(
+ this->p_vn_in.data(),
+ this->rbf_vec_idx_e.data(),
+ this->rbf_vec_blk_e.data(),
+ this->rbf_vec_coeff_e.data(),
+ this->p_vt_out.data(),
+ this->i_startblk,
+ this->i_endblk,
+ this->i_startidx_in,
+ this->i_endidx_in,
+ this->slev,
+ this->elev,
+ this->nlev,
+ this->nproma,
+ this->rbf_vec_dim_e,
+ this->nblks_e,
+ this->lacc,
+ this->acc_async);
+
+ for (int jb = this->i_startblk; jb < this->i_endblk; ++jb) {
+ for (int jk = 0; jk < this->nlev; ++jk) {
+ for (int i = 0; i < this->nproma; ++i) {
+ size_t idx = i + static_cast<size_t>(jk) * this->nproma + static_cast<size_t>(jb) * this->nproma * this->nlev;
+ EXPECT_NEAR(this->p_vt_out[idx], static_cast<T>(this->rbf_vec_dim_e), static_cast<T>(1e-5))
+ << "p_vt_out failure at block " << jb << ", level " << jk << ", index " << i;
+ }
+ }
+ }
+
+}
+
+
// Define a typed test fixture for the functions which have different input and output types
template <typename TypePair>
class RbfVecInterpolMixedTestFixture : public ::testing::Test {
--
GitLab
From 9d414e59bb6ed9057b6335f7a5411f7eba352c30 Mon Sep 17 00:00:00 2001
From: Pradipta Samanta <samanta@dkrz.de>
Date: Thu, 6 Mar 2025 10:57:03 +0100
Subject: [PATCH 25/33] put the instantiations at the end of the file
---
src/interpolation/mo_lib_intp_rbf.cpp | 82 +++++++++++++++------------
1 file changed, 47 insertions(+), 35 deletions(-)
diff --git a/src/interpolation/mo_lib_intp_rbf.cpp b/src/interpolation/mo_lib_intp_rbf.cpp
index 76e4784..f1a9fa4 100644
--- a/src/interpolation/mo_lib_intp_rbf.cpp
+++ b/src/interpolation/mo_lib_intp_rbf.cpp
@@ -89,18 +89,6 @@ void rbf_interpol_c2grad_lib(const T* p_cell_in, const int* rbf_c2grad_idx, cons
}//for
}//void
-template
-void rbf_interpol_c2grad_lib<double>(const double* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const double* rbf_c2grad_coeff, double* grad_x, double* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblks_c, bool lacc);
-
-template
-void rbf_interpol_c2grad_lib<float>(const float* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const float* rbf_c2grad_coeff, float* grad_x, float* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblks_c, bool lacc);
-
//------------------------------------------rbf_vec_interpol_cell_lib---------------------------------------------
@@ -177,19 +165,6 @@ void rbf_vec_interpol_cell_lib(const T* p_vn_in, const int* rbf_vec_idx_c, const
}//for
}//void
-template
-void rbf_vec_interpol_cell_lib<double>(const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
- int rbf_vec_dim_c, bool lacc, bool acc_async);
-template
-void rbf_vec_interpol_cell_lib<float>(const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
- int rbf_vec_dim_c, bool lacc, bool acc_async);
-
//------------------------------------------rbf_vec_interpol_edge_lib---------------------------------------------
template <typename T>
@@ -230,16 +205,53 @@ void rbf_vec_interpol_edge_lib(const T* p_vn_in, const int* rbf_vec_idx_e, const
}
template
-void rbf_vec_interpol_edge_lib<double>(const double* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const double* rbf_vec_coeff_e, double* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async);
+void rbf_vec_interpol_cell_lib<double>(
+ const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+ const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
+ int rbf_vec_dim_c, bool lacc, bool acc_async
+);
+template
+void rbf_vec_interpol_cell_lib<float>(
+ const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+ const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
+ int rbf_vec_dim_c, bool lacc, bool acc_async
+);
+
+template
+void rbf_interpol_c2grad_lib<double>(
+ const double* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const double* rbf_c2grad_coeff, double* grad_x, double* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblks_c, bool lacc
+);
+
+template
+void rbf_interpol_c2grad_lib<float>(
+ const float* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const float* rbf_c2grad_coeff, float* grad_x, float* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblks_c, bool lacc
+);
+
+template
+void rbf_vec_interpol_edge_lib<double>(
+ const double* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
+ const double* rbf_vec_coeff_e, double* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async
+);
template
-void rbf_vec_interpol_edge_lib<float>(const float* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const float* rbf_vec_coeff_e, float* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async);
+void rbf_vec_interpol_edge_lib<float>(
+ const float* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
+ const float* rbf_vec_coeff_e, float* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async
+);
--
GitLab
From a3e149a6fc76fbc261f754c66e18ecb9912073ac Mon Sep 17 00:00:00 2001
From: Pradipta Samanta <samanta@dkrz.de>
Date: Thu, 6 Mar 2025 11:48:00 +0100
Subject: [PATCH 26/33] moved intp_rbf routines from the temporary file to the
main file
fixed a bug
---
src/interpolation/CMakeLists.txt | 2 -
...b_intp_rbf-rbf_vec_interpol_vertex_lib.cpp | 197 ------------------
...b_intp_rbf-rbf_vec_interpol_vertex_lib.hpp | 32 ---
...f-rbf_vec_interpol_vertex_lib_bindings.cpp | 134 ------------
...rbf-rbf_vec_interpol_vertex_lib_bindings.h | 54 -----
src/interpolation/mo_lib_intp_rbf.cpp | 176 ++++++++++++++++
src/interpolation/mo_lib_intp_rbf.hpp | 20 ++
.../mo_lib_intp_rbf_bindings.cpp | 118 +++++++++++
src/interpolation/mo_lib_intp_rbf_bindings.h | 38 ++++
test/c/test_intp_rbf.cpp | 1 -
10 files changed, 352 insertions(+), 420 deletions(-)
delete mode 100644 src/interpolation/mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib.cpp
delete mode 100644 src/interpolation/mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib.hpp
delete mode 100644 src/interpolation/mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib_bindings.cpp
delete mode 100644 src/interpolation/mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib_bindings.h
diff --git a/src/interpolation/CMakeLists.txt b/src/interpolation/CMakeLists.txt
index d5af767..bf5b0ac 100644
--- a/src/interpolation/CMakeLists.txt
+++ b/src/interpolation/CMakeLists.txt
@@ -16,8 +16,6 @@ add_library(
mo_lib_interpolation_vector.F90
mo_lib_interpolation_vector.cpp
mo_lib_intp_rbf.F90
- mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib.cpp
- mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib_bindings.cpp
interpolation_bindings.cpp
mo_lib_intp_rbf_bindings.cpp
mo_lib_intp_rbf.cpp)
diff --git a/src/interpolation/mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib.cpp b/src/interpolation/mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib.cpp
deleted file mode 100644
index c9b776e..0000000
--- a/src/interpolation/mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib.cpp
+++ /dev/null
@@ -1,197 +0,0 @@
-// ICON
-//
-// ---------------------------------------------------------------
-// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
-// Contact information: icon-model.org
-//
-// See AUTHORS.TXT for a list of authors
-// See LICENSES/ for license information
-// SPDX-License-Identifier: BSD-3-Clause
-// ---------------------------------------------------------------
-
-/// Contains the only mo_lib_intp_rbf::rbf_vec_interpol_vertex_lib()
-///
-/// Separate to avoid conflicts with Ali working on rest of mo_lib_intp_rbf
-
-#include <type_traits>
-#include <Kokkos_Core.hpp>
-#include "mo_lib_loopindices.hpp"
-#include "mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib.hpp"
-
-
-constexpr int rbf_vec_dim_v = 6;
-
-//-------------------------------------------------------------------------
-//
-//
-//>
-/// Performs vector RBF reconstruction at triangle vertices.
-///
-/// Theory described in Narcowich and Ward (Math Comp. 1994) and
-/// Bonaventura and Baudisch (Mox Report n. 75).
-/// It takes edge based variables as input and combines them
-/// into three dimensional cartesian vectors at each vertex.
-///
-/// Two templated variables in order to support mixed precision.
-/// Intended that type_traits::is_floating_point(T,S)==TRUE
-/// precision(T) >= precision(S)
-template <typename T, typename S>
-void rbf_vec_interpol_vertex_lib(
- const T* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const T* rbf_vec_coeff_v,
- S* p_u_out,
- S* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- // Dimensions for the arrays.
- const int nlev, const int nblks_e, const int nblks_v
- )
-{
- /*
-#ifdef DIM_ENABLE_GPU
- if (lacc){ using MemSpace = Kokkos::CudaSpace;
- } else { using MemSpace = Kokkos::HostSpace; }
-#else
- using MemSpace = Kokkos::HostSpace;
-#endif
-
- */
-
- // Wrap raw pointers in unmanaged Kokkos Views.
- typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
- typedef Kokkos::View<const T****, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT4D;
- typedef Kokkos::View<const int***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstInt3D;
- typedef Kokkos::View<S***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedS3D;
-
-
-
- // input components of velocity or horizontal vorticity vectors at edge midpoints
- // dim: (nproma,nlev,nblks_e)
- UnmanagedConstT3D p_e_in_view(p_e_in, nproma, nlev, nblks_e);
-
- // index array defining the stencil of surrounding edges for vector rbf interpolation at each triangle vertex
- // (rbf_vec_dim_v,nproma,nblks_v)
- UnmanagedConstInt3D iidx_view(rbf_vec_idx_v, rbf_vec_dim_v, nproma, nblks_v);
- UnmanagedConstInt3D iblk_view(rbf_vec_blk_v, rbf_vec_dim_v, nproma, nblks_v);
-
- // coefficients are working precision array containing the coefficients used for vector rbf interpolation
- // at each tringle vertex (input is normal component),
- // dim: (rbf_vec_dim_v,2,nproma,nblks_v)
- UnmanagedConstT4D ptr_coeff_view(rbf_vec_coeff_v, rbf_vec_dim_v, 2, nproma, nblks_v);
-
- // reconstructed x-component (u) of velocity vector,
- // dim: (nproma,nlev,nblks_v)
- UnmanagedS3D p_u_out_view(p_u_out, nproma, nlev, nblks_v);
- // reconstructed y-component (v) of velocity vector,
- // dim: (nproma,nlev,nblks_v)
- UnmanagedS3D p_v_out_view(p_v_out, nproma, nlev, nblks_v);
-
- // Local vars
- //int jv, jk, jb; // integer over vertices, levels, and blocks,
- int jb; // integer over vertices, levels, and blocks,
- int i_startidx; // start index
- int i_endidx; // end index
-
- for (jb=i_startblk; jb <= i_endblk; ++jb){
-
- get_indices_v_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk,
- i_startidx, i_endidx);
-
- Kokkos::MDRangePolicy<Kokkos::Rank<2>> innerPolicy(
- {slev, i_startidx}, {elev + 1, i_endidx + 1});
-
- Kokkos::parallel_for("rbf_vec_interpol_vertex_lib", innerPolicy,
- KOKKOS_LAMBDA(const int jk, const int jv) {
-
- // NOTE: Static indexes reduced by 1 from Fortran version
- p_u_out_view(jv, jk, jb) =
- ptr_coeff_view(0, 0, jv, jb)*p_e_in_view(iidx_view(0, jv, jb), jk, iblk_view(0, jv, jb)) +
- ptr_coeff_view(1, 0, jv, jb)*p_e_in_view(iidx_view(1, jv, jb), jk, iblk_view(1, jv, jb)) +
- ptr_coeff_view(2, 0, jv, jb)*p_e_in_view(iidx_view(2, jv, jb), jk, iblk_view(2, jv, jb)) +
- ptr_coeff_view(3, 0, jv, jb)*p_e_in_view(iidx_view(3, jv, jb), jk, iblk_view(3, jv, jb)) +
- ptr_coeff_view(4, 0, jv, jb)*p_e_in_view(iidx_view(4, jv, jb), jk, iblk_view(4, jv, jb)) +
- ptr_coeff_view(5, 0, jv, jb)*p_e_in_view(iidx_view(5, jv, jb), jk, iblk_view(5, jv, jb));
- p_v_out_view(jv, jk, jb) =
- ptr_coeff_view(0, 1, jv, jb)*p_e_in_view(iidx_view(0, jv, jb), jk, iblk_view(0, jv, jb)) +
- ptr_coeff_view(1, 1, jv, jb)*p_e_in_view(iidx_view(1, jv, jb), jk, iblk_view(1, jv, jb)) +
- ptr_coeff_view(2, 1, jv, jb)*p_e_in_view(iidx_view(2, jv, jb), jk, iblk_view(2, jv, jb)) +
- ptr_coeff_view(3, 1, jv, jb)*p_e_in_view(iidx_view(3, jv, jb), jk, iblk_view(3, jv, jb)) +
- ptr_coeff_view(4, 1, jv, jb)*p_e_in_view(iidx_view(4, jv, jb), jk, iblk_view(4, jv, jb)) +
- ptr_coeff_view(5, 1, jv, jb)*p_e_in_view(iidx_view(5, jv, jb), jk, iblk_view(5, jv, jb));
- }
- );
- }
-}
-
-// Explicit instantiation - double precision
-template
-void rbf_vec_interpol_vertex_lib<double, double>(
- const double* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const double* rbf_vec_coeff_v,
- double* p_u_out,
- double* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- );
-
-// Explicit instantiation - single precision
-template
-void rbf_vec_interpol_vertex_lib<float, float>(
- const float* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const float* rbf_vec_coeff_v,
- float* p_u_out,
- float* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- );
-
-// Explicit instantiation - mixed precision
-template
-void rbf_vec_interpol_vertex_lib<double, float>(
- const double* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const double* rbf_vec_coeff_v,
- float* p_u_out,
- float* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- );
-
diff --git a/src/interpolation/mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib.hpp b/src/interpolation/mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib.hpp
deleted file mode 100644
index c0b6f05..0000000
--- a/src/interpolation/mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib.hpp
+++ /dev/null
@@ -1,32 +0,0 @@
-// ICON
-//
-// ---------------------------------------------------------------
-// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
-// Contact information: icon-model.org
-//
-// See AUTHORS.TXT for a list of authors
-// See LICENSES/ for license information
-// SPDX-License-Identifier: BSD-3-Clause
-// ---------------------------------------------------------------
-
-#pragma once
-
-template <typename T, typename S>
-void rbf_vec_interpol_vertex_lib(
- const T* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const T* rbf_vec_coeff_v,
- S* p_u_out,
- S* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_c
- );
\ No newline at end of file
diff --git a/src/interpolation/mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib_bindings.cpp b/src/interpolation/mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib_bindings.cpp
deleted file mode 100644
index 06dc467..0000000
--- a/src/interpolation/mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib_bindings.cpp
+++ /dev/null
@@ -1,134 +0,0 @@
-// ICON
-//
-// ---------------------------------------------------------------
-// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
-// Contact information: icon-model.org
-//
-// See AUTHORS.TXT for a list of authors
-// See LICENSES/ for license information
-// SPDX-License-Identifier: BSD-3-Clause
-// ---------------------------------------------------------------
-
-#include "mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib_bindings.h"
-#include "mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib.hpp"
-
-void rbf_vec_interpol_vertex_lib_dp(
- const double* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const double* rbf_vec_coeff_v,
- double* p_u_out,
- double* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- )
-{
- rbf_vec_interpol_vertex_lib<double, double>(
- p_e_in,
- rbf_vec_idx_v,
- rbf_vec_blk_v,
- rbf_vec_coeff_v,
- p_u_out,
- p_v_out,
- i_startblk, // start_block needed for get_indices_c_lib
- i_endblk, // end_block needed for get_indices_c_lib
- i_startidx_in, // start_index needed for get_indices_c_lib
- i_endidx_in, // end_index needed for get_indices_c_lib
- slev, // vertical start level
- elev, // vertical end level
- nproma, // inner loop length/vector length
- lacc, // if true, use Cuda mem-/exec-spaces
- acc_async, // [deprecated] use async acc
- nlev, nblks_e, nblks_v
- );
-}
-
-
-void rbf_vec_interpol_vertex_lib_sp(
- const float* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const float* rbf_vec_coeff_v,
- float* p_u_out,
- float* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- )
-{
- rbf_vec_interpol_vertex_lib<float, float>(
- p_e_in,
- rbf_vec_idx_v,
- rbf_vec_blk_v,
- rbf_vec_coeff_v,
- p_u_out,
- p_v_out,
- i_startblk, // start_block needed for get_indices_c_lib
- i_endblk, // end_block needed for get_indices_c_lib
- i_startidx_in, // start_index needed for get_indices_c_lib
- i_endidx_in, // end_index needed for get_indices_c_lib
- slev, // vertical start level
- elev, // vertical end level
- nproma, // inner loop length/vector length
- lacc, // if true, use Cuda mem-/exec-spaces
- acc_async, // [deprecated] use async acc
- nlev, nblks_e, nblks_v
- );
-
-}
-
-void rbf_vec_interpol_vertex_lib_mixprec(
- const double* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const double* rbf_vec_coeff_v,
- float* p_u_out,
- float* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- )
-{
- rbf_vec_interpol_vertex_lib<double, float>(
- p_e_in,
- rbf_vec_idx_v,
- rbf_vec_blk_v,
- rbf_vec_coeff_v,
- p_u_out,
- p_v_out,
- i_startblk, // start_block needed for get_indices_c_lib
- i_endblk, // end_block needed for get_indices_c_lib
- i_startidx_in, // start_index needed for get_indices_c_lib
- i_endidx_in, // end_index needed for get_indices_c_lib
- slev, // vertical start level
- elev, // vertical end level
- nproma, // inner loop length/vector length
- lacc, // if true, use Cuda mem-/exec-spaces
- acc_async, // [deprecated] use async acc
- nlev, nblks_e, nblks_v
- );
-
-}
-
diff --git a/src/interpolation/mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib_bindings.h b/src/interpolation/mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib_bindings.h
deleted file mode 100644
index 4356f88..0000000
--- a/src/interpolation/mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib_bindings.h
+++ /dev/null
@@ -1,54 +0,0 @@
-// ICON
-//
-// ---------------------------------------------------------------
-// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
-// Contact information: icon-model.org
-//
-// See AUTHORS.TXT for a list of authors
-// See LICENSES/ for license information
-// SPDX-License-Identifier: BSD-3-Clause
-// ---------------------------------------------------------------
-
-#pragma once
-
-extern "C" {
-
-void rbf_vec_interpol_vertex_lib_dp(
- const double* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const double* rbf_vec_coeff_v,
- double* p_u_out,
- double* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- );
-
-void rbf_vec_interpol_vertex_lib_sp(
- const float* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const float* rbf_vec_coeff_v,
- float* p_u_out,
- float* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- );
-
-}
\ No newline at end of file
diff --git a/src/interpolation/mo_lib_intp_rbf.cpp b/src/interpolation/mo_lib_intp_rbf.cpp
index f1a9fa4..361710f 100644
--- a/src/interpolation/mo_lib_intp_rbf.cpp
+++ b/src/interpolation/mo_lib_intp_rbf.cpp
@@ -12,6 +12,119 @@
#include "mo_lib_intp_rbf.hpp"
#include <Kokkos_Core.hpp>
+constexpr int rbf_vec_dim_v = 6;
+
+//-------------------------------------------------------------------------
+//
+//
+//>
+/// Performs vector RBF reconstruction at triangle vertices.
+///
+/// Theory described in Narcowich and Ward (Math Comp. 1994) and
+/// Bonaventura and Baudisch (Mox Report n. 75).
+/// It takes edge based variables as input and combines them
+/// into three dimensional cartesian vectors at each vertex.
+///
+/// Two templated variables in order to support mixed precision.
+/// Intended that type_traits::is_floating_point(T,S)==TRUE
+/// precision(T) >= precision(S)
+template <typename T, typename S>
+void rbf_vec_interpol_vertex_lib(
+ const T* p_e_in,
+ const int* rbf_vec_idx_v,
+ const int* rbf_vec_blk_v,
+ const T* rbf_vec_coeff_v,
+ S* p_u_out,
+ S* p_v_out,
+ const int i_startblk, // start_block needed for get_indices_c_lib
+ const int i_endblk, // end_block needed for get_indices_c_lib
+ const int i_startidx_in, // start_index needed for get_indices_c_lib
+ const int i_endidx_in, // end_index needed for get_indices_c_lib
+ const int slev, // vertical start level
+ const int elev, // vertical end level
+ const int nproma, // inner loop length/vector length
+ const bool lacc, // if true, use Cuda mem-/exec-spaces
+ const bool acc_async, // [deprecated] use async acc
+ // Dimensions for the arrays.
+ const int nlev, const int nblks_e, const int nblks_v
+ )
+{
+ /*
+#ifdef DIM_ENABLE_GPU
+ if (lacc){ using MemSpace = Kokkos::CudaSpace;
+ } else { using MemSpace = Kokkos::HostSpace; }
+#else
+ using MemSpace = Kokkos::HostSpace;
+#endif
+
+ */
+
+ // Wrap raw pointers in unmanaged Kokkos Views.
+ typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
+ typedef Kokkos::View<const T****, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT4D;
+ typedef Kokkos::View<const int***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstInt3D;
+ typedef Kokkos::View<S***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedS3D;
+
+
+
+ // input components of velocity or horizontal vorticity vectors at edge midpoints
+ // dim: (nproma,nlev,nblks_e)
+ UnmanagedConstT3D p_e_in_view(p_e_in, nproma, nlev, nblks_e);
+
+ // index array defining the stencil of surrounding edges for vector rbf interpolation at each triangle vertex
+ // (rbf_vec_dim_v,nproma,nblks_v)
+ UnmanagedConstInt3D iidx_view(rbf_vec_idx_v, rbf_vec_dim_v, nproma, nblks_v);
+ UnmanagedConstInt3D iblk_view(rbf_vec_blk_v, rbf_vec_dim_v, nproma, nblks_v);
+
+ // coefficients are working precision array containing the coefficients used for vector rbf interpolation
+ // at each tringle vertex (input is normal component),
+ // dim: (rbf_vec_dim_v,2,nproma,nblks_v)
+ UnmanagedConstT4D ptr_coeff_view(rbf_vec_coeff_v, rbf_vec_dim_v, 2, nproma, nblks_v);
+
+ // reconstructed x-component (u) of velocity vector,
+ // dim: (nproma,nlev,nblks_v)
+ UnmanagedS3D p_u_out_view(p_u_out, nproma, nlev, nblks_v);
+ // reconstructed y-component (v) of velocity vector,
+ // dim: (nproma,nlev,nblks_v)
+ UnmanagedS3D p_v_out_view(p_v_out, nproma, nlev, nblks_v);
+
+ // Local vars
+ //int jv, jk, jb; // integer over vertices, levels, and blocks,
+ int jb; // integer over vertices, levels, and blocks,
+ int i_startidx; // start index
+ int i_endidx; // end index
+
+ for (jb=i_startblk; jb <= i_endblk; ++jb){
+
+ get_indices_v_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk,
+ i_startidx, i_endidx);
+
+ Kokkos::MDRangePolicy<Kokkos::Rank<2>> innerPolicy(
+ {slev, i_startidx}, {elev + 1, i_endidx + 1});
+
+ Kokkos::parallel_for("rbf_vec_interpol_vertex_lib", innerPolicy,
+ KOKKOS_LAMBDA(const int jk, const int jv) {
+
+ // NOTE: Static indexes reduced by 1 from Fortran version
+ p_u_out_view(jv, jk, jb) =
+ ptr_coeff_view(0, 0, jv, jb)*p_e_in_view(iidx_view(0, jv, jb), jk, iblk_view(0, jv, jb)) +
+ ptr_coeff_view(1, 0, jv, jb)*p_e_in_view(iidx_view(1, jv, jb), jk, iblk_view(1, jv, jb)) +
+ ptr_coeff_view(2, 0, jv, jb)*p_e_in_view(iidx_view(2, jv, jb), jk, iblk_view(2, jv, jb)) +
+ ptr_coeff_view(3, 0, jv, jb)*p_e_in_view(iidx_view(3, jv, jb), jk, iblk_view(3, jv, jb)) +
+ ptr_coeff_view(4, 0, jv, jb)*p_e_in_view(iidx_view(4, jv, jb), jk, iblk_view(4, jv, jb)) +
+ ptr_coeff_view(5, 0, jv, jb)*p_e_in_view(iidx_view(5, jv, jb), jk, iblk_view(5, jv, jb));
+ p_v_out_view(jv, jk, jb) =
+ ptr_coeff_view(0, 1, jv, jb)*p_e_in_view(iidx_view(0, jv, jb), jk, iblk_view(0, jv, jb)) +
+ ptr_coeff_view(1, 1, jv, jb)*p_e_in_view(iidx_view(1, jv, jb), jk, iblk_view(1, jv, jb)) +
+ ptr_coeff_view(2, 1, jv, jb)*p_e_in_view(iidx_view(2, jv, jb), jk, iblk_view(2, jv, jb)) +
+ ptr_coeff_view(3, 1, jv, jb)*p_e_in_view(iidx_view(3, jv, jb), jk, iblk_view(3, jv, jb)) +
+ ptr_coeff_view(4, 1, jv, jb)*p_e_in_view(iidx_view(4, jv, jb), jk, iblk_view(4, jv, jb)) +
+ ptr_coeff_view(5, 1, jv, jb)*p_e_in_view(iidx_view(5, jv, jb), jk, iblk_view(5, jv, jb));
+ }
+ );
+ }
+}
+
template <typename T>
void rbf_interpol_c2grad_lib(const T* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
const T* rbf_c2grad_coeff, T* grad_x, T* grad_y,
@@ -204,6 +317,69 @@ void rbf_vec_interpol_edge_lib(const T* p_vn_in, const int* rbf_vec_idx_e, const
}
}
+// Explicit instantiation - double precision
+template
+void rbf_vec_interpol_vertex_lib<double, double>(
+ const double* p_e_in,
+ const int* rbf_vec_idx_v,
+ const int* rbf_vec_blk_v,
+ const double* rbf_vec_coeff_v,
+ double* p_u_out,
+ double* p_v_out,
+ const int i_startblk, // start_block needed for get_indices_c_lib
+ const int i_endblk, // end_block needed for get_indices_c_lib
+ const int i_startidx_in, // start_index needed for get_indices_c_lib
+ const int i_endidx_in, // end_index needed for get_indices_c_lib
+ const int slev, // vertical start level
+ const int elev, // vertical end level
+ const int nproma, // inner loop length/vector length
+ const bool lacc, // if true, use Cuda mem-/exec-spaces
+ const bool acc_async, // [deprecated] use async acc
+ const int nlev, const int nblks_e, const int nblks_v
+ );
+
+// Explicit instantiation - single precision
+template
+void rbf_vec_interpol_vertex_lib<float, float>(
+ const float* p_e_in,
+ const int* rbf_vec_idx_v,
+ const int* rbf_vec_blk_v,
+ const float* rbf_vec_coeff_v,
+ float* p_u_out,
+ float* p_v_out,
+ const int i_startblk, // start_block needed for get_indices_c_lib
+ const int i_endblk, // end_block needed for get_indices_c_lib
+ const int i_startidx_in, // start_index needed for get_indices_c_lib
+ const int i_endidx_in, // end_index needed for get_indices_c_lib
+ const int slev, // vertical start level
+ const int elev, // vertical end level
+ const int nproma, // inner loop length/vector length
+ const bool lacc, // if true, use Cuda mem-/exec-spaces
+ const bool acc_async, // [deprecated] use async acc
+ const int nlev, const int nblks_e, const int nblks_v
+ );
+
+// Explicit instantiation - mixed precision
+template
+void rbf_vec_interpol_vertex_lib<double, float>(
+ const double* p_e_in,
+ const int* rbf_vec_idx_v,
+ const int* rbf_vec_blk_v,
+ const double* rbf_vec_coeff_v,
+ float* p_u_out,
+ float* p_v_out,
+ const int i_startblk, // start_block needed for get_indices_c_lib
+ const int i_endblk, // end_block needed for get_indices_c_lib
+ const int i_startidx_in, // start_index needed for get_indices_c_lib
+ const int i_endidx_in, // end_index needed for get_indices_c_lib
+ const int slev, // vertical start level
+ const int elev, // vertical end level
+ const int nproma, // inner loop length/vector length
+ const bool lacc, // if true, use Cuda mem-/exec-spaces
+ const bool acc_async, // [deprecated] use async acc
+ const int nlev, const int nblks_e, const int nblks_v
+ );
+
template
void rbf_vec_interpol_cell_lib<double>(
const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
diff --git a/src/interpolation/mo_lib_intp_rbf.hpp b/src/interpolation/mo_lib_intp_rbf.hpp
index 5b9d5e1..53547fc 100644
--- a/src/interpolation/mo_lib_intp_rbf.hpp
+++ b/src/interpolation/mo_lib_intp_rbf.hpp
@@ -14,6 +14,26 @@
#include <Kokkos_Core.hpp>
#include <vector>
+template <typename T, typename S>
+void rbf_vec_interpol_vertex_lib(
+ const T* p_e_in,
+ const int* rbf_vec_idx_v,
+ const int* rbf_vec_blk_v,
+ const T* rbf_vec_coeff_v,
+ S* p_u_out,
+ S* p_v_out,
+ const int i_startblk, // start_block needed for get_indices_c_lib
+ const int i_endblk, // end_block needed for get_indices_c_lib
+ const int i_startidx_in, // start_index needed for get_indices_c_lib
+ const int i_endidx_in, // end_index needed for get_indices_c_lib
+ const int slev, // vertical start level
+ const int elev, // vertical end level
+ const int nproma, // inner loop length/vector length
+ const bool lacc, // if true, use Cuda mem-/exec-spaces
+ const bool acc_async, // [deprecated] use async acc
+ const int nlev, const int nblks_e, const int nblks_c
+ );
+
template <typename T>
void rbf_interpol_c2grad_lib(const T* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
const T* rbf_c2grad_coeff, T* grad_x, T* grad_y,
diff --git a/src/interpolation/mo_lib_intp_rbf_bindings.cpp b/src/interpolation/mo_lib_intp_rbf_bindings.cpp
index b679619..e8461e6 100644
--- a/src/interpolation/mo_lib_intp_rbf_bindings.cpp
+++ b/src/interpolation/mo_lib_intp_rbf_bindings.cpp
@@ -13,6 +13,124 @@
#include "mo_lib_intp_rbf_bindings.h"
#include "mo_lib_intp_rbf.hpp"
+void rbf_vec_interpol_vertex_lib_dp(
+ const double* p_e_in,
+ const int* rbf_vec_idx_v,
+ const int* rbf_vec_blk_v,
+ const double* rbf_vec_coeff_v,
+ double* p_u_out,
+ double* p_v_out,
+ const int i_startblk, // start_block needed for get_indices_c_lib
+ const int i_endblk, // end_block needed for get_indices_c_lib
+ const int i_startidx_in, // start_index needed for get_indices_c_lib
+ const int i_endidx_in, // end_index needed for get_indices_c_lib
+ const int slev, // vertical start level
+ const int elev, // vertical end level
+ const int nproma, // inner loop length/vector length
+ const bool lacc, // if true, use Cuda mem-/exec-spaces
+ const bool acc_async, // [deprecated] use async acc
+ const int nlev, const int nblks_e, const int nblks_v
+ )
+{
+ rbf_vec_interpol_vertex_lib<double, double>(
+ p_e_in,
+ rbf_vec_idx_v,
+ rbf_vec_blk_v,
+ rbf_vec_coeff_v,
+ p_u_out,
+ p_v_out,
+ i_startblk, // start_block needed for get_indices_c_lib
+ i_endblk, // end_block needed for get_indices_c_lib
+ i_startidx_in, // start_index needed for get_indices_c_lib
+ i_endidx_in, // end_index needed for get_indices_c_lib
+ slev, // vertical start level
+ elev, // vertical end level
+ nproma, // inner loop length/vector length
+ lacc, // if true, use Cuda mem-/exec-spaces
+ acc_async, // [deprecated] use async acc
+ nlev, nblks_e, nblks_v
+ );
+}
+
+void rbf_vec_interpol_vertex_lib_sp(
+ const float* p_e_in,
+ const int* rbf_vec_idx_v,
+ const int* rbf_vec_blk_v,
+ const float* rbf_vec_coeff_v,
+ float* p_u_out,
+ float* p_v_out,
+ const int i_startblk, // start_block needed for get_indices_c_lib
+ const int i_endblk, // end_block needed for get_indices_c_lib
+ const int i_startidx_in, // start_index needed for get_indices_c_lib
+ const int i_endidx_in, // end_index needed for get_indices_c_lib
+ const int slev, // vertical start level
+ const int elev, // vertical end level
+ const int nproma, // inner loop length/vector length
+ const bool lacc, // if true, use Cuda mem-/exec-spaces
+ const bool acc_async, // [deprecated] use async acc
+ const int nlev, const int nblks_e, const int nblks_v
+ )
+{
+ rbf_vec_interpol_vertex_lib<float, float>(
+ p_e_in,
+ rbf_vec_idx_v,
+ rbf_vec_blk_v,
+ rbf_vec_coeff_v,
+ p_u_out,
+ p_v_out,
+ i_startblk, // start_block needed for get_indices_c_lib
+ i_endblk, // end_block needed for get_indices_c_lib
+ i_startidx_in, // start_index needed for get_indices_c_lib
+ i_endidx_in, // end_index needed for get_indices_c_lib
+ slev, // vertical start level
+ elev, // vertical end level
+ nproma, // inner loop length/vector length
+ lacc, // if true, use Cuda mem-/exec-spaces
+ acc_async, // [deprecated] use async acc
+ nlev, nblks_e, nblks_v
+ );
+
+}
+
+void rbf_vec_interpol_vertex_lib_mixprec(
+ const double* p_e_in,
+ const int* rbf_vec_idx_v,
+ const int* rbf_vec_blk_v,
+ const double* rbf_vec_coeff_v,
+ float* p_u_out,
+ float* p_v_out,
+ const int i_startblk, // start_block needed for get_indices_c_lib
+ const int i_endblk, // end_block needed for get_indices_c_lib
+ const int i_startidx_in, // start_index needed for get_indices_c_lib
+ const int i_endidx_in, // end_index needed for get_indices_c_lib
+ const int slev, // vertical start level
+ const int elev, // vertical end level
+ const int nproma, // inner loop length/vector length
+ const bool lacc, // if true, use Cuda mem-/exec-spaces
+ const bool acc_async, // [deprecated] use async acc
+ const int nlev, const int nblks_e, const int nblks_v
+ )
+{
+ rbf_vec_interpol_vertex_lib<double, float>(
+ p_e_in,
+ rbf_vec_idx_v,
+ rbf_vec_blk_v,
+ rbf_vec_coeff_v,
+ p_u_out,
+ p_v_out,
+ i_startblk, // start_block needed for get_indices_c_lib
+ i_endblk, // end_block needed for get_indices_c_lib
+ i_startidx_in, // start_index needed for get_indices_c_lib
+ i_endidx_in, // end_index needed for get_indices_c_lib
+ slev, // vertical start level
+ elev, // vertical end level
+ nproma, // inner loop length/vector length
+ lacc, // if true, use Cuda mem-/exec-spaces
+ acc_async, // [deprecated] use async acc
+ nlev, nblks_e, nblks_v
+ );
+
+}
void rbf_interpol_c2grad_lib_sp(const float* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
const float* rbf_c2grad_coeff, float* grad_x, float* grad_y,
diff --git a/src/interpolation/mo_lib_intp_rbf_bindings.h b/src/interpolation/mo_lib_intp_rbf_bindings.h
index cc1dd5c..fb8e87c 100644
--- a/src/interpolation/mo_lib_intp_rbf_bindings.h
+++ b/src/interpolation/mo_lib_intp_rbf_bindings.h
@@ -12,6 +12,44 @@
extern "C" {
+void rbf_vec_interpol_vertex_lib_dp(
+ const double* p_e_in,
+ const int* rbf_vec_idx_v,
+ const int* rbf_vec_blk_v,
+ const double* rbf_vec_coeff_v,
+ double* p_u_out,
+ double* p_v_out,
+ const int i_startblk, // start_block needed for get_indices_c_lib
+ const int i_endblk, // end_block needed for get_indices_c_lib
+ const int i_startidx_in, // start_index needed for get_indices_c_lib
+ const int i_endidx_in, // end_index needed for get_indices_c_lib
+ const int slev, // vertical start level
+ const int elev, // vertical end level
+ const int nproma, // inner loop length/vector length
+ const bool lacc, // if true, use Cuda mem-/exec-spaces
+ const bool acc_async, // [deprecated] use async acc
+ const int nlev, const int nblks_e, const int nblks_v
+ );
+
+void rbf_vec_interpol_vertex_lib_sp(
+ const float* p_e_in,
+ const int* rbf_vec_idx_v,
+ const int* rbf_vec_blk_v,
+ const float* rbf_vec_coeff_v,
+ float* p_u_out,
+ float* p_v_out,
+ const int i_startblk, // start_block needed for get_indices_c_lib
+ const int i_endblk, // end_block needed for get_indices_c_lib
+ const int i_startidx_in, // start_index needed for get_indices_c_lib
+ const int i_endidx_in, // end_index needed for get_indices_c_lib
+ const int slev, // vertical start level
+ const int elev, // vertical end level
+ const int nproma, // inner loop length/vector length
+ const bool lacc, // if true, use Cuda mem-/exec-spaces
+ const bool acc_async, // [deprecated] use async acc
+ const int nlev, const int nblks_e, const int nblks_v
+ );
+
void rbf_interpol_c2grad_lib_sp(
const float* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
const float* rbf_c2grad_coeff, float* grad_x, const float* grad_y,
diff --git a/test/c/test_intp_rbf.cpp b/test/c/test_intp_rbf.cpp
index 426a86b..924ff02 100644
--- a/test/c/test_intp_rbf.cpp
+++ b/test/c/test_intp_rbf.cpp
@@ -14,7 +14,6 @@
#include <vector>
#include <algorithm>
#include <numeric>
-#include "mo_lib_intp_rbf-rbf_vec_interpol_vertex_lib.hpp"
#include "mo_lib_intp_rbf.hpp"
// Free-function helpers for 3D and 4D array sizes (assumed column-major)
--
GitLab
From f0f195141b717ccf9c243161cacab52b114cf535 Mon Sep 17 00:00:00 2001
From: Pradipta Samanta <samanta@dkrz.de>
Date: Thu, 6 Mar 2025 12:15:27 +0100
Subject: [PATCH 27/33] deleted intp_rbf_bindings and merged its contents into
interpolation_bindings
---
src/interpolation/CMakeLists.txt | 5 +-
src/interpolation/interpolation_bindings.cpp | 191 ++++++++++++++++
src/interpolation/interpolation_bindings.h | 78 +++++++
.../mo_lib_intp_rbf_bindings.cpp | 204 ------------------
src/interpolation/mo_lib_intp_rbf_bindings.h | 92 --------
5 files changed, 271 insertions(+), 299 deletions(-)
delete mode 100644 src/interpolation/mo_lib_intp_rbf_bindings.cpp
delete mode 100644 src/interpolation/mo_lib_intp_rbf_bindings.h
diff --git a/src/interpolation/CMakeLists.txt b/src/interpolation/CMakeLists.txt
index bf5b0ac..96f281c 100644
--- a/src/interpolation/CMakeLists.txt
+++ b/src/interpolation/CMakeLists.txt
@@ -16,9 +16,8 @@ add_library(
mo_lib_interpolation_vector.F90
mo_lib_interpolation_vector.cpp
mo_lib_intp_rbf.F90
- interpolation_bindings.cpp
- mo_lib_intp_rbf_bindings.cpp
- mo_lib_intp_rbf.cpp)
+ mo_lib_intp_rbf.cpp
+ interpolation_bindings.cpp)
add_library(${PROJECT_NAME}::interpolation ALIAS iconmath-interpolation)
diff --git a/src/interpolation/interpolation_bindings.cpp b/src/interpolation/interpolation_bindings.cpp
index 628f411..75449d6 100644
--- a/src/interpolation/interpolation_bindings.cpp
+++ b/src/interpolation/interpolation_bindings.cpp
@@ -12,6 +12,7 @@
#include "interpolation_bindings.h"
#include "mo_lib_interpolation_scalar.hpp"
#include "mo_lib_interpolation_vector.hpp"
+#include "mo_lib_intp_rbf.hpp"
// This is the binding for mo_interpolation_vector::edges2cells_vector_lib
// (wp=dp)
@@ -326,3 +327,193 @@ void cell_avg_lib_sp(const float *psi_c, const int *cell_neighbor_idx,
avg_psi_c, i_startblk, i_endblk, i_startidx_in,
i_endidx_in, slev, elev, nproma, nlev, nblks_c, lacc);
}
+
+void rbf_vec_interpol_vertex_lib_dp(
+ const double* p_e_in,
+ const int* rbf_vec_idx_v,
+ const int* rbf_vec_blk_v,
+ const double* rbf_vec_coeff_v,
+ double* p_u_out,
+ double* p_v_out,
+ const int i_startblk, // start_block needed for get_indices_c_lib
+ const int i_endblk, // end_block needed for get_indices_c_lib
+ const int i_startidx_in, // start_index needed for get_indices_c_lib
+ const int i_endidx_in, // end_index needed for get_indices_c_lib
+ const int slev, // vertical start level
+ const int elev, // vertical end level
+ const int nproma, // inner loop length/vector length
+ const bool lacc, // if true, use Cuda mem-/exec-spaces
+ const bool acc_async, // [deprecated] use async acc
+ const int nlev, const int nblks_e, const int nblks_v
+ )
+{
+ rbf_vec_interpol_vertex_lib<double, double>(
+ p_e_in,
+ rbf_vec_idx_v,
+ rbf_vec_blk_v,
+ rbf_vec_coeff_v,
+ p_u_out,
+ p_v_out,
+ i_startblk, // start_block needed for get_indices_c_lib
+ i_endblk, // end_block needed for get_indices_c_lib
+ i_startidx_in, // start_index needed for get_indices_c_lib
+ i_endidx_in, // end_index needed for get_indices_c_lib
+ slev, // vertical start level
+ elev, // vertical end level
+ nproma, // inner loop length/vector length
+ lacc, // if true, use Cuda mem-/exec-spaces
+ acc_async, // [deprecated] use async acc
+ nlev, nblks_e, nblks_v
+ );
+}
+
+void rbf_vec_interpol_vertex_lib_sp(
+ const float* p_e_in,
+ const int* rbf_vec_idx_v,
+ const int* rbf_vec_blk_v,
+ const float* rbf_vec_coeff_v,
+ float* p_u_out,
+ float* p_v_out,
+ const int i_startblk, // start_block needed for get_indices_c_lib
+ const int i_endblk, // end_block needed for get_indices_c_lib
+ const int i_startidx_in, // start_index needed for get_indices_c_lib
+ const int i_endidx_in, // end_index needed for get_indices_c_lib
+ const int slev, // vertical start level
+ const int elev, // vertical end level
+ const int nproma, // inner loop length/vector length
+ const bool lacc, // if true, use Cuda mem-/exec-spaces
+ const bool acc_async, // [deprecated] use async acc
+ const int nlev, const int nblks_e, const int nblks_v
+ )
+{
+ rbf_vec_interpol_vertex_lib<float, float>(
+ p_e_in,
+ rbf_vec_idx_v,
+ rbf_vec_blk_v,
+ rbf_vec_coeff_v,
+ p_u_out,
+ p_v_out,
+ i_startblk, // start_block needed for get_indices_c_lib
+ i_endblk, // end_block needed for get_indices_c_lib
+ i_startidx_in, // start_index needed for get_indices_c_lib
+ i_endidx_in, // end_index needed for get_indices_c_lib
+ slev, // vertical start level
+ elev, // vertical end level
+ nproma, // inner loop length/vector length
+ lacc, // if true, use Cuda mem-/exec-spaces
+ acc_async, // [deprecated] use async acc
+ nlev, nblks_e, nblks_v
+ );
+
+}
+
+void rbf_vec_interpol_vertex_lib_mixprec(
+ const double* p_e_in,
+ const int* rbf_vec_idx_v,
+ const int* rbf_vec_blk_v,
+ const double* rbf_vec_coeff_v,
+ float* p_u_out,
+ float* p_v_out,
+ const int i_startblk, // start_block needed for get_indices_c_lib
+ const int i_endblk, // end_block needed for get_indices_c_lib
+ const int i_startidx_in, // start_index needed for get_indices_c_lib
+ const int i_endidx_in, // end_index needed for get_indices_c_lib
+ const int slev, // vertical start level
+ const int elev, // vertical end level
+ const int nproma, // inner loop length/vector length
+ const bool lacc, // if true, use Cuda mem-/exec-spaces
+ const bool acc_async, // [deprecated] use async acc
+ const int nlev, const int nblks_e, const int nblks_v
+ )
+{
+ rbf_vec_interpol_vertex_lib<double, float>(
+ p_e_in,
+ rbf_vec_idx_v,
+ rbf_vec_blk_v,
+ rbf_vec_coeff_v,
+ p_u_out,
+ p_v_out,
+ i_startblk, // start_block needed for get_indices_c_lib
+ i_endblk, // end_block needed for get_indices_c_lib
+ i_startidx_in, // start_index needed for get_indices_c_lib
+ i_endidx_in, // end_index needed for get_indices_c_lib
+ slev, // vertical start level
+ elev, // vertical end level
+ nproma, // inner loop length/vector length
+ lacc, // if true, use Cuda mem-/exec-spaces
+ acc_async, // [deprecated] use async acc
+ nlev, nblks_e, nblks_v
+ );
+
+}
+
+void rbf_interpol_c2grad_lib_sp(const float* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const float* rbf_c2grad_coeff, float* grad_x, float* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc){
+
+ rbf_interpol_c2grad_lib<float>(p_cell_in, rbf_c2grad_idx, rbf_c2grad_blk,
+ rbf_c2grad_coeff,
+ grad_x, grad_y, i_startblk, i_endblk,
+ i_startidx_in, i_endidx_in, slev, elev,
+ nproma, rbf_c2grad_dim, nlev, nblk_c, lacc);
+}
+
+void rbf_interpol_c2grad_lib_dp(const double* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const double* rbf_c2grad_coeff, double* grad_x, double* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc){
+
+ rbf_interpol_c2grad_lib<double>(p_cell_in, rbf_c2grad_idx, rbf_c2grad_blk,
+ rbf_c2grad_coeff,
+ grad_x, grad_y, i_startblk, i_endblk,
+ i_startidx_in, i_endidx_in, slev, elev,
+ nproma, rbf_c2grad_dim, nlev, nblk_c, lacc);
+}
+
+void rbf_vec_interpol_cell_lib_sp(const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+ const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
+ int rbf_vec_dim_c, bool lacc, bool acc_async){
+
+ rbf_vec_interpol_cell_lib<float>(p_vn_in, rbf_vec_idx_c, rbf_vec_blk_c,
+ rbf_vec_coeff_c, p_u_out, p_v_out,
+ i_startblk, i_endblk, i_startidx_in, i_endidx_in,
+ slev, elev, nproma, nlev, nblks_c, nblks_e, rbf_vec_dim_c, lacc, acc_async);
+}
+
+void rbf_vec_interpol_cell_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+ const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
+ int rbf_vec_dim_c, bool lacc, bool acc_async){
+
+ rbf_vec_interpol_cell_lib<double>(p_vn_in, rbf_vec_idx_c, rbf_vec_blk_c, rbf_vec_coeff_c,
+ p_u_out, p_v_out, i_startblk, i_endblk, i_startidx_in, i_endidx_in,
+ slev, elev, nproma, nlev, nblks_c, nblks_e, rbf_vec_dim_c, lacc, acc_async);
+}
+
+void rbf_vec_interpol_edge_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
+ const double* rbf_vec_coeff_e, double* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async){
+
+ rbf_vec_interpol_edge_lib<double>(p_vn_in, rbf_vec_idx_e, rbf_vec_blk_e,
+ rbf_vec_coeff_e, p_vt_out,
+ i_startblk, i_endblk, i_startidx_in,i_endidx_in,
+ slev, elev, nlev, nproma, rbf_vec_dim_e, nblks_e, lacc, acc_async);
+}
+
+void rbf_vec_interpol_edge_lib_sp(const float* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
+ const float* rbf_vec_coeff_e, float* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async){
+
+ rbf_vec_interpol_edge_lib<float>(p_vn_in, rbf_vec_idx_e, rbf_vec_blk_e,
+ rbf_vec_coeff_e, p_vt_out,
+ i_startblk, i_endblk, i_startidx_in, i_endidx_in,
+ slev, elev, nlev, nproma, rbf_vec_dim_e, nblks_e,lacc, acc_async);
+}
diff --git a/src/interpolation/interpolation_bindings.h b/src/interpolation/interpolation_bindings.h
index 7cb873d..419446a 100644
--- a/src/interpolation/interpolation_bindings.h
+++ b/src/interpolation/interpolation_bindings.h
@@ -185,4 +185,82 @@ void cell_avg_lib_sp(const float *psi_c, const int *cell_neighbor_idx,
const int i_startidx_in, const int i_endidx_in,
const int slev, const int elev, const int nproma,
const int nlev, const int nblks_c, const bool lacc);
+
+void rbf_vec_interpol_vertex_lib_dp(
+ const double* p_e_in,
+ const int* rbf_vec_idx_v,
+ const int* rbf_vec_blk_v,
+ const double* rbf_vec_coeff_v,
+ double* p_u_out,
+ double* p_v_out,
+ const int i_startblk, // start_block needed for get_indices_c_lib
+ const int i_endblk, // end_block needed for get_indices_c_lib
+ const int i_startidx_in, // start_index needed for get_indices_c_lib
+ const int i_endidx_in, // end_index needed for get_indices_c_lib
+ const int slev, // vertical start level
+ const int elev, // vertical end level
+ const int nproma, // inner loop length/vector length
+ const bool lacc, // if true, use Cuda mem-/exec-spaces
+ const bool acc_async, // [deprecated] use async acc
+ const int nlev, const int nblks_e, const int nblks_v
+ );
+
+void rbf_vec_interpol_vertex_lib_sp(
+ const float* p_e_in,
+ const int* rbf_vec_idx_v,
+ const int* rbf_vec_blk_v,
+ const float* rbf_vec_coeff_v,
+ float* p_u_out,
+ float* p_v_out,
+ const int i_startblk, // start_block needed for get_indices_c_lib
+ const int i_endblk, // end_block needed for get_indices_c_lib
+ const int i_startidx_in, // start_index needed for get_indices_c_lib
+ const int i_endidx_in, // end_index needed for get_indices_c_lib
+ const int slev, // vertical start level
+ const int elev, // vertical end level
+ const int nproma, // inner loop length/vector length
+ const bool lacc, // if true, use Cuda mem-/exec-spaces
+ const bool acc_async, // [deprecated] use async acc
+ const int nlev, const int nblks_e, const int nblks_v
+ );
+
+void rbf_interpol_c2grad_lib_sp(
+ const float* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const float* rbf_c2grad_coeff, float* grad_x, const float* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
+
+void rbf_interpol_c2grad_lib_dp(
+ const double* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
+ const double* rbf_c2grad_coeff, double* grad_x, const double* grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
+
+void rbf_vec_interpol_cell_lib_sp(
+ const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+ const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
+ int rbf_vec_dim_c, bool lacc, bool acc_async);
+
+void rbf_vec_interpol_cell_lib_dp(
+ const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
+ const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
+ int rbf_vec_dim_c, bool lacc, bool acc_async);
+
+void rbf_vec_interpol_edge_lib_dp(
+ const double* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
+ const double* rbf_vec_coeff_e, double* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async);
+
+void rbf_vec_interpol_edge_lib_sp(
+ const float* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
+ const float* rbf_vec_coeff_e, float* p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
+ int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async);
}
diff --git a/src/interpolation/mo_lib_intp_rbf_bindings.cpp b/src/interpolation/mo_lib_intp_rbf_bindings.cpp
deleted file mode 100644
index e8461e6..0000000
--- a/src/interpolation/mo_lib_intp_rbf_bindings.cpp
+++ /dev/null
@@ -1,204 +0,0 @@
-// ICON
-//
-// ---------------------------------------------------------------
-// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
-// Contact information: icon-model.org
-//
-// See AUTHORS.TXT for a list of authors
-// See LICENSES/ for license information
-// SPDX-License-Identifier: BSD-3-Clause
-// ---------------------------------------------------------------
-
-
-#include "mo_lib_intp_rbf_bindings.h"
-#include "mo_lib_intp_rbf.hpp"
-
-void rbf_vec_interpol_vertex_lib_dp(
- const double* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const double* rbf_vec_coeff_v,
- double* p_u_out,
- double* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- )
-{
- rbf_vec_interpol_vertex_lib<double, double>(
- p_e_in,
- rbf_vec_idx_v,
- rbf_vec_blk_v,
- rbf_vec_coeff_v,
- p_u_out,
- p_v_out,
- i_startblk, // start_block needed for get_indices_c_lib
- i_endblk, // end_block needed for get_indices_c_lib
- i_startidx_in, // start_index needed for get_indices_c_lib
- i_endidx_in, // end_index needed for get_indices_c_lib
- slev, // vertical start level
- elev, // vertical end level
- nproma, // inner loop length/vector length
- lacc, // if true, use Cuda mem-/exec-spaces
- acc_async, // [deprecated] use async acc
- nlev, nblks_e, nblks_v
- );
-}
-
-void rbf_vec_interpol_vertex_lib_sp(
- const float* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const float* rbf_vec_coeff_v,
- float* p_u_out,
- float* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- )
-{
- rbf_vec_interpol_vertex_lib<float, float>(
- p_e_in,
- rbf_vec_idx_v,
- rbf_vec_blk_v,
- rbf_vec_coeff_v,
- p_u_out,
- p_v_out,
- i_startblk, // start_block needed for get_indices_c_lib
- i_endblk, // end_block needed for get_indices_c_lib
- i_startidx_in, // start_index needed for get_indices_c_lib
- i_endidx_in, // end_index needed for get_indices_c_lib
- slev, // vertical start level
- elev, // vertical end level
- nproma, // inner loop length/vector length
- lacc, // if true, use Cuda mem-/exec-spaces
- acc_async, // [deprecated] use async acc
- nlev, nblks_e, nblks_v
- );
-
-}
-
-void rbf_vec_interpol_vertex_lib_mixprec(
- const double* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const double* rbf_vec_coeff_v,
- float* p_u_out,
- float* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- )
-{
- rbf_vec_interpol_vertex_lib<double, float>(
- p_e_in,
- rbf_vec_idx_v,
- rbf_vec_blk_v,
- rbf_vec_coeff_v,
- p_u_out,
- p_v_out,
- i_startblk, // start_block needed for get_indices_c_lib
- i_endblk, // end_block needed for get_indices_c_lib
- i_startidx_in, // start_index needed for get_indices_c_lib
- i_endidx_in, // end_index needed for get_indices_c_lib
- slev, // vertical start level
- elev, // vertical end level
- nproma, // inner loop length/vector length
- lacc, // if true, use Cuda mem-/exec-spaces
- acc_async, // [deprecated] use async acc
- nlev, nblks_e, nblks_v
- );
-
-}
-
-void rbf_interpol_c2grad_lib_sp(const float* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const float* rbf_c2grad_coeff, float* grad_x, float* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc){
-
- rbf_interpol_c2grad_lib<float>(p_cell_in, rbf_c2grad_idx, rbf_c2grad_blk,
- rbf_c2grad_coeff,
- grad_x, grad_y, i_startblk, i_endblk,
- i_startidx_in, i_endidx_in, slev, elev,
- nproma, rbf_c2grad_dim, nlev, nblk_c, lacc);
-}
-
-void rbf_interpol_c2grad_lib_dp(const double* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const double* rbf_c2grad_coeff, double* grad_x, double* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc){
-
- rbf_interpol_c2grad_lib<double>(p_cell_in, rbf_c2grad_idx, rbf_c2grad_blk,
- rbf_c2grad_coeff,
- grad_x, grad_y, i_startblk, i_endblk,
- i_startidx_in, i_endidx_in, slev, elev,
- nproma, rbf_c2grad_dim, nlev, nblk_c, lacc);
-}
-
-void rbf_vec_interpol_cell_lib_sp(const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
- int rbf_vec_dim_c, bool lacc, bool acc_async){
-
- rbf_vec_interpol_cell_lib<float>(p_vn_in, rbf_vec_idx_c, rbf_vec_blk_c,
- rbf_vec_coeff_c, p_u_out, p_v_out,
- i_startblk, i_endblk, i_startidx_in, i_endidx_in,
- slev, elev, nproma, nlev, nblks_c, nblks_e, rbf_vec_dim_c, lacc, acc_async);
-}
-
-void rbf_vec_interpol_cell_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
- int rbf_vec_dim_c, bool lacc, bool acc_async){
-
- rbf_vec_interpol_cell_lib<double>(p_vn_in, rbf_vec_idx_c, rbf_vec_blk_c, rbf_vec_coeff_c,
- p_u_out, p_v_out, i_startblk, i_endblk, i_startidx_in, i_endidx_in,
- slev, elev, nproma, nlev, nblks_c, nblks_e, rbf_vec_dim_c, lacc, acc_async);
-}
-
-void rbf_vec_interpol_edge_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const double* rbf_vec_coeff_e, double* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async){
-
- rbf_vec_interpol_edge_lib<double>(p_vn_in, rbf_vec_idx_e, rbf_vec_blk_e,
- rbf_vec_coeff_e, p_vt_out,
- i_startblk, i_endblk, i_startidx_in,i_endidx_in,
- slev, elev, nlev, nproma, rbf_vec_dim_e, nblks_e, lacc, acc_async);
-}
-
-void rbf_vec_interpol_edge_lib_sp(const float* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const float* rbf_vec_coeff_e, float* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async){
-
- rbf_vec_interpol_edge_lib<float>(p_vn_in, rbf_vec_idx_e, rbf_vec_blk_e,
- rbf_vec_coeff_e, p_vt_out,
- i_startblk, i_endblk, i_startidx_in, i_endidx_in,
- slev, elev, nlev, nproma, rbf_vec_dim_e, nblks_e,lacc, acc_async);
-}
diff --git a/src/interpolation/mo_lib_intp_rbf_bindings.h b/src/interpolation/mo_lib_intp_rbf_bindings.h
deleted file mode 100644
index fb8e87c..0000000
--- a/src/interpolation/mo_lib_intp_rbf_bindings.h
+++ /dev/null
@@ -1,92 +0,0 @@
-// ICON
-//
-// ---------------------------------------------------------------
-// Copyright (C) 2004-2024, DWD, MPI-M, DKRZ, KIT, ETH, MeteoSwiss
-// Contact information: icon-model.org
-//
-// See AUTHORS.TXT for a list of authors
-// See LICENSES/ for license information
-// SPDX-License-Identifier: BSD-3-Clause
-// ---------------------------------------------------------------
-#pragma once
-
-extern "C" {
-
-void rbf_vec_interpol_vertex_lib_dp(
- const double* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const double* rbf_vec_coeff_v,
- double* p_u_out,
- double* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- );
-
-void rbf_vec_interpol_vertex_lib_sp(
- const float* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const float* rbf_vec_coeff_v,
- float* p_u_out,
- float* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- );
-
-void rbf_interpol_c2grad_lib_sp(
- const float* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const float* rbf_c2grad_coeff, float* grad_x, const float* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
-
-void rbf_interpol_c2grad_lib_dp(
- const double* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const double* rbf_c2grad_coeff, double* grad_x, const double* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
-
-void rbf_vec_interpol_cell_lib_sp(
- const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
- int rbf_vec_dim_c, bool lacc, bool acc_async);
-
-void rbf_vec_interpol_cell_lib_dp(
- const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
- int rbf_vec_dim_c, bool lacc, bool acc_async);
-
-void rbf_vec_interpol_edge_lib_dp(
- const double* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const double* rbf_vec_coeff_e, double* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async);
-
-void rbf_vec_interpol_edge_lib_sp(
- const float* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const float* rbf_vec_coeff_e, float* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async);
-}
--
GitLab
From ed5c21b6247e11166e6d27b3bc5d3741a5664117 Mon Sep 17 00:00:00 2001
From: Pradipta Samanta <samanta@dkrz.de>
Date: Thu, 6 Mar 2025 14:15:28 +0100
Subject: [PATCH 28/33] made changes to test_intp_rbf
---
test/c/test_intp_rbf.cpp | 70 +++++++++++++++++++++-------------------
1 file changed, 36 insertions(+), 34 deletions(-)
diff --git a/test/c/test_intp_rbf.cpp b/test/c/test_intp_rbf.cpp
index 924ff02..983352c 100644
--- a/test/c/test_intp_rbf.cpp
+++ b/test/c/test_intp_rbf.cpp
@@ -39,38 +39,43 @@ typedef ::testing::Types< MixedPrecision<double, double>,
MixedPrecision<double, float>,
MixedPrecision<float, float> > MixedTypes;
-// Define a typed test fixture for the functions which have the same input and output types
-template <typename T>
-class RbfInterpolTypedTestFixture : public ::testing::Test {
+class interp_dimensions {
public:
-
- static constexpr int nproma = 3; // inner loop length
- static constexpr int nlev = 4; // vertical levels
- static constexpr int nblks_c = 2; // number of blocks for the cell arrays
- static constexpr int nblks_e = 2;
- static constexpr int rbf_c2grad_dim = 10; // fixed dimension for the RBF c2 gradient
+ // Constant dimensions.
+ static constexpr int nproma = 3; // inner loop length
+ static constexpr int nlev = 4; // number of vertical levels
+ static constexpr int nblks_c = 2; // number of cell blocks
+ static constexpr int nblks_e = 2; // number of edge blocks
+ static constexpr int nblks_v = 2; // number of vertex blocks
+ static constexpr int rbf_c2grad_dim = 10; // fixed dimension
static constexpr int rbf_vec_dim_c = 9;
static constexpr int rbf_vec_dim_e = 4;
- // Parameters for the index ranges.
- int i_startblk = 0;
- int i_endblk = 1; // Note: loop over jb uses jb < i_endblk, so blocks 0 and 1
- int i_startidx_in = 0;
- int i_endidx_in = nproma-1; // [0, nproma) i.e. 0 .. nproma-1
- int slev = 0;
- int elev = nlev-1;
- bool lacc = false; // Not using any ACC-specific behavior.
- bool acc_async = false;
+ // Parameter values.
+ const int i_startblk = 0;
+ const int i_endblk = 1; // Test blocks [0, 1]
+ const int i_startidx_in = 0;
+ const int i_endidx_in = nproma - 1;
+ const int slev = 0;
+ const int elev = nlev - 1;
+ const bool lacc = false; // Not using ACC-specific behavior.
+ const bool acc_async = false; // No asynchronous execution.
+};
+
+// Define a typed test fixture for the functions which have the same input and output types
+template <typename T>
+class RbfInterpolTypedTestFixture : public ::testing::Test, public interp_dimensions {
+public:
// Data arrays.
- std::vector<T> p_cell_in; // size: nproma * nlev * nblks_c
- std::vector<int> rbf_c2grad_idx; // size: rbf_c2grad_dim * nproma * nblks_c
- std::vector<int> rbf_c2grad_blk; // size: rbf_c2grad_dim * nproma * nblks_c
- std::vector<int> rbf_vec_idx_c; //size: rbf_vec_dim_c * nproma * nblks_c
- std::vector<int> rbf_vec_blk_c; // size: rbf_vec_dim_c * nproma * nblks_c
- std::vector<T> rbf_c2grad_coeff; // size: rbf_c2grad_dim * 2 * nproma * nblks_c
- std::vector<T> grad_x; // size: nproma * nlev * nblks_c
- std::vector<T> grad_y; // size: nproma * nlev * nblks_c
+ std::vector<T> p_cell_in; // size: nproma * nlev * nblks_c
+ std::vector<int> rbf_c2grad_idx; // size: rbf_c2grad_dim * nproma * nblks_c
+ std::vector<int> rbf_c2grad_blk; // size: rbf_c2grad_dim * nproma * nblks_c
+ std::vector<int> rbf_vec_idx_c; //size: rbf_vec_dim_c * nproma * nblks_c
+ std::vector<int> rbf_vec_blk_c; // size: rbf_vec_dim_c * nproma * nblks_c
+ std::vector<T> rbf_c2grad_coeff; // size: rbf_c2grad_dim * 2 * nproma * nblks_c
+ std::vector<T> grad_x; // size: nproma * nlev * nblks_c
+ std::vector<T> grad_y; // size: nproma * nlev * nblks_c
std::vector<T> p_vn_in;
std::vector<T> rbf_vec_coeff_c;
std::vector<T> p_u_out;
@@ -120,7 +125,7 @@ typedef ::testing::Types<float, double> MyTypes;
TYPED_TEST_SUITE(RbfInterpolTypedTestFixture, MyTypes);
-TYPED_TEST(RbfInterpolTypedTestFixture, BasicTest) {
+TYPED_TEST(RbfInterpolTypedTestFixture, C2Grad) {
using T = TypeParam;
rbf_interpol_c2grad_lib<TypeParam>(
this->p_cell_in.data(),
@@ -158,7 +163,7 @@ TYPED_TEST(RbfInterpolTypedTestFixture, BasicTest) {
}
-TYPED_TEST(RbfInterpolTypedTestFixture, cell_lib) {
+TYPED_TEST(RbfInterpolTypedTestFixture, Cell) {
using T = TypeParam;
rbf_vec_interpol_cell_lib<T>(
@@ -188,15 +193,12 @@ TYPED_TEST(RbfInterpolTypedTestFixture, cell_lib) {
size_t idx = i + static_cast<size_t>(jk) * this->nproma + static_cast<size_t>(jb) * this->nproma * this->nlev;
EXPECT_NEAR(this->p_u_out[idx], static_cast<T>(this->rbf_vec_dim_c), static_cast<T>(1e-5))
<< "p_u_out failure at block " << jb << ", level " << jk << ", index " << i;
- // EXPECT_NEAR(this->p_v_out[idx], static_cast<T>(this->rbf_vec_dim_c), static_cast<T>(1e-5))
- // << "p_v_out failure at block " << jb << ", level " << jk << ", index " << i;
}
}
}
}
-
-TYPED_TEST(RbfInterpolTypedTestFixture, edge_lib){
+TYPED_TEST(RbfInterpolTypedTestFixture, Egde){
using T = TypeParam;
@@ -234,7 +236,7 @@ TYPED_TEST(RbfInterpolTypedTestFixture, edge_lib){
// Define a typed test fixture for the functions which have different input and output types
template <typename TypePair>
-class RbfVecInterpolMixedTestFixture : public ::testing::Test {
+class RbfVecInterpolMixedTestFixture : public ::testing::Test, public interp_dimensions {
public:
using InType = typename TypePair::in_type;
using OutType = typename TypePair::out_type;
@@ -279,7 +281,7 @@ public:
TYPED_TEST_SUITE(RbfVecInterpolMixedTestFixture, MixedTypes);
-TYPED_TEST(RbfVecInterpolMixedTestFixture, BasicTest) {
+TYPED_TEST(RbfVecInterpolMixedTestFixture, Vertex) {
using InType = typename TestFixture::InType;
using OutType = typename TestFixture::OutType;
--
GitLab
From eb4096e188f939d5dc311c095cc403a8257a27fe Mon Sep 17 00:00:00 2001
From: Pradipta Samanta <samanta@dkrz.de>
Date: Thu, 6 Mar 2025 14:19:33 +0100
Subject: [PATCH 29/33] made few changes to test_interpolation_scalar
---
test/c/test_interpolation_scalar.cpp | 9 ++++-----
1 file changed, 4 insertions(+), 5 deletions(-)
diff --git a/test/c/test_interpolation_scalar.cpp b/test/c/test_interpolation_scalar.cpp
index 0ee7fa3..507ec3f 100644
--- a/test/c/test_interpolation_scalar.cpp
+++ b/test/c/test_interpolation_scalar.cpp
@@ -48,8 +48,7 @@ public:
static constexpr int nlev = 7; // number of vertical levels
static constexpr int nblks_c = 2; // number of cell blocks
static constexpr int nblks_e = 2; // number of edge blocks (for p_e_in)
- static constexpr int nblks_v =
- 2; // number of vertex blocks (for rbf arrays and outputs)
+ static constexpr int nblks_v = 2; // number of vertex blocks
// Parameter values.
const int i_startblk = 0;
@@ -385,7 +384,7 @@ TYPED_TEST_SUITE(InterpolationScalarMixedTestFixture, MixedTypesSP2DP);
//
////////////////////////////////////////////////////////////////////////////////
-TYPED_TEST(InterpolationScalarMixedTestFixture, cells2edges) {
+TYPED_TEST(InterpolationScalarMixedTestFixture, Cells2Edges) {
using InType = typename TestFixture::InType;
using OutType = typename TestFixture::OutType;
@@ -424,7 +423,7 @@ TYPED_TEST(InterpolationScalarMixedTestFixture, cells2edges) {
//
////////////////////////////////////////////////////////////////////////////////
-TYPED_TEST(InterpolationScalarMixedTestFixture, cells2verts) {
+TYPED_TEST(InterpolationScalarMixedTestFixture, Cells2Verts) {
using InType = typename TestFixture::InType;
using OutType = typename TestFixture::OutType;
@@ -496,7 +495,7 @@ public:
TYPED_TEST_SUITE(Cells2vertsriScalarLibTestFixture, MixedTypes);
// Add test
-TYPED_TEST(Cells2vertsriScalarLibTestFixture, cells2verts_ri) {
+TYPED_TEST(Cells2vertsriScalarLibTestFixture, Cells2VertsRI) {
using InType = typename TestFixture::InType;
using OutType = typename TestFixture::OutType;
--
GitLab
From 1d49b1bde48b2f43dd65c63c04517e4e7332b761 Mon Sep 17 00:00:00 2001
From: Pradipta Samanta <samanta@dkrz.de>
Date: Thu, 6 Mar 2025 15:07:23 +0100
Subject: [PATCH 30/33] formatted the c codes using clang-format
---
src/interpolation/interpolation_bindings.cpp | 262 +++---
src/interpolation/interpolation_bindings.h | 103 +--
src/interpolation/mo_lib_intp_rbf.cpp | 798 ++++++++++---------
src/interpolation/mo_lib_intp_rbf.hpp | 56 +-
test/c/test_intp_rbf.cpp | 319 ++++----
5 files changed, 732 insertions(+), 806 deletions(-)
diff --git a/src/interpolation/interpolation_bindings.cpp b/src/interpolation/interpolation_bindings.cpp
index 75449d6..a13ad0f 100644
--- a/src/interpolation/interpolation_bindings.cpp
+++ b/src/interpolation/interpolation_bindings.cpp
@@ -329,191 +329,121 @@ void cell_avg_lib_sp(const float *psi_c, const int *cell_neighbor_idx,
}
void rbf_vec_interpol_vertex_lib_dp(
- const double* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const double* rbf_vec_coeff_v,
- double* p_u_out,
- double* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- )
-{
- rbf_vec_interpol_vertex_lib<double, double>(
- p_e_in,
- rbf_vec_idx_v,
- rbf_vec_blk_v,
- rbf_vec_coeff_v,
- p_u_out,
- p_v_out,
- i_startblk, // start_block needed for get_indices_c_lib
- i_endblk, // end_block needed for get_indices_c_lib
- i_startidx_in, // start_index needed for get_indices_c_lib
- i_endidx_in, // end_index needed for get_indices_c_lib
- slev, // vertical start level
- elev, // vertical end level
- nproma, // inner loop length/vector length
- lacc, // if true, use Cuda mem-/exec-spaces
- acc_async, // [deprecated] use async acc
- nlev, nblks_e, nblks_v
- );
+ const double *p_e_in, const int *rbf_vec_idx_v, const int *rbf_vec_blk_v,
+ const double *rbf_vec_coeff_v, double *p_u_out, double *p_v_out,
+ const int i_startblk, const int i_endblk, const int i_startidx_in,
+ const int i_endidx_in, const int slev, const int elev, const int nproma,
+ const bool lacc, const bool acc_async, const int nlev, const int nblks_e,
+ const int nblks_v) {
+ rbf_vec_interpol_vertex_lib<double, double>(
+ p_e_in, rbf_vec_idx_v, rbf_vec_blk_v, rbf_vec_coeff_v, p_u_out, p_v_out,
+ i_startblk, i_endblk, i_startidx_in, i_endidx_in, slev, elev, nproma,
+ lacc, acc_async, nlev, nblks_e, nblks_v);
}
void rbf_vec_interpol_vertex_lib_sp(
- const float* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const float* rbf_vec_coeff_v,
- float* p_u_out,
- float* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- )
-{
- rbf_vec_interpol_vertex_lib<float, float>(
- p_e_in,
- rbf_vec_idx_v,
- rbf_vec_blk_v,
- rbf_vec_coeff_v,
- p_u_out,
- p_v_out,
- i_startblk, // start_block needed for get_indices_c_lib
- i_endblk, // end_block needed for get_indices_c_lib
- i_startidx_in, // start_index needed for get_indices_c_lib
- i_endidx_in, // end_index needed for get_indices_c_lib
- slev, // vertical start level
- elev, // vertical end level
- nproma, // inner loop length/vector length
- lacc, // if true, use Cuda mem-/exec-spaces
- acc_async, // [deprecated] use async acc
- nlev, nblks_e, nblks_v
- );
-
+ const float *p_e_in, const int *rbf_vec_idx_v, const int *rbf_vec_blk_v,
+ const float *rbf_vec_coeff_v, float *p_u_out, float *p_v_out,
+ const int i_startblk, const int i_endblk, const int i_startidx_in,
+ const int i_endidx_in, const int slev, const int elev, const int nproma,
+ const bool lacc, const bool acc_async, const int nlev, const int nblks_e,
+ const int nblks_v) {
+ rbf_vec_interpol_vertex_lib<float, float>(
+ p_e_in, rbf_vec_idx_v, rbf_vec_blk_v, rbf_vec_coeff_v, p_u_out, p_v_out,
+ i_startblk, i_endblk, i_startidx_in, i_endidx_in, slev, elev, nproma,
+ lacc, acc_async, nlev, nblks_e, nblks_v);
}
void rbf_vec_interpol_vertex_lib_mixprec(
- const double* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const double* rbf_vec_coeff_v,
- float* p_u_out,
- float* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- )
-{
- rbf_vec_interpol_vertex_lib<double, float>(
- p_e_in,
- rbf_vec_idx_v,
- rbf_vec_blk_v,
- rbf_vec_coeff_v,
- p_u_out,
- p_v_out,
- i_startblk, // start_block needed for get_indices_c_lib
- i_endblk, // end_block needed for get_indices_c_lib
- i_startidx_in, // start_index needed for get_indices_c_lib
- i_endidx_in, // end_index needed for get_indices_c_lib
- slev, // vertical start level
- elev, // vertical end level
- nproma, // inner loop length/vector length
- lacc, // if true, use Cuda mem-/exec-spaces
- acc_async, // [deprecated] use async acc
- nlev, nblks_e, nblks_v
- );
-
+ const double *p_e_in, const int *rbf_vec_idx_v, const int *rbf_vec_blk_v,
+ const double *rbf_vec_coeff_v, float *p_u_out, float *p_v_out,
+ const int i_startblk, const int i_endblk, const int i_startidx_in,
+ const int i_endidx_in, const int slev, const int elev, const int nproma,
+ const bool lacc, const bool acc_async, const int nlev, const int nblks_e,
+ const int nblks_v) {
+ rbf_vec_interpol_vertex_lib<double, float>(
+ p_e_in, rbf_vec_idx_v, rbf_vec_blk_v, rbf_vec_coeff_v, p_u_out, p_v_out,
+ i_startblk, i_endblk, i_startidx_in, i_endidx_in, slev, elev, nproma,
+ lacc, acc_async, nlev, nblks_e, nblks_v);
}
-void rbf_interpol_c2grad_lib_sp(const float* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const float* rbf_c2grad_coeff, float* grad_x, float* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc){
-
- rbf_interpol_c2grad_lib<float>(p_cell_in, rbf_c2grad_idx, rbf_c2grad_blk,
- rbf_c2grad_coeff,
- grad_x, grad_y, i_startblk, i_endblk,
- i_startidx_in, i_endidx_in, slev, elev,
- nproma, rbf_c2grad_dim, nlev, nblk_c, lacc);
+void rbf_interpol_c2grad_lib_sp(const float *p_cell_in,
+ const int *rbf_c2grad_idx,
+ const int *rbf_c2grad_blk,
+ const float *rbf_c2grad_coeff, float *grad_x,
+ float *grad_y, int i_startblk, int i_endblk,
+ int i_startidx_in, int i_endidx_in, int slev,
+ int elev, int nproma, int rbf_c2grad_dim,
+ int nlev, int nblk_c, bool lacc) {
+
+ rbf_interpol_c2grad_lib<float>(
+ p_cell_in, rbf_c2grad_idx, rbf_c2grad_blk, rbf_c2grad_coeff, grad_x,
+ grad_y, i_startblk, i_endblk, i_startidx_in, i_endidx_in, slev, elev,
+ nproma, rbf_c2grad_dim, nlev, nblk_c, lacc);
}
-void rbf_interpol_c2grad_lib_dp(const double* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const double* rbf_c2grad_coeff, double* grad_x, double* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc){
-
- rbf_interpol_c2grad_lib<double>(p_cell_in, rbf_c2grad_idx, rbf_c2grad_blk,
- rbf_c2grad_coeff,
- grad_x, grad_y, i_startblk, i_endblk,
- i_startidx_in, i_endidx_in, slev, elev,
- nproma, rbf_c2grad_dim, nlev, nblk_c, lacc);
+void rbf_interpol_c2grad_lib_dp(const double *p_cell_in,
+ const int *rbf_c2grad_idx,
+ const int *rbf_c2grad_blk,
+ const double *rbf_c2grad_coeff, double *grad_x,
+ double *grad_y, int i_startblk, int i_endblk,
+ int i_startidx_in, int i_endidx_in, int slev,
+ int elev, int nproma, int rbf_c2grad_dim,
+ int nlev, int nblk_c, bool lacc) {
+
+ rbf_interpol_c2grad_lib<double>(
+ p_cell_in, rbf_c2grad_idx, rbf_c2grad_blk, rbf_c2grad_coeff, grad_x,
+ grad_y, i_startblk, i_endblk, i_startidx_in, i_endidx_in, slev, elev,
+ nproma, rbf_c2grad_dim, nlev, nblk_c, lacc);
}
-void rbf_vec_interpol_cell_lib_sp(const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
- int rbf_vec_dim_c, bool lacc, bool acc_async){
+void rbf_vec_interpol_cell_lib_sp(
+ const float *p_vn_in, const int *rbf_vec_idx_c, const int *rbf_vec_blk_c,
+ const float *rbf_vec_coeff_c, float *p_u_out, float *p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in, int slev,
+ int elev, int nproma, int nlev, int nblks_c, int nblks_e, int rbf_vec_dim_c,
+ bool lacc, bool acc_async) {
- rbf_vec_interpol_cell_lib<float>(p_vn_in, rbf_vec_idx_c, rbf_vec_blk_c,
- rbf_vec_coeff_c, p_u_out, p_v_out,
- i_startblk, i_endblk, i_startidx_in, i_endidx_in,
- slev, elev, nproma, nlev, nblks_c, nblks_e, rbf_vec_dim_c, lacc, acc_async);
+ rbf_vec_interpol_cell_lib<float>(
+ p_vn_in, rbf_vec_idx_c, rbf_vec_blk_c, rbf_vec_coeff_c, p_u_out, p_v_out,
+ i_startblk, i_endblk, i_startidx_in, i_endidx_in, slev, elev, nproma,
+ nlev, nblks_c, nblks_e, rbf_vec_dim_c, lacc, acc_async);
}
-void rbf_vec_interpol_cell_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
- int rbf_vec_dim_c, bool lacc, bool acc_async){
+void rbf_vec_interpol_cell_lib_dp(
+ const double *p_vn_in, const int *rbf_vec_idx_c, const int *rbf_vec_blk_c,
+ const double *rbf_vec_coeff_c, double *p_u_out, double *p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in, int slev,
+ int elev, int nproma, int nlev, int nblks_c, int nblks_e, int rbf_vec_dim_c,
+ bool lacc, bool acc_async) {
- rbf_vec_interpol_cell_lib<double>(p_vn_in, rbf_vec_idx_c, rbf_vec_blk_c, rbf_vec_coeff_c,
- p_u_out, p_v_out, i_startblk, i_endblk, i_startidx_in, i_endidx_in,
- slev, elev, nproma, nlev, nblks_c, nblks_e, rbf_vec_dim_c, lacc, acc_async);
+ rbf_vec_interpol_cell_lib<double>(
+ p_vn_in, rbf_vec_idx_c, rbf_vec_blk_c, rbf_vec_coeff_c, p_u_out, p_v_out,
+ i_startblk, i_endblk, i_startidx_in, i_endidx_in, slev, elev, nproma,
+ nlev, nblks_c, nblks_e, rbf_vec_dim_c, lacc, acc_async);
}
-void rbf_vec_interpol_edge_lib_dp(const double* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const double* rbf_vec_coeff_e, double* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async){
-
- rbf_vec_interpol_edge_lib<double>(p_vn_in, rbf_vec_idx_e, rbf_vec_blk_e,
- rbf_vec_coeff_e, p_vt_out,
- i_startblk, i_endblk, i_startidx_in,i_endidx_in,
- slev, elev, nlev, nproma, rbf_vec_dim_e, nblks_e, lacc, acc_async);
+void rbf_vec_interpol_edge_lib_dp(
+ const double *p_vn_in, const int *rbf_vec_idx_e, const int *rbf_vec_blk_e,
+ const double *rbf_vec_coeff_e, double *p_vt_out, int i_startblk,
+ int i_endblk, int i_startidx_in, int i_endidx_in, int slev, int elev,
+ int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async) {
+
+ rbf_vec_interpol_edge_lib<double>(
+ p_vn_in, rbf_vec_idx_e, rbf_vec_blk_e, rbf_vec_coeff_e, p_vt_out,
+ i_startblk, i_endblk, i_startidx_in, i_endidx_in, slev, elev, nlev,
+ nproma, rbf_vec_dim_e, nblks_e, lacc, acc_async);
}
-void rbf_vec_interpol_edge_lib_sp(const float* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const float* rbf_vec_coeff_e, float* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async){
+void rbf_vec_interpol_edge_lib_sp(
+ const float *p_vn_in, const int *rbf_vec_idx_e, const int *rbf_vec_blk_e,
+ const float *rbf_vec_coeff_e, float *p_vt_out, int i_startblk, int i_endblk,
+ int i_startidx_in, int i_endidx_in, int slev, int elev, int nlev,
+ int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc, bool acc_async) {
- rbf_vec_interpol_edge_lib<float>(p_vn_in, rbf_vec_idx_e, rbf_vec_blk_e,
- rbf_vec_coeff_e, p_vt_out,
- i_startblk, i_endblk, i_startidx_in, i_endidx_in,
- slev, elev, nlev, nproma, rbf_vec_dim_e, nblks_e,lacc, acc_async);
+ rbf_vec_interpol_edge_lib<float>(
+ p_vn_in, rbf_vec_idx_e, rbf_vec_blk_e, rbf_vec_coeff_e, p_vt_out,
+ i_startblk, i_endblk, i_startidx_in, i_endidx_in, slev, elev, nlev,
+ nproma, rbf_vec_dim_e, nblks_e, lacc, acc_async);
}
diff --git a/src/interpolation/interpolation_bindings.h b/src/interpolation/interpolation_bindings.h
index 419446a..0f4778d 100644
--- a/src/interpolation/interpolation_bindings.h
+++ b/src/interpolation/interpolation_bindings.h
@@ -187,80 +187,59 @@ void cell_avg_lib_sp(const float *psi_c, const int *cell_neighbor_idx,
const int nlev, const int nblks_c, const bool lacc);
void rbf_vec_interpol_vertex_lib_dp(
- const double* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const double* rbf_vec_coeff_v,
- double* p_u_out,
- double* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- );
+ const double *p_e_in, const int *rbf_vec_idx_v, const int *rbf_vec_blk_v,
+ const double *rbf_vec_coeff_v, double *p_u_out, double *p_v_out,
+ const int i_startblk, const int i_endblk, const int i_startidx_in,
+ const int i_endidx_in, const int slev, const int elev, const int nproma,
+ const bool lacc, const bool acc_async, const int nlev, const int nblks_e,
+ const int nblks_v);
void rbf_vec_interpol_vertex_lib_sp(
- const float* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const float* rbf_vec_coeff_v,
- float* p_u_out,
- float* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- );
+ const float *p_e_in, const int *rbf_vec_idx_v, const int *rbf_vec_blk_v,
+ const float *rbf_vec_coeff_v, float *p_u_out, float *p_v_out,
+ const int i_startblk, const int i_endblk, const int i_startidx_in,
+ const int i_endidx_in, const int slev, const int elev, const int nproma,
+ const bool lacc, const bool acc_async, const int nlev, const int nblks_e,
+ const int nblks_v);
void rbf_interpol_c2grad_lib_sp(
- const float* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const float* rbf_c2grad_coeff, float* grad_x, const float* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
+ const float *p_cell_in, const int *rbf_c2grad_idx,
+ const int *rbf_c2grad_blk, const float *rbf_c2grad_coeff, float *grad_x,
+ const float *grad_y, int i_startblk, int i_endblk, int i_startidx_in,
+ int i_endidx_in, int slev, int elev, int nproma, int rbf_c2grad_dim,
+ int nlev, int nblk_c, bool lacc);
void rbf_interpol_c2grad_lib_dp(
- const double* p_cell_in, const int*rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const double* rbf_c2grad_coeff, double* grad_x, const double* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblk_c, bool lacc);
+ const double *p_cell_in, const int *rbf_c2grad_idx,
+ const int *rbf_c2grad_blk, const double *rbf_c2grad_coeff, double *grad_x,
+ const double *grad_y, int i_startblk, int i_endblk, int i_startidx_in,
+ int i_endidx_in, int slev, int elev, int nproma, int rbf_c2grad_dim,
+ int nlev, int nblk_c, bool lacc);
void rbf_vec_interpol_cell_lib_sp(
- const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
- int rbf_vec_dim_c, bool lacc, bool acc_async);
+ const float *p_vn_in, const int *rbf_vec_idx_c, const int *rbf_vec_blk_c,
+ const float *rbf_vec_coeff_c, float *p_u_out, float *p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in, int slev,
+ int elev, int nproma, int nlev, int nblks_c, int nblks_e, int rbf_vec_dim_c,
+ bool lacc, bool acc_async);
void rbf_vec_interpol_cell_lib_dp(
- const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
- int rbf_vec_dim_c, bool lacc, bool acc_async);
+ const double *p_vn_in, const int *rbf_vec_idx_c, const int *rbf_vec_blk_c,
+ const double *rbf_vec_coeff_c, double *p_u_out, double *p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in, int slev,
+ int elev, int nproma, int nlev, int nblks_c, int nblks_e, int rbf_vec_dim_c,
+ bool lacc, bool acc_async);
void rbf_vec_interpol_edge_lib_dp(
- const double* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const double* rbf_vec_coeff_e, double* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async);
+ const double *p_vn_in, const int *rbf_vec_idx_e, const int *rbf_vec_blk_e,
+ const double *rbf_vec_coeff_e, double *p_vt_out, int i_startblk,
+ int i_endblk, int i_startidx_in, int i_endidx_in, int slev, int elev,
+ int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async);
void rbf_vec_interpol_edge_lib_sp(
- const float* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const float* rbf_vec_coeff_e, float* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async);
+ const float *p_vn_in, const int *rbf_vec_idx_e, const int *rbf_vec_blk_e,
+ const float *rbf_vec_coeff_e, float *p_vt_out, int i_startblk, int i_endblk,
+ int i_startidx_in, int i_endidx_in, int slev, int elev, int nlev,
+ int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc, bool acc_async);
}
diff --git a/src/interpolation/mo_lib_intp_rbf.cpp b/src/interpolation/mo_lib_intp_rbf.cpp
index 361710f..d1178a6 100644
--- a/src/interpolation/mo_lib_intp_rbf.cpp
+++ b/src/interpolation/mo_lib_intp_rbf.cpp
@@ -30,404 +30,446 @@ constexpr int rbf_vec_dim_v = 6;
/// precision(T) >= precision(S)
template <typename T, typename S>
void rbf_vec_interpol_vertex_lib(
- const T* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const T* rbf_vec_coeff_v,
- S* p_u_out,
- S* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
+ const T *p_e_in, const int *rbf_vec_idx_v, const int *rbf_vec_blk_v,
+ const T *rbf_vec_coeff_v, S *p_u_out, S *p_v_out,
+ const int i_startblk, // start_block needed for get_indices_c_lib
+ const int i_endblk, // end_block needed for get_indices_c_lib
+ const int i_startidx_in, // start_index needed for get_indices_c_lib
+ const int i_endidx_in, // end_index needed for get_indices_c_lib
+ const int slev, // vertical start level
+ const int elev, // vertical end level
+ const int nproma, // inner loop length/vector length
+ const bool lacc, // if true, use Cuda mem-/exec-spaces
+ const bool acc_async, // [deprecated] use async acc
// Dimensions for the arrays.
- const int nlev, const int nblks_e, const int nblks_v
- )
-{
- /*
+ const int nlev, const int nblks_e, const int nblks_v) {
+ /*
#ifdef DIM_ENABLE_GPU
- if (lacc){ using MemSpace = Kokkos::CudaSpace;
- } else { using MemSpace = Kokkos::HostSpace; }
+ if (lacc){ using MemSpace = Kokkos::CudaSpace;
+ } else { using MemSpace = Kokkos::HostSpace; }
#else
- using MemSpace = Kokkos::HostSpace;
+ using MemSpace = Kokkos::HostSpace;
#endif
- */
-
- // Wrap raw pointers in unmanaged Kokkos Views.
- typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
- typedef Kokkos::View<const T****, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT4D;
- typedef Kokkos::View<const int***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstInt3D;
- typedef Kokkos::View<S***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedS3D;
-
-
-
- // input components of velocity or horizontal vorticity vectors at edge midpoints
- // dim: (nproma,nlev,nblks_e)
- UnmanagedConstT3D p_e_in_view(p_e_in, nproma, nlev, nblks_e);
-
- // index array defining the stencil of surrounding edges for vector rbf interpolation at each triangle vertex
- // (rbf_vec_dim_v,nproma,nblks_v)
- UnmanagedConstInt3D iidx_view(rbf_vec_idx_v, rbf_vec_dim_v, nproma, nblks_v);
- UnmanagedConstInt3D iblk_view(rbf_vec_blk_v, rbf_vec_dim_v, nproma, nblks_v);
-
- // coefficients are working precision array containing the coefficients used for vector rbf interpolation
- // at each tringle vertex (input is normal component),
- // dim: (rbf_vec_dim_v,2,nproma,nblks_v)
- UnmanagedConstT4D ptr_coeff_view(rbf_vec_coeff_v, rbf_vec_dim_v, 2, nproma, nblks_v);
-
- // reconstructed x-component (u) of velocity vector,
- // dim: (nproma,nlev,nblks_v)
- UnmanagedS3D p_u_out_view(p_u_out, nproma, nlev, nblks_v);
- // reconstructed y-component (v) of velocity vector,
- // dim: (nproma,nlev,nblks_v)
- UnmanagedS3D p_v_out_view(p_v_out, nproma, nlev, nblks_v);
-
- // Local vars
- //int jv, jk, jb; // integer over vertices, levels, and blocks,
- int jb; // integer over vertices, levels, and blocks,
- int i_startidx; // start index
- int i_endidx; // end index
-
- for (jb=i_startblk; jb <= i_endblk; ++jb){
-
- get_indices_v_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk,
- i_startidx, i_endidx);
-
- Kokkos::MDRangePolicy<Kokkos::Rank<2>> innerPolicy(
- {slev, i_startidx}, {elev + 1, i_endidx + 1});
-
- Kokkos::parallel_for("rbf_vec_interpol_vertex_lib", innerPolicy,
- KOKKOS_LAMBDA(const int jk, const int jv) {
-
- // NOTE: Static indexes reduced by 1 from Fortran version
- p_u_out_view(jv, jk, jb) =
- ptr_coeff_view(0, 0, jv, jb)*p_e_in_view(iidx_view(0, jv, jb), jk, iblk_view(0, jv, jb)) +
- ptr_coeff_view(1, 0, jv, jb)*p_e_in_view(iidx_view(1, jv, jb), jk, iblk_view(1, jv, jb)) +
- ptr_coeff_view(2, 0, jv, jb)*p_e_in_view(iidx_view(2, jv, jb), jk, iblk_view(2, jv, jb)) +
- ptr_coeff_view(3, 0, jv, jb)*p_e_in_view(iidx_view(3, jv, jb), jk, iblk_view(3, jv, jb)) +
- ptr_coeff_view(4, 0, jv, jb)*p_e_in_view(iidx_view(4, jv, jb), jk, iblk_view(4, jv, jb)) +
- ptr_coeff_view(5, 0, jv, jb)*p_e_in_view(iidx_view(5, jv, jb), jk, iblk_view(5, jv, jb));
- p_v_out_view(jv, jk, jb) =
- ptr_coeff_view(0, 1, jv, jb)*p_e_in_view(iidx_view(0, jv, jb), jk, iblk_view(0, jv, jb)) +
- ptr_coeff_view(1, 1, jv, jb)*p_e_in_view(iidx_view(1, jv, jb), jk, iblk_view(1, jv, jb)) +
- ptr_coeff_view(2, 1, jv, jb)*p_e_in_view(iidx_view(2, jv, jb), jk, iblk_view(2, jv, jb)) +
- ptr_coeff_view(3, 1, jv, jb)*p_e_in_view(iidx_view(3, jv, jb), jk, iblk_view(3, jv, jb)) +
- ptr_coeff_view(4, 1, jv, jb)*p_e_in_view(iidx_view(4, jv, jb), jk, iblk_view(4, jv, jb)) +
- ptr_coeff_view(5, 1, jv, jb)*p_e_in_view(iidx_view(5, jv, jb), jk, iblk_view(5, jv, jb));
- }
- );
- }
+ */
+
+ // Wrap raw pointers in unmanaged Kokkos Views.
+ typedef Kokkos::View<const T ***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged>
+ UnmanagedConstT3D;
+ typedef Kokkos::View<const T ****, Kokkos::LayoutLeft,
+ Kokkos::MemoryUnmanaged>
+ UnmanagedConstT4D;
+ typedef Kokkos::View<const int ***, Kokkos::LayoutLeft,
+ Kokkos::MemoryUnmanaged>
+ UnmanagedConstInt3D;
+ typedef Kokkos::View<S ***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged>
+ UnmanagedS3D;
+
+ // input components of velocity or horizontal vorticity vectors at edge
+ // midpoints dim: (nproma,nlev,nblks_e)
+ UnmanagedConstT3D p_e_in_view(p_e_in, nproma, nlev, nblks_e);
+
+ // index array defining the stencil of surrounding edges for vector rbf
+ // interpolation at each triangle vertex (rbf_vec_dim_v,nproma,nblks_v)
+ UnmanagedConstInt3D iidx_view(rbf_vec_idx_v, rbf_vec_dim_v, nproma, nblks_v);
+ UnmanagedConstInt3D iblk_view(rbf_vec_blk_v, rbf_vec_dim_v, nproma, nblks_v);
+
+ // coefficients are working precision array containing the coefficients used
+ // for vector rbf interpolation at each tringle vertex (input is normal
+ // component), dim: (rbf_vec_dim_v,2,nproma,nblks_v)
+ UnmanagedConstT4D ptr_coeff_view(rbf_vec_coeff_v, rbf_vec_dim_v, 2, nproma,
+ nblks_v);
+
+ // reconstructed x-component (u) of velocity vector,
+ // dim: (nproma,nlev,nblks_v)
+ UnmanagedS3D p_u_out_view(p_u_out, nproma, nlev, nblks_v);
+ // reconstructed y-component (v) of velocity vector,
+ // dim: (nproma,nlev,nblks_v)
+ UnmanagedS3D p_v_out_view(p_v_out, nproma, nlev, nblks_v);
+
+ // Local vars
+ // int jv, jk, jb; // integer over vertices, levels, and blocks,
+ int jb; // integer over vertices, levels, and blocks,
+ int i_startidx; // start index
+ int i_endidx; // end index
+
+ for (jb = i_startblk; jb <= i_endblk; ++jb) {
+
+ get_indices_v_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk,
+ i_endblk, i_startidx, i_endidx);
+
+ Kokkos::MDRangePolicy<Kokkos::Rank<2>> innerPolicy(
+ {slev, i_startidx}, {elev + 1, i_endidx + 1});
+
+ Kokkos::parallel_for(
+ "rbf_vec_interpol_vertex_lib", innerPolicy,
+ KOKKOS_LAMBDA(const int jk, const int jv) {
+ // NOTE: Static indexes reduced by 1 from Fortran version
+ p_u_out_view(jv, jk, jb) =
+ ptr_coeff_view(0, 0, jv, jb) *
+ p_e_in_view(iidx_view(0, jv, jb), jk, iblk_view(0, jv, jb)) +
+ ptr_coeff_view(1, 0, jv, jb) *
+ p_e_in_view(iidx_view(1, jv, jb), jk, iblk_view(1, jv, jb)) +
+ ptr_coeff_view(2, 0, jv, jb) *
+ p_e_in_view(iidx_view(2, jv, jb), jk, iblk_view(2, jv, jb)) +
+ ptr_coeff_view(3, 0, jv, jb) *
+ p_e_in_view(iidx_view(3, jv, jb), jk, iblk_view(3, jv, jb)) +
+ ptr_coeff_view(4, 0, jv, jb) *
+ p_e_in_view(iidx_view(4, jv, jb), jk, iblk_view(4, jv, jb)) +
+ ptr_coeff_view(5, 0, jv, jb) *
+ p_e_in_view(iidx_view(5, jv, jb), jk, iblk_view(5, jv, jb));
+ p_v_out_view(jv, jk, jb) =
+ ptr_coeff_view(0, 1, jv, jb) *
+ p_e_in_view(iidx_view(0, jv, jb), jk, iblk_view(0, jv, jb)) +
+ ptr_coeff_view(1, 1, jv, jb) *
+ p_e_in_view(iidx_view(1, jv, jb), jk, iblk_view(1, jv, jb)) +
+ ptr_coeff_view(2, 1, jv, jb) *
+ p_e_in_view(iidx_view(2, jv, jb), jk, iblk_view(2, jv, jb)) +
+ ptr_coeff_view(3, 1, jv, jb) *
+ p_e_in_view(iidx_view(3, jv, jb), jk, iblk_view(3, jv, jb)) +
+ ptr_coeff_view(4, 1, jv, jb) *
+ p_e_in_view(iidx_view(4, jv, jb), jk, iblk_view(4, jv, jb)) +
+ ptr_coeff_view(5, 1, jv, jb) *
+ p_e_in_view(iidx_view(5, jv, jb), jk, iblk_view(5, jv, jb));
+ });
+ }
}
template <typename T>
-void rbf_interpol_c2grad_lib(const T* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const T* rbf_c2grad_coeff, T* grad_x, T* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblks_c, bool lacc){
-
- //aliases for unmanaged Kokkos views
- typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
- typedef Kokkos::View<T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedT3D;
- typedef Kokkos::View<const int***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstInt3D;
- typedef Kokkos::View<const T****, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT4D;
-
- //to avoid memory ownership issues
- UnmanagedConstT3D p_cell_in_view(p_cell_in, nproma, nlev, nblks_c);
- UnmanagedT3D grad_x_view(grad_x, nproma, nlev, nblks_c);
- UnmanagedT3D grad_y_view(grad_y, nproma, nlev, nblks_c);
- UnmanagedConstInt3D rbf_c2grad_idx_view(rbf_c2grad_idx, rbf_c2grad_dim, nproma, nblks_c);
- UnmanagedConstInt3D rbf_c2grad_blk_view(rbf_c2grad_blk, rbf_c2grad_dim, nproma, nblks_c);
- UnmanagedConstT4D rbf_c2grad_coeff_view(rbf_c2grad_coeff, rbf_c2grad_dim, 2, nproma, nblks_c);
-
- for (int jb = i_startblk; jb <= i_endblk; ++jb) {
-
- int i_startidx, i_endidx;
- get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk, i_startidx, i_endidx);
-
- Kokkos::parallel_for("rbf_interpol_c2grad",
- Kokkos::MDRangePolicy<Kokkos::Rank<2>>({slev, i_startidx}, {elev+1, i_endidx+1}),
- KOKKOS_LAMBDA(const int jk, const int jc){
-
- grad_x_view(jc, jk, jb) =
- rbf_c2grad_coeff_view(0, 1, jc, jb)*
- p_cell_in_view(jc, jk, jb) +
- rbf_c2grad_coeff_view(1, 1, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(1, jc, jb), jk, rbf_c2grad_blk_view(1, jc, jb)) +
- rbf_c2grad_coeff_view(2, 1, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(2, jc, jb), jk, rbf_c2grad_blk_view(2, jc, jb)) +
- rbf_c2grad_coeff_view(3, 1, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(3, jc, jb), jk, rbf_c2grad_blk_view(3, jc, jb)) +
- rbf_c2grad_coeff_view(4, 1, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(4, jc, jb), jk, rbf_c2grad_blk_view(4, jc, jb)) +
- rbf_c2grad_coeff_view(5, 1, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(5, jc, jb), jk, rbf_c2grad_blk_view(5, jc, jb)) +
- rbf_c2grad_coeff_view(6, 1, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(6, jc, jb), jk, rbf_c2grad_blk_view(6, jc, jb)) +
- rbf_c2grad_coeff_view(7, 1, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(7, jc, jb), jk, rbf_c2grad_blk_view(7, jc, jb)) +
- rbf_c2grad_coeff_view(8, 1, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(8, jc, jb), jk, rbf_c2grad_blk_view(8, jc, jb)) +
- rbf_c2grad_coeff_view(9, 1, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(9, jc, jb), jk, rbf_c2grad_blk_view(9, jc, jb));
-
- grad_y_view(jc, jk, jb) =
- rbf_c2grad_coeff_view(0, 2, jc, jb)*
- p_cell_in_view(jc, jk, jb) +
- rbf_c2grad_coeff_view(1, 2, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(1, jc, jb), jk, rbf_c2grad_blk_view(1, jc, jb)) +
- rbf_c2grad_coeff_view(2, 2, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(2, jc, jb), jk, rbf_c2grad_blk_view(2, jc, jb)) +
- rbf_c2grad_coeff_view(3, 2, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(3, jc, jb), jk, rbf_c2grad_blk_view(3, jc, jb)) +
- rbf_c2grad_coeff_view(4, 2, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(4, jc, jb), jk, rbf_c2grad_blk_view(4, jc, jb)) +
- rbf_c2grad_coeff_view(5, 2, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(5, jc, jb), jk, rbf_c2grad_blk_view(5, jc, jb)) +
- rbf_c2grad_coeff_view(6, 2, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(6, jc, jb), jk, rbf_c2grad_blk_view(6, jc, jb)) +
- rbf_c2grad_coeff_view(7, 2, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(7, jc, jb), jk, rbf_c2grad_blk_view(7, jc, jb)) +
- rbf_c2grad_coeff_view(8, 2, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(8, jc, jb), jk, rbf_c2grad_blk_view(8, jc, jb)) +
- rbf_c2grad_coeff_view(9, 2, jc, jb)*
- p_cell_in_view(rbf_c2grad_idx_view(9, jc, jb), jk, rbf_c2grad_blk_view(9, jc, jb));
- });
-
- }//for
-}//void
-
+void rbf_interpol_c2grad_lib(const T *p_cell_in, const int *rbf_c2grad_idx,
+ const int *rbf_c2grad_blk,
+ const T *rbf_c2grad_coeff, T *grad_x, T *grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in,
+ int i_endidx_in, int slev, int elev, int nproma,
+ int rbf_c2grad_dim, int nlev, int nblks_c,
+ bool lacc) {
+
+ // aliases for unmanaged Kokkos views
+ typedef Kokkos::View<const T ***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged>
+ UnmanagedConstT3D;
+ typedef Kokkos::View<T ***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged>
+ UnmanagedT3D;
+ typedef Kokkos::View<const int ***, Kokkos::LayoutLeft,
+ Kokkos::MemoryUnmanaged>
+ UnmanagedConstInt3D;
+ typedef Kokkos::View<const T ****, Kokkos::LayoutLeft,
+ Kokkos::MemoryUnmanaged>
+ UnmanagedConstT4D;
+
+ // to avoid memory ownership issues
+ UnmanagedConstT3D p_cell_in_view(p_cell_in, nproma, nlev, nblks_c);
+ UnmanagedT3D grad_x_view(grad_x, nproma, nlev, nblks_c);
+ UnmanagedT3D grad_y_view(grad_y, nproma, nlev, nblks_c);
+ UnmanagedConstInt3D rbf_c2grad_idx_view(rbf_c2grad_idx, rbf_c2grad_dim,
+ nproma, nblks_c);
+ UnmanagedConstInt3D rbf_c2grad_blk_view(rbf_c2grad_blk, rbf_c2grad_dim,
+ nproma, nblks_c);
+ UnmanagedConstT4D rbf_c2grad_coeff_view(rbf_c2grad_coeff, rbf_c2grad_dim, 2,
+ nproma, nblks_c);
+
+ for (int jb = i_startblk; jb <= i_endblk; ++jb) {
+
+ int i_startidx, i_endidx;
+ get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk,
+ i_endblk, i_startidx, i_endidx);
+
+ Kokkos::MDRangePolicy<Kokkos::Rank<2>> innerPolicy(
+ {slev, i_startidx}, {elev + 1, i_endidx + 1});
+
+ Kokkos::parallel_for(
+ "rbf_interpol_c2grad", innerPolicy,
+ KOKKOS_LAMBDA(const int jk, const int jc) {
+ grad_x_view(jc, jk, jb) =
+ rbf_c2grad_coeff_view(0, 1, jc, jb) * p_cell_in_view(jc, jk, jb) +
+ rbf_c2grad_coeff_view(1, 1, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(1, jc, jb), jk,
+ rbf_c2grad_blk_view(1, jc, jb)) +
+ rbf_c2grad_coeff_view(2, 1, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(2, jc, jb), jk,
+ rbf_c2grad_blk_view(2, jc, jb)) +
+ rbf_c2grad_coeff_view(3, 1, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(3, jc, jb), jk,
+ rbf_c2grad_blk_view(3, jc, jb)) +
+ rbf_c2grad_coeff_view(4, 1, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(4, jc, jb), jk,
+ rbf_c2grad_blk_view(4, jc, jb)) +
+ rbf_c2grad_coeff_view(5, 1, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(5, jc, jb), jk,
+ rbf_c2grad_blk_view(5, jc, jb)) +
+ rbf_c2grad_coeff_view(6, 1, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(6, jc, jb), jk,
+ rbf_c2grad_blk_view(6, jc, jb)) +
+ rbf_c2grad_coeff_view(7, 1, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(7, jc, jb), jk,
+ rbf_c2grad_blk_view(7, jc, jb)) +
+ rbf_c2grad_coeff_view(8, 1, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(8, jc, jb), jk,
+ rbf_c2grad_blk_view(8, jc, jb)) +
+ rbf_c2grad_coeff_view(9, 1, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(9, jc, jb), jk,
+ rbf_c2grad_blk_view(9, jc, jb));
+
+ grad_y_view(jc, jk, jb) =
+ rbf_c2grad_coeff_view(0, 2, jc, jb) * p_cell_in_view(jc, jk, jb) +
+ rbf_c2grad_coeff_view(1, 2, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(1, jc, jb), jk,
+ rbf_c2grad_blk_view(1, jc, jb)) +
+ rbf_c2grad_coeff_view(2, 2, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(2, jc, jb), jk,
+ rbf_c2grad_blk_view(2, jc, jb)) +
+ rbf_c2grad_coeff_view(3, 2, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(3, jc, jb), jk,
+ rbf_c2grad_blk_view(3, jc, jb)) +
+ rbf_c2grad_coeff_view(4, 2, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(4, jc, jb), jk,
+ rbf_c2grad_blk_view(4, jc, jb)) +
+ rbf_c2grad_coeff_view(5, 2, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(5, jc, jb), jk,
+ rbf_c2grad_blk_view(5, jc, jb)) +
+ rbf_c2grad_coeff_view(6, 2, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(6, jc, jb), jk,
+ rbf_c2grad_blk_view(6, jc, jb)) +
+ rbf_c2grad_coeff_view(7, 2, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(7, jc, jb), jk,
+ rbf_c2grad_blk_view(7, jc, jb)) +
+ rbf_c2grad_coeff_view(8, 2, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(8, jc, jb), jk,
+ rbf_c2grad_blk_view(8, jc, jb)) +
+ rbf_c2grad_coeff_view(9, 2, jc, jb) *
+ p_cell_in_view(rbf_c2grad_idx_view(9, jc, jb), jk,
+ rbf_c2grad_blk_view(9, jc, jb));
+ });
+
+ } // for
+} // void
//------------------------------------------rbf_vec_interpol_cell_lib---------------------------------------------
template <typename T>
-void rbf_vec_interpol_cell_lib(const T* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const T* rbf_vec_coeff_c, T* p_u_out, T* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
- int rbf_vec_dim_c, bool lacc, bool acc_async){
-
- typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
- typedef Kokkos::View<T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedT3D;
- typedef Kokkos::View<const int***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstInt3D;
- typedef Kokkos::View<const T****, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT4D;
-
- UnmanagedConstT3D p_vn_in_view(p_vn_in, nproma, nlev, nblks_e);
- UnmanagedConstInt3D rbf_vec_idx_c_view(rbf_vec_idx_c, rbf_vec_dim_c, nproma, nblks_c);
- UnmanagedConstInt3D rbf_vec_blk_c_view(rbf_vec_blk_c, rbf_vec_dim_c, nproma, nblks_c);
- UnmanagedConstT4D rbf_vec_coeff_c_view(rbf_vec_coeff_c, nproma, nblks_c); //TODO
- UnmanagedT3D p_u_out_view(p_u_out, nproma, nlev, nblks_c);
- UnmanagedT3D p_v_out_view(p_u_out, nproma, nlev, nblks_c);
-
-
- for (int jb = i_startblk; jb <= i_endblk; ++jb) {
-
- int i_startidx, i_endidx;
- get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk, i_startidx, i_endidx);
-
- Kokkos::parallel_for("rbf_vec_interpol_cell_lib",
- Kokkos::MDRangePolicy<Kokkos::Rank<2>>({slev, i_startidx}, {elev+1, i_endidx+1}),
- KOKKOS_LAMBDA(const int jk, const int jc){
-
- p_u_out_view(jc, jk, jb) =
- rbf_vec_coeff_c_view(0, 1, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(0, jc, jb), jk, rbf_vec_blk_c_view(0, jc, jb)) +
- rbf_vec_coeff_c_view(1, 1, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(1, jc, jb), jk, rbf_vec_blk_c_view(1, jc, jb)) +
- rbf_vec_coeff_c_view(2, 1, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(2, jc, jb), jk, rbf_vec_blk_c_view(2, jc, jb)) +
- rbf_vec_coeff_c_view(3, 1, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(3, jc, jb), jk, rbf_vec_blk_c_view(3, jc, jb)) +
- rbf_vec_coeff_c_view(4, 1, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(4, jc, jb), jk, rbf_vec_blk_c_view(4, jc, jb)) +
- rbf_vec_coeff_c_view(5, 1, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(5, jc, jb), jk, rbf_vec_blk_c_view(5, jc, jb)) +
- rbf_vec_coeff_c_view(6, 1, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(6, jc, jb), jk, rbf_vec_blk_c_view(6, jc, jb)) +
- rbf_vec_coeff_c_view(7, 1, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(7, jc, jb), jk, rbf_vec_blk_c_view(7, jc, jb)) +
- rbf_vec_coeff_c_view(8, 1, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(8, jc, jb), jk, rbf_vec_blk_c_view(8, jc, jb));
-
- p_v_out_view(jc, jk, jb) =
- rbf_vec_coeff_c_view(0, 2, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(0, jc, jb), jk, rbf_vec_blk_c_view(0, jc, jb)) +
- rbf_vec_coeff_c_view(1, 2, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(1, jc, jb), jk, rbf_vec_blk_c_view(1, jc, jb)) +
- rbf_vec_coeff_c_view(2, 2, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(2, jc, jb), jk, rbf_vec_blk_c_view(2, jc, jb)) +
- rbf_vec_coeff_c_view(3, 2, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(3, jc, jb), jk, rbf_vec_blk_c_view(3, jc, jb)) +
- rbf_vec_coeff_c_view(4, 2, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(4, jc, jb), jk, rbf_vec_blk_c_view(4, jc, jb)) +
- rbf_vec_coeff_c_view(5, 2, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(5, jc, jb), jk, rbf_vec_blk_c_view(5, jc, jb)) +
- rbf_vec_coeff_c_view(6, 2, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(6, jc, jb), jk, rbf_vec_blk_c_view(6, jc, jb)) +
- rbf_vec_coeff_c_view(7, 2, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(7, jc, jb), jk, rbf_vec_blk_c_view(7, jc, jb)) +
- rbf_vec_coeff_c_view(8, 2, jc, jb)*
- p_vn_in_view(rbf_vec_idx_c_view(8, jc, jb), jk, rbf_vec_blk_c_view(8, jc, jb));
- });
- Kokkos::fence();
- }//for
-}//void
+void rbf_vec_interpol_cell_lib(const T *p_vn_in, const int *rbf_vec_idx_c,
+ const int *rbf_vec_blk_c,
+ const T *rbf_vec_coeff_c, T *p_u_out, T *p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in,
+ int i_endidx_in, int slev, int elev, int nproma,
+ int nlev, int nblks_c, int nblks_e,
+ int rbf_vec_dim_c, bool lacc, bool acc_async) {
+
+ typedef Kokkos::View<const T ***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged>
+ UnmanagedConstT3D;
+ typedef Kokkos::View<T ***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged>
+ UnmanagedT3D;
+ typedef Kokkos::View<const int ***, Kokkos::LayoutLeft,
+ Kokkos::MemoryUnmanaged>
+ UnmanagedConstInt3D;
+ typedef Kokkos::View<const T ****, Kokkos::LayoutLeft,
+ Kokkos::MemoryUnmanaged>
+ UnmanagedConstT4D;
+
+ UnmanagedConstT3D p_vn_in_view(p_vn_in, nproma, nlev, nblks_e);
+ UnmanagedConstInt3D rbf_vec_idx_c_view(rbf_vec_idx_c, rbf_vec_dim_c, nproma,
+ nblks_c);
+ UnmanagedConstInt3D rbf_vec_blk_c_view(rbf_vec_blk_c, rbf_vec_dim_c, nproma,
+ nblks_c);
+ UnmanagedConstT4D rbf_vec_coeff_c_view(rbf_vec_coeff_c, nproma,
+ nblks_c); // TODO
+ UnmanagedT3D p_u_out_view(p_u_out, nproma, nlev, nblks_c);
+ UnmanagedT3D p_v_out_view(p_u_out, nproma, nlev, nblks_c);
+
+ for (int jb = i_startblk; jb <= i_endblk; ++jb) {
+
+ int i_startidx, i_endidx;
+ get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk,
+ i_endblk, i_startidx, i_endidx);
+
+ Kokkos::MDRangePolicy<Kokkos::Rank<2>> innerPolicy(
+ {slev, i_startidx}, {elev + 1, i_endidx + 1});
+
+ Kokkos::parallel_for(
+ "rbf_vec_interpol_cell_lib", innerPolicy,
+ KOKKOS_LAMBDA(const int jk, const int jc) {
+ p_u_out_view(jc, jk, jb) =
+ rbf_vec_coeff_c_view(0, 1, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(0, jc, jb), jk,
+ rbf_vec_blk_c_view(0, jc, jb)) +
+ rbf_vec_coeff_c_view(1, 1, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(1, jc, jb), jk,
+ rbf_vec_blk_c_view(1, jc, jb)) +
+ rbf_vec_coeff_c_view(2, 1, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(2, jc, jb), jk,
+ rbf_vec_blk_c_view(2, jc, jb)) +
+ rbf_vec_coeff_c_view(3, 1, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(3, jc, jb), jk,
+ rbf_vec_blk_c_view(3, jc, jb)) +
+ rbf_vec_coeff_c_view(4, 1, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(4, jc, jb), jk,
+ rbf_vec_blk_c_view(4, jc, jb)) +
+ rbf_vec_coeff_c_view(5, 1, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(5, jc, jb), jk,
+ rbf_vec_blk_c_view(5, jc, jb)) +
+ rbf_vec_coeff_c_view(6, 1, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(6, jc, jb), jk,
+ rbf_vec_blk_c_view(6, jc, jb)) +
+ rbf_vec_coeff_c_view(7, 1, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(7, jc, jb), jk,
+ rbf_vec_blk_c_view(7, jc, jb)) +
+ rbf_vec_coeff_c_view(8, 1, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(8, jc, jb), jk,
+ rbf_vec_blk_c_view(8, jc, jb));
+
+ p_v_out_view(jc, jk, jb) =
+ rbf_vec_coeff_c_view(0, 2, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(0, jc, jb), jk,
+ rbf_vec_blk_c_view(0, jc, jb)) +
+ rbf_vec_coeff_c_view(1, 2, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(1, jc, jb), jk,
+ rbf_vec_blk_c_view(1, jc, jb)) +
+ rbf_vec_coeff_c_view(2, 2, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(2, jc, jb), jk,
+ rbf_vec_blk_c_view(2, jc, jb)) +
+ rbf_vec_coeff_c_view(3, 2, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(3, jc, jb), jk,
+ rbf_vec_blk_c_view(3, jc, jb)) +
+ rbf_vec_coeff_c_view(4, 2, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(4, jc, jb), jk,
+ rbf_vec_blk_c_view(4, jc, jb)) +
+ rbf_vec_coeff_c_view(5, 2, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(5, jc, jb), jk,
+ rbf_vec_blk_c_view(5, jc, jb)) +
+ rbf_vec_coeff_c_view(6, 2, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(6, jc, jb), jk,
+ rbf_vec_blk_c_view(6, jc, jb)) +
+ rbf_vec_coeff_c_view(7, 2, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(7, jc, jb), jk,
+ rbf_vec_blk_c_view(7, jc, jb)) +
+ rbf_vec_coeff_c_view(8, 2, jc, jb) *
+ p_vn_in_view(rbf_vec_idx_c_view(8, jc, jb), jk,
+ rbf_vec_blk_c_view(8, jc, jb));
+ });
+ Kokkos::fence();
+ } // for
+} // void
//------------------------------------------rbf_vec_interpol_edge_lib---------------------------------------------
template <typename T>
-void rbf_vec_interpol_edge_lib(const T* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const T* rbf_vec_coeff_e, T* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async){
-
- typedef Kokkos::View<const T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstT3D;
- typedef Kokkos::View<T***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedT3D;
- typedef Kokkos::View<const int***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged> UnmanagedConstInt3D;
-
- UnmanagedConstT3D p_vn_in_view(p_vn_in, nproma,nlev,nblks_e);
- UnmanagedConstInt3D rbf_vec_idx_e_view(rbf_vec_idx_e, rbf_vec_dim_e,nproma,nblks_e);
- UnmanagedConstInt3D rbf_vec_blk_e_view(rbf_vec_blk_e, rbf_vec_dim_e,nproma,nblks_e);
- UnmanagedConstT3D rbf_vec_coeff_e_view(rbf_vec_coeff_e, rbf_vec_dim_e,nproma,nblks_e);
- UnmanagedT3D p_vt_out_view(p_vt_out, nproma,nlev,nblks_e);
-
-
- for (int jb = i_startblk; jb <= i_endblk; ++jb) {
-
- int i_startidx, i_endidx;
- get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk, i_endblk, i_startidx, i_endidx);
-
- Kokkos::parallel_for("rbf_vec_interpol_edge_lib",
- Kokkos::MDRangePolicy<Kokkos::Rank<2>>({slev, i_startidx}, {elev+1, i_endidx+1}),
- KOKKOS_LAMBDA(const int jk, const int je){
-
- p_vt_out_view(je, jk, jb) =
- rbf_vec_coeff_e_view(0, je, jb) * p_vn_in_view(rbf_vec_idx_e_view(0, je, jb), jk, rbf_vec_blk_e_view(0, je, jb)) +
- rbf_vec_coeff_e_view(1, je, jb) * p_vn_in_view(rbf_vec_idx_e_view(1, je, jb), jk, rbf_vec_blk_e_view(1, je, jb)) +
- rbf_vec_coeff_e_view(2, je, jb) * p_vn_in_view(rbf_vec_idx_e_view(2, je, jb), jk, rbf_vec_blk_e_view(2, je, jb)) +
- rbf_vec_coeff_e_view(3, je, jb) * p_vn_in_view(rbf_vec_idx_e_view(3, je, jb), jk, rbf_vec_blk_e_view(3, je, jb));
-
- });
- }
+void rbf_vec_interpol_edge_lib(const T *p_vn_in, const int *rbf_vec_idx_e,
+ const int *rbf_vec_blk_e,
+ const T *rbf_vec_coeff_e, T *p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in,
+ int i_endidx_in, int slev, int elev, int nlev,
+ int nproma, int rbf_vec_dim_e, int nblks_e,
+ bool lacc, bool acc_async) {
+
+ typedef Kokkos::View<const T ***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged>
+ UnmanagedConstT3D;
+ typedef Kokkos::View<T ***, Kokkos::LayoutLeft, Kokkos::MemoryUnmanaged>
+ UnmanagedT3D;
+ typedef Kokkos::View<const int ***, Kokkos::LayoutLeft,
+ Kokkos::MemoryUnmanaged>
+ UnmanagedConstInt3D;
+
+ UnmanagedConstT3D p_vn_in_view(p_vn_in, nproma, nlev, nblks_e);
+ UnmanagedConstInt3D rbf_vec_idx_e_view(rbf_vec_idx_e, rbf_vec_dim_e, nproma,
+ nblks_e);
+ UnmanagedConstInt3D rbf_vec_blk_e_view(rbf_vec_blk_e, rbf_vec_dim_e, nproma,
+ nblks_e);
+ UnmanagedConstT3D rbf_vec_coeff_e_view(rbf_vec_coeff_e, rbf_vec_dim_e, nproma,
+ nblks_e);
+ UnmanagedT3D p_vt_out_view(p_vt_out, nproma, nlev, nblks_e);
+
+ for (int jb = i_startblk; jb <= i_endblk; ++jb) {
+
+ int i_startidx, i_endidx;
+ get_indices_c_lib(i_startidx_in, i_endidx_in, nproma, jb, i_startblk,
+ i_endblk, i_startidx, i_endidx);
+
+ Kokkos::MDRangePolicy<Kokkos::Rank<2>> innerPolicy(
+ {slev, i_startidx}, {elev + 1, i_endidx + 1});
+
+ Kokkos::parallel_for(
+ "rbf_vec_interpol_edge_lib", innerPolicy,
+ KOKKOS_LAMBDA(const int jk, const int je) {
+ p_vt_out_view(je, jk, jb) =
+ rbf_vec_coeff_e_view(0, je, jb) *
+ p_vn_in_view(rbf_vec_idx_e_view(0, je, jb), jk,
+ rbf_vec_blk_e_view(0, je, jb)) +
+ rbf_vec_coeff_e_view(1, je, jb) *
+ p_vn_in_view(rbf_vec_idx_e_view(1, je, jb), jk,
+ rbf_vec_blk_e_view(1, je, jb)) +
+ rbf_vec_coeff_e_view(2, je, jb) *
+ p_vn_in_view(rbf_vec_idx_e_view(2, je, jb), jk,
+ rbf_vec_blk_e_view(2, je, jb)) +
+ rbf_vec_coeff_e_view(3, je, jb) *
+ p_vn_in_view(rbf_vec_idx_e_view(3, je, jb), jk,
+ rbf_vec_blk_e_view(3, je, jb));
+ });
+ }
}
// Explicit instantiation - double precision
-template
-void rbf_vec_interpol_vertex_lib<double, double>(
- const double* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const double* rbf_vec_coeff_v,
- double* p_u_out,
- double* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- );
+template void rbf_vec_interpol_vertex_lib<double, double>(
+ const double *p_e_in, const int *rbf_vec_idx_v, const int *rbf_vec_blk_v,
+ const double *rbf_vec_coeff_v, double *p_u_out, double *p_v_out,
+ const int i_startblk, const int i_endblk, const int i_startidx_in,
+ const int i_endidx_in, const int slev, const int elev, const int nproma,
+ const bool lacc, const bool acc_async, const int nlev, const int nblks_e,
+ const int nblks_v);
// Explicit instantiation - single precision
-template
-void rbf_vec_interpol_vertex_lib<float, float>(
- const float* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const float* rbf_vec_coeff_v,
- float* p_u_out,
- float* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- );
+template void rbf_vec_interpol_vertex_lib<float, float>(
+ const float *p_e_in, const int *rbf_vec_idx_v, const int *rbf_vec_blk_v,
+ const float *rbf_vec_coeff_v, float *p_u_out, float *p_v_out,
+ const int i_startblk, const int i_endblk, const int i_startidx_in,
+ const int i_endidx_in, const int slev, const int elev, const int nproma,
+ const bool lacc, const bool acc_async, const int nlev, const int nblks_e,
+ const int nblks_v);
// Explicit instantiation - mixed precision
-template
-void rbf_vec_interpol_vertex_lib<double, float>(
- const double* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const double* rbf_vec_coeff_v,
- float* p_u_out,
- float* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_v
- );
-
-template
-void rbf_vec_interpol_cell_lib<double>(
- const double* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const double* rbf_vec_coeff_c, double* p_u_out, double* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
- int rbf_vec_dim_c, bool lacc, bool acc_async
-);
-
-template
-void rbf_vec_interpol_cell_lib<float>(
- const float* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const float* rbf_vec_coeff_c, float* p_u_out, float* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
- int rbf_vec_dim_c, bool lacc, bool acc_async
-);
-
-template
-void rbf_interpol_c2grad_lib<double>(
- const double* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const double* rbf_c2grad_coeff, double* grad_x, double* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblks_c, bool lacc
-);
-
-template
-void rbf_interpol_c2grad_lib<float>(
- const float* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const float* rbf_c2grad_coeff, float* grad_x, float* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblks_c, bool lacc
-);
-
-template
-void rbf_vec_interpol_edge_lib<double>(
- const double* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const double* rbf_vec_coeff_e, double* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async
-);
-
-template
-void rbf_vec_interpol_edge_lib<float>(
- const float* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const float* rbf_vec_coeff_e, float* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async
-);
+template void rbf_vec_interpol_vertex_lib<double, float>(
+ const double *p_e_in, const int *rbf_vec_idx_v, const int *rbf_vec_blk_v,
+ const double *rbf_vec_coeff_v, float *p_u_out, float *p_v_out,
+ const int i_startblk, const int i_endblk, const int i_startidx_in,
+ const int i_endidx_in, const int slev, const int elev, const int nproma,
+ const bool lacc, const bool acc_async, const int nlev, const int nblks_e,
+ const int nblks_v);
+
+template void rbf_vec_interpol_cell_lib<double>(
+ const double *p_vn_in, const int *rbf_vec_idx_c, const int *rbf_vec_blk_c,
+ const double *rbf_vec_coeff_c, double *p_u_out, double *p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in, int slev,
+ int elev, int nproma, int nlev, int nblks_c, int nblks_e, int rbf_vec_dim_c,
+ bool lacc, bool acc_async);
+
+template void rbf_vec_interpol_cell_lib<float>(
+ const float *p_vn_in, const int *rbf_vec_idx_c, const int *rbf_vec_blk_c,
+ const float *rbf_vec_coeff_c, float *p_u_out, float *p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in, int slev,
+ int elev, int nproma, int nlev, int nblks_c, int nblks_e, int rbf_vec_dim_c,
+ bool lacc, bool acc_async);
+
+template void rbf_interpol_c2grad_lib<double>(
+ const double *p_cell_in, const int *rbf_c2grad_idx,
+ const int *rbf_c2grad_blk, const double *rbf_c2grad_coeff, double *grad_x,
+ double *grad_y, int i_startblk, int i_endblk, int i_startidx_in,
+ int i_endidx_in, int slev, int elev, int nproma, int rbf_c2grad_dim,
+ int nlev, int nblks_c, bool lacc);
+
+template void rbf_interpol_c2grad_lib<float>(
+ const float *p_cell_in, const int *rbf_c2grad_idx,
+ const int *rbf_c2grad_blk, const float *rbf_c2grad_coeff, float *grad_x,
+ float *grad_y, int i_startblk, int i_endblk, int i_startidx_in,
+ int i_endidx_in, int slev, int elev, int nproma, int rbf_c2grad_dim,
+ int nlev, int nblks_c, bool lacc);
+
+template void rbf_vec_interpol_edge_lib<double>(
+ const double *p_vn_in, const int *rbf_vec_idx_e, const int *rbf_vec_blk_e,
+ const double *rbf_vec_coeff_e, double *p_vt_out, int i_startblk,
+ int i_endblk, int i_startidx_in, int i_endidx_in, int slev, int elev,
+ int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
+ bool acc_async);
+
+template void rbf_vec_interpol_edge_lib<float>(
+ const float *p_vn_in, const int *rbf_vec_idx_e, const int *rbf_vec_blk_e,
+ const float *rbf_vec_coeff_e, float *p_vt_out, int i_startblk, int i_endblk,
+ int i_startidx_in, int i_endidx_in, int slev, int elev, int nlev,
+ int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc, bool acc_async);
diff --git a/src/interpolation/mo_lib_intp_rbf.hpp b/src/interpolation/mo_lib_intp_rbf.hpp
index 53547fc..8a85502 100644
--- a/src/interpolation/mo_lib_intp_rbf.hpp
+++ b/src/interpolation/mo_lib_intp_rbf.hpp
@@ -16,41 +16,35 @@
template <typename T, typename S>
void rbf_vec_interpol_vertex_lib(
- const T* p_e_in,
- const int* rbf_vec_idx_v,
- const int* rbf_vec_blk_v,
- const T* rbf_vec_coeff_v,
- S* p_u_out,
- S* p_v_out,
- const int i_startblk, // start_block needed for get_indices_c_lib
- const int i_endblk, // end_block needed for get_indices_c_lib
- const int i_startidx_in, // start_index needed for get_indices_c_lib
- const int i_endidx_in, // end_index needed for get_indices_c_lib
- const int slev, // vertical start level
- const int elev, // vertical end level
- const int nproma, // inner loop length/vector length
- const bool lacc, // if true, use Cuda mem-/exec-spaces
- const bool acc_async, // [deprecated] use async acc
- const int nlev, const int nblks_e, const int nblks_c
- );
+ const T *p_e_in, const int *rbf_vec_idx_v, const int *rbf_vec_blk_v,
+ const T *rbf_vec_coeff_v, S *p_u_out, S *p_v_out, const int i_startblk,
+ const int i_endblk, const int i_startidx_in, const int i_endidx_in,
+ const int slev, const int elev, const int nproma, const bool lacc,
+ const bool acc_async, const int nlev, const int nblks_e, const int nblks_c);
template <typename T>
-void rbf_interpol_c2grad_lib(const T* p_cell_in, const int* rbf_c2grad_idx, const int* rbf_c2grad_blk,
- const T* rbf_c2grad_coeff, T* grad_x, T* grad_y,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int rbf_c2grad_dim, int nlev, int nblks_c, bool lacc);
-
+void rbf_interpol_c2grad_lib(const T *p_cell_in, const int *rbf_c2grad_idx,
+ const int *rbf_c2grad_blk,
+ const T *rbf_c2grad_coeff, T *grad_x, T *grad_y,
+ int i_startblk, int i_endblk, int i_startidx_in,
+ int i_endidx_in, int slev, int elev, int nproma,
+ int rbf_c2grad_dim, int nlev, int nblks_c,
+ bool lacc);
template <typename T>
-void rbf_vec_interpol_cell_lib(const T* p_vn_in, const int* rbf_vec_idx_c, const int* rbf_vec_blk_c,
- const T* rbf_vec_coeff_c, T* p_u_out, T* p_v_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nproma, int nlev, int nblks_c, int nblks_e,
+void rbf_vec_interpol_cell_lib(const T *p_vn_in, const int *rbf_vec_idx_c,
+ const int *rbf_vec_blk_c,
+ const T *rbf_vec_coeff_c, T *p_u_out, T *p_v_out,
+ int i_startblk, int i_endblk, int i_startidx_in,
+ int i_endidx_in, int slev, int elev, int nproma,
+ int nlev, int nblks_c, int nblks_e,
int rbf_vec_dim_c, bool lacc, bool acc_async);
template <typename T>
-void rbf_vec_interpol_edge_lib(const T* p_vn_in, const int* rbf_vec_idx_e, const int* rbf_vec_blk_e,
- const T* rbf_vec_coeff_e, T* p_vt_out,
- int i_startblk, int i_endblk, int i_startidx_in, int i_endidx_in,
- int slev, int elev, int nlev, int nproma, int rbf_vec_dim_e, int nblks_e, bool lacc,
- bool acc_async);
+void rbf_vec_interpol_edge_lib(const T *p_vn_in, const int *rbf_vec_idx_e,
+ const int *rbf_vec_blk_e,
+ const T *rbf_vec_coeff_e, T *p_vt_out,
+ int i_startblk, int i_endblk, int i_startidx_in,
+ int i_endidx_in, int slev, int elev, int nlev,
+ int nproma, int rbf_vec_dim_e, int nblks_e,
+ bool lacc, bool acc_async);
diff --git a/test/c/test_intp_rbf.cpp b/test/c/test_intp_rbf.cpp
index 983352c..cdee264 100644
--- a/test/c/test_intp_rbf.cpp
+++ b/test/c/test_intp_rbf.cpp
@@ -9,45 +9,43 @@
// SPDX-License-Identifier: BSD-3-Clause
// ---------------------------------------------------------------
-#include <gtest/gtest.h>
+#include "mo_lib_intp_rbf.hpp"
#include <Kokkos_Core.hpp>
-#include <vector>
#include <algorithm>
+#include <gtest/gtest.h>
#include <numeric>
-#include "mo_lib_intp_rbf.hpp"
+#include <vector>
// Free-function helpers for 3D and 4D array sizes (assumed column-major)
-template<typename T>
-size_t num_elements_3d(int d1, int d2, int d3) {
+template <typename T> size_t num_elements_3d(int d1, int d2, int d3) {
return static_cast<size_t>(d1) * d2 * d3;
}
-template<typename T>
-size_t num_elements_4d(int d1, int d2, int d3, int d4) {
+template <typename T> size_t num_elements_4d(int d1, int d2, int d3, int d4) {
return static_cast<size_t>(d1) * d2 * d3 * d4;
}
// Define a helper struct that holds the two types.
-template<typename InT, typename OutT>
-struct MixedPrecision {
- using in_type = InT;
+template <typename InT, typename OutT> struct MixedPrecision {
+ using in_type = InT;
using out_type = OutT;
};
// Define the list of type pairs we want to test.
-typedef ::testing::Types< MixedPrecision<double, double>,
- MixedPrecision<double, float>,
- MixedPrecision<float, float> > MixedTypes;
+typedef ::testing::Types<MixedPrecision<double, double>,
+ MixedPrecision<double, float>,
+ MixedPrecision<float, float>>
+ MixedTypes;
class interp_dimensions {
public:
// Constant dimensions.
- static constexpr int nproma = 3; // inner loop length
- static constexpr int nlev = 4; // number of vertical levels
- static constexpr int nblks_c = 2; // number of cell blocks
- static constexpr int nblks_e = 2; // number of edge blocks
- static constexpr int nblks_v = 2; // number of vertex blocks
- static constexpr int rbf_c2grad_dim = 10; // fixed dimension
+ static constexpr int nproma = 3; // inner loop length
+ static constexpr int nlev = 4; // number of vertical levels
+ static constexpr int nblks_c = 2; // number of cell blocks
+ static constexpr int nblks_e = 2; // number of edge blocks
+ static constexpr int nblks_v = 2; // number of vertex blocks
+ static constexpr int rbf_c2grad_dim = 10; // fixed dimension
static constexpr int rbf_vec_dim_c = 9;
static constexpr int rbf_vec_dim_e = 4;
@@ -58,22 +56,24 @@ public:
const int i_endidx_in = nproma - 1;
const int slev = 0;
const int elev = nlev - 1;
- const bool lacc = false; // Not using ACC-specific behavior.
+ const bool lacc = false; // Not using ACC-specific behavior.
const bool acc_async = false; // No asynchronous execution.
};
-// Define a typed test fixture for the functions which have the same input and output types
+// Define a typed test fixture for the functions which have the same input and
+// output types
template <typename T>
-class RbfInterpolTypedTestFixture : public ::testing::Test, public interp_dimensions {
+class RbfInterpolTypedTestFixture : public ::testing::Test,
+ public interp_dimensions {
public:
-
// Data arrays.
- std::vector<T> p_cell_in; // size: nproma * nlev * nblks_c
+ std::vector<T> p_cell_in; // size: nproma * nlev * nblks_c
std::vector<int> rbf_c2grad_idx; // size: rbf_c2grad_dim * nproma * nblks_c
std::vector<int> rbf_c2grad_blk; // size: rbf_c2grad_dim * nproma * nblks_c
- std::vector<int> rbf_vec_idx_c; //size: rbf_vec_dim_c * nproma * nblks_c
+ std::vector<int> rbf_vec_idx_c; // size: rbf_vec_dim_c * nproma * nblks_c
std::vector<int> rbf_vec_blk_c; // size: rbf_vec_dim_c * nproma * nblks_c
- std::vector<T> rbf_c2grad_coeff; // size: rbf_c2grad_dim * 2 * nproma * nblks_c
+ std::vector<T>
+ rbf_c2grad_coeff; // size: rbf_c2grad_dim * 2 * nproma * nblks_c
std::vector<T> grad_x; // size: nproma * nlev * nblks_c
std::vector<T> grad_y; // size: nproma * nlev * nblks_c
std::vector<T> p_vn_in;
@@ -86,17 +86,20 @@ public:
std::vector<T> rbf_vec_coeff_e;
std::vector<T> p_vt_out;
-
RbfInterpolTypedTestFixture() {
size_t size3d = static_cast<size_t>(nproma) * nlev * nblks_c;
size_t size3d_idx = static_cast<size_t>(rbf_c2grad_dim) * nproma * nblks_c;
size_t size4d = static_cast<size_t>(rbf_c2grad_dim) * 2 * nproma * nblks_c;
- size_t size3d_vec_dim = static_cast<size_t>(rbf_vec_dim_c) * nproma * nblks_c;
- size_t size_4d_vec_dim = static_cast<size_t>(rbf_vec_dim_c) * 2 * nproma * nblks_c;
+ size_t size3d_vec_dim =
+ static_cast<size_t>(rbf_vec_dim_c) * nproma * nblks_c;
+ size_t size_4d_vec_dim =
+ static_cast<size_t>(rbf_vec_dim_c) * 2 * nproma * nblks_c;
- size_t size3d_edge_lib = static_cast<size_t>(rbf_vec_dim_e) * nproma * nblks_c;
- size_t size_4d_edge_lib = static_cast<size_t>(rbf_vec_dim_e) * 2 * nproma * nblks_c;
+ size_t size3d_edge_lib =
+ static_cast<size_t>(rbf_vec_dim_e) * nproma * nblks_c;
+ size_t size_4d_edge_lib =
+ static_cast<size_t>(rbf_vec_dim_e) * 2 * nproma * nblks_c;
p_cell_in.resize(size3d, static_cast<T>(1));
p_vn_in.resize(size3d, static_cast<T>(1));
@@ -104,7 +107,7 @@ public:
rbf_vec_idx_c.resize(size3d_vec_dim, 1);
rbf_vec_blk_c.resize(size3d_vec_dim, 0);
rbf_c2grad_idx.resize(size3d_idx, 1);
- rbf_c2grad_blk.resize(size3d_idx, 0); // Set block indices to 0 for testing.
+ rbf_c2grad_blk.resize(size3d_idx, 0); // Set block indices to 0 for testing.
rbf_vec_idx_e.resize(size3d_vec_dim, 1);
rbf_vec_blk_e.resize(size3d_vec_dim, 0);
@@ -128,23 +131,12 @@ TYPED_TEST_SUITE(RbfInterpolTypedTestFixture, MyTypes);
TYPED_TEST(RbfInterpolTypedTestFixture, C2Grad) {
using T = TypeParam;
rbf_interpol_c2grad_lib<TypeParam>(
- this->p_cell_in.data(),
- this->rbf_c2grad_idx.data(),
- this->rbf_c2grad_blk.data(),
- this->rbf_c2grad_coeff.data(),
- this->grad_x.data(),
- this->grad_y.data(),
- this->i_startblk,
- this->i_endblk,
- this->i_startidx_in,
- this->i_endidx_in,
- this->slev,
- this->elev,
- this->nproma,
- this->rbf_c2grad_dim,
- this->nlev,
- this->nblks_c,
- this->lacc);
+ this->p_cell_in.data(), this->rbf_c2grad_idx.data(),
+ this->rbf_c2grad_blk.data(), this->rbf_c2grad_coeff.data(),
+ this->grad_x.data(), this->grad_y.data(), this->i_startblk,
+ this->i_endblk, this->i_startidx_in, this->i_endidx_in, this->slev,
+ this->elev, this->nproma, this->rbf_c2grad_dim, this->nlev, this->nblks_c,
+ this->lacc);
// For each block from i_startblk to i_endblk-1, and for each (i, level)
// the kernel sums rbf_c2grad_dim contributions, each equal to 1.
@@ -152,171 +144,160 @@ TYPED_TEST(RbfInterpolTypedTestFixture, C2Grad) {
for (int jb = this->i_startblk; jb < this->i_endblk; ++jb) {
for (int jk = 0; jk < this->nlev; ++jk) {
for (int i = 0; i < this->nproma; ++i) {
- size_t idx = i + static_cast<size_t>(jk) * this->nproma + static_cast<size_t>(jb) * this->nproma * this->nlev;
- EXPECT_NEAR(this->grad_x[idx], static_cast<TypeParam>(this->rbf_c2grad_dim), static_cast<TypeParam>(1e-5))
- << "grad_x failure at block " << jb << ", level " << jk << ", index " << i;
- EXPECT_NEAR(this->grad_y[idx], static_cast<TypeParam>(this->rbf_c2grad_dim), static_cast<TypeParam>(1e-5))
- << "grad_y failure at block " << jb << ", level " << jk << ", index " << i;
+ size_t idx = i + static_cast<size_t>(jk) * this->nproma +
+ static_cast<size_t>(jb) * this->nproma * this->nlev;
+ EXPECT_NEAR(this->grad_x[idx],
+ static_cast<TypeParam>(this->rbf_c2grad_dim),
+ static_cast<TypeParam>(1e-5))
+ << "grad_x failure at block " << jb << ", level " << jk
+ << ", index " << i;
+ EXPECT_NEAR(this->grad_y[idx],
+ static_cast<TypeParam>(this->rbf_c2grad_dim),
+ static_cast<TypeParam>(1e-5))
+ << "grad_y failure at block " << jb << ", level " << jk
+ << ", index " << i;
}
}
}
}
-
TYPED_TEST(RbfInterpolTypedTestFixture, Cell) {
- using T = TypeParam;
-
- rbf_vec_interpol_cell_lib<T>(
- this->p_vn_in.data(),
- this->rbf_vec_idx_c.data(),
- this->rbf_vec_blk_c.data(),
- this->rbf_vec_coeff_c.data(),
- this->p_u_out.data(),
- this->p_v_out.data(),
- this->i_startblk,
- this->i_endblk,
- this->i_startidx_in,
- this->i_endidx_in,
- this->slev,
- this->elev,
- this->nproma,
- this->rbf_c2grad_dim,
- this->nlev,
- this->nblks_c,
- this->nblks_e,
- this->lacc,
- this->acc_async);
-
- for (int jb = this->i_startblk; jb < this->i_endblk; ++jb) {
- for (int jk = 0; jk < this->nlev; ++jk) {
- for (int i = 0; i < this->nproma; ++i) {
- size_t idx = i + static_cast<size_t>(jk) * this->nproma + static_cast<size_t>(jb) * this->nproma * this->nlev;
- EXPECT_NEAR(this->p_u_out[idx], static_cast<T>(this->rbf_vec_dim_c), static_cast<T>(1e-5))
- << "p_u_out failure at block " << jb << ", level " << jk << ", index " << i;
- }
- }
+ using T = TypeParam;
+
+ rbf_vec_interpol_cell_lib<T>(
+ this->p_vn_in.data(), this->rbf_vec_idx_c.data(),
+ this->rbf_vec_blk_c.data(), this->rbf_vec_coeff_c.data(),
+ this->p_u_out.data(), this->p_v_out.data(), this->i_startblk,
+ this->i_endblk, this->i_startidx_in, this->i_endidx_in, this->slev,
+ this->elev, this->nproma, this->rbf_c2grad_dim, this->nlev, this->nblks_c,
+ this->nblks_e, this->lacc, this->acc_async);
+
+ for (int jb = this->i_startblk; jb < this->i_endblk; ++jb) {
+ for (int jk = 0; jk < this->nlev; ++jk) {
+ for (int i = 0; i < this->nproma; ++i) {
+ size_t idx = i + static_cast<size_t>(jk) * this->nproma +
+ static_cast<size_t>(jb) * this->nproma * this->nlev;
+ EXPECT_NEAR(this->p_u_out[idx], static_cast<T>(this->rbf_vec_dim_c),
+ static_cast<T>(1e-5))
+ << "p_u_out failure at block " << jb << ", level " << jk
+ << ", index " << i;
}
+ }
}
+}
-TYPED_TEST(RbfInterpolTypedTestFixture, Egde){
- using T = TypeParam;
-
-
- rbf_vec_interpol_edge_lib<T>(
- this->p_vn_in.data(),
- this->rbf_vec_idx_e.data(),
- this->rbf_vec_blk_e.data(),
- this->rbf_vec_coeff_e.data(),
- this->p_vt_out.data(),
- this->i_startblk,
- this->i_endblk,
- this->i_startidx_in,
- this->i_endidx_in,
- this->slev,
- this->elev,
- this->nlev,
- this->nproma,
- this->rbf_vec_dim_e,
- this->nblks_e,
- this->lacc,
- this->acc_async);
-
- for (int jb = this->i_startblk; jb < this->i_endblk; ++jb) {
- for (int jk = 0; jk < this->nlev; ++jk) {
- for (int i = 0; i < this->nproma; ++i) {
- size_t idx = i + static_cast<size_t>(jk) * this->nproma + static_cast<size_t>(jb) * this->nproma * this->nlev;
- EXPECT_NEAR(this->p_vt_out[idx], static_cast<T>(this->rbf_vec_dim_e), static_cast<T>(1e-5))
- << "p_vt_out failure at block " << jb << ", level " << jk << ", index " << i;
- }
- }
- }
+TYPED_TEST(RbfInterpolTypedTestFixture, Egde) {
+ using T = TypeParam;
-}
+ rbf_vec_interpol_edge_lib<T>(
+ this->p_vn_in.data(), this->rbf_vec_idx_e.data(),
+ this->rbf_vec_blk_e.data(), this->rbf_vec_coeff_e.data(),
+ this->p_vt_out.data(), this->i_startblk, this->i_endblk,
+ this->i_startidx_in, this->i_endidx_in, this->slev, this->elev,
+ this->nlev, this->nproma, this->rbf_vec_dim_e, this->nblks_e, this->lacc,
+ this->acc_async);
+ for (int jb = this->i_startblk; jb < this->i_endblk; ++jb) {
+ for (int jk = 0; jk < this->nlev; ++jk) {
+ for (int i = 0; i < this->nproma; ++i) {
+ size_t idx = i + static_cast<size_t>(jk) * this->nproma +
+ static_cast<size_t>(jb) * this->nproma * this->nlev;
+ EXPECT_NEAR(this->p_vt_out[idx], static_cast<T>(this->rbf_vec_dim_e),
+ static_cast<T>(1e-5))
+ << "p_vt_out failure at block " << jb << ", level " << jk
+ << ", index " << i;
+ }
+ }
+ }
+}
-// Define a typed test fixture for the functions which have different input and output types
+// Define a typed test fixture for the functions which have different input and
+// output types
template <typename TypePair>
-class RbfVecInterpolMixedTestFixture : public ::testing::Test, public interp_dimensions {
+class RbfVecInterpolMixedTestFixture : public ::testing::Test,
+ public interp_dimensions {
public:
- using InType = typename TypePair::in_type;
+ using InType = typename TypePair::in_type;
using OutType = typename TypePair::out_type;
// Constant dimensions.
- static constexpr int nproma = 3; // inner loop length
- static constexpr int nlev = 4; // number of vertical levels
- static constexpr int nblks_e = 2; // number of edge blocks (for p_e_in)
- static constexpr int nblks_v = 2; // number of vertex blocks (for rbf arrays and outputs)
- static constexpr int rbf_vec_dim = 6; // fixed dimension for rbf vector (stencil points)
+ static constexpr int nproma = 3; // inner loop length
+ static constexpr int nlev = 4; // number of vertical levels
+ static constexpr int nblks_e = 2; // number of edge blocks (for p_e_in)
+ static constexpr int nblks_v =
+ 2; // number of vertex blocks (for rbf arrays and outputs)
+ static constexpr int rbf_vec_dim =
+ 6; // fixed dimension for rbf vector (stencil points)
// Parameter values.
- int i_startblk = 0;
- int i_endblk = 1; // Test blocks [0, 1]
+ int i_startblk = 0;
+ int i_endblk = 1; // Test blocks [0, 1]
int i_startidx_in = 0;
- int i_endidx_in = nproma - 1; // Full range: 0 .. nproma-1
- int slev = 0;
- int elev = nlev - 1; // Full vertical range (0 .. nlev-1)
- bool lacc = false; // Not using ACC-specific behavior.
- bool acc_async = false; // No asynchronous execution.
+ int i_endidx_in = nproma - 1; // Full range: 0 .. nproma-1
+ int slev = 0;
+ int elev = nlev - 1; // Full vertical range (0 .. nlev-1)
+ bool lacc = false; // Not using ACC-specific behavior.
+ bool acc_async = false; // No asynchronous execution.
// Arrays stored in std::vector.
- std::vector<InType> p_e_in; // Dimensions: (nproma, nlev, nblks_e)
- std::vector<int> rbf_vec_idx_v; // Dimensions: (rbf_vec_dim, nproma, nblks_v)
- std::vector<int> rbf_vec_blk_v; // Dimensions: (rbf_vec_dim, nproma, nblks_v)
- std::vector<InType> rbf_vec_coeff_v; // Dimensions: (rbf_vec_dim, 2, nproma, nblks_v)
- std::vector<OutType> p_u_out; // Dimensions: (nproma, nlev, nblks_v)
- std::vector<OutType> p_v_out; // Dimensions: (nproma, nlev, nblks_v)
+ std::vector<InType> p_e_in; // Dimensions: (nproma, nlev, nblks_e)
+ std::vector<int> rbf_vec_idx_v; // Dimensions: (rbf_vec_dim, nproma, nblks_v)
+ std::vector<int> rbf_vec_blk_v; // Dimensions: (rbf_vec_dim, nproma, nblks_v)
+ std::vector<InType>
+ rbf_vec_coeff_v; // Dimensions: (rbf_vec_dim, 2, nproma, nblks_v)
+ std::vector<OutType> p_u_out; // Dimensions: (nproma, nlev, nblks_v)
+ std::vector<OutType> p_v_out; // Dimensions: (nproma, nlev, nblks_v)
RbfVecInterpolMixedTestFixture() {
// Allocate and initialize inputs.
- p_e_in.resize(num_elements_3d<InType>(nproma, nlev, nblks_e), static_cast<InType>(1));
+ p_e_in.resize(num_elements_3d<InType>(nproma, nlev, nblks_e),
+ static_cast<InType>(1));
rbf_vec_idx_v.resize(num_elements_3d<int>(rbf_vec_dim, nproma, nblks_v), 1);
rbf_vec_blk_v.resize(num_elements_3d<int>(rbf_vec_dim, nproma, nblks_v), 0);
- rbf_vec_coeff_v.resize(num_elements_4d<InType>(rbf_vec_dim, 2, nproma, nblks_v), static_cast<InType>(1));
+ rbf_vec_coeff_v.resize(
+ num_elements_4d<InType>(rbf_vec_dim, 2, nproma, nblks_v),
+ static_cast<InType>(1));
// Allocate output arrays and initialize to zero.
- p_u_out.resize(num_elements_3d<OutType>(nproma, nlev, nblks_v), static_cast<OutType>(0));
- p_v_out.resize(num_elements_3d<OutType>(nproma, nlev, nblks_v), static_cast<OutType>(0));
+ p_u_out.resize(num_elements_3d<OutType>(nproma, nlev, nblks_v),
+ static_cast<OutType>(0));
+ p_v_out.resize(num_elements_3d<OutType>(nproma, nlev, nblks_v),
+ static_cast<OutType>(0));
}
};
TYPED_TEST_SUITE(RbfVecInterpolMixedTestFixture, MixedTypes);
TYPED_TEST(RbfVecInterpolMixedTestFixture, Vertex) {
- using InType = typename TestFixture::InType;
+ using InType = typename TestFixture::InType;
using OutType = typename TestFixture::OutType;
// Call the function with mixed precision.
rbf_vec_interpol_vertex_lib<InType, OutType>(
- this->p_e_in.data(),
- this->rbf_vec_idx_v.data(),
- this->rbf_vec_blk_v.data(),
- this->rbf_vec_coeff_v.data(),
- this->p_u_out.data(),
- this->p_v_out.data(),
- this->i_startblk,
- this->i_endblk,
- this->i_startidx_in,
- this->i_endidx_in,
- this->slev,
- this->elev,
- this->nproma,
- this->lacc,
- this->acc_async,
- this->nlev,
- this->nblks_e,
- this->nblks_v);
+ this->p_e_in.data(), this->rbf_vec_idx_v.data(),
+ this->rbf_vec_blk_v.data(), this->rbf_vec_coeff_v.data(),
+ this->p_u_out.data(), this->p_v_out.data(), this->i_startblk,
+ this->i_endblk, this->i_startidx_in, this->i_endidx_in, this->slev,
+ this->elev, this->nproma, this->lacc, this->acc_async, this->nlev,
+ this->nblks_e, this->nblks_v);
// Check the outputs only for blocks in the range [i_startblk, i_endblk].
for (int block = this->i_startblk; block <= this->i_endblk; ++block) {
for (int level = 0; level < this->nlev; ++level) {
for (int i = 0; i < this->nproma; ++i) {
// Compute the linear index for a 3D array in column-major order:
- size_t idx = i + level * this->nproma + block * this->nproma * this->nlev;
- // Since every contribution is 1 and there are 6 stencil points, expect 6.
- EXPECT_NEAR(this->p_u_out[idx], static_cast<OutType>(6), static_cast<OutType>(1e-5))
- << "Failure at block " << block << ", level " << level << ", index " << i;
- EXPECT_NEAR(this->p_v_out[idx], static_cast<OutType>(6), static_cast<OutType>(1e-5))
- << "Failure at block " << block << ", level " << level << ", index " << i;
+ size_t idx =
+ i + level * this->nproma + block * this->nproma * this->nlev;
+ // Since every contribution is 1 and there are 6 stencil points,
+ // expect 6.
+ EXPECT_NEAR(this->p_u_out[idx], static_cast<OutType>(6),
+ static_cast<OutType>(1e-5))
+ << "Failure at block " << block << ", level " << level << ", index "
+ << i;
+ EXPECT_NEAR(this->p_v_out[idx], static_cast<OutType>(6),
+ static_cast<OutType>(1e-5))
+ << "Failure at block " << block << ", level " << level << ", index "
+ << i;
}
}
}
--
GitLab
From 2cbc4234c219596f9f90816724d6cb5555ddea3a Mon Sep 17 00:00:00 2001
From: Pradipta Samanta <samanta@dkrz.de>
Date: Thu, 6 Mar 2025 18:36:01 +0100
Subject: [PATCH 31/33] added some comments to the bindings
---
src/interpolation/interpolation_bindings.cpp | 11 ++++++++++-
src/interpolation/interpolation_bindings.h | 8 ++++++++
2 files changed, 18 insertions(+), 1 deletion(-)
diff --git a/src/interpolation/interpolation_bindings.cpp b/src/interpolation/interpolation_bindings.cpp
index a13ad0f..4524ad7 100644
--- a/src/interpolation/interpolation_bindings.cpp
+++ b/src/interpolation/interpolation_bindings.cpp
@@ -328,6 +328,7 @@ void cell_avg_lib_sp(const float *psi_c, const int *cell_neighbor_idx,
i_endidx_in, slev, elev, nproma, nlev, nblks_c, lacc);
}
+// This is the binding for mo_intp_rbf::rbf_vec_interpol_vertex_dp_lib
void rbf_vec_interpol_vertex_lib_dp(
const double *p_e_in, const int *rbf_vec_idx_v, const int *rbf_vec_blk_v,
const double *rbf_vec_coeff_v, double *p_u_out, double *p_v_out,
@@ -341,6 +342,7 @@ void rbf_vec_interpol_vertex_lib_dp(
lacc, acc_async, nlev, nblks_e, nblks_v);
}
+// This is the binding for mo_intp_rbf::rbf_vec_interpol_vertex_sp_lib
void rbf_vec_interpol_vertex_lib_sp(
const float *p_e_in, const int *rbf_vec_idx_v, const int *rbf_vec_blk_v,
const float *rbf_vec_coeff_v, float *p_u_out, float *p_v_out,
@@ -354,7 +356,8 @@ void rbf_vec_interpol_vertex_lib_sp(
lacc, acc_async, nlev, nblks_e, nblks_v);
}
-void rbf_vec_interpol_vertex_lib_mixprec(
+// This is the binding for mo_intp_rbf::rbf_vec_interpol_vertex_dpsp_lib
+void rbf_vec_interpol_vertex_lib_dpsp(
const double *p_e_in, const int *rbf_vec_idx_v, const int *rbf_vec_blk_v,
const double *rbf_vec_coeff_v, float *p_u_out, float *p_v_out,
const int i_startblk, const int i_endblk, const int i_startidx_in,
@@ -367,6 +370,7 @@ void rbf_vec_interpol_vertex_lib_mixprec(
lacc, acc_async, nlev, nblks_e, nblks_v);
}
+// This is the binding for mo_intp_rbf::rbf_vec_interpol_cell_lib (wp=dp)
void rbf_interpol_c2grad_lib_sp(const float *p_cell_in,
const int *rbf_c2grad_idx,
const int *rbf_c2grad_blk,
@@ -382,6 +386,7 @@ void rbf_interpol_c2grad_lib_sp(const float *p_cell_in,
nproma, rbf_c2grad_dim, nlev, nblk_c, lacc);
}
+// This is the binding for mo_intp_rbf::rbf_vec_interpol_cell_lib (wp=sp)
void rbf_interpol_c2grad_lib_dp(const double *p_cell_in,
const int *rbf_c2grad_idx,
const int *rbf_c2grad_blk,
@@ -397,6 +402,7 @@ void rbf_interpol_c2grad_lib_dp(const double *p_cell_in,
nproma, rbf_c2grad_dim, nlev, nblk_c, lacc);
}
+// This is the binding for mo_intp_rbf::rbf_vec_interpol_cell_lib (wp=dp)
void rbf_vec_interpol_cell_lib_sp(
const float *p_vn_in, const int *rbf_vec_idx_c, const int *rbf_vec_blk_c,
const float *rbf_vec_coeff_c, float *p_u_out, float *p_v_out,
@@ -410,6 +416,7 @@ void rbf_vec_interpol_cell_lib_sp(
nlev, nblks_c, nblks_e, rbf_vec_dim_c, lacc, acc_async);
}
+// This is the binding for mo_intp_rbf::rbf_vec_interpol_cell_lib (wp=sp)
void rbf_vec_interpol_cell_lib_dp(
const double *p_vn_in, const int *rbf_vec_idx_c, const int *rbf_vec_blk_c,
const double *rbf_vec_coeff_c, double *p_u_out, double *p_v_out,
@@ -423,6 +430,7 @@ void rbf_vec_interpol_cell_lib_dp(
nlev, nblks_c, nblks_e, rbf_vec_dim_c, lacc, acc_async);
}
+// This is the binding for mo_intp_rbf::rbf_vec_interpol_edge_lib (wp=dp)
void rbf_vec_interpol_edge_lib_dp(
const double *p_vn_in, const int *rbf_vec_idx_e, const int *rbf_vec_blk_e,
const double *rbf_vec_coeff_e, double *p_vt_out, int i_startblk,
@@ -436,6 +444,7 @@ void rbf_vec_interpol_edge_lib_dp(
nproma, rbf_vec_dim_e, nblks_e, lacc, acc_async);
}
+// This is the binding for mo_intp_rbf::rbf_vec_interpol_edge_lib (wp=sp)
void rbf_vec_interpol_edge_lib_sp(
const float *p_vn_in, const int *rbf_vec_idx_e, const int *rbf_vec_blk_e,
const float *rbf_vec_coeff_e, float *p_vt_out, int i_startblk, int i_endblk,
diff --git a/src/interpolation/interpolation_bindings.h b/src/interpolation/interpolation_bindings.h
index 0f4778d..64c6a8c 100644
--- a/src/interpolation/interpolation_bindings.h
+++ b/src/interpolation/interpolation_bindings.h
@@ -202,6 +202,14 @@ void rbf_vec_interpol_vertex_lib_sp(
const bool lacc, const bool acc_async, const int nlev, const int nblks_e,
const int nblks_v);
+void rbf_vec_interpol_vertex_lib_dpsp(
+ const double *p_e_in, const int *rbf_vec_idx_v, const int *rbf_vec_blk_v,
+ const double *rbf_vec_coeff_v, float *p_u_out, float *p_v_out,
+ const int i_startblk, const int i_endblk, const int i_startidx_in,
+ const int i_endidx_in, const int slev, const int elev, const int nproma,
+ const bool lacc, const bool acc_async, const int nlev, const int nblks_e,
+ const int nblks_v);
+
void rbf_interpol_c2grad_lib_sp(
const float *p_cell_in, const int *rbf_c2grad_idx,
const int *rbf_c2grad_blk, const float *rbf_c2grad_coeff, float *grad_x,
--
GitLab
From 399e08caf80cc42ca7925ae1b7f7f921e9df0aef Mon Sep 17 00:00:00 2001
From: Pradipta Samanta <samanta@dkrz.de>
Date: Thu, 6 Mar 2025 18:42:27 +0100
Subject: [PATCH 32/33] fixed a typo
---
test/c/test_intp_rbf.cpp | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/test/c/test_intp_rbf.cpp b/test/c/test_intp_rbf.cpp
index cdee264..040d440 100644
--- a/test/c/test_intp_rbf.cpp
+++ b/test/c/test_intp_rbf.cpp
@@ -186,7 +186,7 @@ TYPED_TEST(RbfInterpolTypedTestFixture, Cell) {
}
}
-TYPED_TEST(RbfInterpolTypedTestFixture, Egde) {
+TYPED_TEST(RbfInterpolTypedTestFixture, Edge) {
using T = TypeParam;
rbf_vec_interpol_edge_lib<T>(
--
GitLab
From 609c5e5804c1b75ccff24ec3ce53126f0cc1a7d7 Mon Sep 17 00:00:00 2001
From: Pradipta Samanta <samanta@dkrz.de>
Date: Thu, 6 Mar 2025 18:48:50 +0100
Subject: [PATCH 33/33] Revert "Fixed strange typo"
This reverts commit 8175944d3f080ec145d4ce59e2e6fdddb8d6153d.
---
src/horizontal/mo_lib_divrot.F90 | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/src/horizontal/mo_lib_divrot.F90 b/src/horizontal/mo_lib_divrot.F90
index c1360f6..a2f2ad9 100644
--- a/src/horizontal/mo_lib_divrot.F90
+++ b/src/horizontal/mo_lib_divrot.F90
@@ -561,7 +561,7 @@ CONTAINS
!!
!! !LITERATURE
!! Ollivier-Gooch et al (2002): A High-Order-Accurate Unstructured Mesh
-!! Finite-Volume Scheme for the Advection-Diffusion Equation, J. Compute. Phys.,
+!! Finite-Volume Scheme for the Advection-Diffusion Equation, J. Comput. Phys.,
!! 181, 729-752
!!
SUBROUTINE recon_lsq_cell_q_lib(p_cc, lsq_idx_c, lsq_blk_c, &
@@ -800,7 +800,7 @@ CONTAINS
!!
!! !LITERATURE
!! Ollivier-Gooch et al (2002): A High-Order-Accurate Unstructured Mesh
-!! Finite-Volume Scheme for the Advection-Diffusion Equation, J. Compute. Phys.,
+!! Finite-Volume Scheme for the Advection-Diffusion Equation, J. Comput. Phys.,
!! 181, 729-752
!!
SUBROUTINE recon_lsq_cell_q_svd_lib(p_cc, lsq_idx_c, lsq_blk_c, &
@@ -1008,7 +1008,7 @@ CONTAINS
!!
!! !LITERATURE
!! Ollivier-Gooch et al (2002): A High-Order-Accurate Unstructured Mesh
-!! Finite-Volume Scheme for the Advection-Diffusion Equation, J. Compute. Phys.,
+!! Finite-Volume Scheme for the Advection-Diffusion Equation, J. Comput. Phys.,
!! 181, 729-752
!!
SUBROUTINE recon_lsq_cell_c_lib(p_cc, lsq_idx_c, lsq_blk_c, &
@@ -1295,7 +1295,7 @@ CONTAINS
!!
!! !LITERATURE
!! Ollivier-Gooch et al (2002): A High-Order-Accurate Unstructured Mesh
-!! Finite-Volume Scheme for the Advection-Diffusion Equation, J. Compute. Phys.,
+!! Finite-Volume Scheme for the Advection-Diffusion Equation, J. Comput. Phys.,
!! 181, 729-752
!!
SUBROUTINE recon_lsq_cell_c_svd_lib(p_cc, lsq_idx_c, lsq_blk_c, &
--
GitLab