Venado optimizations

.devcontainer/devcontainer.json

@@ -6,7 +6,9 @@
 				"fortran-lang.linter-gfortran",
 				"GitHub.vscode-pull-request-github",
 				"ms-vscode.cmake-tools",
+				"ms-vscode.cpptools-extension-pack",
 				"ms-vscode.cpptools",
+				"redhat.vscode-yaml",
 				"twxs.cmake"
 			]
 		}

.github/workflows/CI.yaml

@@ -3,7 +3,7 @@
 name: CI
 on:
   release:
-    type:
+    types:
       - created
   push:
     branches:
@@ -26,13 +26,13 @@ jobs:
     runs-on: ubuntu-20.04
     steps:
       - name: Check out sources
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
       - name: Prepare container
         run: ./scripts/prepare-container-focal.sh
       - run: bundle install
       - run: bundle exec danger || true
       - run: BML_OPENMP=no EMACS=emacs27 ./build.sh --debug check_indent
-      - uses: actions/upload-artifact@v3
+      - uses: actions/upload-artifact@v4
         if: failure()
         with:
           name: lint-artifacts
@@ -45,7 +45,7 @@ jobs:
     runs-on: ubuntu-20.04
     steps:
       - name: Check out sources
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
       - name: Prepare container
         run: ./scripts/prepare-container-focal.sh
       - name: Build docs
@@ -58,25 +58,25 @@ jobs:
       fail-fast: false
       matrix:
         include:
-          - JOBNAME: gcc-11 C single real
-            TEST_SCRIPT: ./scripts/ci-gcc-11-C-single-real.sh
-          - JOBNAME: gcc-11 C double real
-            TEST_SCRIPT: ./scripts/ci-gcc-11-C-double-real.sh
-          - JOBNAME: gcc-11 C single complex
-            TEST_SCRIPT: ./scripts/ci-gcc-11-C-single-complex.sh
-          - JOBNAME: gcc-11 C double complex
-            TEST_SCRIPT: ./scripts/ci-gcc-11-C-double-complex.sh
-          - JOBNAME: gcc-11 Fortran single real
-            TEST_SCRIPT: ./scripts/ci-gcc-11-Fortran-single-real.sh
-          - JOBNAME: gcc-11 Fortran double real
-            TEST_SCRIPT: ./scripts/ci-gcc-11-Fortran-double-real.sh
-          - JOBNAME: gcc-11 Fortran single complex
-            TEST_SCRIPT: ./scripts/ci-gcc-11-Fortran-single-complex.sh
-          - JOBNAME: gcc-11 Fortran double complex
-            TEST_SCRIPT: ./scripts/ci-gcc-11-Fortran-double-complex.sh
+          - JOBNAME: gcc-14 C single real
+            TEST_SCRIPT: ./scripts/ci-gcc-14-C-single-real.sh
+          - JOBNAME: gcc-14 C double real
+            TEST_SCRIPT: ./scripts/ci-gcc-14-C-double-real.sh
+          - JOBNAME: gcc-14 C single complex
+            TEST_SCRIPT: ./scripts/ci-gcc-14-C-single-complex.sh
+          - JOBNAME: gcc-14 C double complex
+            TEST_SCRIPT: ./scripts/ci-gcc-14-C-double-complex.sh
+          - JOBNAME: gcc-14 Fortran single real
+            TEST_SCRIPT: ./scripts/ci-gcc-14-Fortran-single-real.sh
+          - JOBNAME: gcc-14 Fortran double real
+            TEST_SCRIPT: ./scripts/ci-gcc-14-Fortran-double-real.sh
+          - JOBNAME: gcc-14 Fortran single complex
+            TEST_SCRIPT: ./scripts/ci-gcc-14-Fortran-single-complex.sh
+          - JOBNAME: gcc-14 Fortran double complex
+            TEST_SCRIPT: ./scripts/ci-gcc-14-Fortran-double-complex.sh
     steps:
       - name: Check out sources
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
         with:
           fetch-depth: 0
       - name: Prepare container
@@ -141,7 +141,7 @@ jobs:
             TEST_SCRIPT: ./scripts/ci-clang-16-gfortran-10-C-double-complex.sh
     steps:
       - name: Check out sources
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
         with:
           fetch-depth: 0
       - name: Prepare container

.github/workflows/container.yaml

@@ -19,7 +19,7 @@ jobs:
     if: ${{ github.repository == 'lanl/bml' }}
     steps:
       - name: Checkout
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
       - name: Login to DockerHub
         uses: docker/login-action@v2
         with:

scripts/ci-gcc-11-C-double-complex.sh → scripts/ci-gcc-14-C-double-complex.sh

@@ -10,9 +10,9 @@ done
 
 basedir=$(${READLINK} --canonicalize $(dirname $0)/..)
 
-export CC=${CC:-gcc-11}
-export CXX=${CXX:-g++-11}
-export FC=${FC:-gfortran-11}
+export CC=${CC:-gcc-14}
+export CXX=${CXX:-g++-14}
+export FC=${FC:-gfortran-14}
 export BUILD_SHARED_LIBS=${BUILD_SHARED_LIBS:-no}
 export BML_OPENMP=${BML_OPENMP:-no}
 export BML_INTERNAL_BLAS=${BML_INTERNAL_BLAS:-no}

scripts/ci-gcc-11-C-double-real.sh → scripts/ci-gcc-14-C-double-real.sh

@@ -10,9 +10,9 @@ done
 
 basedir=$(${READLINK} --canonicalize $(dirname $0)/..)
 
-export CC=${CC:-gcc-11}
-export CXX=${CXX:-g++-11}
-export FC=${FC:-gfortran-11}
+export CC=${CC:-gcc-14}
+export CXX=${CXX:-g++-14}
+export FC=${FC:-gfortran-14}
 export BUILD_SHARED_LIBS=${BUILD_SHARED_LIBS:-no}
 export BML_OPENMP=${BML_OPENMP:-no}
 export BML_INTERNAL_BLAS=${BML_INTERNAL_BLAS:-no}

scripts/ci-gcc-11-C-single-complex.sh → scripts/ci-gcc-14-C-single-complex.sh

@@ -10,9 +10,9 @@ done
 
 basedir=$(${READLINK} --canonicalize $(dirname $0)/..)
 
-export CC=${CC:-gcc-11}
-export CXX=${CXX:-g++-11}
-export FC=${FC:-gfortran-11}
+export CC=${CC:-gcc-14}
+export CXX=${CXX:-g++-14}
+export FC=${FC:-gfortran-14}
 export BUILD_SHARED_LIBS=${BUILD_SHARED_LIBS:-no}
 export BML_OPENMP=${BML_OPENMP:-no}
 export BML_INTERNAL_BLAS=${BML_INTERNAL_BLAS:-no}

scripts/ci-gcc-11-C-single-real.sh → scripts/ci-gcc-14-C-single-real.sh

@@ -10,9 +10,9 @@ done
 
 basedir=$(${READLINK} --canonicalize $(dirname $0)/..)
 
-export CC=${CC:-gcc-11}
-export CXX=${CXX:-g++-11}
-export FC=${FC:-gfortran-11}
+export CC=${CC:-gcc-14}
+export CXX=${CXX:-g++-14}
+export FC=${FC:-gfortran-14}
 export BUILD_SHARED_LIBS=${BUILD_SHARED_LIBS:-no}
 export BML_OPENMP=${BML_OPENMP:-no}
 export BML_INTERNAL_BLAS=${BML_INTERNAL_BLAS:-no}

scripts/ci-gcc-11-Fortran-double-complex.sh → scripts/ci-gcc-14-Fortran-double-complex.sh

@@ -10,9 +10,9 @@ done
 
 basedir=$(${READLINK} --canonicalize $(dirname $0)/..)
 
-export CC=${CC:-gcc-11}
-export CXX=${CXX:-g++-11}
-export FC=${FC:-gfortran-11}
+export CC=${CC:-gcc-14}
+export CXX=${CXX:-g++-14}
+export FC=${FC:-gfortran-14}
 export BUILD_SHARED_LIBS=${BUILD_SHARED_LIBS:-no}
 export BML_OPENMP=${BML_OPENMP:-yes}
 export BML_INTERNAL_BLAS=${BML_INTERNAL_BLAS:-no}

scripts/ci-gcc-11-Fortran-double-real.sh → scripts/ci-gcc-14-Fortran-double-real.sh

@@ -10,9 +10,9 @@ done
 
 basedir=$(${READLINK} --canonicalize $(dirname $0)/..)
 
-export CC=${CC:-gcc-11}
-export CXX=${CXX:-g++-11}
-export FC=${FC:-gfortran-11}
+export CC=${CC:-gcc-14}
+export CXX=${CXX:-g++-14}
+export FC=${FC:-gfortran-14}
 export BUILD_SHARED_LIBS=${BUILD_SHARED_LIBS:-no}
 export BML_OPENMP=${BML_OPENMP:-yes}
 export BML_INTERNAL_BLAS=${BML_INTERNAL_BLAS:-no}

scripts/ci-gcc-11-Fortran-single-complex.sh → scripts/ci-gcc-14-Fortran-single-complex.sh

@@ -10,9 +10,9 @@ done
 
 basedir=$(${READLINK} --canonicalize $(dirname $0)/..)
 
-export CC=${CC:-gcc-11}
-export CXX=${CXX:-g++-11}
-export FC=${FC:-gfortran-11}
+export CC=${CC:-gcc-14}
+export CXX=${CXX:-g++-14}
+export FC=${FC:-gfortran-14}
 export BUILD_SHARED_LIBS=${BUILD_SHARED_LIBS:-no}
 export BML_OPENMP=${BML_OPENMP:-yes}
 export BML_INTERNAL_BLAS=${BML_INTERNAL_BLAS:-no}

scripts/ci-gcc-11-Fortran-single-real.sh → scripts/ci-gcc-14-Fortran-single-real.sh

@@ -10,9 +10,9 @@ done
 
 basedir=$(${READLINK} --canonicalize $(dirname $0)/..)
 
-export CC=${CC:-gcc-11}
-export CXX=${CXX:-g++-11}
-export FC=${FC:-gfortran-11}
+export CC=${CC:-gcc-14}
+export CXX=${CXX:-g++-14}
+export FC=${FC:-gfortran-14}
 export BUILD_SHARED_LIBS=${BUILD_SHARED_LIBS:-no}
 export BML_OPENMP=${BML_OPENMP:-yes}
 export BML_INTERNAL_BLAS=${BML_INTERNAL_BLAS:-no}

src/C-interface/dense/bml_allocate_dense_typed.c

@@ -98,6 +98,7 @@ bml_matrix_dense_t *TYPED_FUNC(
     A->matrix_type = dense;
     A->matrix_precision = MATRIX_PRECISION;
     A->N = matrix_dimension.N_rows;
+    A->N_allocated = A->N;
     A->distribution_mode = distrib_mode;
 #ifdef BML_USE_MAGMA
     A->ld = magma_roundup(matrix_dimension.N_rows, 32);
@@ -154,6 +155,7 @@ bml_matrix_dense_t *TYPED_FUNC(
     A->matrix_type = dense;
     A->matrix_precision = MATRIX_PRECISION;
     A->N = matrix_dimension.N_rows;
+    A->N_allocated = A->N;
     A->distribution_mode = distrib_mode;
 #ifdef BML_USE_MAGMA
     A->ld = magma_roundup(matrix_dimension.N_rows, 32);

src/C-interface/dense/bml_introspection_dense.c

@@ -200,3 +200,22 @@ bml_get_data_ptr_dense(
 {
     return A->matrix;
 }
+
+/** Return the dense matrix ld parameter.
+ *
+ * \param A The dense matrix.
+ * \return The matrix ld parameter.
+ */
+int
+bml_get_ld_dense(
+    bml_matrix_dense_t * A)
+{
+    if (A != NULL)
+    {
+        return A->ld;
+    }
+    else
+    {
+        return -1;
+    }
+}

src/C-interface/dense/bml_setters_dense.c

@@ -2,6 +2,37 @@
 #include "../bml_logger.h"
 #include "bml_setters_dense.h"
 #include "bml_types_dense.h"
+#include "bml_allocate_dense.h"
+
+#ifdef BML_USE_MAGMA
+#include "magma_v2.h"
+#endif
+
+void
+bml_set_N_dense(
+    bml_matrix_dense_t * A,
+    int N)
+{
+    if (A->N <= A->N_allocated)
+    {
+        A->N = N;
+#ifdef BML_USE_MAGMA
+        A->ld = magma_roundup(A->N, 32);
+#else
+        A->ld = A->N;
+#endif
+    }
+    else
+    {
+        bml_matrix_dense_t *B;
+        bml_matrix_dimension_t matrix_dimension = { A->N, A->N, A->N };
+
+        B = bml_noinit_matrix_dense(A->matrix_precision, matrix_dimension,
+                                    A->distribution_mode);
+        bml_deallocate_dense(A);
+        A = B;
+    }
+}
 
 void
 bml_set_element_dense(

src/C-interface/dense/bml_setters_dense.h

@@ -7,6 +7,10 @@
 
 #include <complex.h>
 
+void bml_set_N_dense(
+    bml_matrix_dense_t * A,
+    int N);
+
 void bml_set_element_dense(
     bml_matrix_dense_t * A,
     int i,

src/C-interface/dense/bml_types_dense.h

@@ -22,6 +22,8 @@ struct bml_matrix_dense_t
     bml_distribution_mode_t distribution_mode;
     /** The number of rows/columns. */
     int N;
+    /** The number of rows/columns originally allocated. */
+    int N_allocated;
     /** The dense matrix. */
     void *matrix;
     /** The leading dimension of the array matrix. */

src/Fortran-interface/bml_c_interface_m.F90

@@ -371,6 +371,18 @@ function bml_get_bandwidth_C(a) bind(C, name="bml_get_bandwidth")
       integer(C_INT) :: bml_get_bandwidth_C
     end function bml_get_bandwidth_C
 
+    function bml_get_data_ptr_dense_C(a) bind(C, name="bml_get_data_ptr_dense")
+      import :: C_PTR
+      type(C_PTR), value, intent(in) :: a
+      type(C_PTR) :: bml_get_data_ptr_dense_C
+    end function bml_get_data_ptr_dense_C
+
+    function bml_get_ld_dense_C(a) bind(C, name="bml_get_ld_dense")
+      import :: C_PTR, C_INT
+      type(C_PTR), value, intent(in) :: a
+      integer(C_INT) :: bml_get_ld_dense_C
+    end function bml_get_ld_dense_C
+
     function bml_get_sparsity_C(a, threshold) bind(C, name="bml_get_sparsity")
       import :: C_PTR, C_DOUBLE, C_INT
       type(C_PTR), value, intent(in) :: a
@@ -448,6 +460,12 @@ subroutine bml_scale_inplace_C(alpha, a) bind(C, name="bml_scale_inplace")
       type(C_PTR), value :: a
     end subroutine bml_scale_inplace_C
 
+    subroutine bml_set_N_dense_C(a,n) bind(C, name="bml_set_N_dense")
+      import :: C_PTR, C_INT
+      type(C_PTR), value, intent(in) :: a
+      integer(C_INT), value, intent(in) :: n
+    end subroutine bml_set_N_dense_C
+
     subroutine bml_set_row_C(a, i, row, threshold) bind(C, name="bml_set_row")
       import :: C_PTR, C_INT, C_DOUBLE
       type(C_PTR), value, intent(in) :: a
@@ -616,6 +634,11 @@ function bml_transpose_new_C(a) bind(C, name="bml_transpose_new")
       type(C_PTR) :: bml_transpose_new_C
     end function bml_transpose_new_C
 
+    subroutine bml_transpose_C(a) bind(C, name="bml_transpose")
+      import :: C_PTR
+      type(C_PTR), value, intent(in) :: a
+    end subroutine bml_transpose_C
+
     subroutine bml_write_bml_matrix_C(a, filename) &
          & bind(C, name="bml_write_bml_matrix")
       import :: C_PTR, C_CHAR

src/Fortran-interface/bml_introspection_m.F90

@@ -18,6 +18,8 @@ module bml_introspection_m
   public :: bml_get_bandwidth
   public :: bml_get_distribution_mode
   public :: bml_get_sparsity
+  public :: bml_get_data_ptr_dense
+  public :: bml_get_ld_dense
 
 contains
 
@@ -247,4 +249,25 @@ function bml_get_sparsity(a, threshold) result(sparsity)
 
   end function bml_get_sparsity
 
+  function bml_get_data_ptr_dense(a)
+    type(bml_matrix_t), intent(inout) :: a
+    type(C_PTR) :: bml_get_data_ptr_dense
+
+    bml_get_data_ptr_dense = bml_get_data_ptr_dense_C(a%ptr)
+
+  end function bml_get_data_ptr_dense
+
+  !> Return the dense matrix ld parameter.
+  !!
+  !!\param a The matrix.
+  !!\return The matrix ld parameter.
+  function bml_get_ld_dense(a)
+
+    type(bml_matrix_t), intent(in) :: a
+    integer :: bml_get_ld_dense
+
+    bml_get_ld_dense = bml_get_ld_dense_C(a%ptr)
+
+  end function bml_get_ld_dense
+
 end module bml_introspection_m