Update armadillo version 12.8.4

Include/armadillo/armadillo

@@ -19,9 +19,9 @@
 #ifndef ARMA_INCLUDES
 #define ARMA_INCLUDES
 
-// NOTE: functions that are designed to be user accessible are described in the documentation (docs.html).
-// NOTE: all other functions and classes (ie. not explicitly described in the documentation)
-// NOTE: are considered as internal implementation details, and may be changed or removed without notice.
+// WARNING: the documentation (docs.html) describes the public API (functions, classes, constants);
+// WARNING: any functionality which is _not explicitly_ described in the documentation
+// WARNING: is considered as internal implementation detail, and may be changed or removed without notice.
 
 #include "armadillo_bits/config.hpp"
 #include "armadillo_bits/compiler_check.hpp"

Include/armadillo/armadillo_bits/SpSubview_meat.hpp

@@ -44,18 +44,28 @@ SpSubview<eT>::SpSubview(const SpMat<eT>& in_m, const uword in_row1, const uword
 
   m.sync_csc();
 
-  // There must be a O(1) way to do this
-  uword lend     = m.col_ptrs[in_col1 + in_n_cols];
-  uword lend_row = in_row1 + in_n_rows;
-  uword count   = 0;
+  // count the number of non-zeros in the subview
+  uword count = 0;
 
-  for(uword i = m.col_ptrs[in_col1]; i < lend; ++i)
+  if(n_rows == m.n_rows)
     {
-    const uword m_row_indices_i = m.row_indices[i];
+    count = m.col_ptrs[aux_col1 + n_cols] - m.col_ptrs[aux_col1];
+    }
+  else
+    {
+    arma_extra_debug_print("counting non-zeros in sparse subview");
 
-    const bool condition = (m_row_indices_i >= in_row1) && (m_row_indices_i < lend_row);
+    uword lend     = m.col_ptrs[in_col1 + in_n_cols];
+    uword lend_row = in_row1 + in_n_rows;
 
-    count += condition ? uword(1) : uword(0);
+    for(uword i = m.col_ptrs[in_col1]; i < lend; ++i)
+      {
+      const uword m_row_indices_i = m.row_indices[i];
+
+      const bool condition = (m_row_indices_i >= in_row1) && (m_row_indices_i < lend_row);
+
+      count += condition ? uword(1) : uword(0);
+      }
     }
 
   access::rw(n_nonzero) = count;

Include/armadillo/armadillo_bits/arma_version.hpp

@@ -23,7 +23,7 @@
 
 #define ARMA_VERSION_MAJOR 12
 #define ARMA_VERSION_MINOR 8
-#define ARMA_VERSION_PATCH 3
+#define ARMA_VERSION_PATCH 4
 #define ARMA_VERSION_NAME  "Cortisol Injector"
 
 

Include/armadillo/armadillo_bits/fn_conv_to.hpp

@@ -30,7 +30,7 @@ class conv_to
 
   template<typename in_eT, typename T1>
   inline static out_eT from(const Base<in_eT, T1>& in, const typename arma_not_cx<in_eT>::result* junk = nullptr);
-
+  
   template<typename in_eT, typename T1>
   inline static out_eT from(const Base<in_eT, T1>& in, const typename arma_cx_only<in_eT>::result* junk = nullptr);
 

Include/armadillo/armadillo_bits/fn_dot.hpp

@@ -289,13 +289,33 @@ dot
   {
   arma_extra_debug_sigprint();
 
+  typedef typename T1::elem_type eT;
+
+  if(is_SpSubview_col<T2>::value)
+    {
+    // TODO: refactor to use C++17 "if constexpr" to avoid reinterpret_cast shenanigans
+
+    const SpSubview_col<eT>& yy = reinterpret_cast< const SpSubview_col<eT>& >(y);
+
+    if(yy.n_rows == yy.m.n_rows)
+      {
+      arma_extra_debug_print("using sparse column vector specialisation");
+
+      const quasi_unwrap<T1> U(x);
+
+      arma_debug_assert_same_size(U.M.n_elem, uword(1), yy.n_elem, uword(1), "dot()");
+
+      yy.m.sync();
+
+      return dense_sparse_helper::dot(U.M.memptr(), yy.m, yy.aux_col1);
+      }
+    }
+
   const   Proxy<T1> pa(x);
   const SpProxy<T2> pb(y);
 
   arma_debug_assert_same_size(pa.get_n_rows(), pa.get_n_cols(), pb.get_n_rows(), pb.get_n_cols(), "dot()");
 
-  typedef typename T1::elem_type eT;
-
   eT result = eT(0);
 
   typename SpProxy<T2>::const_iterator_type it     = pb.begin();

README.md

@@ -268,7 +268,7 @@ Additional libraries used in **suanPan** are listed as follows.
 - [**VTK**](https://vtk.org/) version 9.2.6
 - [**CUDA**](https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/) version 12.0
 - [**MAGMA**](https://icl.utk.edu/magma/) version 2.8.0
-- [**Armadillo**](http://arma.sourceforge.net/) version 12.8.3
+- [**Armadillo**](http://arma.sourceforge.net/) version 12.8.4
 - [**ensmallen**](https://ensmallen.org/) version 2.21.1
 - [**oneMKL**](https://software.intel.com/content/www/us/en/develop/tools/oneapi/components/onemkl.html) version 2024.1.0
 - [**Catch2**](https://github.com/catchorg/Catch2) version 3.5.4