Merge pull request #435 from RcppCore/feature/armadillo_12.8.1

RcppArmadillo 0.12.8.1.0

DESCRIPTION

@@ -1,8 +1,8 @@
 Package: RcppArmadillo
 Type: Package
 Title: 'Rcpp' Integration for the 'Armadillo' Templated Linear Algebra Library
-Version: 0.12.8.0.0.1
-Date: 2024-02-19
+Version: 0.12.8.1.0
+Date: 2024-03-02
 Author: Dirk Eddelbuettel, Romain Francois, Doug Bates, Binxiang Ni, and Conrad Sanderson
 Maintainer: Dirk Eddelbuettel <[email protected]>
 Description: 'Armadillo' is a templated C++ linear algebra library (by Conrad

configure

@@ -1,6 +1,6 @@
 #! /bin/sh
 # Guess values for system-dependent variables and create Makefiles.
-# Generated by GNU Autoconf 2.71 for RcppArmadillo 0.12.8.0.0.
+# Generated by GNU Autoconf 2.71 for RcppArmadillo 0.12.8.1.0.
 #
 # Report bugs to <[email protected]>.
 #
@@ -610,8 +610,8 @@ MAKEFLAGS=
 # Identity of this package.
 PACKAGE_NAME='RcppArmadillo'
 PACKAGE_TARNAME='rcpparmadillo'
-PACKAGE_VERSION='0.12.8.0.0'
-PACKAGE_STRING='RcppArmadillo 0.12.8.0.0'
+PACKAGE_VERSION='0.12.8.1.0'
+PACKAGE_STRING='RcppArmadillo 0.12.8.1.0'
 PACKAGE_BUGREPORT='[email protected]'
 PACKAGE_URL=''
 
@@ -1229,7 +1229,7 @@ if test "$ac_init_help" = "long"; then
   # Omit some internal or obsolete options to make the list less imposing.
   # This message is too long to be a string in the A/UX 3.1 sh.
   cat <<_ACEOF
-\`configure' configures RcppArmadillo 0.12.8.0.0 to adapt to many kinds of systems.
+\`configure' configures RcppArmadillo 0.12.8.1.0 to adapt to many kinds of systems.
 
 Usage: $0 [OPTION]... [VAR=VALUE]...
 
@@ -1291,7 +1291,7 @@ fi
 
 if test -n "$ac_init_help"; then
   case $ac_init_help in
-     short | recursive ) echo "Configuration of RcppArmadillo 0.12.8.0.0:";;
+     short | recursive ) echo "Configuration of RcppArmadillo 0.12.8.1.0:";;
    esac
   cat <<\_ACEOF
 
@@ -1372,7 +1372,7 @@ fi
 test -n "$ac_init_help" && exit $ac_status
 if $ac_init_version; then
   cat <<\_ACEOF
-RcppArmadillo configure 0.12.8.0.0
+RcppArmadillo configure 0.12.8.1.0
 generated by GNU Autoconf 2.71
 
 Copyright (C) 2021 Free Software Foundation, Inc.
@@ -1486,7 +1486,7 @@ cat >config.log <<_ACEOF
 This file contains any messages produced by compilers while
 running configure, to aid debugging if configure makes a mistake.
 
-It was created by RcppArmadillo $as_me 0.12.8.0.0, which was
+It was created by RcppArmadillo $as_me 0.12.8.1.0, which was
 generated by GNU Autoconf 2.71.  Invocation command line was
 
   $ $0$ac_configure_args_raw
@@ -3944,7 +3944,7 @@ cat >>$CONFIG_STATUS <<\_ACEOF || ac_write_fail=1
 # report actual input values of CONFIG_FILES etc. instead of their
 # values after options handling.
 ac_log="
-This file was extended by RcppArmadillo $as_me 0.12.8.0.0, which was
+This file was extended by RcppArmadillo $as_me 0.12.8.1.0, which was
 generated by GNU Autoconf 2.71.  Invocation command line was
 
   CONFIG_FILES    = $CONFIG_FILES
@@ -3999,7 +3999,7 @@ ac_cs_config_escaped=`printf "%s\n" "$ac_cs_config" | sed "s/^ //; s/'/'\\\\\\\\
 cat >>$CONFIG_STATUS <<_ACEOF || ac_write_fail=1
 ac_cs_config='$ac_cs_config_escaped'
 ac_cs_version="\\
-RcppArmadillo config.status 0.12.8.0.0
+RcppArmadillo config.status 0.12.8.1.0
 configured by $0, generated by GNU Autoconf 2.71,
   with options \\"\$ac_cs_config\\"
 

configure.ac

@@ -11,7 +11,7 @@
 AC_PREREQ([2.69])
 
 ## Process this file with autoconf to produce a configure script.
-AC_INIT([RcppArmadillo],[0.12.8.0.0],[[email protected]])
+AC_INIT([RcppArmadillo],[0.12.8.1.0],[[email protected]])
 
 ## Set R_HOME, respecting an environment variable if one is set
 : ${R_HOME=$(R RHOME)}

inst/NEWS.Rd

@@ -3,6 +3,19 @@
 \newcommand{\ghpr}{\href{https://github.com/RcppCore/RcppArmadillo/pull/#1}{##1}}
 \newcommand{\ghit}{\href{https://github.com/RcppCore/RcppArmadillo/issues/#1}{##1}}
 
+\section{Changes in RcppArmadillo version 0.12.8.1.0 (2024-03-02)}{
+  \itemize{
+    \item Upgraded to Armadillo release 12.8.1 (Cortisol Injector)
+    \itemize{
+      \item Workaround in \code{norm()} for yet another bug in macOS
+      accelerate framework
+    }
+    \item Update README for RcppArmadillo usage counts
+    \item Update examples to use '#include <RcppArmadillo/Lighter>' for
+    faster compilation excluding unused Rcpp features
+  }
+}
+
 \section{Changes in RcppArmadillo version 0.12.8.0.0 (2024-02-06)}{
   \itemize{
     \item Upgraded to Armadillo release 12.8.0 (Cortisol Injector)

inst/include/armadillo_bits/arma_version.hpp

@@ -23,7 +23,7 @@
 
 #define ARMA_VERSION_MAJOR 12
 #define ARMA_VERSION_MINOR 8
-#define ARMA_VERSION_PATCH 0
+#define ARMA_VERSION_PATCH 1
 #define ARMA_VERSION_NAME  "Cortisol Injector"
 
 

inst/include/armadillo_bits/compiler_setup.hpp

@@ -115,7 +115,7 @@
 
 #if defined(__APPLE__) || defined(__apple_build_version__)
   // NOTE: Apple accelerate framework has broken implementations of functions that return a float value,
-  // NOTE: such as sdot(), slange(), clange(), slansy(), clanhe(), slangb()
+  // NOTE: such as sdot(), slange(), clange(), slansy(), clanhe(), slangb(), snrm2(), sasum()
   #undef  ARMA_BLAS_FLOAT_BUG
   #define ARMA_BLAS_FLOAT_BUG
 

inst/include/armadillo_bits/op_norm_meat.hpp

@@ -174,7 +174,7 @@ op_norm::vec_norm_1(const Proxy<T1>& P, const typename arma_cx_only<typename T1:
     }
   else
     {
-    arma_extra_debug_print("op_norm::vec_norm_1(): detected possible underflow or overflow");
+    arma_extra_debug_print("detected possible underflow or overflow");
 
     const quasi_unwrap<typename Proxy<T1>::stored_type> R(P.Q);
 
@@ -224,26 +224,32 @@ op_norm::vec_norm_1_direct_std(const Mat<eT>& X)
   const uword N = X.n_elem;
   const eT*   A = X.memptr();
 
-  if(N < uword(32))
+  eT out_val = eT(0);
+
+  #if defined(ARMA_USE_ATLAS)
     {
-    return op_norm::vec_norm_1_direct_mem(N,A);
+    arma_extra_debug_print("atlas::cblas_asum()");
+    out_val = atlas::cblas_asum(N,A);
     }
-  else
+  #elif defined(ARMA_USE_BLAS)
     {
-    #if defined(ARMA_USE_ATLAS)
-      {
-      return atlas::cblas_asum(N,A);
-      }
-    #elif defined(ARMA_USE_BLAS)
+    if(has_blas_float_bug<eT>::value)
       {
-      return blas::asum(N,A);
+      out_val = op_norm::vec_norm_1_direct_mem(N,A);
       }
-    #else
+    else
       {
-      return op_norm::vec_norm_1_direct_mem(N,A);
+      arma_extra_debug_print("blas::asum()");
+      out_val = blas::asum(N,A);
       }
-    #endif
     }
+  #else
+    {
+    out_val = op_norm::vec_norm_1_direct_mem(N,A);
+    }
+  #endif
+
+  return (out_val <= eT(0)) ? eT(0) : out_val;
   }
 
 
@@ -397,7 +403,7 @@ op_norm::vec_norm_2(const Proxy<T1>& P, const typename arma_not_cx<typename T1::
     }
   else
     {
-    arma_extra_debug_print("op_norm::vec_norm_2(): detected possible underflow or overflow");
+    arma_extra_debug_print("detected possible underflow or overflow");
 
     const quasi_unwrap<typename Proxy<T1>::stored_type> tmp(P.Q);
 
@@ -476,7 +482,7 @@ op_norm::vec_norm_2(const Proxy<T1>& P, const typename arma_cx_only<typename T1:
     }
   else
     {
-    arma_extra_debug_print("op_norm::vec_norm_2(): detected possible underflow or overflow");
+    arma_extra_debug_print("detected possible underflow or overflow");
 
     const quasi_unwrap<typename Proxy<T1>::stored_type> R(P.Q);
 
@@ -519,36 +525,38 @@ op_norm::vec_norm_2_direct_std(const Mat<eT>& X)
   const uword N = X.n_elem;
   const eT*   A = X.memptr();
 
-  eT result;
+  eT out_val = eT(0);
 
-  if(N < uword(32))
+  #if defined(ARMA_USE_ATLAS)
     {
-    result = op_norm::vec_norm_2_direct_mem(N,A);
+    arma_extra_debug_print("atlas::cblas_nrm2()");
+    out_val = atlas::cblas_nrm2(N,A);
     }
-  else
+  #elif defined(ARMA_USE_BLAS)
     {
-    #if defined(ARMA_USE_ATLAS)
+    if(has_blas_float_bug<eT>::value)
       {
-      result = atlas::cblas_nrm2(N,A);
+      out_val = op_norm::vec_norm_2_direct_mem(N,A);
       }
-    #elif defined(ARMA_USE_BLAS)
-      {
-      result = blas::nrm2(N,A);
-      }
-    #else
+    else
       {
-      result = op_norm::vec_norm_2_direct_mem(N,A);
+      arma_extra_debug_print("blas::nrm2()");
+      out_val = blas::nrm2(N,A);
       }
-    #endif
     }
+  #else
+    {
+    out_val = op_norm::vec_norm_2_direct_mem(N,A);
+    }
+  #endif
 
-  if( (result != eT(0)) && arma_isfinite(result) )
+  if( (out_val != eT(0)) && arma_isfinite(out_val) )
     {
-    return result;
+    return (out_val < eT(0)) ? eT(0) : out_val;
     }
   else
     {
-    arma_extra_debug_print("op_norm::vec_norm_2_direct_std(): detected possible underflow or overflow");
+    arma_extra_debug_print("detected possible underflow or overflow");
 
     return op_norm::vec_norm_2_direct_robust(X);
     }
@@ -563,7 +571,7 @@ op_norm::vec_norm_2_direct_mem(const uword N, const eT* A)
   {
   arma_extra_debug_sigprint();
 
-  eT acc;
+  eT acc = eT(0);
 
   #if (defined(ARMA_SIMPLE_LOOPS) || defined(__FAST_MATH__))
     {
@@ -673,7 +681,9 @@ op_norm::vec_norm_2_direct_robust(const Mat<eT>& X)
     acc1 += val_i * val_i;
     }
 
-  return ( std::sqrt(acc1 + acc2) * max_val ); 
+  const eT out_val = std::sqrt(acc1 + acc2) * max_val;
+
+  return (out_val <= eT(0)) ? eT(0) : out_val;
   }
 
 
@@ -882,9 +892,12 @@ op_norm::mat_norm_2(const Mat<eT>& X)
   if(X.internal_has_nonfinite())  { arma_debug_warn_level(1, "norm(): given matrix has non-finite elements"); }
 
   Col<T> S;
+
   svd(S, X);
 
-  return (S.n_elem > 0) ? S[0] : T(0);
+  const T out_val = (S.n_elem > 0) ? S[0] : T(0);
+
+  return (out_val <= T(0)) ? T(0) : out_val;
   }
 
 

inst/include/armadillo_bits/spop_norm_meat.hpp

@@ -53,9 +53,12 @@ spop_norm::mat_norm_2(const SpMat<eT>& X, const typename arma_real_only<eT>::res
   const SpMat<eT>  C = (A.n_rows <= A.n_cols) ? (A*B) : (B*A);
 
   Col<T> eigval;
+
   eigs_sym(eigval, C, 1);
 
-  return (eigval.n_elem > 0) ? T(std::sqrt(eigval[0])) : T(0);
+  const T out_square_val = (eigval.n_elem > 0) ? T(eigval[0]) : T(0);
+
+  return (out_square_val <= T(0)) ? T(0) : T(std::sqrt(out_square_val));
   }
 
 
@@ -84,9 +87,12 @@ spop_norm::mat_norm_2(const SpMat<eT>& X, const typename arma_cx_only<eT>::resul
   const SpMat<eT>  C = (A.n_rows <= A.n_cols) ? (A*B) : (B*A);
 
   Col<eT> eigval;
+
   eigs_gen(eigval, C, 1);
 
-  return (eigval.n_elem > 0) ? T(std::sqrt(std::real(eigval[0]))) : T(0);
+  const T out_square_val = (eigval.n_elem > 0) ? T(std::real(eigval[0])) : T(0);
+
+  return (out_square_val <= T(0)) ? T(0) : T(std::sqrt(out_square_val));
   }
 
 

inst/include/armadillo_bits/translate_blas.hpp

@@ -145,6 +145,9 @@ namespace blas
 
         eT result[2];  // paranoia: using two elements instead of one
 
+        result[0] = eT(0);
+        result[1] = eT(0);
+
         blas::gemv(&trans, &m, &n, &alpha, x, &m, y, &inc, &beta, &result[0], &inc);
 
         return result[0];
@@ -186,11 +189,14 @@ namespace blas
 
       eT result[2];  // paranoia: using two elements instead of one
 
+      result[0] = eT(0);
+      result[1] = eT(0);
+
       blas::gemv(&trans, &m, &n, &alpha, x, &m, y, &inc, &beta, &result[0], &inc);
 
       return result[0];
       }
-
+    
     return eT(0);
     }
 
@@ -205,6 +211,8 @@ namespace blas
 
     if(is_float<eT>::value)
       {
+      // WARNING: sasum() from Accelerate framework (macOS) may return 'double' instead of 'float'
+
       blas_int n   = blas_int(n_elem);
       blas_int inc = 1;
 
@@ -235,6 +243,8 @@ namespace blas
 
     if(is_float<eT>::value)
       {
+      // WARNING: snrm2() from Accelerate framework (macOS) may return 'double' instead of 'float'
+
       blas_int n   = blas_int(n_elem);
       blas_int inc = 1;