Merge branch 'master' of https://github.com/Reference-LAPACK/lapack

Author: scr2016

.travis.yml

@@ -18,15 +18,8 @@ matrix:
   - os: linux
     env: CMAKE_BUILD_TYPE=Coverage
   - os: osx
+    osx_image: xcode10.3
     env: CMAKE_BUILD_TYPE=Release
-    before_install:
-    - brew update > /dev/null
-    - brew upgrade gcc > /dev/null
-  - os: osx
-    env: CMAKE_BUILD_TYPE=Coverage
-    before_install:
-    - brew update > /dev/null
-    - brew upgrade gcc > /dev/null
 
 script:
   - export PR=https://api.github.com/repos/$TRAVIS_REPO_SLUG/pulls/$TRAVIS_PULL_REQUEST

CMakeLists.txt

@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 2.8.12)
+cmake_minimum_required(VERSION 3.2)
 
 project(LAPACK Fortran C)
 
@@ -348,15 +348,42 @@ function(_display_cpp_implementation_msg name)
   message(STATUS "----------------")
   message(STATUS "Thank you for your interest in ${name}++, a newly developed C++ API for ${name} library")
   message(STATUS "The objective of ${name}++ is to provide a convenient, performance oriented API for development in the C++ language, that, for the most part, preserves established conventions, while, at the same time, takes advantages of modern C++ features, such as: namespaces, templates, exceptions, etc.")
-  message(STATUS "We are still working on integrating ${name}++ in our library. For the moment, you can download directly ${name_lc}++ from https://bitbucket.org/icl/${name_lc}pp")
   message(STATUS "For support ${name}++ related question, please email: slate-user@icl.utk.edu")
   message(STATUS "----------------")
 endfunction()
-if(BLAS++)
+if (BLAS++)
   _display_cpp_implementation_msg("BLAS")
+  include(ExternalProject)
+  ExternalProject_Add(blaspp
+	  URL https://bitbucket.org/icl/blaspp/downloads/blaspp-2020.10.02.tar.gz
+	  #	  CMAKE_ARGS "-DUSE_CMAKE_FIND_BLAS=ON -DLIBRARY_PATH=${CMAKE_BINARY_DIR}"
+	  CONFIGURE_COMMAND env LIBRARY_PATH=${CMAKE_BINARY_DIR}/lib LD_LIBRARY_PATH=${PROJECT_BINARY_DIR}/lib ${CMAKE_COMMAND} -DCMAKE_INSTALL_PREFIX=${PROJECT_BINARY_DIR} -DCMAKE_INSTALL_LIBDIR=lib -DBUILD_SHARED_LIBS=${BUILD_SHARED_LIBS} ${PROJECT_BINARY_DIR}/blaspp-prefix/src/blaspp
+	  BUILD_COMMAND env LIBRARY_PATH=${PROJECT_BINARY_DIR}/lib LIB_SUFFIX="" make
+	  INSTALL_COMMAND make PREFIX=${PROJECT_BINARY_DIR} LIB_SUFFIX="" install
+	  )
+  ExternalProject_Add_StepDependencies(blaspp build ${BLAS_LIBRARIES})
 endif()
-if(LAPACK++)
+if (LAPACK++)
+	message (STATUS "linking lapack++ against ${LAPACK_LIBRARIES}")
   _display_cpp_implementation_msg("LAPACK")
+  include(ExternalProject)
+  if (BUILD_SHARED_LIBS)
+  ExternalProject_Add(lapackpp
+	  URL https://bitbucket.org/icl/lapackpp/downloads/lapackpp-2020.10.02.tar.gz
+	  CONFIGURE_COMMAND env LIBRARY_PATH=${CMAKE_BINARY_DIR}/lib LD_LIBRARY_PATH=${PROJECT_BINARY_DIR}/lib ${CMAKE_COMMAND} -DCMAKE_INSTALL_PREFIX=${PROJECT_BINARY_DIR} -DCMAKE_INSTALL_LIBDIR=lib -DLAPACK_LIBRARIES=${LAPACK_LIBRARIES} -DBUILD_SHARED_LIBS=${BUILD_SHARED_LIBS}  ${PROJECT_BINARY_DIR}/lapackpp-prefix/src/lapackpp
+	  BUILD_COMMAND env LIBRARY_PATH=${PROJECT_BINARY_DIR}/lib LIB_SUFFIX="" make
+	  INSTALL_COMMAND make PREFIX=${PROJECT_BINARY_DIR} LIB_SUFFIX="" install
+	  )
+  else ()
+# FIXME this does not really work as the libraries list gets converted to a semicolon-separated list somewhere in the lapack++ build files
+  ExternalProject_Add(lapackpp
+	  URL https://bitbucket.org/icl/lapackpp/downloads/lapackpp-2020.10.02.tar.gz
+	  CONFIGURE_COMMAND env LIBRARY_PATH=${CMAKE_BINARY_DIR}/lib LD_LIBRARY_PATH=${PROJECT_BINARY_DIR}/lib ${CMAKE_COMMAND} -DCMAKE_INSTALL_PREFIX=${PROJECT_BINARY_DIR} -DCMAKE_INSTALL_LIBDIR=lib -DLAPACK_LIBRARIES="${PROJECT_BINARY_DIR}/lib/liblapack.a -lgfortran"  -DBUILD_SHARED_LIBS=${BUILD_SHARED_LIBS}  ${PROJECT_BINARY_DIR}/lapackpp-prefix/src/lapackpp
+	  BUILD_COMMAND env LIBRARY_PATH=${PROJECT_BINARY_DIR}/lib LIB_SUFFIX="" make
+	  INSTALL_COMMAND make PREFIX=${PROJECT_BINARY_DIR} LIB_SUFFIX="" install
+  )
+  endif()
+  ExternalProject_Add_StepDependencies(lapackpp build blaspp ${BLAS_LIBRARIES} ${LAPACK_LIBRARIES})
 endif()
 
 # --------------------------------------------------
@@ -464,4 +491,48 @@ install(FILES
   DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/${LAPACKLIB}-${LAPACK_VERSION}
   COMPONENT Development
   )
-  
+if (LAPACK++)
+  install(
+  DIRECTORY "${LAPACK_BINARY_DIR}/lib/"
+  DESTINATION ${CMAKE_INSTALL_LIBDIR}
+  FILES_MATCHING REGEX "liblapackpp.(a|so)$"
+  )
+  install(
+    DIRECTORY "${PROJECT_BINARY_DIR}/lapackpp-prefix/src/lapackpp/include/"
+    DESTINATION "${CMAKE_INSTALL_INCLUDEDIR}"
+    FILES_MATCHING REGEX "\\.(h|hh)$"
+  )
+  write_basic_package_version_file(
+    "lapackppConfigVersion.cmake"
+    VERSION 2020.10.02
+    COMPATIBILITY AnyNewerVersion
+  )
+  install(
+    FILES "${CMAKE_CURRENT_BINARY_DIR}/lib/lapackpp/lapackppConfig.cmake"
+          "${CMAKE_CURRENT_BINARY_DIR}/lib/lapackpp/lapackppConfigVersion.cmake"
+    DESTINATION "${CMAKE_INSTALL_LIBDIR}/cmake/"
+  )
+
+endif()
+if (BLAS++)
+  write_basic_package_version_file(
+    "blasppConfigVersion.cmake"
+    VERSION 2020.10.02
+    COMPATIBILITY AnyNewerVersion
+  )
+  install(
+    FILES "${CMAKE_CURRENT_BINARY_DIR}/lib/blaspp/blasppConfig.cmake"
+          "${CMAKE_CURRENT_BINARY_DIR}/lib/blaspp/blasppConfigVersion.cmake"
+    DESTINATION "${CMAKE_INSTALL_LIBDIR}/cmake/"
+  )
+  install(
+  DIRECTORY "${LAPACK_BINARY_DIR}/lib/"
+  DESTINATION ${CMAKE_INSTALL_LIBDIR}
+  FILES_MATCHING REGEX "libblaspp.(a|so)$"
+  )
+  install(
+    DIRECTORY "${PROJECT_BINARY_DIR}/blaspp-prefix/src/blaspp/include/"
+    DESTINATION "${CMAKE_INSTALL_INCLUDEDIR}"
+    FILES_MATCHING REGEX "\\.(h|hh)$"
+  )
+endif()

INSTALL/CMakeLists.txt

@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 2.8.7)
+cmake_minimum_required(VERSION 3.2)
 project(TIMING Fortran)
 add_executable(secondtst_NONE second_NONE.f secondtst.f)
 add_executable(secondtst_EXT_ETIME second_EXT_ETIME.f secondtst.f)

LAPACKE/mangling/CMakeLists.txt

@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 2.8.7)
+cmake_minimum_required(VERSION 3.2)
 project(MANGLING C Fortran)
 
 add_executable(xintface Fintface.f Cintface.c)

SRC/cgesdd.f

@@ -281,9 +281,9 @@ SUBROUTINE CGESDD( JOBZ, M, N, A, LDA, S, U, LDU, VT, LDVT,
      $                   CUNGQR, CUNMBR, SBDSDC, SLASCL, XERBLA
 *     ..
 *     .. External Functions ..
-      LOGICAL            LSAME
+      LOGICAL            LSAME, SISNAN
       REAL               SLAMCH, CLANGE
-      EXTERNAL           LSAME, SLAMCH, CLANGE
+      EXTERNAL           LSAME, SLAMCH, CLANGE, SISNAN
 *     ..
 *     .. Intrinsic Functions ..
       INTRINSIC          INT, MAX, MIN, SQRT
@@ -647,6 +647,10 @@ SUBROUTINE CGESDD( JOBZ, M, N, A, LDA, S, U, LDU, VT, LDVT,
 *     Scale A if max element outside range [SMLNUM,BIGNUM]
 *
       ANRM = CLANGE( 'M', M, N, A, LDA, DUM )
+      IF( SISNAN ( ANRM ) ) THEN
+          INFO = -4
+          RETURN
+      END IF
       ISCL = 0
       IF( ANRM.GT.ZERO .AND. ANRM.LT.SMLNUM ) THEN
          ISCL = 1

SRC/chgeqz.f

@@ -319,13 +319,14 @@ SUBROUTINE CHGEQZ( JOB, COMPQ, COMPZ, N, ILO, IHI, H, LDH, T, LDT,
       REAL               ABSB, ANORM, ASCALE, ATOL, BNORM, BSCALE, BTOL,
      $                   C, SAFMIN, TEMP, TEMP2, TEMPR, ULP
       COMPLEX            ABI22, AD11, AD12, AD21, AD22, CTEMP, CTEMP2,
-     $                   CTEMP3, ESHIFT, RTDISC, S, SHIFT, SIGNBC, T1,
-     $                   U12, X
+     $                   CTEMP3, ESHIFT, S, SHIFT, SIGNBC,
+     $                   U12, X, ABI12, Y
 *     ..
 *     .. External Functions ..
+      COMPLEX            CLADIV
       LOGICAL            LSAME
       REAL               CLANHS, SLAMCH
-      EXTERNAL           LSAME, CLANHS, SLAMCH
+      EXTERNAL           CLADIV, LSAME, CLANHS, SLAMCH
 *     ..
 *     .. External Subroutines ..
       EXTERNAL           CLARTG, CLASET, CROT, CSCAL, XERBLA
@@ -350,6 +351,7 @@ SUBROUTINE CHGEQZ( JOB, COMPQ, COMPZ, N, ILO, IHI, H, LDH, T, LDT,
          ILSCHR = .TRUE.
          ISCHUR = 2
       ELSE
+         ILSCHR = .TRUE.
          ISCHUR = 0
       END IF
 *
@@ -363,6 +365,7 @@ SUBROUTINE CHGEQZ( JOB, COMPQ, COMPZ, N, ILO, IHI, H, LDH, T, LDT,
          ILQ = .TRUE.
          ICOMPQ = 3
       ELSE
+         ILQ = .TRUE.
          ICOMPQ = 0
       END IF
 *
@@ -376,6 +379,7 @@ SUBROUTINE CHGEQZ( JOB, COMPQ, COMPZ, N, ILO, IHI, H, LDH, T, LDT,
          ILZ = .TRUE.
          ICOMPZ = 3
       ELSE
+         ILZ = .TRUE.
          ICOMPZ = 0
       END IF
 *
@@ -729,22 +733,34 @@ SUBROUTINE CHGEQZ( JOB, COMPQ, COMPZ, N, ILO, IHI, H, LDH, T, LDT,
             AD22 = ( ASCALE*H( ILAST, ILAST ) ) /
      $             ( BSCALE*T( ILAST, ILAST ) )
             ABI22 = AD22 - U12*AD21
+            ABI12 = AD12 - U12*AD11
 *
-            T1 = HALF*( AD11+ABI22 )
-            RTDISC = SQRT( T1**2+AD12*AD21-AD11*AD22 )
-            TEMP = REAL( T1-ABI22 )*REAL( RTDISC ) +
-     $             AIMAG( T1-ABI22 )*AIMAG( RTDISC )
-            IF( TEMP.LE.ZERO ) THEN
-               SHIFT = T1 + RTDISC
-            ELSE
-               SHIFT = T1 - RTDISC
+            SHIFT = ABI22
+            CTEMP = SQRT( ABI12 )*SQRT( AD21 )
+            TEMP = ABS1( CTEMP )
+            IF( CTEMP.NE.ZERO ) THEN
+               X = HALF*( AD11-SHIFT )
+               TEMP2 = ABS1( X )
+               TEMP = MAX( TEMP, ABS1( X ) )
+               Y = TEMP*SQRT( ( X / TEMP )**2+( CTEMP / TEMP )**2 )
+               IF( TEMP2.GT.ZERO ) THEN
+                  IF( REAL( X / TEMP2 )*REAL( Y )+
+     $                AIMAG( X / TEMP2 )*AIMAG( Y ).LT.ZERO )Y = -Y
+               END IF
+               SHIFT = SHIFT - CTEMP*CLADIV( CTEMP, ( X+Y ) )
             END IF
          ELSE
 *
 *           Exceptional shift.  Chosen for no particularly good reason.
 *
-            ESHIFT = ESHIFT + (ASCALE*H(ILAST,ILAST-1))/
-     $                        (BSCALE*T(ILAST-1,ILAST-1))
+            IF( ( IITER / 20 )*20.EQ.IITER .AND. 
+     $         BSCALE*ABS1(T( ILAST, ILAST )).GT.SAFMIN ) THEN
+               ESHIFT = ESHIFT + ( ASCALE*H( ILAST,
+     $            ILAST ) )/( BSCALE*T( ILAST, ILAST ) )
+            ELSE
+               ESHIFT = ESHIFT + ( ASCALE*H( ILAST,
+     $            ILAST-1 ) )/( BSCALE*T( ILAST-1, ILAST-1 ) )
+            END IF
             SHIFT = ESHIFT
          END IF
 *

SRC/chseqr.f

@@ -320,10 +320,10 @@ SUBROUTINE CHSEQR( JOB, COMPZ, N, ILO, IHI, H, LDH, W, Z, LDZ,
 *     .    CLAHQR because of insufficient subdiagonal scratch space.
 *     .    (This is a hard limit.) ====
       INTEGER            NTINY
-      PARAMETER          ( NTINY = 11 )
+      PARAMETER          ( NTINY = 15 )
 *
 *     ==== NL allocates some local workspace to help small matrices
-*     .    through a rare CLAHQR failure.  NL > NTINY = 11 is
+*     .    through a rare CLAHQR failure.  NL > NTINY = 15 is
 *     .    required and NL <= NMIN = ILAENV(ISPEC=12,...) is recom-
 *     .    mended.  (The default value of NMIN is 75.)  Using NL = 49
 *     .    allows up to six simultaneous shifts and a 16-by-16

SRC/clahqr.f

@@ -194,6 +194,7 @@
 *  =====================================================================
       SUBROUTINE CLAHQR( WANTT, WANTZ, N, ILO, IHI, H, LDH, W, ILOZ,
      $                   IHIZ, Z, LDZ, INFO )
+      IMPLICIT NONE
 *
 *  -- LAPACK auxiliary routine (version 3.7.0) --
 *  -- LAPACK is a software package provided by Univ. of Tennessee,    --
@@ -218,14 +219,16 @@ SUBROUTINE CLAHQR( WANTT, WANTZ, N, ILO, IHI, H, LDH, W, ILOZ,
       PARAMETER          ( RZERO = 0.0e0, RONE = 1.0e0, HALF = 0.5e0 )
       REAL               DAT1
       PARAMETER          ( DAT1 = 3.0e0 / 4.0e0 )
+      INTEGER            KEXSH
+      PARAMETER          ( KEXSH = 6 )
 *     ..
 *     .. Local Scalars ..
       COMPLEX            CDUM, H11, H11S, H22, SC, SUM, T, T1, TEMP, U,
      $                   V2, X, Y
       REAL               AA, AB, BA, BB, H10, H21, RTEMP, S, SAFMAX,
      $                   SAFMIN, SMLNUM, SX, T2, TST, ULP
       INTEGER            I, I1, I2, ITS, ITMAX, J, JHI, JLO, K, L, M,
-     $                   NH, NZ
+     $                   NH, NZ, KDEFL
 *     ..
 *     .. Local Arrays ..
       COMPLEX            V( 2 )
@@ -315,6 +318,10 @@ SUBROUTINE CLAHQR( WANTT, WANTZ, N, ILO, IHI, H, LDH, W, ILOZ,
 *
       ITMAX = 30 * MAX( 10, NH )
 *
+*     KDEFL counts the number of iterations since a deflation
+*
+      KDEFL = -2
+*
 *     The main loop begins here. I is the loop index and decreases from
 *     IHI to ILO in steps of 1. Each iteration of the loop works
 *     with the active submatrix in rows and columns L to I.
@@ -374,6 +381,7 @@ SUBROUTINE CLAHQR( WANTT, WANTZ, N, ILO, IHI, H, LDH, W, ILOZ,
 *
          IF( L.GE.I )
      $      GO TO 140
+         KDEFL = KDEFL + 1
 *
 *        Now the active submatrix is in rows and columns L to I. If
 *        eigenvalues only are being computed, only the active submatrix
@@ -384,18 +392,18 @@ SUBROUTINE CLAHQR( WANTT, WANTZ, N, ILO, IHI, H, LDH, W, ILOZ,
             I2 = I
          END IF
 *
-         IF( ITS.EQ.10 ) THEN
+         IF( MOD(KDEFL,2*KEXSH).EQ.0 ) THEN
 *
 *           Exceptional shift.
 *
-            S = DAT1*ABS( REAL( H( L+1, L ) ) )
-            T = S + H( L, L )
-         ELSE IF( ITS.EQ.20 ) THEN
+            S = DAT1*ABS( REAL( H( I, I-1 ) ) )
+            T = S + H( I, I )
+         ELSE IF( MOD(KDEFL,KEXSH).EQ.0 ) THEN
 *
 *           Exceptional shift.
 *
-            S = DAT1*ABS( REAL( H( I, I-1 ) ) )
-            T = S + H( I, I )
+            S = DAT1*ABS( REAL( H( L+1, L ) ) )
+            T = S + H( L, L )
          ELSE
 *
 *           Wilkinson's shift.
@@ -556,6 +564,8 @@ SUBROUTINE CLAHQR( WANTT, WANTZ, N, ILO, IHI, H, LDH, W, ILOZ,
 *     H(I,I-1) is negligible: one eigenvalue has converged.
 *
       W( I ) = H( I, I )
+*     reset deflation counter
+      KDEFL = 0
 *
 *     return to start of the main loop with new value of I.
 *

SRC/claqr0.f

@@ -260,7 +260,7 @@ SUBROUTINE CLAQR0( WANTT, WANTZ, N, ILO, IHI, H, LDH, W, ILOZ,
 *     .    CLAHQR because of insufficient subdiagonal scratch space.
 *     .    (This is a hard limit.) ====
       INTEGER            NTINY
-      PARAMETER          ( NTINY = 11 )
+      PARAMETER          ( NTINY = 15 )
 *
 *     ==== Exceptional deflation windows:  try to cure rare
 *     .    slow convergence by varying the size of the
@@ -355,22 +355,22 @@ SUBROUTINE CLAQR0( WANTT, WANTZ, N, ILO, IHI, H, LDH, W, ILOZ,
          END IF
 *
 *        ==== NWR = recommended deflation window size.  At this
-*        .    point,  N .GT. NTINY = 11, so there is enough
+*        .    point,  N .GT. NTINY = 15, so there is enough
 *        .    subdiagonal workspace for NWR.GE.2 as required.
 *        .    (In fact, there is enough subdiagonal space for
-*        .    NWR.GE.3.) ====
+*        .    NWR.GE.4.) ====
 *
          NWR = ILAENV( 13, 'CLAQR0', JBCMPZ, N, ILO, IHI, LWORK )
          NWR = MAX( 2, NWR )
          NWR = MIN( IHI-ILO+1, ( N-1 ) / 3, NWR )
 *
 *        ==== NSR = recommended number of simultaneous shifts.
-*        .    At this point N .GT. NTINY = 11, so there is at
+*        .    At this point N .GT. NTINY = 15, so there is at
 *        .    enough subdiagonal workspace for NSR to be even
 *        .    and greater than or equal to two as required. ====
 *
          NSR = ILAENV( 15, 'CLAQR0', JBCMPZ, N, ILO, IHI, LWORK )
-         NSR = MIN( NSR, ( N+6 ) / 9, IHI-ILO )
+         NSR = MIN( NSR, ( N-3 ) / 6, IHI-ILO )
          NSR = MAX( 2, NSR-MOD( NSR, 2 ) )
 *
 *        ==== Estimate optimal workspace ====
@@ -418,7 +418,7 @@ SUBROUTINE CLAQR0( WANTT, WANTZ, N, ILO, IHI, H, LDH, W, ILOZ,
 *        ==== NSMAX = the Largest number of simultaneous shifts
 *        .    for which there is sufficient workspace. ====
 *
-         NSMAX = MIN( ( N+6 ) / 9, 2*LWORK / 3 )
+         NSMAX = MIN( ( N-3 ) / 6, 2*LWORK / 3 )
          NSMAX = NSMAX - MOD( NSMAX, 2 )
 *
 *        ==== NDFL: an iteration count restarted at deflation. ====
@@ -659,7 +659,7 @@ SUBROUTINE CLAQR0( WANTT, WANTZ, N, ILO, IHI, H, LDH, W, ILOZ,
 *              .      (NVE-by-KDU) vertical work WV arrow along
 *              .      the left-hand-edge. ====
 *
-               KDU = 3*NS - 3
+               KDU = 2*NS
                KU = N - KDU + 1
                KWH = KDU + 1
                NHO = ( N-KDU+1-4 ) - ( KDU+1 ) + 1