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

BLAS/SRC/crotg.f90

@@ -11,7 +11,7 @@
 !  CROTG constructs a plane rotation
 !     [  c         s ] [ a ] = [ r ]
 !     [ -conjg(s)  c ] [ b ]   [ 0 ]
-!  where c is real, s ic complex, and c**2 + conjg(s)*s = 1.
+!  where c is real, s is complex, and c**2 + conjg(s)*s = 1.
 !
 !> \par Purpose:
 !  =============
@@ -28,8 +28,8 @@
 !>    r = sgn(a)*sqrt(|a|**2 + |b|**2)
 !>    c = a / r
 !>    s = b / r
-!> the same as in CROTG when |a| > |b|.  When |b| >= |a|, the
-!> sign of c and s will be different from those computed by CROTG
+!> the same as in SROTG when |a| > |b|.  When |b| >= |a|, the
+!> sign of c and s will be different from those computed by SROTG
 !> if the signs of a and b are not the same.
 !>
 !> \endverbatim

BLAS/SRC/zrotg.f90

@@ -11,7 +11,7 @@
 !  ZROTG constructs a plane rotation
 !     [  c         s ] [ a ] = [ r ]
 !     [ -conjg(s)  c ] [ b ]   [ 0 ]
-!  where c is real, s ic complex, and c**2 + conjg(s)*s = 1.
+!  where c is real, s is complex, and c**2 + conjg(s)*s = 1.
 !
 !> \par Purpose:
 !  =============
@@ -28,8 +28,8 @@
 !>    r = sgn(a)*sqrt(|a|**2 + |b|**2)
 !>    c = a / r
 !>    s = b / r
-!> the same as in ZROTG when |a| > |b|.  When |b| >= |a|, the
-!> sign of c and s will be different from those computed by ZROTG
+!> the same as in DROTG when |a| > |b|.  When |b| >= |a|, the
+!> sign of c and s will be different from those computed by DROTG
 !> if the signs of a and b are not the same.
 !>
 !> \endverbatim

CBLAS/CMakeLists.txt

@@ -11,9 +11,7 @@ FortranCInterface_HEADER(${LAPACK_BINARY_DIR}/include/cblas_mangling.h
                          MACRO_NAMESPACE "F77_"
                          SYMBOL_NAMESPACE "F77_")
 if(NOT FortranCInterface_GLOBAL_FOUND OR NOT FortranCInterface_MODULE_FOUND)
-  message(WARNING "Reverting to pre-defined include/lapacke_mangling.h")
-    configure_file(include/lapacke_mangling_with_flags.h.in
-                  ${LAPACK_BINARY_DIR}/include/lapacke_mangling.h)
+  message(WARNING "Reverting to pre-defined include/cblas_mangling.h")
     configure_file(include/cblas_mangling_with_flags.h.in
                  ${LAPACK_BINARY_DIR}/include/cblas_mangling.h)
 endif()

CBLAS/include/cblas.h

@@ -1,6 +1,7 @@
 #ifndef CBLAS_H
 #define CBLAS_H
 #include <stddef.h>
+#include <stdint.h>
 
 
 #ifdef __cplusplus
@@ -15,10 +16,12 @@ extern "C" {            /* Assume C declarations for C++ */
 /*
  * Integer type
  */
+#ifndef CBLAS_INT
 #ifdef WeirdNEC
-   #define CBLAS_INT long
+   #define CBLAS_INT int64_t
 #else
-   #define CBLAS_INT int
+   #define CBLAS_INT int32_t
+#endif
 #endif
 
 typedef enum CBLAS_LAYOUT {CblasRowMajor=101, CblasColMajor=102} CBLAS_LAYOUT;

CBLAS/include/cblas_f77.h

@@ -10,6 +10,7 @@
 #define CBLAS_F77_H
 
 #include <stdarg.h>
+#include <stdint.h>
 
 /* It seems all current Fortran compilers put strlen at end.
 *  Some historical compilers put strlen after the str argument
@@ -24,10 +25,12 @@
    #define F77_STRLEN(a) (_fcdlen)
 #endif
 
+#ifndef F77_INT
 #ifdef WeirdNEC
-   #define F77_INT long
+   #define F77_INT int64_t
 #else
-   #define F77_INT int
+   #define F77_INT int32_t
+#endif
 #endif
 
 #ifdef  F77_CHAR
@@ -527,7 +530,11 @@
 extern "C" {
 #endif
 
-void F77_xerbla(FCHAR, void *);
+#ifdef BLAS_FORTRAN_STRLEN_END
+  #define F77_xerbla(...) F77_xerbla_base(__VA_ARGS__, 1)
+#else
+  #define F77_xerbla(...) F77_xerbla_base(__VA_ARGS__)
+#endif
 void F77_xerbla_base(FCHAR, void *
    #ifdef BLAS_FORTRAN_STRLEN_END
       , size_t

CBLAS/src/xerbla.c

@@ -10,13 +10,16 @@ void
 #ifdef HAS_ATTRIBUTE_WEAK_SUPPORT
 __attribute__((weak))
 #endif
-F77_xerbla
+F77_xerbla_base
 #ifdef F77_CHAR
-(F77_CHAR F77_srname, void *vinfo)
+(F77_CHAR F77_srname, void *vinfo
 #else
-(char *srname, void *vinfo)
+(char *srname, void *vinfo
 #endif
-
+#ifdef BLAS_FORTRAN_STRLEN_END
+, size_t len
+#endif
+)
 {
 #ifdef F77_CHAR
    char *srname;

CMAKE/CheckLAPACKCompilerFlags.cmake

@@ -14,6 +14,26 @@ macro( CheckLAPACKCompilerFlags )
 
 set( FPE_EXIT FALSE )
 
+# FORTRAN ILP default
+if ( FORTRAN_ILP )
+    if( CMAKE_Fortran_COMPILER_ID STREQUAL "Intel" )
+        if ( WIN32 )
+            set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} /integer-size:64")
+        else ()
+            set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -integer-size 64")
+        endif()
+    else()
+        set(CPE_ENV $ENV{PE_ENV})
+        if(CPE_ENV STREQUAL "CRAY")
+          set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -sinteger64")
+        elseif(CPE_ENV STREQUAL "NVIDIA")
+          set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -i8")
+        else()
+          set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -fdefault-integer-8")  
+        endif()        
+    endif()
+endif()
+
 # GNU Fortran
 if( CMAKE_Fortran_COMPILER_ID STREQUAL "GNU" )
   if( "${CMAKE_Fortran_FLAGS}" MATCHES "-ffpe-trap=[izoupd]")

CMakeLists.txt

@@ -53,14 +53,7 @@ if(BUILD_INDEX64)
   set(LAPACKELIB "lapacke64")
   set(TMGLIB "tmglib64")
   set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -DWeirdNEC -DLAPACK_ILP64 -DHAVE_LAPACK_CONFIG_H")
-  set(CPE_ENV $ENV{PE_ENV})
-  if(CPE_ENV STREQUAL "CRAY")
-    set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -sinteger64")
-  elseif(CPE_ENV STREQUAL "NVIDIA")
-    set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -i8")
-  else()
-    set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -fdefault-integer-8")  
-  endif()
+  set(FORTRAN_ILP TRUE)
 else()
   set(BLASLIB "blas")
   set(CBLASLIB "cblas")
@@ -98,28 +91,44 @@ include(PreventInBuildInstalls)
 
 # Check if recursive flag exists
 include(CheckFortranCompilerFlag)
-check_fortran_compiler_flag("-recursive" _recursiveFlag)
-check_fortran_compiler_flag("-frecursive" _frecursiveFlag)
-check_fortran_compiler_flag("-Mrecursive" _MrecursiveFlag)
+if(CMAKE_Fortran_COMPILER_ID STREQUAL Flang)
+  check_fortran_compiler_flag("-Mrecursive" _MrecursiveFlag)
+elseif(CMAKE_Fortran_COMPILER_ID STREQUAL GNU)
+  check_fortran_compiler_flag("-frecursive" _frecursiveFlag)
+elseif(CMAKE_Fortran_COMPILER_ID STREQUAL Intel)
+  check_fortran_compiler_flag("-recursive" _recursiveFlag)
+elseif(CMAKE_Fortran_COMPILER_ID STREQUAL XL)
+  check_fortran_compiler_flag("-qrecur" _qrecurFlag)
+else()
+  message(WARNING "Fortran local arrays should be allocated on the stack."
+    " Please use a compiler which guarantees that feature."
+    " See https://github.com/Reference-LAPACK/lapack/pull/188 and references therein.")
+endif()
 
 # Add recursive flag
-if(_recursiveFlag)
-  string(REGEX MATCH "-recursive" output_test <string> "${CMAKE_Fortran_FLAGS}")
+if(_MrecursiveFlag)
+  string(REGEX MATCH "-Mrecursive" output_test <string> "${CMAKE_Fortran_FLAGS}")
   if(NOT output_test)
-    set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -recursive"
+    set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -Mrecursive"
       CACHE STRING "Recursive flag must be set" FORCE)
   endif()
 elseif(_frecursiveFlag)
   string(REGEX MATCH "-frecursive" output_test <string> "${CMAKE_Fortran_FLAGS}")
   if(NOT output_test)
     set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -frecursive"
-    CACHE STRING "Recursive flag must be set" FORCE)
+      CACHE STRING "Recursive flag must be set" FORCE)
   endif()
-elseif(_MrecursiveFlag)
-  string(REGEX MATCH "-Mrecursive" output_test <string> "${CMAKE_Fortran_FLAGS}")
+elseif(_recursiveFlag)
+  string(REGEX MATCH "-recursive" output_test <string> "${CMAKE_Fortran_FLAGS}")
   if(NOT output_test)
-    set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -Mrecursive"
-    CACHE STRING "Recursive flag must be set" FORCE)
+    set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -recursive"
+      CACHE STRING "Recursive flag must be set" FORCE)
+  endif()
+elseif(_qrecurFlag)
+  string(REGEX MATCH "-qrecur" output_test <string> "${CMAKE_Fortran_FLAGS}")
+  if(NOT output_test)
+    set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -qrecur"
+      CACHE STRING "Recursive flag must be set" FORCE)
   endif()
 endif()
 
@@ -128,7 +137,7 @@ if(UNIX)
     set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -fp-model strict")
   endif()
   if(CMAKE_Fortran_COMPILER_ID STREQUAL XL)
-    set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -qnosave -qstrict=none")
+    set(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} -qnosave -qstrict")
   endif()
 # Delete libmtsk in linking sequence for Sun/Oracle Fortran Compiler.
 # This library is not present in the Sun package SolarisStudio12.3-linux-x86-bin

INSTALL/make.inc.XLF

@@ -14,10 +14,10 @@ CFLAGS = -O3 -qnosave
 #  the compiler options desired when NO OPTIMIZATION is selected.
 #
 FC = xlf
-FFLAGS = -O3 -qfixed -qnosave
+FFLAGS = -O3 -qfixed -qnosave -qrecur
 # For -O2, add -qstrict=none
 FFLAGS_DRV = $(FFLAGS)
-FFLAGS_NOOPT = -O0 -qfixed -qnosave
+FFLAGS_NOOPT = -O0 -qfixed -qnosave -qrecur
 
 #  Define LDFLAGS to the desired linker options for your machine.
 #

INSTALL/make.inc.gfortran

@@ -79,3 +79,8 @@ CBLASLIB     = $(TOPSRCDIR)/libcblas.a
 LAPACKLIB    = $(TOPSRCDIR)/liblapack.a
 TMGLIB       = $(TOPSRCDIR)/libtmglib.a
 LAPACKELIB   = $(TOPSRCDIR)/liblapacke.a
+
+#  DOCUMENTATION DIRECTORY
+# If you generate html pages (make html), documentation will be placed in $(DOCSDIR)/explore-html
+# If you generate man pages (make man), documentation will be placed in $(DOCSDIR)/man
+DOCSDIR       = $(TOPSRCDIR)/DOCS

LAPACKE/include/lapacke_config.h

@@ -41,12 +41,13 @@ extern "C" {
 #endif /* __cplusplus */
 
 #include <stdlib.h>
+#include <stdint.h>
 
 #ifndef lapack_int
 #if defined(LAPACK_ILP64)
-#define lapack_int              long
+#define lapack_int              int64_t
 #else
-#define lapack_int              int
+#define lapack_int              int32_t
 #endif
 #endif
 

SRC/cgeqrf.f

@@ -95,7 +95,8 @@
 *> \param[in] LWORK
 *> \verbatim
 *>          LWORK is INTEGER
-*>          The dimension of the array WORK.  LWORK >= max(1,N).
+*>          The dimension of the array WORK.
+*>          LWORK >= 1, if MIN(M,N) = 0, and LWORK >= N, otherwise.
 *>          For optimum performance LWORK >= N*NB, where NB is
 *>          the optimal blocksize.
 *>
@@ -175,30 +176,35 @@ SUBROUTINE CGEQRF( M, N, A, LDA, TAU, WORK, LWORK, INFO )
 *
 *     Test the input arguments
 *
+      K = MIN( M, N )
       INFO = 0
       NB = ILAENV( 1, 'CGEQRF', ' ', M, N, -1, -1 )
-      LWKOPT = N*NB
-      WORK( 1 ) = LWKOPT
       LQUERY = ( LWORK.EQ.-1 )
       IF( M.LT.0 ) THEN
          INFO = -1
       ELSE IF( N.LT.0 ) THEN
          INFO = -2
       ELSE IF( LDA.LT.MAX( 1, M ) ) THEN
          INFO = -4
-      ELSE IF( LWORK.LT.MAX( 1, N ) .AND. .NOT.LQUERY ) THEN
-         INFO = -7
+      ELSE IF( .NOT.LQUERY ) THEN
+         IF( LWORK.LE.0 .OR. ( M.GT.0 .AND. LWORK.LT.MAX( 1, N ) ) )
+     $      INFO = -7
       END IF
       IF( INFO.NE.0 ) THEN
          CALL XERBLA( 'CGEQRF', -INFO )
          RETURN
       ELSE IF( LQUERY ) THEN
+         IF( K.EQ.0 ) THEN
+            LWKOPT = 1
+         ELSE
+            LWKOPT = N*NB
+         END IF
+         WORK( 1 ) = LWKOPT
          RETURN
       END IF
 *
 *     Quick return if possible
 *
-      K = MIN( M, N )
       IF( K.EQ.0 ) THEN
          WORK( 1 ) = 1
          RETURN

SRC/cgerqf.f

@@ -88,7 +88,8 @@
 *> \param[in] LWORK
 *> \verbatim
 *>          LWORK is INTEGER
-*>          The dimension of the array WORK.  LWORK >= max(1,M).
+*>          The dimension of the array WORK.
+*>          LWORK >= 1, if MIN(M,N) = 0, and LWORK >= M, otherwise.
 *>          For optimum performance LWORK >= M*NB, where NB is
 *>          the optimal blocksize.
 *>
@@ -188,8 +189,9 @@ SUBROUTINE CGERQF( M, N, A, LDA, TAU, WORK, LWORK, INFO )
          END IF
          WORK( 1 ) = LWKOPT
 *
-         IF( LWORK.LT.MAX( 1, M ) .AND. .NOT.LQUERY ) THEN
-            INFO = -7
+         IF ( .NOT.LQUERY ) THEN
+            IF( LWORK.LE.0 .OR. ( N.GT.0 .AND. LWORK.LT.MAX( 1, M ) ) )
+     $         INFO = -7
          END IF
       END IF
 *

SRC/cla_gbrfsx_extended.f

@@ -54,7 +54,7 @@
 *> In addition to normwise error bound, the code provides maximum
 *> componentwise error bound if possible. See comments for ERR_BNDS_NORM
 *> and ERR_BNDS_COMP for details of the error bounds. Note that this
-*> subroutine is only resonsible for setting the second fields of
+*> subroutine is only responsible for setting the second fields of
 *> ERR_BNDS_NORM and ERR_BNDS_COMP.
 *> \endverbatim
 *

SRC/cla_gerfsx_extended.f

@@ -54,7 +54,7 @@
 *> In addition to normwise error bound, the code provides maximum
 *> componentwise error bound if possible. See comments for ERRS_N
 *> and ERRS_C for details of the error bounds. Note that this
-*> subroutine is only resonsible for setting the second fields of
+*> subroutine is only responsible for setting the second fields of
 *> ERRS_N and ERRS_C.
 *> \endverbatim
 *

SRC/cla_herfsx_extended.f

@@ -55,7 +55,7 @@
 *> In addition to normwise error bound, the code provides maximum
 *> componentwise error bound if possible. See comments for ERR_BNDS_NORM
 *> and ERR_BNDS_COMP for details of the error bounds. Note that this
-*> subroutine is only resonsible for setting the second fields of
+*> subroutine is only responsible for setting the second fields of
 *> ERR_BNDS_NORM and ERR_BNDS_COMP.
 *> \endverbatim
 *

SRC/cla_porfsx_extended.f

@@ -54,7 +54,7 @@
 *> In addition to normwise error bound, the code provides maximum
 *> componentwise error bound if possible. See comments for ERR_BNDS_NORM
 *> and ERR_BNDS_COMP for details of the error bounds. Note that this
-*> subroutine is only resonsible for setting the second fields of
+*> subroutine is only responsible for setting the second fields of
 *> ERR_BNDS_NORM and ERR_BNDS_COMP.
 *> \endverbatim
 *

SRC/cla_syrfsx_extended.f

@@ -55,7 +55,7 @@
 *> In addition to normwise error bound, the code provides maximum
 *> componentwise error bound if possible. See comments for ERR_BNDS_NORM
 *> and ERR_BNDS_COMP for details of the error bounds. Note that this
-*> subroutine is only resonsible for setting the second fields of
+*> subroutine is only responsible for setting the second fields of
 *> ERR_BNDS_NORM and ERR_BNDS_COMP.
 *> \endverbatim
 *