Index-64 extended API overhaul

.github/workflows/cmake.yml

@@ -48,7 +48,7 @@ jobs:
 
   test-install-release:
     # Use GNU compilers
-  
+
     # The CMake configure and build commands are platform agnostic and should work equally
     # well on Windows or Mac.  You can convert this to a matrix build if you need
     # cross-platform coverage.
@@ -66,9 +66,9 @@ jobs:
         fflags: [
           "-Wall -Wno-unused-dummy-argument -Wno-unused-variable -Wno-unused-label -Werror=conversion -fimplicit-none -frecursive -fcheck=all",
           "-Wall -Wno-unused-dummy-argument -Wno-unused-variable -Wno-unused-label -Werror=conversion -fimplicit-none -frecursive -fcheck=all -fopenmp" ]
-    
+
     steps:
-    
+
     - name: Checkout LAPACK
       uses: actions/checkout@8e5e7e5ab8b370d6c329ec480221332ada57f0ab # v3.5.2
 
@@ -103,7 +103,7 @@ jobs:
         -D BUILD_SHARED_LIBS:BOOL=ON
 
     - name: Build
-      # Execute tests defined by the CMake configuration.  
+      # Execute tests defined by the CMake configuration.
       # See https://cmake.org/cmake/help/latest/manual/ctest.1.html for more detail
       run: cmake --build build --config ${{env.BUILD_TYPE}}
 
@@ -120,13 +120,52 @@ jobs:
     - name: Install
       run: cmake --build build --target install -j2
 
+  test-extended-api-only:
+    runs-on: ubuntu-latest
+
+    env:
+      BUILD_TYPE: Release
+      FFLAGS: "-Wall -Wno-unused-dummy-argument -Wno-unused-variable -Wno-unused-label -Werror=conversion -fimplicit-none -frecursive -fcheck=all"
+
+    strategy:
+      fail-fast: true
+      matrix:
+        shared_libs: [ OFF, ON ]
+
+    steps:
+
+    - name: Checkout LAPACK
+      uses: actions/checkout@8e5e7e5ab8b370d6c329ec480221332ada57f0ab # v3.5.2
+
+    - name: Install ninja-build tool
+      uses: seanmiddleditch/gha-setup-ninja@16b940825621068d98711680b6c3ff92201f8fc0 # v3
+
+    - name: Configure CMake
+      run: >
+        cmake -B build -G Ninja
+        -D CMAKE_BUILD_TYPE=${{env.BUILD_TYPE}}
+        -D BUILD_SHARED_LIBS:BOOL=${{matrix.shared_libs}}
+        -D BUILD_DEFAULT_API:BOOL=OFF
+        -D BUILD_INDEX64_EXT_API:BOOL=ON
+        -D BUILD_TESTING:BOOL=ON
+        -D CBLAS:BOOL=ON
+        -D LAPACKE:BOOL=ON
+        -D LAPACKE_WITH_TMG:BOOL=ON
+
+    - name: Build
+      run: cmake --build build --config ${{env.BUILD_TYPE}}
+
+    - name: Test
+      working-directory: ${{github.workspace}}/build
+      run: ctest -C ${{env.BUILD_TYPE}} --schedule-random -j2 --output-on-failure --timeout 100
+
   coverage:
-    runs-on: ubuntu-latest 
+    runs-on: ubuntu-latest
     env:
       BUILD_TYPE: Coverage
       FFLAGS: "-fopenmp"
-    steps:     
-    
+    steps:
+
     - name: Checkout LAPACK
       uses: actions/checkout@8e5e7e5ab8b370d6c329ec480221332ada57f0ab # v3.5.2
 
@@ -198,7 +237,7 @@ jobs:
 
     - name: Install ninja-build tool
       uses: seanmiddleditch/gha-setup-ninja@16b940825621068d98711680b6c3ff92201f8fc0 # v3
-    
+
     - name: Install APT packages
       run: |
         sudo apt update

BLAS/SRC/CMakeLists.txt

@@ -31,31 +31,31 @@
 
 set(LAPACK_INSTALL_EXPORT_NAME ${BLASLIB}-targets)
 
-set(SBLAS1 isamax.f sasum.f saxpy.f saxpby.f scopy.f sdot.f snrm2.f90
-	srot.f srotg.f90 sscal.f sswap.f sdsdot.f srotmg.f srotm.f)
+set(SBLAS1
+  isamax.f sasum.f saxpy.f saxpby.f scopy.f sdot.f snrm2.f90 srot.f srotg.f90
+  sscal.f sswap.f sdsdot.f srotmg.f srotm.f)
 
-set(CBLAS1 scabs1.f scasum.f scnrm2.f90 icamax.f90 caxpy.f caxpby.f ccopy.f
-	cdotc.f cdotu.f csscal.f crotg.f90 cscal.f cswap.f csrot.f)
+set(CBLAS1
+  scabs1.f scasum.f scnrm2.f90 icamax.f90 caxpy.f caxpby.f ccopy.f cdotc.f
+  cdotu.f csscal.f crotg.f90 cscal.f cswap.f csrot.f)
 
-set(DBLAS1 idamax.f dasum.f daxpy.f daxpby.f dcopy.f ddot.f dnrm2.f90
-	drot.f drotg.f90 dscal.f dsdot.f dswap.f drotmg.f drotm.f)
+set(DBLAS1
+  idamax.f dasum.f daxpy.f daxpby.f dcopy.f ddot.f dnrm2.f90 drot.f drotg.f90
+  dscal.f dsdot.f dswap.f drotmg.f drotm.f)
 
 set(DB1AUX sscal.f isamax.f)
 
-set(ZBLAS1 dcabs1.f dzasum.f dznrm2.f90 izamax.f90 zaxpy.f zaxpby.f zcopy.f
-	zdotc.f zdotu.f zdscal.f zrotg.f90 zscal.f zswap.f zdrot.f)
+set(ZBLAS1
+  dcabs1.f dzasum.f dznrm2.f90 izamax.f90 zaxpy.f zaxpby.f zcopy.f zdotc.f
+  zdotu.f zdscal.f zrotg.f90 zscal.f zswap.f zdrot.f)
 
 set(CB1AUX
-    isamax.f idamax.f
-    sasum.f saxpy.f scopy.f sdot.f sgemm.f sgemv.f snrm2.f90 srot.f sscal.f
-    sswap.f)
+  isamax.f idamax.f sasum.f saxpy.f scopy.f sdot.f sgemm.f sgemv.f snrm2.f90
+  srot.f sscal.f sswap.f)
 
 set(ZB1AUX
-    icamax.f90 idamax.f
-    cgemm.f cherk.f cscal.f ctrsm.f
-    dasum.f daxpy.f dcopy.f ddot.f dgemm.f dgemv.f dnrm2.f90 drot.f dscal.f
-    dswap.f
-    scabs1.f)
+  icamax.f90 idamax.f cgemm.f cherk.f cscal.f ctrsm.f dasum.f daxpy.f dcopy.f
+  ddot.f dgemm.f dgemv.f dnrm2.f90 drot.f dscal.f dswap.f scabs1.f)
 
 #---------------------------------------------------------------------
 #  Auxiliary routines needed by both the Level 2 and Level 3 BLAS
@@ -65,38 +65,40 @@ set(ALLBLAS lsame.f xerbla.f xerbla_array.f)
 #---------------------------------------------------------
 #  Level 2 BLAS
 #---------------------------------------------------------
-set(SBLAS2 sgemv.f sgbmv.f ssymv.f ssbmv.f sspmv.f
-	strmv.f stbmv.f stpmv.f strsv.f stbsv.f stpsv.f
-	sger.f ssyr.f sspr.f ssyr2.f sspr2.f
-	sskewsymv.f sskewsyr2.f)
+set(SBLAS2
+  sgemv.f sgbmv.f ssymv.f ssbmv.f sspmv.f strmv.f stbmv.f stpmv.f strsv.f
+  stbsv.f stpsv.f sger.f ssyr.f sspr.f ssyr2.f sspr2.f sskewsymv.f sskewsyr2.f)
 
-set(CBLAS2 cgemv.f cgbmv.f chemv.f chbmv.f chpmv.f
-	ctrmv.f ctbmv.f ctpmv.f ctrsv.f ctbsv.f ctpsv.f
-	cgerc.f cgeru.f cher.f chpr.f cher2.f chpr2.f)
+set(CBLAS2
+  cgemv.f cgbmv.f chemv.f chbmv.f chpmv.f ctrmv.f ctbmv.f ctpmv.f ctrsv.f
+  ctbsv.f ctpsv.f cgerc.f cgeru.f cher.f chpr.f cher2.f chpr2.f)
 
-set(DBLAS2 dgemv.f dgbmv.f dsymv.f dsbmv.f dspmv.f
-	dtrmv.f dtbmv.f dtpmv.f dtrsv.f dtbsv.f dtpsv.f
-	dger.f dsyr.f dspr.f dsyr2.f dspr2.f
-	dskewsymv.f dskewsyr2.f)
+set(DBLAS2
+  dgemv.f dgbmv.f dsymv.f dsbmv.f dspmv.f dtrmv.f dtbmv.f dtpmv.f dtrsv.f
+  dtbsv.f dtpsv.f dger.f dsyr.f dspr.f dsyr2.f dspr2.f dskewsymv.f dskewsyr2.f)
 
-set(ZBLAS2 zgemv.f zgbmv.f zhemv.f zhbmv.f zhpmv.f
-	ztrmv.f ztbmv.f ztpmv.f ztrsv.f ztbsv.f ztpsv.f
-	zgerc.f zgeru.f zher.f zhpr.f zher2.f zhpr2.f)
+set(ZBLAS2
+  zgemv.f zgbmv.f zhemv.f zhbmv.f zhpmv.f ztrmv.f ztbmv.f ztpmv.f ztrsv.f
+  ztbsv.f ztpsv.f zgerc.f zgeru.f zher.f zhpr.f zher2.f zhpr2.f)
 
 #---------------------------------------------------------
 #  Level 3 BLAS
 #---------------------------------------------------------
-set(SBLAS3 sgemm.f ssymm.f ssyrk.f ssyr2k.f strmm.f strsm.f sgemmtr.f
-	sskewsymm.f sskewsyr2k.f)
+set(SBLAS3
+  sgemm.f ssymm.f ssyrk.f ssyr2k.f strmm.f strsm.f sgemmtr.f sskewsymm.f
+  sskewsyr2k.f)
 
-set(CBLAS3 cgemm.f csymm.f csyrk.f csyr2k.f ctrmm.f ctrsm.f
-	chemm.f cherk.f cher2k.f cgemmtr.f)
+set(CBLAS3
+  cgemm.f csymm.f csyrk.f csyr2k.f ctrmm.f ctrsm.f chemm.f cherk.f cher2k.f
+  cgemmtr.f)
 
-set(DBLAS3 dgemm.f dsymm.f dsyrk.f dsyr2k.f dtrmm.f dtrsm.f dgemmtr.f
-	dskewsymm.f dskewsyr2k.f)
+set(DBLAS3
+  dgemm.f dsymm.f dsyrk.f dsyr2k.f dtrmm.f dtrsm.f dgemmtr.f dskewsymm.f
+  dskewsyr2k.f)
 
-set(ZBLAS3 zgemm.f zsymm.f zsyrk.f zsyr2k.f ztrmm.f ztrsm.f
-	zhemm.f zherk.f zher2k.f zgemmtr.f)
+set(ZBLAS3
+  zgemm.f zsymm.f zsyrk.f zsyr2k.f ztrmm.f ztrsm.f zhemm.f zherk.f zher2k.f
+  zgemmtr.f)
 
 
 set(SOURCES)
@@ -115,65 +117,41 @@ if(BUILD_COMPLEX16)
 endif()
 list(REMOVE_DUPLICATES SOURCES)
 
-add_library(${BLASLIB}_obj OBJECT ${SOURCES})
-set_target_properties(${BLASLIB}_obj PROPERTIES POSITION_INDEPENDENT_CODE ON)
+if(BUILD_DEFAULT_API)
+  add_library(${BLASLIB}_obj OBJECT ${SOURCES})
+endif()
 
 if(BUILD_INDEX64_EXT_API)
-  set(SOURCES_64_F)
-  # Copy files so we can set source property specific to /${BLASLIB}_64_obj target
-  file(MAKE_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/${BLASLIB}_64_obj)
-  file(COPY ${SOURCES} DESTINATION ${CMAKE_CURRENT_BINARY_DIR}/${BLASLIB}_64_obj)
-  file(GLOB SOURCES_64_F ${CMAKE_CURRENT_BINARY_DIR}/${BLASLIB}_64_obj/*.f*)
-  add_library(${BLASLIB}_64_obj OBJECT ${SOURCES_64_F})
+  include(ExtendedAPIHelpers)
+  generate_64bit_suffixed_sources(${BLASLIB} SOURCES SOURCES_64)
+
+  add_library(${BLASLIB}_64_obj OBJECT ${SOURCES_64})
   target_compile_options(${BLASLIB}_64_obj PRIVATE ${FOPT_ILP64})
-  set_target_properties(${BLASLIB}_64_obj PROPERTIES POSITION_INDEPENDENT_CODE ON)
-  #Add _64 suffix to all Fortran functions via macros
-  foreach(F IN LISTS SOURCES_64_F)
-    if(CMAKE_Fortran_COMPILER_ID STREQUAL "NAG" OR CMAKE_Fortran_COMPILER_ID MATCHES "Intel")
-      set_source_files_properties(${F} PROPERTIES COMPILE_FLAGS "-fpp")
-    else()
-      set_source_files_properties(${F} PROPERTIES COMPILE_FLAGS "-cpp")
-    endif()
-    file(STRINGS ${F} ${F}.lst)
-    list(FILTER ${F}.lst INCLUDE REGEX "subroutine|SUBROUTINE|external|EXTERNAL|function|FUNCTION")
-    list(FILTER ${F}.lst EXCLUDE REGEX "^!.*")
-    list(FILTER ${F}.lst EXCLUDE REGEX "^[*].*")
-    list(FILTER ${F}.lst EXCLUDE REGEX "end|END")
-    foreach(FUNC IN LISTS ${F}.lst)
-      string(REGEX REPLACE "^[a-zA-Z0-9_ *]*(subroutine|SUBROUTINE|external|EXTERNAL|function|FUNCTION)[ ]*[*]?" "" FUNC ${FUNC})
-      string(REGEX REPLACE "[(][a-zA-Z0-9_, )]*$" "" FUNC ${FUNC})
-      string(STRIP ${FUNC} FUNC)
-      list(APPEND COPT_64_F "${FUNC}=${FUNC}_64")
-    endforeach()
-    list(REMOVE_DUPLICATES COPT_64_F)
-    set_source_files_properties(${F} PROPERTIES COMPILE_DEFINITIONS "${COPT_64_F}")
-  endforeach()
 endif()
 
 add_library(${BLASLIB}
-        $<TARGET_OBJECTS:${BLASLIB}_obj>
-        $<$<BOOL:${BUILD_INDEX64_EXT_API}>: $<TARGET_OBJECTS:${BLASLIB}_64_obj>>)
+  $<$<BOOL:${BUILD_DEFAULT_API}>: $<TARGET_OBJECTS:${BLASLIB}_obj>>
+  $<$<BOOL:${BUILD_INDEX64_EXT_API}>: $<TARGET_OBJECTS:${BLASLIB}_64_obj>>)
 
 # For flang, use C linker instead of Fortran linker to avoid macOS-specific flags
 # that CMake adds (tested CMake 4.2).
-if (CMAKE_Fortran_COMPILER_ID STREQUAL "LLVMFlang")
-   set_target_properties (${BLASLIB} PROPERTIES LINKER_LANGUAGE C)
+if(CMAKE_Fortran_COMPILER_ID STREQUAL "LLVMFlang")
+  set_target_properties (${BLASLIB} PROPERTIES LINKER_LANGUAGE C)
 endif()
 
 set_target_properties(
   ${BLASLIB} PROPERTIES
   VERSION ${LAPACK_VERSION}
   SOVERSION ${LAPACK_MAJOR_VERSION}
-  POSITION_INDEPENDENT_CODE ON
   )
 lapack_install_library(${BLASLIB})
 
 add_library(BLAS::BLAS ALIAS ${BLASLIB})
 install(EXPORT ${BLASLIB}-targets
-    FILE ${BLASLIB}-targets.cmake
-    NAMESPACE BLAS::
-    DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/${LAPACKLIB}-${LAPACK_VERSION}
-    COMPONENT Development
+  FILE ${BLASLIB}-targets.cmake
+  NAMESPACE BLAS::
+  DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/${LAPACKLIB}-${LAPACK_VERSION}
+  COMPONENT Development
 )
 
 if( TEST_FORTRAN_COMPILER )

BLAS/TESTING/CMakeLists.txt

@@ -1,12 +1,10 @@
-macro(add_blas_test name src)
-  get_filename_component(baseNAME ${src} NAME_WE)
-  set(TEST_INPUT "${CMAKE_CURRENT_SOURCE_DIR}/${baseNAME}.in")
+function(_add_blas_test name src test_input)
   add_executable(${name} ${src})
   target_link_libraries(${name} ${BLASLIB})
-  if(EXISTS "${TEST_INPUT}")
+  if(EXISTS "${test_input}")
     add_test(NAME BLAS-${name} COMMAND "${CMAKE_COMMAND}"
       -DTEST=$<TARGET_FILE:${name}>
-      -DINPUT=${TEST_INPUT}
+      -DINPUT=${test_input}
       -DINTDIR=${CMAKE_CFG_INTDIR}
       -P "${LAPACK_SOURCE_DIR}/TESTING/runtest.cmake")
   else()
@@ -21,7 +19,34 @@ macro(add_blas_test name src)
   if(CMAKE_Fortran_COMPILER_ID STREQUAL "NAG")
     target_compile_options(${name} PRIVATE "-Onopropagate")
   endif()
-endmacro()
+endfunction()
+
+function(add_blas_test name source)
+  get_filename_component(baseNAME ${source} NAME_WE)
+  set(test_input "${CMAKE_CURRENT_SOURCE_DIR}/${baseNAME}.in")
+
+  if(BUILD_DEFAULT_API)
+    _add_blas_test(${name} ${source} "${test_input}")
+  endif()
+
+  if(BUILD_INDEX64_EXT_API)
+    include(ExtendedAPIHelpers)
+    generate_64bit_suffixed_sources(${name} source source_64 NO_STRING_REPLACEMENTS)
+
+    # Create 64-bit version of test input if it exists, replacing output
+    # file names with their 64-bit counterparts
+    if(EXISTS "${test_input}")
+      file(READ "${test_input}" test_input_content)
+      string(REPLACE "'${baseNAME}.out'" "'${baseNAME}_64.out'"
+        test_input_content "${test_input_content}")
+      set(test_input_64 "${CMAKE_CURRENT_BINARY_DIR}/${baseNAME}_64.in")
+      file(WRITE "${test_input_64}" "${test_input_content}")
+    endif()
+
+    _add_blas_test(${name}_64 "${source_64}" "${test_input_64}")
+    target_compile_options(${name}_64 PRIVATE ${FOPT_ILP64})
+  endif()
+endfunction()
 
 if(BUILD_SINGLE)
   add_blas_test(xblat1s sblat1.f)

CBLAS/examples/CMakeLists.txt

@@ -1,11 +1,13 @@
-add_executable(xexample1_CBLAS cblas_example1.c)
-add_executable(xexample2_CBLAS cblas_example2.c)
+if(BUILD_DEFAULT_API)
+  add_executable(xexample1_CBLAS cblas_example1.c)
+  add_executable(xexample2_CBLAS cblas_example2.c)
 
-target_link_libraries(xexample1_CBLAS ${CBLASLIB})
-target_link_libraries(xexample2_CBLAS ${CBLASLIB} ${BLAS_LIBRARIES})
+  target_link_libraries(xexample1_CBLAS ${CBLASLIB})
+  target_link_libraries(xexample2_CBLAS ${CBLASLIB} ${BLAS_LIBRARIES})
 
-add_test(example1_CBLAS ${CMAKE_RUNTIME_OUTPUT_DIRECTORY}/xexample1_CBLAS)
-add_test(example2_CBLAS ${CMAKE_RUNTIME_OUTPUT_DIRECTORY}/xexample2_CBLAS)
+  add_test(example1_CBLAS ${CMAKE_RUNTIME_OUTPUT_DIRECTORY}/xexample1_CBLAS)
+  add_test(example2_CBLAS ${CMAKE_RUNTIME_OUTPUT_DIRECTORY}/xexample2_CBLAS)
+endif()
 
 if(BUILD_INDEX64_EXT_API)
   add_executable(xexample1_64_CBLAS cblas_example1_64.c)

CBLAS/include/CMakeLists.txt

@@ -17,8 +17,6 @@ if(CBLAS)
   install(FILES ${CBLAS_INCLUDE}
     DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}
     COMPONENT Development)
-
-  file(COPY cblas_test.h DESTINATION ${LAPACK_BINARY_DIR}/test_include)
 endif()
 
 # Create a header file for CBLAS routine mangling (cblas_mangling.h)

CBLAS/include/cblas_64.h

@@ -109,7 +109,7 @@ void cblas_zcopy_64(const int64_t N, const void *X, const int64_t incX,
                  void *Y, const int64_t incY);
 void cblas_zaxpy_64(const int64_t N, const void *alpha, const void *X,
                  const int64_t incX, void *Y, const int64_t incY);
-void cblas_zaxbpy_64(const int64_t N, const void *alpha, const void *X,
+void cblas_zaxpby_64(const int64_t N, const void *alpha, const void *X,
                  const int64_t incX, const void *beta, void *Y, const int64_t incY);
 
 

CBLAS/include/cblas_globals.h

@@ -0,0 +1,13 @@
+#ifndef CBLAS_GLOBALS_H
+#define CBLAS_GLOBALS_H
+
+#if defined(CBLAS_DLL_IMPORTS)
+  #define CBLAS_GLOBAL_SYMBOL __declspec(dllimport)
+#else
+  #define CBLAS_GLOBAL_SYMBOL
+#endif
+
+extern CBLAS_GLOBAL_SYMBOL int CBLAS_CallFromC;
+extern CBLAS_GLOBAL_SYMBOL int RowMajorStrg;
+
+#endif