Release v3.4.5

Author: grisuthedragon

.editorconfig

@@ -0,0 +1,33 @@
+root = true
+
+[*]
+end_of_line = lf
+trim_trailing_whitespace = true
+charset = utf-8
+
+[Makefile]
+indent_style = tab
+
+[*.{c,cpp,h,hpp,f,f90,F,F90}]
+indent_style = space
+indent_size = 4
+insert_final_newline = true
+
+[*.{f,F}]
+max_line_length = 72
+
+[*.{f90,f95,f03,F90,F95,F03}]
+max_line_length = 132
+
+[*{c,cpp,h,hpp}]
+curly_bracket_next_line = true
+indent_brace_style = Allman
+
+[CMakeLists.txt]
+indent_style = space
+indent_size = 4
+
+[*.cmake]
+indent_style = space
+indent_size = 4
+

.gitignore

@@ -6,6 +6,7 @@
 *.o
 *.oct
 *.pyc
+*.rej
 *.swp
 *.tar.gz
 *octave-workspace*

CHANGELOG

@@ -1,9 +1,16 @@
-next
-----
+3.4.5 (January 30, 2025)
+------------------------
 - Fix cblas scnrm2 typo (See GH #54)
 - Consistent indentation (See GH #55)
 - Fix some typos (See GH #56)
-
+- Support for NVPL BLAS
+- Fix bug in xAXPBY with ALPHA = BETA = 0.0
+- Add different symbols nameing schemens in parallel
+- Add support for xGEMMT/xGEMMTR
+- Fix all disclaimers to match the SPDX identifier
+- Fix building the auxillary library on Apple MacOS X (GH#61)
+- Fix underlinkage with GCC 15 and glibc, (GH #62)
+- Fix: detect wrong MKL/TBB without crashing the build process.
 
 3.4.4 (May 24, 2024)
 --------------------

CITATION.cff

@@ -8,9 +8,9 @@ authors:
     given-names: "Jens"
     orcid: "https://orcid.org/0000-0001-5567-9637"
 title: "FlexiBLAS - A BLAS and LAPACK wrapper library with runtime exchangable backends"
-version: 3.4.4
-doi: 10.5281/zenodo.11260086
-date-released: 2024-05-24
+version: 3.4.5
+doi: 10.5281/zenodo.14764701
+date-released: 2025-01-30
 url: "https://github.com/mpimd-csc/flexiblas"
 license: LGPL-3.0-or-later
 keywords:

CMakeLists.txt

@@ -1,10 +1,11 @@
-cmake_minimum_required(VERSION 3.9.0)
+cmake_minimum_required(VERSION 3.10.0)
 INCLUDE(libcscutils/CMakeModules/CheckPolicy.cmake)
 CheckPolicy(CMP0048 NEW)
+CheckPolicy(CMP0054 NEW)
+CheckPolicy(CMP0069 NEW)
+CheckPolicy(CMP0075 NEW)
 
-PROJECT(FlexiBLAS LANGUAGES C Fortran VERSION 3.4.4)
-cmake_policy(SET CMP0054 NEW)
-cmake_policy(SET CMP0069 NEW)
+PROJECT(FlexiBLAS LANGUAGES C Fortran VERSION 3.4.5)
 
 
 #
@@ -105,7 +106,6 @@ INCLUDE(ccache)
 
 
 #SET(BUILD_SHARED_LIBS ON)
-SET(LIBS m)
 INCLUDE(CheckCCompilerFlag)
 INCLUDE(CheckFortranCompilerFlag)
 
@@ -195,7 +195,10 @@ IF (LINK_OPENMP)
     endif()
 ENDIF()
 
-
+INCLUDE(cmake/libm.cmake)
+IF (NEED_LINKING_AGAINST_LIBM)
+    SET(LIBS ${LIBS} m)
+ENDIF()
 
 
 
@@ -204,6 +207,10 @@ IF (NOT DEFINED ABI)
         SET(ABI "Intel" CACHE STRING "ABI of the Fortran Interface")
     ELSEIF ( ${CMAKE_Fortran_COMPILER_ID} STREQUAL "PGI" )
         SET(ABI "Intel" CACHE STRING "ABI of the Fortran Interface")
+    ELSEIF ( ${CMAKE_Fortran_COMPILER_ID} STREQUAL "NVHPC" AND CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64" )
+        SET(ABI "GNU" CACHE STRING "ABI of the Fortran Interface")
+    ELSEIF ( ${CMAKE_Fortran_COMPILER_ID} STREQUAL "NVHPC" AND CMAKE_SYSTEM_PROCESSOR MATCHES "x86_64" )
+        SET(ABI "Intel" CACHE STRING "ABI of the Fortran Interface")
     ELSEIF ( ${CMAKE_Fortran_COMPILER_ID} STREQUAL "XL" )
         SET(ABI "XL" CACHE STREQUAL "ABI of the Fortran Interface")
     ELSE()
@@ -228,11 +235,34 @@ ENDIF()
 
 # Name Mangling
 include(FortranCInterface)
-FortranCInterface_HEADER(${PROJECT_BINARY_DIR}/include/flexiblas_fortran_mangle.h
+FortranCInterface_HEADER(${PROJECT_BINARY_DIR}/fc.h
     MACRO_NAMESPACE "FC_"
     SYMBOL_NAMESPACE "FC_")
 FortranCInterface_VERIFY()
 
+MESSAGE(STATUS "Fortran - C - Interface -- FortranCInterface_GLOBAL_SUFFIX = ${FortranCInterface_GLOBAL_SUFFIX}")
+MESSAGE(STATUS "Fortran - C - Interface -- FortranCInterface_GLOBAL_CASE = ${FortranCInterface_GLOBAL_CASE}")
+
+IF ( FortranCInterface_GLOBAL_SUFFIX STREQUAL "_" AND FortranCInterface_GLOBAL_CASE STREQUAL "LOWER")
+    SET(FLEXIBLAS_MANGLE_DEFAULT 1)
+    SET(FLEXIBLAS_MANGLE_WINDOWS 0)
+    SET(FLEXIBLAS_MANGLE_UPPPERCASE 0)
+    MESSAGE(STATUS "Use default Fortran name mangling.")
+ELSEIF(FortranCInterface_GLOBAL_SUFFIX STREQUAL "" AND FortranCInterface_GLOBAL_CASE STREQUAL "LOWER")
+    SET(FLEXIBLAS_MANGLE_DEFAULT 0)
+    SET(FLEXIBLAS_MANGLE_WINDOWS 1)
+    SET(FLEXIBLAS_MANGLE_UPPPERCASE 0)
+    MESSAGE(STATUS "Use Windows-like Fortran name mangling.")
+ELSEIF(FortranCInterface_GLOBAL_SUFFIX STREQUAL "" AND FortranCInterface_GLOBAL_CASE STREQUAL "UPPER")
+    SET(FLEXIBLAS_MANGLE_DEFAULT 0)
+    SET(FLEXIBLAS_MANGLE_WINDOWS 0)
+    SET(FLEXIBLAS_MANGLE_UPPPERCASE 1)
+    MESSAGE(STATUS "Use upper-case Fortran name mangling.")
+ENDIF()
+
+
+
+
 include(CheckThreadLocal)
 
 IF (DEFINED FALLBACK)
@@ -329,7 +359,9 @@ SET(LIBS ${LIBS})
 SET(INCLUDE_DIR ${INCLUDE_DIR} "${PROJECT_SOURCE_DIR}/libcscutils/include" "${PROJECT_BINARY_DIR}/libcscutils/include")
 
 IF (CBLAS STREQUAL ON)
-    ADD_DEFINITIONS(-DFLEXIBLAS_CBLAS)
+    SET(FLEXIBLAS_CBLAS TRUE)
+ELSE()
+    SET(FLEXIBLAS_CBLAS FALSE)
 ENDIF()
 
 
@@ -422,7 +454,6 @@ INSTALL(FILES   ${CMAKE_SOURCE_DIR}/include/blas_gnu.h
     ${CMAKE_SOURCE_DIR}/include/blas_intel.h
     ${CMAKE_BINARY_DIR}/include/flexiblas_config.h
     ${CMAKE_BINARY_DIR}/include/flexiblas_api.h
-    ${CMAKE_BINARY_DIR}/include/flexiblas_fortran_mangle.h
     DESTINATION ${CMAKE_INSTALL_FULL_INCLUDEDIR}/${FLEXIBLAS_INCLUDE_PREFIX}/)
 IF(LAPACK STREQUAL ON)
     INSTALL(FILES   ${CMAKE_BINARY_DIR}/include/lapack.h
@@ -431,6 +462,7 @@ IF(LAPACK STREQUAL ON)
 
 ENDIF()
 INSTALL(FILES   ${CMAKE_SOURCE_DIR}/src/flexiblas_backend.h
+    ${CMAKE_SOURCE_DIR}/src/flexiblas_fortran_mangle.h
     ${CMAKE_SOURCE_DIR}/src/flexiblas_real_calls.h
     ${CMAKE_SOURCE_DIR}/src/flexiblas_real_cblas_calls.h
     ${CMAKE_SOURCE_DIR}/src/flexiblas_dummy_fortran.h

CODE

@@ -1,8 +1,8 @@
 name: FlexiBLAS - A BLAS and LAPACK wrapper library with runtime exchangable backends
 shortname: flexiblas
-version: 3.4.4
-release-date: 2024-05-24
-id: 10.5281/zenodo.11260086
+version: 3.4.5
+release-date: 2025-01-30
+id: 10.5281/zenodo.14764701
 id-type: doi
 author: Martin Köhler, Jens Saak
 orcid: 0000-0003-2338-9904, 0000-0001-5567-9637

CONTRIBUTING.md

@@ -0,0 +1,71 @@
+# FlexiBLAS Contribution Guide
+
+Thank you for considering contributing to the FlexiBLAS open source project! 
+We welcome contributions from everyone. Please follow these guidelines to ensure 
+a smooth contribution process.
+
+## How to Contribute
+
+1. **Fork the Repository**: Start by forking the repository to your GitHub account.
+
+2. **Clone the Repository**: Clone your forked repository to your local machine.
+
+   ```bash
+   git clone https://github.com/your-username/flexiblas.git
+   ```
+
+3. **Create a Branch**: Create a new branch for your feature or bug fix.
+
+   ```bash
+   git checkout -b feature-or-bugfix-name
+   ```
+
+4. **Make Changes**: Implement your changes in the codebase and add yourself to the list 
+   of contributors in the `CONTRIBUTORS.md` file in the root of the project. By doing this 
+   you accept the conditions from this contribution guide. 
+
+5. **Commit Changes**: Commit your changes with a clear and descriptive commit message.
+
+   ```bash
+   git commit -m "Description of changes"
+   ```
+
+6. **Push Changes**: Push your changes to your forked repository.
+
+   ```bash
+   git push origin feature-or-bugfix-name
+   ```
+
+7. **Submit a Pull Request**: Go to the [FlexiBLAS repository](https://github.com/mpimd-csc/flexiblas)
+   and submit a pull request. Provide a detailed description of your changes and any related issues.
+   
+**Note**: Merge requests are not directly accepted into the main repository. Instead, they are first 
+reviewed and, if approved, moved to an internal repository for further testing and integration before
+ being merged into the main codebase.
+
+## Code Style Guide
+
+To maintain consistency across the codebase, please adhere to the code style guidelines specified 
+in the `.editorconfig` file located in the root directory of the repository. This file contains 
+settings for indentation, line endings, and other code style preferences. Most modern code editors 
+support `.editorconfig` and will automatically apply these settings. For more details, we refer To
+[https://editorconfig.org/](https://editorconfig.org/)
+
+## Copyright and Licensing
+
+By contributing to this project, you agree to the following terms regarding copyright and licensing:
+
+- **Copyright Assignment**: You agree that the authors of the software retain the copyright to the
+  entire codebase, including your contributions. This means that while you will be acknowledged as 
+  a contributor, the copyright of the code remains with the original authors.
+
+- **Right to Re-license**: The authors of the project reserve the right to re-license the entire
+  codebase, including your contributions, under different terms. This could involve changing the 
+  licensing terms to better suit the project's goals or to comply with legal requirements.
+
+- **Acknowledgment**: Your contributions will be acknowledged in the project's documentation
+  or release notes, giving credit for your work. Additionally, contributors may be listed in 
+  the `CONTRIBUTORS.md` file as a token of appreciation for their efforts.
+
+We appreciate your contributions and look forward to collaborating with you! If you have any 
+questions or concerns about these terms, please feel free to open an issue or contact us directly.

CONTRIBUTORS.md

@@ -0,0 +1,14 @@
+Maintainer and Main Author
+--------------------------
+
+* Martin Köhler, MPI Magdeburg,  koehlerm<|AT|>mpi-magdeburg<|DOT|>mpg<|DOT|>.de
+* Jens Saak, MPI Magdeburg,  saak<|DOT|>mpi-magdeburg<|DOT|>mpg<|DOT|>.de
+
+Contributors
+------------
+Please list yourself, optionally with an affiliation, and a one-liner of your contribution
+
+* Christian Himpe, Universität Münster, testing and bug hunting
+* Jörn Papenbroock, parts of the hook framework 
+* Iñaki Úcar, packaging for Fedora and bug hunting related to the build-system
+

ISSUES.md

@@ -1,7 +1,7 @@
 Issues using FlexiBLAS
 ======================
 
-Last Update: Jun 16, 2020
+Last Update: Oct 10, 2024
 
 Table of Contents
 -----------------
@@ -14,6 +14,9 @@ Table of Contents
 6. CBLAS Interface of BLIS
 7. Building with pre-built BLAS and LAPACK on MacOSX
 8. gcc >= 10.0 and LAPACK 3.7/3.8
+9. LTO Type mismatch with gcc
+10. NVHPC
+
 
 1. Profiling Numpy/Scipy with linked against FlexiBLAS
 ------------------------------------------------------
@@ -110,4 +113,31 @@ cmake ........ -DCMAKE_Fortran_FLAGS="-fallow-argument-mismatch"
 ```
 while configuring FlexiBLAS. Alternatively, the tests can be disabled.
 
+9. LTO Type mismatch with gcc
+-----------------------------
+Some newer versions of gcc (at least version 13.x and 14.x) produce wrong LTO
+warnings. They accidently assume Fortran's hidden string length argument to be a
+`long int` instead of a `size_t` aka `flexiblas_fortran_charlen_t`. This results
+in a warning like:
+```
+    type ‘long int’ should match type ‘flexiblas_fortran_charlen_t’
+```
+while linking the LAPACK fallback library. This error can safely be ignored
+and is already subject to a GCC bug:
+
+    https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102079
+
+10. NVHPC
+---------
+The up to at least version 24.7 of the NVHPC compiler on Aarch64, the `-z now`
+flag is added to all linker calls. This avoids many parts of the hook framework
+to work and even disturbs some backend detection. The problem is known to
+Nvidia, see https://forums.developer.nvidia.com/t/nvc-adds-z-now-to-linking-causes-error-in-plugin-based-codes-on-arm64-gh200/307243
+
+Furthermore, the NVHPC compiler suite leads to severval errors in the testing
+suite, which also appear if one compiles the reference LAPACK with the NVHPC
+compilers. Even simple routines like `dznrm2` are affected by error introduced
+by the NVHPC compiler.
+
+
 

README.md

@@ -1,13 +1,13 @@
 FlexiBLAS - A BLAS and LAPACK wrapper library with runtime exchangeable backends
 ================================================================================
 
-**Version 3.4.4** - DOI: 10.5281/zenodo.11260086
+**Version 3.4.5** - DOI: 10.5281/zenodo.14764701
 
 **Project Website:** https://www.mpi-magdeburg.mpg.de/projects/flexiblas
 
 **Git Repository:** https://gitlab.mpi-magdeburg.mpg.de/software/flexiblas-release
 
-Copyright 2013-2024 by *Martin Köhler* (0000-0003-2338-9904)
+Copyright 2013-2025 by *Martin Köhler* (0000-0003-2338-9904)
                    and *Jens Saak* (0000-0001-5567-9637)
 
 ## Documentation
@@ -212,6 +212,7 @@ The following compilers are tested:
 * Intel icx/ifx oneAPI >= 2022
 * LLVM CLANG/FLANG >= 18
 * AMD AOCC 4.2.0
+* NVHPC on x86_64 (On ARM64 some issues exists.)
 
 Other compilers might work, but not tested.
 

cmake/libm.cmake

@@ -0,0 +1,17 @@
+include(CheckFunctionExists)
+
+if(NOT COS_FUNCTION_EXISTS AND NOT NEED_LINKING_AGAINST_LIBM)
+  CHECK_FUNCTION_EXISTS(cos COS_FUNCTION_EXISTS)
+  if(NOT COS_FUNCTION_EXISTS)
+      MESSAGE(STATUS "Check if cos needs libm")
+      unset(COS_FUNCTION_EXISTS CACHE)
+      list(APPEND CMAKE_REQUIRED_LIBRARIES m)
+      CHECK_FUNCTION_EXISTS(cos COS_FUNCTION_EXISTS)
+      if(COS_FUNCTION_EXISTS)
+          set(NEED_LINKING_AGAINST_LIBM True CACHE BOOL "" FORCE)
+      else()
+          message(FATAL_ERROR "Failed making the cos() or libm does not work properly.")
+      endif()
+  endif()
+endif()
+