change to dynamic linking and add interface and threading layers (#72)

* change to dynamic linking and add interface and threading layers

* update changelog

* fix typos

* switch to lp64

Author: vtavana

.github/workflows/build_pip.yaml

@@ -42,10 +42,9 @@ jobs:
 
       - name: Install Compiler and MKL
         run: |
-          conda install mkl-devel mkl-service dpcpp_linux-64
+          conda install mkl-devel tbb-devel dpcpp_linux-64
           python -c "import sys; print(sys.executable)"
           which python
-          python -c "import mkl; print(mkl.__file__)"
 
       - name: Build conda package
         run: |

CHANGELOG.md

@@ -4,13 +4,16 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
-## [0.2.0] (05/DD/2025)
+## [0.2.0] (06/DD/2025)
+This release updates `mkl_umath` to be aligned with both numpy-1.26.x and numpy-2.x.x.
 
 ### Added
-* `mkl_umath` is now aligned with both numpy-1.26,x and numpy-2.x.x [gh-65](https://github.com/IntelPython/mkl_umath/pull/65)
+* The definition of `sign` function for complex floating point data types is updated to match numpy-2.x.x [gh-65](https://github.com/IntelPython/mkl_umath/pull/65)
+* `ldexp` function is updated to allow `int64` explicitly similar to numpy-2.x.x behavior [gh-73](https://github.com/IntelPython/mkl_umath/pull/73)
 
 ### Changed 
 * Migrated from `setup.py` to `pyproject toml` [gh-63](https://github.com/IntelPython/mkl_umath/pull/63)
+* Changed to dynamic linking and added interface and threading layers [gh-72](https://github.com/IntelPython/mkl_umath/pull/72)
 
 ### Fixed
 * Fixed a bug for `mkl_umath.is_patched` function [gh-66](https://github.com/IntelPython/mkl_umath/pull/66)

CMakeLists.txt

@@ -23,7 +23,11 @@ cmake_print_variables(Python_NumPy_INCLUDE_DIRS)
 set(CYTHON_FLAGS "-t -w \"${CMAKE_SOURCE_DIR}\"")
 find_package(Cython REQUIRED)
 
-set(MKL_LINK sdl)
+find_package(TBB REQUIRED)
+set(MKL_ARCH "intel64")
+set(MKL_LINK "dynamic")
+set(MKL_THREADING "tbb_thread")
+set(MKL_INTERFACE "lp64")
 find_package(MKL REQUIRED)
 
 if(WIN32)

README.md

@@ -47,6 +47,7 @@ If these are installed as part of a `oneAPI` installation, the following package
 
 If build dependencies are to be installed with Conda, the following packages must be installed from the Intel(R) channel
 - `mkl-devel`
+- `tbb-devel`
 - `dpcpp_linux-64` (or `dpcpp_win-64` for Windows)
 - `numpy-base`
 
@@ -56,7 +57,7 @@ then the remaining dependencies
 - `cython`
 - `scikit-build`
 
-and for `mkl-devel` and `dpcpp_linux-64` in a Conda environment, `MKLROOT` environment variable must be set
+and for `mkl-devel`, `tbb-devel` and `dpcpp_linux-64` in a Conda environment, `MKLROOT` environment variable must be set
 On Linux
 ```sh
 export MKLROOT=$CONDA_PREFIX

conda-recipe-cf/meta.yaml

@@ -28,11 +28,10 @@ requirements:
       - scikit-build
       - python
       - mkl-devel
+      - tbb-devel
       - numpy
     run:
       - python
-      - mkl
-      - mkl-service
       - {{ pin_compatible('intel-cmplr-lib-rt') }}
 
 test:

conda-recipe/meta.yaml

@@ -28,11 +28,10 @@ requirements:
       - scikit-build
       - python
       - mkl-devel
+      - tbb-devel
       - numpy-base
     run:
       - python
-      - mkl
-      - mkl-service
       - {{ pin_compatible('intel-cmplr-lib-rt') }}
       - {{ pin_compatible('numpy') }}