What's New: Intel® Software Development Tools 2025.0

[StefR_Intel]

The Intel® Distribution for Python has been updated to include functional and security updates. Users should update to the latest version. For more details on security and known issues, please check out Intel® Distribution for Python* Known Issues.

Deprecation notice

Starting with the 2025.0 release, the Data Parallel Extension for Numba* will no longer be part of the Intel® distribution for Python*. However, all previously published packages are still available on the package channel (https://software.repos.intel.com/python/conda).

What’s new

The Intel® Distribution for Python* now ships with Python 3.12, although Python 3.11 packages are readily available in Intel’s conda package channel (https://software.repos.intel.com/python/conda), which will already be configured in one’s .condarc if Intel® Distribution for Python’s installer was used. 

Data Parallel Extension for Numpy* expands compatibility, adding NumPy 2.0 support in the runtime and providing asynchronous execution of offloaded operations. This update provides significant new functionality with support for more than 25 new functions and keywords providing 90% functional combability with CuPy*. 

• Execution of offloaded operations is now asynchronous. 

• Deprecated dpnp.dparray interface was removed. 

• Support for DLPack 1.0 exchange. 

• Implemented dpnp.fft module for Discrete Fourier Transform which functions support all keywords available in NumPy and CuPy. 

• dpnp.graidient supports all keywords available in NumPy and CuPy. 

• dpnp.unwrap supports all keywords available in NumPy and CuPy. 

• dpnp.ediff1d supports all keywords available in NumPy and CuPy. 

• dpnp.allclose and dpnp.isclose support all keywords available in NumPy and CuPy. 

• dpnp.extract supports input arrays of different shapes. 

• Removed limitations on input arguments in dpnp.take and dpnp.place functions. 

• Added support of device keyword to dpnp.astype. 

• Added support of out keyword in dpnp.all/dpnp.any. 

• New implemented functions: 
  • dpnp.unique 
  • dpnp.sort_complex 
  • dpnp.nextafter 
  • dpnp.trim_zeros 
  • dpnp.flatnonzero 
  • dpnp.nan_to_num 
  • dpnp.isneginf 
  • dpnp.isposinf 
  • dpnp.iscomplex 
  • dpnp.iscomplexobj 
  • dpnp.isreal 
  • dpnp.isrealobj 
  • dpnp.isscalar 
  • dpnp.array_equal 
  • dpnp.array_equiv 

Data Parallel Control Library* also expands compatibility, adding NumPy 2.0 support in the runtime and providing asynchronous execution of offloaded operations. This update expands the functionality of Python Array API support. 

• Execution of offloaded operations is now asynchronous. 

• Size of struct for dpctl.memory objects has changed. 

• DPCTLSyclInterface library changed from being an SO-versioned shared library on Linux to being an ordinary shared library. 

• Support for DLPack 1.0 exchange as mandated by Python Array API 2023.12 

• Python scalar support for dpctl.tensor.where function. 

• New Python Array API functions added:  
  • dpctl.tensor.nextafter 
  • dpctl.tensor.diff 
  • dpctl.tensor.count_nonzero 
  • dpctl.tensor.take_along_axis 

The Intel® Distribution for Python*, 2025 includes an online-based installer that fetches the latest conda packages available from Intel’s and conda-forge’s conda channels (https://software.repos.intel.com/python/conda and https://anaconda.org/conda-forge, respectively). This ensures our users get packages that are security-compliant and most up to date. We are providing instructions for those who require an offline installation as part of the “Getting Started guide”. 
​​