How to improve the performance of your Python applications
Blog|by Leanne Bevan|6 February 2019
As mentioned in previous posts, Python is fast becoming the most popular programming language these days. In this article I will take a look at a range of tools, libraries and other assets that Intel has on offer to help support those programming in Python.
All of these tools and libraries are ideal for:
- Machine Learning developers
- Data scientists
- Data analysts
- Numerical and scientific computing developers
- High-performance computing (HPC) developers
Intel Distribution for Python
Though Python is easy to use and ideal for cloud-based web applications due to its single-threaded nature (bringing flexibility), it executes scripts slowly.
Intel Distribution for Python helps to overcome this issue by enabling you to create Python applications that run almost as fast as native code. You can accelerate NumPy, SciPy and scikit-learn with the help of Intel’s various performance libraries. Plus, access vectorisation and multi-threading instructions such as Numba and Cython and composable parallelism with Threading Building Blocks.
Intel Distribution for Python is included within Intel Parallel Studio XE, a product that contains a collection of Intel’s developer tools and libraries, and is compatible with Visual Studio and PyCharm.
Watch this overview video of the key benefits of the Intel Distribution for Python. There’s also an on-demand video you can watch to learn how to speed up Python applications and soar core computations.
Python Profiling
You can use Intel VTune Amplifier XE in conjunction with Intel Distribution for Python to easily develop and deploy Python applications. This tool can help you to profile your code for performance, identify inefficient synchronisations and wait times (with very low overheads), as well as attach/detach with a running application. Watch this video to find out more.
Gain Performance Insights
Use the Intel Advisor analysis tool (within Intel Parallel Studio XE) to find out data insights such as which loops should be threaded and vectorised first, whether the threading performance will scale with higher core counts, if there is a dependency that prevents vectorisation, identify trip counts and memory access patterns and whether you have vectorised efficiently. Intel Advisor stores its data in a proprietary database that is now accessible via a Python API. You can flexibly generate customised reports on program metrics. Find out how in this article.
Grey Matter is proud to be an Intel Software Elite Reseller. We can support you with licensing and enablement, and offer competitive pricing. We run events throughout the year with Intel Software specialists to help you take full advantage of their tools and new technologies.
Intel Parallel Studio XE contains a collection of Intel developer tools including Intel VTune Amplifier, C and Fortran compilers, numerical libraries, profilers and Intel Distribution for Python. You also receive priority support. You can purchase Intel Parallel Studio XE right away from our online store. Alternatively, try it out now with a free trial.
We are hosting a webinar on 30 April about how you can use parallelism and profiling to improve the performance of Python code. Find out more here.