The use of Python in HPC is growing at a considerable rate. Python is easy to use and readable: the perfect solution for rapid development of complex workflows. However, because of its interpreted nature, the Python code itself can often fail to meet performance requirements. Fortunately, Python is flexible and can be extended using modules like NumPy or SciPy. Sometimes shipped with Python distributions for scientific computing, they rely on widely used HPC libraries like BLAS or LAPACK and offer the best performance. So how do you diagnose when an application can benefit from these modules, decide which sections to rewrite, and optimize the application further when scaling up?
This webinar offers a masterclass around the topic from Arm’s application experts, using new capability in Arm forge for profiling Python applications at any scale.
- Latest diagnostic techniques for determining problematic Python-based codes
- Practical tips for targeting rewrite and optimization work
- Effective use of libraries and modules for dealing with inefficient code sections