Berkeley UPC

Berkeley UPC is a cool extension of the C programming language. It's specifically made for high-performance computing on big parallel machines. What’s awesome about this language is that it offers a uniform programming model that works for both shared and distributed memory hardware.

Understanding Shared Memory

So, here’s how it works: programmers get a single shared, partitioned address space. This means any processor can directly read and write to variables, but each variable is actually linked to just one processor.

The SPMD Model

UPC uses something called the Single Program Multiple Data (SPMD) model. This means the level of parallelism is set when the program starts up, and typically there’s just one thread of execution per processor.

Key Features of UPC

If you want to express parallelism in your code, UPC makes it easy by extending ISO C 99 with some handy features:

• An explicitly parallel execution model


• A shared address space


• Synchronization primitives and a memory consistency model


• Memory management tools

This combination allows you to enjoy the flexibility of shared memory programming while also having control over data layout and performance—just like message passing programming!

What Makes Berkeley UPC Special?

Berkeley UPC stands out because it's not only high-performance but also portable and completely open-source! It was built for large-scale multiprocessors, PC clusters, and clusters with shared memory multiprocessors.

Your Mac Can Handle It!

This is actually the first UPC implementation that works on Mac OS X. It gives supercomputing applications competitive performance on some of the largest OS X clusters around. Plus, you can even develop UPC code right on your Mac OS X laptops and desktops! If you're interested in learning more or want to download it here!