One of the key takeaways of last lecture was that the implementation of parallelism can be vastly different from the abstraction that's exposed to the programmer. That's apparent here because each of these functions seem to be doing the same computation in parallel but how they are called is vastly different.

An interesting thing to note about this slide is that the abstractions presented to the programmer are different level of abstract. This allows for different amounts of software-based optimization by the programmer based on what/how resources are used. What unites these abstractions is that they all provide a way for the programmer to declare what kind of parallelism they want, and the compiler/system handles how the parallelism is implemented