Run time vs. state
Issue #99
resolved
In the overview section we make a big deal about how it's more correct to talk about state rather than a upc++ run time. But on lines 1392-3 (Chapter 9, progress) we say
Progress in UPC++ refers to how the calling application allows the UPC++ internal run time to advance the state of its outstanding asynchronous operations.
We use the word "runtime" freely throughout the spec. We should clarify at the start that our use of the word run time is an abuse of the notation, or change the word "runtime" to "state" throughout the document.
Comments (2)
-
-
- changed status to resolved
fix issue
#99: remove a confusing overview sentence→ <<cset e4f94bc95887>>
- Log in to comment
I propose we fix this by striking the overview sentence that makes this sound like an important distinction:
The literature uses "runtime" as a nebulous term (eg see this early paper), and there's no strong correlation between "runtime" and "hidden threads of execution". The sentence right before this one makes it clear UPC++ does not use hidden progress threads, so IMO there's already no ambiguity here, and no reason we should avoid the term runtime.