Stream: t-compiler/wg-polonius

Topic: Intuition for `DatafrogOpt`


ecstatic-morse (Jan 01 2021 at 19:49, on Zulip):

Am I correct that rust-lang/polonius#23 is the genesis of the DatafrogOpt ruleset? I stared at it for a while but couldn't figure out what was going on before looking through the git history and finding that PR.

lqd (Jan 01 2021 at 22:37, on Zulip):

absolutely

lqd (Jan 01 2021 at 22:45, on Zulip):

the intuition I have (and note that I don't understand this variant as well as I should) is that the vast majority of subsets are local and involve dead origins (that is, they mostly participate in propagating loans at a single point), and the origins of interest (especially in the NLL case) at least involved dying along an edge (so that both subsets and loans were to be propagated down the CFG) -- and which I like to call "differential liveness" -- so this is where this variant focuses the TC

Last update: Jun 20 2021 at 00:15UTC