|
@emeryberger | |||||
|
“Mesh: Compacting Memory Management for C/C++ Applications” arxiv.org/abs/1902.04738, #PLDI 2019 pic.twitter.com/6qIAzeesdY
|
||||||
|
||||||
|
@vanderZwan@vis.social
@JobvdZwan
|
17. velj |
|
Is the shuffle vector a novel data structure? It almost looks too elegant and simple to not have been discovered yet!
I wonder what other uses it might have? pic.twitter.com/jd16LKXhwO
|
||
|
|
||
|
@vanderZwan@vis.social
@JobvdZwan
|
18. velj |
|
So to all the gamedev and procgen prodigies I follow on Twitter, like @mrj_games, @GridSageGames, @redblobgames, @gavanw, @eigenbom: can you see a use for this? A really elegant and low-overhead randomization data structure.
|
||
|
|
||
|
Mike Conley @mconley@mastodon.social
@mike_conley
|
17. velj |
|
Hey @MikeHommey and @froydnj, did you see this?
|
||
|
|
||
|
glandium
@MikeHommey
|
17. velj |
|
I just realized I had a pre-print of the paper sitting on my to-read list... since June.
|
||
|
|
||
|
Marek D
@marekdenis
|
17. velj |
|
Paper candidate for @adriancolyer ?:)
|
||
|
|
||
|
Kees Cook
@kees_cook
|
18. velj |
|
"Mesh’s write barrier is implemented with [read-only] page protections and a segfault trap handler [... which] waits on a lock for the current meshing operation to complete, the last step of which is remapping the source virtual span as read/write." 😍
|
||
|
|
||
|
Robert O'Callahan
@rocallahan
|
17. velj |
|
Very nice! There are a few issues I would worry about that you didn't mention in the paper...
|
||
|
|
||
|
Robert O'Callahan
@rocallahan
|
17. velj |
|
VMA fragmentation. As the number of VMAs grows Linux VM operations get slower, and there is a hard limit of 65536 per process. Seems like you have to keep a mapped view of a span around as long as there is at least one object in the span.
|
||
|
|
||
|
Katharina 🍪🏴 Fey
@spacekookie
|
17. velj |
|
The paper seems to have some syntax errors (Linux if MacOS, section 3.2 has a half sentence dangling in it,...) but is otherwise very well written!
|
||
|
|
||
|
Olaf Marzocchi
@profsen
|
17. velj |
|
Is this something a new compiler could just apply blindly?
|
||
|
|
||