You'll never need to make it smaller, but this requires the same efforts.
But good to see someone finally explaining the simplest and fastest GC, because the objects are compacted, usually in the same cache line. MMTk is way underused
The intent is to help improve writing clarity because, to lawyers and us amateur unsolicited grammar "mall security guards", words and word choices can vary in accuracy and precision based on how they were chosen to increase or decrease confusion. Clarity is important because being imprecise or inaccurate imposes a costlier burden on multiple people other than yourself. English is a difficult and messy language, with slightly different interpretations conventions British and American flavoⓊrs, seemingly arbitrary odd rules, and much imprecision; it's basically Indo-Greek-Latin-Germanic "Creole gumbo". (I can't even learn Spanish, much less Arabic, Hindi, Korean, Mandarin, Russian, or Urdu.) And beware of singulare tantum without plural forms like "feedback", for they are sneaky too.
C4 2011 [PDF] https://www.azul.com/sites/default/files/images/c4_paper_acm...
ORCA 2013 [PDF] https://www.ponylang.io/media/papers/opsla237-clebsch.pdf
Pony+ORCA 2017 [PDF] https://www.ponylang.io/media/papers/orca_gc_and_type_system...
Generational GC (as seen in Erlang BEAM) 2008-2010? [PDF, presentation not paper] https://www.cs.purdue.edu/homes/hosking/661-12/20120216.pdf
[PDF, different presentation] https://pdfs.semanticscholar.org/737d/041822cca60a341e4058ba...
Also: https://gchandbook.org
An immediate question I have is whether one could extend compacting GCs to also "sort" or "cluster" the live objects for better cache access in some way (in a practical sense I mean, where the extra runtime overhead is worth it). Also what kind clustering would actually work in that case.
I'm guessing naively tracking how often each object is accessed in a certain window of time and sorting by that wouldn't work, since the most accessed objects could be used at completely different times - some sort of "accessed together" algorithm probably makes more sense. But then we're in the territory of a combinatorial explosion of options if we're not careful.
Maybe something really stupid would work:
1. use a single global integer
that increases by one with
each heap memory access,
2. assign the counter's value to
GC metadata of the accessed
object in question as a sort
of "time stamp"
3. during compacting, "sort"
all living objects by whatever
they have as their value
(this is probably really hard)
4. after compacting, reset counter
and time stamps
Worst case this would put objects right next to at least two objects that were accessed around the same time.I presume that this question must have been researched to death, does anyone have any suggestions?