Hi,
On 2018-01-24 17:07:01 -0300, Alvaro Herrera wrote:
> Andres Freund wrote:
> > glibc's malloc does add a header. My half-informed suspicion is that
> > most newer malloc backing allocators will have a header, because
> > maintaining a shared lookup-by-address table is pretty expensive to
> > maintain. A bit of metadata indicating size and/or source of the
> > allocation makes using thread-local information a lot easier.
>
> Sounds like it'd be smart to allocate something closer to
> M_MMAP_THRESHOLD (which with typical values would be about double the
> amount of memory the current RAW_BUF_SIZE value), minus a few dozen
> bytes to allow for palloc's and malloc's respective headers. That way
> we bet for a mmap()ed allocation with minimum space wastage across all
> layers.
In general there might be cases where that's worthwhile, although
M_MMAP_THRESHOLD IIRC isn't a constant anymore, complicating things. The
specific case this thread is discussing doesn't seem worthy of such
attention though, there's afaict no actual practical problem here.
> Not sure whether we want to try to minimize wastage through
> clever use of malloc_usable_size() on each backend's first allocation --
> that'd probably just be overengineering.
Likely also dangerous, I don't think this is a constant.
Greetings,
Andres Freund