Horton

Re: Reducing the chunk header sizes on all memory context types

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: David Rowley <dgrowleyml@gmail.com>
Cc: Tomas Vondra <tomas.vondra@enterprisedb.com>, Amit Kapila <amit.kapila16@gmail.com>, Andres Freund <andres@anarazel.de>, Robert Haas <robertmhaas@gmail.com>, Yura Sokolov <y.sokolov@postgrespro.ru>, PostgreSQL Developers <pgsql-hackers@lists.postgresql.org>
Date: 2022-08-30T03:15:17Z
Lists: pgsql-hackers 
David Rowley <dgrowleyml@gmail.com> writes:
> On Tue, 30 Aug 2022 at 03:39, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> I'd suggest reverting df0f4feef as it seems to be
>> a red herring.

> I think it's useless providing that a 64-bit variable will always be
> aligned to 8 bytes on all of our supported 32-bit platforms as,
> without the padding, the uint64 hdrmask in MemoryChunk will always be
> aligned to 8 bytes meaning the memory following that will be aligned
> too.  If we have a platform where a uint64 isn't aligned to 8 bytes
> then we might need the padding.

It's not so much "8 bytes".  The question is this: is there any
platform on which uint64 has less than MAXALIGN alignment
requirement?  If it is maxaligned then the compiler will insert any
required padding automatically, so the patch accomplishes little.

AFAICS that could only happen if "double" has 8-byte alignment
requirement but int64 does not.  I recall some discussion about
that possibility a month or two back, but I think we concluded
that we weren't going to support it.

I guess what I mostly don't like about df0f4feef is the hardwired "8"
constants.  Yeah, it's hard to see how sizeof(uint64) isn't 8, but
it's not very readable like this IMO.

			regards, tom lane

Commits

  1. Harden memory context allocators against bogus chunk pointers.

  2. Improve our ability to detect bogus pointers passed to pfree et al.

  3. Remove MemoryContextContains().

  4. Remove uses of MemoryContextContains in nodeAgg.c and nodeWindowAgg.c.

  5. Temporarily make MemoryContextContains return false

  6. Make MemoryContextContains work correctly again

  7. Make more effort to put a sentinel at the end of allocated memory

  8. Fix some possibly latent bugs in slab.c

  9. Various cleanups of the new memory context header code

  10. Revert "Add missing padding from MemoryChunk struct"

  11. Use MAXALIGN() in calculations using sizeof(SlabBlock)

  12. Add missing padding from MemoryChunk struct

  13. Improve performance of and reduce overheads of memory management