712835 - Add a memory reporter for the nsAtomTable

Status: RESOLVED FIXED

(Core :: XPCOM, defect)

Description

[Nicholas Nethercote [inactive]]

DMD says it's reasonable prominent:

 Unreported: 1 block(s) in record 9 of 17967
  262,144 bytes (262,144 requested / 0 slop) 
  0.25% of the heap (7.19% cumulative unreported)
    at 0x402A063: malloc (vg_replace_malloc.c:263)
    by 0x403C0A4: moz_malloc (mozalloc.cpp:113)
    by 0x7A4126C: PL_DHashAllocTable (pldhash.cpp:114)
    by 0x7A41AAB: ChangeTable(PLDHashTable*, int) (pldhash.cpp:564)
    by 0x7A41D26: PL_DHashTableOperate (pldhash.cpp:649)
    by 0x7A565D9: GetAtomHashEntry(unsigned short const*, unsigned int) (nsAtomT
able.cpp:403)
    by 0x7A56A0F: NS_NewAtom(nsAString_internal const&) (nsAtomTable.cpp:534)
    by 0x68B4D95: do_GetAtom(nsAString_internal const&) (nsIAtom.h:199)

That's just the hash table storage, which I'm sure can get substantially bigger with more browsing.  There's also the AtomImpls pointed to by the hash table entries, and then the strings hanging off the AtomImpls, but I'll need to look more closely to see if they actually belong to the AtomImpls.

Comment 1

[Nicholas Nethercote [inactive]]

Attachment: patch 1

This patch just adds nsStringBuffer::SizeOfIncludingThisIfUnshared(), which the next patch will use.

Comment 2

[Nicholas Nethercote [inactive]]

Attachment: patch 2

This adds the "explicit/atom-table" memory reporter.  With Gmail and
TechCrunch open it's just over 1MB worth.

I created EnsureTableExists() for an earlier version where I registered the
memory reporter in nsAtomTable.cpp.  That ended up having to change, but I
left EnsureTableExists() in because I think it's an improvement.

Comment 3

[Nicholas Nethercote [inactive]]

Attachment: patch 3

This patch moves AtomImpl and PermanentAtomImpl out of the header, because they don't need to be there.

Comment 4

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

Comment on attachment 583977 [details] [diff] [review]
patch 3

Review of attachment 583977 [details] [diff] [review]:
-----------------------------------------------------------------

Why do we want this?

Comment 5

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

Comment on attachment 583977 [details] [diff] [review]
patch 3

Review of attachment 583977 [details] [diff] [review]:
-----------------------------------------------------------------

Ok, I guess hiding these is fine.

Comment 6

[Nicholas Nethercote [inactive]]

> Why do we want this?

Basic information hiding, that's all.  I just figured I'd fix it while I was in the area.

Comment 7

[Nicholas Nethercote [inactive]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/365161a949b5
https://hg.mozilla.org/integration/mozilla-inbound/rev/4b655c46d071
https://hg.mozilla.org/integration/mozilla-inbound/rev/fa8de405cc87

Comment 8

[Matt Brubeck (:mbrubeck)]

https://hg.mozilla.org/mozilla-central/rev/365161a949b5
https://hg.mozilla.org/mozilla-central/rev/4b655c46d071
https://hg.mozilla.org/mozilla-central/rev/fa8de405cc87