Jesse saw the following output. It looks very likely that the images/* reporters and the gfx/surface/image reporter are double-counting some memory. CC'ing some gfx people. Main Process Mapped Memory 3,606,937,600 B (100.0%) -- mapped ├──3,183,395,936 B (88.26%) -- other ├────417,708,960 B (11.58%) -- heap │ ├──366,023,808 B (10.15%) -- used │ └───51,685,152 B (01.43%) -- unused └──────5,832,704 B (00.16%) -- js ├──5,308,416 B (00.15%) -- mjit-code └────524,288 B (00.01%) -- tjit-code Used Heap Memory 366,023,808 B (100.0%) -- heap-used ├──138,558,573 B (37.86%) -- images │ ├──138,237,501 B (37.77%) -- content │ │ ├──135,305,282 B (36.97%) -- used │ │ │ ├──119,908,256 B (32.76%) -- uncompressed │ │ │ └───15,397,026 B (04.21%) -- raw │ │ └────2,932,219 B (00.80%) -- unused │ │ ├──2,594,860 B (00.71%) -- uncompressed │ │ └────337,359 B (00.09%) -- raw │ └──────321,072 B (00.09%) -- chrome │ ├──321,072 B (00.09%) -- used │ │ ├──321,072 B (00.09%) -- uncompressed │ │ └────────0 B (00.00%) -- raw │ └────────0 B (00.00%) -- unused │ ├──0 B (00.00%) -- raw │ └──0 B (00.00%) -- uncompressed ├──122,852,976 B (33.56%) -- gfx │ └──122,852,976 B (33.56%) -- surface │ └──122,852,976 B (33.56%) -- image ├───80,783,240 B (22.07%) -- js │ ├──77,594,624 B (21.20%) -- gc-heap │ ├───2,320,224 B (00.63%) -- tjit-data │ │ └──2,320,224 B (00.63%) -- allocators │ │ ├──1,776,000 B (00.49%) -- reserve │ │ └────544,224 B (00.15%) -- main │ └─────868,392 B (00.24%) -- mjit-data ├───49,715,672 B (13.58%) -- storage │ └──49,715,672 B (13.58%) -- sqlite │ ├──40,946,592 B (11.19%) -- urlclassifier3.sqlite │ │ ├──40,841,408 B (11.16%) -- cache-used │ │ ├─────100,944 B (00.03%) -- stmt-used │ │ └───────4,240 B (00.00%) -- schema-used │ ├───5,191,008 B (01.42%) -- places.sqlite │ │ ├──4,717,368 B (01.29%) -- cache-used │ │ ├────395,280 B (00.11%) -- stmt-used │ │ └─────78,360 B (00.02%) -- schema-used │ ├─────984,128 B (00.27%) -- other │ ├─────733,512 B (00.20%) -- cookies.sqlite │ │ ├──718,040 B (00.20%) -- cache-used │ │ ├───12,528 B (00.00%) -- stmt-used │ │ └────2,944 B (00.00%) -- schema-used │ ├─────613,256 B (00.17%) -- extensions.sqlite │ │ ├──429,848 B (00.12%) -- cache-used │ │ ├──172,192 B (00.05%) -- stmt-used │ │ └───11,216 B (00.00%) -- schema-used │ ├─────341,480 B (00.09%) -- addons.sqlite │ │ ├──264,760 B (00.07%) -- cache-used │ │ ├───69,760 B (00.02%) -- stmt-used │ │ └────6,960 B (00.00%) -- schema-used │ ├─────335,680 B (00.09%) -- chromeappsstore.sqlite │ │ ├──265,144 B (00.07%) -- cache-used │ │ ├───64,184 B (00.02%) -- stmt-used │ │ └────6,352 B (00.00%) -- schema-used │ ├─────265,904 B (00.07%) -- downloads.sqlite │ │ ├──256,312 B (00.07%) -- cache-used │ │ ├────6,768 B (00.00%) -- stmt-used │ │ └────2,824 B (00.00%) -- schema-used │ ├─────206,152 B (00.06%) -- webappsstore.sqlite │ │ ├──135,616 B (00.04%) -- cache-used │ │ ├───64,184 B (00.02%) -- stmt-used │ │ └────6,352 B (00.00%) -- schema-used │ ├──────33,336 B (00.01%) -- formhistory.sqlite │ │ ├──19,832 B (00.01%) -- stmt-used │ │ ├──10,824 B (00.00%) -- cache-used │ │ └───2,680 B (00.00%) -- schema-used │ ├──────21,936 B (00.01%) -- content-prefs.sqlite │ │ ├───9,632 B (00.00%) -- stmt-used │ │ ├───8,280 B (00.00%) -- cache-used │ │ └───4,024 B (00.00%) -- schema-used │ ├──────17,264 B (00.00%) -- signons.sqlite │ │ ├───8,264 B (00.00%) -- cache-used │ │ ├───4,528 B (00.00%) -- stmt-used │ │ └───4,472 B (00.00%) -- schema-used │ ├──────13,296 B (00.00%) -- permissions.sqlite │ │ ├───6,776 B (00.00%) -- stmt-used │ │ ├───4,464 B (00.00%) -- cache-used │ │ └───2,056 B (00.00%) -- schema-used │ └──────12,128 B (00.00%) -- search.sqlite │ ├───5,672 B (00.00%) -- stmt-used │ ├───4,448 B (00.00%) -- cache-used │ └───2,008 B (00.00%) -- schema-used ├────8,640,641 B (02.36%) -- layout │ ├──8,640,641 B (02.36%) -- all │ └──────────0 B (00.00%) -- bidi └──-34,527,294 B (-9.43%) -- other Other Measurements 544,038,912 B -- resident 366,025,952 B -- heap-zone0-committed 413,515,776 B -- heap-zone0-used

Bug 660577 comment 27 has a similar example. The problem shows up on any page that has lots of images.

This is the simple, brute force approach -- just move all the gfx/surface/* reporters out of the "explicit allocations" tree into the "other measurements" tree.

Comment on attachment 536822 [details] [diff] [review] patch This is not a patch :)

Attempt 2!

Attempt 3!

I set the tracking-firefox6? flag because this is a low-risk fix to a bug that was introduced by the about:memory revamp, which was first introduced in Firefox 6.

not going to track. please nominate the patch with a risk assessment.

Comment on attachment 537232 [details] [diff] [review] patch v3 Requesting approval-mozilla-aurora. The benefit of this patch is that it avoids some nonsensical negative numbers in about:memory. The risk is extremely low, it just renames and reclassifies the gfx memory reporters so they are shown in the "Other Measurements" part of about:memory instead of the "Explicit Allocations" part.

Comment on attachment 537232 [details] [diff] [review] patch v3 Giddyup

Setting status-firefox6=fixed per comment 11, so that this will disappear from the "approved but not yet landed" bug list.