1307768 - Regression from Sep 22 2016 caused Memory tool to display asm.js heap area with the whole virtually allocated size instead of the actual allocated size in JavaScript.

Status: RESOLVED FIXED

(DevTools :: Memory, defect, P2)

Description

[Jukka Jylänki]

Attachment: hello_world.tar.gz

STR:

1. Open hello_world.html
2. Open Devtools Memory tab and do a snapshot.

Observed:

The resulting snapshot shows a 6GB memory allocation for the asm.js heap.

Expected:

The actual heap size allocated in JS code is 16MB, the memory tool is reporting the 64bit virtually reserved guardband area in the size.

Mozregression points the cause at https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=89cf1ca5cf6c88f356edbc246521cdf0853411c2&tochange=1e3b7a0422b6533d5aca29d92fec1e3fd31149ec

Comment 1

[Benjamin Bouvier [:bbouvier] (inactive)]

Thanks for the regression range!

Comment 2

[Benjamin Bouvier [:bbouvier] (inactive)]

Attachment: memory-reporting.patch

Using the macro that defines all the fields has a few drawbacks; in particular, in ClassInfo::sizeOfAllThings, all the fields member get added altogether:

http://searchfox.org/mozilla-central/source/js/public/MemoryMetrics.h#191

and this gets called by the memory reporter devtool through this:

http://searchfox.org/mozilla-central/source/js/src/jsobj.cpp#3826

I am not sure to understand all the consequences of keeping the wasmGuardPageSize in this set of fields, so I think it might be simpler and better to just put it outside of the meta-macro. There is special handling of this field already for the about:memory reporting, which I assume is the most important thing we want to be able to keep track of, for the guard page sizes.

I've considered making other macros, like ADD_SIZE_TO_N_IF_NOT_IGNORE (using the servoKind Ignore), but it was unclear in which members this should be used or not.

Verified that with this patch:
- the wasm vsize doesn't appear anymore in the memory reporter devtool
- about:memory still shows the wasm vsize at the bottom end.

Comment 3

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/2ef2b3f6e7fa
Don't include wasmGuardPages with the rest of the ClassInfo fields; r=njn

Comment 4

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/2ef2b3f6e7fa