1348821 - Change CacheIR spewer to log IC mode

Status: RESOLVED FIXED

(Core :: JavaScript Engine: JIT, enhancement, P3)

Description

[Jan de Mooij [:jandem]]

Bug 1328140 improves our IC heuristics in a number of ways. We should make sure the spewer still gives us useful information, for instance when we stop trying to attach stubs we should probably still log the fallback hits so we know which ICs are hot.

Comment 1

[Tom S [:evilpie]]

Attachment: For MegamorphicNativeSlot differentiate between HandleMissing

I am going to work on a few things that would be useful to have now.

Comment 2

[Tom S [:evilpie]]

Attachment: Log IC mode

Comment 3

[Jan de Mooij [:jandem]]

Comment on attachment 8849526 [details] [diff] [review]
For MegamorphicNativeSlot differentiate between HandleMissing

Review of attachment 8849526 [details] [diff] [review]:
-----------------------------------------------------------------

Thanks.

Comment 4

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/e7a7282cf4f8
For MegamorphicNativeSlot differentiate between HandleMissing. r=jandem
https://hg.mozilla.org/integration/mozilla-inbound/rev/30c3b52e7892
Log CacheIR IC mode. r=jandem

Comment 5

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/e7a7282cf4f8
https://hg.mozilla.org/mozilla-central/rev/30c3b52e7892

Comment 6

[Tom S [:evilpie]]

Attachment: Log address and shape of objects

Comment 7

[Jan de Mooij [:jandem]]

Comment on attachment 8851280 [details] [diff] [review]
Log address and shape of objects

Review of attachment 8851280 [details] [diff] [review]:
-----------------------------------------------------------------

Nice. Would it be useful to log v.toObject().getClass()->name too?

Comment 8

[Tom S [:evilpie]]

Thanks for the review. We already log the class-name as type, because we are using InformalValueTypeName for that.

Comment 9

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/bbd6c0e8298e
Log address and shape of objects in CacheIR. r=jandem

Comment 10

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/99bc59df7d85
probably fix unified build failure. a=bustage

Comment 11

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/32fd9c7b21ad
Fix it. IGNORE CLOSED TREE

Comment 12

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

https://hg.mozilla.org/mozilla-central/rev/bbd6c0e8298e
https://hg.mozilla.org/mozilla-central/rev/99bc59df7d85
https://hg.mozilla.org/mozilla-central/rev/32fd9c7b21ad

Comment 13

[Tom S [:evilpie]]

To log fallback hits we would add a else block to all the canAttachStub conditions. However I would like to reuse trackNotAttached and the IRGenerators are only created inside the if block.

Comment 14

[BMO Automation]

Per policy at https://wiki.mozilla.org/Bug_Triage/Projects/Bug_Handling/Bug_Husbandry#Inactive_Bugs. If this bug is not an enhancement request or a bug not present in a supported release of Firefox, then it may be reopened.