1710884 - Add IPC glue code for self-hosted XDR content [Android]

Status: NEW

(Core :: IPC, task)

Description

[Nicolas B. Pierron [:nbp — off until 29-09]]

Fork of the second part of Bug 1698045.

Comment 1

[Nicolas B. Pierron [:nbp — off until 29-09]]

Attachment: WIP: Bug 1710884 - Use shared memory for JS initialization on Android.

Comment 2

[Nicolas B. Pierron [:nbp — off until 29-09]]

Note to self, the self-hosted shared-memory is optional and can be toggle by a preference in about:config. Due to the not-yet-implemented-highly-needed-cleanup required by Android, we might have to pass a fake fd (-1) even when the content is not present, to satisfy the fact that Android matches fd to their purpose by looking up at specific indexes in the fd-array.

Comment 3

[Chris Peterson [:cpeterson]]

Ted Campbell says we should fix this bug for Android Fission because it will reduce memory usage, probably about 400 kB per process. See bug 1710987 comment 5 for the memory improvements seen on desktop.

Comment 4

[Markus Stange [:mstange]]

Randell, I saw you updated the patch recently - did you push it to try and see if the failures / timeouts are still happening?

Comment 5

[Nicolas B. Pierron [:nbp — off until 29-09]]

I have no intent to push these changes forward, if anybody more knowledgeable about GeckoView development is willing to have a look, feel free to take over the patch.

This change is expected to bring a ~40ms improvement (Bug 1618391) to the start of the content process(-es), by using what the parent process already computed.

Comment 6

[Dave Hunt [:davehunt] [he/him] ⌚BST]

The Performance Impact Calculator has determined this bug's performance impact to be low. If you'd like to request re-triage, you can reset the Performance Impact flag to "?" or needinfo the triage sheriff.

Platforms: Android
Impact on browser: Causes noticeable startup delay

Comment 7

[Randell Jesup [:jesup] (needinfo me)]

I have a copy of this patch locally, but it's low on my personal queue

Comment 8

[[:owlish] 🦉 PST]

Hey Dave, you mentioned that the performance calculator suggests the performance impact of this bug will be low. How does that calculator work? Has it proved to be reliable?

Comment 9

[Dave Hunt [:davehunt] [he/him] ⌚BST]

(In reply to [:owlish] 🦉 PST from comment #8)

Hey Dave, you mentioned that the performance calculator suggests the performance impact of this bug will be low. How does that calculator work? Has it proved to be reliable?

Hey. You can find details on the calculator at https://wiki.mozilla.org/Performance/Triage#Triage_calculator. Since this bug was triaged we have updated the calculator so that Android has the same weight as Windows. This upgrades the impact from low to medium.

The Performance Impact Calculator has determined this bug's performance impact to be medium. If you'd like to request re-triage, you can reset the Performance Impact flag to "?" or needinfo the triage sheriff.

Platforms: Android
Impact on browser: Causes noticeable startup delay

Comment 10

[Dave Hunt [:davehunt] [he/him] ⌚BST]

I should add that the performance triage and impact calculator are more suited to defects than tasks.