1171612 - Use C++11 features to make Statistics module nicer

Status: RESOLVED FIXED

(Core :: JavaScript: GC, defect)

Description

[Terrence Cole [:terrence]]

Attachment: stats_use_cpp11-v0.diff

This is mostly centered around MAX_MULTIPARENT_PHASES. First, the nicer |using name = type| syntax allows us to easily extract the const from the PhaseTimeTable, allowing us to use replace even the declarations. Then we always used MAX_MULTIPARENT_PHASES with +1, so I just moved that into the def. Then we were not being consistent with iterating with size_t vs. int, so I made these use auto + MakeRange so that we get the right type for free.

Comment 1

[Steve Fink [:sfink] [:s:]]

Comment on attachment 8615484 [details] [diff] [review]
stats_use_cpp11-v0.diff

Review of attachment 8615484 [details] [diff] [review]:
-----------------------------------------------------------------

So, the patch is good but I'm hung up on the naming. Bumping MAX_MULTIPARENT_PHASES by 1 invalidates the meaning. On the other hand, the previous meaning -- while accurate -- was not how it was being used, so your patch plus a new name would be much better.

The weird thing is that index 0 is used for regular, single-parent phases. Or rather, the timings stored at index 0 correspond to the subset of the tree that never pass through a multi-parent phase. (The difference is that a multi-parent phase could have a single-parent child, and that child's timing would be stored at a nonzero index.)

NUM_DAG_INDEXES would be accurate for the new MAX_MULTIPARENT_PHASES+1 value, but it's pretty cryptic. NUM_TIMING_ARRAYS, maybe?

Comment 2

[Terrence Cole [:terrence]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=93203e3e4b9f

Comment 3

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/53d75ce7f7fd

Comment 4

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

https://hg.mozilla.org/mozilla-central/rev/53d75ce7f7fd