949878 - Rooting hazard analysis timeouts

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

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

When h4writer pushed bug 939614 this morning, suddenly the SM(Hf) static rooting analysis became insanely slow. Where before it took 10 minutes to load the callgraph, it started taking over an hour (and getting killed for nonresponsiveness).

I still don't know why that patch triggered it, but the slowness is due to bug 949724.

Comment 1

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

Attachment: Switch to integer indexes to avoid pathological array slowdown

Fortunately, the workaround is trivial. With this patch, bug 939614 completes the analysis run just fine. (The computeGCFunctions phase drops from its usual 10 min to about 1 minute.)

Comment 2

[Terrence Cole [:terrence]]

Comment on attachment 8347026 [details] [diff] [review]
Switch to integer indexes to avoid pathological array slowdown

Review of attachment 8347026 [details] [diff] [review]:
-----------------------------------------------------------------

Great! r=me

This is a great real-world stress test for our array machinery. Could you file a followup bug to make this workload into a benchmark?

Comment 3

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

(In reply to Terrence Cole [:terrence] from comment #2)
> This is a great real-world stress test for our array machinery. Could you
> file a followup bug to make this workload into a benchmark?

That's bug 949724.

Comment 4

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

http://hg.mozilla.org/integration/mozilla-inbound/rev/52a6e01a91d9

Comment 5

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/52a6e01a91d9