Closed Bug 661770 Opened 12 years ago Closed 12 years ago

TM: inline Atomize into js_AtomizeChars


(Core :: JavaScript Engine, defect)

Not set





(Reporter: cdleary, Assigned: cdleary)


(Whiteboard: fixed-in-tracemonkey)


(3 files)

This Atomize interface is super hot through getTokenInternal, and valgrind likes the results on parsemark for x86.
Attachment #537104 - Flags: review?(nnethercote)
I'm hoping to get this one as a ride-along, since I use it to measure the valgrind diff results. A bunch of things change about the parse routine for compile-and-go, one example being use of the globalMap to understand global index slots for property accesses. Since parsing content works this way, I think the shell should work this way as well. Maybe we could make it an option later if we find that necessary (and maybe let users pass a JSObject to use as a fake global?), but I think the default being compile and go is more reflective of what we'd like to optimize.
Attachment #537107 - Flags: review?(nnethercote)
Comment on attachment 537104 [details] [diff] [review]
Inline Atomize into js_AtomizeChars.

Review of attachment 537104 [details] [diff] [review]:
Attachment #537104 - Flags: review?(nnethercote) → review+
Comment on attachment 537107 [details] [diff] [review]
Make shell's |compile| function compile-and-go.

Review of attachment 537107 [details] [diff] [review]:

I don't understand compile-and-go well enough to review this.  Maybe Luke can review?
Attachment #537107 - Flags: review?(nnethercote) → review?(luke)
Attachment #537107 - Flags: review?(luke) → review?(dvander)
Attachment #537107 - Flags: review?(dvander) → review+
You need to log in before you can comment on or make changes to this bug.