Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

Why there is no "javascript.options.ion.chrome" option?

202 views
Skip to first unread message

Tetsuharu OHZEKI [:saneyuki_s]

unread,
Apr 7, 2013, 11:23:59 AM4/7/13
to dev-tech-...@lists.mozilla.org
Hello.

I have a question.
Why do we have no "javascript.options.ion.chrome" option? Other jit options
has each options like "javascript.options.*.chrome" and
"javascript.options.*.chrome".

cheers.

--
Tetsuharu OHZEKI (a.k.a. saneyuki_s)
saneyuk...@gmail.com

Jeff Walden

unread,
Apr 13, 2013, 10:48:44 PM4/13/13
to
On 04/07/2013 08:23 AM, Tetsuharu OHZEKI [:saneyuki_s] wrote:
> Why do we have no "javascript.options.ion.chrome" option? Other jit options
> has each options like "javascript.options.*.chrome" and
> "javascript.options.*.chrome".

Probably no one went to the trouble to add it, but I don't know for sure. (Which, as far as I'm concerned, is a good thing. People shouldn't need to change engine options to get optimal performance: we should be fast by default. And to the extent someone thinks one mode or another is truly faster for their workload, 1) that's not going to be true forever, and 2) workloads change over time, so unless you're constantly testing you're eventually going to choose the wrong options.)

Jeff

Nicolas B. Pierron

unread,
Apr 13, 2013, 11:07:56 PM4/13/13
to
On 04/07/2013 08:23 AM, Tetsuharu OHZEKI [:saneyuki_s] wrote:
> Why do we have no "javascript.options.ion.chrome" option? Other jit options
> has each options like "javascript.options.*.chrome" and
> "javascript.options.*.chrome".

Because IonMonkey has been optimized for features which are only available
in the content without any special version mention to the script tag.

For example, there is no support for let in IonMonkey, even if this should
not be hard to support. So just that would cause most function to switch
back-and-forth between JITs, which was a big concern with JägerMonkey. This
should no longer be a big concern with the Baseline Compiler.

--
Nicolas B. Pierron

Boris Zbarsky

unread,
Apr 14, 2013, 10:35:28 AM4/14/13
to
On 4/13/13 11:07 PM, Nicolas B. Pierron wrote:
> For example, there is no support for let in IonMonkey, even if this
> should not be hard to support. So just that would cause most function
> to switch back-and-forth between JITs

Note that the ion.content switch just controls ion in both chrome and
content, as far as I can tell. As in, ion is enabled in chrome.

-Boris

Nicolas B. Pierron

unread,
Apr 14, 2013, 8:30:58 PM4/14/13
to
Hum … I don't know if this was intended then.
I don't remember seeing any trace of Ion compilation of the chrome so far.

--
Nicolas B. Pierron

0 new messages