503461 - TM: inline FinalizeObject into jsgc and make sure Scope hold() and drop() are inline too

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Andreas Gal :gal]

Quite noticable speedup for the gc path.

Comment 1

[Andreas Gal :gal]

Attachment: patch

Comment 2

[David Mandelin [:dmandelin]]

Comment on attachment 387816 [details] [diff] [review]
patch

I'm not super familiar with these finalizers but from what I see you simply moved some functions from one file to another so they can be inlined at the point where they are called the most, so it looks good to me.

Comment 3

[Andreas Gal :gal]

http://hg.mozilla.org/tracemonkey/rev/81af09df8033

Comment 4

[Igor Bukanov]

Comment on attachment 387816 [details] [diff] [review]
patch

>+static void
>+FinalizeObject(JSContext *cx, JSObject *obj)
>+{
>+    /* Cope with stillborn objects that have no map. */
>+    if (!obj->map)
>+        return;
>+
>+    if (JS_UNLIKELY(cx->debugHooks->objectHook != NULL)) {
>+        cx->debugHooks->objectHook(cx, obj, JS_FALSE,
>+                                   cx->debugHooks->objectHookData);
>+    }
>+
>+    /* Finalize obj first, in case it needs map and slots. */
>+    STOBJ_GET_CLASS(obj)->finalize(cx, obj);

For the default JS objects and arrays we are calling here the empty JS_FinalizeStub method. My guess is that avoiding calling this method via simple check can make the finalization faster. But this can go to a separated bug.

Comment 5

[Robert Sayre]

http://hg.mozilla.org/mozilla-central/rev/81af09df8033