gecko/js
Rafael Ávila de Espíndola 74a6dda7b8 Bug 771683 - Mark some functions MOZ_ALWAYS_INILINE. r=bzbarsky.
When compiling with gcc 4.2 these functions get inlined, but then
again, gcc 4.2 inlines almost anything.

With clang, newer versions of gcc and msvc marking these functions
with MOZ_ALWAYS_INLINE is a win. I assume part of the perf win
gcc 4.5 and msvc get with pgo is from noticing these are hot
and inlining them.

Without PGO it is hard for the compiler to know that this is a
win. With clang this does increase the codesize by about 6%.

The wins from inlining are
* Stack variables (rv in particular) which have their address taken
and are passed as outparams to these functions can go in registers.
* We find that nothing set mCcxToDestroy and ~XPCLazyCallContext
(which is already inlined) can be simplified.
2012-07-06 21:18:17 -04:00
..
ductwork/debugger Bug 758992 - Make the classes which use the XPCOM nsISupports implementation macros final, to avoid the warning about deleting using a pointer to a base class with virtual functions and no virtual dtor (more jsd parts); r=Waldo 2012-06-19 23:43:15 -04:00
examples Bug 716478 - update licence to MPL 2. 2012-05-21 12:12:37 +01:00
ipc Bug 770759 - Add mutable handles (r=bhackett) 2012-07-04 11:12:16 -07:00
jsd Bug 755186 - s/JS_NewCompartmentAndGlobalObject/JS_NewGlobalObject/ (r=jorendorff) 2012-06-20 08:57:29 -07:00
public Bug 687724 (part 3) - Report JS memory consumption for compartments that are associated with |window| objects under "window-objects". r=bholley,jlebar,luke. 2012-07-05 21:12:37 -07:00
src Bug 770407 - Remove overzealous assert in CrossCompartmentWrapper::nativeCall (r=waldo) 2012-07-03 18:51:02 -07:00
xpconnect Bug 771683 - Mark some functions MOZ_ALWAYS_INILINE. r=bzbarsky. 2012-07-06 21:18:17 -04:00