[Bug 599546] Faster read/write macros

[GStreamer (bugzilla.gnome.org)]

https://bugzilla.gnome.org/show_bug.cgi?id=599546
  GStreamer | gstreamer (core) | git

--- Comment #18 from Tim-Philipp Müller <t.i.m at zen.co.uk> 2011-05-31 10:59:16 UTC ---

 - changing existing macros in a way that requires code
   changes elsewhere is an API break, isn't it? Don't think
   we can do that (re. the 'fix code for new macros' patches)

 - adding additional READ_FAST() macros is theoretically
   not a problem of course, but it just feels a bit fail to me.
   It feels like it means we can't be arsed to find a better
   more general solution

 - _FAST() macros are bound to be used in cases where
    they shouldn't be used (or are initially used correctly
    and then something is changed elsewhere and it's not
    correct any longer). Which is going to lead to hard to
    track down bugs (who's got 100% code path coverage?)

 - doing stuff based on HAVE_CPU_XYZ doesn't seem
   right either: this is re. the host system, isn't it?

 - I would still like to know how the inline-function + memcpy()
   approach fares with gcc + MSVC. I seem to remember that
   the memcpy() is equally efficient. Why try to be smarter
   than the compiler? Wouldn't it be better in most cases to
   just let the compiler optimise this?

-- 
Configure bugmail: https://bugzilla.gnome.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the bug.
You are the assignee for the bug.