[FFmpeg-devel] [PATCH] define _DARWIN_C_SOURCE for MAP_ANON

Reimar Döffinger Reimar.Doeffinger
Wed Jul 21 21:24:57 CEST 2010


On Wed, Jul 21, 2010 at 08:19:24PM +0100, M?ns Rullg?rd wrote:
> Reimar D?ffinger <Reimar.Doeffinger at gmx.de> writes:
> 
> > On Wed, Jul 21, 2010 at 11:37:27AM -0700, Alexander Strange wrote:
> >> sys_icache_invalidate doesn't do anything on x86.
> >> x86 (at least Intel and AMD) don't even have an icache in the first
> >> place,
> >
> > That's nonsense.
> >
> >> so they don't need to be flushe and I don't think could test
> >> code to flush one.
> >
> > They have some special hacks to ensure D and I-cache are coherent.
> > Actually they have some even worse hacks, truly horrible things
> > in order to make self-modifying code continue to work (like the
> > debugger-defeating stuff where one instruction write to the one
> > directly following it).
> 
> I didn't know that.  Since the dyncode is x86 only, there is no need
> for a change then.

Yeah, it's one of those really strange things, even I forgot about
it.
It's also a bit more complex, since it seems things like Intel's
trace cache that caches the micro-ops actually breaks this to a
degree, though it shouldn't be in a way that matters to us (still
only a rather small handful of instructions).



More information about the ffmpeg-devel mailing list