On Thu, Jun 30, 2011 at 10:36 AM, Ronald S. Bultje <[email protected]> wrote: > Hi, > > On Thu, Jun 30, 2011 at 10:29 AM, Jason Garrett-Glaser <[email protected]> wrote: >> On Thu, Jun 30, 2011 at 10:22 AM, Ronald S. Bultje <[email protected]> >> wrote: >>> On Wed, Jun 29, 2011 at 3:51 PM, Jason Garrett-Glaser <[email protected]> >>> wrote: >>>> --- >>>> libavcodec/h264.h | 247 >>>> +++++++++++++++++++++++++++-------------------------- >>>> 1 files changed, 124 insertions(+), 123 deletions(-) >>> >>> The commit msg here could be adapted a little to say that you prevent >>> gcc from reloading h->xx repeatedly (which is what this patch does, >>> afaics). >>> >>> Patch itself is OK. I wonder why gcc does this (reloading h->xx) for >>> you, I don't think it does for me... New "optimization" in a later >>> version of gcc? >> >> It does because the C standard requires it, basically. > > So why doesn't it do it for things like s->current_picture.xx? Because > it's marked const and no other pointer of that type exists (thus > forcing restrict)?
Probably. Also consider the rule that "uint8_t *t can alias anything". Jason _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
