On Fri, Jun 22, 2018 at 04:51:09PM +0200, Tomas Härdin wrote: > fre 2018-06-22 klockan 14:07 +0000 skrev Eran Kornblau: > > First, regarding the if you added, it's redundant - if you look a few lines > > above, you'll see 'if (atom.size < atom.header_size)'. > > atom.header_size is either 8 or 16, it can't be anything else, so atom.size > > can't be < 8 at this point. > > If you look closely you'll see that check is after subtracting > atom.header_size. > > > I'll leave it to the maintainers to decide whether this tool is helpful or > > not, IMHO, all these comments make the > > code less readable, and some of the changes make it less efficient. I don't > > think this slight reduction of performance > > matters much in the context of faststart, but in other parts of ffmpeg it > > can be significant. > > > > Few examples to why I think it's less efficient - > > 1. the change of macros to functions - maybe the compiler will inline them, > > but maybe it won't... > > You're assuming inlining actually makes the code faster. It's not the > 80's anymore.
Until someone tests it, both claims are assumtations. The date on the calender surely is not a good argument though even though it has a somewhat "authorative" vibe to it. Personally i prefer inline/always inline functions over macros when they are equally fast though ... Speaking about the date. I would have thought the need to manually annotate the code for analyzers was a thing more for the past and for academia. But quite possibly i have missed something ... If you compare the time it takes a human to anotate a piece of code (some of this anotation itself may be incorrect) and subsequently a analyzer to find bugs. To the alternative of a human manually reviewing the code line by line and a analyzer running over the code which does not need anotations which finds more bugs ? The question matters, because i think we want the maximum return for the time that is invested [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB Asymptotically faster algorithms should always be preferred if you have asymptotical amounts of data
signature.asc
Description: PGP signature
_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
