home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.amiga.programmer
- Path: chang.unx.sas.com!walker
- From: walker@chang.unx.sas.com (Doug Walker)
- Subject: Re: Is a real good mpeg player possible on the amiga ?
- Sender: news@unx.sas.com (Noter of Newsworthy Events)
- Message-ID: <DM5HrH.IA4@unx.sas.com>
- Date: Fri, 2 Feb 1996 13:35:41 GMT
- X-Nntp-Posting-Host: chang.unx.sas.com
- References: <4e2gqu$5ca@caers3.unicaen.fr> <DLqtLz.FFL@unx.sas.com> <4eedlr$pte@maureen.teleport.com> <4efhj3$p7s@serpens.rhein.de>
- Organization: SAS Institute Inc.
-
- In article <4efhj3$p7s@serpens.rhein.de>,
- Michael van Elst <mlelstv@serpens.rhein.de> wrote:
- >sschaem@teleport.com (Stephan Schaem) writes:
- >
- >> support all this correctly. YUV do not take the major amount
- >> of time, and could be overlap with write to the buffer.
- >
- >Actually it takes quite some time because you have to handle
- >pixels byte by byte. Overlaps to the buffer are often not possible
- >(there is no write buffer in Buster) and a separate copy pass
- >is often _faster_.
- >
- >> I played with a PC and the 8bit dither mode was the fastest...
- >
- >It is possible that they reduce the quality of the decoding to
- >get some more speed.
- >
-
- I'm sure they do - in fact, NO reasonable-speed software player
- that I am aware of handles MPEG motion vectors perfectly. To
- fully handle motion vectors, you need to handle half-pel
- increments, which in the worst case would force you to
- average the values of four pels in the past reference frame
- and four pels in the future reference frame (assuming motion
- vectors in both X and Y that were not an even number of pels
- and both future and past references) for each pel. This is
- simply unreasonable to expect in software. Many players
- average only two values for each of the past and future
- reference vectors, the diagonally- opposite ones; others
- probably round off the half-pel and take whatever comes up.
- There are lots of other compromises that are typically made
- for speed.
-
- It is also probable that they spent a lot more time optimizing for
- 8-bit dither mode. 16-bit mode is a lot faster in our MPEG player,
- assuming you have a Pentium. I don't know what is faster on a 486.
-
- Remember that when you test a software player, you are testing the
- hardware it's running on, the software itself, and the codec it is
- using all at the same time. Don't leap to conclusions about the
- codec when the problem may be in the implementation.
-
- --
- ***** / walker@unx.sas.com
- *|_o_o|\\ Doug Walker< BIX, Portal: djwalker
- *|. o.| || \ AOL: weissblau
- | o |//
- ======
- Any opinions expressed are mine, not those of SAS Institute, Inc.
-
-