Recherche avancée

Sur d’autres sites (5180)

• Understanding The Dreamcast GD-ROM Layout

  24 mars 2022, par Multimedia Mike — Sega Dreamcast

  I’m finally completing something I set out to comprehend over a decade ago. I wanted to understand how data is actually laid out on a Sega Dreamcast GD-ROM drive. I’m trying to remember why I even still care. There was something about how I wanted to make sure the contents of a set of Dreamcast demo discs was archived for study.

  
  
  

  I eventually figured it out. Read on, if you are interested in the technical details. Or, if you would like to examine the fruits of this effort, check out the Dreamcast demo discs that I took apart and uploaded to the Internet Archive.

  If you care to read some geeky technical details of some of the artifacts on these sampler discs, check out this followup post on Dreamcast Finds.

  Motivation
  Why do I still care about this ? Well, see the original charter of this blog above. It’s mostly about studying multimedia formats, as well as the general operation of games and their non-multimedia data formats. It’s also something that has nagged at me ever since I extracted a bunch of Dreamcast discs years ago and tried to understand why the tracks were arranged the way they were, and how I could systematically split the files out of the filesystem. This turns out not to be as easy as it might sound, even if you can get past the obstacle of getting at the raw data.
  

  CD/CD-ROM Refresher
  As I laid out in my Grand Unified Theory of Compact Disc, every compact disc can be viewed conceptually as a string of sectors, where each sector is 2352 bytes long. The difference among the various CD types (audio CDs, various CD-ROM types) boils down to the format of contents of the 2352-byte sectors. For an audio CD, every sector’s 2352 bytes represents 1/75 of a second of CD-quality audio samples.

  Meanwhile, there are various sector layouts for different CD-ROM modes, useful for storing computer data. This post is most interested in “mode 1/form 1”, which uses 2048 of the 2352 bytes for data, while using the remaining bytes for error detection and correction codes. A filesystem (usually ISO-9660) is overlaid on these 2048-byte sectors in order to create data structures for organizing strings of sectors into files.

  A CD has between 1 and 99 tracks. A pure CD-ROM will have a single data track. Pure audio CDs tend to have numerous audio tracks, usually 1 per song. Mixed CDs are common. For software, this usually manifests as the first track being data and containing an ISO-9660 filesystem, followed by a series of audio tracks, sometimes for in-game music. For audio CDs, there is occasionally a data track at the end of the disc with some extra media types.

  GD-ROM Refresher
  The Dreamcast used optical discs called GD-ROMs, where the GD stands for “gigadisc”. These discs were designed to hold about 1 gigabyte of data, vs. the usual 650-700MB offered by standard CD solutions, while using the same laser unit as is used for CDs. I’m not sure how it achieved this exactly. I always assumed it was some sort of “double density” sector scheme. According to Wikipedia, the drive read the disc at a slower rate which allowed it to read more data (presumably the “pits” vs. “lands” which comprise the surface of an optical disc). This might be equivalent to my theory.

  The GD-ROM discs cannot be read in a standard optical drive. It is necessary to get custom software onto the Dreamcast which will ask the optical hardware to extract the sectors and exfiltrate them off of the unit somehow. There are numerous methods for this. Alternatively, just find rips that are increasingly plentiful around the internet. However, just because you might be able to find the data for a given disc does not mean that you can easily explore the contents.

  Typical Layout Patterns
  Going back to my study of the GD-ROM track layouts, 2 clear patterns emerge :

  All of the game data is packed into track 3 :

  
  
  

  Track 3 has data, the last track has data, and the tracks in between contain standard CD audio :

  
  
  

  Also, the disc is always, always 100% utilized.

  Track 1 always contains an ISO-9660 filesystem and can be read by any standard CD-ROM drive. And it usually has nothing interesting. Track 3 also contains what appears to be an ISO-9660 filesystem. However, if you have a rip of the track and try to mount the image with standard tools, it will not work. In the second layout, the data follows no obvious format.

  Cracking The Filesystem Code
  I figured out quite a few years ago that in the case of the consolidated data track 3, that’s simply a standard ISO-9660 filesystem that would work fine with standard ISO-9660 reading software… if the data track were located beginning at sector 45000. The filesystem data structures contain references to absolute sector numbers. Thus, if it were possible to modify some ISO-9660 software to assume the first sector is 45000, it ought to have no trouble interpreting the data.

  
  
  

  How about the split data track format ? Actually, it works the same way. If all the data were sitting on its original disc, track 3 would have data structures pointing to strings of contiguous sectors (extents) in the final track, and those are the files.

  To express more succinctly : track 3 contains the filesystem root structure and the directory structures, while the final track contains the actual file data. How is the filesystem always 100% full ? Track 3 gets padded out with 0-sectors until the beginning of any audio sectors.

  
  
  

  Why Lay Things Out Like This ?
  Why push the data as far out on the disc as possible ? A reasonable explanation for this would be for read performance. Compact discs operate on Constant Linear Velocity (CLV), vs. Constant Angular Velocity (CAV). The implication of this is that data on the outside of the disc is read faster than data on the inside. I once profiled this characteristic in order to prove it to myself, using both PC CD drives as well as a Dreamcast. By pushing the data to the outer sectors, graphical data gets loaded into RAM faster, and full motion videos, which require a certain minimum bitrate for a good experience, have a better guarantee that playback will be smooth.

  Implications For Repacking
  Once people figured out how to boot burned CDs in the Dreamcast, they had a new problem : Squeeze as much as 1 gigabyte down to around 650 megabytes at the most. It looks like the most straightforward strategy was to simply rework the filesystem to remove the often enormous amount of empty space in track 3.

  My understanding is that another major strategy is to re-encode certain large assets. Full motion video (FMV) assets are a good target here since the prevailing FMV middleware format used on Sega Dreamcast games was Sofdec, which is basically just MPEG-1 video. There is ample opportunity to transcode these files to lower bitrate settings to squeeze some bits (and a lot of visual quality) out of them.

  Further, if you don’t really care about the audio tracks, you could just replace them with brief spurts of silence.

  Making A Tool
  So I could make a tool that would process these collections of files representing a disc. I could also adapt it for various forms that a Dreamcast rip might take (I have found at least 3 so far). I could eventually expand it to handle lots of other disc formats (you know, something like Aaru does these days). And that would have been my modus operandi perhaps 10 or more years ago. And of course, the ambitious tool would have never seen daylight as I got distracted by other ideas.

  I wanted to get a solution up and running as quickly as possible this time. Here was my initial brainstorm : assemble all the tracks into a single, large disc while pretending the audio tracks consist of 2048-byte sectors. In doing so, I ought to be able to use fuseiso to mount the giant image, with a modification to look for the starting sector at a somewhat nonstandard location.

  To achieve the first part I wrote a quick Python script that processed the contents of a GDI file, which was stored alongside the ISO (data) and RAW (audio) track track rips from when I extracted the disc. The GDI is a very matter-of-fact listing of the tracks and their properties, e.g. :

  5
  1 0 4 2048 track01.iso 0
  2 721 0 2352 track02.raw 0
  3 45000 4 2048 track03.iso 0
  4 338449 0 2352 track04.raw 0
  5 349096 4 2048 track05.iso 0
  

  track number / starting sector / track type (4=data, 0=audio) / bytes per sector / filename / ??

  The script skips the first 2 filenames, instead writing 45000 zero sectors in order to simulate the CD-compatible area. Then, for each file, if it’s an ISO, append the data to the final data file ; if it’s audio, compute the number of sectors occupied, and then append that number of 2048-byte zero sectors to the final data file.

  Finally, to interpret the filesystem, I used an old tool that I’ve relied upon for a long time– fuseiso. This is a program that leverages Filesystem in Userspace (FUSE) to mount ISO-9660 filesystems as part of the local filesystem, without needing root privileges. The original source hasn’t been updated for 15 years, but I found a repo that attempts to modernize it slightly. I forked a version which fixes a few build issues.

  Anyway, I just had to update a table to ask it to start looking for the root ISO-9660 filesystem at a different location than normal. Suddenly, after so many years, I was able to freely browse a GD-ROM filesystem directly under Linux !

  Conclusion And Next Steps
  I had to hack the fuseiso3 tool a bit in order to make this work. I don’t think it’s especially valuable to make sure anyone can run with the same modifications since the tool assumes that a GD-ROM rip has been processed through the exact pipeline I described above.

  I have uploaded all of the North American Dreamcast demo discs to archive.org. See this post for a more granular breakdown of what this entails. In the course of this exercise, I also found some European demo discs that could use the same extraction.

  What else ? Should I perform the same extraction experiment for all known Dreamcast games ? Would anyone care ? Maybe if there’s a demand for it.

  Here is a followup on the interesting and weird things I have found on these discs so far.

  The post Understanding The Dreamcast GD-ROM Layout first appeared on Breaking Eggs And Making Omelettes.

• H.264 and VP8 for still image coding : WebP ?

  1er octobre 2010, par Dark Shikari — H.264, VP8, google, psychovisual optimizations

  JPEG is a very old lossy image format. By today’s standards, it’s awful compression-wise : practically every video format since the days of MPEG-2 has been able to tie or beat JPEG at its own game. The reasons people haven’t switched to something more modern practically always boil down to a simple one — it’s just not worth the hassle. Even if JPEG can be beaten by a factor of 2, convincing the entire world to change image formats after 20 years is nigh impossible. Furthermore, JPEG is fast, simple, and practically guaranteed to be free of any intellectual property worries. It’s been tried before : JPEG-2000 first, then Microsoft’s JPEG XR, both tried to unseat JPEG. Neither got much of anywhere.

  Now Google is trying to dump yet another image format on us, “WebP”. But really, it’s just a VP8 intra frame. There are some obvious practical problems with this new image format in comparison to JPEG ; it doesn’t even support all of JPEG’s features, let alone many of the much-wanted features JPEG was missing (alpha channel support, lossless support). It only supports 4:2:0 chroma subsampling, while JPEG can handle 4:2:2 and 4:4:4. Google doesn’t seem interested in adding any of these features either.

  But let’s get to the meat and see how these encoders stack up on compressing still images. As I explained in my original analysis, VP8 has the advantage of H.264′s intra prediction, which is one of the primary reasons why H.264 has such an advantage in intra compression. It only has i4x4 and i16x16 modes, not i8x8, so it’s not quite as fancy as H.264′s, but it comes close.

  The test files are all around 155KB ; download them for the exact filesizes. For all three, I did a binary search of quality levels to get the file sizes close. For x264, I encoded with --tune stillimage --preset placebo. For libvpx, I encoded with --best. For JPEG, I encoded with ffmpeg, then applied jpgcrush, a lossless jpeg compressor. I suspect there are better JPEG encoders out there than ffmpeg ; if you have one, feel free to test it and post the results. The source image is the 200th frame of Parkjoy, from derf’s page (fun fact : this video was shot here ! More info on the video here.).

  Files : (x264 [154KB], vp8 [155KB], jpg [156KB])

  Results (decoded to PNG) : (x264, vp8, jpg)

  This seems rather embarrassing for libvpx. Personally I think VP8 looks by far the worst of the bunch, despite JPEG’s blocking. What’s going on here ? VP8 certainly has better entropy coding than JPEG does (by far !). It has better intra prediction (JPEG has just DC prediction). How could VP8 look worse ? Let’s investigate.

  VP8 uses a 4×4 transform, which tends to blur and lose more detail than JPEG’s 8×8 transform. But that alone certainly isn’t enough to create such a dramatic difference. Let’s investigate a hypothesis — that the problem is that libvpx is optimizing for PSNR and ignoring psychovisual considerations when encoding the image… I’ll encode with --tune psnr --preset placebo in x264, turning off all psy optimizations. 

  Files : (x264, optimized for PSNR [154KB]) [Note for the technical people : because adaptive quantization is off, to get the filesize on target I had to use a CQM here.]

  Results (decoded to PNG) : (x264, optimized for PSNR)

  What a blur ! Only somewhat better than VP8, and still worse than JPEG. And that’s using the same encoder and the same level of analysis — the only thing done differently is dropping the psy optimizations. Thus we come back to the conclusion I’ve made over and over on this blog — the encoder matters more than the video format, and good psy optimizations are more important than anything else for compression. libvpx, a much more powerful encoder than ffmpeg’s jpeg encoder, loses because it tries too hard to optimize for PSNR.

  These results raise an obvious question — is Google nuts ? I could understand the push for “WebP” if it was better than JPEG. And sure, technically as a file format it is, and an encoder could be made for it that’s better than JPEG. But note the word “could”. Why announce it now when libvpx is still such an awful encoder ? You’d have to be nuts to try to replace JPEG with this blurry mess as-is. Now, I don’t expect libvpx to be able to compete with x264, the best encoder in the world — but surely it should be able to beat an image format released in 1992 ?

  Earth to Google : make the encoder good first, then promote it as better than the alternatives. The reverse doesn’t work quite as well.

  [155KB]

• FFMPEG hwaccel convert WEBM to MP4 video failed : could not find codec parameters

  1er juillet 2020, par Yen Nguyen

  I use FFMPEG to convert webm videos into mp4 format using a Nvidia GPU GeoForce GTX 1660 card on Windows 10. It has worked perfectly since the beginning for few months but now it's stopped working with strange error as below :

  



  [h264 @ 0000013d8b5eee00] missing picture in access unit with size 431
[extract_extradata @ 0000013d8b6aef40] No start code is found.
big-buck-bunny-trailer.webm: could not find codec parameters


  



  FFMPEG version and Nvidia driver version are the same as before it just crashed overnight without a clue. I tried Googling around and did update to the latest ffmpeg version as well Nvidia driver version but it still doesn't help.
If I switch to CPU decoder (without hwaccel arg) using libx264 it just works fine.

  



  Environment Windows 10 64bit | ffmpeg version git-2020-04-15-51db0a4 (latest) | Nvidia GeForce GTX 1660, 
 - Geoforce Game Ready Driver version:445.87 Release Date Apr 15, 2020

  



   ffmpeg -y -vsync 0 -hwaccel cuvid -c:v h264_cuvid -i big-buck-bunny-trailer.webm -c:v h264_nvenc output.mp4

ffmpeg version git-2020-04-15-51db0a4 Copyright (c) 2000-2020 the FFmpeg developers
  built with gcc 9.3.1 (GCC) 20200328
  configuration: --enable-gpl --enable-version3 --enable-sdl2 --enable-fontconfig --enable-gnutls --enable-iconv --enable-libass --enable-libdav1d --enable-libbluray --enable-libfreetype --enable-libmp3lame --enable-libopencore-amrnb --enable-libopencore-amrwb --enable-libopenjpeg --enable-libopus --enable-libshine --enable-libsnappy --enable-libsoxr --enable-libsrt --enable-libtheora --enable-libtwolame --enable-libvpx --enable-libwavpack --enable-libwebp --enable-libx264 --enable-libx265 --enable-libxml2 --enable-libzimg --enable-lzma --enable-zlib --enable-gmp --enable-libvidstab --enable-libvmaf --enable-libvorbis --enable-libvo-amrwbenc --enable-libmysofa --enable-libspeex --enable-libxvid --enable-libaom --disable-w32threads --enable-libmfx --enable-ffnvcodec --enable-cuda-llvm --enable-cuvid --enable-d3d11va --enable-nvenc --enable-nvdec --enable-dxva2 --enable-avisynth --enable-libopenmpt --enable-amf
  libavutil      56. 42.102 / 56. 42.102
  libavcodec     58. 78.102 / 58. 78.102
  libavformat    58. 42.100 / 58. 42.100
  libavdevice    58.  9.103 / 58.  9.103
  libavfilter     7. 77.101 /  7. 77.101
  libswscale      5.  6.101 /  5.  6.101
  libswresample   3.  6.100 /  3.  6.100
  libpostproc    55.  6.100 / 55.  6.100
[h264 @ 0000013d8b5eee00] missing picture in access unit with size 431
[extract_extradata @ 0000013d8b6aef40] No start code is found.
big-buck-bunny-trailer.webm: could not find codec parameters
WARNING: defaulting hwaccel_output_format to cuda for compatibility with old commandlines. This behaviour is DEPRECATED and will be removed in the future. Please explicitly set "-hwaccel_output_format cuda".
Guessed Channel Layout for Input Stream #0.1 : mono
Input #0, matroska,webm, from 'big-buck-bunny-trailer.webm':
  Metadata:
    encoder         : http://sourceforge.net/projects/yamka
    creation_time   : 2010-05-20T08:21:12.000000Z
  Duration: 00:00:32.48, bitrate: N/A
    Stream #0:0(eng): Video: h264, none, 640x360, SAR 1:1 DAR 16:9, 25 fps, 25 tbr, 1k tbn (default)
    Stream #0:1(eng): Audio: vorbis, 44100 Hz, mono (default)
Stream mapping:
  Stream #0:0 -> #0:0 (h264 (h264_cuvid) -> h264 (h264_nvenc))
  Stream #0:1 -> #0:1 (vorbis (native) -> aac (native))
Press [q] to stop, [?] for help
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 66
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1339
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1415
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1908
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2123
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2504
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2294
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2684
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2468
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2622
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2867
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2737
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 3017
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2556
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 7811
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1880
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1722
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2309
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1215
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2080
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1775
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2247
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1782
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1945
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1529
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1807
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2015
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1568
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2159
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1861
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2397
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1627
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2364
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1821
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2282
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2068
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1699
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2129
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2268
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2197
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1981
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2169
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2888
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1600
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2958
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1518
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2738
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2319
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2421
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2841
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 1949
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2422
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2288
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2559
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2332
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2403
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2257
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2573
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2320
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2231
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2408
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2415
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2194
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2697
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2480
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2256
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2245
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2519
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2203
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2332
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2341
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2439
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2201
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2882
[NULL @ 0000013d8b5eee00] missing picture in access unit with size 2467
Too many packets buffered for output stream 0:1.
[aac @ 0000013d8b6bf5c0] Qavg: 115.929
[aac @ 0000013d8b6bf5c0] 2 frames left in the queue on closing
Conversion failed!


  



  Does anyone know how to fix it, i am getting stuck for few days ?
Much appreciated !