Recherche avancée

Sur d’autres sites (9771)

• Translating Return To Ringworld

  17 août 2016, par Multimedia Mike — Game Hacking

  As indicated in my previous post, the Translator has expressed interest in applying his hobby towards another DOS adventure game from the mid 1990s : Return to Ringworld (henceforth R2RW) by Tsunami Media. This represents significantly more work than the previous outing, Phantasmagoria.

  
  
  Return to Ringworld Title Screen
  

  I have been largely successful thus far in crafting translation tools. I have pushed the fruits of these labors to a Github repository named improved-spoon (named using Github’s random name generator because I wanted something more interesting than ‘game-hacking-tools’).

  Further, I have recorded everything I have learned about the game’s resource format (named RLB) at the XentaxWiki.

  New Challenges
  The previous project mostly involved scribbling subtitle text on an endless series of video files by leveraging a separate software library which took care of rendering fonts. In contrast, R2RW has at least 30k words of English text contained in various blocks which require translation. Further, the game encodes its own fonts (9 of them) which stubbornly refuse to be useful for rendering text in nearly any other language.

  Thus, the immediate 2 challenges are :
  

  1. Translating volumes of text to Spanish
  2. Expanding the fonts to represent Spanish characters

  Normally, “figuring out the file format data structures involved” is on the list as well. Thankfully, understanding the formats is not a huge challenge since the folks at the ScummVM project already did all the heavy lifting of reverse engineering the file formats.

  The Pitch
  Here was the plan :

  • Create a tool that can dump out the interesting data from the game’s master resource file.
  • Create a tool that can perform the elaborate file copy described in the previous post. The new file should be bit for bit compatible with the original file.
  • Modify the rewriting tool to repack some modified strings into the new resource file.
  • Unpack the fonts and figure out a way to add new characters.
  • Repack the new fonts into the resource file.
  • Repack message strings with Spanish characters.

  Showing The Work : Modifying Strings
  First, I created the tool to unpack blocks of message string resources. I elected to dump the strings to disk as JSON data since it’s easy to write and read JSON using Python, and it’s quick to check if any mistakes have crept in.

  The next step is to find a string to focus on. So I started the game and looked for the first string I could trigger :

  
  
  

  This shows up in the JSON string dump as :

    
      "Spanish" : " !0205Your quarters on the Lance of Truth are spartan, in accord with your mercenary lifestyle.",
      "English" : " !0205Your quarters on the Lance of Truth are spartan, in accord with your mercenary lifestyle."
    ,
  

  As you can see, many of the strings are encoded with an ID key as part of the string which should probably be left unmodified. I changed the Spanish string :

    
      "Spanish" : " !0205Hey, is this thing on ?",
      "English" : " !0205Your quarters on the Lance of Truth are spartan, in accord with your mercenary lifestyle."
    ,
  

  And then I wrote the repacking tool to substitute this message block for the original one. Look ! The engine liked it !

  
  
  

  Little steps, little steps.

  Showing The Work : Modifying Fonts
  The next little step is to find a place to put the new characters. First, a problem definition : The immediate goal is to translate the game into Spanish. The current fonts encoded in the game resource only support 128 characters, corresponding to 7-bit ASCII. In order to properly express Spanish, 16 new characters are required : á, é, í, ó, ú, ü, ñ (each in upper and lower case for a total of 14 characters) as well as the inverted punctuation symbols : ¿, ¡.

  Again, ScummVM already documents (via code) the font coding format. So I quickly determined that each of the 9 fonts is comprised of 128 individual bitmaps with either 1 or 2 bits per pixel. I wrote a tool to unpack each character into an individual portable grey map (PGM) image. These can be edited with graphics editors or with text editors since they are just text files.

  Where to put the 16 new Spanish characters ? ASCII characters 1-31 are non-printable, so my first theory was that these characters would be empty and could be repurposed. However, after dumping and inspecting, I learned that they represent the same set of characters as seen in DOS Code Page 437. So that’s a no-go (so I assumed ; I didn’t check if any existing strings leveraged those characters).

  My next plan was hope that I could extend the font beyond index 127 and use positions 128-143. This worked superbly. This is the new example string :

    
      "Spanish" : " !0205¿Ves esto ? ¡La puntuacion se hace girar !",
      "English" : " !0205Your quarters on the Lance of Truth are spartan, in accord with your mercenary lifestyle."
    ,
  

  Fortunately, JSON understands UTF-8 and after mapping the 16 necessary characters down to the numeric range of 128-143, I repacked the new fonts and the new string :

  
  
  Translation : “See this ? The punctuation is rotated !”
  

  Another victory. Notice that there are no diacritics in this string. None are required for this translation (according to Google Translate). But adding the diacritics to the 14 characters isn’t my department. My tool does help by prepopulating [aeiounAEIOUN] into the right positions to make editing easier for the Translator. But the tool does make the effort to rotate the punctuation since that is easy to automate.

  Next Steps and Residual Weirdness
  There is another method for storing ASCII text inside the R2RW resource called strip resources. These store conversation scripts. There are plenty of fields in the data structures that I don’t fully understand. So, following the lessons I learned from my previous translation outing, I was determined to modify as little as possible. This means copying over most of the original data structures intact, but changing the field representing the relative offset that points to the corresponding string. This works well since the strings are invariably stored NULL-terminated in a concatenated manner.

  I wanted to document for the record that the format that R2RW uses has some weirdness in they way it handles residual bytes in a resource. The variant of the resource format that R2RW uses requires every block to be aligned on a 16-byte boundary. If there is space between the logical end of the resource and the start of the next resource, there are random bytes in that space. This leads me to believe that these bytes were originally recorded from stale/uninitialized memory. This frustrates me because when I write the initial file copy tool which unpacks and repacks each block, I want the new file to be identical to the original. However, these apparent nonsense bytes at the end thwart that effort.

  But leaving those bytes as 0 produces an acceptable resource file.

  Text On Static Images
  There is one last resource type we are working on translating. There are various bits of text that are rendered as images. For example, from the intro :

  
  
  

  It’s possible to locate and extract the exact image that is overlaid on this scene, though without the colors :

  
  
  

  The palettes are stored in a separate resource type. So it seems the challenge is to figure out the palette in use for these frames and render a transparent image that uses the same palette, then repack the new text-image into the new resource file.

• Announcing the world’s fastest VP8 decoder : ffvp8

  24 juillet 2010, par Dark Shikari — ffmpeg, google, speed, VP8

  Back when I originally reviewed VP8, I noted that the official decoder, libvpx, was rather slow. While there was no particular reason that it should be much faster than a good H.264 decoder, it shouldn’t have been that much slower either ! So, I set out with Ronald Bultje and David Conrad to make a better one in FFmpeg. This one would be community-developed and free from the beginning, rather than the proprietary code-dump that was libvpx. A few weeks ago the decoder was complete enough to be bit-exact with libvpx, making it the first independent free implementation of a VP8 decoder. Now, with the first round of optimizations complete, it should be ready for primetime. I’ll go into some detail about the development process, but first, let’s get to the real meat of this post : the benchmarks.

  We tested on two 1080p clips : Parkjoy, a live-action 1080p clip, and the Sintel trailer, a CGI 1080p clip. Testing was done using “time ffmpeg -vcodec libvpx or vp8 -i input -vsync 0 -an -f null -”. We all used the latest SVN FFmpeg at the time of this posting ; the last revision optimizing the VP8 decoder was r24471.

  

  As these benchmarks show, ffvp8 is clearly much faster than libvpx, particularly on 64-bit. It’s even faster by a large margin on Atom, despite the fact that we haven’t even begun optimizing for it. In many cases, ffvp8′s extra speed can make the difference between a video that plays and one that doesn’t, especially in modern browsers with software compositing engines taking up a lot of CPU time. Want to get faster playback of VP8 videos ? The next versions of FFmpeg-based players, like VLC, will include ffvp8. Want to get faster playback of WebM in your browser ? Lobby your browser developers to use ffvp8 instead of libvpx. I expect Chrome to switch first, as they already use libavcodec for most of their playback system.

  Keep in mind ffvp8 is not “done” — we will continue to improve it and make it faster. We still have a number of optimizations in the pipeline that aren’t committed yet.

  Developing ffvp8

  The initial challenge, primarily pioneered by David and Ronald, was constructing the core decoder and making it bit-exact to libvpx. This was rather challenging, especially given the lack of a real spec. Many parts of the spec were outright misleading and contradicted libvpx itself. It didn’t help that the suite of official conformance tests didn’t even cover all the features used by the official encoder ! We’ve already started adding our own conformance tests to deal with this. But I’ve complained enough in past posts about the lack of a spec ; let’s get onto the gritty details.

  The next step was adding SIMD assembly for all of the important DSP functions. VP8′s motion compensation and deblocking filter are by far the most CPU-intensive parts, much the same as in H.264. Unlike H.264, the deblocking filter relies on a lot of internal saturation steps, which are free in SIMD but costly in a normal C implementation, making the plain C code even slower. Of course, none of this is a particularly large problem ; any sane video decoder has all this stuff in SIMD.

  I tutored Ronald in x86 SIMD and wrote most of the motion compensation, intra prediction, and some inverse transforms. Ronald wrote the rest of the inverse transforms and a bit of the motion compensation. He also did the most difficult part : the deblocking filter. Deblocking filters are always a bit difficult because every one is different. Motion compensation, by comparison, is usually very similar regardless of video format ; a 6-tap filter is a 6-tap filter, and most of the variation going on is just the choice of numbers to multiply by.

  The biggest challenge in an SIMD deblocking filter is to avoid unpacking, that is, going from 8-bit to 16-bit. Many operations in deblocking filters would naively appear to require more than 8-bit precision. A simple example in the case of x86 is abs(a-b), where a and b are 8-bit unsigned integers. The result of “a-b” requires a 9-bit signed integer (it can be anywhere from -255 to 255), so it can’t fit in 8-bit. But this is quite possible to do without unpacking : (satsub(a,b) | satsub(b,a)), where “satsub” performs a saturating subtract on the two values. If the value is positive, it yields the result ; if the value is negative, it yields zero. Oring the two together yields the desired result. This requires 4 ops on x86 ; unpacking would probably require at least 10, including the unpack and pack steps.

  After the SIMD came optimizing the C code, which still took a significant portion of the total runtime. One of my biggest optimizations was adding aggressive “smart” prefetching to reduce cache misses. ffvp8 prefetches the reference frames (PREVIOUS, GOLDEN, and ALTREF)… but only the ones which have been used reasonably often this frame. This lets us prefetch everything we need without prefetching things that we probably won’t use. libvpx very often encodes frames that almost never (but not quite never) use GOLDEN or ALTREF, so this optimization greatly reduces time spent prefetching in a lot of real videos. There are of course countless other optimizations we made that are too long to list here as well, such as David’s entropy decoder optimizations. I’d also like to thank Eli Friedman for his invaluable help in benchmarking a lot of these changes.

  What next ? Altivec (PPC) assembly is almost nonexistent, with the only functions being David’s motion compensation code. NEON (ARM) is completely nonexistent : we’ll need that to be fast on mobile devices as well. Of course, all this will come in due time — and as always — patches welcome !

  Appendix : the raw numbers

  Here’s the raw numbers (in fps) for the graphs at the start of this post, with standard error values :

  Core i7 620QM (1.6Ghz), Windows 7, 32-bit :
  Parkjoy ffvp8 : 44.58 0.44
  Parkjoy libvpx : 33.06 0.23
  Sintel ffvp8 : 74.26 1.18
  Sintel libvpx : 56.11 0.96

  Core i5 520M (2.4Ghz), Linux, 64-bit :
  Parkjoy ffvp8 : 68.29 0.06
  Parkjoy libvpx : 41.06 0.04
  Sintel ffvp8 : 112.38 0.37
  Sintel libvpx : 69.64 0.09

  Core 2 T9300 (2.5Ghz), Mac OS X 10.6.4, 64-bit :
  Parkjoy ffvp8 : 54.09 0.02
  Parkjoy libvpx : 33.68 0.01
  Sintel ffvp8 : 87.54 0.03
  Sintel libvpx : 52.74 0.04

  Core Duo (2Ghz), Mac OS X 10.6.4, 32-bit :
  Parkjoy ffvp8 : 21.31 0.02
  Parkjoy libvpx : 17.96 0.00
  Sintel ffvp8 : 41.24 0.01
  Sintel libvpx : 29.65 0.02

  Atom N270 (1.6Ghz), Linux, 32-bit :
  Parkjoy ffvp8 : 15.29 0.01
  Parkjoy libvpx : 12.46 0.01
  Sintel ffvp8 : 26.87 0.05
  Sintel libvpx : 20.41 0.02

• Announcing the world’s fastest VP8 decoder : ffvp8

  24 juillet 2010, par Dark Shikari — VP8, ffmpeg, google, speed

  Back when I originally reviewed VP8, I noted that the official decoder, libvpx, was rather slow. While there was no particular reason that it should be much faster than a good H.264 decoder, it shouldn’t have been that much slower either ! So, I set out with Ronald Bultje and David Conrad to make a better one in FFmpeg. This one would be community-developed and free from the beginning, rather than the proprietary code-dump that was libvpx. A few weeks ago the decoder was complete enough to be bit-exact with libvpx, making it the first independent free implementation of a VP8 decoder. Now, with the first round of optimizations complete, it should be ready for primetime. I’ll go into some detail about the development process, but first, let’s get to the real meat of this post : the benchmarks.

  We tested on two 1080p clips : Parkjoy, a live-action 1080p clip, and the Sintel trailer, a CGI 1080p clip. Testing was done using “time ffmpeg -vcodec libvpx or vp8 -i input -vsync 0 -an -f null -”. We all used the latest SVN FFmpeg at the time of this posting ; the last revision optimizing the VP8 decoder was r24471.

  

  As these benchmarks show, ffvp8 is clearly much faster than libvpx, particularly on 64-bit. It’s even faster by a large margin on Atom, despite the fact that we haven’t even begun optimizing for it. In many cases, ffvp8′s extra speed can make the difference between a video that plays and one that doesn’t, especially in modern browsers with software compositing engines taking up a lot of CPU time. Want to get faster playback of VP8 videos ? The next versions of FFmpeg-based players, like VLC, will include ffvp8. Want to get faster playback of WebM in your browser ? Lobby your browser developers to use ffvp8 instead of libvpx. I expect Chrome to switch first, as they already use libavcodec for most of their playback system.

  Keep in mind ffvp8 is not “done” — we will continue to improve it and make it faster. We still have a number of optimizations in the pipeline that aren’t committed yet.

  Developing ffvp8

  The initial challenge, primarily pioneered by David and Ronald, was constructing the core decoder and making it bit-exact to libvpx. This was rather challenging, especially given the lack of a real spec. Many parts of the spec were outright misleading and contradicted libvpx itself. It didn’t help that the suite of official conformance tests didn’t even cover all the features used by the official encoder ! We’ve already started adding our own conformance tests to deal with this. But I’ve complained enough in past posts about the lack of a spec ; let’s get onto the gritty details.

  The next step was adding SIMD assembly for all of the important DSP functions. VP8′s motion compensation and deblocking filter are by far the most CPU-intensive parts, much the same as in H.264. Unlike H.264, the deblocking filter relies on a lot of internal saturation steps, which are free in SIMD but costly in a normal C implementation, making the plain C code even slower. Of course, none of this is a particularly large problem ; any sane video decoder has all this stuff in SIMD.

  I tutored Ronald in x86 SIMD and wrote most of the motion compensation, intra prediction, and some inverse transforms. Ronald wrote the rest of the inverse transforms and a bit of the motion compensation. He also did the most difficult part : the deblocking filter. Deblocking filters are always a bit difficult because every one is different. Motion compensation, by comparison, is usually very similar regardless of video format ; a 6-tap filter is a 6-tap filter, and most of the variation going on is just the choice of numbers to multiply by.

  The biggest challenge in an SIMD deblocking filter is to avoid unpacking, that is, going from 8-bit to 16-bit. Many operations in deblocking filters would naively appear to require more than 8-bit precision. A simple example in the case of x86 is abs(a-b), where a and b are 8-bit unsigned integers. The result of “a-b” requires a 9-bit signed integer (it can be anywhere from -255 to 255), so it can’t fit in 8-bit. But this is quite possible to do without unpacking : (satsub(a,b) | satsub(b,a)), where “satsub” performs a saturating subtract on the two values. If the value is positive, it yields the result ; if the value is negative, it yields zero. Oring the two together yields the desired result. This requires 4 ops on x86 ; unpacking would probably require at least 10, including the unpack and pack steps.

  After the SIMD came optimizing the C code, which still took a significant portion of the total runtime. One of my biggest optimizations was adding aggressive “smart” prefetching to reduce cache misses. ffvp8 prefetches the reference frames (PREVIOUS, GOLDEN, and ALTREF)… but only the ones which have been used reasonably often this frame. This lets us prefetch everything we need without prefetching things that we probably won’t use. libvpx very often encodes frames that almost never (but not quite never) use GOLDEN or ALTREF, so this optimization greatly reduces time spent prefetching in a lot of real videos. There are of course countless other optimizations we made that are too long to list here as well, such as David’s entropy decoder optimizations. I’d also like to thank Eli Friedman for his invaluable help in benchmarking a lot of these changes.

  What next ? Altivec (PPC) assembly is almost nonexistent, with the only functions being David’s motion compensation code. NEON (ARM) is completely nonexistent : we’ll need that to be fast on mobile devices as well. Of course, all this will come in due time — and as always — patches welcome !

  Appendix : the raw numbers

  Here’s the raw numbers (in fps) for the graphs at the start of this post, with standard error values :

  Core i7 620QM (1.6Ghz), Windows 7, 32-bit :
  Parkjoy ffvp8 : 44.58 0.44
  Parkjoy libvpx : 33.06 0.23
  Sintel ffvp8 : 74.26 1.18
  Sintel libvpx : 56.11 0.96

  Core i5 520M (2.4Ghz), Linux, 64-bit :
  Parkjoy ffvp8 : 68.29 0.06
  Parkjoy libvpx : 41.06 0.04
  Sintel ffvp8 : 112.38 0.37
  Sintel libvpx : 69.64 0.09

  Core 2 T9300 (2.5Ghz), Mac OS X 10.6.4, 64-bit :
  Parkjoy ffvp8 : 54.09 0.02
  Parkjoy libvpx : 33.68 0.01
  Sintel ffvp8 : 87.54 0.03
  Sintel libvpx : 52.74 0.04

  Core Duo (2Ghz), Mac OS X 10.6.4, 32-bit :
  Parkjoy ffvp8 : 21.31 0.02
  Parkjoy libvpx : 17.96 0.00
  Sintel ffvp8 : 41.24 0.01
  Sintel libvpx : 29.65 0.02

  Atom N270 (1.6Ghz), Linux, 32-bit :
  Parkjoy ffvp8 : 15.29 0.01
  Parkjoy libvpx : 12.46 0.01
  Sintel ffvp8 : 26.87 0.05
  Sintel libvpx : 20.41 0.02