Back when I gave my arguments I why don’t consider Rust a mature language, one of those arguments was that is lacks proper assembler support and systems programming language requires it since some of the tasks you need to perform (including optimisation) require as low level access as you can get. Here I would like to argue why asm!{} may be enough for most cases it’s definitely not for mine.
Read the rest of this entry »
Rust needs proper stand-alone assembler support
July 27th, 2021Looking for formats to look at
July 21st, 2021As I mentioned in one of the previous posts, I’ve achieved all the goals for NihAV that I initially set except for trying to write a proper encoder (and no, world domination has never been on my list). Unfortunately this will be not so easy thing to do and I’d like to have a distraction time from time.
Usually I distracted myself with reverse-engineering some format and maybe implementing decoding support for it in NihAV but recently I realized that I ran out of low-hanging fruit. There should be interesting codecs and game containers out there still waiting for their chance but I could not remember anything. I even went through ScummVM code and documented video formats from there in The Wiki, that’s how bored I was.
So I’d be grateful if somebody can point me out to a thing to RE. Last time when Peter drew my attention to VGM/XVD it turned out to be a very fulfilling experience.
On FMV games video compression
July 13th, 2021Back in the day I’ve stumbled on a post called Archaic Video Compression Algorithms of FMV Games Era (in Russian) giving a concise review of VGA era video. For a long time I wanted to discuss it but I lost the link to it and managed to locate it just recently.
For those unable to read Russian here’s a gist of it: all the variety of those formats used just three methods to compress video, those methods being lossless redundancy removal via RLE or LZ77-based method (Huffman codes not employed), vector quantisation, and block truncation coding. Inter-frame compression was rather primitive, often without any motion predictions; fade-ins/outs were not handled either. And later, when computers became powerful enough, everybody switched to DCT coding and later to in-engine animation.
It is a very good first approximation but as a person who looked into several formats myself and even documented a couple of them, I have something to add or correct.
First of all, ‘archaic’ means something not merely old but also not in common use—and that does not apply neither to LZ77 (which is still the base of most of the practical compressors and archivers; and there’s WebP using its variant in lossless mode) nor vector quantisation (even if it’s mostly employed in texture compression nowadays).
Second, I consider block truncation coding to be a specialised form of vector quantisation for a case of two colours per block. And considering that the input data could be palettised already it’s more likely the encoder searched for the most fitting two colours in the palette instead of generating two RGB triplets and handing it all to the palettisation process.
Third, I’d argue that back in the day codecs did not need sub-pixel motion compensation and motion vectors were often not needed since you could simply code an offset in LZ77 compression. And some of the codecs allowed copying data from either already decoded parts of the current frame or some not yet touched part (or any part in some cases) of the previous frame. This archaic method of copying from already decoded part of the frame is present in AV1 for example (under IntraBC name).
Fourth, I think that recursive block division deserves to be mentioned there.
A minor nit would be that even if very few codecs supported explicit fading, a lot of them had support for palette change event (partial or full). Yet it mostly depended on people developing the format and the encoder for it how well it was handled.
So in my opinion the final list of the techniques most FMV games of the VGA period used should look like this:
- lossless data compression with RLE or LZ77-based method
- vector quantisation—both whole tiles and inside single tile;
- recursive block splitting (either horizontal/vertical or as a quadtree);
- and fullpel motion compensation usually done by skipping pixels, blocks, or reusing some already decoded region.
And typical video codec of that era should look like one of those two:
- line-based codec employing RLE, probably with skips for inter-frames and optional LZ77 compressing of whole frame data afterwards (or extending LZ77 approach to work with the forward data in the buffer as well);
- tile-based codec usually working with 4×4 tiles, employing vector quantisation to paint the tile with 1-8 colours, sometimes sub-dividing tile for better detail preservation, sometimes having a dictionary of tile configurations, sometimes even employing block-based motion compensation; those codecs usually employ 1-4 byte opcodes to code the tile type of whole sequences of tiles of the same type.
Of course there are deviations from those two schemes but the majority of game codecs of that era should follow either of those two designs.
Later, as the original article said, those codecs were supplanted by DCT-based ones (I’ve seen the ones based on MJPEG and MPEG-1/2; and there’s hybrid Bink). On the other hand, DCT-based Bink2 is still in use.
Overall, if you look at game formats section in the wiki you’ll see a lot of various formats and many of them are built on a handful of the idea described above while still being quite different from each other. Originality lies not only in the methods employed but also in how you combine them together. And the modern codecs are the example of how different organisations build essentially the same codec using slightly different methods. MPEG-5 LCEVC was a nice change though.
Here’s a ship being loaded
July 8th, 2021Since I admitted myself that I feel old, I guess it’s my solemn duty to yell at clouds time from time.
I consider satire to be the most realistic depiction of the world (unless it’s a pasquinade) and as Shakespeare put a phrase in a mouth of one of the King Lear characters, jesters do oft prove prophets.
For example, I still like to re-read various satirical pieces by Ilf and Petrov (probably the best Soviet satirists in the first half of XXth century) and one of those pieces gave a title to this post because of its similarity to what I want to talk about.
That short story mentions a game “loading a ship” (here’s the title) where a group of bored people tries to “load a ship” by calling things starting with the same pre-defined letter—lamps, Lilliputians, locomotives, liquors etc etc—until people can’t recall any more things starting with that letter or somebody suggest a name and people start to argue if that should be allowed. The story itself is about one man who metaphorically loaded a ship by inventing new and new social activities until at some point it was discovered that nobody remembers what are his direct duties should be and he was fired.
And here we transition to the Linux Foundation. Previously I thought it should promote Linux adoption, make some Linux-related standards and pay money to main Linux developers and maintainers so they can work on it full-time. But news in last couple of months showed me I was wrong.
First there was a piece of news about AgStack (Linux for Agriculture). Then there was a piece of news about Linux Foundation organising some unrelated initiative for Microsoft (while not participating in it itself). And finally there’s a piece of news about Open Voice Network, an initiative for ethical standards of voice assistants.
This made me wonder not just why I encounter such news but also what does the foundation really do. The answer did not make me more optimistic.
So it seems that Linux Foundation is transitioning from a foundation that does what I expected it to do to an organisation that does IaaS (initiative-as-a-service): you pay them and they prepare all required infrastructure to have it all running, just invite some organisations to participate. Beside that what are those member companies are paying for? I don’t know the official explanation but to me it looks like three major categories to put it very bluntly and impolitely: bribes, extortion money and indulgences. Bribes is what you pay to affect the politics of the foundation in a way favourable to you (maybe it’s called lobbying, maybe such things do not happen at all—give me facts and I’ll amend this post). Extortion money is what you have to pay to participate in various standardising activity (aka membership fees, no different from any other standardising activity). Indulgences are money you pay to avert attention from your GPL violations regarding the kernel sources (just search for GPL violators and Linux Foundation, you’ll find not just Chinese but American companies mentioned there; at least in one case the known GPL violator hasn’t published its modified kernel sources after becoming a member).
I could not find any reports about the foundation beside their 2020 annual report so I can only speculate how it has changed in the past regarding income, spending and stuff. Yet I can foresee three scenarios on how it may develop in the future:
- Shifting to a different activity. Even now actual Linux-related things seem to be less than a half of what the foundation does nowadays, so maybe in the future it will realise that Linux is a legacy they don’t care any longer and maybe even rename themselves to reflect their new main occupation;
- Fossilising. Linux Foundation may exist in the future but both its activity and being a member will become a tradition that nobody understands but they’ll keep doing it because it’s a tradition;
- Withering. While Linux is relatively popular kernel, nobody can guarantee it will remain popular in the future. Linus himself is not immortal and his successor may be not as skilled, we have IBM trying to replace Linux kernel with systemd while keeping the name—or maybe some other reason will hurt Linux popularity. Or some other kernel will replace it in its major niches (I can easily imagine Android being rebased on Fuchsia; the same may happen to servers as well). In result Linux and Linux-related things will become not interesting to most companies and they’ll drop their financial support.
You may say that there might be a scenario where Linux Foundation will concentrate on Linux-only stuff but I’m sceptical. AgStack is a clear signal they run out of ideas where to promote Linux but in accordance with Parkinson’s law they’ll still keep growing by inventing new stuff as long as there’s enough income to employ more people, organise more conferences and such.
As usual, I’ll be happy to be proved wrong.
NihAV: feature complete
June 13th, 2021A year ago I wrote a post about NihAV being conceptually done, now it’s even closer to perfection since I’ve finished writing a useful video player.
Writing it (and especially debugging it) was surprisingly hard so in most cases after looking at it I thought that I’d rather do something else—RE some game format, work on deflate support, just anything but that. But at last it’s done and working adequately. The previous player could just play video until it’s over (or deadlock occasionally), this one supports pausing and seeking plus it seems not to deadlock.
Conceptually it’s very simple as any other player concept, it’s just various features and limitations complicate the design. You simply open input, read packets, discard them or decode with a corresponding decoder and present the result (draw the frame or send audio to the sound card). Now what to do in order to make it offer all the features I wanted from it?
First, interactivity and low(er) response latency. The latter is easy to achieve, you just put audio and video decoding into separate threads so the main loop just does the demuxing and checking for user commands. So now you have user commands that might affect decoding threads (for example, seeking or going to play the next file). I implemented it by setting a special flag that decoding thread checks and discards all queued input packets until some command is received.
Another fun thing was volume control. In my audio player I simply queue samples and let SDL deal with them, here I switched to callback and fill the requested buffer with samples adjusted by current volume. This way you have volume change applied almost immediately instead of it taking effect in a second or more depending in audio queue fill.
And speaking about queues, that was also a fun thing to manage. In a default implementation (demux-send-repeat) you either end up sending all packets before the decoding thread processes them all (and this will consume a lot of memory) or you block on sending via a limited message channel (which either requires a separate thread with more complicated interaction between them all or you can forget about interactivity). The answer is obviously to keep check of how full the channel is and do not demux new packets unless you’re sure you can send them. I keep a queue for the packets or events that should be sent but there’s no space in communication channel for that yet. Additionally I make audio part report the current buffer fill so I know there’s no reason to send more packets yet. Similarly for video I keep a queue of ready to display frames (in SDL textures, so drawing them is just one SDL blit call).
Overall, nothing particularly tricky but debugging it was still not fun. I actually ended up adding a compile-time debugging feature that will dump a lot of internal playback information into debug.log so I can figure out what actually happened there (i.e. NIHing a logger but you should not be surprised by that).
Of course it still lacks a lot of features for a serious player like proper synchronisation (with automatic framedropping and audio underrun corrections), more features (like playback at different speed, taking screenshots and switching between different input streams) and GUI. But I’m fine with the current state of my player and maybe enhance it later if the need arises.
Fun thing is that my player seemed to stall on MP4s. As it turned out, the problem was in MP4 demuxer producing packets in interleaved order (one for audio stream, one for video stream, one for audio stream again, one for video stream…) while it should’ve output packets for various streams for more or less the same time position (which means usually two AAC frames between video frames instead of just one). After the change my player works as expected on MP4s as well. And if I ever get to fixing and optimising H.264 decoder it should be good enough to serve as an everyday video player (I use nihav-sndplay to listen to my music collection already).
I guess there’s just one of the original ideas (of what I wanted to try at the time of public NihAV release) left that I haven’t tried yet, namely to experiment on writing some DCT-based video encoder with rate control and such (maybe even for VP6). This should keep me occupied for a couple of years. Or at least inspire me to do something else instead.
On the Origin of Bloatware
June 11th, 2021This is inspired by both a private discussion on why modern computing is so complex and my migration from Ubuntu 12.04LTS to systemd 20.04LTS.
Since I’ve finally changed from my less than ten years old operating system to something more modern I’ve noticed that it became noticeably slower (not irritatingly slower though but slower nevertheless) except for Firefox (which is probably not because of JS engine improvements but rather because of native execution of now supported APIs instead of polyfills). And trying various desktop environments before settling on Cinnamon I’m horrified by how bloated and unusable (to me) they are. My friends complain about modern technology demanding more effort to maintain because of complexity and weird interdependencies—while it’s supposed to make your life easier. So why it is like that?
For a keen reader the title of this post contains the answer. For the rest I’ll elaborate it below.
Read the rest of this entry »
Revisiting legendary Q format
May 30th, 2021Since I had nothing better to do I decided to look again at Q format and try to write a decoder for NihAV while at it.
It turns out there are three versions of the format are known: the one in Death Gate (version 3), the one in Shannara (version 4) and the one in Mission Control (not an adventure game this time; it’s version 5 obviously). Versions 4 and 5 differ only in minor details, the compression is the same. Version 3 uses the same principles but some coding details are different.
The main source of confusion was the fact that you have two context-dependent opcodes, namely 0xF9 and 0xFB. The first one either repeats the previous block several times or reuses motion vector from that block. The second one is even trickier. For version 5 frames and for version 4 with mode 7 signalled it signals a series of blocks with 3-16 colours in each. But for mode 6 it signals a series of blocks with the same type as the previous one but with some parameters changed. If it was preceded by a fill block, these blocks with have fill value. After a block with patterns you have a series of blocks with patterns reusing the same colours as the original block. For motion-compensated block you have the same kind of motion information transmitted.
But the weirdest thing IMO is the interlaced coding in version 5. For some reason (scalability? lower latency?) they decided to code frame in two part, so frame type 9 codes even rows of the frame and frame type 11 codes odd rows—and in this cases rows are four pixels high as it is one block height. That is definitely not something I was expecting.
All in all, the format turned out to be even weirder than I expected it to be.
Why I still like C and strongly dislike C++
May 26th, 2021This comes up in my conversations surprisingly often so I thought it’s worth to write my thoughts down instead of repeating them again and again.
As it is common with C programmers, C was not my first nor my last language, but I still like it and when I have to write programs I do it in C. Meanwhile I try to be aware of modern (and not so modern) programming languages and their trends and write my own multimedia-related hobby project in Rust. So why I have not moved to anything else yet and how C++ comes to all this?
Read the rest of this entry »
ZMBV support in NihAV and deflate format fun
May 22nd, 2021As I said in the previous post, I wanted to add ZMBV support to NihAV, mostly because it is rather simple codec (which means I can write a decoder and an encoder for it without spending too much time), it’s lossless and supports various bit-depths too (which means I can encode various content into it preserving the original format).
I still had to improve my deflate support (both decompressing and compressing) a bit to support the way the data is stored there. At least now I mostly understand what various flags are for.
First of all, by itself deflate format specifies just a bitstream split into blocks of data that may contain any amount of coded data. And these blocks start at the next bit after the previous block has ended, no byte aligning except by chance or after a copy block (which aligns bitstream before storing length and block contents).
Then, there is raw format used in various formats (like Zip or gzip) and there’s zlib format used for most cases data is stored as part of some other format (that means you have two initial bytes like 0x78 0x5E and 2×2 bytes of checksum in the end).
So, ZMBV uses unterminated stream format: first frame contains zlib header plus one or several blocks of data padded with an empty copy block to the byte limit, next frame contains continuation of that stream (also one or more blocks padded to the byte boundary) and so on. This is obviously done so you can decode frames one after another and still exploit the redundancy from the previously coded frame data if you’re lucky.
Normally you would start decoding data and keep decoding it until the final block (there’s a flag in block header for that) has been decoded—or error out earlier for insufficient data. In this case though we need to decode data block, check if we are at the end of input data and then return the decoded data. Similarly during data compression we need to encode all current data and pad output stream to the byte boundary if needed.
This is not hard or particularly tricky but it demonstrates that deflated data can be stored in different ways. At least now I really understand what that Z_SYNC_FLUSH flag is for.
Adding deflate support to NihAV
May 18th, 2021Since I wanted to do something different I decided to finally implement deflate support for NihAV—by which I mean compression support in addition to decompression. Here is how well it went.
As usual, my goal was to implement it in mostly straightforward way but with reasonable speed instead of having something completely on par with zlib or better.
At first I implemented the simplest form of compression – copying data without compression (but with the proper headers and ADLER-32 checksum at the end). Then I added a simple encoding with fixed codes that simply output symbols as it—no compression yet but at least it tests how well bitstream is written. Then I moved to dynamic codes. Then I added a brute force search and started encoding matches. So by the end of the weekend I had something working already and then I could make it faster and/or better.
Of course the first thing to remember is that you can reduce search time by using some structure for a faster text search. I think suffix trie is now popular but I settled for an old-fashioned hash by three bytes. Initially it was twice as slow since while the number of string comparisons decreased hundredfold, updating hash table on each step consumed too much time. So I switched to linked-list hash that resembles FAT somewhat (i.e. for each position in the input you have a pointer to the next location of the same three-letter hash plus an additional table pointing to the start of chain for each hash key). And I calculated it once per a large block just discarding matches outside of the desired range. Of course this can be done better but it worked fast enough for me.
Now the compression. There are three main strategies I tried: naïve aka greedy one (you simply output the longest match you can find at the current step), lazy (you also check the next position if it produces even better match and use it if possible—surprisingly enough it gives a significant benefit) and theoretically optimal (you construct a trellis and see which combination and literals can give you the best coding; it has issues but theoretically it’s the best one).
So why it’s “theoretically optimal” and not just optimal? Because it needs to calculate the accurate bit cost and you can’t know it until you produce all the symbols to be encoded and calculate the actual lengths for them. Of course you can do it in an iterative process or employ a heuristic to predict bit length somehow but I simply used “9 bits for the symbol and 5 bits plus escape bits for distance additionally if it’s present”. I think for some cases it even produced larger files than lazy decoding.
Here is a list from the top of my head of things than can be improved (but I guess anybody who has written a LZ77-based compressor knows it better than me):
- method selection—sometimes copying data verbatim is better (in the case of noise) or using fixed codes (because the overhead from transmitting dynamic codes eats all the advantage);
- partitioning—currently I use 64kB blocks but depending on content (usually detected by the symbol frequency variations) it’s better to cut block earlier or make it larger. I played a bit with the block size but changing it (in either direction) currently leads to compression ratio drops;
- faster search for the matching strings;
- heuristics for either faster optimal parsing or better-compressing other method.
Of course some of it can be sped up by simply using unsafe Rust so no checks on array access are performed but I don’t think it’s worth it now.
And finally here are some benchmarks for the curious ones performed on a source file of the program:
- copy: 32156 bytes (from 32145 bytes)
- fixed codes and greedy search: 7847 bytes, 80ms
- dynamic codes and greedy search: 6818 bytes, 80ms
- dynamic codes and lazy search: 6665 bytes, 100ms
- dynamic codes and “optimal” search: 6529 bytes, 690ms
gzip -9for the reference: 6466 bytes, <10ms
As you can see it’s not fast but it works. I also checked that the resulting compressed data is decoded fine (plus some tests on large files that will be split into several blocks). Now all that’s left is to implement ZMBV decoder and encoder.