Sur d’autres sites (4726)

• Adventures in Unicode

  29 novembre 2012, par Multimedia Mike — Programming, php, Python, sqlite3, unicode

  Tangential to multimedia hacking is proper metadata handling. Recently, I have gathered an interest in processing a large corpus of multimedia files which are likely to contain metadata strings which do not fall into the lower ASCII set. This is significant because the lower ASCII set intersects perfectly with my own programming comfort zone. Indeed, all of my programming life, I have insisted on covering my ears and loudly asserting “LA LA LA LA LA ! ALL TEXT EVERYWHERE IS ASCII !” I suspect I’m not alone in this.

  Thus, I took this as an opportunity to conquer my longstanding fear of Unicode. I developed a self-learning course comprised of a series of exercises which add up to this diagram :

  
  
  

  Part 1 : Understanding Text Encoding
  Python has regular strings by default and then it has Unicode strings. The latter are prefixed by the letter ‘u’. This is what ‘ö’ looks like encoded in each type.

  python

  < view plain text >

  1. >>> ’ö’, u’ö’
  2. (’\xc3\xb6’, u’\xf6’)

  A large part of my frustration with Unicode comes from Python yelling at me about UnicodeDecodeErrors and an inability to handle the number 0xc3 for some reason. This usually comes when I’m trying to wrap my head around an unrelated problem and don’t care to get sidetracked by text encoding issues. However, when I studied the above output, I finally understood where the 0xc3 comes from. I just didn’t understand what the encoding represents exactly.
  

  I can see from assorted tables that ‘ö’ is character 0xF6 in various encodings (in Unicode and Latin-1), so u’\xf6′ makes sense. But what does ‘\xc3\xb6′ mean ? It’s my style to excavate straight down to the lowest levels, and I wanted to understand exactly how characters are represented in memory. The UTF-8 encoding tables inform us that any Unicode code point above 0x7F but less than 0×800 will be encoded with 2 bytes :

   110xxxxx 10xxxxxx
  

  Applying this pattern to the \xc3\xb6 encoding :

              hex : 0xc3      0xb6
             bits : 11000011  10110110
   important bits : ---00011  —110110
        assembled : 00011110110
       code point : 0xf6
  

  I was elated when I drew that out and made the connection. Maybe I’m the last programmer to figure this stuff out. But I’m still happy that I actually understand those Python errors pertaining to the number 0xc3 and that I won’t have to apply canned solutions without understanding the core problem.

  I’m cheating on this part of this exercise just a little bit since the diagram implied that the Unicode text needs to come from a binary file. I’ll return to that in a bit. For now, I’ll just contrive the following Unicode string from the Python REPL :

  python

  < view plain text >

  1. >>> u = u’Üñìçôđé’
  2. >>> u
  3. u’\xdc\xf1\xec\xe7\xf4\u0111\xe9’

  Part 2 : From Python To SQLite3
  The next step is to see what happens when I use Python’s SQLite3 module to dump the string into a new database. Will the Unicode encoding be preserved on disk ? What will UTF-8 look like on disk anyway ?

  python

  < view plain text >

  1. >>> import sqlite3
  2. >>> conn = sqlite3.connect(’unicode.db’)
  3. >>> conn.execute("CREATE TABLE t (t text)")
  4. >>> conn.execute("INSERT INTO t VALUES (?)", (u, ))
  5. >>> conn.commit()
  6. >>> conn.close()

  Next, I manually view the resulting database file (unicode.db) using a hex editor and look for strings. Here we go :

  000007F0   02 29 C3 9C  C3 B1 C3 AC  C3 A7 C3 B4  C4 91 C3 A9
  

  Look at that ! It’s just like the \xc3\xf6 encoding we see in the regular Python strings.

  Part 3 : From SQLite3 To A Web Page Via PHP
  Finally, use PHP (love it or hate it, but it’s what’s most convenient on my hosting provider) to query the string from the database and display it on a web page, completing the outlined processing pipeline.

  php

  < view plain text >

  1. < ?php
  2. $dbh = new PDO("sqlite:unicode.db") ;
  3. foreach ($dbh->query("SELECT t from t") as $row) ;
  4. $unicode_string = $row[’t’] ;
  5.  ?>
  6.  
  7. <html>
  8. <head><meta http-equiv="Content-Type" content="text/html ; charset=utf-8"></meta></head>
  9. <body><h1>< ?=$unicode_string ?></h1></body>
  10. </html>

  I tested the foregoing PHP script on 3 separate browsers that I had handy (Firefox, Internet Explorer, and Chrome) :

  
  
  

  I’d say that counts as success ! It’s important to note that the “meta http-equiv” tag is absolutely necessary. Omit and see something like this :

  
  
  

  Since we know what the UTF-8 stream looks like, it’s pretty obvious how the mapping is operating here : 0xc3 and 0xc4 correspond to ‘Ã’ and ‘Ä’, respectively. This corresponds to an encoding named ISO/IEC 8859-1, a.k.a. Latin-1. Speaking of which…

  Part 4 : Converting Binary Data To Unicode
  At the start of the experiment, I was trying to extract metadata strings from these binary multimedia files and I noticed characters like our friend ‘ö’ from above. In the bytestream, this was represented simply with 0xf6. I mistakenly believed that this was the on-disk representation of UTF-8. Wrong. Turns out it’s Latin-1.

  However, I still need to solve the problem of transforming such strings into Unicode to be shoved through the pipeline diagrammed above. For this experiment, I created a 9-byte file with the Latin-1 string ‘Üñìçôdé’ couched by 0′s, to simulate yanking a string out of a binary file. Here’s unicode.file :

  00000000   00 DC F1 EC  E7 F4 64 E9  00         ......d..
  

  (Aside : this experiment uses plain ‘d’ since the ‘đ’ with a bar through it doesn’t occur in Latin-1 ; shows up all over the place in Vietnamese, at least.)

  I’ve been mashing around Python code via the REPL, trying to get this string into a Unicode-friendly format. This is a successful method but it’s probably not the best :

  python

  < view plain text >

  1. >>> import struct
  2. >>> f = open(’unicode.file’, ’r’).read()
  3. >>> u = u’’
  4. >>> for c in struct.unpack("B"*7, f[1 :8]) :
  5. ... u += unichr(c)
  6. ...
  7. >>> u
  8. u’\xdc\xf1\xec\xe7\xf4d\xe9’
  9. >>> print u
  10. Üñìçôdé

  Conclusion
  Dealing with text encoding matters reminds me of dealing with integer endian-ness concerns. When you’re just dealing with one system, you probably don’t need to think too much about it because the system is usually handling everything consistently underneath the covers.

  However, when the data leaves one system and will be interpreted by another system, that’s when a programmer needs to be cognizant of matters such as integer endianness or text encoding.

• ffmpeg audio out of sync with video stacking

  3 avril 2024, par s0mbre

  The problem

  &#xA;

  When trying to do horizontal stacking of two videos in ffmpeg, the combined audio track loses sync with the video on the second input. As far as I've look it up, this problem is very common, not to say notorious with ffmpeg.

  &#xA;

  I do hstack muxing like this :

  &#xA;

  ffmpeg -i 1.mp4 -i 2.mp4 -filter_complex \
"[0:v]scale=1280:-2,crop=w=640:h=720:x=0[v1]; \
[1:v]scale=1280:-2,crop=w=640:h=720:x=0[v2]; \
[v1][v2]hstack=shortest=1[v]; \
[0:a][1:a]amix=duration=shortest[a]" \
-map [v] -map [a] -c:v libx264 -c:a libmp3lame -r 30 -y stuff/out.mp4


  &#xA;

  It encodes just fine as far as the hsplit goes. But the resulting video is out of sync with the audio : the second input (located on the right side in the resulting split) demonstrates about 3 sec. audio off-syncking, where the audio track is ahead of the picture. I realize this is somehow connected with the source videos' timestamps, but no popular remediation recipes helped (see below).

  &#xA;

  What I expect

  &#xA;

  I'd expect a resulting stacked video where the audio track is muxed with exact correspondence to the original input pictures.

  &#xA;

  What I tried (all in vain !)

  &#xA;

  Something I've tried but to no avail :

  &#xA;

  &#xA;
  1. Appending -async 1 option as suggested here and here

  &#xA;

  2. Using the aresample=async=1 or aresample=async=1000 filter on each audio input as suggested here and here

  &#xA;

  3. Padding each audio track with apad as suggested here

  &#xA;

  4. Using the adelay=0 and adelay=[delay]s filters on the failing input

  &#xA;

  5. Changing the audio codec to a number of alternatives (aac etc.)

  &#xA;

  6. Infinite combinations of 1-5 above...

  &#xA;

  &#xA;

  What seems indeed to work is manual passing a delay value to the -itsoffset filter as suggested here and in the docs and using the offset track as an extra (pure audio) input. But how do I find the exact offset with a different set of videos ?

  &#xA;

  In short, I am at a standstill after 7+ days of ravenous search-and-try.

  &#xA;

• Anatomy of an optimization : H.264 deblocking

  26 mai 2010, par Dark Shikari — H.264, assembly, development, speed, x264

  As mentioned in the previous post, H.264 has an adaptive deblocking filter. But what exactly does that mean — and more importantly, what does it mean for performance ? And how can we make it as fast as possible ? In this post I’ll try to answer these questions, particularly in relation to my recent deblocking optimizations in x264.

  H.264′s deblocking filter has two steps : strength calculation and the actual filter. The first step calculates the parameters for the second step. The filter runs on all the edges in each macroblock. That’s 4 vertical edges of length 16 pixels and 4 horizontal edges of length 16 pixels. The vertical edges are filtered first, from left to right, then the horizontal edges, from top to bottom (order matters !). The leftmost edge is the one between the current macroblock and the left macroblock, while the topmost edge is the one between the current macroblock and the top macroblock.

  Here’s the formula for the strength calculation in progressive mode. The highest strength that applies is always selected.

  If we’re on the edge between an intra macroblock and any other macroblock : Strength 4
  If we’re on an internal edge of an intra macroblock : Strength 3
  If either side of a 4-pixel-long edge has residual data : Strength 2
  If the motion vectors on opposite sides of a 4-pixel-long edge are at least a pixel apart (in either x or y direction) or the reference frames aren’t the same : Strength 1
  Otherwise : Strength 0 (no deblocking)

  These values are then thrown into a lookup table depending on the quantizer : higher quantizers have stronger deblocking. Then the actual filter is run with the appropriate parameters. Note that Strength 4 is actually a special deblocking mode that performs a much stronger filter and affects more pixels.

  One can see somewhat intuitively why these strengths are chosen. The deblocker exists to get rid of sharp edges caused by the block-based nature of H.264, and so the strength depends on what exists that might cause such sharp edges. The strength calculation is a way to use existing data from the video stream to make better decisions during the deblocking process, improving compression and quality.

  Both the strength calculation and the actual filter (not described here) are very complex if naively implemented. The latter can be SIMD’d with not too much difficulty ; no H.264 decoder can get away with reasonable performance without such a thing. But what about optimizing the strength calculation ? A quick analysis shows that this can be beneficial as well.

  Since we have to check both horizontal and vertical edges, we have to check up to 32 pairs of coefficient counts (for residual), 16 pairs of reference frame indices, and 128 motion vector values (counting x and y as separate values). This is a lot of calculation ; a naive implementation can take 500-1000 clock cycles on a modern CPU. Of course, there’s a lot of shortcuts we can take. Here’s some examples :

  • If the macroblock uses the 8×8 transform, we only need to check 2 edges in each direction instead of 4, because we don’t deblock inside of the 8×8 blocks.
  • If the macroblock is a P-skip, we only have to check the first edge in each direction, since there’s guaranteed to be no motion vector differences, reference frame differences, or residual inside of the macroblock.
  • If the macroblock has no residual at all, we can skip that check.
  • If we know the partition type of the macroblock, we can do motion vector checks only along the edges of the partitions.
  • If the effective quantizer is so low that no deblocking would be performed no matter what, don’t bother calculating the strength.

  But even all of this doesn’t save us from ourselves. We still have to iterate over a ton of edges, checking each one. Stuff like the partition-checking logic greatly complicates the code and adds overhead even as it reduces the number of checks. And in many cases decoupling the checks to add such logic will make it slower : if the checks are coupled, we can avoid doing a motion vector check if there’s residual, since Strength 2 overrides Strength 1.

  But wait. What if we could do this in SIMD, just like the actual loopfilter itself ? Sure, it seems more of a problem for C code than assembly, but there aren’t any obvious things in the way. Many years ago, Loren Merritt (pengvado) wrote the first SIMD implementation that I know of (for ffmpeg’s decoder) ; it is quite fast, so I decided to work on porting the idea to x264 to see if we could eke out a bit more speed here as well.

  Before I go over what I had to do to make this change, let me first describe how deblocking is implemented in x264. Since the filter is a loopfilter, it acts “in loop” and must be done in both the encoder and decoder — hence why x264 has it too, not just decoders. At the end of encoding one row of macroblocks, x264 goes back and deblocks the row, then performs half-pixel interpolation for use in encoding the next frame.

  We do it per-row for reasons of cache coherency : deblocking accesses a lot of pixels and a lot of code that wouldn’t otherwise be used, so it’s more efficient to do it in a single pass as opposed to deblocking each macroblock immediately after encoding. Then half-pixel interpolation can immediately re-use the resulting data.

  Now to the change. First, I modified deblocking to implement a subset of the macroblock_cache_load function : spend an extra bit of effort loading the necessary data into a data structure which is much simpler to address — as an assembly implementation would need (x264_macroblock_cache_load_deblock). Then I massively cleaned up deblocking to move all of the core strength-calculation logic into a single, small function that could be converted to assembly (deblock_strength_c). Finally, I wrote the assembly functions and worked with Loren to optimize them. Here’s the result.

  And the timings for the resulting assembly function on my Core i7, in cycles :

  deblock_strength_c : 309
  deblock_strength_mmx : 79
  deblock_strength_sse2 : 37
  deblock_strength_ssse3 : 33

  Now that is a seriously nice improvement. 33 cycles on average to perform that many comparisons–that’s absurdly low, especially considering the SIMD takes no branchy shortcuts : it always checks every single edge ! I walked over to my performance chart and happily crossed off a box.

  But I had a hunch that I could do better. Remember, as mentioned earlier, we’re reloading all that data back into our data structures in order to address it. This isn’t that slow, but takes enough time to significantly cut down on the gain of the assembly code. And worse, less than a row ago, all this data was in the correct place to be used (when we just finished encoding the macroblock) ! But if we did the deblocking right after encoding each macroblock, the cache issues would make it too slow to be worth it (yes, I tested this). So I went back to other things, a bit annoyed that I couldn’t get the full benefit of the changes.

  Then, yesterday, I was talking with Pascal, a former Xvid dev and current video hacker over at Google, about various possible x264 optimizations. He had seen my deblocking changes and we discussed that a bit as well. Then two lines hit me like a pile of bricks :

  <_skal_> tried computing the strength at least ?
  <_skal_> while it’s fresh

  Why hadn’t I thought of that ? Do the strength calculation immediately after encoding each macroblock, save the result, and then go pick it up later for the main deblocking filter. Then we can use the data right there and then for strength calculation, but we don’t have to do the whole deblock process until later.

  I went and implemented it and, after working my way through a horde of bugs, eventually got a working implementation. A big catch was that of slices : deblocking normally acts between slices even though normal encoding does not, so I had to perform extra munging to get that to work. By midday today I was able to go cross yet another box off on the performance chart. And now it’s committed.

  Sometimes chatting for 10 minutes with another developer is enough to spot the idea that your brain somehow managed to miss for nearly a straight week.

  NB : the performance chart is on a specific test clip at a specific set of settings (super fast settings) relevant to the company I work at, so it isn’t accurate nor complete for, say, default settings.

  Update : Here’s a higher resolution version of the current chart, as requested in the comments.
  

  http://git.videolan.org/?p=x264.git ;a=commit ;h=b9ec3ea27812288485329ac7771143021cd4917b