Recherche avancée

Sur d’autres sites (3599)

• lavu/x86 : add FFT assembly

  10 avril 2021, par Lynne

  lavu/x86 : add FFT assembly
  This commit adds a pure x86 assembly SIMD version of the FFT in libavutil/tx.
  
  The design of this pure assembly FFT is pretty unconventional.
  On the lowest level, instead of splitting the complex numbers into
  
  real and imaginary parts, we keep complex numbers together but split
  
  them in terms of parity. This saves a number of shuffles in each transform,
  
  but more importantly, it splits each transform into two independent
  
  paths, which we process using separate registers in parallel.
  
  This allows us to keep all units saturated and lets us use all available
  
  registers to avoid dependencies.
  
  Moreover, it allows us to double the granularity of our per-load permutation,
  
  skipping many expensive lookups and allowing us to use just 4 loads per register,
  
  rather than 8, or in case FMA3 (and by extension, AVX2), use the vgatherdpd
  
  instruction, which is at least as fast as 4 separate loads on old hardware,
  
  and quite a bit faster on modern CPUs).
  Higher up, we go for a bottom-up construction of large transforms, foregoing
  
  the traditional per-transform call-return recursion chains. Instead, we always
  
  start at the bottom-most basis transform (in this case, a 32-point transform),
  
  and continue constructing larger and larger transforms until we return to the
  
  top-most transform.
  
  This way, we only touch the stack 3 times per a complete target transform :
  
  once for the 1/2 length transform and two times for the 1/4 length transform.
  The combination algorithm we use is a standard Split-Radix algorithm,
  
  as used in our C code. Although a version with less operations exists
  
  (Steven G. Johnson and Matteo Frigo's "A modified split-radix FFT with fewer
  
  arithmetic operations", IEEE Trans. Signal Process. 55 (1), 111–119 (2007),
  
  which is the one FFTW uses), it only has 2% less operations and requires at least 4x
  
  the binary code (due to it needing 4 different paths to do a single transform).
  
  That version also has other issues which prevent it from being implemented
  
  with SIMD code as efficiently, which makes it lose the marginal gains it offered,
  
  and cannot be performed bottom-up, requiring many recursive call-return chains,
  
  whose overhead adds up.
  We go through a lot of effort to minimize load/stores by keeping as much in
  
  registers in between construcring transforms. This saves us around 32 cycles,
  
  on paper, but in reality a lot more due to load/store aliasing (a load from a
  
  memory location cannot be issued while there's a store pending, and there are
  
  only so many (2 for Zen 3) load/store units in a CPU).
  
  Also, we interleave coefficients during the last stage to save on a store+load
  
  per register.
  Each of the smallest, basis transforms (4, 8 and 16-point in our case)
  
  has been extremely optimized. Our 8-point transform is barely 20 instructions
  
  in total, beating our old implementation 8-point transform by 1 instruction.
  
  Our 2x8-point transform is 23 instructions, beating our old implementation by
  
  6 instruction and needing 50% less cycles. Our 16-point transform's combination
  
  code takes slightly more instructions than our old implementation, but makes up
  
  for it by requiring a lot less arithmetic operations.
  Overall, the transform was optimized for the timings of Zen 3, which at the
  
  time of writing has the most IPC from all documented CPUs. Shuffles were
  
  preferred over arithmetic operations due to their 1/0.5 latency/throughput.
  On average, this code is 30% faster than our old libavcodec implementation.
  
  It's able to trade blows with the previously-untouchable FFTW on small transforms,
  
  and due to its tiny size and better prediction, outdoes FFTW on larger transforms
  
  by 11% on the largest currently supported size.
  

  • [DH] libavutil/tx.c
  • [DH] libavutil/tx_priv.h
  • [DH] libavutil/x86/Makefile
  • [DH] libavutil/x86/tx_float.asm
  • [DH] libavutil/x86/tx_float_init.c

• Encoder (codec png) not found for output stream #0:0 [duplicate]

  7 juin 2016, par Anubhav Dhawan

  This question already has an answer here :

  • ffmpeg & png watermark issue
    
    2 answers
    

  I’m trying to create a NodeJS app that converts a video into a GIF image.

  I’m using node-gify plugin for this purpose, which uses FFmpeg and GraphicsMagick.

  Here’s my sample code :

  var gify = require('./');
  
  var http = require('http');
  
  var fs = require('fs');
  
  
  
  var opts = {
  
    height: 300,
  
    rate: 10
  
  };
  
  
  
  console.time('convert');
  
  gify('out.mp4', 'out.gif', opts, function(err) {
  
    if (err) throw err;
  
    console.timeEnd('convert');
  
    var s = fs.statSync('out.gif');
  
    console.log('size: %smb', s.size / 1024 / 1024 | 0);
  
  });

  And here’s my console error :

  > gify@0.2.0 start /home/daffodil/repos/node-gify-master
  
  > node example.js
  
  
  
  /home/daffodil/repos/node-gify-master/example.js:24
  
    if (err) throw err;
  
             ^
  
  
  
  Error: Command failed: /bin/sh -c ffmpeg -i out.mp4 -filter:v scale=-1:300 -r 10 /tmp/IP5OXJZELd/%04d.png
  
  ffmpeg version 3.0.2 Copyright (c) 2000-2016 the FFmpeg developers
  
    built with gcc 4.8 (Ubuntu 4.8.4-2ubuntu1~14.04)
  
    configuration: --disable-yasm
  
    libavutil      55. 17.103 / 55. 17.103
  
    libavcodec     57. 24.102 / 57. 24.102
  
    libavformat    57. 25.100 / 57. 25.100
  
    libavdevice    57.  0.101 / 57.  0.101
  
    libavfilter     6. 31.100 /  6. 31.100
  
    libswscale      4.  0.100 /  4.  0.100
  
    libswresample   2.  0.101 /  2.  0.101
  
  Input #0, mov,mp4,m4a,3gp,3g2,mj2, from 'out.mp4':
  
    Metadata:
  
      major_brand     : mp42
  
      minor_version   : 1
  
      compatible_brands: mp42mp41
  
      creation_time   : 2005-02-25 02:35:57
  
    Duration: 00:01:10.00, start: 0.000000, bitrate: 106 kb/s
  
      Stream #0:0(eng): Audio: aac (LC) (mp4a / 0x6134706D), 8000 Hz, stereo, fltp, 19 kb/s (default)
  
      Metadata:
  
        creation_time   : 2005-02-25 02:35:57
  
        handler_name    : Apple Sound Media Handler
  
      Stream #0:1(eng): Video: mpeg4 (Advanced Simple Profile) (mp4v / 0x7634706D), yuv420p, 192x242 [SAR 1:1 DAR 96:121], 76 kb/s, 15 fps, 15 tbr, 600 tbn, 1k tbc (default)
  
      Metadata:
  
        creation_time   : 2005-02-25 02:35:57
  
        handler_name    : Apple Video Media Handler
  
      Stream #0:2(eng): Data: none (rtp  / 0x20707472), 4 kb/s (default)
  
      Metadata:
  
        creation_time   : 2005-02-25 02:35:57
  
        handler_name    : hint media handler
  
      Stream #0:3(eng): Data: none (rtp  / 0x20707472), 3 kb/s (default)
  
      Metadata:
  
        creation_time   : 2005-02-25 02:35:57
  
        handler_name    : hint media handler
  
  Output #0, image2, to '/tmp/IP5OXJZELd/%04d.png':
  
    Metadata:
  
      major_brand     : mp42
  
      minor_version   : 1
  
      compatible_brands: mp42mp41
  
      Stream #0:0(eng): Video: png, none, q=2-31, 128 kb/s (default)
  
      Metadata:
  
        creation_time   : 2005-02-25 02:35:57
  
        handler_name    : Apple Video Media Handler
  
  Stream mapping:
  
    Stream #0:1 -> #0:0 (mpeg4 (native) -> ? (?))
  
  Encoder (codec png) not found for output stream #0:0
  
  
  
      at ChildProcess.exithandler (child_process.js:213:12)
  
      at emitTwo (events.js:100:13)
  
      at ChildProcess.emit (events.js:185:7)
  
      at maybeClose (internal/child_process.js:827:16)
  
      at Socket.<anonymous> (internal/child_process.js:319:11)
  
      at emitOne (events.js:90:13)
  
      at Socket.emit (events.js:182:7)
  
      at Pipe._onclose (net.js:471:12)
  
  </anonymous>

  PS : I had a couple of problems installing FFmpeg on my Ubuntu 14.04.

  • First, FFmpeg is removed from Ubuntu 14.04 (legal issues AFAIK). But I managed to apt-get it through this.
  • Second, when I tried to ./configure (as mentioned in its README.md), I got this error - yasm/nasm not found or too old. Use --disable-yasm for a crippled build.. So I used ./configure --disable-yasm instead, and it (somehow) worked.

  Update #1

  After read this log a couple of times, I managed to produce a sample GIF from my mp4 file, by changing the command, which example.js tries to run :

  From

  ffmpeg -i out.mp4 -filter:v scale=-1:300 -r 10 /tmp/Lz43nx6wv1/%04d.png

  To

  ffmpeg -i out.mp4 -filter:v scale=-1:300 -r 10 out.gif

  But it’s still using command line, I need to do this by code.

  So I dived into the code and found that this wrong url is coming from the plugin’s index.js :

  ...
  
  
  
  // tmpfile(s)
  
    var id = uid(10);
  
    var dir = path.resolve('/tmp/' + id);
  
    var tmp  = path.join(dir, '/%04d.png');
  
  
  
  ...

  Is this an issue with the plugin, or am I doing something wrong here ?
  In any case, please put the correct stub here, because I don’t want to touch this part unless I know what I’m doing ?

  Update #2

  Now I installed zlib1g-dev, and then reinstalled both FFmpeg and graphicsMagick, and now I see this error :

  gm convert: No decode delegate for this image format (/tmp/ZQbEAynAcf/0702.png).

  Thanks in advance :)

• A/B Testing Platform for Piwik Analytics

  10 novembre 2016, par InnoCraft — Plugins, Press Releases

  We are very excited to launch A/B Testing – a powerful new product designed to help you maximise your success and increase your conversions.

  What is A/B Testing and why is it so important ?

  Experimenting with A/B Testing is an important marketing strategy that goes a long way in maximising your conversions.

  With A/B Testing you can start improving your conversions and sales in your existing website or app by creating different versions of a web page or mobile app screen to find out how to modify your website to optimize business success.

  How does A/B Testing work ?

  To explain how A/B Testing works, let’s take an example. Imagine one of your most important landing pages has been redesigned to hopefully get more email signups. Now you need to make sure the new version is “better” than the current version, or at least does result in less conversions. To compare the performance of these two versions, you can then run an A/B test whereby the A/B testing product tells you whether the redesigned version is producing more email signups compared to the original version.

  How it works is that visitors on your website during the experiment will see a randomly chosen version (variation) and A/B Testing makes sure that your users will always see the same variation for all subsequent visits. The behaviour and goals converted by each user is measured continuously by Piwik A/B Testing. The beautifully designed and easy to use dashboard shows how each variation impacts your success metrics such as conversions, sales or page views.

  When the A/B test is complete, you will then know which is the winning version getting you closer to reaching your goals ! A/B Testing includes an advanced statistical analysis engine to provide a reliable indicator of which of your versions is actually improving your bottom line.

  Examples of successful A/B tests

  Below are a just a few interesting A/B tests results from the article 7 Dead-Simple A/B Tests You Should Run on Your Homepage (HubSpot).

  Increase click on CTA button by 21%

  

  We actually conducted an A/B test on button color and found that changing the CTA button from green to red increased clicks by 21%. In the world of website optimization, that’s huge !

  47.7% more clicks on the navigation bar

  

  An A/B test comparing the phrases used on the navigation bar discovered that the menu link “How It Works” earned 47.7% more clicks than “Why Use Us.”

  5% more conversions using a different image

  

  Highrise, a small business CRM, A/B tested their website’s imagery by featuring two different people – Jocelyn on the left, and Michael on the right. They found Michael helped improve conversions by almost 5%.

  Changing the headline to increase conversions by 38%

  

  WhichTestWon ran an A/B test on AwayFind’s homepage and found that Version B increased signups by 38%. This could be because the headline is more concise and the subheader includes bolded words to clearly communicate the product’s value

  Where do I start testing ?

  Your testing strategy is tied to your overall business strategy. What do you want people to achieve on your website ? What are the goals that you want your users to convert ? What are your one or two Key Performance Indicators, and what are your targets for these KPIs ?

  When you start thinking of improving your targets you can begin by looking at the most important parts that can be improved :

  • What are your call to actions ?
  • Which are your most popular pages, and landing pages ?
  • What elements of my website is above the fold and can be seen by all visitors within a few seconds ?

  How do I get A/B Testing ?

  All premium plugins come with our 14 day money back guarantee and 1-click installation & updates. Customers get all product updates for free.

  Learn more about A/B Testing Get it now on the Marketplace

  A/B Testing is available for purchase and download on the Marketplace.

  If you are not using Piwik yet, you can also signup for a free trial of Piwik Cloud (including A/B Testing !).

  Have a question about this product ? Get in touch.