Pure-data Alternatives

pure-data reviews and mentions

• Pure Data as a plugin, with a new GUI
  14 projects | news.ycombinator.com | 22 Oct 2022

  > The other advantage is because these things were implemented in the 80s

  Pd was developed in the mid 90s

  > they are very computationally efficient

  Not as efficient as it could be, though. For example, instead of proper SIMD instructions, the DSP perform routines only use manual loop unrolling, praying that the compiler will auto-vectorize it.

  Finally, everything is single-threaded, leaving lots of performance on the table. FWIW, I have a PR for an asynchronous task API (https://github.com/pure-data/pure-data/pull/1357) and also a branch for multi-threaded DSP (https://github.com/Spacechild1/pure-data/tree/multi-threadin...).

• Show HN: Glicol(Graph-Oriented Live Coding Language) and DSP Lib Written in Rust
  15 projects | news.ycombinator.com | 7 Apr 2022

  FWIW, Pd and Max/MSP always had sample-level control in the sense that subpatches can be reblocked. For example, if you put a [block~ 1] object in a Pd subpatch, the process function will be called for every sample, so you can have single-sample feedback paths. Pd also has the [fexpr~] object which allows users to write FIR and IIR filters in a simple expression-syntax. Finally, Max/MSP offers the very powerful [gen~] object. You can check it out for inspiration (if you haven't already).

  Pd (and Max/MSP) also allow to upsample/resample subpatches, which is important for minimizing aliasing (caused by certain kinds of processing, such as distortion).

  Pd also uses the reblocking mechanism to implement FFT processing. The output of [rfft~] is just an ordinary signal that can be manipulated by the usual signal objects. You can also write the output to a table, manipulate it in the control domain with [bang~], and then read it back in the next DSP tick. IMO, this is a very powerful and elegant approach. SuperCollider, on the other hand, only supports a single global blocksize and samplerate which prevents temporary upsampling + anti-aliasing, severly limits single-sample feedback and leads to a rather awkward FFT implementation (you need dedicated PV_* objects for the most basic operations, such as addition and multiplication).

  Another thing to think about is multi-threaded DSP. With Supernova, Tim Blechmann miraculously managed to retrofit multi-threading onto scsynth. Max/MSP offers some support for multi-threading (IIRC, top level patches and poly~ instances run in parallel). Recently, I have been working on adding multi-threading to Pd (it's working, but still very much experimental): https://github.com/Spacechild1/pure-data/tree/multi-threadin.... If you design an audio engine in 2022, multi-threading should be considered from the start; you don't have to implement it yet, but at least leave the door open to do it at a later stage.

  ---

  I'm not sure how far you want to go with Glicol. I guess for the typical Algorave live coder all these things are probably not important. But if you want Glicol to be a flexible modern audio engine/library, you will have to think about FFT, upsampling, single-sample feedback, multi-processing etc. at some point. My advice is to not leave these things as an afterthought; you should at least think about it from the start while designing your engine - if you want to avoid some of the mistakes that other existing audio engines made. This is just a word of "warning" from someone having spent countless of hours in Pd and SuperCollider source code :-)

• How a Single Line of Code Made a 24-Core Server Slower Than a Laptop
  4 projects | news.ycombinator.com | 31 Dec 2021

  Great write up!

  What I like about Pd is that you can freely reblock and resample any subpatch. Want some section with single-sample-feedback? Just put a [block~ 1]. You can also increase the blocksize. Usually, this is done for upsampling and FFT processing. Finally, reblocking can be nested, meaning that you can reblock to 1024 samples and inside have another subpatch running at 1 sample blocksize.

  SuperCollider, on the other hand, has a fixed global blocksize and samplerate, which I think is one of its biggest limitations. (Needless to say, there are many things it does better than Pd!)

  ---

  In the last few days I have been experimenting with adding multi-threading support to Pd (https://github.com/Spacechild1/pure-data/tree/multi-threadin...). With the usual blocksize of 64 sample, you can definitely observe the scheduling overhead in the CPU meter. If you have a few (heavy-weight) subpatches running in parallel, the overhead is neglible. But for [clone] with a high number of (light-weight) copies, the overhead becomes rather noticable. In my quick tests, reblocking to 256 samples already reduces the overhead significantly, at the cost of increased latency, of course.

  ---

  Also, in my plugin host for Pd/Supercollider (https://git.iem.at/pd/vstplugin/) I have a multi-threading and bridging/sandboxing option. If the plugin itself is rather lightweight and the blocksize is small, the scheduling overhead becomes quite noticable. In Pd you can just put [vstplugin~] in a subpatch + [block~]. For the SuperCollider version I have added a "reblock" argument to process the plugin at a higher blocksize, at the cost of increased latency.

• A note from our sponsor - InfluxDB
  www.influxdata.com | 27 Apr 2024

  Get real-time insights from all types of time series data with InfluxDB. Ingest, query, and analyze billions of data points in real-time with unbounded cardinality. Learn more →