> Is it possible to do audio processing using a GPU?
This thread fascinates me. A sibling comment addresses why it can't be done in realtime (barring some pedestrian-sounding hardware improvements), but the question of "how" fascinates me. I know a little bit about audio, I'm a classical musician with some experience with big halls, I know some physics, and I wrote a raytracer in JS back in the early aughts. Can we raytrace sound? It's a wave, it's got a spectrum, different materials reflect, refract, and absorb various parts of the spectrum differently... sounds pretty similar! The main differences I see is that we usually only have a handful of "single pixel cameras" (I expect to lose a bit of parallelism to that) and more problematic, we can't treat the speed of sound as infinite. I can imagine a cool hack to preprocess rays for a static scene, but my imagined approach breaks down entirely if the mic moves erratically.
We can already do this (without GPUs, and quite likely on them). It is done, for example, by a certain class of reverb algorithm which more or less literally "raytraces" the sound in room. It's not the most efficient way to implement a reverb (convolution algorithms are better for this). But it does allow for a large degree of control that convolution does not.
In addition, you may also consider physical modelling synthesis to be an example of this. Look up Pianoteq. It is a physical modelled piano that literally does all the math for the sound from a hammer of particular hardness striking a string of given size and tension at a particular speed, and then that sound generating resonances within the body of the piano and other strings, and then out into the room and interacting with microphone placement, lid opening etc. etc.
Most (not all) people agree that Pianoteq is the best available synthetic piano. It's all math, no samples.
There are similar physical modelling synth engines for many other kinds of sounds, notably drums. They are not particular popular because they use more CPU than samples and can be complicated to tweak. They also really benefit from a more sophisticated control surface/instrument than most people have access to (i.e. playing a physically modelled drum from a MIDI keyboard doesn't really allow you to get into the nuances of the drum).
Yes, sound can be raytraced - in realtime on current GPUs even. Using a ray model for sound waves is called Geometric Acoustics and works reasonably well in most cases, although it has a few shortcomings, mostly when dealing with low frequencies (room modes, diffraction).
I think it's too pessimistic to say it can't be done in real time on current GPUs. I think it should be possible to do reliable low latency readback with modern explicit graphics APIs. The harder problem is scheduling the audio work at high priority so it always finishes before the deadline, but this too has improved recently because VR compositors also have strict deadlines they must meet.
I remember hearing that Nvidia were working on an audio spatialiser using their RTX hardware/expertise, but I can't recall hearing about it after 2018 or so, which is where the project I was working on that I wanted audio spatialisation for ended.
There are a number of clever audio processing algorithms implemented on GPUs already. But that's not the issue at hand, which is doing so in a low-latency realtime context. GPUs are insanely powerful, and if you're just doing playback of material from disk, the tiny bit of extra latency they may cause isn't a problem, so go for whatever they can do. But if you're using the computer for synthesis or FX processing, you need better than GPUs can offer in terms of latency (for now).
This thread fascinates me. A sibling comment addresses why it can't be done in realtime (barring some pedestrian-sounding hardware improvements), but the question of "how" fascinates me. I know a little bit about audio, I'm a classical musician with some experience with big halls, I know some physics, and I wrote a raytracer in JS back in the early aughts. Can we raytrace sound? It's a wave, it's got a spectrum, different materials reflect, refract, and absorb various parts of the spectrum differently... sounds pretty similar! The main differences I see is that we usually only have a handful of "single pixel cameras" (I expect to lose a bit of parallelism to that) and more problematic, we can't treat the speed of sound as infinite. I can imagine a cool hack to preprocess rays for a static scene, but my imagined approach breaks down entirely if the mic moves erratically.