Continuing the jack/pipewire debate...

Discuss anything new and newsworthy! See http://planet.linuxaudio.org and https://libreav.org/news for more Linux Audio News!

Announcements of proprietary software may fit better in the Marketplace.


Moderators: raboof, MattKingUSA, khz

User avatar
Impostor
Established Member
Posts: 1333
Joined: Wed Aug 17, 2022 1:55 pm
Has thanked: 138 times
Been thanked: 344 times

Re: Food for thought -- continuing the jack/pipewire debate

Post by Impostor »

Audiojunkie wrote: Fri Jun 02, 2023 7:13 pm

I think we all need to get more involved in conversations with the developers to try to help them to understand better our needs and why we need them, and get them on board with truly making pipewire a true jack replacement.

God forbid we make 'em feel incompetent though.

User avatar
Audiojunkie
Established Member
Posts: 392
Joined: Thu Feb 21, 2019 4:27 pm
Has thanked: 382 times
Been thanked: 153 times

Re: Food for thought -- continuing the jack/pipewire debate

Post by Audiojunkie »

Impostor wrote: Fri Jun 02, 2023 6:36 pm
merlyn wrote: Fri Jun 02, 2023 6:29 pm

I'm not going to be using Pipewire any time soon, in the same way that Wayland is the new thing but I still use X11. Another negative is that x42 expressed his reservations. That's not a good sign to me. But ... a lot can happen before version 1.0.

Yeah. Pipewire could even become the nail in the coffin of Linux pro audio.
Don't take this as a prediction though.

I have serious doubts about that--I still think Pipewire is the future. However, I think if the developers don't understand more than superficially what we do and why it is important, then it will take much longer for low latency and 0 xruns to become a priority.

User avatar
Audiojunkie
Established Member
Posts: 392
Joined: Thu Feb 21, 2019 4:27 pm
Has thanked: 382 times
Been thanked: 153 times

Re: Food for thought -- continuing the jack/pipewire debate

Post by Audiojunkie »

Impostor wrote: Fri Jun 02, 2023 7:24 pm
Audiojunkie wrote: Fri Jun 02, 2023 7:13 pm

I think we all need to get more involved in conversations with the developers to try to help them to understand better our needs and why we need them, and get them on board with truly making pipewire a true jack replacement.

God forbid we make 'em feel incompetent though.

Yeah, I felt like I offended one of the guys I was communicating with when I was chatting (which wasn't my intention), but it does become a problem if they don't fully understand more than superficially these issues that are unique to the linux pro audio world and they shrug it off with little concern.

wtay
Established Member
Posts: 21
Joined: Mon Apr 17, 2023 9:56 am
Been thanked: 23 times

Re: Food for thought -- continuing the jack/pipewire debate

Post by wtay »

Audiojunkie wrote: Fri Jun 02, 2023 7:26 pm
Impostor wrote: Fri Jun 02, 2023 6:36 pm
merlyn wrote: Fri Jun 02, 2023 6:29 pm

I'm not going to be using Pipewire any time soon, in the same way that Wayland is the new thing but I still use X11. Another negative is that x42 expressed his reservations. That's not a good sign to me. But ... a lot can happen before version 1.0.

Yeah. Pipewire could even become the nail in the coffin of Linux pro audio.
Don't take this as a prediction though.

I have serious doubts about that--I still think Pipewire is the future. However, I think if the developers don't understand more than superficially what we do and why it is important, then it will take much longer for low latency and 0 xruns to become a priority.

You want the lowest possible latency with as little xruns as possible, I think we all understand the requirement...

User avatar
Audiojunkie
Established Member
Posts: 392
Joined: Thu Feb 21, 2019 4:27 pm
Has thanked: 382 times
Been thanked: 153 times

Re: Food for thought -- continuing the jack/pipewire debate

Post by Audiojunkie »

Thank you Wim for popping in!!! But it's more than:

"You want the lowest possible latency with as little xruns as possible, I think we all understand the requirement..."

It's being able to maintain that lowest possible latency (at least equivalent to JACK) with as little xruns as possible (at least equivalent to JACK) "under a proper DAW workload". In addition, it's about consistency for MOST users with equivalent hardware, not just SOME users. Also, if there is any tuning involved to achieve that, we are wanting to know this too--we need clear documentation on this.

I really appreciate all the work you do, and your willingness to engage with us about these things. Please don't take offense to our concerns. Please see them as just passionate, rather than offensive. :)

wtay
Established Member
Posts: 21
Joined: Mon Apr 17, 2023 9:56 am
Been thanked: 23 times

Re: Food for thought -- continuing the jack/pipewire debate

Post by wtay »

Audiojunkie wrote: Fri Jun 02, 2023 7:45 pm

Thank you Wim for popping in!!! But it's more than:

"You want the lowest possible latency with as little xruns as possible, I think we all understand the requirement..."

It's being able to maintain that lowest possible latency with as little xruns as possible under a proper DAW workload. In addition, it's about consistency for MOST users with equivalent hardware, nut just some. Also, if there is any tuning involved to achieve that, we are wanting to know this too.

Yes, of course, a constant latency under various load.

I think the main question is why PipeWire does not deliver that yet. My answer is to be patient, it's a lot of work and there are only so many hours in a day. I believe the design can deliver it eventually, you just need to be patient. I have a few more ideas to work out...

I think we can conclude that the timer based scheduling doesn't work reliably enough for pro audio so we're going to have to add IRQ based scheduling like what jack does. It just takes some time to nicely design this into the existing stuff. It's not finished yet, we're working on it...

User avatar
Audiojunkie
Established Member
Posts: 392
Joined: Thu Feb 21, 2019 4:27 pm
Has thanked: 382 times
Been thanked: 153 times

Re: Food for thought -- continuing the jack/pipewire debate

Post by Audiojunkie »

wtay wrote: Fri Jun 02, 2023 7:55 pm
Audiojunkie wrote: Fri Jun 02, 2023 7:45 pm

Thank you Wim for popping in!!! But it's more than:

"You want the lowest possible latency with as little xruns as possible, I think we all understand the requirement..."

It's being able to maintain that lowest possible latency with as little xruns as possible under a proper DAW workload. In addition, it's about consistency for MOST users with equivalent hardware, nut just some. Also, if there is any tuning involved to achieve that, we are wanting to know this too.

Yes, of course, a constant latency under various load.

I think the main question is why PipeWire does not deliver that yet. My answer is to be patient, it's a lot of work and there are only so many hours in a day. I believe the design can deliver it eventually, you just need to be patient. I have a few more ideas to work out...

I think we can conclude that the timer based scheduling doesn't work reliably enough for pro audio so we're going to have to add IRQ based scheduling like what jack does. It just takes some time to nicely design this into the existing stuff. It's not finished yet, we're working on it...

This is VERY heartening news! Thank you again for responding and addressing our concerns about pipewire!!! I really appreciate it!! Do you have a rough idea as to when we might see that change? Are we talking about 2 or more years away, or are we talking about less than a year? I'm just looking for a rough timeline (nothing to hold you to).

Thank you again!!

novalix
Established Member
Posts: 97
Joined: Wed Aug 11, 2021 1:12 pm
Has thanked: 6 times
Been thanked: 35 times

Re: Food for thought -- continuing the jack/pipewire debate

Post by novalix »

Had a good laugh at the it-is-all-debians-fault part. Although i don 't know what "mispachaging" really means.

User avatar
Impostor
Established Member
Posts: 1333
Joined: Wed Aug 17, 2022 1:55 pm
Has thanked: 138 times
Been thanked: 344 times

Re: Food for thought -- continuing the jack/pipewire debate

Post by Impostor »

Audiojunkie wrote: Fri Jun 02, 2023 8:29 pm

Do you have a rough idea as to when we might see that change? Are we talking about 2 or more years away, or are we talking about less than a year? I'm just looking for a rough timeline (nothing to hold you to).

Patience I have in spades. I just hope the same is true of distro publishers.

wtay
Established Member
Posts: 21
Joined: Mon Apr 17, 2023 9:56 am
Been thanked: 23 times

Re: Food for thought -- continuing the jack/pipewire debate

Post by wtay »

Audiojunkie wrote: Fri Jun 02, 2023 8:29 pm

Do you have a rough idea as to when we might see that change? Are we talking about 2 or more years away, or are we talking about less than a year? I'm just looking for a rough timeline (nothing to hold you to).

I'm wrapping up netjack2 support, I added ffado support and a jackdbus module... After that I'll try to revisit the ALSA driver rework. We're talking less than 6 months, I plan to release 1.0 later this year and I would like to have feature parity (or almost) with JACK2, including an equivalent ALSA driver.

User avatar
Audiojunkie
Established Member
Posts: 392
Joined: Thu Feb 21, 2019 4:27 pm
Has thanked: 382 times
Been thanked: 153 times

Re: Continuing the jack/pipewire debate...

Post by Audiojunkie »

Very cool! Thank you for responding to our group!!

folderol
Established Member
Posts: 2069
Joined: Mon Sep 28, 2015 8:06 pm
Location: Here, of course!
Has thanked: 224 times
Been thanked: 400 times
Contact:

Re: Continuing the jack/pipewire debate...

Post by folderol »

This is indeed very good news.
Does pipewire handle MIDI as well, or is the left alone?

The Yoshimi guy {apparently now an 'elderly'}
x42
Established Member
Posts: 35
Joined: Sun Apr 29, 2012 2:59 pm
Has thanked: 2 times
Been thanked: 5 times

Re: Food for thought -- continuing the jack/pipewire debate

Post by x42 »

wtay wrote: Fri Jun 02, 2023 7:55 pm

I think we can conclude that the timer based scheduling doesn't work reliably enough for pro audio

macOS CoreAudio (and CoreMIDI) begs to differ.

wtay
Established Member
Posts: 21
Joined: Mon Apr 17, 2023 9:56 am
Been thanked: 23 times

Re: Food for thought -- continuing the jack/pipewire debate

Post by wtay »

x42 wrote: Sat Jun 03, 2023 1:28 pm
wtay wrote: Fri Jun 02, 2023 7:55 pm

I think we can conclude that the timer based scheduling doesn't work reliably enough for pro audio

macOS CoreAudio (and CoreMIDI) begs to differ.

The concept certainly works very well on Linux as well for some hardware, not so much for other hardware. My behringer works very reliably with a 32 sample period at 4.9 ms roundtrip latency using the timer based scheduling, JACK gets 4.6ms out of the same hardware with IRQ based scheduling. On my laptop I can't reliably get the same results and need a larger period. Some people report +10ms more latency with timers (and the same periods) than with IRQ based scheduling.

I believe it is driver/hardware dependent and not necessarily a software problem. macOS uses a different kernel and drivers and they control a large part of the hardware so the comparison is not entirely fair.

Gps
Established Member
Posts: 1136
Joined: Mon Mar 09, 2015 3:09 pm
Has thanked: 331 times
Been thanked: 112 times

Re: Food for thought -- continuing the jack/pipewire debate

Post by Gps »

Audiojunkie wrote: Fri Jun 02, 2023 7:29 pm
Impostor wrote: Fri Jun 02, 2023 7:24 pm
Audiojunkie wrote: Fri Jun 02, 2023 7:13 pm

I think we all need to get more involved in conversations with the developers to try to help them to understand better our needs and why we need them, and get them on board with truly making pipewire a true jack replacement.

God forbid we make 'em feel incompetent though.

Yeah, I felt like I offended one of the guys I was communicating with when I was chatting (which wasn't my intention), but it does become a problem if they don't fully understand more than superficially these issues that are unique to the linux pro audio world and they shrug it off with little concern.

This happened with me and a LMMS dev. I totally unintended offended him.
He did not understood my question 100% and took it as me unfairly criticizing LMMS.

To keep it short this was before I learned to use plugins like reverb.

I was asking why the presets of an Yamaha DX7 sounded better then the LMMS sounds.
I also still think he never messed with an Yamaha DX7 and cubase on an Atari ST.
Then he would not be offended, and probably could have answered my question.
To his defense, the stock LMMS drum sounds do sound damn good on my studio monitors.

I also know I can be quite direct, and had to tune that down in online communication. :)

Post Reply