I'll start a ticket.  Thanks for confirming!  And apologies for botching up
the list reply.


Brady Sharp
[email protected]

On Wed, Sep 10, 2025, 7:31 PM Christof Ressi <[email protected]> wrote:

> (Bringing it back to the list.)
>
> There has been a performance regression in Pd 0.56-test1 that sounds very
> similar to yours: https://github.com/pure-data/pure-data/issues/2654.
> However, that one has been resolved since the final 0.56 release. (I've
> just tested to make sure.)
>
> If I create a new empty GOP, and put 4 of these buttons in it, and
> duplicate that GOP 32 times, I start getting glitches when mousing around.
>
> I just made a patch with 64 GOP subpatches, each containing 4 instances of
> your abstraction. Then I tried to quickly move the mouse over the canvas.
> Here are my results:
>
> Pd 0.56-1: [iemlib/dsp~] reports an average load of 7-10% and occasional
> peak loads of up to 28%.
>
> Pd 0.55-2: [iemlib/dsp~] reports an average load of 0.5-1% and occasional
> peaks of up to 3%.
>
> In both cases, the average idle load (with DSP turned on) is 0.5-1%.
>
> (Tested on a Windows 10 laptop with an AMD Ryzen 7 PRO 5850U.)
>
> So there is indeed a noticable performance regression! Please open a
> ticket on GitHub!
>
> Christof
> On 11.09.2025 00:56, Brady Sharp wrote:
>
> So yes, I've got LOTS of GOP's, and GOP's within GOP's, probably up to 3
> layers deep in some places.
> I've eliminated the input streaming factor and just added a [osc~ 440]
> going direct to [dac~] within the patch.
>
> In PD-0.56-1:
> Running iem/dsp~, it shows a 38%-ish CPU load, and a 35%-ish Mean CPU Load
> when idle.
> As I mouse over the canvas, it goes crazy, like 190% Peak, 85% Mean, adn
> audio glitches, until I stop moving the mouse, or move off the canvas.
> But the oddest part is, I can freely drag around these multi-layer GOP's,
> and there are no audio glitches and the DSP metering stays the same.
> It's only when moving the mouse without dragging anything that it has the
> issue.
>
> In PD-0.55-1:
> DSP peaks around 40%-ish, with a mean of about 36%-ish, no matter what I
> do with the mouse.  Audio stays consistent.
>
> Going back to PD-0.56-1:
> I put in the [osc~ 440] into the [dac~] again, and just started deleting
> swaths of my GOP abstractions.
> Picture these as modular guitar effects.  Lets say I loosely have about 32
> of them (with vastly different DSP requirements, but I really don't think
> that's the issue).  These are GOP's within GOP's, mostly 2 levels deep for
> the most part.
> I have to reduce them to about 6 before I stop getting dropouts while
> moving the mouse.
>
> After starting from scratch with a new patch with just an [osc~]->[dac~],
> I think I've found an abstraction that triggers it.  It's basically a Midi
> Assign button.  It IS a very small GOP abstraction, and of course I've got
> probably 120 of them in my performance patch.  Basically [asnbtn 5] creates
> a button labelled "5." on the GOP,  that can be used to assign a midi
> command to an effect parameter.  There really isn't much to it.  there is
> no audio DSP on it at all, just message routing.  If I create a new empty
> GOP, and put 4 of these buttons in it, and duplicate that GOP 32 times, I
> start getting glitches when mousing around.
>
> Again, ONLY in pd-0.56-1.
>
> Abstraction is attached for reference.
>
> If this is appropriate to post to the github, I'm happy to, I'm just
> making sure it's nothing exclusive to my PC.
> I'm happy for more testing suggestions if there are any.
>
> Thanks,
>
> Brady
>
>
> On Wed, Sep 10, 2025 at 2:27 PM Christof Ressi <[email protected]>
> wrote:
>
>> Hi,
>>
>> Only starting in 0.56-1, whenever I move my mouse cursor, I'm getting
>> audio glitching that corresponds to the mouse movement as it moves across
>> the PD Canvas.
>>
>> Does it only happen if the canvas contains many objects or many GOPs? Or
>> does it also happen with an empty canvas?
>>
>> What's the CPU usage of the "pd" process and the "wish" process when
>> you're moving the mouse?
>>
>> Christof
>> On 10.09.2025 19:29, Brady Sharp wrote:
>>
>> Hi Folks,
>>
>> Only starting in 0.56-1, whenever I move my mouse cursor, I'm getting
>> audio glitching that corresponds to the mouse movement as it moves across
>> the PD Canvas.
>> Ironically, this is just when hovering the mouse over the canvas and
>> moving it.  If I pick up a GOP subpatch and drag it, the audio glitching
>> goes away!  If I move the mouse outside of the PD Canvas, it also goes away.
>> I'm running on an old Core 2 duo 6300 laptop (8gb RAM) with a reasonably
>> complicated patch (takes about 10 seconds to load), but adding any other
>> apps in the background doesn't seem to affect it in any way.
>> Increasing audio delay or audio block size settings has no effect
>> either.  PD's DPS block size always stays at 64.  I tried recreating the
>> preferences file as well.
>> Ruled out a USB Mouse and a USB Hub and all MIDI devices as well.  Those
>> don't seem to matter.
>>
>> Running latest Manjaro with XFCE desktop (X11) with pd 0.56-1, and using
>> Pipewire (JACK)
>>
>> 0.55-1 still works great and has none of these symptoms under the same
>> conditions.
>> Both versions are compiled, ie. not using the pd packaged with Manjaro.
>>
>> Because of the behavior above, it leads me to think something has changed
>> with either PD or TCL...
>>
>> Any thoughts?
>>
>> Thanks in advance!
>>
>>
>> --
>> Brady Sharp
>> [email protected]
>> bradysharp.com
>>
>> [email protected] - the Pure Data 
>> mailinglisthttps://lists.iem.at/hyperkitty/list/[email protected]/message/SD4MIGEPYJY2B6UJRGZTVN6CI6YJZH6A/
>>
>> To unsubscribe send an email to [email protected] mailing list
>> UNSUBSCRIBE and account-management -> https://lists.iem.at/
>>
>> ---
>> [email protected] - the Pure Data mailinglist
>>
>> https://lists.iem.at/hyperkitty/list/[email protected]/message/4OHWZI4OTVHSWCTRHFB5SIAKJGILVJKB/
>>
>> To unsubscribe send an email to [email protected] mailing list
>> UNSUBSCRIBE and account-management -> https://lists.iem.at/
>
>
>
> --
> Brady Sharp
> [email protected]
> bradysharp.com
>
> ---
> [email protected] - the Pure Data mailinglist
>
> https://lists.iem.at/hyperkitty/list/[email protected]/message/XCFZFDV6BSCSR7DJEI3IZVUCVP32ACWA/
>
> To unsubscribe send an email to [email protected] mailing list
> UNSUBSCRIBE and account-management -> https://lists.iem.at/
---
[email protected] - the Pure Data mailinglist
https://lists.iem.at/hyperkitty/list/[email protected]/message/L2JWJ3EBIW33GH5TB3KKO5EFGBXGHSFW/

To unsubscribe send an email to [email protected] mailing list
UNSUBSCRIBE and account-management -> https://lists.iem.at/

Reply via email to