Re: Win11 ??
"David McClain (as dbm at refined-audiometrics dot com)" <[email protected]>
| Newsgroups | gmane.lisp.lispworks.general |
|---|---|
| Message-ID | <[email protected]> |
Sorry… my incoming data sample rate is 48 kHz, but I asked Win11 to downsample the data coming into the collection buffers to 8 kHz. The data is completely band-limited to 3 kHz, so no problem there. Hence, why at 48 kHz, it takes 512 ms to fill a buffer of 4096 samples. Win11 might have its attention diverted by multiples of 20 micro-sec, but I’m seeing dropouts of entire samples at the 8 kHz rate, which corresponds to multiples of 125 micro-sec. - DM > On Sep 3, 2025, at 05:25, David McClain <[email protected]> wrote: > > I have noticed a regular disruption of long running code, performing realtime data collection, when running on Win11. I haven’t yet been able to compare against MacOS because I don’t yet have an audio reader on the Mac. > > I cannot fathom anything in the Lisp code producing this artifact (see attached pic), which occurs about every 76 minutes. I set a timer to remind myself to watch closely at the next predicted event, and sure enough, the data collection was disrupted. The spectral display showed what looked like a data collection with dropouts - phase disruptions amid the contiguous data vector samples. > > My processing is totally asynchronous and only occurs when a full buffer of data arrives. So the system thought it had a full buffer of data and called the processing. My processing takes at most 20-50 ms, while filling a buffer takes exactly 512 ms. > > Buffering is 5 deep, so the hand-off to the next buffer isn’t the problem here. If it were then I’d be seeing this issue much more frequently. (about every 2.5 minutes) > > My suspicion is that there is some CRON event in Win11 happening every 76 minutes, which diverts the attention of the processor for some extended period. My sample rate is 48 kHz, so anything pulling attention longer than multiples of 20 microsec could produce phase discontinuities in the data collection. > > This is all running on an AMD Ryzen 5 5600G at 3.9 GHz - multicore CPU with very light loading, about 2-5% max CPU load during data collection. The actual data collection happens in an external C routine running M$ Multimedia handlers, so should be granted realtime performance. > > Anyone else seeing anything like this? > > - DM<Screenshot 2025-09-03 050746.jpg> _______________________________________________ Lisp Hug - the mailing list for LispWorks users [email protected] http://www.lispworks.com/support/lisp-hug.html