00:37 <mtrbot-ml> [mattermost] <astro> I now have an async executor switching between yielding tasks on the zynq

00:54 <mtrbot-ml> [mattermost] <sb10q> @astro not yet but I was planning to do it today

00:54 <mtrbot-ml> [mattermost] <astro> it might be not necessary after all

00:55 <mtrbot-ml> [mattermost] <astro> the rdid data looks good to me

00:55 <mtrbot-ml> [mattermost] <sb10q> very good

00:55 <mtrbot-ml> [mattermost] <astro> could you perhaps verify with vivado tools?

01:27 <mtrbot-ml> [mattermost] <astro> what are the requirements regarding SMP-safety? are we going to run async code on both cores?

01:32 <mtrbot-ml> [mattermost] <sb10q> no, only on the comms (master) CPU

05:35 <mtrbot-ml> [mattermost] <sb10q> @astro do you still need the nrf52 board?

07:10 <mtrbot-ml> [mattermost] <sb10q> @astro you now have 10k and 5ohm on thermostat ch0 and nothing on ch1

13:53 <airwoodix> rjo, sb10q: is there any experience in seamless compression of ARTIQ datasets? (e.g. lossless compression of camera images) Does it make sense at all?

13:54 <airwoodix> Naive expectation: same interface (set_dataset, get_dataset, maybe even append_to_dataset) but with lossless compression in the background

13:55 <airwoodix> *seamless -> transparent

15:18 <rjo> airwoodix : not specifically. You can certainly use opaque binary array with compressed data and the algorithm of Our choice. But my guess would be that improving pyon, simpyco and the h5 layer are worthier targets.

15:23 <airwoodix> rjo: thanks. The primary goal is to reduce the disk size of the hdf5 files without throwing data away.

15:25 <tpw_rules> doesn't hdf5 have several available compression options?

15:25 <rjo> airwoodix i think there are some specific lossless compression options for datasets in h5 or the h5 file itself. Intrinsic to h5.

15:26 <rjo> There is also an issue in ARTIQ proposing a more spce efficient storage for small datasets.

15:29 <airwoodix> Sounds great, thanks. I'll have a look.

16:20 <rjo> airwoodix: what's the image size nowadays? back in zurich we had 512x512 EMCCD shots, three per camera, three cameras, per minute. that was not much storage space back then.

16:54 <Eldra> Sometimes I have the following error : Urukul AD9910 AUX_DAC mismatch. And i cannot run codes anymore. Is there a way to reset/reboot everything without disconnected power ?

16:55 <Eldra> artiq_coremgmt --device-db ../config/device_db.py reboot _ doesn't work

17:07 <rjo> The AD9910 DDS chips are fragile. There is an ARTIQ issue about that. They require a power cycle if they are in that state.

17:29 <Eldra> Yeah I saw it. So I need to power off the entier system.

18:18 <airwoodix> rjo: for this particular case, 1384x1032 so 2.9 MB per image raw from the camera (16 bits) but it gets worse if one wants to store pre-processed data on double precision floating point numbers...

20:25 <rjo> airwoodix: image based lossless compression is pretty different for integers than for HDR images. there is jpeg_xt or openexr for the latter. alternatively standard opaque binary compression algorithms. and always consider whether 8 bytes per pixel is actually required to represent SNR. I remember that for our thomas-fermi/BEC fits it wasn't.