06:19 <tomeu> ok, after a night of good sleep and some thinking time before getting up, finally got it

06:20 <tomeu> things are rock solid again with hw_job_limit=1

06:20 <tomeu> I had been assuming all this time that we were only submitting a single job to the HW at a time

06:21 <tomeu> austriancoder: you might want to try that out in CI

07:51 <tomeu> does anybody know of a recent forward port of galcore?

08:10 <austriancoder> tomeu: 6.1: https://github.com/nxp-imx/linux-imx .. i think they will switch sooner or later to 6.6

12:40 <tomeu> thanks!

15:31 <gbisson> actually there's already a 6.6 version for imx95 out there

15:31 <gbisson> https://github.com/nxp-imx/linux-imx/tree/lf-6.6.y-imx95-er1/drivers/mxc/gpu-viv

15:35 <gbisson> but I don't know where the libaries for it are (version seems to be 6.4.11p2.x)

16:12 <tomeu> oh, that's great

16:12 <tomeu> what I need are the patches for the forward port, let's see if I find them

16:29 <sravn> tomeu: Asking out of pure ignorance here.. How far are what you do from being able to run tensorflow? I ask because we have something (done by colleagues of mine, I do not know the code) that runs tensorflow on an i.MX8 and I wonder if we maybe in the future can benefit from your work

16:30 <sravn> They tell me the tensorflow code runs mostly/only on the CPU - most likely using a downstream kernel :-(