| ↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |
| Who | What | When |
|---|---|---|
| tfiga | ndufresne: controls to query profiles/levels - yes :)
I need to understand more relations between profiles and levels, though generally we already have menu controls for profiles of most codecs, so one can only report the supported values inside the menu for decoder | [01:49] |
| ........... (idle for 50mn) | ||
| ndufresne: the colorimetry is something I need to think a bit more too | [02:40] | |
| ................................. (idle for 2h44mn) | ||
| *** | TrishoolinKing has left | [05:24] |
| ................................ (idle for 2h39mn) | ||
| svarbanov | ndufresne: yes, venus can detect current profile/level and also can preliminary say the supported profile/levels but I'm not sure that this is implemented correctly now :( | [08:03] |
| ........................................................................ (idle for 5h56mn) | ||
| ndufresne | tfiga, sorry, wasn't clear, for profile/levels, we don't need new controls, all we need is to enable these controls on decoders, so that we can report the profile, and then the extended menu control query is enough to get the supported values
svarbanov, don't worry about implementation now, we are just trying to see how to fill the gap, I think there is very little to be done here I think there will be some static set of profiles/levels in the drivers, for the menu enumeration but I don't know what we do for profiles/level pair validation, this is pretty well defined in the spec, so we could do that once, or do that in userspace right now, I believe we just let anything be set, and fix it later | [13:59] |
| ............ (idle for 57mn) | ||
| *** | benjiG has left | [14:58] |
| ndufresne | pH5, my first attempt, https://paste.fedoraproject.org/paste/8r-jYpghOk~FEG12wrqVbg
It's a partial one, since we still rejects non-default colorimetry, also, we don't read-back the colorimetry later, we assume no csc taking place does coda encoder do csc ? probably not, everything is 4:2:0, it just supports different layout, which isn't changing the csc | [15:01] |
| pH5 | ndufresne: you are right, there is no csc support in coda. | [15:05] |
| lyakh | hi, has the meeting taken place last Tuesday? Any notes available yet? | [15:06] |
| ndufresne | which means a csc enabled encoder simply does not exist for now
lyakh, sure, let me find it for you lyakh, https://paste.fedoraproject.org/paste/wBj8qeYAe5ZBe6AZtsSOvQ | [15:06] |
| lyakh | ndufresne: thanks!
oh, that's just a copy of the mail(s) and a list of attendees :-) | [15:07] |
| ndufresne | oops, sorry, maybe it's the wrong one
wrong one indeed, sorry https://paste.fedoraproject.org/paste/8124R7GzfC7EVczOTpwwPA lyakh, this one ^ | [15:10] |
| lyakh | ndufresne: thanks! | [15:11] |
| ndufresne | lyakh, I also picked some screenshot of some of the presentation | [15:12] |
| lyakh | are they online? | [15:12] |
| ndufresne | not yet, I can zip that if you want | [15:12] |
| lyakh | sure, everything helps, thanks. I think usually presentation originals are collected on linuxtv... Let me check | [15:13] |
| ndufresne | https://people.collabora.com/~nicolas/LinuxMediaWorkshopScreenshots.zip
this has not been shared yet ^ but I'm sure mchehab will ask for the originals, I just took them to re-read them later | [15:15] |
| lyakh | no, don't see any originals yet, thanks for the screenshots | [15:16] |
| ............................ (idle for 2h18mn) | ||
| ndufresne | pH5, so, did you manage to reach 60fps on the drm driver ? I cannot get it to go faster then 30 here, even after dropping the sync in kmssink | [17:34] |
| ................. (idle for 1h20mn) | ||
| pH5 | ndufresne: is that only on the overlay plane (drmModeSetPlane) or also on the primary plane (drmModePageFlip)?
our setplane does a blocking commit, so that + vblank wait halves frame rate pageflip should work properly kmssink should really be updated to atomic modesetting | [18:54] |
| ndufresne | pH5, I haven't tried pageFilp yes, but I have commented out the vlank wait as this is a atomic driver using blocking setplane
pH5, yeah, but that's not as simple as it looks like | [18:55] |
| pH5 | I believe it | [18:56] |
| ndufresne | ayaka has started working on it, but it looks like something I'll have to take care at some point, he managed to get something working, but to make it really usable, there is another step
pH5, while you are around, quick question, how to I get the m2m converter to be probed ? do I need to add something in the dtb ? ndufresne just included the patchset you sent today, but am blocked on how to get the driver to load pH5, I confirm that pageFlip goes 60fps actually, both are 60 now, the delta is that I've moved away from linuxtv/master to linus master | [18:56] |
| pH5 | hmm, imx-media binds to the capture-subsystem node (which is in the dtsi). imx-media only completes if all subdevices (csis and everything that hangs off them in the DT) are loaded | [19:01] |
| ndufresne | that's what I thought, but for some reason, I don't get this m2m node ... | [19:02] |
| pH5 | VIDEO_IMX_CSI enabled?
do you get /dev/media0 ? | [19:02] |
| ndufresne | yes for CONFIG_VIDEO_IMX_CSI=y
ph, weird, I don't have /dev/media0 do I need to disable CSI to free the IC ? | [19:02] |
| pH5 | I think you'll also need VIDEO_MUX and OV5642 or disable them in the DT. | [19:04] |
| ndufresne | you mean need to *disable* these or to add these ? | [19:05] |
| pH5 | ndufresne: enable config options and load the drivers or add status = "disabled" property in &ipu1_csi0_mux and &ipu2_csi1_mux. | [19:07] |
| hverkuil | ezequielg: I updated v4l2-compliance in the v4l-utils git repo to improve the output.
Can you update it and run 'v4l2-compliance -M0 -v' and mail me the output? | [19:07] |
| pH5 | ndufresne: I think we currently need the CSIs enabled as they are used to link imx-media with the ipu-v3 instances. | [19:08] |
| ndufresne | pH5, ok, so mux was enabled, but not OV5642
adding ov didn't change anything (I don't have a sensor, ov5640 1-0040: ov5640_init_slave_id: failed with -6" so let's try option 2 | [19:08] |
| pH5 | ndufresne: you could also just disable the sensor in DT or remove the of-graph link between sensor and mux | [19:13] |
| ndufresne | ndufresne learning this, slowly
there is no &ipu1_csi0_mux, I got &ipu1_csi0_mux_from_parallel_sensor instead, same ? | [19:14] |
| pH5 | ndufresne: that is an endpoint inside a port inside that mux. the mux is in imx6q.dtsi. | [19:18] |
| ndufresne | imx6dl.dtsi in my case | [19:18] |
| pH5 | oh, ok
I'd add status = "disabled" to the ov5642 node. | [19:19] |
| ndufresne | was already there, added it for the other ov too, not sure it was required | [19:20] |
| pH5 | oh, there's more. yes, that needs to go, too, if you don't have sensors. | [19:21] |
| ndufresne | I really just has a bare sabre-lite, no extensions at all ...
ok, got /dev/media0 now and v4l2video8convert ndufresne should implement stable name for the one converter found ... | [19:21] |
| ↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |