| ↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |
| Who | What | When |
|---|---|---|
| *** | svarbanov has quit IRC (Ping timeout: 248 seconds) | [01:21] |
| ......... (idle for 44mn) | ||
| gnurou has quit IRC (Quit: gnurou) | [02:05] | |
| ..................................................................................... (idle for 7h1mn) | ||
| bbrezillon | paulk-leonov: ping on H264 uAPI updates :P | [09:06] |
| paulk-leonov | bbrezillon: oh shit | [09:10] |
| mripard | (for the record, paulk-leonov did say this morning "oh, bbrezillon is going to be pissed I still haven't reviewed his patches") | [09:13] |
| bbrezillon | mripard: so, he's actually faking the surprise :)
paulk-leonov: I can offer you a beer if that helps ;) | [09:17] |
| paulk-leonov | bbrezillon: it's an ever-renewed surprise each time I think about it :)
bbrezillon: thanks for the motivational help hehe | [09:19] |
| ............................... (idle for 2h30mn) | ||
| bbrezillon: mhh I think I should quit drugs
I guess I'll dive back into it to get clear thoughts on the topic -- IIRC ndufresne also did some thinking about the DPB vs ref list thing (but too many context switches since that discussion took place) (on my side) | [11:49] | |
| .... (idle for 16mn) | ||
| bbrezillon | paulk-leonov: hehe
no problem, and I think your initial motivations (assuming I got them right) were sound | [12:08] |
| paulk-leonov | indeed I remember that DPB is more of a decoder-specific concept and that I wanted to stay closer to the bitstream rather than decoder-specific concepts, but not sure how it turned out in practice | [12:14] |
| *** | bbrezillon has quit IRC (Quit: WeeChat 1.9.1) | [12:28] |
| ...... (idle for 27mn) | ||
| hverkuil | ezequielg: I'm ready | [12:55] |
| ezequielg | hverkuil: hey!
give me one second. | [13:06] |
| hverkuil | ok | [13:07] |
| ezequielg | seems Boris left us :-)
hverkuil: so, regarding the VP8 uAPI. Boris was wondering about 64-bit alignments on fields. i think he was concerned about backwards compatibility on these uAPIs, when adding more fields. | [13:12] |
| hverkuil | fields of the compound controls?
such as struct v4l2_ctrl_vp8_frame_header? | [13:13] |
| ezequielg | yes | [13:15] |
| hverkuil | ezequielg: well, you can't add new fields to a struct used in a compound control. If new information is needed in the future, then you need to make new controls to store that data in. | [13:16] |
| ezequielg | in other words, we are explicitly dropping backwards compatibility. | [13:17] |
| hverkuil | Yes. | [13:17] |
| ezequielg | if i understand correctly, it's what we've done on the mpeg-2 and h264 cases. | [13:17] |
| hverkuil | I guess it is possible to add reserved fields, but there has to be a good reason for it.
correct. Basically all this information comes from the standards. If there are future extensions to the standards, then it is perfectly fine to add support for that via new controls. | [13:17] |
| ezequielg | yes.
ndufresne reviewed it before i submitted, and mentioned he was OK with it. | [13:18] |
| hverkuil | Those will then only be used by applications that actually want to use those extensions.
My main concern with these structs is that there should be no holes in them, either by choosing types for fields and ordering the fields in such a way that everything is aligned correctly, or by adding padding fields (still not really happy about that, but it looks that I'm losing that battle). And of course that the layout is the same for 32 and 64 bit architectures. | [13:19] |
| ezequielg | the layout is the same for 32 and 64 in this case. and i believe we could add a test somewhere to confirm it.
i mean continuously. | [13:22] |
| hverkuil | All the fields in those structs should also relate to the standard, not to any HW specifics. | [13:22] |
| ezequielg | right. | [13:23] |
| hverkuil | ezequielg: anything else? :-) | [13:28] |
| ezequielg | hverkuil: i think i will re-review the controls, just to make sure there aren't any hw specifics left. make sure there aren't any holes, and submit a v2 (there are some documentation style nitpicks). | [13:29] |
| hverkuil | Sounds good. | [13:30] |
| ezequielg | feel free to review this v1, perhaps you spot anything worth fixing. | [13:30] |
| hverkuil | Can anyone here test the m2m-deinterlace driver? | [13:42] |
| ezequielg | hverkuil: what do you have in mind?
i haven't used it before, but iirc, all it needs is a generic dmaengine. | [13:49] |
| hverkuil | I'm making a patch to switch it over to v4l2_fh as it is one of the few drivers that still doesn't use it. So I'm trying to find a way to test it.
I'll try first to see if I can get it to run on an intel laptop. | [13:52] |
| Hmm, no. The dma engine needs DMA_INTERLEAVE capability. | [13:58] | |
| ezequielg | imx has that.
i wonder if i can try on my wandboard - i am doing some experiments here, guess i can do more experiments as well and try m2m. if not deinterlace, i can do a simple memcpy. | [14:07] |
| hverkuil | I looked at imx, but I'm not sure if it is supported beyond i.MX2.
I think the TI pandaboard can do it, though. I'll try that. | [14:11] |
| ezequielg | well, a am335x would work. but i don't have one on my desk.
i have to stand and walk 1 meter... | [14:22] |
| *** | hverkuil has quit IRC (Quit: ZNC 1.7.2+deb2 - https://znc.in) | [14:33] |
| paulk-leonov | ezequielg: hverkuil: by the way, we have a gdb script (from mripard's h264 series) to check alignment on various archs
it's quite handy and easy to use :) ezequielg: also feel free to CC me for VP8/9 uapi | [14:39] |
| ezequielg | paulk-leonov: as i mentioned in the vp8 cover letter, https://gitlab.collabora.com/ezequiel/v4l2-ctrl-abi-check | [14:40] |
| paulk-leonov | I'd like to get a look at how it will concern cedrus | [14:40] |
| ezequielg | based on your script. | [14:40] |
| paulk-leonov | ah, hadn't seen the series | [14:40] |
| ezequielg | haven't Cced you? shame on me. | [14:43] |
| .......... (idle for 49mn) | ||
| jernej | paulk-leonov: did you do any work on Cedrus VP8?
I made initial driver and while register values match to CedarX driver/library, I think content of auxiliary buffers is not correct and thus I get garbage on screen I'm not sure how I can dump them from BSP driver making gdb script is one way, but I would like to avoid that | [15:32] |
| *** | benjiG has left | [15:39] |
| .............. (idle for 1h5mn) | ||
| paulk-leonov | ezequielg: mhh not sure I was CC-ed indeed, feel free to add me for the next revision :)
jernej: no work yet on my side, but I'm interested in it nice to know you got a head start :) but not sure how to go regarding tracing indeed | [16:44] |
| jernej | paulk-leonov: I will push my work later, if you are interested in it | [16:50] |
| ............ (idle for 59mn) | ||
| *** | _abbenormal has quit IRC (Quit: Leaving)
aballier has quit IRC (Ping timeout: 258 seconds) | [17:49] |
| .................... (idle for 1h39mn) | ||
| Muzer has quit IRC (Ping timeout: 252 seconds) | [19:32] | |
| ......... (idle for 40mn) | ||
| APic has quit IRC (Ping timeout: 245 seconds) | [20:12] | |
| .................. (idle for 1h27mn) | ||
| jernej has quit IRC (Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net) | [21:39] | |
| ......................... (idle for 2h2mn) | ||
| mmattice has quit IRC (Ping timeout: 244 seconds) | [23:41] | |
| ↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |