| ↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |
| Who | What | When |
|---|---|---|
| *** | rubdos has quit IRC (Ping timeout: 245 seconds) | [04:53] |
| ......................................... (idle for 3h24mn) | ||
| fling has quit IRC (Quit: ZNC 1.7.2+deb3 - https://znc.in) | [08:17] | |
| ..... (idle for 21mn) | ||
| hverkuil | ribalda: thanks! I have some work planned that can go on top of your upcoming patch, so I am looking forward to this.
ndufresne: V4L2_FMT_FLAG_DYN_RESOLUTION was introduced instead of relying on failing event subscription because it is format specific. I.e. a driver that supports MPEG2 and H264 might support dyn res changes for one, but not the other. But event subscription would always work since it is not per pixelformat, it is global. Suppose dyn res change is not supported for MPEG2. The event subscription would still succeed, you just would never get the event for that pixelformat. | [08:38] |
| ribalda | hverkuil: If you want to expedite the job I can start by sending the patches for the kernel. | [08:41] |
| hverkuil | ndufresne: sorry, that was a reply to a quite old question. Probably obsolete by now.
ribalda: I'd like to have something by the end of next week, if possible. jernej: dynamic arrays are not supported in the control framework. it's quite hard to implement that. Or to define the precise semantics. | [08:45] |
| jernej | ok, so I guess we will have to limit entry point size so some value
*array size | [08:49] |
| hverkuil | jernej: is this a control that will change a lot (i.e. every frame), or something that is fairly static? | [08:49] |
| jernej | I think it's per frame, but I have to check again | [08:50] |
| hverkuil | and how big will it be? | [08:50] |
| jernej | 4 bytes per entry | [08:50] |
| hverkuil | how many entries in the worst case? | [08:51] |
| jernej | I'm using my own hack with 256 entries maximum and it seems to be enough
and BSP Allwinner driver seems to have same limitation | [08:51] |
| hverkuil | OK. That's not too bad. | [08:51] |
| jernej | but I'm not sure if this is true maximum or just one value which seems to be good enough | [08:52] |
| hverkuil | how many entries are typically needed when decoding 1080p? I'm trying to get a feel of the overhead involved. | [08:52] |
| jernej | I didn't collect any statistics | [08:53] |
| hverkuil | It's not urgent, but if you have the opportunity to get some statistics, then that would be nice. | [08:54] |
| jernej | ok, I'll check my sample videos, but only a few are 1080p
mostly less or 4k | [08:55] |
| hverkuil | well, for 4k it is interesting as well to get statistics.
It's mainly to have an idea of how much overkill 256 elements is in practice. BTW, thank you very much for all your efforts! Much appreciated. | [08:55] |
| jernej | I guess we can still put pointer in control...
no problem :) I did most Cedrus improvements long ago, but only now it's appropriate time to mainline it | [08:57] |
| hverkuil: will be there any report from ELCE media meeting for those who didn't attend? | [09:03] | |
| hverkuil | Yes. It was quite productive. I plan/hope to post reports for the codecs and 'future work' sessions by the end of next week.
a libcamera report will be up to pinchartl. | [09:04] |
| jernej | ok, great to hear that it was productive | [09:05] |
| hverkuil | of particular interest to you will be that we decided on how to tackle the final issues that are blocking the stateful encoding spec to be merged.
It should be possible to get that done before the end of this year. | [09:06] |
| ................ (idle for 1h16mn) | ||
| ribalda | hverkuil: hi!, fast question. for "struct v4l2_ext_control", shall I add the field *p_area ?
otherwise userspace has to make a casting | [10:22] |
| hverkuil | yes | [10:22] |
| ribalda | ok | [10:23] |
| ............. (idle for 1h4mn) | ||
| hverkuil: I just sent the patchset for v4l-utils and kernel. The first patch from v4l-utils should not be merged, you need to take the "real" headers from media/master | [11:27] | |
| ............................................................. (idle for 5h2mn) | ||
| *** | b-rad has quit IRC (Remote host closed the connection) | [16:29] |
| ............ (idle for 59mn) | ||
| jernej | hverkuil: kwiboo: I think we should make a note in stateless codecs docs that v4l2_m2m_buf_copy_metadata() should be called after references are resolved
otherwise it gets tricky with interlaced content, where you decode top and bottom field in same capture buffer | [17:28] |
| ..................................................................... (idle for 5h40mn) | ||
| *** | neg has quit IRC (Quit: quit) | [23:09] |
| ↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |