| ↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |
| Who | What | When |
|---|---|---|
| *** | digetx has joined #linux-media | [02:36] |
| ........ (idle for 38mn) | ||
| KitsuWhooa has quit IRC (Quit: Unable to handle kernel NULL pointer dereference at null)
KitsuWhooa has joined #linux-media | [03:14] | |
| .......... (idle for 45mn) | ||
| eelstrebor has quit IRC (Quit: Ex-Chat) | [04:00] | |
| ......... (idle for 42mn) | ||
| Mo has joined #linux-media | [04:42] | |
| ten1572377432463050661821 has quit IRC (Remote host closed the connection)
ten1572377432463050661821 has joined #linux-media | [04:51] | |
| ...... (idle for 28mn) | ||
| jmassot_ has quit IRC (Remote host closed the connection) | [05:19] | |
| ..... (idle for 23mn) | ||
| Artea has quit IRC (Remote host closed the connection) | [05:42] | |
| ....... (idle for 34mn) | ||
| frieder has joined #linux-media | [06:16] | |
| GBenji has joined #linux-media
shoragan has joined #linux-media Mo has quit IRC (Quit: Leaving) | [06:28] | |
| ...... (idle for 25mn) | ||
| Mo has joined #linux-media | [06:55] | |
| ao2 has joined #linux-media | [07:05] | |
| .... (idle for 17mn) | ||
| svarbanov has joined #linux-media | [07:22] | |
| svarbanov_ has quit IRC (Ping timeout: 480 seconds) | [07:28] | |
| sgruszka has joined #linux-media | [07:36] | |
| ..... (idle for 21mn) | ||
| thansen is now known as Guest2582
thansen has joined #linux-media Guest2582 has quit IRC (Ping timeout: 480 seconds) | [07:57] | |
| tomba_ has joined #linux-media
hverkuil1 has joined #linux-media hverkuil has quit IRC (Read error: Connection reset by peer) hverkuil1 is now known as hverkuil hverkuil has left Artea has joined #linux-media tomba_ has quit IRC () tomba has joined #linux-media | [08:06] | |
| tomba has quit IRC ()
tomba has joined #linux-media | [08:22] | |
| ........ (idle for 36mn) | ||
| svarbanov_ has joined #linux-media | [08:59] | |
| svarbanov has quit IRC (Ping timeout: 480 seconds) | [09:05] | |
| ........ (idle for 39mn) | ||
| darkapex has quit IRC (Remote host closed the connection)
darkapex has joined #linux-media | [09:44] | |
| ........... (idle for 50mn) | ||
| svarbanov_ has quit IRC (Ping timeout: 480 seconds) | [10:35] | |
| svarbanov_ has joined #linux-media | [10:48] | |
| ..... (idle for 22mn) | ||
| svarbanov has joined #linux-media | [11:10] | |
| prabhakalad | tomba: Continuing from https://patchwork.kernel.org/project/linux-renesas-soc/cover/20240904210719.52466-1-prabhakar.mahadev-lad.rj@bp.renesas.com/#26014467 bascially I want RZ/G2L CRU to support capturing from all the VCs and not just VC0. | [11:10] |
| *** | svarbanov_ has quit IRC (Read error: Connection reset by peer) | [11:11] |
| pinchartl | prabhakalad: the OV5645 can output a single VC at a time. if the goal is only to test the implementation on the CRU side, I would hardcode the VC in the ov5645 driver for testing purpose instead of making it configurable from userspace | [11:12] |
| *** | sgruszka has quit IRC (Ping timeout: 480 seconds) | [11:12] |
| prabhakalad | pinchartl: OK, but any reason why we dont want to allow the ov5645 sensor driver to be configurable from user space? | [11:15] |
| tomba | prabhakalad, what is the use case? how does the user know what VC to configure? | [11:17] |
| pinchartl | prabhakalad: we currently don't have an API to assign a specific VC from userspace, because that has never been needed
if there's a good use case, then we can design one | [11:18] |
| prabhakalad | pinchartl: also how will the rz/g2l cru driver know on which VC the data is coming from if I hardcode it in the ov5645 driver? | [11:18] |
| tomba | prabhakalad, the driver calls get_frame_desc on ov5645 | [11:19] |
| pinchartl | what Tomi said :-) the ov5645 driver tells which VC it uses through .get_frame_desc() | [11:21] |
| tomba | pratyush, the routing & streams describe the logical side of this all, and then at stream start time get_frame_desc can be used to get the hardware level details of the streams | [11:21] |
| prabhakalad | ok got you, by hardcoding I misunderstood that you meant just write the regsiter and I didnt consider about get_frame_desc :)
pinchartl: tomba: thanks for the inputs, I'll update patch 12/12 accordingly. Do you have any other comments on rest of the other patches? | [11:23] |
| pinchartl | I haven't had time to read them yet I'm afraid
so no comments yet :-) | [11:30] |
| tomba | prabhakalad, same here, sorry. I only had a quick look, mainly at the intro mail =) | [11:32] |
| prabhakalad | no worries, I'll rework on the series and send a v2. | [11:33] |
| .......... (idle for 49mn) | ||
| *** | mvaittin has joined #linux-media
tales-aparecida275 has quit IRC () tonyk has quit IRC (Quit: The Lounge - https://thelounge.chat) | [12:22] |
| tonyk has joined #linux-media | [12:30] | |
| ...... (idle for 26mn) | ||
| mvaittin has quit IRC (Ping timeout: 480 seconds) | [12:56] | |
| eelstrebor has joined #linux-media
gnuiyl_ has joined #linux-media Mo has quit IRC (Remote host closed the connection) gnuiyl has quit IRC (Ping timeout: 480 seconds) | [13:08] | |
| sgruszka has joined #linux-media
eelstrebor has quit IRC (Quit: Ex-Chat) | [13:21] | |
| ..... (idle for 21mn) | ||
| linkmauve | paulk, thanks! I'll review your changes later and possibly incorporate them into a v2. :) | [13:43] |
| *** | sgruszka has quit IRC (Ping timeout: 480 seconds) | [13:56] |
| ........ (idle for 37mn) | ||
| hverkuil has joined #linux-media
digetx has quit IRC (Read error: Connection reset by peer) digetx has joined #linux-media | [14:33] | |
| digetx has quit IRC (Read error: Connection reset by peer)
digetx has joined #linux-media | [14:42] | |
| ............ (idle for 58mn) | ||
| frieder has quit IRC (Remote host closed the connection) | [15:40] | |
| GBenji has left | [15:50] | |
| prabhakalad | pinchartl: to clarify CSI2 bridge sits between the ov5645 senosr and CRU. For the CRU to get VC from sensor it will need to call get_frame_desc to CSI2 subdev and then CSI2 subdev will have to implement .get_frame_desc which inturn calls get_frame_desc into ov5645 sensor driver. Is this the correct approach or am I missing something? | [15:59] |
| pinchartl | if the CRU needs the VC, yes, that's how it should work
does the CSI-2 RX output all VCs and DTs on a single output bus, with each sample tagged with the VC and DT IDs ? | [16:01] |
| prabhakalad | Yes the CSI2 can output all the VCs and DTs on a single output bus, with each sample tagged. | [16:03] |
| pinchartl | and do you need that information to program ICnMC.VCSEL ? | [16:04] |
| prabhakalad | Yes ICnMC.VCSEL to capture from a given VCx | [16:05] |
| pinchartl | yes then I think that's the way to go | [16:05] |
| prabhakalad | thanks for the clarification. | [16:06] |
| ........................ (idle for 1h59mn) | ||
| *** | cyrinux has quit IRC ()
cyrinux has joined #linux-media Moarc_ has quit IRC (Quit: i znowu NADMUCHAÅ BALONA) Moarc has joined #linux-media ao2 has quit IRC (Remote host closed the connection) | [18:05] |
| .............. (idle for 1h8mn) | ||
| gouchi has joined #linux-media | [19:23] | |
| ............... (idle for 1h14mn) | ||
| sailus has quit IRC (Ping timeout: 480 seconds) | [20:37] | |
| ..... (idle for 21mn) | ||
| sailus has joined #linux-media | [20:58] | |
| gouchi has quit IRC (Quit: Quitte) | [21:04] | |
| ....... (idle for 30mn) | ||
| linkmauve | paulk, oh, so it’s also possible to decode 4:4:4 now with your patches?
And downsampling happens when the user selected a 4:2:0 or 4:2:2 format as the capture format, interesting. Is that the common way of doing it? I’ll test that on my A64, and maybe on an A10 (but remotely, I am not home until some weeks later, hopefully I won’t crash anything). | [21:34] |
| Hmm… % git fetch paulk sunxi/cedrus/jpeg-nv16
fatal: pack has junk at the end error: Unable to find 5acc2f557495298391ee585237dfe2fd69b5024e under https://git.paulk.fr/projects/linux.git Cannot obtain needed object 5acc2f557495298391ee585237dfe2fd69b5024e error: fetch failed. Is this something you have to fix on your end? Does fetching work for you? | [21:46] | |
| *** | ten1572377432463050661821 has quit IRC (Remote host closed the connection)
ten1572377432463050661821 has joined #linux-media | [21:55] |
| ...... (idle for 27mn) | ||
| tmerciai2 has quit IRC (Ping timeout: 480 seconds) | [22:22] | |
| .... (idle for 19mn) | ||
| tmerciai2 has joined #linux-media | [22:41] | |
| paulk | linkmauve: let me check
linkmauve: I would assume 4:4:4 should work, probably both tiled and linear, but haven't tried it I'm not sure you could do 4:4:4 to 4:2:2 though, there only seems to be a single switch for downsampling so it probably always goes to 4:2:0 | [22:46] |
| linkmauve | When I tried, 4:4:4 could only get out as 4:2:2 without the flag, or 4:2:0 with it. | [22:50] |
| paulk | ah | [22:50] |
| *** | tmerciai2 has quit IRC (Remote host closed the connection) | [22:50] |
| linkmauve | Except on A10/A20 where the flag didn’t exist. | [22:50] |
| paulk | so on A10/A20 it always outputs the same subsampling as what is coded in the JPEG? | [22:51] |
| linkmauve | No, always 4:2:0 for 4:2:0, and always 4:2:2 for 4:2:2 or 4:4:4, IIRC. | [22:52] |
| *** | tmerciai2 has joined #linux-media | [22:52] |
| paulk | ok good to know | [22:52] |
| linkmauve | It will need to be handled differently from the newer SoCs. | [22:53] |
| paulk | So one thing I added with the v4l2 format info table is to check that the sub-sampling is compatible with the provided format
To make sure that we error out for cases that cannot be supported But given what you just said it probably boils down to refusing 4:2:2 formats for 4:2:0 coded data | [22:53] |
| linkmauve | And 4:2:0 for 4:2:2 coded data. | [22:55] |
| paulk | Well in that case we just transparently switch the downsampling bit
So the user gets what is in the format We don't need to have a 1:1 match between coded and user format | [22:55] |
| linkmauve | There is no downsampling bit on the A10 and A20.
(That I know of.) Or, and something I want to do next is to play with the ISP VE engine, it should allow format conversion from what I understand, but I found no sample code so far. But from what I could get from https://linux-sunxi.org/VE_Register_guide#MACC_ISP_CTRL it should allow NV16 to NV12 conversion (and possibly NV16_32L32 to linear NV12). Not sure which SoC this documentation is based on though, perhaps that didn’t exist on the A10 and A20 either. | [22:58] |
| paulk | Oh right yeah there should be a capability bit for jpeg downsampling
and that would have to affect the format checking indeed linkmauve: the sunxi wiki docs is pretty dated, typically a10/a20 era IIRC the h264 encoder uses the ISP as front-end and yeah it does have some format / scaling / rotation abilities but I don't know if it can be connected to the decode part my understanding sofar was that it was only for encode https://github.com/bootlin/linux/commit/4e9497947c56f10351ed8c000605dbe47d4aee46#diff-0dd9d7bda0966930521375b4729a460c33c6043222217844f46ba89228894c62R722 that's what I have figured out of the isp registers on v3s | [23:02] |
| https://github.com/bootlin/linux/blob/4e9497947c56f10351ed8c000605dbe47d4aee46/drivers/staging/media/sunxi/cedrus/cedrus_regs.h#L724 rather | [23:13] | |
| linkmauve | And that didn’t work at all standalone, by passing it an already-decoded buffer? | [23:15] |
| *** | darkapex has quit IRC (Remote host closed the connection)
darkapex has joined #linux-media | [23:20] |
| linkmauve | If it did, would it be more sensible to expose it as a different /dev/video* device, or would it be better to just extend the output capabilities of other decoders by running it after each decoded frame? | [23:24] |
| paulk | linkmauve: mhh don't think I really tried it out
yeah that's a good question if it involves configuring an input dma address I guess it's better to expose it as a separate thing and that can be useful for general-purpose scaling or format conversion | [23:24] |
| linkmauve | Yeah. | [23:25] |
| paulk | also if you want to do it all kernel side as decoder post-proc you'd need to allocate a bounce buffer
and you also couldn't use returned decoded frames as references | [23:25] |
| linkmauve | That might be why the ST12 format got kept on the A64 as a capture format, so that it outputs a tiled buffer which is then used as the input for the ISP engine. | [23:26] |
| paulk | so you'd need to have a single bounce buffer per capture buffer | [23:26] |
| linkmauve | Right. | [23:26] |
| paulk | so endless pain for little benefit | [23:26] |
| linkmauve | Userland should learn to use that kind of device eventually anyway. :) | [23:26] |
| paulk | it's quite possible that untiling goes through the ISP yes
IIRC you could feed it the tiled format as input for encoding https://github.com/aodzip/libcedarc/blob/master/include/vencoder.h#L234 so surely there's a switch to have it FIFO-connected to either the encoder or the decoder | [23:27] |
| linkmauve | Oh, there are AllWinner SoCs which support AFBC now?
I’m pretty certain none of mine do. Also since this JPEG decoder we now know that tile 32x32 can be either 4:2:0 or 4:2:2 at least. | [23:31] |
| paulk | I think that has been the case on the display side for a while, but I don't know about the VE
linkmauve: indeed there's also a 128x32 format that the decoder can output but there's about no interest in using it maybe it has slightly more efficient AXI transactions and decodes a bit faster | [23:32] |
| linkmauve | Aren’t we always limited by the memory bandwidth anyway? | [23:35] |
| paulk | I tend to believe this is often the case yeah. It's hard to have a clear idea of what's going on though. | [23:36] |
| linkmauve | Yeah… | [23:37] |
| paulk, ah you didn’t develop against mainline? Your second patch doesn’t apply, the format of the drivers/media/v4l2-core/v4l2-common.c table now has .bpp_div added.
I’ll fix it shortly and then test on A64. | [23:42] | |
| *** | ten1572377432463050661821 has quit IRC (Remote host closed the connection)
ten1572377432463050661821 has joined #linux-media | [23:46] |
| ↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |