06:46 <memeka> mszyprow: ping

06:46 <mszyprow> memeka: pong

06:47 <memeka> mszyprow: gst-launch-1.0 filesrc location=grill-mjpeg.mov ! parsebin ! v4l2jpegdec ! v4l2convert capture-io-mode=dmabuf ! kmssink

06:47 <memeka> CPU usage was 17% for the fully accelerated pipeline, while using jpegdec ! videoconvert results in 30% CPU usage and v4l2jpegdec ! videoconvert results in 20% CPU usage.

06:48 <memeka> mszyprow: https://github.com/mihailescu2m/linux/commit/ad6fe4f26d2de7704f9f961d0b4c2a0ff435f167

06:48 <memeka> It’s hacky but seems to work

06:49 <mszyprow> memeka: sounds good :)

06:49 <mszyprow> memeka: feel free to submit it fo ml

06:50 <memeka> Will do after more testing

06:51 <memeka> It’s also only for 542x (exynos3250_jpeg_irq) for now, although I think it will work for s5p variants too

06:51 <memeka> Exynos4 needs one line to make it work for sure tho

07:19 <PabloPL> memeka: Will it work on s5pv210 ;)?

07:27 <memeka> PabloPL: what jpeg hardware does it have?

07:28 <PabloPL> meek: SJPEG_S5P

07:28 <PabloPL> memeka: https://github.com/torvalds/linux/blob/master/drivers/media/platform/s5p-jpeg/jpeg-core.c#L3172

07:29 <memeka> PabloPL: samsung,s5pv210-jpeg?

07:30 <memeka> It should work pls test

07:31 <memeka> PabloPL: patch is on top of 2 other jpeg patches from ml on top of 5.0 mainline

07:32 <PabloPL> memeka: Could You give me link do patchwork?

07:32 <memeka> PabloPL: nah it’s your patches

07:33 <PabloPL> memeka: Will test (but need to wait aroung 8h when i will have my device back)

07:33 <memeka> https://usercontent.irccloud-cdn.com/file/2mRrvNq8/1548315218.JPG

07:34 <memeka> PabloPL: ^^

07:35 <PabloPL> memeka: Thanks! Will do the testing later

07:35 <memeka> Let me know, thx

07:39 <PabloPL> memeka: by the way, those my two s5p-jpeg patches are already accepted in mainline (like others). Now need to get them all merged from linux-next...

07:40 <memeka> PabloPL: they’re not in 5.0-rc3 :)

07:41 <PabloPL> memeka: Like other 14 patches :(. not yet in Linus tree

12:36 <memeka> mszyprow: ping

12:36 <memeka> [ 4.326639] cpu cpu0: Linked as a consumer to regulator.44

12:36 <memeka> [ 4.331204] cpu cpu0: Dropping the link to regulator.44

12:36 <memeka> [ 4.337033] cpu cpu0: Linked as a consumer to regulator.44

12:36 <memeka> ^^^ why this happens?

12:36 <mszyprow> memeka: useless debug messages

12:38 <mszyprow> memeka: someone should disable those messages, as they don't really provide any usefull information

12:38 <memeka> guess so :)

12:38 <mszyprow> memeka: it is about the way regulator core implements its dependencies between devices

12:39 <memeka> mszyprow: also: https://github.com/mihailescu2m/linux/commit/cec6a9f80c0fcd2e463f47a281e9b4a16c7e9479

12:40 <memeka> w/o this patch, sometimes you get washed out colors playing video :)

12:40 <memeka> confirmed with ffplay on SDL2 backed

12:40 <mszyprow> memeka: okay, but the patch is not really correct

12:41 <memeka> what's the correct way?

12:41 <mszyprow> memeka: you need to make it depend on the configured zpos

12:41 <mszyprow> memeka: by changing zpos you can change the order of planes

12:43 <memeka> if (plane->index) then?

12:43 <memeka> and i need to get the plane from somewhere...

12:44 <mszyprow> memeka: see priority = state->base.normalized_zpos + 1;

12:44 <mszyprow> memeka: window with priority == 0 should have alpha disabled then

12:45 <mszyprow> priority == 1 of course, 0 means disabled afair

12:51 <memeka> state = &ctx->crtc->base.state ?

12:51 <memeka> no

12:52 <memeka> state = &ctx->planes[win]->base.state ?

12:53 <memeka> state = to_exynos_plane_state(&ctx->planes[win]->base.state) ?

12:53 <memeka> then priority = state->base.normalized_zpos + 1;

12:54 <memeka> if (priority) instead of if (win) ?

14:11 <mszyprow> memeka: you already have everything needed in mixer_graph_buffer(), you may need only to change the parameters for mixer_cfg_layer()/mixer_cfg_gfx_blend() call depending on the 'priority'