Chromium Code Reviews
|
|
Chromium Code Reviews
Issue
880643004:
Use GSC0 instead of GSC1 for video encoding. (Closed)
|
Created:
5 years, 11 months ago by ynovikov Modified:
5 years, 11 months ago CC:
chromium-reviews, posciak+watch_chromium.org, jam, mcasas+watch_chromium.org, feature-media-reviews_chromium.org, darin-cc_chromium.org, piman+watch_chromium.org, wjia+watch_chromium.org, jln+watch_chromium.org, marcheu Base URL:
https://chromium.googlesource.com/chromium/src.git@master Target Ref:
refs/pending/heads/master Project:
chromium Visibility:
Public. |
DescriptionUse GSC0 instead of GSC1 for video encoding.
BUG=368775
TEST=Casting WebRTC works on daisy and peach_pit
Committed: https://crrev.com/0c700e62b4574a1d63d61b500c34a710baa0f5ad
Cr-Commit-Position: refs/heads/master@{#313367}
Patch Set 1 #
Messages
Total messages: 19 (3 generated)
ynovikov@chromium.org changed reviewers: + jln@chromium.org, posciak@chromium.org
Need to free up GSC1 for YUV overlays usage.
marcheu@chromium.org changed reviewers: + marcheu@chromium.org
yup it seems wise to do this while we can (i.e. while we haven't introduced gsc1/2/3 consumers yet). LGTM.
lgtm
lgtm
When you tested, did you verify that it was using HW encode?
Is there a simple way to check that? I checked that /dev/video12 is open by gpu process. On 26 Jan 2015 18:47, <posciak@chromium.org> wrote: > When you tested, did you verify that it was using HW encode? > > https://codereview.chromium.org/880643004/ > To unsubscribe from this group and stop receiving emails from it, send an email to chromium-reviews+unsubscribe@chromium.org.
On 2015/01/27 00:24:46, ynovikov wrote: > Is there a simple way to check that? I checked that /dev/video12 is open by > gpu process. That should be sufficient then thanks. Just checking if webrtc worked per TEST= line in description would not have been sufficient. The easy way to check is to check about:histograms/Media.RTCVideoEncoderInitEncodeSuccess and verify it was 1.
On 2015/01/27 00:29:12, Pawel Osciak wrote: > On 2015/01/27 00:24:46, ynovikov wrote: > > Is there a simple way to check that? I checked that /dev/video12 is open by > > gpu process. > > That should be sufficient then thanks. Just checking if webrtc worked per TEST= > line in description would not have been sufficient. > > The easy way to check is to check > about:histograms/Media.RTCVideoEncoderInitEncodeSuccess and verify it was 1. I don't have Media.RTCVideoEncoderInitEncodeSuccess, but I do have Cast.Sender.VideoEncodeAcceleratorInitializeSuccess==1. I think it's sufficient, as I've verified that V4L2ImageProcessor::Initialize() is called from that path.
The CQ bit was checked by ynovikov@chromium.org
CQ is trying da patch. Follow status at https://chromium-cq-status.appspot.com/patch-status/880643004/1
Message was sent while issue was closed.
Committed patchset #1 (id:1)
Message was sent while issue was closed.
Patchset 1 (id:??) landed as https://crrev.com/0c700e62b4574a1d63d61b500c34a710baa0f5ad Cr-Commit-Position: refs/heads/master@{#313367}
Message was sent while issue was closed.
On 2015/01/27 21:32:28, ynovikov wrote: > On 2015/01/27 00:29:12, Pawel Osciak wrote: > > On 2015/01/27 00:24:46, ynovikov wrote: > > > Is there a simple way to check that? I checked that /dev/video12 is open by > > > gpu process. > > > > That should be sufficient then thanks. Just checking if webrtc worked per > TEST= > > line in description would not have been sufficient. > > > > The easy way to check is to check > > about:histograms/Media.RTCVideoEncoderInitEncodeSuccess and verify it was 1. > > I don't have Media.RTCVideoEncoderInitEncodeSuccess, but I do have > Cast.Sender.VideoEncodeAcceleratorInitializeSuccess==1. > I think it's sufficient, as I've verified that V4L2ImageProcessor::Initialize() > is called from that path. Oh so you were not testing webrtc. Cast does not use webrtc. But yes, it's sufficient. Thanks.
Message was sent while issue was closed.
On 2015/01/28 00:22:57, Pawel Osciak wrote: > On 2015/01/27 21:32:28, ynovikov wrote: > > On 2015/01/27 00:29:12, Pawel Osciak wrote: > > > On 2015/01/27 00:24:46, ynovikov wrote: > > > > Is there a simple way to check that? I checked that /dev/video12 is open > by > > > > gpu process. > > > > > > That should be sufficient then thanks. Just checking if webrtc worked per > > TEST= > > > line in description would not have been sufficient. > > > > > > The easy way to check is to check > > > about:histograms/Media.RTCVideoEncoderInitEncodeSuccess and verify it was 1. > > > > I don't have Media.RTCVideoEncoderInitEncodeSuccess, but I do have > > Cast.Sender.VideoEncodeAcceleratorInitializeSuccess==1. > > I think it's sufficient, as I've verified that > V4L2ImageProcessor::Initialize() > > is called from that path. > > Oh so you were not testing webrtc. Cast does not use webrtc. But yes, it's > sufficient. Thanks. I was casting https://apprtc.appspot.com/?debug=loopback. That's the only thing I found that actually used GSC hardware.
Message was sent while issue was closed.
On 2015/01/28 00:44:00, ynovikov wrote: > On 2015/01/28 00:22:57, Pawel Osciak wrote: > > On 2015/01/27 21:32:28, ynovikov wrote: > > > On 2015/01/27 00:29:12, Pawel Osciak wrote: > > > > On 2015/01/27 00:24:46, ynovikov wrote: > > > > > Is there a simple way to check that? I checked that /dev/video12 is open > > by > > > > > gpu process. > > > > > > > > That should be sufficient then thanks. Just checking if webrtc worked per > > > TEST= > > > > line in description would not have been sufficient. > > > > > > > > The easy way to check is to check > > > > about:histograms/Media.RTCVideoEncoderInitEncodeSuccess and verify it was > 1. > > > > > > I don't have Media.RTCVideoEncoderInitEncodeSuccess, but I do have > > > Cast.Sender.VideoEncodeAcceleratorInitializeSuccess==1. > > > I think it's sufficient, as I've verified that > > V4L2ImageProcessor::Initialize() > > > is called from that path. > > > > Oh so you were not testing webrtc. Cast does not use webrtc. But yes, it's > > sufficient. Thanks. > > I was casting https://apprtc.appspot.com/?debug=loopback. > That's the only thing I found that actually used GSC hardware. Which device? If it was on Pit, you should have actually have had two instances of HW encoder running and you should've seen both the Cast histogram you saw and Webrtc histogram I mentioned above. If you didn't see both, something is wrong.
Message was sent while issue was closed.
On 2015/01/28 00:46:27, Pawel Osciak wrote: > On 2015/01/28 00:44:00, ynovikov wrote: > > On 2015/01/28 00:22:57, Pawel Osciak wrote: > > > On 2015/01/27 21:32:28, ynovikov wrote: > > > > On 2015/01/27 00:29:12, Pawel Osciak wrote: > > > > > On 2015/01/27 00:24:46, ynovikov wrote: > > > > > > Is there a simple way to check that? I checked that /dev/video12 is > open > > > by > > > > > > gpu process. > > > > > > > > > > That should be sufficient then thanks. Just checking if webrtc worked > per > > > > TEST= > > > > > line in description would not have been sufficient. > > > > > > > > > > The easy way to check is to check > > > > > about:histograms/Media.RTCVideoEncoderInitEncodeSuccess and verify it > was > > 1. > > > > > > > > I don't have Media.RTCVideoEncoderInitEncodeSuccess, but I do have > > > > Cast.Sender.VideoEncodeAcceleratorInitializeSuccess==1. > > > > I think it's sufficient, as I've verified that > > > V4L2ImageProcessor::Initialize() > > > > is called from that path. > > > > > > Oh so you were not testing webrtc. Cast does not use webrtc. But yes, it's > > > sufficient. Thanks. > > > > I was casting https://apprtc.appspot.com/?debug=loopback. > > That's the only thing I found that actually used GSC hardware. > > Which device? If it was on Pit, you should have actually have had two instances > of HW encoder running and you should've seen both the Cast histogram you saw and > Webrtc histogram I mentioned above. If you didn't see both, something is wrong. Ah, sorry, I've initially tested on both daisy and peach_pit, but checked the histogram only on daisy. Now I've checked on peach_pit, too, and, indeed, I see both the histograms. Thanks to your explanation, now I understand why I saw two /dev/video12 fds open on peach_pit, and only one on daisy.
Message was sent while issue was closed.
On 2015/01/28 01:43:25, ynovikov wrote: > On 2015/01/28 00:46:27, Pawel Osciak wrote: > > On 2015/01/28 00:44:00, ynovikov wrote: > > > On 2015/01/28 00:22:57, Pawel Osciak wrote: > > > > On 2015/01/27 21:32:28, ynovikov wrote: > > > > > On 2015/01/27 00:29:12, Pawel Osciak wrote: > > > > > > On 2015/01/27 00:24:46, ynovikov wrote: > > > > > > > Is there a simple way to check that? I checked that /dev/video12 is > > open > > > > by > > > > > > > gpu process. > > > > > > > > > > > > That should be sufficient then thanks. Just checking if webrtc worked > > per > > > > > TEST= > > > > > > line in description would not have been sufficient. > > > > > > > > > > > > The easy way to check is to check > > > > > > about:histograms/Media.RTCVideoEncoderInitEncodeSuccess and verify it > > was > > > 1. > > > > > > > > > > I don't have Media.RTCVideoEncoderInitEncodeSuccess, but I do have > > > > > Cast.Sender.VideoEncodeAcceleratorInitializeSuccess==1. > > > > > I think it's sufficient, as I've verified that > > > > V4L2ImageProcessor::Initialize() > > > > > is called from that path. > > > > > > > > Oh so you were not testing webrtc. Cast does not use webrtc. But yes, it's > > > > sufficient. Thanks. > > > > > > I was casting https://apprtc.appspot.com/?debug=loopback. > > > That's the only thing I found that actually used GSC hardware. > > > > Which device? If it was on Pit, you should have actually have had two > instances > > of HW encoder running and you should've seen both the Cast histogram you saw > and > > Webrtc histogram I mentioned above. If you didn't see both, something is > wrong. > > Ah, sorry, I've initially tested on both daisy and peach_pit, but checked the > histogram only on daisy. > Now I've checked on peach_pit, too, and, indeed, I see both the histograms. > Thanks to your explanation, now I understand why I saw two /dev/video12 fds open > on peach_pit, and only one on daisy. Sure, thank you so much for retesting, very much appreciated! |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
