Manifests in chromiumos/manifest-versions.git repository

[Lingyun Cai]

Hi all,

I find that there are different trees in  https://chromium.googlesource.com/chromiumos/manifest-versions/+/master , like paladin, chrome-LKGM-candidates, etc.. If I want to sync the repo at ChromiumOS version 9176.0.0, I could find the corresponding manifest files in paladin/, chrome-LKGM-candidates/ and android-LKGM-candidates/ directories; besides, in these directories, there are more than one manifest files corresponding to 9176.0.0 with the suffix rc1/2/3/4. What's the difference among these trees and which trees/files are more appropriate for syncing the chromiumos repo and building the image?

Thanks,

Lingyun

[Mike Frysinger]

the official 9176.0.0 manifest isn't published as it's only an internal manifest, and we don't strip down and create a public Chromium OS one.  hence the closest you'll get is the paladin one with the highest rc or the next version with the lowest rc.  i.e. 9176.0.0-rc(largest) or 9177.0.0-rc(smallest).

paladin == the commit queue

chrome/android lkgm are related to the respective PFQs and aren't relevant for what you want.

-mike

--
--
Chromium OS Developers mailing list: chromiu...@chromium.org
View archives, change email options, or unsubscribe:
http://groups.google.com/a/chromium.org/group/chromium-os-dev?hl=en

[Cai, Lingyun]

Hey Mike, thanks for your help before and now we can build the images locally.

But we find that there is performance gap between official and local build images with the same manifest version. For example, the performance gap of page_cycler_v2.typical_25 between official (9592.0.0,dev-channel) and local (9592.0.0-rc5.xml) images  is ~12%. Is this reasonable and expected?

 

Thanks,

Lingyun

[Mike Frysinger]

my guess is you're seeing the effect of afdo.  we build chrome with USE=afdo_use.

-mike

[Cai, Lingyun]

Thanks Mike.

Based on your suggestion, we’ve tried to build chrome package with USE=afdo_use and deploy it to image. To evaluate the impact of afdo, we run page_cycler_v2 on two boards (elm and reef) using our local build images, and the results show that reef is improved by ~5% but no obvious performance change observed on elm. Is this expected that afdo only impacts some boards or boards w/ specific architectures (i.e., x86, arm…)?

 

Actually, even without afdo, the performance we observed on our local build image is better than that of the corresponding official build images (by running page_cycler_v2). It’s confused that official image w/ afdo performs worse than local image w/o afdo, we wonder if there are other features on local or official build images that would introduce the performance gap.

 

Besides, is afdo enabled for the official images in all channels (dev/beta/stable) ?

[Yunlian Jiang]

Yes, afdo is enabled for al the offical images. Did you pass the USE flag '-cros-debug' to build_packages?

That may cause some performance difference too.

---
You received this message because you are subscribed to the Google Groups "Chromium OS dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to chromium-os-dev+unsubscribe@chromium.org.

[Cai, Lingyun]

Thanks Yunlian.

We didn’t pass the USE flag ‘-cros-debug’ to ‘build_packages’. We followed the developer guide and below are the instructions we use to build local images, do you have any suggestions on this?

 

cros_sdk --enter

export BOARD=xxxx  

./setup_board --board=$BOARD

./build_packages --board=$BOARD  --accept_licenses=Google-TOS

./build_image --board=$BOARD --noenable_rootfs_verification --adjust_part='ROOT-A:+1G'  test

 

Besides, about the other two issues (afdo impact and performance gap w/ official images), do you have more insights?

 

Thanks,

Lingyun

To unsubscribe from this group and stop receiving emails from it, send an email to chromium-os-d...@chromium.org.

 

--
--
Chromium OS Developers mailing list: chromiu...@chromium.org
View archives, change email options, or unsubscribe:
http://groups.google.com/a/chromium.org/group/chromium-os-dev?hl=en

---
You received this message because you are subscribed to the Google Groups "Chromium OS dev" group.

To unsubscribe from this group and stop receiving emails from it, send an email to chromium-os-d...@chromium.org.