Bug P3
Status Update
No update yet.
Comments
bu...@chops-service-accounts.iam.gserviceaccount.com
For consideration as a feature
bu...@chops-service-accounts.iam.gserviceaccount.com
@abhishekbh do you know what would be involved in supporting OpenCL? Especially for anything ML-related, being able to use the GPU for computation will be valuable.
cy...@google.com
I'm not sure at all what the OpenCL API is like and what it would to virtualize it. Definitely not on our radar and would require some staffing.
vr...@google.com
Could this be tied into the #crostini-gpu-support feature flag?
bu...@chops-service-accounts.iam.gserviceaccount.com
This is doable eventually when 1. crostini gets vulkan and 2. we put clvk in crostini
bu...@chops-service-accounts.iam.gserviceaccount.com
While waiting for Vulkan to get into crostini (and then clvk, which is an OpenCL implementation on top of Vulkan), know that it is possible to have OpenCL over CPU (not GPU) by using project like
It is an open-source OpenCL compliant implementation that will run OpenCL applications on the CPU.
From what I remember, it does not have strong dependencies on unusual packages.
vr...@google.com
This would be interesting to try. I wonder if this would be sufficient to enable Davinci Resolve and if it would be performant enough to run. Even on high-end Chromebooks this would be a big win.
vi...@google.com
> 1. It is interesting to note that some devices fail all of their runs - not even a single pass. This maybe fixed by adjusting the timeouts used, thoughts on this ???
We should understand why those devices behave differently. I thought the timeout is generous enough (I didn't check again)?
We should understand why those devices behave differently. I thought the timeout is generous enough (I didn't check again)?
vi...@google.com
Any update?
vr...@google.com
I debugged further for the 1st reason - this test always failing on some devices.
Wondering if it is a dependency issue? Am I correct in thinking that devices should at least pass the security.Bootlockbox(http://cs/chromeos_public/src/platform/tast-tests/src/chromiumos/tast/remote/bundles/cros/security/boot_lockbox.go ) test, and only then pass this arc.AdbSideloadwarningUI test. But bootlock box test also has tpm2 in its dependencies.
Is it correct that I should also add tpm2 dependency in this test as well?
Wondering if it is a dependency issue? Am I correct in thinking that devices should at least pass the security.Bootlockbox(
Is it correct that I should also add tpm2 dependency in this test as well?
vi...@google.com
boot lockbox requires tpm2. If the dependency is not transitive (BootLockboxService already requires tpm2), then adding tpm2 manually may help.
bu...@chops-service-accounts.iam.gserviceaccount.com
The following revision refers to this bug:
https://chromium.googlesource.com/chromiumos/platform/tast-tests/+/df4e32a5621198e090e3c8824396ccc58847c6b3
commit df4e32a5621198e090e3c8824396ccc58847c6b3
Author: Vaibhav Raheja <vraheja@google.com>
Date: Thu Aug 27 21:33:02 2020
tast-tests: add tpm dependency for arc.AdbSideloadWarningUI test
ADB Sideloading Warning UI test requires BootLockboxService which
further requires tpm2 dependency. Adding it in this test's dependencies
to prevent running the test on devices that don't have that capability.
BUG=chromium:1041137
TEST=tast run $DUT arc.ADBSideloadWarningUI
Change-Id: Ib336340bf2c5947c2446f9e9b3ff9aca60cff9bd
Reviewed-on:https://chromium-review.googlesource.com/c/chromiumos/platform/tast-tests/+/2378448
Reviewed-by: Victor Hsieh <victorhsieh@chromium.org>
Reviewed-by: Hidehiko Abe <hidehiko@chromium.org>
Tested-by: Vaibhav Raheja <vraheja@chromium.org>
Auto-Submit: Vaibhav Raheja <vraheja@chromium.org>
Commit-Queue: Vaibhav Raheja <vraheja@chromium.org>
[modify]https://crrev.com/df4e32a5621198e090e3c8824396ccc58847c6b3/src/chromiumos/tast/remote/bundles/cros/arc/adb_sideload_warning_ui.go
commit df4e32a5621198e090e3c8824396ccc58847c6b3
Author: Vaibhav Raheja <vraheja@google.com>
Date: Thu Aug 27 21:33:02 2020
tast-tests: add tpm dependency for arc.AdbSideloadWarningUI test
ADB Sideloading Warning UI test requires BootLockboxService which
further requires tpm2 dependency. Adding it in this test's dependencies
to prevent running the test on devices that don't have that capability.
BUG=chromium:1041137
TEST=tast run $DUT arc.ADBSideloadWarningUI
Change-Id: Ib336340bf2c5947c2446f9e9b3ff9aca60cff9bd
Reviewed-on:
Reviewed-by: Victor Hsieh <victorhsieh@chromium.org>
Reviewed-by: Hidehiko Abe <hidehiko@chromium.org>
Tested-by: Vaibhav Raheja <vraheja@chromium.org>
Auto-Submit: Vaibhav Raheja <vraheja@chromium.org>
Commit-Queue: Vaibhav Raheja <vraheja@chromium.org>
[modify]
vi...@google.com
I guess the SSH connection flakiness also happens to other tests that require reboot? If the problem has already been happening in CQ, this shouldn't make it worse.
vr...@google.com
Yes, problem occurs in other tests as well and not specific to only security.BootLockbox.
vi...@google.com
Let's promote the test from informational then? The test seems quite stable now (and all previous problems are not from the test itself).
Description
Now that grpc support is implemented in Tast, let's add a Tast test to adb sideloading.
We may as well create a test for boot lockbox, as we'll need the service anyway.