-
Notifications
You must be signed in to change notification settings - Fork 43
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Encoder: RADV not supported #71
Comments
Hi @dabrain34 - #57 originally fixed this, and also more recently #69 - hopefully #69 can be merged soon to fix this issue for encode (although it exists in the decoder too) |
Yes, we need a separate transfer queue for both AMD and Intel implementations. Can you please try the ToT. It should be fixed now. |
Since the last update (20240329) I get this kind of error with AMD(RADV) and INTEL(ANV)
|
Thank you, Stephane! Can you please tell us if this is happening during the pre-processing stage or the encode stage? An applicaiton call- stack would be very helpful. |
The assert is located in the radv driver, here is the backtrace:
|
Here is the version for ANV:
|
Thank you, for the backtrace, Stephane!
I'll add this as an explicit option, but I would highly recommend that drivers support the creation of views against individual planes. I'm not even sure if this is not required by the spec. However, the issue here could also be that the encoder is requesting the VK_IMAGE_USAGE_SAMPLED_BIT bit. result = m_linearInputImagePool->Configure(... with VK_IMAGE_USAGE_TRANSFER_SRC_BIT? Essentially, removing the VK_IMAGE_USAGE_SAMPLED_BIT and VK_IMAGE_USAGE_STORAGE_BIT. Sorry, but I do not have a setup with the open source driver and will not be able to try it. |
@dabrain34 @zlatinski, I took at look at the location of that assert. It will trigger if the VkImage was not created with I was able to work around this by setting After that, RADV (and ANV as well probably) segfaults in vkQueueSubmit. pWaitDstStageMask has to be a pointer to an array of |
Thank you, lolzballs!
Yes, indeed, this is the correct use when the individual planes of YCbCr MP formats are used. There is no need to check for support for that, because this feature is required for MP formats.
Yes, I've missed that for the encoder - sorry! Fixed both issues at the ToT of main. |
I confirm that RADV is now properly supported with I created a PR #95 to set the default value for layeredDPB to true |
When trying to encode using RADV encoder,
vk-video-enc-test
fails with:The text was updated successfully, but these errors were encountered: