2024-03-22 03:11 PM
I saw one other post on this but it isn't clear anyone was actually successful putting chromium for Wayland on successfully with hardware acceleration. I'm able to get it to load but it's very slow and the chrome://GPU makes it seem unlikely.
Fwiw I'm using poky with the meta-st layer instead of stopenlinux. I'm using core-image-weston. It does not complain about the Eula not being accepted so curious if that's a hint?
Most things I read discuss chromium 101 instead of the most recent release but I think that's more imx8 related.
I am using the stm32mp157-dk2.
Solved! Go to Solution.
2024-03-25 01:56 AM
Hello @Thechef99 ,
2 points to mention regarding your message :
1) If the EULA for the GPU is not accepted, this is not a blocker for the compilation and the image generation. The fallback is that instead of using GPU binaries delivered by ST, the distribution will use Etnaviv / Mesa stacks to interact with the GPU.
2) As far as I know, in theory, recent versions of Chromium layer are able to interface themselves with Etnativ for GPU use. However, an internal study was made on our side, and the result for MP15 is not as good as expected. The rendering is very low, and Chromium is reaching GPU performances that MP15 can offer.
I hope it will help you to go forward.
Kind regards,
Erwan.
PS : If you want to check how does the GPU is used on your platform (GCNANO or Etnaviv stack), please perform a perf top command while having the GPU working, and check if you see things like *LibGAL* or *mesa* (LibGAL = gcnano/galcore, mesa=etnativ).
2024-03-25 01:56 AM
Hello @Thechef99 ,
2 points to mention regarding your message :
1) If the EULA for the GPU is not accepted, this is not a blocker for the compilation and the image generation. The fallback is that instead of using GPU binaries delivered by ST, the distribution will use Etnaviv / Mesa stacks to interact with the GPU.
2) As far as I know, in theory, recent versions of Chromium layer are able to interface themselves with Etnativ for GPU use. However, an internal study was made on our side, and the result for MP15 is not as good as expected. The rendering is very low, and Chromium is reaching GPU performances that MP15 can offer.
I hope it will help you to go forward.
Kind regards,
Erwan.
PS : If you want to check how does the GPU is used on your platform (GCNANO or Etnaviv stack), please perform a perf top command while having the GPU working, and check if you see things like *LibGAL* or *mesa* (LibGAL = gcnano/galcore, mesa=etnativ).
2024-03-25 05:51 AM
Thanks for such a detailed and relevant response!
1. Just to clarify you are saying that regardless of LibGAL or mesa, the performance you observed is not very good. I would assume LibGAL>mesa, but who knows.
2. This may be adding up. According to this presentation
https://www.st.com/resource/en/product_training/STM32MP1-System-Graphics_Processing_Unit_GPU.pdf
The supported OpenGL 2.0 is from 2007! So its possible the stm32mp1 designers just really never defined something like a kiosk as a good use case. The Cortex A7 is from 2011 as well so maybe this is just the nature of the development cycle of an MPU released in 2020.
3. Any thoughts on this scenario using the stm32mp2. Unfortunately I still don't see dev boards to evaluate?
2024-03-25 06:40 AM
@Thechef99 ,
This is not really in term of performances that this point blocks, but most in term of features. Chromium is using renderD128 to interact with the GPU, but Galcore does not expose it, while Etnaviv does. But anyway, Chromium is a little bit too heavy for MP15, and rendering are not good from what we could have seen.
Concerning the STM32MP2, this is not the same. Its GPU can indeed make Chromium work like a charm on it. But you are right, no evaluation board is available for now, only the press release has been made. More information to come at the Embedded World in 2 weeks... ;)
Thank you for your interest.
Kind regards,
Erwan.