Freedreno's MSM DRM Driver In Linux 4.19 Getting "DPU1" Support For SDM845+ Support

[phoronix]

Phoronix: Freedreno's MSM DRM Driver In Linux 4.19 Getting "DPU1" Support For SDM845+ Support

Freedreno lead developer Rob Clark at Red Hat has sent in his batch of feature updates to DRM-Next ahead of the imminent Linux 4.19 kernel development cycle kicking off...

http://www.phoronix.com/scan.php?pag...x-4.19-Updates

[You-]

You have to hand it to Rob Clarke. When he started working on Freedreno, Qualcomm's level of support for Linux graphics was similar to that of nVidia's, if not more hostile (as most mobile chipset manufacturers were).

He made the effort successful enough for there to be unofficial backing Qualcomm and now it seems Qualcomm will be upstreaming their own work as part of this effort.

Hopefully this will mean full timely upstream support for future Qualcomm platforms.

We have also gone from zero support to projects tackling graphics on Arm Mali (Lima, Panfrost), Broadcom's Video Core (Whatever VC5 was renamed to) and Vivante (Etnaviv). Not a bad improvement.

[boxie]

You have to hand it to Rob Clarke. When he started working on Freedreno, Qualcomm's level of support for Linux graphics was similar to that of nVidia's, if not more hostile (as most mobile chipset manufacturers were).

He made the effort successful enough for there to be unofficial backing Qualcomm and now it seems Qualcomm will be upstreaming their own work as part of this effort.

Hopefully this will mean full timely upstream support for future Qualcomm platforms.

We have also gone from zero support to projects tackling graphics on Arm Mali (Lima, Panfrost), Broadcom's Video Core (Whatever VC5 was renamed to) and Vivante (Etnaviv). Not a bad improvement.

I am guessing that Googles Project Treble also helped in their decision for a more open source approach and with Rob being a part it just made sense to go down this path.

I wonder how much of that 100k LoC was reduced because of shared kernel code between all the display drivers and how much more (if possible) could be shared across drivers?

[bemerk]

Is there any chip series other than PowerVR left that hasn't an open source linux driver project?

[You-]

I am guessing that Googles Project Treble also helped in their decision for a more open source approach

AFAIK Project Treble is to help drivers that are either closed source or not maintained in the kernel. It probably has little benefit to code that is upstreamed.

[boxie]

AFAIK Project Treble is to help drivers that are either closed source or not maintained in the kernel. It probably has little benefit to code that is upstreamed.

Project Treble is many things. A separation layer within android (think HAL), Working with chipset vendors to supply a working system to the mobile device manufacturers (including the encouraging of not maintaining so much out of tree code) and making sure that the phone vendors make it all work. It really is a massive undertaking.

A really good interview with lots of info on it is here https://arstechnica.com/gadgets/2018...fragmentation/

[You-]

I understand that but the main beneficiaries are for drivers and modules that are not in the mainline kernel (and is currently limited to Android fork of the kernel). For drivers in the mainline kernel, no separation layer is needed.

I am supporting of Project Treble. It should help the Android landscape transform into something much more open and upgradeable.

By moving upstream, Qualcomm are providing a solution here that can be around longer than the 2 years they currently are willing to support (but without "support"). that should work for all users and not just android.