On 2020-11-26 17:15 +0000, Jessica Clarke wrote:
> Source: mali-midgard
> Version: 16.0+pristine-4
> Severity: serious
>
> mali_base_hwconfig_issues.h currently has:
>
> > /* AUTOMATICALLY GENERATED FILE. If you want to amend the issues/features,
> > * please update base/tools/hwconfig_generator/hwc_{issues,features}.py
> > * For more information see base/tools/hwconfig_generator/README
> > */
>
> Unless patches to this file are accepted, this is not the preferred
> form, and the mentioned scripts do not exist (I assume they are
> Arm-internal scripts).Hmm. Well spotted. Patches to this file would definitely be accepted (insofar as anyone cares about this old version of the driver anyway (The last one with X support)) It looks like an embodiment of the (horrible) way that the mali driver supports different hardware variants (both changes in the GPU itself and how it is powered and clocked in a particular chip/board). The code is full of ifdefs for different hardware because it's assumed that it will be built for each target device. A properly generic driver is not supported, except those devices which work using devicetree info only. That file is obviously a list of the HW features relevent to each GPU variant. It is pretty cryptic though and it would be nice to know what they referred to. I'll investigate. The public linux driver is done as a release from an internal project which is aimed at delivering an eclipse DDK toolkit setup product largely targetting android. That script may well be part of generating the linux release to make sure things match. So far as us linux users are concerned it's just a header. I think it's fair to say that the linux driver is still a standalone piece of software which can be modified and updated on its own so I'm not convinced that this is actually grounds to throw it out. Wookey -- Principal hats: Linaro, Debian, Wookware, ARM http://wookware.org/
signature.asc
Description: PGP signature
