[Mesa-dev] [PATCH 2/2] radv: Add new flag for LLVM dependecies with vulkan

Tobias Droste tdroste at gmx.de
Wed Oct 5 21:40:46 UTC 2016


Hi Emil,

the approach Dave used is ok and makes this obsolete. I talked about this to 
him in IRC. 
I actually _don't_ want a gallium like approach and didn't like this patch too 
much. This was just he least amount of changes without touching anything else 
and still make it work.

Having said that, I think the whole LLVM stuff should be handled differently. 
I wait for radv beeing merged before proposing any chances to not make it more 
difficult than it should be.

Tobias

Am Mittwoch, 5. Oktober 2016, 13:16:45 CEST schrieb Emil Velikov:
> On 2 October 2016 at 20:45, Tobias Droste <tdroste at gmx.de> wrote:
> > This reuse the same logic gallium uses to determine if LLVM is needed or
> > not:
> > --enable-vulkan-llvm is set to yes if at least one vulkan driver is
> > active and the host is i3*6 or x86_64.
> > To build vulkan drivers without LLVM (e.g. intel) one has to add
> > --disable-vulkan-llvm.
> > 
> > In order to make this all work the vulkan driver check has to move
> > either diretcly below or directly above the gallium driver checks.
> > Move them below the gallium driver stuff.
> 
> Having a blond moment here - is there an actual issue with the way
> (Dave/Bas) did the Vulkan/LLVM detection ?
> 
> If the goal is to have a more gallium-like approach that isn't
> applicable since the gallium aux modules (using LLVM) are linked in
> every driver - regardless if it uses them or not. Admittedly one could
> untangle this and make the LLVM requirement implicit as requesting a
> driver which uses LLVM.
> 
> Even then, some modules can optionally make use of it (LLVM) for which
> we'll still need the toggle :-\
> 
> Afaict RADV does not use the gallium aux module so the above should
> not be needed, no ?
> 
> -Emil


More information about the mesa-dev mailing list