On 7/11/18 7:40 am, Timothy Arceri wrote:
On 7/11/18 5:00 am, Eric Anholt wrote:
Timothy Arceri writes:
On Mon, Nov 5, 2018, at 4:44 PM, Timothy Arceri wrote:
On 6/11/18 11:30 am, Eric Anholt wrote:
Timothy Arceri writes:
For now I have only enabled this for RADV we can do it
also for
On 7/11/18 5:00 am, Eric Anholt wrote:
Timothy Arceri writes:
On Mon, Nov 5, 2018, at 4:44 PM, Timothy Arceri wrote:
On 6/11/18 11:30 am, Eric Anholt wrote:
Timothy Arceri writes:
For now I have only enabled this for RADV we can do it
also for radeonsi also but we need to add a CAP for
Timothy Arceri writes:
> On Mon, Nov 5, 2018, at 4:44 PM, Timothy Arceri wrote:
>> On 6/11/18 11:30 am, Eric Anholt wrote:
>> > Timothy Arceri writes:
>> >
>> >> For now I have only enabled this for RADV we can do it
>> >> also for radeonsi also but we need to add a CAP for it.
>> >
>> > If
On Mon, Nov 5, 2018, at 4:44 PM, Timothy Arceri wrote:
> On 6/11/18 11:30 am, Eric Anholt wrote:
> > Timothy Arceri writes:
> >
> >> For now I have only enabled this for RADV we can do it
> >> also for radeonsi also but we need to add a CAP for it.
> >
> > If we're doing this at link time, why
On 6/11/18 11:30 am, Eric Anholt wrote:
Timothy Arceri writes:
For now I have only enabled this for RADV we can do it
also for radeonsi also but we need to add a CAP for it.
If we're doing this at link time, why not push the constant value from
the producer into the consumer shader and then
Timothy Arceri writes:
> For now I have only enabled this for RADV we can do it
> also for radeonsi also but we need to add a CAP for it.
If we're doing this at link time, why not push the constant value from
the producer into the consumer shader and then cut out the varying
entirely?
For now I have only enabled this for RADV we can do it
also for radeonsi also but we need to add a CAP for it.
vkpipeline-db results:
Totals from affected shaders:
SGPRS: 4104 -> 3728 (-9.16 %)
VGPRS: 3604 -> 3472 (-3.66 %)
Spilled SGPRs: 0 -> 0 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Private
