Re: [Mesa-dev] [PATCH] nv50/ir: remove DUMMY edge type

2019-10-14 Thread Ilia Mirkin
Oh. Probably. Let me have another look then. On Mon, Oct 14, 2019 at 5:27 PM Karol Herbst wrote: > > isn't that what "UNKNOWN" is for? > > On Mon, Oct 14, 2019 at 11:16 PM Ilia Mirkin wrote: > > > > The idea was that this type would be used when you're not sure, and > > then run the classifier a

Re: [Mesa-dev] [PATCH] nv50/ir: remove DUMMY edge type

2019-10-14 Thread Karol Herbst
isn't that what "UNKNOWN" is for? On Mon, Oct 14, 2019 at 11:16 PM Ilia Mirkin wrote: > > The idea was that this type would be used when you're not sure, and > then run the classifier afterwards. Otherwise the classifier doesn't > know which edges to classify... > > On Mon, Oct 14, 2019 at 5:10 P

Re: [Mesa-dev] [PATCH] nv50/ir: remove DUMMY edge type

2019-10-14 Thread Ilia Mirkin
The idea was that this type would be used when you're not sure, and then run the classifier afterwards. Otherwise the classifier doesn't know which edges to classify... On Mon, Oct 14, 2019 at 5:10 PM Karol Herbst wrote: > > it was never used > > Signed-off-by: Karol Herbst > --- > src/gallium/

[Mesa-dev] Anyone interested in CoC training courtesy of X.Org?

2019-10-14 Thread Lyude Paul
Hi! This year we decided to start putting members of the CoC teams for Xorg and XDC through code of conduct training, paid for by the X.Org foundation and run by Otter Tech: https://otter.technology/code-of-conduct-training/ The training goes over practicing taking incident reports, following up

[Mesa-dev] [PATCH] nv50/ir: remove DUMMY edge type

2019-10-14 Thread Karol Herbst
it was never used Signed-off-by: Karol Herbst --- src/gallium/drivers/nouveau/codegen/nv50_ir_bb.cpp| 3 --- src/gallium/drivers/nouveau/codegen/nv50_ir_graph.cpp | 8 +--- src/gallium/drivers/nouveau/codegen/nv50_ir_graph.h | 1 - src/gallium/drivers/nouveau/codegen/nv50_ir_ra.cpp

Re: [Mesa-dev] [Nouveau] [PATCH] nv50/ir: mark STORE destination inputs as used

2019-10-14 Thread Karol Herbst
Reviewed-by: Karol Herbst On Mon, Oct 14, 2019 at 8:47 AM Ilia Mirkin wrote: > > Observed an issue when looking at the code generatedy by the > image-vertex-attrib-input-output piglit test. Even though the test > itself worked fine (due to TIC 0 being used for the image), this needs > to be fixe

Re: [Mesa-dev] [Nouveau] [PATCH] gm107/ir: fix loading z offset for layered 3d image bindings

2019-10-14 Thread Karol Herbst
I don't think this is a good idea overall. The way simpler solution would be to disable tiling on the z axis for 3d images so that we don't hurt the most common case, 2d images. And that's what I was seeing nvidia doing anyway. So with that we would end up adding a bunch of instructions hurting t