projects / mesa.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: fbac9f8) | patch
gm107/ir: use lane 0 for manual textureGrad handling
authorIlia Mirkin <imirkin@alum.mit.edu>
Wed, 20 Dec 2017 04:37:25 +0000 (23:37 -0500)
committerIlia Mirkin <imirkin@alum.mit.edu>
Fri, 22 Dec 2017 05:17:15 +0000 (00:17 -0500)
commit05944a392ebe30f8a67bf70e1fbc4eb088fb67a0
tree5ca43e1e92d3e0c72c27f6f0dc2878ccae33ff94tree
parentfbac9f86aaffb102ee9dca6ade0fdf04fcd28d84commit | diff
gm107/ir: use lane 0 for manual textureGrad handling

This is parallel to the pre-SM50 change which does this. Adjusts the
shuffles / quadops to make the values correct relative to lane 0, and
then splat the results to all lanes for the final move into the target
register.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Tested-By: Karol Herbst <kherbst@redhat.com>
src/gallium/drivers/nouveau/codegen/nv50_ir_lowering_gm107.cpp diff | blob | history