When you set PIPE_CAP_FS_POINT_IS_SYSVAL, we'd fall out of
is_allowed_invariant(), despite it being allowed on special builtins.
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22950>
This adds a new limit, changing PIPE_SHADER_TYPES seems frought
with danger and very tricky. Currently only lava/llvmpipe plan to
offer task/mesh in GL land, so let's limit the fallout for now.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23291>
There is no good reason of having it too big and radeonsi returns an
absurd high value through PIPE_SHADER_CAP_MAX_CONST_BUFFER0_SIZE.
This speeds up the CTS testing this device limit.
Signed-off-by: Karol Herbst <git@karolherbst.de>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23378>
In commit 284f0c9a57 I refactored the
handling of the data source to just call a helper rather than special
casing opcodes with 0 or 2 sources. Unfortunately, I also dropped the
"else return 1", creating a fallthrough for all sources other than
SURFACE_LOGICAL_SRC_ADDRESS and SURFACE_LOGICAL_SRC_DATA.
The case below happened to return the correct value for all cases except
SURFACE_LOGICAL_SRC_SURFACE, which has been returning 2 instead of 1
since that commit.
Restore the else case. Thanks to Marcin Ślusarz for catching this.
Fixes: 284f0c9a57 ("intel/compiler: Add an lsc_op_num_data_values() helper")
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23347>
Handle the other pipeline stats counters in order to implement
GL_ARB_pipeline_statistics_query. Note that this does away with
collecting *all* the counters if DEBUG_COUNTERS is enabled, other-
wise it was getting over-complicated.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23301>
This was originally intended for earlier gens where certain draw-time
state needed a WFI (because it wasn't banked, etc). This isn't the case
on a6xx. So just directly emit WFIs where they are needed.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23301>
I found those issues while testing DOOM eternal and Ian also ran into
it with other shaders.
We write the desc register in SIMD1 exec_all, so all the data is in
the first component. We need to make sure to pass that component in
the lower SEND instructions.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Tested-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23354>
Because is_ngg wasn't set early enough, radv_get_wave_size() would always
return 64.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23336>
We draw the gallium hud directly to the rendered buffer, meaning that if
the buffer age is queried and then a partial redraw is done, we get a
ghosting effect from the hud drawn in previous frames.
Since we need to draw the hud with updated values every frame anyway,
there's no harm in disabling the buffer age and partial redraw.
Signed-off-by: Italo Nicola <italonicola@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Tested-by: Chris Healy <cphealy@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20387>
Actually set the packing behavior as the intended one when using this attribute,
even on mingw.
Otherwise, the default behavior is to use the ms_struct which makes the structures
use more space in some situations.
Signed-off-by: Corentin Noël <corentin.noel@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23338>
