Drivers can provide the inputs required for the backend to call the
compute function.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34872>
This affects piglit/bin/ext_texture_array-compressed teximage pbo -auto -fbo
Routine create_surface() at src/gallium/drivers/crocus/crocus_state.c
does not suppport compressed format and returns NULL in that case.
Since MR https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34054
routine st_try_pbo_compressed_texsubimage() at src/mesa/state_tracker/st_cb_texture.c is missing a test via pipe->create_surface() and does
not fallback to _mesa_store_compressed_texsubimage() which causes a SIGSEGV abort.
Regression is solved by introducing the variable surface_no_compress at struct pipe_caps, causing the fallback.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/13426
Signed-off-by: GKraats <vd.kraats@hccnet.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37150>
We can get into a situation where many BLT operations after a draw
cause a force flush of the current command stream. Calling
etna_state_update(..) without an active draw results in a segfault
due to inconsistent context state.
Track whether we're currently in draw_vbo to avoid updating derived
states during force flush outside of draw operations.
Passes dEQP-GLES3.functional.samplers.multi_tex_3d.* on GC7000.
Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36506>
This involves also emitting the CF for the blocks in the
sfn-assembler and we have to remove the special handling in the
old backend assembler.
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37096>
This is a step to emitting VTX and TEX block start instructions in the
sfn assembler instead of relying on the old backend asm code.
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37096>
To simplify the scheduling we need the method to be callable for all
instruction types.
v2: remove now useless override in ExportInstr
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37096>
The previous implementation was using temporal clamping variable within geometry shader.
It caused the framebuffer_layer_id input to be ignored, so gl_Layer would end up with a value of 0.
The fix removes the use of the temp variable.
Tested using CTS 4.6.6.0 test cases:
./glcts --deqp-case=dEQP-GL45-ES31.functional.geometry_shading.layered.fragment_layer_cubemap
./glcts --deqp-case=dEQP-GL45-ES31.functional.geometry_shading.layered.fragment_layer_3d
./glcts --deqp-case=dEQP-GL45-ES31.functional.geometry_shading.layered.fragment_layer_2d_array
./glcts --deqp-case=dEQP-GL45-ES31.functional.geometry_shading.layered.fragment_layer_2d_multisample_array
These tests fail before and pass after the change.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37069>
Fixes: 5f757bb95c ("nir: Make the load_store_vectorizer provide align_mul + align_offset.")
This is found when I am trying to narrow bit_size and num_components to uint8_t
Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37042>
Implement HSD 16028171704/14025112257:
LSC state cache livelock:- Once state cache entries are full,
subsequent walker dispatches with two threads per thread group maybe
gets stuck infinitely because of state cache live lock.
One thread continuously stuck in loop doing UGM fence + evict and UGM
read is waiting on UGM read to have certain value. while other thread
supposed to update the value that first thread is waiting for. But
since entries are full in state cache, there is second thread never
make progress.
Closes: #12352
Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37128>
The hardware support for tg4 is unclear from RE and feature databases.
Enable this extension on halti5 GPUs as a conservative starting point.
Support for 128 bit formats is missing, so it's gated behind ETNA_MESA_DEBUG=deqp.
Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36622>
Implement offset lowering by converting pixel offsets to normalized
coordinate space and adjusting coordinates accordingly.
Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36622>
With the threaded submit code, there's a small window where
the fence can be signalled but the buffer lists cleanup code
from amdgpu_cs_submit_ib has not executed yet.
This is harmless in general, because num_active_ioctls is
an optimization.
When the glFinish is called though, it's best to avoid this
inconsistencies because it can create odd behavior as shown
by the flakes being fixed by this commit.
The test runs a query, uses glFinish and expect the results
to be available. Since results availability depends on
num_active_ioctls, it made the test flaky.
Fix this by syncing when PIPE_FLUSH_HINT_FINISH is set.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36930>
