If a dimension in workgroup_size is 1 we can simply emit a 0 instead of
loading a system value. Helps a couple of compute shaders.
Totals:
CodeSize: 934788544 -> 934788048 (-0.00%)
Static cycle count: 222856984 -> 222856908 (-0.00%)
Totals from 8 (0.01% of 87622) affected shaders:
CodeSize: 34992 -> 34496 (-1.42%)
Static cycle count: 8593 -> 8517 (-0.88%)
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36428>
The two backends (ir3, dxil) using the lowering have
info->fs.uses_sample_shading matching set when sample shading is used -- the VK
drivers pass the rasterization state flag into the compiler, while
freedreno and d3d12 have caps->force_persample_interp so the frontend
creates a shader variant with info->fs.uses_sample_shading set.
This means that we can drop the sample-id SHL/AND in the pixel-rate
shading case, which in turn means that drivers don't need to have a
load_sample_id() that doesn't trigger sample-rate shading (which Adreno
doesn't have).
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36429>
... which is still needed to keep the end2end tests happy
when one thread creates and destroys instances multiple times
(which happens in the MultiThreadedShutdown test).
Test: GfxstreamEnd2EndTests
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36471>
We were using the maximum of all attachment sizes as the bounding box
for the frame buffer. But in fact we want the minimum, as we do not
want to draw outside of any attachment boundaries.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36360>
The untyped pointer types only have a storage class associated, and the
operations using them would carry the necessary "data type" information.
Untyped pointers themselves are identified by "vtn_type::pointed" being
NULL. For the NIR lowering the operations will have explicit casts
before them when applicable and the nir_derefs representing untyped
pointers will use the "void" glsl_type.
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36427>
There are a few things that could cause us to return NULL here and we
should just fail to parse the SPIR-V if that ever happens instead of
crashing when someone tries to access it.
Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36427>
This adds a vtn_pointer_ssa_is_desc_index() helper and makes both
to/from_ssa work in terms of it, ensuring we never screw up the
condition and turn a deref into a block index or vice versa.
Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36427>
These haven't existed in a while but the vtn_pointer fields stayed
around. Drop offset since no one uses it and rename block_index to
desc_idx with a better comment.
Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36427>
Introduce a new virgl_framebuffer_state to be able to store surface handles
directly.
Signed-off-by: Corentin Noël <corentin.noel@collabora.com>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35681>
This is the last piece to expose KHR_robustness. The main bulk of the
job is done already in the broadcom compiler through lowerings, so we
only need to set the proper key in order to enable the robustness.
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Reviewed-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Maíra Canal <mcanal@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35965>
This uniform will be used when creating context with robust buffer
access.
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Reviewed-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35965>
This implements supports for GPU reset notifications, which is part of
the KHR_robustness extension.
It requires support from kernel in order to expose it.
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Reviewed-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Maíra Canal <mcanal@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35965>
As this is a simulator, it will always report no reset.
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Reviewed-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Maíra Canal <mcanal@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35965>
This checks if kernel supports GPU reset counters, which are actually
two counters: one for resets caused by current context, and another one
for resets caused in general.
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Reviewed-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Maíra Canal <mcanal@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35965>
It is obvious that we check the return value to see if kernel supports
the feature, so no need to add a comment.
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Maíra Canal <mcanal@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35965>
AddressSanitizer: global-buffer-overflow ../src/util/u_debug.c:331 in debug_parse_flags_option
Fixes: 5c92fe45a1 ("radeonsi: support more than 64 options for AMD_DEBUG")
Reviewed-by: Qiang Yu <yuq825@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36434>
It's mostly used for SEND messages and fully defines the register data
(that's its purpose after all).
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36457>
The delivered values in the coarse pixel size are 0 when coarse pixel
dispatch is disabled and that is screwing up our half pixel offset
adjustment.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: mesa-stable
Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36457>
