fran/mesa
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
188,335 Commits 63 Branches 889 Tags
24.1-branchpoint
Commit Graph

188335 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Timur Kristóf
12c0c20ed1 radv: Remove superfluous bool arg from slot_to_ps_input. 
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28764>
 2024-04-18 18:35:07 +00:00
Timur Kristóf
a4cf36998f radv: Refactor emitting PS input types. 
Instead of taking a lot of booleans, refactor offset_to_ps_input
to take an enum which represents the PS input type. This helps
code readability as we keep adding more and more input types.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28764>
 2024-04-18 18:35:07 +00:00
Timur Kristóf
c471aed748 radv: Only consider interpolated inputs as 16-bit float. 
Enabling FP16_INTERP_MODE makes no sense for other types of inputs.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28764>
 2024-04-18 18:35:07 +00:00
Timur Kristóf
afb08ee244 radv: Run DCE before deleting I/O variables. 
Apparently, nir_lower_io leaves dead code in shaders, which
prevented us from deleting the IO variables properly.

Fixes: dbfb96f08f
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28764>
 2024-04-18 18:35:07 +00:00
Boris Brezillon
0c1cf20e83 panvk/ci: Re-enable copy_and_blit tests 
Apparently something fixed the copy tests that were failing.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28811>
 2024-04-18 18:09:58 +00:00
Boris Brezillon
21775a459f panvk: Add support for KHR_push_descriptor 
Most of the panvk_descriptor_set logic can be re-used, we just need
to provide intermediate helpers that can be used for both the push set
and regular set population logic (which implies passing our own storage
for the SW descriptor UBO instead of assuming it's always backed by a
BO).

Once this is done, we add temporary storage to the panvk_cmd_buffer
object, and populate the push sets at draw time.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28178>
 2024-04-18 17:26:49 +00:00
Mike Blumenkrantz
160dd5bf2b zink: add VK_PIPELINE_CREATE_CAPTURE_STATISTICS_BIT_KHR for shaderdb 
cc: mesa-stable

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28815>
 2024-04-18 16:34:37 +00:00
Mike Blumenkrantz
fd6468a5ae zink: destroy shaderdb pipelines 
cc: mesa-stable

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28815>
 2024-04-18 16:34:37 +00:00
Mike Blumenkrantz
b7b51295bf zink: add ZINK_DEBUG=nopc to completely disable precompilation 
I hack this in all the time for debugging anyway

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28766>
 2024-04-18 16:07:07 +00:00
Mike Blumenkrantz
a83fd26d0e nir/print: stop trying to match i/o vars using base/driver_location 
this is broken

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28552>
 2024-04-18 15:30:47 +00:00
Boris Brezillon
52cc12b8d2 panvk/ci: Enable dEQP-VK.pipeline.monolithic.* 
We regressed a pipeline.monolithic test when switching to dynamic
rendering and this went unnoticed because
dEQP-VK.pipeline.monolithic.* is not run.

With this addition, execution time is too long to fit in a single job,
so we split VK tests in 2 and only run half of the tests, which should
be okay as long as we keep the tests manual.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
a8a0051086 panvk/ci: Make sure we catch GPU faults 
PAN_MESA_DEBUG=sync allows us to abort on faults.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
d49ca71554 panvk: Abort on fault when PANVK_DEBUG=sync 
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
9a8dae1bf8 panvk: Make sure we use the proper format for views of depth+stencil images 
The view format doesn't take into account the internal image format
layout when only one aspect is selected. We need to patch the pipe_format
manually in that case.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
7d8bfe405d panvk: Skip tiler jobs when the vertex shader doesn't write the position 
Fixes the dEQP-VK.pipeline.monolithic.no_position.* tests.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Erik Faye-Lund
d5c2d152a5 panfrost: implement a driver-specific max-miplevel 
The previous define was common for all panfrost code, but for Vulkan we
should probably use the real hardware-limits. So let's introduce a
GL-driver specific define here, and use that instead. This lets us
change the other define to the real HW limitation.

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
2b1db3de7b panvk: Fill maxCustomBorderColorSamplers 
dEQP-VK doesn't seem to test it, but the validation layer complains.
Pick a randomly high value, since there's no physical limit on the
custom border color (this is part of the sampler descriptor).

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
10a631f767 panvk: Re-order things in panvk_physical_device_init() 
We need to know the GPU architecture when we initialize the physical
device properties/features. This implies creating the kmod device and
querying its properties early.

While at it, simplify the error path so we just have one entry point.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
a261e6c4ea panvk: Swizzle the border color on v7 when the format is BGR 
pan_texture.c tweaks the texture swizzle for BGR formats on v7 to
support AFBC(BGR). We need to take that into account when preparing
the border color.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
5e07edd4f6 panvk: Don't advertize vertex_buffer cap on sRGB formats 
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
294406bba2 panvk: Make sure we have a decode context created when we need one 
The sync and dump options also need a valid decode context.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
0eec8b0ff6 panvk: Fix depth/stencil image views 
Vulkan wants an R001 pattern, where the depth/stencil is stored in
the red component, but the pan_format/texture logic gives us RRRR.
Tweak the swizzle so we get what Vulkan wants.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
b40dad1d59 panvk: Make sure we pick a valid wrap_mode_r value for unnormalizedCoordinates 
When unnormalizedCoordinates is true, we can only use CLAMP_TO_EDGE or
CLAMP_TO_BORDER. This also puts a constraint on the viewType (1D or 2D),
which means we should ignore addressModeW and pick a default that works
fine with the unnormalizedCoordinates case.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
b9fe15e47f panvk: Fix has_non_vs_attribute() test in panvk_draw_prepare_vs_attribs() 
attrib_count should be checked against pipeline->attribs.attrib_count
not pipeline->attribs.buf_count.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
58a5ae3f72 panvk: Fix img2buf copies with image X offset not aligned on 16 pixels 
There are probably better ways of handling copies, but the whole thing
will disappear with the transition to vk_meta. The only reason we fix
it now is so we can enable more tests in CI and catch potential
regressions.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
7b06393257 panvk: Don't assume pViewportState != NULL 
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
f1d8a65d7f pan/bi: Extend bi_emit_texc() to support wider direct tex/sampler idx 
The current code was doing some assumptions on the maximum
texture/sampler sizes which no longer stand with panvk. Add explicit
range checks on the direct texture/sampler_index.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
feffe7fd2a panvk: We don't support resolve operations yet 
Make sure we crash when resolveMode != VK_RESOLVE_MODE_NONE in the
vkCmdBeginRendering() path instead of silently ignoring resolve
arguments.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
3b6c2cf0f0 panvk: Make sure the sample_pattern is set in the tiler descriptor 
This field should match the one defined in the framebuffer descriptor.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:57 +00:00
Boris Brezillon
47a98bbe34 panvk: Fill pan_tls_info::wls::instances 
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:56 +00:00
Boris Brezillon
3c3476bae8 panvk: Fix input attachment support 
Lower input attachments to texture operations as it was meant to be
according to the set layout accounting, and fix the descriptor set
logic to fill texture descriptors instead of image descriptors.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:56 +00:00
Boris Brezillon
502a1ca300 pan/bi: Allow subpass sampler dims 
Those are just regular 2D[MS] textures.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:56 +00:00
Boris Brezillon
51ab0a4c25 panvk: Don't assume VkGraphicsPipelineCreateInfo::pColorBlendState != NULL 
When the subpass being associated to a graphics pipeline has no color
attachment, pColorBlendState can be NULL.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:56 +00:00
Boris Brezillon
303f04d346 pan/bi: Support fragment store_output() with a non-zero offset 
Fragment color outputs can be defined as a vec4 array. As long as the
indexing is not dynamic, we can add the constant offset passed to
the store_output() intrinsic to the RT index we extracted from the
location.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:56 +00:00
Boris Brezillon
df4e3e4777 panvk: Fix the colorAttachmentCount check in begin_rendering_init_fbinfo() 
colorAttachmentCount should be less that or equal the number of RT slots,
not strictly less than.

Fixes: 595d362d4b ("panvk: Implement dynamic rendering entry points")
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:56 +00:00
Boris Brezillon
cb9ad5e44b panvk: Fix attach-less rendering 
VkRenderingInfo without any attachment is a perfectly valid use case.
Make sure we don't end up asserting on fbinfo width/height being zero
in that case.

Fixes: 595d362d4b panvk: Implement dynamic rendering entry points
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
 2024-04-18 15:04:56 +00:00
Eric Engestrom
b3e6ef964f util: simplify loop logic in util_format_get_first_non_void_channel() 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28800>
 2024-04-18 14:26:37 +00:00
Eric Engestrom
71b93f63dd radv: initialize a couple of variables 
Because musl will lose (in the next commit) the ability to see that they are initialized...

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28800>
 2024-04-18 14:26:37 +00:00
Eric Engestrom
05c4c17f91 wsi/x11: drop unused param in x11_present_to_x11_sw() 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28806>
 2024-04-18 13:51:48 +00:00
Eric Engestrom
8b53127e9f docs/rusticl: add an intro explaining what Rusticl is 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28807>
 2024-04-18 13:32:45 +00:00
Eric Engestrom
ae5231a0de vk/overlay-layer: simplify print and make it more readable 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28805>
 2024-04-18 12:56:39 +00:00
Eric Engestrom
0e61560266 vk/overlay-layer: fix None checks 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28805>
 2024-04-18 12:56:39 +00:00
Eric Engestrom
fc5fa08d89 vk/overlay-layer: drop unused imports 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28805>
 2024-04-18 12:56:39 +00:00
Eric Engestrom
acf1c7dc73 lavapipe: add 1 new failure and 1 new timeout since CTS uprev to 1.3.8.0 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28803>
 2024-04-18 12:22:19 +00:00
Eric Engestrom
aeab27cc24 egl: drop dead dri2_dpy param in dri2_wl_visual_idx_from_config() 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28528>
 2024-04-18 11:22:49 +00:00
Eric Engestrom
7350b65669 etnaviv: avoid re-defining prog_python 
It's already defined to the exact same thing in the root `meson.build`

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28801>
 2024-04-18 10:03:08 +00:00
Samuel Pitoiset
b67f9b4d0d radV/ci: reduce the parallelism of navi21 to 3 
With CTS 1.3.8.0 the execution time has been reduced a lot.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28795>
 2024-04-18 09:24:22 +00:00
Samuel Pitoiset
6ee39b1870 radv/ci: update lists for TAHITI and Zink/Polaris10 
1f4662cc4e introduced regressions.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28795>
 2024-04-18 09:24:22 +00:00
Samuel Pitoiset
e85b2641a1 radv/ci: add one more flake since CTS 1.3.8.0 for RENOIR 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28795>
 2024-04-18 09:24:22 +00:00
Iago Toral Quiroga
1070c9b0e7 broadcom/compiler: enable perquad with uses_wide_subgroup_intrinsics 
This fixes a number of regressions in Vulkan subgroups tests in CTS.

Fixes: 97f5721bfc ('broadcom/compiler: needs_quad_helper_invocation enable PER_QUAD TMU access')
cc: mesa-stable

Reviewed-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28797>
 2024-04-18 09:00:56 +00:00