We want to give developers the option to run their jobs on devices
that are still being stabilized in the CI infrastructure.
These jobs should be optional and not prevent merging from happening.
The is-forked-branch-or-pre-merge anchor was not being used anywhere,
so it was changed to is-forked-branch-or-pre-merge-not-for-marge to
create this new rule.
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10405>
The prototype uses a pointer and the actual function definition had an
array. For some reason, GCC never complained about this until GCC 11.
This fixes a compile warning when building with GCC 11.
Fixes: 09ced65420 "intel/isl: Add format conversion code"
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10537>
Not all varying fetches could be pulled into the start block.
If there are fetches we couldn't pull, like load_interpolated_input
with offset which depends on a non-reorderable ssbo load or on a
phi node, this pass is skipped since it would be hard to find a place
to set (ei) flag (beside at the very end).
We also don't have to manually set (ei) in such cases since a5xx and
a6xx do automatically release varying storage at the end.
Earlier gens need further testing, however they do not support
interpolateAt* functions at moment, so unless we would like to support
sample shading on them - they are fine.
Fixes crash in GTA V.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10483>
This patch should not be backported directly to 21.1, as master already
fixed one failure. I'll post a backport of this series with the
additional failure documented when this one gets merged.
v2:
- remove dEQP-VK.synchronization.* from the skip list (Hakzsam)
- drop dEQP-VK.memory.pipeline_barrier.* from the skip list (Hakzsam, me)
v3:
- re-introduce dEQP-VK.memory.pipeline_barrier.transfer_src_transfer_dst.1048576
Signed-off-by: Martin Peres <martin.peres@mupuf.org>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10539>
Some extensions have renamed their property and/or feature structs,
listing the "correct" type first. So we should prefer that one rather
than overwriting it with a later one.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10101>
This finally enables the 16-bit float feature.
Ideally we would also check VK_KHR_shader_float16_int8, but the python
code for that is giving me some issues now, so let's deal with that
later.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10101>
SPIR-V doesn't seem to have any opcodes to sample textures using lower
precision directly, so let's sample and downcast later instead.
Ideally, we'd do this as a NIR-pass first, but this does the trick for
now.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10101>
This prepares us for being able to support using 16-bit float types
in shaders, which might help performance in some cases.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10101>
The mysterious support_16bit_int_alu-option doesn't really mean what it
says. Instead it means "we support 16 bit compares, if 16 bit ALU
operations occur". And since 16 bit operations only appear if we're
lowering mediump/lowp, we can always set this option.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10101>
This prepares us for being able to support using 16-bit int types in
shaders, which might help performance in some cases.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10101>
We're about to need this in order to support 16-bit floats, because the
lowering code for that emits function-temp derefs, and we don't handle
it.
A better long-term solution would be to just support function-temp
variables and indirect derefs. But that's more work, and kinda
orthogonal to what this patchset tries to accomplish, so let's save that
for another day.
Fixes the following piglit:
- spec@arb_gl_spirv@execution@ubo@array-inside-ubo-copy
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10101>
This pass is needed to finish off the [ui]2imp lowering. Follow what
other drivers do and perform some dead-code elimimation etc when
lowering happens.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10101>
This is a 32-bit argument, so on platforms where UINT_MAX is larger,
this is going to... well, do exactly the same thing, but this is
slightly clearer why.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10101>
With the missing else, this prints the compacted hex followed by hex
for an uncompacted version of the compacted instruction. It also
doesn't print hex for instructions that are not compacted.
Fixes: bc4a127d6e ("intel/disasm: Label support in shader disassembly for UIP/JIP")
Fixes: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4245
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10535>
the only case in which this is nonzero is if a multidraw gets split by the frontend,
i.e., mesa core, and in all other cases it can be ignored. the value can also be ignored
for all indirect draws, though it seems many (most?) gallium drivers are not aware of this
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10166>
this moves index_bias into the multidraw struct, enabling draws where the value
changes to be merged; the draw_info struct member is renamed and moved to the end
of the struct for tc use
u_vbuf still has some checks to split draws if index_bias changes, maybe
this can be removed at some point?
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10166>
Section 7.1 (SHADER OBJECTS) of the OpenGL 4.6 spec says:
"An INVALID_VALUE error is generated if count is negative."
However a count of 0 is not an error. Previously it would cause a
GL_OUT_OF_MEMORY error.
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10477>
This change moves the SRVs associated with read-only SSBOs to be emitted
before any other UAV. We do this because the validator expects resources
to be emitted in a specific order, as noted by `emit_module`.
Previously, we emitted SSBOs as SRVs (read-only) or UAVs (read-write)
after other UAVs.
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10514>
We'd like to avoid __glXInitialize as much as possible since it involves
taking a global lock. This means converting internal APIs to operate as
much as possible in terms of something other than a Display *, since if
that's all you have then you're forced to call __glXInitialize to get to
the glx_display.
The contortions in DRI2 displease me, but DRI2 displeases me, so.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10418>
1. pre-allocate command pools at device creation if anb enabled
2. force VK_SHARING_MODE_CONCURRENT for wsi image if necessary
3. pre-allocate and record command buffers at android wsi image creation
4. transfer in the ownership at vkAcquireImageANDROID
5. transfer out the ownership at vkQueueSignalReleaseImageANDROID
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10335>
Whenver VK_IMAGE_LAYOUT_PRESENT_SRC_KHR is used, replace it with
VK_IMAGE_LAYOUT_GENERAL as required for proper layout and ownership
transfer for external memory backed swapchain images.
This will be Android only until common WSI is fixed.
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10335>
If we give radv_pipeline_cache_insert_shaders() a variant which was
retrived from the cache, it will free it.
I haven't encountered this issue, but it seems possible.
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/10433>
If radv_pipeline_cache_insert_shaders() finds a GS copy shader in the
cache, it will free the variant in gs_variants and replace it with the one
in the cache.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10433>
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Fixes: bd1705a480 ("vulkan: Make vk_debug_report_callback derive from vk_object_base")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10433>
