meson: Bump libdrm for amdgpu to 2.4.107.

For capture/replay va stability.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10570>

glsl: Prohibit implicit conversion of mem parameter in atomicOP functions

Per OpenGL Shading Language, section 8.11. "Atomic Memory Functions"
first argument "mem" of all atomicOP functions is inout.
The same is true for ARB_shader_storage_buffer_object and
GL_INTEL_shader_atomic_float_minmax

For implicit conversion of inout parameters it is required for type
to support bi-directional conversion, since there is no such types
in glsl - implicit conversion is effectively prohibited.

Alternatively we could have marked atomic_var parameter of built-in
atomicOP functions as inout, however it opens another can of worms
during NIR lowerings.

Fixes: ea0a1f5beb22982a886ba862ba95f92c9e35165a
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/2837
Signed-off-by: Danylo Piliaiev <danylo.piliaiev@globallogic.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4887>

zink: ci updates for wideline fails

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11041>

lavapipe: wideLines support

easy enough

Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11041>

panvk: Add vkEvents support

Use syncobjs to implement vkEvents (as suggested by Boris).

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11709>

panfrost: Specify alignment for the Job Header descriptor

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11709>

panvk: Make panvk_queue_transfer_sync more generic

Have it accept a syncobj so it can be used in a future commit.

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11709>

panvk: Support calls to CreateDescriptorSetLayout without bindings

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11709>

ci: Don't artifact rendered images when job succeeds

Signed-off-by: Rohan Garg <rohan.garg@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11088>

ci: remove few CTS that are now skipped with RADV

Some expected list of failures were actually outdated.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11712>

ci: update list of expected failures against CTS 1.2.6.2 for RADV

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11712>

lavapipe: implement VK_EXT_line_rasterization

rectangular and strict lines aren't supported in this, and multisampling
must be disabled for correct line rasterization

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11066>

lavapipe: store the geometry shader prim type to render state

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11066>

lavapipe: store whether the geometry shader outputs GL_LINES

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11066>

panvk: Add VkCommandPool support

Mostly just copied from turnip.

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11695>

panvk: Support returning BOs allocated by panvk_pool to a 'free BO' pool

So all CommandBuffers in a given CommandPool can reuse BOs for their
memory pools.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11695>

panfrost: Fork pan_pool for Gallium and Vulkan

This commit adds the actual implementations, allowing to diverge while
still sharing code that depends on pool functionality.

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Suggested-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11695>

panfrost: Start splitting the panfrost pool logic

The Gallium and Vulkan drivers will soon use different memory pool
implementation, but some pieces in libpanfrost depend on pan_pool. Let's
split the implementation so we have common bits still available while
letting the drivers implement what really matters: the allocation logic.

All the generic pieces are prefixed pan_pool, and what will become the
gallium implementation is prefixed panfrost_pool. We'll then duplicate
the panfrost_pool bits in panvk and prefix it with panvk_pool, and
implementations will start diverging from there.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11695>

panvk: Use the desc alloctor when we can

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11695>

panfrost: Allocate WRITE_VALUE jobs with panfrost_pool_alloc_desc()

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11695>

panfrost: Add alignment info to write-value and cache-flush jobs

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11695>

panfrost: Pass a memory pool to pan_blit_ctx_init()

Pass a memory pool to pan_blit_ctx_init() instead of creating a new pool.
Useful for Vulkan since the descriptor pool is at the command buffer
level and is thus shared by all blit batches. Doing this will save us a
BO ownership transfer.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11695>

panfrost: Don't add blit context BOs twice

The transient_bo has already been added to the BO list, no need to call
panfrost_batch_add_bo() a second time on the same BO.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11695>

crocus: restrict prim_restart on index buffer check to pre-hsw

This code has no use on hsw or chv

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11707>

crocus: reorder version checks on indirect xfb

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11707>

crocus: inline group_index<->bti

this is on a fastpath for ubo emission

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11707>

crocus: optimise bo_unref path a little.

This just splits it into the atomic/non-atomic paths

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11707>

crocus: don't update draw parameters unless needed

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11707>

crocus: inline the d/s resource handling functions

These are pretty simple, so inlining is fine and helps drawoverhead

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11707>

draw/tess: write correct primitive id into vertices

The code was using a prim assembler after the tess stage, however
tess prims aren't necessarily the output prim types, so just put
the prim ids into the vertices at tess stage, and skip prim assembly.

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11000>

draw: fix tessellation output vertex size calculation

This ensures space for the extra outputs is calculated in the
tes vertex outputs.

dEQP-VK.pipeline.misc.primitive_id_from_tess

Fixes: dacf8f5f5c82 ("draw: hook up final bits of tessellation")
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11000>

crocus: fixup index buffer dirtying.

This fixes a possible problem if a non-indexed draw comes in first
in a new batch, then the batch might not emit the index buffer.

I'm unsure if we see this, I just spotted it trying to fix alacritty

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11705>

crocus: fix crash on index buffer rebinding.

This was crashing in plasmashell.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11704>

crocus/gen5: enable support for GL_EXT_gpu_shader4

There has been a few requests for this and I've got patches for glamor
to use this to speed up the paths that use GLSL 1.30 now.

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11693>

crocus: expose ARB_blend_func_extended on gen 45/50

In theory the docs say 965gm can support this but the original
965 can't, but we can't distinguish that yet.

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11693>

crocus: cleanup some deadcode in the gen5 blend emit

The rt->blend_enable is always 1 at this point, and the gen5
specific code isn't being used anymore, it just looked like it
was.

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11693>

nir/ifind_msb_rev: fix input check

ifind_msb_rev was introduced in a5747f8ab357ff00c630b937b221e5fb59d90289.

ifind_msb_rev guards against src0 being both 0 or -1 at the same time.
That is always true. This patch changes it to check for those values
individually.

Spotted from a compile warning.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Fixes: a5747f8ab35 (\"nir: add opcodes for *find_msb_rev and lowering\")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11630>

iris: make iris_bind_reserve_3d and Wa_1604061319 only check for dirty render bindings

+    9.31%  drawover:gdrv0  iris_dri.so      [.] iris_binder_reserve_3d

+    2.36%  drawover:gdrv0  iris_dri.so         [.] iris_binder_reserve_3d

If the app never uses compute, then the compute bindings bit will always
be dirty causing these two paths never get shortcuts.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11699>

panfrost: Only upload UBOs when needed

If all of the used values from a UBO are pushed, it doesn't need to be
uploaded.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11700>

pan/mdg: Create a mask of UBOs that need to be uploaded

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11700>

pan/bi: Create a mask of UBOs that need to be uploaded

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11700>

panfrost: Don't set dirty_mask for constant buffers

It is unused.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11700>

crocus: Make the driver loader use PCI IDs for crocus

Add PCI IDs based in pci_ids/i965_pci_ids.h and move crocus before
iris in driver_map[].

This allows Xorg to load the crocus driver since iris would claim
the devices handled by crocus (because the i915 kernel driver is
used for all Intel devices) then fail during initialization.

Signed-off-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11694>

crocus: Add pipe loader driver

Signed-off-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11694>

vulkan/wsi: Fix prime blits to use system memory for the destination

The intention here was to pass VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT to
select_memory_types() when requesting device local memory, or simply
pass 0 for the prime blit destination which should be in system memory.

Unfortunately, that meant we did (type.propertyFlags & 0) == 0 which
was vacuously true, causing us to not filter out device local types.

Fixes hybrid display of Vulkan apps on Intel TGL+DG1 systems.

Tested-by: Luis Felipe Strano Moraes <luis.strano@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11680>

i915g: Make sure the 1D texture Y channel is initialized.

Even with the wrap mode forced to REPEAT, we get undefined results in
CelShading when the Y channel is unwritten since the beginning of the
program.

I dropped the coords==0 case in the process, since that's not possible and
made the 1D case confusing.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11457>

i915g: Force 1D textures to use wrap mode for the Y coordinate.

There are no 1D textures in HW, so we use 2D, but at the shader level
there no Y coordinate to 1D sampling, so we need that value to be ignored.
WRAP mode can get us that.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11457>

meson: Make prefer-{crocus,iris} always take effect

As written this would require that the driver be built before we looked
at the option. This is wrong because it affects code outside of the
driver, it's in libGL's PCI ID table. This is sort of harmless for
crocus at the moment, but for iris you would need to build it in order
to remove it from the table; if you built just i965 and tried to run it
against gen9, the libGL you just built would direct the loader to the
iris driver you just didn't, and setup would fail, which is: goofy.

Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11655>

iris: Delete unused bo->cache_coherent flag

This was used in the heuristics for deciding whether to CPU map,
but those have since been deleted, so this is now unused.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11644>

iris: Fail BO allocation if we can't enable snooping properly.

If the caller has asked for a coherent BO with snooping, and the kernel
fails to set it for whatever reason, we were happily returning them a
non-coherent buffer.  This isn't what they wanted and could lead to
surprising results.

Better to simply fail the allocation.  Probably.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11644>

iris: Stop calling I915_GEM_SET_CACHING on discrete GPUs

On integrated GPUs without LLC, we enable snooping when someone requests
coherency for a buffer.  (With LLC, it's already coherent.)

For discrete GPUs...if someone requests coherency, we allocate the
buffer in SMEM and resort to WC maps rather than WB maps with CPU
caches enabled.  There's no snooping to enable, and calling this ioctl
is nonsensical, and may fail.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11644>

i915g: whitespace fixup from the cube map fix.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11668>

i915g: Add support for per-vertex point size.

Closes: #4973
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11668>

radv: allow more fast clears for depth surfaces without TC-compat HTILE

With HTILE only, all values between 0.0 and 1.0 are fetchable.
This should allow more fast clears for depth surfaces where
TC-compat HTILE is disabled.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10035>

radv: prevent fast clearing HTILE depth for unrestricted ranges

VK_EXT_depth_range_unrestricted removes the restriction that the
clear value must be between 0.0 and 1.0.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10035>

radv: add support for more HTILE clear codes

The HTILE clear code is now computed based on the floating point value.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10035>

radv: advertise VK_EXT_color_write_enable

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11530>

radv: implement VK_EXT_color_write_enable

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11530>

Fix up leftover "state_trackers" references to "frontends"

Fixes: d6287a94b69 "gallium: rename 'state tracker' to 'frontend'"
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11676>

nir_lower_readonly_images: Clear variable data when changing the type

For images, variable data includes the format. For samplers, variable
data is used for OpenCL inline samplers. When converting a variable
from one to the other, zero out the data so we don't accidentally
interpret a converted image as an inline sampler.

Fixes: fa677c86 ("nir_lower_readonly_images_to_tex: Support non-CL semantics")
Acked-by: Enrico Galli <enrico.galli@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11674>

zink: check last_finished before timeline waiting

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11686>

zink: add a screen function for waiting on a batch id

this is super hacky for the non-timeline case but whatever

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11686>

zink: stop screwing up buffer offsets during for maps

stop using the memory bind offset

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11687>

zink: use 0 as the offset when mapping qbos

zink_resource_object::offset is the memory offset, which is unrelated to
the offset in a buffer

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11687>

zink: use pipe_resource::width0 for clamping ssbo sizes

zink_resource_object::size is the size of the memory allocation, which
is effectively unrelated

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11687>

zink: don't add mem allocation offset when copying buf2image

this doesn't apply to any image/buffer operations, it's solely for memory
allocation/binding/mapping

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11687>

zink: only flag persistent resource maps for invalidation if they aren't coherent

coherent is already synchronized

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11688>

zink: use vkresult helper for map return

consistency++

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11689>

util/disk_cache: Don't leak when cache is empty

When we exit early having failed to find any candidate cache files to
evict, don't leak the list head whilst doing so.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Fixes: f58e6fee7452 ("util/disk_cache: delete more cache items in one go when full")
Ref: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11523>

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11682>

zink: add resource refs after last descriptor unbind

redundant at this point; no functional changes

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11684>

zink: split samplerview/imageview usage/refcounting calls

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11684>

zink: split batch usage setting from refcounting

batch tracking has until now been the same as refcounting, but this
is inefficient and incurs additional overhead from constant atomic ops,
so make them separate

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11684>

zink: split out resource tracking into more incremental functions

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11684>

zink: remove stencil resource batch tracking

I think maybe I misunderstood how things worked, or there were issues
generally with resource lifetimes, or whatever, but this isn't needed anymore

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11684>

zink: don't add batch tracking during buffer rebinds if refs are dirty

this will be handled automatically

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11684>

zink: add more explicit casts to draw code

this makes c++ happy

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11685>

zink: add c++ header guards

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11685>

zink: remove unnecessary return from zink_desc_type_from_vktype()

this doesn't compile in c++

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11685>

zink: rename 'template' struct member

this is a c++ keyword

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11685>

zink: make prim type a bitfield in pipeline info

match draw info here

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11685>

nir: Fix constant folding for irhadd/urhadd

This should be a subtract, not an add. The comment's proof is correct,
but the (wrong) expression we actually use isn't what it's in the
comment! Correct the discrepancy.

The lowering in nir_opt_algebraic was correctly typed.

Fixes: 272e927d0e9 ("nir/spirv: initial handling of OpenCL.std extension opcodes")
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11671>

zink: rework pipeline cache implementation

this is now a screen-based queue which can be triggered to serialize
cache updates, ensuring synchronization

the cache is on the program object, enabling incremental updates as well as
variant loading for an entire pipeline collection at once

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11595>

zink: merge some streamout state emission into the same draw conditional

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11667>

zink: move line width and depth bias updating into conditional during draw

only do calculations when needed

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11667>

zink: unify fb surface unbinding

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11666>

zink: add a ref for flush resource

the lifetime of this needs to always be preserved

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11665>

zink: break up ctx descriptor state updating to per-type functions

this is a bit easier to read/manage going forward

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11664>

zink: unify/consolidate some barrier queuing

this queuing is based on bind counts, and it's all the same, so it
can mostly be combined for better reuse

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11663>

zink: avoid unnecessarily rewriting gl_DrawID

with tc enabled, we get genuine multidraws with valid drawids, so we can
update this to reflect the new capabilities...which are the same as the old
ones except that potentially some drivers can now do direct multidraws

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11662>

microsoft/compiler: Change behavior for emitting inexpressible barriers

If the barrier tries to apply to memory that we can't express, just
don't apply the memory portion of the barrier. Similarly, if it tries
to apply a global memory barrier at invocation level, upgrade it to
thread-group.

Reviewed-by: Enrico Galli <enrico.galli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11670>

docs/isl/tiling: Fix swizzle pattern for X-tiling

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11679>

lavapipe: Add a missing VKAPI_ATTR.

Signed-off-by: Georg Lehmann's avatarGeorg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11114>

lavapipe: Use common default allocator.

Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11114>

overlay_layer: add missing undef

reviewed-by: Dylan Baker <dylan@pnwbakers.com>

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11673>

meson: print information about layers being built as part of summary

reviewed-by: Dylan Baker <dylan@pnwbakers.com>

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11673>

gallium/u_threaded: do not apply start twice

We already apply start when setting the bits in writable_buffer, so
shifting the bits up by start just makes us apply the offset twice.

Caught by Coverity.

Fixes: 988d0917208 ("gallium/u_threaded: clear valid buffer range only if it's not bound for write")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11635>

radv: disable DCC for DOOM 2016 and Wolfenstein II

Both games perform two image layout transitions with the same image
in the same pipeline barrier with UNDEFINED and this re-initializes
DCC to the uncompressed state. No ideal solution sadly. Note that
both games declare all images as CONCURRENT.

This fixes rendering issues on GFX10+ because DCC for stores is
supported and this implicitly enables DCC for concurrent.

Fixes: da166f648f6 ("radv: enable DCC for concurrent images on GFX10")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4927
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4607
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11656>

zink: drop getenv hacking now that gallium is fixed.

This drops the getenv hacks which cause problems testing multithread apps

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11643>

gallium/sw: add sw_vk bit to avoid having to futz with env vars for lavapipe

lavapipe really only currently works with llvmpipe, and likely for the forseeable
future.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11643>

radv: enable VK_KHR_shader_subgroup_uniform_control_flow

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/11626>