nvk/image: Add image address helpers

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk/copy: Use nvk_buffer_address in CmdCopyBuffer

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk/buffer: Add a push_buffer_ref helper

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk/buffer: Take an offset in nvk_buffer_address

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: report maxMipLevels as 1

the CTS uses that value in some places

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: add support for blits

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: simple format table

this is enough to run the image copy and blit tests on a couple of formats

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: add basic support for images

enough to get basic operations working

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: use fermi class definitions

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nouveau/ws: bind 2D class

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nouveau/ws: set GPU object class

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nouveau/ws: PUSH_IMMD only works with 16 bit values

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: Re-format descriptor sets some more

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: Re-format pipeline layouts

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: Re-format descriptor set layouts

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: descriptor set freeing fix

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: implement buffer address.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: add descriptor set bo allocation.

This allocates the backing bo for the pool, and allocates the
sets from it.

This is mostly ported from radv.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: Expose nvk_descriptor_stride_align_for_type

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: some boilerplate for descriptor sets

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: add initial descriptor pool framework.

destruct has internal impl for create to call on errors later.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: reindent descriptor sets to mesa std.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: Implement vkUpdateDescriptorSets

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: Add initial descriptor set lowering

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: Implement VkPipelineLayout

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: Implement VkDescriptorSetLayout

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: Reset pushbufs on command buffer reset

Karol: also reset on VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nouveau/ws: fix setting push bo domains

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: impl nvk_CmdCopyBuffer

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: add nvk_CmdPipelineBarrier2 stub

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: add nvk_bo_sync

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: add cmd buffer framework

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: allocate a GPU context for each VkDevice

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: Add initial queue

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

novueau/bo: add nouveau_ws_bo_wait

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

novueau/bo: refcount it

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nouveau/ws: add a cmd buffer

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nouveau/ws: add context support

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nouveau/headers: add host classes

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nouveau/headers: typedef Nv void types

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nouveau/headers: add nvtypes.h

Nvidia does provide an nvtypes.h file within their open-gpu-kernel-modules,
but that one is painful to port over to userspace.

Just provide the defines we will need instead.

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: add bind buffer memory

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: fix header guards to be less generic.

You could imagine want to use these defines for something
that isn't header guards, so just make the headers guards
more explicit.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: add some initial wsi framework.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: add missing finish calls

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: add basic device memory support

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nouveau/ws: add bo API

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: set nonCoherentAtomSize as the CTS divides with this value

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nouveau/ws: add a field for the SM version

This is very useful when looking up capabitilites in Nvidia CUDA docs

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nouveau/ws: dup the fd

We do the same in gallium

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nouveau/ws: reorganize a little

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: advertize memory heaps and types

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: implement GetPhysicalDeviceQueueFamilyProperties2 to make the CTS happy

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nouveau/headers: initial sync of headers

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nouveau/headers: add script to sync in-tree headers with open-gpu-doc

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: fix nvk_buffer include guards

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: Add a stub implementation of VkBuffer

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: Add stub implementation of VkSampler

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: Add stub implementations of VkImage and VkImageView

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: Implement EnumerateInstanceVersion

This is required for future Vulkan versions.  The common code also calls
it for validating Vulkan versions.  May as well implement it now.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: use winsys lib

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nouveau/winsys: add the new winsys implementation

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

nvk: add vulkan skeleton

This is enough to run vulkaninfo without crashing.

Jason:
* Drop a redundant nvk_device_entrypoints
* Add some VKAPI_ATTR and VKAPI_CALL
* nvk: Move EnumerateInstanceExtensionProperties to the top
  This way things are more-or-less in initialization order.  First the
  version then extensions then create the instance.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24326>

r600/sfn: Use clause local registers in RA

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24480>

r600/sfn: Track whether a register is ALU clause local

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24480>

r600/sfn: Renumber shader blocks in scheduler

Consistent block numbering is required for identifying registers
that are used clause-locally.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24480>

radv/ci: skip some mesh shader tests on GFX1100

This lets me do a GFX11 CTS run without RADV_DEBUG=nocompute.

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/24499>

zink/ci: set the default timeout for zink jobs to 30min instead of 1h

Except zink-anv-tgl which takes a lot more, and keeps the 1h timeout
(for now, future changes might improve this).

Suggested-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24500>

egl/dri2: Add a couple of missing mutex release in error path

This is causing a hang in the intel compute runtime when
trying to cl/gl share an unsupported texture format.

Fixes: f1efe037dfd82d95b025 ("egl/dri2: Add display lock")

Signed-off-by: Sylvain Munaut <tnt@246tNt.com>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24487>

clang-format: Set ColumnLimit to 78

This matches the tree-wide .editorconfig.

Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24429>

ci: use mold to build deqp

deqp is probably the biggest thing that we're building (definitely in
size of binary linked), so it's also the one that benefits the most from
using mold for linking.

Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24490>

ci: include mold in x86_64_test-base & rootfs images

Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24490>

vulkan: Use VkBufferUsageFlags2 in vk_buffer

Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24497>

n50/compute: submit initial compute state in nv50_screen_create

This fixes some CACHE_ERROR caused by proper multi-threading support. The
bug is a bit older though, just never triggered because there was only one
push buffer to begin with.

Without this change the compute initialization stayed unpushed in the
screen push buffer causing random issues.

Fixes: ff72440b402 ("nv50: implement a basic compute support")
Signed-off-by: Karol Herbst <git@karolherbst.de>
Reviewed-by: M Henning <drawoc@darkrefraction.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24496>

vulkan/format: add VK_FORMAT_{A8_UNORM,A1B5G5R5_UNORM_PACK16}_KHR

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24365>

vulkan/runtime: add common implementation for GetImageSubresourceLayout()

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24365>

vulkan/runtime: add a layered implementation of vkCmdBindIndexBuffer

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24365>

vulkan: bump header register to 1.3.261

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Acked-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24365>

vulkan: skip non required extension structures

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24365>

Revert "compiler: Combine duplicated implementation of is_gl_identifier into glsl_types.h"

This reverts commit f9860a84b3cf1735ece78bc5907b2be4e2878f8e.  It's a
bit annoying having this scattered around but it's 100% a GLSL thing and
there's no reason why it should go in glsl_types.h.  The fact that
glsl_print_type() even uses it is a bit sketchy.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24491>

Revert "mesa, compiler: Move gl_texture_index to glsl_types.h"

This reverts commit 1b836a52ea2dcc0b77a48a2a3d1f9052321cf325.  This
patch, while claiming to decouple things, actually increases coupling
because it leaks two OpenGL state tracker limits and an OpenGL state
tracker fixed binding enum into the entire compiler.  Nothing wants to
know these outside the OpenGL state tracker and the GL-specific compiler
passes.  Put them back where they were.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24491>

rusticl/kernel: mark can_remove_var as unsafe and document it

Signed-off-by: Karol Herbst <git@karolherbst.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24470>

rusticl/kernel: get rid of initial function_temp type lowering

It's not needed and causes some issues with generic memory.

Fixes: 01cb01a7ee9 ("rusticl/kernel: only handle function_temp memory before lowering printf")
Signed-off-by: Karol Herbst <git@karolherbst.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24470>

rusticl/kernel: move things around in lower_and_optimize_nir

Signed-off-by: Karol Herbst <git@karolherbst.de>
Reviewed-by: @LingMan <18294-LingMan@users.noreply.gitlab.freedesktop.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24470>

rusticl/kernel: merge lower_and_optimize_nir_pre_inputs and lower_and_optimize_nir_late

Signed-off-by: Karol Herbst <git@karolherbst.de>
Reviewed-by: @LingMan <18294-LingMan@users.noreply.gitlab.freedesktop.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24470>

rusticl/kernel: rename res to internal_args inside lower_and_optimize_nir_late

Signed-off-by: Karol Herbst <git@karolherbst.de>
Reviewed-by: @LingMan <18294-LingMan@users.noreply.gitlab.freedesktop.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24470>

radv: track if vertex binding stride is dynamic from the cmdbuf state

This allows us to remove one more pipeline occurence during cmdbuf
recording. Note that shader object always uses dynamic vertex input.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24473>

radv: re-emit binning state if the framebuffer is dirty

This used to depend on the graphics pipeline, but now that everything
is dynamic it should be fine to trigger it on fb changes.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24473>

radv: track if patch control points is dynamic from the cmdbuf state

This allows us to remove one more pipeline occurence during cmdbuf
recording. Note that patch control points is always dynamic with
shader object.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24473>

radv: remove redundant check in radv_cmd_buffer_after_draw()

RADV_CMD_FLAG_PS_PARTIAL_FLUSH is only used for draws with
RADV_DEBUG=syncshaders, which implies a valid graphics pipeline.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24473>

radv: don't set vertex_attribute_strides on GFX8+

Bounds checking no longer compares against the stride.

fossil-db (navi21):
Totals from 5 (0.00% of 133461) affected shaders:
Instrs: 1532 -> 1520 (-0.78%)
CodeSize: 8144 -> 8084 (-0.74%)
Latency: 5104 -> 5102 (-0.04%)
InvThroughput: 686 -> 680 (-0.87%)
VClause: 19 -> 17 (-10.53%)
PreVGPRs: 118 -> 116 (-1.69%)

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/24115>

virgl: Only PIPE_BUFFER with VIRGL_BIND_CUSTOM flag is considered busy during creation

When the virglrenderer performs the attach action for the PIPE_BUFFER
with the VIRGL_BIND_CUSTOM flag, it will synchronize the data from
res->ptr to res->iov, at this time we need to treat it as busy,
otherwise it will cause some race conditions.

This optimizes commit:
fe9333f7b5e ("virgl: Set res->maybe_busy to true when creating resources")

Signed-off-by: Feng Jiang <jiangfeng@kylinos.cn>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18098>

intel/compiler: Fix sparse cube map array coordinate lowering

Brown paper bag fix for my untested review feedback comments.

Cube array images use a coordinate of the form <X, Y, 6*Slice+Face>,
while cube array textures use a <X, Y, Slice, Face> style coordinate.

This code tried to convert one to the other, but instead of writing
Z / 6 and Z % 6, we tried to reuse our original division result.  What
we wanted was Z - (Z/6) * 6, but instead we botched it and wrote Z-Z*6
which produced...totally invalid cube faces.

Fixes: fe81d40bff26 ("intel/nir: add lower for sparse images & textures")
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Tested-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24481>

radv/video: Match function definitions to declarations

This adds the same function attributes from definitions to declarations
and fixes the following compiler errors for 32 bit Windows OS.

../src/amd/vulkan/radv_video.c:270:1: error: conflicting types for 'radv_CreateVideoSessionKHR'
src/amd/vulkan/radv_entrypoints.h:3963:34: note: previous declaration of 'radv_CreateVideoSessionKHR'
../src/amd/vulkan/radv_video.c:319:1: error: conflicting types for 'radv_DestroyVideoSessionKHR'
src/amd/vulkan/radv_entrypoints.h:3971:30: note: previous declaration of 'radv_DestroyVideoSessionKHR'

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24460>

bin/pick-ui: use venv wrapper

Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24367>

bin/ci/update_traces_checksum: use venv wrapper

Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24367>

bin/ci/gitlab_gql: use venv wrapper

Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24367>

bin/ci/ci_run_n_monitor: use venv wrapper

Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24367>

bin: add wrapper to run scripts in a python venv

This isolates the script environment from the rest of the machine,
avoiding missing/incompatible dependencies and avoiding polluting the
rest of the machine with python packages.

Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24367>

nir: Simplify alu_instr_is_copy

mov is considered vec.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24432>