zink: cache render-passes

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: simplify renderpass/framebuffer logic a tad

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: implement batching

This reduces stalling quite a bit.

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: return after blitting

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: remove unusual alignment

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: tweak state handling

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: move primitive-topology stuff into program

The primitive topology is a bit of an odd-ball, as it's the only
truly draw-call specific state that needs to be passed to the program to
get a pipeline.

So let's make this a bit more explict, by passing it separately. This
makes the flow of data a bit easier to wrap your head around.

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: assign increasing locations to varyings

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: ensure textures are transitioned properly

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: ensure non-fragment shaders use lod-versions of texture

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: emit dedicated block for variables

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: use uvec for undefs

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: do not destroy staging-resource, deref it

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: track used resources

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: implement fmod

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: store shader_info in zink_shader

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: texture-rects?

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: delete samplers after the current cmdbuf

This makes them zombies for a little while.

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: add curr_cmdbuf-helper

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: reference blit/copy-region resources

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: whitespace cleanup

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: wait for idle on context-destroy

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: reference ubos and textures

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: reference vertex and index buffers

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: return old fence from zink_flush

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: reference renderpass and framebuffer from cmdbuf

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: cache those pipelines

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: move renderpass inside gfx pipeline state

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: cache programs

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: pass zink_render_pass to pipeline-creation

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: prepare for multiple cmdbufs

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: move cmdbuf-resetting into a helper

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: do not leak image-views

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: move render-pass begin to helper

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: prepare for caching of renderpases/framebuffers

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink/spirv: implement loops

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink/spirv: implement discard

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink/spirv: implement if-statements

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink/spirv: prepare for control-flow

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink/spirv: handle reading registers

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink/spirv: implement some integer ops

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink/spirv: store all values as uint.

This adds bitcasting to uint everywhere for now,
and stores all spir-v ssa values as uints.

It also casts bool to 0/0xffffffff for now
(nir 1-bit bools may be coming in the future).

This fixes a lot of piglit tests to pass now

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: remove discard_if

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: query support (v2)

This at least passes piglit occlusion_query test for me here now.

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: transform z-range

In vulkan, the Z-range of clip-space goes from 0..W instead of -W..+W
as is the case in OpenGL. So we need to transform the Z-range to
account for this.

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: add dri loader

export MESA_LOADER_DRIVER_OVERRIDE=zink should now work without using
swrast paths

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink/spirv: implement point-sprites

This passes glsl-fs-pointcoord_gles2 from piglit.

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: ask for flatshade lowering

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: detect presence of VK_KHR_maintenance1

Acked-by: Jordan Justen <jordan.l.justen@intel.com>

zink: introduce opengl over vulkan

Here's zink, a so far pretty simple vulkan-gallium driver that is able
to translate some applications from OpenGL to Vulkan.

The compiler is quite limited for now, this will be improved on later.

Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Acked-by: Jordan Justen <jordan.l.justen@intel.com>

radv: fix OpQuantizeToF16 for NaN on GFX6-7

Do not flush NaN to 0.

Fixes
dEQP-VK.spirv_assembly.instruction.compute.opquantize.propagated_nans

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: enable fast depth/stencil clears with separate aspects on GFX8

It's similar to GFX9+. Shadow of Mordor (Vulkan beta) hits that
path and it works fine.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

iris: Mark aux-map BO as used by all batches

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

iris/gen12: Write GFX_AUX_TABLE base address register

Rework:
 * Move last_aux_map_state to iris_batch. (Nanley, Ken)

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

iris: Map each surf to it's aux-surf in the aux-map tables

Rework: Nanley Chery
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

isl/gen12: 64k surface alignment

Reworks:
 * Update size for aux map change (Nanley)

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

iris/bufmgr: Initialize aux map context for gen12

Reworks:
 * free gen_buffer in gen_aux_map_buffer_free. (Rafael)
 * lock around aux_map_bos accesses. (Ken)

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

anv: Add aux-map translation for gen12+

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

anv/gen12: Write GFX_AUX_TABLE base address register

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

genxml/gen12: Add AUX MAP register definitions

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

anv/gen12: Initialize aux map context

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

intel/common: Add surface to aux map translation table support

Reworks:
 * Add ISL_FORMAT_B8G8R8X8_UNORM_SRGB to get_format_encoding (Nanley)
 * ralloc_free aux_map_buffer entries in gen_aux_map_finish. (Rafael)
 * verify_aligned_space => align_and_verify_space (Rafael)
 * Add mutex to aux-map code. (Rafael, Nanley)
 * Add gen_aux_map_fill_bos (Ken)
 * Make gen_aux_map_get_state_num lockless

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

anv: Implement aux-map allocator interface

This interface allows the aux-map code in the intel/common library to
allocate and free buffers.

Reworks:
 * free gen_buffer in gen_aux_map_buffer_free. (Rafael)

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

intel/common: Add interface to allocate device buffers

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/dev: store whether the device uses an aux map tables on devinfo

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965: setup sized internalformat for MESA_FORMAT_R10G10B10A2_UNORM

Commit d2b60e433e5 introduced restrictions (as per GLES spec) on the
internal format. We need to setup a sized format for the texture image
so framebuffers created with that are considered complete.

This change fixes following Android CTS test in AHardwareBufferNativeTests
category:

   SingleLayer_ColorTest_GpuColorOutputAndSampledImage_R10G10B10A2_UNORM

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Fixes: d2b60e433e5 ("mesa/main: R10G10B10_(A2) formats are not color renderable in ES")
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

tu: fix empty-body instruction

Fixes: 8d43e2b2ded0fe3c82d4 ("meson: add -Werror=empty-body to disallow `if(x);`")
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

v3d: fix empty-body instruction

Fixes: 8d43e2b2ded0fe3c82d4 ("meson: add -Werror=empty-body to disallow `if(x);`")
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

radv: fix empty-body instruction

Fixes: 8d43e2b2ded0fe3c82d4 ("meson: add -Werror=empty-body to disallow `if(x);`")
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv: fix empty-body instruction

Fixes: 8d43e2b2ded0fe3c82d4 ("meson: add -Werror=empty-body to disallow `if(x);`")
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

freedreno/a2xx: use sysval for pointcoord

Fixes a problem with shaders using gl_PointCoord.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reported-by: Fabio Estevam <festevam@gmail.com>
Tested-by: Fabio Estevam <festevam@gmail.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>

pan/midgard: Disable precise occlusion queries

I thought there was hardware support for this, but it seems to broken,
or at least more complex than I believed.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>

panfrost: allocate bo for occlusion query results

This memory needs to still be available after all the drawing is done
and forgotten about, so cannot be transient.
Also clear the result so that no rendering returns a zero.

Signed-off-by: Urja Rannikko <urjaman@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>

panfrost: Expose serialized NIR support

Serialized NIR is required for clover with the SPIR-V pipeline. With
this change and PAN_MESA_DEBUG=deqp, clinfo is able to successfully
probe panfrost.

Code from Nouveau (commit 7955fabcf89c7265f7f4244e46c5bcb83b9687fa by
Karol Herbst).

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>

pipe-loader: Default to kmsro if probe fails

A device supported by kmsro will not automatically probe kmsro since the
driver name will be panfrost/lima/v3d/..., not "kmsro". Since kmsro is a
bit of a catch-all for generic (mostly embedded) GPUs, add a fallback on
kmsro for the dynamic loader.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Acked-by: Karol Herbst <kherbst@redhat.com>

pipe-loader: Add kmsro pipe_loader target

kmsro is used by numerous embedded GPUs for a common winsys abstraction.
Let's add support for it for the dynamic pipe loader, so clover can
probe on these drivers.

We build the target with Panfrost. When other drivers need kmsro+clover,
we can revisit the build system part; my mesonfu is wanting.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Acked-by: Karol Herbst <kherbst@redhat.com>

scons: Fix force_scons parsing.

- Use parsed options instead of using ARGUMENTS directly.
- Handle the case of mingw cross compilation.

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Closes: https://gitlab.freedesktop.org/mesa/mesa/issues/2003

radv: enable secure compile support

Can be enabled via the environment variable which tells the
driver how many compilation threads are expected to be called,
and therefore how many forked processes the driver should
create.

For example we would expect to call fossilize replay with
something like this:

RADV_SECURE_COMPILE_THREADS=8 ./fossilize-replay --num-threads 8 \
--shader-cache-size 0 --ignore-derived-pipelines pipeline_cache.foz

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: a support for a secure compile fork at device creation

This added support for the fork, the installation of the seccomp
filter, and the main loop for the actual compilation to be called
from i.e. run_secure_compile_device().

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: add radv_secure_compile()

This function will be called by the parent process when doing a
secure compile. It first selects a free process to work with then
passes it all the information it needs to compile the pipeline.

Once the pipeline information has been passed to the secure
process, it then waits around to read/write any disk cache entries
required before exiting.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: for secure compile exit early from radv_shader_variant_create()

We don't have permission to be creating shared memory etc.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: allow the secure process to read and write from disk cache

This allows the secure process to read and write to the disk cache
via the parent process. This commit just adds the functionality
needed for the secure process, the following commit will add the
functionality for the parent process.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: add radv_device_use_secure_compile() helper

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: add some new members to radv device and instance for secure compile

These will be used by the following commits to hold information about
the forked secure compile processes.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: add radv_secure_compile_type enum

This will be used to identify information being passed between the
parent and secure process during a secure compile.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: add radv_create_shaders() to radv_shader.h

In a follwing commit we want to be able to call this for secure
compiles from radv_device.c

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: add debug option to turn off in memory cache

This can be usefull for debugging the on disk cache, but is also
useful in the following patch for secure compiles which will be
used to compile huge pipeline collections. These pipeline
collections can be multiple GBs and the in memory cache grows to
multiple GBs very quickly when they are compiled so we want to
be able to turn off the in memory cache.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

radv: get topology from pipeline key rather than VkGraphicsPipelineCreateInfo

This is cleaner and avoids having to read/write an additional copy of
topology for use with secure compile.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

docs: document new feature EGL_EXT_image_flush_external

egl: implement new functions from EGL_EXT_image_flush_external

Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Reviewed-By: Tapani Pälli <tapani.palli@intel.com>

egl: handle EGL_IMAGE_EXTERNAL_FLUSH_EXT

Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Reviewed-By: Tapani Pälli <tapani.palli@intel.com>

st/dri: add support for EGL_EXT_image_flush_external

Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Reviewed-By: Tapani Pälli <tapani.palli@intel.com>

st/dri: assume external consumers of back buffers can write to the buffers

Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Reviewed-By: Tapani Pälli <tapani.palli@intel.com>

dri_interface: add interface for EGL_EXT_image_flush_external

Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Reviewed-By: Tapani Pälli <tapani.palli@intel.com>

include: add the definition of EGL_EXT_image_flush_external

Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Reviewed-By: Tapani Pälli <tapani.palli@intel.com>

gitlab-ci: Add a job for meson on windows

This adds a new CI job that runs on windows with MSVC. It currently
builds softpipe and osmesa, and runs the related unit tests. It does
rely on meson's wraps for zlib, but I've set up caching of the wrap
dependencies so hopefully that wont be a problem.

I really wanted to user powershell for this, but there just isn't an
easy way to do that, it's much easier to use batch scripts, so thats
what I used.

The leading `/` for .gitlab-ci/lava... must be removed because windows
doesn't understand it, and when it reads the file the job ends in error.

Reviewed-by: Eric Engestrom <eric@engestrom.ch>

gitlab-ci: refactor out some common stuff for Windows and Linux

Reviewed-by: Eric Engestrom <eric@engestrom.ch>

nir: Fix invalid code for MSVC

Fixes: ee2050b1111b65594f3470035f7b6f1330824684
       ("nir: Use BITSET for tracking varyings in lower_io_arrays")

docs: update releasing process to use new scripts and gitlab

There were several out of date entries in this document, update them to
current practices.

Reviewed-by: Juan A. Suarez <jasuarez@igalia.com>

bin/gen_release_notes.py: Add a warning if new features are introduced in a point release

Fixes: 86079447da1e00d49db0cbff9a102eb4e71e8702
       ("scripts: Add a gen_release_notes.py script")
Reviewed-by: Juan A. Suarez <jasuarez@igalia.com>