pan/va: Add disassembler generator

When we bring up the Valhall compiler in Mesa, we will like to have a
disassembler in native code, so we shouldn't write our disassembler in
Python. Instead, we write a disassembler generator in Python with mako
templates, which will produce a va_disasm_instr entrypoint from the
architecture defined in ISA.xml and valhall.py.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12025>

pan/va: Check for FAU conflicts in the assembler

Logic described in the "Uniform/constant restrictions" section of the
Valhall specification. (You know, my Valhall specification. Is there
another one?)

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12025>

pan/va: Add Valhall assembler

This Python script acts as a standalone assembler. It takes Valhall
instructions with Mesa-flavour syntax, parses them, errors out if there
are syntax errors, and writes out an assembled binary if there are not.
It also is available as a programmatic interface for automated testing.

While this attempts to handle syntax errors, it does not check
semantics. It will happily compile programs that fault, provided each
instruction locally 'looks' plausible.

The code itself is quite small, despite supporting most of the known
ISA, because the syntax is regular and the heavylifting is done by
ISA.xml and valhall.py.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12025>

pan/va: Assert no instructions are duplicated

Caught a bug in the published PDF. Whoops!

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12025>

pan/va: Add ISA.xml parser and support code

This Python module parses the ISA.xml file added in the previous
commits, extracts all the useful information, and combines it with
extra annotations. In total, it provides a programmatic way to interface
with the Valhall instruction set.

   from valhall import instructions, enums

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12025>

pan/va: Add initial ISA.xml for Valhall

This handwritten file is the product of over a hundred hours of
reverse-engineering and represents the sum of what I've learned about
the Valhall architecture. It will be used in the next commits as the
backbone of a Valhall toolchain.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12025>

llvmpipe: correct the debug information printed with GALLIVM_PERF=nopt

GALLIVM_PERF_NO_OPT is "belond" to GALLIVM_PERF environment variable,
not GALLIVM_DEBUG. when GALLIVM_PERF=nopt is passed to llvmpipe,
"-sroa -early-cse -simplifycfg -reassociate -constprop -instcombine -gvn"
should not be printed. Those llvm optimation techniques should only be
printed when the default(-O2) optimation is enabled.

$ GALLIVM_PERF=nopt GALLIVM_DEBUG=dumpbc glmark2

Before apply this patch:

  Invoke as "opt -sroa -early-cse -simplifycfg -reassociate -mem2reg -constprop -instcombine -gvn ir_setup_variant_0.bc | llc -O2 [-mcpu=<-mcpu option>] [-mattr=<-mattr option(s)>]"

After apply this patch:

  Invoke as "opt -mem2reg ir_fs304_variant0.bc | llc -O0 [-mcpu=<-mcpu option>] [-mattr=<-mattr option(s)>]"

Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Signed-off-by: suijingfeng <suijingfeng@loongson.cn>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11994>

gallium: Reset {d,r}Priv in dri_unbind_context

The code in dri_make_current just checks the value of the pointers
to decide to update texture_stamp or not. This is buggy since a new
allocated drawable could share the same address with the previous
released drawable. Fix the stale pointer issue by always resetting
these pointers to NULL in dri_unbind_context.

v2:
   Move the reset codes to the end of the function.

Signed-off-by: Lepton Wu <lepton@chromium.org>
Cc: mesa-stable
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12050>

anv/android: Rework our handling of AHardwareBuffer imports

The current code we have for this is a bit of a mess, likely due to
trying too hard to put it in anv_android.c.  The external_format bit in
anv_image, for instance, really means "quit creation early" which is
something we want to do for AHardwareBuffer imports regardless of
whether or not they use a native format.  It gets set both by declaring
an AHardwareBuffer external handle type and by VkExternalFormatANDROID.
However, VkExternalFormatANDROID is only allowed for AHardwareBuffer
imports.  If we ever did get an external format outside the context of
an AHardwareBuffer import, we would end up with a useless partially
created image.

When we detect an AHardwareBuffer import, we punt off to a function in
anv_android.c that does nothing interesting but call anv_create_image
with AUX disabled and external_format = true.  The aux disable here is
useless because the actual isl_surf layout is done by resolve_ahw_image
which also sets ISL_SURF_USAGE_DISABLE_AUX_BIT.  As far as external
formats go, anv_image_from_external() sets it regardless of whether or
not there is actually an external format.

This commit replaces anv_image::external_format with anv_image::from_ahb
which is the thing we actually want to track for this.  We delete
anv_image_from_external and a bunch of the external_format handling
because it's all useless.  The end result is massively simpler and,
while it appears to blur the boundary between Android code and the rest
of the driver, it makes the whole flow more obvious.

Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12040>

tu: Triage some CTS failures

These should be fixed by the linked VK-GL-CTS change. I couldn't
reproduce the other renderpass failure on a650 though.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12065>

panvk: Implement vkCmdClearAttachments

For now it only supports one RT and color attachments.

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/12019>

panvk: Expose panvk_cmd_alloc_fb_desc and panvk_cmd_alloc_tls_desc

For use in panvk_meta.c.

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/12019>

panvk: A pipeline might not be bound when the render pass is ended

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/12019>

panvk: Move check for fragment requirement up to the draw

As the panvk_cmd_alloc_fb_desc function might be called outside a draw,
without a pipeline having been bound to the command buffer.

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/12019>

panvk: Close batch when ending a command buffer

In some cases, there will be an open batch because an event operation
has been recorded, or a non-rendering command.

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/12019>

panvk: Start a new batch when the job index gets above the limit

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/12019>

panvk: Don't try to update samplers if they are immutable

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/12019>

ci/v3dv: update vulkan expected results

Add a couple of flakes.

Reviewed-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12068>

glsl: Properly handle .length() of an unsized array

There are two distinct cases:
- The last member of a shader storage block (length determined at run-time)
- Implicitly-sized array (length determined at link-time)

Fixes: 273f61a0051a ("glsl: Add parser/compiler support for unsized array's length()")
Signed-off-by: Yevhenii Kolesnikov <yevhenii.kolesnikov@globallogic.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11952>

glsl: Add operator for .length() method on implicitly-sized arrays

ARB_shader_storage_buffer_object extension (promoted to core in 4.3) allows us
to call .length() method on arrays declared without an explicit size. The length is
determined at link time as a maximum array access.

Fixes: 273f61a0051a ("glsl: Add parser/compiler support for unsized array's length()")
Signed-off-by: Yevhenii Kolesnikov <yevhenii.kolesnikov@globallogic.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11952>

lima: avoid crash with negative viewport values

The viewport value computations done in lima_set_viewport_states
can result in a negative value for viewport.
These could end up converted to unsigned values in
lima_clip_scissor_to_viewport causing crashes from invalid
scissor commands.
Prevent this by limiting the minimum value to zero as is already
done for the left and bottom values.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/2938
Cc: mesa-stable
Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Vasily Khoruzhick <anarsoul@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12055>

dlist: don't handle unmerged draws as merged

The comment was incorrect: we can have N draws using the
same mode with N > 1 (eg: GL_QUAD_STRIP draws
cannot be merged).

This commit fixes the drawing code to use the correct draw
function.

This fixes a hang in Starsector (see issue #5086).

Fixes: b328d8e9bc9 ("dlist: use an union instead of allocating a 1-sized array")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11978>

v3dv: expose VK_KHR_multiview

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12034>

v3dv: implement interaction of queries with multiview

When multiview is enabled, queries must begin and end in the
same subpass and N consecutive queries are implicitly used,
where N is the number of views enabled in the subpass.
Implementations decide how results are split across queries.

In our case, only one query is really used, but we still need
to flag all N queries as available by the time we flag the one
we use so that the application doesn't receive unexpected errors
when trying to retrieve values from them.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12034>

v3dv: fix query error handling

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12034>

v3dv: track first and last subpass that use a view index

When multiview is enabled, we no longer care about when a particular
attachment is first or last used in a render pass, since not all views
in the attachment will meet that criteria. Instead, we need to track
each individual view (layer) in each attachment and emit our stores,
loads and clears accordingly.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12034>

v3dv: skip processing tiles for layers that are not in the view mask

If a multiview subpass doesn't use a particular layer then we can ignore
that layer completely.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12034>

v3dv: use correct number of layers for multiview

The Vulkan spec states that when multiview is enabled the number of
layers in the framebuffer is set to one and that each attachment
must then have at least as many layers as referenced by view masks
in the subpasses in which is used.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12034>

v3dv: don't merge subpasses with different view masks

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12034>

v3dv: broadcast multiview draw commands

We implement multiview by replicating draw commands for all enabled
views and setting a command buffer state for the currently active
view we are broadcasting to.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12034>

broadcom/compiler: implement nir_intrinsic_load_view_index

This is used for multiview's gl_ViewIndex built-in.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12034>

v3dv: inject a custom passthrough geometry shader for multiview pipelines

This allows us to use layered rendering to broadcast draw calls to the
appropriate views (layers).

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12034>

v3dv: move all our NIR pre-processing to preprocess_nir

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12034>

v3dv: store multiview info in our render pass data

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12034>

v3dv: drop unused parameters

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12034>

radv: implement VK_EXT_shader_atomic_float2

Some floating atomic instructions are not available on GFX8-9.
No LLVM support.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12060>

aco: implement VK_EXT_shader_atomic_float2

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12060>

venus: log more WSI messages

They can be useful before resorting to gdb or perfetto.

Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12073>

turnip: Add support for VK_VALVE_mutable_descriptor_type

v1.  Hyunjun Ko <zzoon@igalia.com>
- Add to hanlde VK_DESCRIPTOR_POOL_CREATE_HOST_ONLY_BIT_VALVE
- Don't support VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT and
VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER

v2.  Hyunjun Ko <zzoon@igalia.com>
- Fix some indentations and nitpicks.
- Add the extension to features.txt

v3.  Hyunjun Ko <zzoon@igalia.com>
- Remove unnecessary asserts.

Signed-off-by: Eduardo Lima Mitev <elima@igalia.com>
Signed-off-by: Hyunjun Ko <zzoon@igalia.com>
Reviewed-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9322>

venus: suballocate memory in more cases

When a dedicated allocation is not required, ignore it and suballocate.

Fixes dEQP-VK.api.invariance.random.

Cc: 21.2 mesa-stable
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12072>

venus: clean up vn_AllocateMemory

Mainly to add vn_device_memory_should_suballocate.

Cc: 21.2 mesa-stable
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12072>

llvmpipe: fix nir dot products (fsum op)

When the dot product uses a source which can be optimized to a scalar,
after a bunch of nir optimization steps the source to fsum will be a scalar
with a x replicate swizzle. Hence nir_src_num_components is just 1 and the
fsum was just a no-op which is not correct. Arguably this could be optimized
a bit better, but just determine the number of addends by using nir_op_infos
instead (the operand fetch was fixed already by 39a938ecf41b doing the same).

Fixes: 4eb0475b5a00 ("gallivm/nir: add fsum support")
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12066>

zink: EXT_vertex_input_dynamic_state

this eliminates vertex attributes from the pipeline state, massively
deduplicating the number of pipelines needed

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12009>

zink: improve oom flushing

flush on 30k works and also check oom_flush

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12010>

zink: merge draw_count and compute_count, move to batch struct

one fewer deref and now it's a single counter

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12010>

zink: more explicitly check shader stages during compile

this is a bit more obvious

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12013>

zink: remove no longer used internal resource function

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12008>

zink: replace some direct batch_usage calls with resource abstractions

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12008>

zink: use new resource batch usage utils for is_resource_busy

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12008>

zink: simplify some dumb code in invalidate_buffer

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12008>

zink: use resource batch usage helpers in invalidate_buffer()

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12008>

zink: collapse a conditional in zink_batch_resource_usage_set()

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12008>

zink: add some resource util functions for batch usage

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12008>

zink: force batch completion check on query result

non-timeline drivers have no screen-based method of "checking" batch
completion, so the context method has to be used here to avoid an infinite
loop

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12008>

zink: add screen function for checking usage completion

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12008>

freedreno/a6xx: Add missing PC_CCU_INVALIDATE_x

The kernel normally inserts these between submits, but when we merge
submits in userspace we need to add them ourselves.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5114
Fixes: cccdc513e3e ("freedreno/drm/sp: Implement deferred submit merging")
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12049>

zink: use array size in spirv bo length calculations

I don't know why I was dropping this, but doing so breaks drivers that have
optimization passes based on the lengths of these variables

Fixes: c1cdf30a119 ("zink: apply Delete All The Code methodology to the ubo/ssbo variables")

Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12070>

kmsro: Add 'kirin' driver support

Kirin is used by hisilicon SOCs (Hikey boards, etc.)
Driver is available in the mainline kernel [1]

[1]: https://elixir.bootlin.com/linux/v5.13/source/drivers/gpu/drm/hisilicon/kirin/kirin_drm_ade.c#L928

Signed-off-by: Roman Stratiienko <r.stratiienko@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11755>

crocus/gen4: restrict memcpy mapping to gen5

This is due to gen4 + 4.5 having some rather strange swizzling
that we can't actually detect properly in userspace

Fixes: f3630548f1da ("crocus: initial gallium driver for Intel gfx 4-7")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12058>

frontends/va: change to per-layer rate control

Allows for each layer in a Scalable Video Coding (SVC) video to have its
own rate control setting, when encoding H.264.

Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11850>

frontends/omx: change rate ctrl struct to array

Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11850>

radeon/vcn/enc: change to per-temporal layer rate control

Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11850>

radeon/vce: change rate ctrl struct to array

Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11850>

r600: change rate ctrl struct to array

Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11850>

gallium: change rate ctrl struct to array

Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11850>

frontends/va: handle h264 num_temporal_layers for SVC encoding

Allows for the number of temporal layers to be specified when encoding
Scalable Video Coding (SVC) H.264 videos.

Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11850>

radeonsi: enable H.264 temporal encoding support for VCN

Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11850>

radeon/vcn/enc: H.264 SVC encode

Implement H.264 temporal, Scalable Video Coding (SVC) for VCN devices by
sending the required parameters to the firmware, and creating H.264 NALU
prefix and SEI scalability_info headers.

Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11850>

gallium: update h264 struct to track temporal layers

Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11850>

frontends/va: check number of temporal layers supported by encoder

Checks the encoder to determine the number of temporal layers supported,
and returns max_num_temporal_layers_minus1, along with setting the
corresponding control flag if multiple layers are supported.

Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11850>

gallium: add temporal layers cap enum

Determine the number of temporal layers the encoder supports. Used for
encoding Scalable Video Coding (SVC) videos.

Signed-off-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11850>

tu: handle half-reg fs outputs

This would allow to enable translation of RelaxedPrecision spirv
variable decorator into mediump which for us means fp16.

Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12067>

nv30: fix emulated vertex index buffers

We ended up applying the offset twice. Quite embarrassing.

This fixes a bunch of vertex shader related issues like the gnome desktop
is less broken and probably a lot of other applications.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5061
Cc: mesa-stable
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12054>

radv: allow unused VkSpecializationMapEntries

Fixes future CTS: dEQP-VK.pipeline.spec_constant.*.basic.*unused_*

Cc: 21.2 mesa-stable
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/12062>

nir/shrink_vectors: shrink vecN properly

This patch allows to shrink vecN instructions where
one or more components at any position are unused.

Stat changes for softpipe:
total instructions in shared programs: 2986101 -> 2985416 (-0.02%)
instructions in affected programs: 51216 -> 50531 (-1.34%)

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11411>

nir/shrink_vectors: shrink ALU properly

ALU instructions of which not all components are read,
can be shrunk to the number of read components.
Previously, this would only remove trailing components.

This patch enables to remove components from any position.

Stat changes for softpipe:
total instructions in shared programs: 3001291 -> 2984698 (-0.55%)
instructions in affected programs: 225585 -> 208992 (-7.36%)
total loops in shared programs: 1389 -> 1358 (-2.23%)
loops in affected programs: 36 -> 5 (-86.11%)

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11411>

nir/opt_shrink_vectors: reverse iteration order

This pass should be backwards in order to reach the fixed point
in linear time.

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11411>

nir: consider write_mask in nir_ssa_def_components_read()

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11411>

nir/opt_shrink_vectors: don't shrink vectors used by intrinsics

Store intrinsics shrink the sources by creating a new vecN.
Other intrinsics cannot shrink their sources.

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11411>

nir/lower_alu_to_scalar: don't skip gaps in write_mask

Otherwise, this may lead to segmentation faults.

Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11411>

v3dv: assert job->cmd_buffer is valid

In v3dv_write_uniforms_wg_offsets() function, the job's cmd_buffer is a
valid command buffer, so there is no reason to check if its NULL or not.

This fixes CID#1487441 ("Dereference after null check") error.

v1:
 - `job->cmd_buffer` is the same as `cmd_buffer` (Alejandro)

v2:
 - Use `cmd_buffer` instead of `job->cmd_buffer` (Iago)

Fixes: 31a786c80ad ("v3dv: handle QUNIFORM_FB_LAYERS")
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11999>

isl: fix mapping of format->stringname

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: ed6e586562f444 ("intel: properly constify isl_format_layouts")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5110
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12051>

vulkan/overlay: don't display histogram and range for device and format

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4320
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12056>

teximage: return correct desktop GL error for compressedteximage

For GL4.6 the spec says to return GL_INVALID_OPERATION for this.

Cc: mesa-stable
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12032>

asahi: Generalize src_offset for non-4byte formats

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12053>

asahi: Add integers to agx_vertex_formats

Handles all the easy cases.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12053>

agx: Shift vertex buffer stride in the compiler

Required to support non-32-bit vertex formats efficiently.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12053>

agx: Add agx_format_shift routine

Required to calculate alignments for vertex buffers correctly.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12053>

asahi: Pass instance_divisor to the compiler

Passes dEQP-GLES3.functional.instanced.*

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12053>

agx: Define p_extract for type converts

Useful for grabbing the high 32-bit word of a 64-bit value.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12053>

agx: Implement instanced arrays

Divide by instance divisor if needed. Same strategy as ACO.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12053>

agx: Include divisors in the vertex shader key

Needed to lower the divisions.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12053>

agx: Add udiv-by-constant routine

Uses the ridiculousfish algorithm, will be used to lower instanced
arrays into something efficient.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12053>

agx: Add agx_ushr helper

Syntax sugar for the underlying bitfield manipulation instruction.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12053>

agx: Handle load_instance_id

Preloaded into r6, as predicted.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12053>

agx: Drop dated /* TODO: RA */

We skip over vertex ID in RA now, it's fine.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12053>

asahi: Enable instancing

Passes dEQP-GLES3.functional.draw.draw_arrays_instanced.lines.single_attribute

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12053>

agx: Plug memory leak in register allocator

Fixes: 85e18deb18a ("agx: Assign registers locally")
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reported-by: Coverity
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12053>

agx: Use consistent ncomps

Fixes register allocation failure in:
dEQP-GLES3.functional.ubo.single_basic_array.shared.row_major_mat4_fragment

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12053>

agx: Dump register file when failing to allocate

Usually shows a bug.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12053>