gitlab-ci: Check the Mesa version used for tracie tests

Verify that the Mesa version used when running tracie tests is the
one that was built by CI, rather than any installed distro version.

Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Reviewed-by: Rohan Garg <rohan.garg@collabora.com>
Reviewed-by: Andres Gomez <agomez@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3894>

aco: clear moved operands in get_reg_create_vector()

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4507>

aco: improve p_create_vector RA for sub-dword operands

These's still improvements needed for sub-dword definitions, but that's
not as simple.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4507>

aco: fix p_extract_vector validation

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4507>

aco: improve vector optimization with sub-dword vectors

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4507>

radv: use RMW packets for updating the maximum sample distance

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/4531>

radv: add radeon_set_context_reg_rmw() helper

For emitting RMW packets in the command stream. This new helper
will be useful for implementing extended dynamic states to only
overwrite the fields that need to be updated instead of storing
more values in the pipeline.

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/4531>

scons/windows: Support build with LLVM 10.

LLVM engine component contains 2 new libraries, LLVMCFGuard and LLVMTextAPI.
Without them build fails like in this run: https://ci.appveyor.com/project/Alexpux/mingw-packages/builds/32102425/job/wkmb4gimfqkkb3cg

Cc: <mesa-stable@lists.freedesktop.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4521>

meson: Link Gallium Nine with ld_args_build_id

This fixes an assertion in iris_disk_cache_init() when the initialization
goes through drm_create_adapter(), which lives in d3dadapter9.so.
In this case build_id_find_nhdr_for_addr() fails and returns NULL, since
the shared library does not include a build ID.

The issue can be reproduced with an iris capable GPU and Xnine, while
removing the shader cache prior to launching the application.

Fix this by doing the same as in 29ea92e6a1e8f5cb3295011d907ea211d6f8f644.

Fixes: 4756864cdc5f "iris: Start wiring up on-disk shader cache"
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4499>

svga: fix build on FreeBSD

MADV_HUGEPAGE only exists on Linux

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3966>

freedreno/a5xx+: Skip compiling the old gmem blit programs.

Saves a bunch of noise for me to sort through in IR3_SHADER_DEBUG=vs,fs
shader-db/run <single shader_test>.  I found that we were crashing on
destroy of NULL programs in fd_prog_fini, so I replicated the gpu_id < 300
early exit from fd_prog_init() down to _fini as well.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4538>

pan/bit: Add BI_CONVERT tests

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4539>

pan/bit: Add BI_CONVERT interpretation

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4539>

pan/bi: ADD packing for CONVERT

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4539>

pan/bi: Rewrite conversion packing

To support roundmodes and other goodies.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4539>

pan/bi: Fix incorrect swizzle packing assert

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4539>

pan/bi: Set BI_ROUNDMODE for BI_CONVERT

It's rarely used in GL but converts do have roundmodes.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4539>

pan/midgard: Fix f2u naming confusion

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4539>

meson: Specify the maximum required libdrm in dri.pc

When dealing with a regression in libdrm-2.4.101, I masked the package
in Gentoo. In doing so, we discovered that Mesa's dri.pc specifies a
version requirement in dri.pc for >= the version of libdrm Mesa was
built against, thus preventing packages from being rebuilt with the
older version of libdrm installed.

Let's reduce this version requirement to the latest libdrm required by
Mesa instead, since libdrm is backward compatible.

Fixes: a3a16d4aa7e ("meson: use dep_libdrm version for pkg-config")
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4534>

freedreno/ir3/ra: cleanup some leftovers

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4440>

freedreno/ir3: rename depth->dce

Since DCE is the only remaining function of this pass, after the pre-RA
scheduler rewrite.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4440>

freedreno/ir3: better cleanup when removing unused instructions

Do a better job of pruning when removing unused instructions, including
cleaning up dangling false-deps.

This introduces a new ssa src pointer iterator, which makes it easy to
clear links without having to think about whether it is a normal ssa
src or a false-dep.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4440>

freedreno/ir3/ra: handle array case for SFU select_reg opt

The src of the SFU instruction could also be array/reg (non-SSA).
Handle this case too.

The postsched cp pass makes this scenario more likely.

Fixes: cc82521de4e ("freedreno/ir3: round-robin RA")
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4440>

freedreno/ir3: add mov/cov stats

While not always avoidable, cov instructions are a useful thing to look
at to see if we could fold into src.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4440>

freedreno/ir3/postsched: avoid moving tex ahead of kill

Add extra dependencies of tex/mem instructions on previous kill
instructions to avoid moving them ahead of kills.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4440>

freedreno/ir3/postsched: remove some leftovers

These aren't used in postsched.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4440>

freedreno/ir3/sched: awareness of partial liveness

Realize that certain instructions make a vecN live, and account for
this, in hopes of scheduling the remaining components of the vecN
sooner.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4440>

freedreno/ir3: new pre-RA scheduler

This replaces the depth-first search scheduler with a more traditional
ready-list scheduler.  It primarily tries to reduce register pressure
(number of live values), with the exception of trying to schedule kills
as early as possible.  (Earlier iterations of this scheduler had a
tendency to push kills later, and in particular moving texture fetches
which may not be necessary ahead of kills.)

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4440>

freedreno/ir3: fix location of inserted mov's

If the group pass must insert a mov to resolve conflicts, avoid the mov
appearing *after* the meta:collect whose src it is.

The current pre-RA scheduler doesn't really care about the initial
instruction order, but the new one will in some cases.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4440>

freedreno/ir3: simplify grouping pass

Since bdf6b7018cedf95b554e21953d5a1935d3067ce7 the logic only needs to
handle grouping collect srcs, So remove the now unnecessary indirection.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4440>

freedreno/ir3: make falsedep use's optional

Add option when collecting uses to control whether they include
falsedeps or not.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4440>

freedreno/ir3: spiff out disasm a bit

for verbose mode, print also the instruction "cycle" (which takes into
account (rptN) and (nopN)) in addition to instruction offset.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4440>

freedreno/computerator: support bindless sampler instructions

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4526>

freedreno/computerator: support nop prefix

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4526>

freedreno/ir3: CSE the up/downconversion of SEL's cond's size.

Not many programs hit this, but if you were, say, selecting between vec4s,
you'd convert the cond 4 times.

instructions in affected programs: 2957 -> 2717 (-8.12%)
nops in affected programs: 989 -> 899 (-9.10%)
non-nops in affected programs: 1968 -> 1818 (-7.62%)
dwords in affected programs: 3232 -> 2752 (-14.85%)
last-baryf in affected programs: 102 -> 90 (-11.76%)
full in affected programs: 5 -> 4 (-20.00%)
sstall in affected programs: 329 -> 329 (0.00%)
(ss) in affected programs: 86 -> 105 (22.09%)
(sy) in affected programs: 14 -> 12 (-14.29%)

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4516>

freedreno/ir3: Stop doing b2n on the SEL condition.

SEL_B32 (and presumably B16) checks for 0 or nonzero in the condition
(tested by just stuffing a uniform's value into it), so there's no need to
do ir3_b2n() on it, or any preceding ir3_n2b().

instructions in affected programs: 664444 -> 659927 (-0.68%)
nops in affected programs: 267898 -> 266312 (-0.59%)
non-nops in affected programs: 420260 -> 417329 (-0.70%)
dwords in affected programs: 144032 -> 137568 (-4.49%)
last-baryf in affected programs: 10801 -> 10321 (-4.44%)
full in affected programs: 2003 -> 2002 (-0.05%)
sstall in affected programs: 76670 -> 77405 (0.96%)
(ss) in affected programs: 4515 -> 4525 (0.22%)
(sy) in affected programs: 612 -> 604 (-1.31%)

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4516>

tnl: Code formatting in t_rebase.c

Many little changes.  Almost everything is indentation or removal of
trailing whitespace.  Some places move a loop variable declaration to
the loop.  Some comments either re-wrapped or converted to single
line.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4512>

tnl: Code formatting in t_draw.c

So many little changes.  Almost everything is indentation or removal of
trailing whitespace.  There's a couple places where an "else" is moved
to the previous line with the "}".  Some places move a loop variable
declaration to the loop.  One change of assert to unreachable.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4512>

tnl: Silence unused parameter warnings in _tnl_split_inplace

Unused since db0eb3a437f ("vbo: Fix up in-place splitting for non-contiguous/indexed primitives.") which landed in 2010.

src/mesa/tnl/t_split_inplace.c: In function ‘_tnl_split_inplace’:
src/mesa/tnl/t_split_inplace.c:270:27: warning: unused parameter ‘min_index’ [-Wunused-parameter]
  270 |                    GLuint min_index,
      |                    ~~~~~~~^~~~~~~~~
src/mesa/tnl/t_split_inplace.c:271:27: warning: unused parameter ‘max_index’ [-Wunused-parameter]
  271 |                    GLuint max_index,
      |                    ~~~~~~~^~~~~~~~~

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4512>

tnl: Silence unused parameter warnings in dump_draw_info

src/mesa/tnl/t_split_copy.c: In function ‘dump_draw_info’:
src/mesa/tnl/t_split_copy.c:149:35: warning: unused parameter ‘ctx’ [-Wunused-parameter]
  149 | dump_draw_info(struct gl_context *ctx,
      |                ~~~~~~~~~~~~~~~~~~~^~~
src/mesa/tnl/t_split_copy.c:154:23: warning: unused parameter ‘min_index’ [-Wunused-parameter]
  154 |                GLuint min_index,
      |                ~~~~~~~^~~~~~~~~
src/mesa/tnl/t_split_copy.c:155:23: warning: unused parameter ‘max_index’ [-Wunused-parameter]
  155 |                GLuint max_index)
      |                ~~~~~~~^~~~~~~~~

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4512>

tnl: Silence unused parameter warnings in _tnl_draw_prims

A tangled mess of a couple parameters that nobody wanted.

src/mesa/tnl/t_draw.c: In function ‘_tnl_draw_prims’:
src/mesa/tnl/t_draw.c:440:42: warning: unused parameter ‘tfb_vertcount’ [-Wunused-parameter]
  440 |     struct gl_transform_feedback_object *tfb_vertcount,
      |     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~
src/mesa/tnl/t_draw.c:441:35: warning: unused parameter ‘stream’ [-Wunused-parameter]
  441 |                          unsigned stream)
      |                          ~~~~~~~~~^~~~~~

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4512>

tnl: Silence unused parameter 'attrib' warning in convert_half_to_float

Also clean up some trivial code / whitespace style issues.

src/mesa/tnl/t_draw.c: In function ‘convert_half_to_float’:
src/mesa/tnl/t_draw.c:121:57: warning: unused parameter ‘attrib’ [-Wunused-parameter]
  121 |                       const struct gl_array_attributes *attrib,
      |                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4512>

tnl: Don't dereference NULL obj pointer in t_rebase_prims

Structurally the code is now similar to the handling of other
gl_buffer_object::obj pointers elsewhere in TNL.  The fixes tag is a
little bit misleading.  I think the change in that commit just exposes a
previously existing bug.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/2746
Fixes: f3cce7087a5 ("mesa: don't ever bind NullBufferObj for glBindBuffer targets")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4512>

tnl: Don't dereference NULL obj pointer in replay_init

Structurally the code is now similar to the handling of other
gl_buffer_object::obj pointers elsewhere in TNL.  The fixes tag is a
little bit misleading.  I think the change in that commit just exposes a
previously existing bug.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/2746
Fixes: f3cce7087a5 ("mesa: don't ever bind NullBufferObj for glBindBuffer targets")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4512>

tnl: Don't dereference NULL obj pointer in bind_indices

Structurally the code is now similar to bind_inputs.  The fixes tag is a
little bit misleading.  I think the change in that commit just exposes a
previously existing bug.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/2746
Fixes: f3cce7087a5 ("mesa: don't ever bind NullBufferObj for glBindBuffer targets")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4512>

aco: fix p_extract_vector optimization in presence of unequally sized vector operands

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4506>

pan/bi: Lower fsqrt

For G72+ anyway. G71 will want something a bit more fine grained. I hope
this has enough precision for GL (the blob apparently does some exponent
fixup).

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4530>

panfrost: Drop dependency on nonexistant write_value

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4530>

glsl: stop processing function parameters if error happened

Fixes: d1fa69ed61d ("glsl: do not attempt assignment if operand type not parsed correctly")
Closes: https://gitlab.freedesktop.org/mesa/mesa/issues/2696
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4341>

aco: fix nir_op_pack_32_2x16_split if one operand is a constant

Because 16-bit constants are represented with the s1 RegClass, we
have to extract the low half.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4509>

aco: implement 16-bit nir_op_f2i64/nir_op_f2u64

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4509>

aco: fix f2i64/f2u64 with sgprs if the exponent computation overflow

This fixes f16->{i64,u64} conversions for +0/-0.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4509>

gitlab-ci: Use all_paths in .test-manual rules

Without this, the .test-manual jobs could end up as 'when: manual' when
the jobs they depend on were 'when: never', which was flagged as
invalid YAML, preventing the pipeline from being created.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4522>

nvc0: add NV_viewport_swizzle support for GM200+

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4519>

st/mesa: add NV_viewport_swizzle support

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4519>

mesa: add GL_NV_viewport_swizzle support

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4519>

gallium: add viewport swizzling state and cap

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4519>

glsl: remove redudant assignment

CID: 1461087
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Andres Gomez <agomez@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4500>

mesa: remove redudant assignment

CID: 1461397
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Andres Gomez <agomez@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4500>

mesa: remove redudant check

CID: 1461410
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Andres Gomez <agomez@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4500>

lima: set offset when export resource

We missed set reource offset.

Reviewed-by: Vasily Khoruzhick <anarsoul@gmail.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4517>

i965: share buffer managers across screens

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: <mesa-stable@lists.freedesktop.org>
Closes: https://gitlab.freedesktop.org/mesa/mesa/issues/1373
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4086>

i965: store DRM fd on intel_screen

v2: Fix storing of drm fd (Ajax)

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: <mesa-stable@lists.freedesktop.org>
Closes: https://gitlab.freedesktop.org/mesa/mesa/issues/1373
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4086>

iris: make resources take a ref on the screen object

Because St creates resources from a screen and attach them onto
another we need to ensure the resources associated to a screen &
bufmgr stay around until we don't need them anymore.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: <mesa-stable@lists.freedesktop.org>
Closes: https://gitlab.freedesktop.org/mesa/mesa/issues/1373
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4086>

iris: share buffer managers accross screens

St happilly uses pipe_resources created with one screen with other
screens. Unfortunately our resources have a single identifier that
related to a given screen and its associated DRM file descriptor.

To workaround this, let's share the buffer manager between screens for
a given DRM device. That way handles are always valid.

v2: Don't forget to close the fd that bufmgr now owns
    Take a copy of the fd to ensure it stays alive even if the dri
    layer closes it

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: <mesa-stable@lists.freedesktop.org>
Closes: https://gitlab.freedesktop.org/mesa/mesa/issues/1373
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4086>

iris: properly free resources on BO allocation failure

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Rafael Antognolli <rafael.antognolli@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4086>

freedreno/log: better decoding for multiple chunks per batch

For larger render targets or smaller GMEM size, we could end up needing
multiple chunks of tracepoints per batch.  But we still want to decode
the traces as single batch.  So we need a bit of a state across
process_chunk() calls to accumulate timestamp information.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4510>

freedreno/log: spiff out parser some more

Extract breakdown of time in GMEM passes.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4510>

freedreno/log: android support

In particular, when stdout doesn't go anywhere useful we need to log to
file.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4510>

freedreno: Fix leak of binning shader variants.

The v->binning variant is never added to shader->variants, so just free
each one as we free the nonbinning variant.

Noticed from drm-shim mode running out of open fds, since each bo ends up
with an fd.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4502>

freedreno/ir3: Fix sz vs class confusion

Add bounds checking to make sure we don't silently access out of
bounds again.

Fixes: 90f7d12236c ("freedreno/ir3/ra: pick higher numbered scalars in first pass")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4503>

pan/decode: Print Bifrost blend descriptor

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

pan/bi: Let !b2b imply branch_cond

Like the blob. Probably doesn't matter.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

panfrost: Fix BI_BLEND packing

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

pan/bi: Fix backwards registers ports

Will matter when packing multiple instructions per bundle.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

panfrost: Pass compiler-appropriate options

FMAs need to fuse for Bifrost.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

panfrost: Move uniform_count to pan_assemble

Again, not Midgard specific.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

panfrost: Move varying linking to cmdstream

This isn't ISA/compiler specific, it's just looking at the NIR. So let's
move it from midgard to pan_assemble.c so it runs for Bifrost too.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

pan/midgard: Remove unused max_varying variable

I don't know why this was here.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

pan/bi: Fix nondeterministic register packing

Uninitialized read.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

panfrost: Call the Bifrost compiler on bi devices

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

panfrost: Set mfbd.msaa.sample_locations on Bifrost

And mfbd.shared_memory only on Midgard.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

panfrost: On Bifrost, set the right tiler descriptor

On both fragment and tiler jobs.

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

panfrost: Don't emit write_value jobs on Bifrost

As on Bifrost GPUs there's a different mechanism for reusing the tiler
data structures.

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

panfrost: Pass IS_BIFROST to pandecode_jc

So we can decode the right structures on Bifrost hw.

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

panfrost: Remove most usage of midgard_payload_vertex_tiler

By passing the prefix and postfix structs around, we can use most of the
cmdstream functions as well for bifrost, as those structs haven't
changed between midgard and bifrost.

Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

panfrost: Unify vertex/tiler structures

Some fields were shuffled but these are essentially the same across the
generations.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

panfrost: Staticize a few cmdstream functions

They are only used within the same source file.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

panfrost: Populate bifrost-specific structs within mali_shader_meta

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

panfrost: Add IS_BIFROST quirk

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Signed-off-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4505>

etnaviv: remove the "active" member of queries

The state tracker only gets to begin/query/destroy when !active and end
when active, so we have no need to try to track this ourselves.

Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4456>

etnaviv: change begin_query(..) to a void function

We always return true.

Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4456>

etnaviv: drop redundant calls to etna_acc_query_suspend(..)

Introduced by accident during rebase.

Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4456>

v3d: Primitive Counts Feedback needs an extra 32-bit padding.

Store Primitive Counts operations write 7 counters in 32-bit words
but also a padding 32-bit with 0. So we need 8 32-bit words instead
of the current 7 allocated.

This was causing an corruption in the next buffer when Transform
Feedback was enabled that were exposed on tests like:
dEQP-GLES3.functional.transform_feedback.*.points.*

This patch fixes 196 tests that were failing when they were run isolated
but they were passing when run using cts-runner.

Fixes: 0f2d1dfe65bf ("v3d: use the GPU to record primitives written to transform feedback")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/2674
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4501>

aco: make some reg_file helpers private and fix their uses

Fixes various subdword RA issues

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4492>

aco: rename aco_lower_bool_phis() -> aco_lower_phis()

We also lower subdword phis, now.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4492>

aco: lower subdword phis with SGPR operands

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4492>

aco: don't constant-propagate into subdword PSEUDO instructions

PSEUDO instructions are lowered using SDWA, and thus,
cannot take literals and before GFX9 cannot take constants
at all. As the in-register representation differs between
32bit and 16bit floats, we first need to ensure correct
behavior.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4492>

aco: ensure correct bit representation of subdword constants

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4492>

aco: setup subdword regclasses for ssa_undef & load_const

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4492>