mesa.git
14 months agoradv,aco: disable opts if VK_PIPELINE_CREATE_DISABLE_OPTIMIZATION_BIT
Samuel Pitoiset [Fri, 28 Aug 2020 06:41:20 +0000 (08:41 +0200)]
radv,aco: disable opts if VK_PIPELINE_CREATE_DISABLE_OPTIMIZATION_BIT

Sounds useful to determine if ACO breaks a specific pipeline
because of various optimizations.

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/6487>

14 months agopanfrost: Fix gnu-empty-initializer errors.
Vinson Lee [Tue, 1 Sep 2020 23:21:47 +0000 (16:21 -0700)]
panfrost: Fix gnu-empty-initializer errors.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3473
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6551>

14 months agoiris: Fix aux assertion in resource_get_handle
Nanley Chery [Wed, 17 Jul 2019 20:19:43 +0000 (13:19 -0700)]
iris: Fix aux assertion in resource_get_handle

iris_resource_get_handle currently asserts that the resource has an aux
state that is suitable for sharing. However, the caller of this function
can pass a flag to specify that it will handle flushing/resolving the
resource as needed for sharing. Take this flag into account when
asserting the state of the aux buffer.

Fixes: e81392868e6 ("iris/resource: Drop redundant checks for aux support")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/128
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/1371>

14 months agoci/bare-metal: Use python for handling fastboot booting and parsing
Eric Anholt [Mon, 31 Aug 2020 20:00:14 +0000 (13:00 -0700)]
ci/bare-metal: Use python for handling fastboot booting and parsing

Modeling after what I did for cros_servo_run.py, this gives us easy
support for restarting the test run a530 when we detect a spontaneous
reboot.  I had to touch up serial_buffer.py to handle buffering in from a
file instead of a serial device, to support the upcoming etnaviv CI
(tested by running it against a serial log from db410c and seeing it step
to calling "fastboot")

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6529>

14 months agoci/bare-metal: Fix capturing of serial output as job artifacts.
Eric Anholt [Mon, 31 Aug 2020 20:04:28 +0000 (13:04 -0700)]
ci/bare-metal: Fix capturing of serial output as job artifacts.

I tried to put them in the wrong directory -- everything needs to go in
results/, which we want clean and ready before we start our job.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6529>

14 months agoci/bare-metal: Log why our run restarts when it does.
Eric Anholt [Mon, 31 Aug 2020 20:41:57 +0000 (13:41 -0700)]
ci/bare-metal: Log why our run restarts when it does.

It would be confusing to see a job quietly restart itself in the middle.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6529>

14 months agoci/bare-metal: Include a timestamp in our serial reads.
Eric Anholt [Mon, 31 Aug 2020 19:41:57 +0000 (12:41 -0700)]
ci/bare-metal: Include a timestamp in our serial reads.

gitlab CI doesn't include timestamps in its logs by default, but it's
really useful for finding delays in our CI so stuff one in on the lines
coming in from serial and being output to the gitlab log.  The artifacts
file is still the raw serial output.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6529>

14 months agoci/bare-metal: Fix detection of "POWER_GOOD not seen in time" fails
Eric Anholt [Mon, 31 Aug 2020 19:08:34 +0000 (12:08 -0700)]
ci/bare-metal: Fix detection of "POWER_GOOD not seen in time" fails

We were only reading from the CPU serial, not EC, so we'd never notice
these sources of job timeouts.  I couldn't find a cleaner solution, so I
spawned two threads to do the blocking reads from our serial line fifos
and merge them together in a single queue to read.

Closes: #3470
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6529>

14 months agoci/bare-metal: Use re.search() instead re.match() for our line matching.
Eric Anholt [Mon, 31 Aug 2020 20:38:15 +0000 (13:38 -0700)]
ci/bare-metal: Use re.search() instead re.match() for our line matching.

match() looks for the start of the line to match our regex, while search
just looks for the regex anywhere in the line.  I messed this up when
converting our greps in shell to python, which was part of breaking the
POWER_GOOD flake detection.  Most of our matches worked, but let's
consistently use this one so we don't mess this up in the future.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6529>

14 months agoamd/common: Fix various non-critical integer overflows
Tony Wasserka [Wed, 2 Sep 2020 17:15:01 +0000 (19:15 +0200)]
amd/common: Fix various non-critical integer overflows

The result of 0xf << 28 is a signed integer and hence overflows into the sign
bit. In practice compilers did the right thing here, since the intent of the
code was unsigned arithmetic anyway.

Cc: mesa-stable
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6568>

14 months agoaco: Fix integer overflows when emitting parallel copies during RA
Tony Wasserka [Wed, 2 Sep 2020 16:28:36 +0000 (18:28 +0200)]
aco: Fix integer overflows when emitting parallel copies during RA

32-bit shifts were accidentally used before this change despite the intended
output being 64 bits.

This was observed when compiling Dolphin's ubershaders.

Cc: mesa-stable
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6568>

14 months agoradv: Fix various non-critical integer overflows
Tony Wasserka [Wed, 2 Sep 2020 16:13:57 +0000 (18:13 +0200)]
radv: Fix various non-critical integer overflows

The result of 0xf << 28 is a signed integer and hence overflows into the sign
bit. In practice compilers did the right thing here, since the intent of the
code was unsigned arithmetic anyway.

These conditions were observed in:
* dEQP-VK.pipeline.image.suballocation.sampling_type.combined.view_type.1d.format.r4g4b4a4_unorm_pack16.count_8.size.512x1
* dEQP-VK.binding_model.descriptorset_random.sets32.noarray.ubolimitlow.sbolimitlow.sampledimglow.outimgonly.noiub.nouab.frag.ialimithigh.0

Cc: mesa-stable
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6568>

14 months agoaco: remove omod_success/clamp_success
Rhys Perry [Wed, 12 Aug 2020 14:58:32 +0000 (15:58 +0100)]
aco: remove omod_success/clamp_success

This simplifies the optimizer and should make SDWA optimizations easier.

No fossil-db changes on Navi.

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/6293>

14 months agoaco: fix mad splitting after applying output modifiers
Rhys Perry [Tue, 17 Dec 2019 14:55:24 +0000 (14:55 +0000)]
aco: fix mad splitting after applying output modifiers

Previously, this wasn't done because the mad label wasn't passed to the
new definition.

fossil-db (Navi):
Totals from 5770 (4.24% of 135946) affected shaders:
SGPRs: 391920 -> 391872 (-0.01%)
VGPRs: 349084 -> 348424 (-0.19%); split: -0.20%, +0.01%
CodeSize: 34639636 -> 34637496 (-0.01%); split: -0.02%, +0.01%
MaxWaves: 58828 -> 58862 (+0.06%)
Instrs: 6723436 -> 6723297 (-0.00%); split: -0.02%, +0.02%
Cycles: 197594168 -> 197591968 (-0.00%); split: -0.02%, +0.02%

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/6293>

14 months agoradv: remove descriptor_indexing fails from expected fails
Rhys Perry [Thu, 3 Sep 2020 18:06:39 +0000 (19:06 +0100)]
radv: remove descriptor_indexing fails from expected fails

I think these should be fixed since
https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6451. They also
pass for me on polaris and navi.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6586>

14 months agoanv: Set alignments on UBO/SSBO root derefs
Jason Ekstrand [Fri, 28 Aug 2020 20:42:45 +0000 (15:42 -0500)]
anv: Set alignments on UBO/SSBO root derefs

This doesn't really do anything for us today.  One day, I suppose we
could use it to do something with wide loads with non-uniform offsets.
The big reason to do this is to get better testing to make sure that NIR
doesn't blow up on the deref paths.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agospirv: Drop the OpenCL type layout code
Jason Ekstrand [Thu, 27 Aug 2020 21:00:26 +0000 (16:00 -0500)]
spirv: Drop the OpenCL type layout code

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agoclover/nir: Use lower_vars_to_explicit for uniform and global
Jason Ekstrand [Thu, 27 Aug 2020 20:56:38 +0000 (15:56 -0500)]
clover/nir: Use lower_vars_to_explicit for uniform and global

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agospirv: Stop counting inputs in entry_point_wrapper
Jason Ekstrand [Tue, 1 Sep 2020 01:09:42 +0000 (20:09 -0500)]
spirv: Stop counting inputs in entry_point_wrapper

nir_shader::num_inputs isn't supposed to be a count of how many input
variables we have.  It's a size of the lowered input space.

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agoclover: Use args.size() to compute new var locations
Jason Ekstrand [Tue, 1 Sep 2020 15:30:17 +0000 (10:30 -0500)]
clover: Use args.size() to compute new var locations

This is better than using num_uniforms as it guarantees what we want: a
mapping from nir_variable to the args vector.

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agonir: Allow uniform in nir_lower_vars_to_explicit_types
Jason Ekstrand [Thu, 27 Aug 2020 21:00:00 +0000 (16:00 -0500)]
nir: Allow uniform in nir_lower_vars_to_explicit_types

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agonir: Allow var_mem_global in nir_lower_vars_to_explicit_types
Jason Ekstrand [Thu, 27 Aug 2020 21:12:57 +0000 (16:12 -0500)]
nir: Allow var_mem_global in nir_lower_vars_to_explicit_types

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agospirv: Propagate alignments to deref chains via casts
Jason Ekstrand [Thu, 27 Aug 2020 23:34:50 +0000 (18:34 -0500)]
spirv: Propagate alignments to deref chains via casts

This commit propagates the alignment information provided either through
the Alignment decoration on pointers or via the alignment mem operands
to OpLoad, OpStore, and OpCopyMemory to the NIR deref chain.  It does so
by wrapping the deref in a cast.  NIR should be able to clean up most
unnecessary casts only leaving us with the useful alignment information.

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agospirv: Add pointer helper vars to OpCopyMemory
Jason Ekstrand [Thu, 27 Aug 2020 23:33:50 +0000 (18:33 -0500)]
spirv: Add pointer helper vars to OpCopyMemory

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agonir/opt_deref: Remove restrictive alignment information from casts
Jason Ekstrand [Mon, 24 Aug 2020 16:27:31 +0000 (11:27 -0500)]
nir/opt_deref: Remove restrictive alignment information from casts

If we have a cast deref with alignment information and we can get equal
or better alignment information from something further up the deref
chain, we can strip the alignment information from the cast and allow
other optimizations to potentially eliminate the cast.

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agonir/opt_deref: Don't remove casts with alignment information
Jason Ekstrand [Mon, 24 Aug 2020 16:26:18 +0000 (11:26 -0500)]
nir/opt_deref: Don't remove casts with alignment information

Generally, if a cast has alignment information, that information is
useful and we don't want to loose it.

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agonir/lower_io: Apply alignments from derefs when available
Jason Ekstrand [Mon, 24 Aug 2020 15:57:57 +0000 (10:57 -0500)]
nir/lower_io: Apply alignments from derefs when available

If the deref has no explicit alignment in the chain, we assume component
alignment which is what we currently assume for all derefs today.  This
should be correct for all APIs in the sense that we can usually assume
at least component alignment.  However, for some APIs such as OpenCL, we
could potentially make larger alignment assumptions.  The intention is
that those will be handled via alignment-increasing casts.

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agonir: Add a helper for getting the alignment of a deref
Jason Ekstrand [Mon, 24 Aug 2020 15:48:51 +0000 (10:48 -0500)]
nir: Add a helper for getting the alignment of a deref

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agonir: Handle all array stride cases in nir_deref_instr_array_stride
Jason Ekstrand [Thu, 27 Aug 2020 16:59:54 +0000 (11:59 -0500)]
nir: Handle all array stride cases in nir_deref_instr_array_stride

This renames it to drop the ptr_as and makes it handle all of the stride
cases.  There's a bit of a tricky bit in here around Booleans but we
currently use 32-bit for those always.

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agonir: Add alignment information to cast derefs
Jason Ekstrand [Mon, 24 Aug 2020 14:51:04 +0000 (09:51 -0500)]
nir: Add alignment information to cast derefs

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agonir/glsl: Add an explicit_alignment field to glsl_type
Jason Ekstrand [Sat, 29 Aug 2020 23:01:46 +0000 (18:01 -0500)]
nir/glsl: Add an explicit_alignment field to glsl_type

When creating explicit type, the alignment information is lost, thus
forcing explicit type users to recalculate the alignment using the same
size_align() function. Let's add a new field to cache this information.

Only structs, matrices, and vectors have and explicit alignment.  Arrays
alignment is implicitly set to its element alignment and matrices are
required to have an alignment that matches that of its vector columns.
the concept of alignment simply doesn't apply to other types.

We make the strategic choice to not allow explicit alignments on
scalars.  This is for a couple of reasons:

 1. There are no cases today where we use explicit types where we want
    any other alignment for scalars than natural alignment.

 2. Vectors don't have a component alignment that's separate from the
    explicit_alignment so it's impossible to get an explicitly aligned
    scalar type which is the component of the explicitly aligned vector
    type.

This choice may cause problems if we ever want to use explicitly laid
out types for things like varyings where we sometimes want vec4
alignment of scalars.  We can deal with that when the time comes.

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agocompiler/types: Make booleans 32-bit for cl_size/align
Jason Ekstrand [Tue, 1 Sep 2020 21:24:19 +0000 (16:24 -0500)]
compiler/types: Make booleans 32-bit for cl_size/align

OpenCL doesn't mandate a size and this is consistent with the rest of
the glsl_type system.  While we're here, we also clean ::cl_size() up a
bit and use a new explicit_type_scalar_byte_size() helper.

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agonir: Expose the packed attribute attached to glsl_type objects
Boris Brezillon [Mon, 29 Jun 2020 07:29:09 +0000 (09:29 +0200)]
nir: Expose the packed attribute attached to glsl_type objects

This should help code calculating field offsets to get it right when
the structure is marked packed.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agonir/glsl: Consider block interfaces as structs when it comes to size/align calculation
Boris Brezillon [Tue, 30 Jun 2020 10:03:00 +0000 (12:03 +0200)]
nir/glsl: Consider block interfaces as structs when it comes to size/align calculation

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agoglsl: Propagate packed info in get_explicit_type_for_size_align()
Boris Brezillon [Fri, 26 Jun 2020 08:27:16 +0000 (10:27 +0200)]
glsl: Propagate packed info in get_explicit_type_for_size_align()

Right now, when calling get_explicit_type_for_size_align() on a packed
struct, the packed attribute is lost and field offsets are wrong.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agospirv: Propagate packed information to glsl_type
Boris Brezillon [Tue, 23 Jun 2020 10:51:49 +0000 (12:51 +0200)]
spirv: Propagate packed information to glsl_type

We need to parse the CPacked decoration early enough to apply it when
calculating field offsets and creating the struct type.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agospirv: Don't accept CPacked decoration on struct members
Boris Brezillon [Tue, 23 Jun 2020 10:49:23 +0000 (12:49 +0200)]
spirv: Don't accept CPacked decoration on struct members

CPacked decoration is only allowed on struct definitions, not struct
members.

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agoclover: Call nir_lower_mem_constant_vars
Jason Ekstrand [Wed, 2 Sep 2020 22:43:07 +0000 (17:43 -0500)]
clover: Call nir_lower_mem_constant_vars

Fixes: 26a4c8f375e "clover/nir: Use nir_var_mem_constant for..."
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agonir: Don't bail too early in lower_mem_constant_vars
Jason Ekstrand [Wed, 2 Sep 2020 22:38:23 +0000 (17:38 -0500)]
nir: Don't bail too early in lower_mem_constant_vars

If there were no constant variables, we would bail out entirely.
However, we may still have constant input pointers coming in from the
client.

Fixes: 4360a8a2b3fce "nir/lower_io: Add support for nir_var_mem_constant"
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6472>

14 months agointel/nir: Stop using nir_lower_vars_to_scratch
Jason Ekstrand [Tue, 14 Jul 2020 21:01:18 +0000 (16:01 -0500)]
intel/nir: Stop using nir_lower_vars_to_scratch

Instead, we do a limited indirect deref lowering and then use
nir_lower_vars_to_explicit_types and nir_lower_explicit_io to lower it
as if it were SSBO or global memory access.  Among other things, this
should enable pointer arithmetic on local variables.  Fun!

The only shader-db change from this change on ICL was a few tiny cycle
count changes in 7 Aztec Ruins compute shaders.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5909>

14 months agonir/lower_indirect_derefs: Add a threshold
Jason Ekstrand [Tue, 14 Jul 2020 18:55:19 +0000 (13:55 -0500)]
nir/lower_indirect_derefs: Add a threshold

Instead of always lowering everything, we add a threshold such that if
the total indirected array size (AoA size) is above that threshold, it
won't lower.  It's assumed that the driver will sort things out somehow
by, for instance, lowering to scratch.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5909>

14 months agointel/compiler: Handle all indirect lowering choices in brw_nir.c
Jason Ekstrand [Tue, 14 Jul 2020 20:19:59 +0000 (15:19 -0500)]
intel/compiler: Handle all indirect lowering choices in brw_nir.c

Since everything flows through NIR and we're doing all of our indirect
deref lowering there now, there's no reason to keep making those
decisions in brw_compiler and stuffing them in the GLSL compiler
structs.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5909>

14 months agozink: generically handle matrix types
Mike Blumenkrantz [Tue, 23 Jun 2020 18:50:37 +0000 (14:50 -0400)]
zink: generically handle matrix types

there's a bunch of glsl 1.10 tests for this

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6268>

14 months agogallium/util: use uint sampler for stencil-reads
Erik Faye-Lund [Mon, 31 Aug 2020 08:14:02 +0000 (10:14 +0200)]
gallium/util: use uint sampler for stencil-reads

Some drivers can't use float-samplers to read integer textures, so let's
make sure the stenicil-sampler has the right type.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6512>

14 months agoradeonsi: optimize out the loop in si_get_ps_input_cntl
Marek Olšák [Sat, 15 Aug 2020 08:39:30 +0000 (04:39 -0400)]
radeonsi: optimize out the loop in si_get_ps_input_cntl

Use a remap table from a semantic to an index instead of searching
for the correct index.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6340>

14 months agoradeonsi: replace TGSI_SEMANTIC with VARYING_SLOT and FRAG_RESULT
Marek Olšák [Sat, 15 Aug 2020 07:09:34 +0000 (03:09 -0400)]
radeonsi: replace TGSI_SEMANTIC with VARYING_SLOT and FRAG_RESULT

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6340>

14 months agoradeonsi: replace TGSI_INTERPOLATE with INTERP_MODE
Marek Olšák [Sat, 15 Aug 2020 06:46:35 +0000 (02:46 -0400)]
radeonsi: replace TGSI_INTERPOLATE with INTERP_MODE

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6340>

14 months agocompiler: add INTERP_MODE_COLOR for radeonsi
Marek Olšák [Sat, 15 Aug 2020 06:45:05 +0000 (02:45 -0400)]
compiler: add INTERP_MODE_COLOR for radeonsi

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6340>

14 months agoradeonsi: remove si_shader_selector::type
Marek Olšák [Sat, 15 Aug 2020 05:01:08 +0000 (01:01 -0400)]
radeonsi: remove si_shader_selector::type

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6340>

14 months agoradeonsi: change PIPE_SHADER to MESA_SHADER (si_dump_descriptors)
Marek Olšák [Sat, 15 Aug 2020 03:56:05 +0000 (23:56 -0400)]
radeonsi: change PIPE_SHADER to MESA_SHADER (si_dump_descriptors)

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6340>

14 months agoradeonsi: precompute si_*_descriptors_idx in si_shader_selector
Marek Olšák [Sat, 15 Aug 2020 04:56:39 +0000 (00:56 -0400)]
radeonsi: precompute si_*_descriptors_idx in si_shader_selector

It helps remove one use of sel->type.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6340>

14 months agoradeonsi: change PIPE_SHADER to MESA_SHADER (si_shader_dump_disassembly)
Marek Olšák [Sat, 15 Aug 2020 03:56:05 +0000 (23:56 -0400)]
radeonsi: change PIPE_SHADER to MESA_SHADER (si_shader_dump_disassembly)

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6340>

14 months agoradeonsi: remove unused si_shader_context::type
Marek Olšák [Sat, 15 Aug 2020 04:50:27 +0000 (00:50 -0400)]
radeonsi: remove unused si_shader_context::type

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6340>

14 months agoradeonsi: change PIPE_SHADER to MESA_SHADER (si_get_shader_part)
Marek Olšák [Sat, 15 Aug 2020 03:56:05 +0000 (23:56 -0400)]
radeonsi: change PIPE_SHADER to MESA_SHADER (si_get_shader_part)

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6340>

14 months agoradeonsi: change PIPE_SHADER to MESA_SHADER (si_compile_llvm)
Marek Olšák [Sat, 15 Aug 2020 03:56:05 +0000 (23:56 -0400)]
radeonsi: change PIPE_SHADER to MESA_SHADER (si_compile_llvm)

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6340>

14 months agoradeonsi: change PIPE_SHADER to MESA_SHADER (debug flags)
Marek Olšák [Sat, 15 Aug 2020 03:56:05 +0000 (23:56 -0400)]
radeonsi: change PIPE_SHADER to MESA_SHADER (debug flags)

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6340>

14 months agoradeonsi: change PIPE_SHADER to MESA_SHADER (si_shader_context::type)
Marek Olšák [Sat, 15 Aug 2020 03:56:05 +0000 (23:56 -0400)]
radeonsi: change PIPE_SHADER to MESA_SHADER (si_shader_context::type)

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6340>

14 months agoradeonsi: change PIPE_SHADER to MESA_SHADER (si_shader_selector::type)
Marek Olšák [Sat, 15 Aug 2020 03:56:05 +0000 (23:56 -0400)]
radeonsi: change PIPE_SHADER to MESA_SHADER (si_shader_selector::type)

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6340>

14 months agoradeonsi: simplify handling color interp modes in si_emit_spi_map
Marek Olšák [Sat, 15 Aug 2020 06:48:13 +0000 (02:48 -0400)]
radeonsi: simplify handling color interp modes in si_emit_spi_map

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6340>

14 months agoradeonsi: don't execute LDS stores for TCS outputs that are never read
Marek Olšák [Sat, 15 Aug 2020 03:41:13 +0000 (23:41 -0400)]
radeonsi: don't execute LDS stores for TCS outputs that are never read

This is a per-component version of the previous mechanism.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6340>

14 months agoradeonsi: don't lower indirect IO in GLSL
Marek Olšák [Sat, 22 Aug 2020 12:41:23 +0000 (08:41 -0400)]
radeonsi: don't lower indirect IO in GLSL

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6445>

14 months agoradeonsi: remove in/out/uniform variables from NIR after lowering IO
Marek Olšák [Fri, 14 Aug 2020 23:11:37 +0000 (19:11 -0400)]
radeonsi: remove in/out/uniform variables from NIR after lowering IO

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6445>

14 months agoradeonsi: lower IO intrinsics - complete rewrite of input/output scanning
Marek Olšák [Fri, 14 Aug 2020 22:09:39 +0000 (18:09 -0400)]
radeonsi: lower IO intrinsics - complete rewrite of input/output scanning

Input and output info is gathered from intrinsics. nir_variables are
ignored (and we'll remove them anyway).

This is a prerequisite for ACO, but also makes the IR prettier.
The ac_nir_to_llvm change has to be in this commit.

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6445>

14 months agoac/nir: handle all lowered IO intrinsics
Marek Olšák [Fri, 14 Aug 2020 22:08:20 +0000 (18:08 -0400)]
ac/nir: handle all lowered IO intrinsics

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6445>

14 months agoradeonsi: clean up code for loading VS inputs
Marek Olšák [Fri, 14 Aug 2020 21:56:42 +0000 (17:56 -0400)]
radeonsi: clean up code for loading VS inputs

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6445>

14 months agoradeonsi: get color interpolation info from shader_info
Marek Olšák [Fri, 14 Aug 2020 23:13:21 +0000 (19:13 -0400)]
radeonsi: get color interpolation info from shader_info

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6445>

14 months agoradeonsi: don't crash if input_usage_mask is 0 for a VS input
Marek Olšák [Fri, 14 Aug 2020 02:12:12 +0000 (22:12 -0400)]
radeonsi: don't crash if input_usage_mask is 0 for a VS input

This will start happening with the lowered IO intrinstics and new scanning
code.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6445>

14 months agofreedreno: fence_server_sync() fixes
Rob Clark [Tue, 1 Sep 2020 22:24:38 +0000 (15:24 -0700)]
freedreno: fence_server_sync() fixes

Two potential problems, batch re-ordering doesn't really play nicely
with fence_server_sync(), so when we switch away from one batch, detect
the case that we need to sync, and if so flush.  The alternative of
trying to track that later batches depend on an earlier batch that had
an in-fence is hairy, and the normal use-case would be to sync at the
beginning of the frame.

But this brings up the second problem, which is that typically we'll get
told to sync on an in-fence before the first draw, which means before
mesa/st flushes down the framebuffer state to the driver.  Which means
we don't yet have the correct batch to attach the fence to.  So we need
to track the in-fence on the context, and transfer it to the batch
before draws, etc.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6575>

14 months agofreedreno: Fix missing rsc->seqno updates
Rob Clark [Tue, 1 Sep 2020 18:23:49 +0000 (11:23 -0700)]
freedreno: Fix missing rsc->seqno updates

There were a couple paths where we weren't getting valid seqno's, which
are supposed to be updated whenever the backing bo is set/changed.  So
wrap that up in a helper to make it harder to mess up.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6575>

14 months agodocs: shift 20.2 rc dates by two weeks to match reality
Eric Engestrom [Wed, 2 Sep 2020 20:44:42 +0000 (22:44 +0200)]
docs: shift 20.2 rc dates by two weeks to match reality

The release candidates have slipped by a couple of weeks, so let's fix
the dates in the calendar.

Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6574>

14 months agodocs: update calendar and link releases notes for 20.1.7
Eric Engestrom [Wed, 2 Sep 2020 20:10:09 +0000 (22:10 +0200)]
docs: update calendar and link releases notes for 20.1.7

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6573>

14 months agodocs: add release notes for 20.1.7
Eric Engestrom [Wed, 2 Sep 2020 19:52:02 +0000 (21:52 +0200)]
docs: add release notes for 20.1.7

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6573>

14 months agoiris: Re-emit push constants if we have a varying workgroup size
Jason Ekstrand [Wed, 2 Sep 2020 18:48:51 +0000 (13:48 -0500)]
iris: Re-emit push constants if we have a varying workgroup size

Fixes: 33c61eb2f10526 "iris: Implement ARB_compute_variable_group_size"
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6570>

14 months agointel/nir: Lower load_num_work_groups to 32-bit if needed
Jason Ekstrand [Tue, 1 Sep 2020 19:46:25 +0000 (14:46 -0500)]
intel/nir: Lower load_num_work_groups to 32-bit if needed

For OpenCL-style kernels, this builtin is 64-bit.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6570>

14 months agointel/fs: Use a single untyped surface read for load_num_work_groups
Jason Ekstrand [Tue, 1 Sep 2020 19:34:44 +0000 (14:34 -0500)]
intel/fs: Use a single untyped surface read for load_num_work_groups

There's no good reason to split this into three.  Sure, CS indirects are
only guaranteed by the spec to be DWORD aligned, but that's all untyped
surface reads require anyway.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6570>

14 months agointel/fs: Don't copy-propagate stride=0 sources into ddx/ddy
Jason Ekstrand [Wed, 2 Sep 2020 15:13:16 +0000 (10:13 -0500)]
intel/fs: Don't copy-propagate stride=0 sources into ddx/ddy

This can come up if, for instance, the shader does a derivative of a
uniform or flat input.  Ideally, NIR would use divergence analysis to
get rid of the derivative in this case but it doesn't right now.  This
fixes a crash in F1 2017.

Cc: mesa-stable@lists.freedesktop.org
Reported-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Tested-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6564>

14 months agost/mesa: fix lowered IO - don't call st_nir_assign_vs_in_locations twice
Marek Olšák [Tue, 1 Sep 2020 16:19:32 +0000 (12:19 -0400)]
st/mesa: fix lowered IO - don't call st_nir_assign_vs_in_locations twice

If IO is lowered, the second call is a no-op, which breaks:
   spec@!opengl 1.1@gl-1.1-color-material-unused-normal-array

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6540>

14 months agonir: fix a bug in is_dual_slot in nir_io_add_const_offset_to_base
Marek Olšák [Tue, 1 Sep 2020 15:37:07 +0000 (11:37 -0400)]
nir: fix a bug in is_dual_slot in nir_io_add_const_offset_to_base

Fixes: 01ab308edc "nir: update IO semantics in nir_io_add_const_offset_to_base"
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6540>

14 months agoiris: Patch constant data pointers into shaders
Jason Ekstrand [Wed, 12 Aug 2020 00:43:17 +0000 (19:43 -0500)]
iris: Patch constant data pointers into shaders

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6244>

14 months agoanv: Patch constant data pointers into shaders with using softpin
Jason Ekstrand [Sat, 8 Aug 2020 21:29:36 +0000 (16:29 -0500)]
anv: Patch constant data pointers into shaders with using softpin

When we have softpin, we know the address of the shader constant data at
shader upload time because it's sitting at the end of the shader.  This
commit changes ANV to use patch constants to embed the address in the
shader patch the right address in at upload time.  This allows us to
avoid having to set up a UBO binding on-the-fly for shader constants.

This commit uses an A64 message but it's quite possible that we could
also use an A32 message and make the dataport do the 64-bit add for us.
However, load_global is what we have right now so it was easier to just
use that.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6244>

14 months agonir/builder: Add load/store_global helpers
Jason Ekstrand [Tue, 19 May 2020 18:53:41 +0000 (13:53 -0500)]
nir/builder: Add load/store_global helpers

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6244>

14 months agoanv: Properly cache brw_stage_prog_data::relocs
Jason Ekstrand [Mon, 10 Aug 2020 16:46:12 +0000 (11:46 -0500)]
anv: Properly cache brw_stage_prog_data::relocs

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6244>

14 months agointel/fs: Add support for a new load_reloc_const intrinsic
Jason Ekstrand [Sat, 8 Aug 2020 18:56:16 +0000 (13:56 -0500)]
intel/fs: Add support for a new load_reloc_const intrinsic

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6244>

14 months agointel/eu: Add a mechanism for emitting relocatable constant MOVs
Jason Ekstrand [Sat, 8 Aug 2020 17:55:29 +0000 (12:55 -0500)]
intel/eu: Add a mechanism for emitting relocatable constant MOVs

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6244>

14 months agointel/eu: Include brw_compiler.h in brw_eu.h
Jason Ekstrand [Mon, 10 Aug 2020 16:09:10 +0000 (11:09 -0500)]
intel/eu: Include brw_compiler.h in brw_eu.h

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6244>

14 months agoanv: Stop storing the shader constant data side-band
Jason Ekstrand [Sat, 8 Aug 2020 03:31:44 +0000 (22:31 -0500)]
anv: Stop storing the shader constant data side-band

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6244>

14 months agointel/fs,vec4: Stuff the constant data from NIR in the end of the program
Jason Ekstrand [Sat, 8 Aug 2020 03:26:07 +0000 (22:26 -0500)]
intel/fs,vec4: Stuff the constant data from NIR in the end of the program

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6244>

14 months agointel/eu: Add some new helpers
Jason Ekstrand [Sat, 8 Aug 2020 02:59:12 +0000 (21:59 -0500)]
intel/eu: Add some new helpers

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6244>

14 months agoiris: Use gen_disassemble
Jason Ekstrand [Tue, 1 Sep 2020 23:57:52 +0000 (18:57 -0500)]
iris: Use gen_disassemble

This one doesn't require the program size and so it won't mess up if we
have a bunch of constant data at the end.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6244>

14 months agointel/compiler: Get rid of struct gen_disasm
Jason Ekstrand [Tue, 1 Sep 2020 23:39:54 +0000 (18:39 -0500)]
intel/compiler: Get rid of struct gen_disasm

It's just a container around a devinfo.  The one useful purpose it did
serve is that gen_disasm_create initialized the compaction table
singletons.  Now that those no longer exist, this isn't necessary.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6244>

14 months agointel/compiler: Get rid of the global compaction table pointers
Jason Ekstrand [Tue, 1 Sep 2020 23:28:58 +0000 (18:28 -0500)]
intel/compiler: Get rid of the global compaction table pointers

With discrete GPUs, it's going to be possible to have GPUs from two
different hardware generations in the machine at the same time.  Global
singletons like this aren't going to fly.  Have a struct containing the
pointers which gets initialized once per shader disassemble instead.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6244>

14 months agospirv: Deal with glslang not setting NonUniform on constructors.
Bas Nieuwenhuizen [Mon, 31 Aug 2020 19:58:12 +0000 (21:58 +0200)]
spirv: Deal with glslang not setting NonUniform on constructors.

Especially a problem for OpImage/OpSampledImage. Note that the problem
doesn't seem to be propagation through glslang, but only in emitting
the SPIR-V. So it is fine if we are somewhat lossy in handling this, as
long as direct Op(Sampled)Image -> texture op chains work.

Fixes: af81486a8cd "spirv: Simplify our handling of NonUniform"
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3406
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6451>

14 months agospirv: Deal with glslang bug not setting the decoration for stores.
Bas Nieuwenhuizen [Mon, 24 Aug 2020 22:20:18 +0000 (00:20 +0200)]
spirv: Deal with glslang bug not setting the decoration for stores.

Fixes: af81486a8cd "spirv: Simplify our handling of NonUniform"
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6451>

14 months agoradv: Avoid deadlock on bo_list.
Bas Nieuwenhuizen [Sun, 30 Aug 2020 23:57:36 +0000 (01:57 +0200)]
radv: Avoid deadlock on bo_list.

With the kernel timeline sysncobj changes, the kernel submits do
not necessarily happen in global vkQueueSubmit order. Which should
be fine, we added the appropriate waits for that. (See
DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT in the winsys)

However, all kernel submissions take a lock on the bo_list mutex,
and since we do the wait in the winsys, we wait while having the
bo_list mutex held. This means that as soon as a wait and a signal
submission are out of order we have a deadlock on the bo_list mutex
and the wait.

Solution is to use a shared reader lock during the kernel submission,
as we only need read access for the submission.

Fixes: 6bc5ce7a91d "radv: Add timeline syncobj for timeline semaphores."
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3446
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6478>

14 months agoradv: Fix threading issue with submission refcounts.
Bas Nieuwenhuizen [Sat, 29 Aug 2020 01:25:02 +0000 (03:25 +0200)]
radv: Fix threading issue with submission refcounts.

If decrement == 0 then:

- it isn't safe to access the submission
- even if it is, checking that the result of the atomic_sub is 0
  doesn't given an unique owner anymore.

So skip it. The submission always starts out with refcount >= 1,
so first one to decrement to 0 still get dibs on executing it.

Fixes: 4aa75bb3bdd "radv: Add wait-before-submit support for timelines."
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6478>

14 months agointel/tools: Disassemble WAIT's argument as a destination
Matt Turner [Tue, 1 Sep 2020 19:09:16 +0000 (12:09 -0700)]
intel/tools: Disassemble WAIT's argument as a destination

WAIT takes a notification register as a destination and a src0 argument.
Since the same notification register is specified in both fields, we
treat it as a special case and disassemble it only once.

If we disassemble it as if it is a source register, its scalar region
will be printed as <0,1,0>. This causes difficulties round-tripping
through the assembler <-> disassembler because that is not an acceptable
destination region. If we instead disassemble the destination, we
instead get a <1> region which is an acceptable and equivalent region
for source and destination.

The test .asm files are regenerated by round-tripping them through the
assembler/disassembler. Note that the <0> region in the tests was a
harmless mistake: the compiler translated it to a <0,1,0> source region
and a <1> destination region, since <0> isn't valid.

Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6543>

14 months agogallium/tgsi_exec: Fix up NumOutputs counting
Eric Anholt [Tue, 25 Aug 2020 23:01:47 +0000 (16:01 -0700)]
gallium/tgsi_exec: Fix up NumOutputs counting

We can get duplicate declarations for an index (for example dvec3 + float
packed into 2 vec4s, the second one won't pack into the first's array
decl), and we'd end up stepping by the wrong amount in GS vtx/prim emit.

Fixes vs-gs-fs-double, sso-vs-gs-fs-array-interleave piglit tests.

Fixes: 49155c3264d0 ("draw/tgsi: fix geometry shader input/output swizzling")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6567>

14 months agogallium/tgsi_exec: Add missing DFLR opcode support.
Eric Anholt [Fri, 24 Jul 2020 19:06:40 +0000 (12:06 -0700)]
gallium/tgsi_exec: Add missing DFLR opcode support.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6567>

14 months agonir/clone: Add a helper for cloning most instruction types
Jason Ekstrand [Mon, 18 May 2020 15:39:43 +0000 (10:39 -0500)]
nir/clone: Add a helper for cloning most instruction types

@anholt needed it for nir_to_tgsi, and the desire comes up frequently.

Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6567>

14 months agonir/opt_vectorize: Add a callback for filtering of vectorizing.
Eric Anholt [Thu, 27 Aug 2020 19:49:13 +0000 (12:49 -0700)]
nir/opt_vectorize: Add a callback for filtering of vectorizing.

For NIR-to-TGSI, we don't want to revectorize 64-bit ops that we split to
scalar beyond vec2 width.  We even have some ops that we would rather
retain as scalar due to TGSI opcodes being scalar, or having more unusual
requirements.

This could be used to do the vectorize_vec2_16bit filtering, but that
shader compiler option is also used in algebraic so leave it in place for
now.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6567>