platform/upstream/mesa.git
19 months agoanv/blorp: support surf generation for addresses
Felix DeGrood [Wed, 22 Mar 2023 18:04:42 +0000 (18:04 +0000)]
anv/blorp: support surf generation for addresses

Already have support for anv_buff. Extended to support addresses.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22178>

19 months agodzn: Enable VK_KHR_get_memory_requirements2
Raun [Thu, 6 Apr 2023 19:31:43 +0000 (14:31 -0500)]
dzn: Enable VK_KHR_get_memory_requirements2

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22349>

19 months agodzn: Enable VK_KHR_bind_memory2
Raun [Thu, 6 Apr 2023 19:23:42 +0000 (14:23 -0500)]
dzn: Enable VK_KHR_bind_memory2

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22349>

19 months agoradv: import retained NIR shaders later in the compilation process
Samuel Pitoiset [Fri, 7 Apr 2023 09:36:08 +0000 (11:36 +0200)]
radv: import retained NIR shaders later in the compilation process

This allows us to remove the intermediate NIR shader pointer.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22356>

19 months agoradv: do not retain noop FS for libs when a cache hit happened
Samuel Pitoiset [Fri, 7 Apr 2023 09:05:57 +0000 (11:05 +0200)]
radv: do not retain noop FS for libs when a cache hit happened

Determine if the graphics pipeline needs a noop FS later instead of
retaining it. This was also suboptimal.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22356>

19 months agoradv: simplify a check when retaining NIR shaders
Samuel Pitoiset [Fri, 7 Apr 2023 08:55:24 +0000 (10:55 +0200)]
radv: simplify a check when retaining NIR shaders

The RETAIN flag is only allowed with graphics libs.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22356>

19 months agoradv: move the serialized NIR to radv_graphics_lib_pipeline
Samuel Pitoiset [Fri, 7 Apr 2023 08:53:38 +0000 (10:53 +0200)]
radv: move the serialized NIR to radv_graphics_lib_pipeline

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22356>

19 months agoradv: add a helper for retaining NIR shaders
Samuel Pitoiset [Fri, 7 Apr 2023 08:44:55 +0000 (10:44 +0200)]
radv: add a helper for retaining NIR shaders

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22356>

19 months agozink: don't access non_fs part of zink_shader from fs
Mike Blumenkrantz [Thu, 6 Apr 2023 18:49:31 +0000 (14:49 -0400)]
zink: don't access non_fs part of zink_shader from fs

Fixes: a6de15eff56 ("zink: add flags to `zink_gfx_program` and `zink_context`")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22347>

19 months agozink: more explicitly track/check rp optimizing per-context
Mike Blumenkrantz [Wed, 5 Apr 2023 15:35:53 +0000 (11:35 -0400)]
zink: more explicitly track/check rp optimizing per-context

if tc creation fails for whatever reason, rp optimizing must be
marked as disabled for that context to avoid erroneous assumptions
about rp operation

fixes #8787

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22319>

19 months agoac/llvm: remove some unused code replaced by nir
Qiang Yu [Wed, 5 Apr 2023 02:09:28 +0000 (10:09 +0800)]
ac/llvm: remove some unused code replaced by nir

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22304>

19 months agoac/nir/ngg: fix store shared alignment
Qiang Yu [Wed, 5 Apr 2023 01:52:01 +0000 (09:52 +0800)]
ac/nir/ngg: fix store shared alignment

For stream!=0, this align_mul=4 is not true. Not observe any
problem yet, just for correctness.

Fixes: 60ac5dda82e ("ac: Add NIR lowering for NGG GS.")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22304>

19 months agoac/nir/ngg: fix gs culling vertex liveness check for odd vertices
Qiang Yu [Wed, 5 Apr 2023 01:27:50 +0000 (09:27 +0800)]
ac/nir/ngg: fix gs culling vertex liveness check for odd vertices

If vertex does not complete a primitive, it should not set the odd
flag which miss lead liveness check when culling is enabled.

For example, if odd flag is set regardless of complete flag, when
culling is enabled, 3 vertices of a triangle's init prim flag:
  [0x00 0x04 0x01]

then after culling, this triangle has been culled, their prim flag:
  [0x00 0x04 0x00]

the second vertex is miss treat as live because its odd flag (code
check prim_flag!=0 for liveness).

Fixes: 1bdeb961bd8 ("ac/nir/ngg: add gs culling")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8725
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22304>

19 months agoradeonsi: fix max scrach lds size calculation when ngg
Qiang Yu [Tue, 4 Apr 2023 01:42:24 +0000 (09:42 +0800)]
radeonsi: fix max scrach lds size calculation when ngg

Fixes: 028d0590f85 ("radeonsi: replace llvm ngg vs/tes with nir lowering")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22304>

19 months agoasahi: Allow explicit non-LINEAR modifiers for scanout
Asahi Lina [Wed, 5 Apr 2023 12:42:14 +0000 (21:42 +0900)]
asahi: Allow explicit non-LINEAR modifiers for scanout

The compositor is responsible for picking the right supported modifiers
for scanout. If we get no modifiers, we have to assume linear, but if we
do, just roll with it and don't attempt to force things.

Signed-off-by: Asahi Lina <lina@asahilina.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Flip kmsro around to allocate on the GPU
Asahi Lina [Wed, 5 Apr 2023 12:41:27 +0000 (21:41 +0900)]
asahi: Flip kmsro around to allocate on the GPU

Our display controller can handle arbitrary GPU imports, so there is no
reason to use dumb KMS buffers. Allocate everything on the GPU instead.

This also allows us to be lazy about mapping things to the KMS side, so
only clients that really want a KMS handle actually do that, which stops
us from ending up with a bunch of junk mapped to DCP (e.g. X11 clients
always request SCANOUT even under XWayland).

Signed-off-by: Asahi Lina <lina@asahilina.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Fix compressed ZS support
Asahi Lina [Wed, 5 Apr 2023 16:42:51 +0000 (01:42 +0900)]
asahi: Fix compressed ZS support

Depth/stencil formats are "not renderable" but do support compression.

I swear I already fixed this at some point and the commit must've fallen
through the cracks...

Signed-off-by: Asahi Lina <lina@asahilina.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Print reasons why compression is disabled
Asahi Lina [Wed, 5 Apr 2023 07:17:50 +0000 (16:17 +0900)]
asahi: Print reasons why compression is disabled

For resource debug. Found a regression in compressed depth this way.

Signed-off-by: Asahi Lina <lina@asahilina.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Add resource debugging
Asahi Lina [Wed, 5 Apr 2023 06:54:04 +0000 (15:54 +0900)]
asahi: Add resource debugging

I keep re-implementing this every time I look at resource-related
issues. Let's just make it official so we can turn it on with a flag
instead of having to add printfs every time ^^

Signed-off-by: Asahi Lina <lina@asahilina.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Add a helper macro for debug/error messages
Asahi Lina [Wed, 5 Apr 2023 06:39:46 +0000 (15:39 +0900)]
asahi: Add a helper macro for debug/error messages

This includes the program short name in the message, which is useful
when running entire desktop sessions with a single log to figure out who
is doing what.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Make BO import path failures more robust
Asahi Lina [Wed, 29 Mar 2023 09:48:53 +0000 (18:48 +0900)]
asahi: Make BO import path failures more robust

These operations can fail for complex reasons through no fault of mesa,
so we should have proper runtime checks for them even in release builds.

Signed-off-by: Asahi Lina <lina@asahilina.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Implement valid buffer range tracking
Asahi Lina [Wed, 29 Mar 2023 10:20:12 +0000 (19:20 +0900)]
asahi: Implement valid buffer range tracking

A common pattern is to allocate a vertex/etc buffer and write to it in
subsets. Some games interleave this with draw calls using the buffer.
This causes very expensive flushing for every draw call.

Fix this by tracking which range of a buffer has been written to, and
elide syncs when the range was previously uninitialized.

Fixes Source engine game performance and probably helps a bunch of
others.

Signed-off-by: Asahi Lina <lina@asahilina.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Fix style nits
Asahi Lina [Wed, 5 Apr 2023 10:48:12 +0000 (19:48 +0900)]
asahi: Fix style nits

Found with a grep abomination which is probably too broken/silly to
actually implement in CI... but hey, at least it found some.

Signed-off-by: Asahi Lina <lina@asahilina.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Locate low VA BOs correctly
Asahi Lina [Wed, 5 Apr 2023 14:48:42 +0000 (23:48 +0900)]
asahi: Locate low VA BOs correctly

These need the shader_base added to them.

Signed-off-by: Asahi Lina <lina@asahilina.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Enable glthread
Asahi Lina [Wed, 29 Mar 2023 09:45:45 +0000 (18:45 +0900)]
asahi: Enable glthread

This helps a lot with FEX, since the GPU driver runs emulated (and only
64bit supports thunking).

Signed-off-by: Asahi Lina <lina@asahilina.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Make agx_alloc_staging() take a screen instead of a context
Asahi Lina [Wed, 5 Apr 2023 11:17:51 +0000 (20:17 +0900)]
asahi: Make agx_alloc_staging() take a screen instead of a context

This makes it clear that it is thread-safe.

Signed-off-by: Asahi Lina <lina@asahilina.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoRevert "asahi: Advertise dual-source blending"
Asahi Lina [Wed, 5 Apr 2023 11:59:50 +0000 (20:59 +0900)]
Revert "asahi: Advertise dual-source blending"

This reverts commit f4e2b2264663482f76681dbc0a8e9fc1574a3ce7.

This is broken until GL3 is enabled, possibly due to a core Mesa bug,
but it's a corner case not worth fixing.

Fixes Chromium.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoagx: Make signal_pix instructions explicit
Alyssa Rosenzweig [Thu, 30 Mar 2023 02:17:18 +0000 (22:17 -0400)]
agx: Make signal_pix instructions explicit

Rather than implicitly packing them with the sample_mask. Again, this is just
changing where they're emitted, no functional changes yet. Bug for bug
compatibility with the old behaviour.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoagx: Rename writeout to wait_pix
Alyssa Rosenzweig [Thu, 30 Mar 2023 02:08:47 +0000 (22:08 -0400)]
agx: Rename writeout to wait_pix

This is the name applegpu is currently using, to capture the semantics of a
pixel fence. I'm not sure what Apple calls this but wait_pix is closer than
writeout for sure.

This commit just does the rename. It doesn't fix the broken semantics we've had,
this is to ease review and bisection.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoagx: Tease apart some sample_mask packing magic
Alyssa Rosenzweig [Thu, 30 Mar 2023 01:58:59 +0000 (21:58 -0400)]
agx: Tease apart some sample_mask packing magic

There's a second instruction here, and a second source in the first instruction.
applegpu has known about the encodings for a while but I never updated the
packing code. We will need to stop hardcoding this for multisampling support, as
preparation tease apart the magic pieces.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Clamp texture buffer sizes
Alyssa Rosenzweig [Sat, 18 Mar 2023 21:34:17 +0000 (17:34 -0400)]
asahi: Clamp texture buffer sizes

Per the spec / freedreno. Fixes
arb_texture_buffer_object-texture-buffer-size-clamp

Fixes: 6b22a02f908 ("asahi,agx: Implement buffer textures with gnarly NIR")
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Dirty track depth bias uploads
Alyssa Rosenzweig [Sun, 26 Mar 2023 21:12:21 +0000 (17:12 -0400)]
asahi: Dirty track depth bias uploads

Reduces how much we upload in SuperTuxKart.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Lower 1D to 2D
Alyssa Rosenzweig [Sat, 18 Mar 2023 20:42:35 +0000 (16:42 -0400)]
asahi: Lower 1D to 2D

Khronos APIs require that we support mipmapping even for 1D textures. However,
it isn't clear if this is supported in the hardware, and how it would work even
if it is. But 1D textures are pretty useless, so we just lower 1D textures to 2D
textures instead of worrying about that.

Fixes piles of Piglits relating to 1D textures.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Implement null textures
Alyssa Rosenzweig [Sat, 18 Mar 2023 20:58:22 +0000 (16:58 -0400)]
asahi: Implement null textures

Use the same silly workaround that Metal does, to fill in texture descriptors
when there's nothing bound in the interest of robust behaviour.

Fixes null pointer dereference in
arb_shading_language_420pack-active-sampler-conflict.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Honour sampler count
Alyssa Rosenzweig [Tue, 28 Mar 2023 04:05:07 +0000 (00:05 -0400)]
asahi: Honour sampler count

It may not be equal to the texture count. Prevents a regression from the next
commit.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoagx: Don't overallocate registers
Alyssa Rosenzweig [Tue, 7 Mar 2023 04:09:38 +0000 (23:09 -0500)]
agx: Don't overallocate registers

We need to account for the full vector lengths. Especially important once we
start restricting the reg file.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoagx: Coalesce more collects
Alyssa Rosenzweig [Tue, 28 Feb 2023 21:44:12 +0000 (16:44 -0500)]
agx: Coalesce more collects

Try harder to coalesce collects, by trying to allocate collects only to regions
of the register file where we actually have a full vector worth of registers
free. If we already know that the vector will be blocked later, it's not a good
base register to pick since we'd be force to shuffle later. So, this tweak to
the collect coalescing heuristic lets us eliminate a pile of pointless copying.

shader-db results are excellent. Note that, although we use more registers,
none of the shaders tested had their thread count affected, likely because the
max HURT isn't too high and most of the scary % here is from using a few more
registers when the register pressure is already low. In the near future, that
property will become guaranteed thanks to live range splitting, too.

total instructions in shared programs: 1507337 -> 1500562 (-0.45%)
instructions in affected programs: 428137 -> 421362 (-1.58%)
helped: 2658
HURT: 167
helped stats (abs) min: 1.0 max: 34.0 x̄: 2.63 x̃: 2
helped stats (rel) min: 0.10% max: 25.00% x̄: 3.04% x̃: 2.14%
HURT stats (abs)   min: 1.0 max: 10.0 x̄: 1.24 x̃: 1
HURT stats (rel)   min: 0.20% max: 23.81% x̄: 3.90% x̃: 3.57%
95% mean confidence interval for instructions value: -2.49 -2.31
95% mean confidence interval for instructions %-change: -2.76% -2.51%
Instructions are helped.

total bytes in shared programs: 10333670 -> 10293172 (-0.39%)
bytes in affected programs: 2996682 -> 2956184 (-1.35%)
helped: 2660
HURT: 175
helped stats (abs) min: 2.0 max: 204.0 x̄: 15.70 x̃: 12
helped stats (rel) min: 0.08% max: 23.08% x̄: 2.64% x̃: 1.83%
HURT stats (abs)   min: 2.0 max: 60.0 x̄: 7.26 x̃: 6
HURT stats (rel)   min: 0.12% max: 22.39% x̄: 3.19% x̃: 2.78%
95% mean confidence interval for bytes value: -14.81 -13.76
95% mean confidence interval for bytes %-change: -2.39% -2.18%
Bytes are helped.

total halfregs in shared programs: 417284 -> 427363 (2.42%)
halfregs in affected programs: 49814 -> 59893 (20.23%)
helped: 95
HURT: 3018
helped stats (abs) min: 1.0 max: 8.0 x̄: 2.29 x̃: 2
helped stats (rel) min: 2.44% max: 28.57% x̄: 9.20% x̃: 6.06%
HURT stats (abs)   min: 1.0 max: 14.0 x̄: 3.41 x̃: 4
HURT stats (rel)   min: 2.08% max: 150.00% x̄: 36.54% x̃: 27.27%
95% mean confidence interval for halfregs value: 3.17 3.31
95% mean confidence interval for halfregs %-change: 34.05% 36.23%
Halfregs are HURT.

total threads in shared programs: 16465280 -> 16465280 (0.00%)
threads in affected programs: 0 -> 0
helped: 0
HURT: 0

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Set PIPE_CAP_LOAD_CONSTBUF
Alyssa Rosenzweig [Sat, 18 Mar 2023 18:45:08 +0000 (14:45 -0400)]
asahi: Set PIPE_CAP_LOAD_CONSTBUF

The CAP is a bit of a misnomer, what it really does is relax the alignment
requirements for UBO packing. It should work fine and save us some memory.
Noticed while debugging piglit fails.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi/decode: Print VDM barriers
Alyssa Rosenzweig [Sat, 18 Mar 2023 18:33:10 +0000 (14:33 -0400)]
asahi/decode: Print VDM barriers

Instead of just decoding silently.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi/decode: Remove agxdecode_dump_bo
Alyssa Rosenzweig [Sat, 18 Mar 2023 18:30:31 +0000 (14:30 -0400)]
asahi/decode: Remove agxdecode_dump_bo

Now that we have proper parsing this is more of a nuissance than not.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoagx: Add helper for calculating occupancy
Alyssa Rosenzweig [Sat, 11 Mar 2023 20:39:09 +0000 (15:39 -0500)]
agx: Add helper for calculating occupancy

Add information about the relationship between program register usage and
program occupancy (the maximum number of threads that may execute concurrently
on a single shader core). This table is derived from studying the
maxTotalThreadsPerThreadgroup property in Metal while varying the register
usage, something I blogged about a few years back. It's probably not 100%
accurate and it hasn't been tested against hardware, but it matters "only" for
performance (not correctness) so I'm not super stressed about the details.

In the (near) future, RA will be able to make use of this information to know
exactly when it can use more registers without hurting performance. In the
present, it's just used for better shader-db statistics.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoagx: Set loads_varying accurately
Alyssa Rosenzweig [Tue, 7 Mar 2023 04:33:11 +0000 (23:33 -0500)]
agx: Set loads_varying accurately

Instead of just always mashing to true. Should be better for depth-only passes.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Add perf debug for shader variants
Alyssa Rosenzweig [Sat, 11 Mar 2023 22:48:56 +0000 (17:48 -0500)]
asahi: Add perf debug for shader variants

Compiling this can cause jank. This is still an issue in Quake3. There is a way
to solve it but it's rather involved and certainly not this weekend's project.
Better perf debugging on the other hand apparently is ^_^

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Add perf debug for generate_mipmap
Alyssa Rosenzweig [Sat, 11 Mar 2023 21:46:45 +0000 (16:46 -0500)]
asahi: Add perf debug for generate_mipmap

The current implementation leaves a lot of perf on the table, so call it out on
ASAHI_MESA_DEBUG=perf to help debugging perf problems, especially if this
ever happens in a real application (i.e. not a benchmark).

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoagx: Don't destroy usub_sat with constant
Alyssa Rosenzweig [Sun, 12 Mar 2023 03:16:52 +0000 (22:16 -0500)]
agx: Don't destroy usub_sat with constant

Fixes KHR-GLES31.core.shader_storage_buffer_object.advanced-unsizedArrayLength-cs-std430-vec-pad

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoagx: Don't allow uniform source to local_atomic
Alyssa Rosenzweig [Sun, 12 Mar 2023 03:07:18 +0000 (22:07 -0500)]
agx: Don't allow uniform source to local_atomic

Fixes KHR-GLES31.core.compute_shader.atomic-case3

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoagx: Constify agx_{read,write}_registers
Alyssa Rosenzweig [Sun, 5 Mar 2023 03:17:29 +0000 (22:17 -0500)]
agx: Constify agx_{read,write}_registers

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoagx: Assert that we don't overflow registers
Alyssa Rosenzweig [Tue, 7 Mar 2023 03:59:23 +0000 (22:59 -0500)]
agx: Assert that we don't overflow registers

This will become particularly important when we bound to smaller register files.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoagx: DCE even with noopt
Alyssa Rosenzweig [Sat, 4 Mar 2023 22:13:30 +0000 (17:13 -0500)]
agx: DCE even with noopt

To simplify live range splitting, RA will soon assume that DCE has run (removing
extraneous vectors). So run DCE even when otherwise disabling backend
optimizations. AGX_MESA_DEBUG=noopt is still useful for disabling instruction
combining, which is the more-likely-to-be-buggy pass anyway.

This also fixes IR not being printed with noopt.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agoasahi: Support more renderable formats
Alyssa Rosenzweig [Sun, 12 Mar 2023 00:47:27 +0000 (19:47 -0500)]
asahi: Support more renderable formats

Fixes KHR-GLES3.copy_tex_image_conversions.forbidden.*

Arguably working around a mesa/st issue but more format support is good for
compatibility and performance anyway.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Janne Grunau <j@jannau.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22353>

19 months agovenus/docs: sync to latest venus supported extensions
Yiwei Zhang [Fri, 31 Mar 2023 06:53:58 +0000 (06:53 +0000)]
venus/docs: sync to latest venus supported extensions

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22243>

19 months agovenus: add VK_EXT_rasterization_order_attachment_access support
Yiwei Zhang [Fri, 31 Mar 2023 06:26:01 +0000 (06:26 +0000)]
venus: add VK_EXT_rasterization_order_attachment_access support

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22243>

19 months agovenus: add VK_EXT_load_store_op_none support
Yiwei Zhang [Fri, 31 Mar 2023 08:28:04 +0000 (08:28 +0000)]
venus: add VK_EXT_load_store_op_none support

There's no feature/properties structs associated with this extension.

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22243>

19 months agovenus: sync latest protocol for layering extensions
Yiwei Zhang [Fri, 31 Mar 2023 06:16:36 +0000 (06:16 +0000)]
venus: sync latest protocol for layering extensions

- VK_EXT_load_store_op_none
- VK_EXT_rasterization_order_attachment_access

Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22243>

19 months agoradeonsi/vcn: optimize bitstream buffer resize logic
Sajeesh Sidharthan [Wed, 5 Apr 2023 08:58:59 +0000 (14:28 +0530)]
radeonsi/vcn: optimize bitstream buffer resize logic

bitstream buffer is unmapped, resized and mapped again if new size
is greater than the current bitstream buffer size. This will be done
for each input buffer. This patch will avoid that and do resize
only once irrespective of number of input buffers. With the new logic,
total size is calculated first and call unmap, resize and map only once.

Signed-off-by: Sajeesh Sidharthan <sajeesh.sidharthan@amd.com>
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Reviewed-by: Veerabadhran Gopalakrishnan <Veerabadhran.Gopalakrishnan@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22308>

19 months agonir/print: Don't print sampler_index for txf
Alyssa Rosenzweig [Thu, 30 Mar 2023 18:08:38 +0000 (14:08 -0400)]
nir/print: Don't print sampler_index for txf

NIR's docs for sampler_index say

    The following operations do not require a sampler and, as such, this
    field should be ignored:
       - nir_texop_txf
       - nir_texop_txf_ms
       - nir_texop_txs
       - nir_texop_query_levels
       - nir_texop_texture_samples
       - nir_texop_samples_identical

Contrary to this documentation, we were still printing the sampler_index anyway,
even though the value is formally undefined. This was helpful for
PIPE_CAP_TEXTURE_BUFFER_SAMPLER drivers that (despite the NIR docs) respected the
sampler_index anyway. There are no longer any such drivers, so we should stop
printing sampler_index for txf to avoid confusion (and noise).

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22223>

19 months agodocs/gallium: Note samplers are not used for txf
Alyssa Rosenzweig [Fri, 31 Mar 2023 01:02:45 +0000 (21:02 -0400)]
docs/gallium: Note samplers are not used for txf

Now that PIPE_CAP_TEXTURE_BUFFER_SAMPLER is gone, txf does not require samplers
for any texture on any Gallium driver. NIR already requires drivers to ignore
sampler_index for non-sampler operation (mainly txf), and nowadays all Gallium
drivers ingest NIR. So, document that samplers aren't bound for txf (etc) as
part of the Gallium frontend-driver contract.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Suggested-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22223>

19 months agogallium: Remove PIPE_CAP_TEXTURE_BUFFER_SAMPLER
Alyssa Rosenzweig [Thu, 30 Mar 2023 17:24:16 +0000 (13:24 -0400)]
gallium: Remove PIPE_CAP_TEXTURE_BUFFER_SAMPLER

No more users. It was already not respected by rusticl so you couldn't set it if
you wanted OpenCL support. I regret introducing the CAP in the first place, and
no more drivers should use it.

Reverts d5d3f77e4ac ("gallium: Add new cap PIPE_CAP_TEXTURE_BUFFER_SAMPLER").

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22223>

19 months agopanfrost: Unset TEXTURE_BUFFER_SAMPLERS
Alyssa Rosenzweig [Thu, 30 Mar 2023 17:21:14 +0000 (13:21 -0400)]
panfrost: Unset TEXTURE_BUFFER_SAMPLERS

We no longer need this CAP, as we can easily synthesize our own internal sampler
for this case. Gallium doesn't need to know about this quirk of our hardware.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Italo Nicola <italonicola@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22223>

19 months agopan/{mdg,bi}: Always use sampler 0 for txf
Alyssa Rosenzweig [Thu, 30 Mar 2023 18:00:46 +0000 (14:00 -0400)]
pan/{mdg,bi}: Always use sampler 0 for txf

Now that we upload workaround samplers for txf, sampler 0 is guaranteed to be
valid but other samplers are not. So ignore whatever the current sampler_index
value is (it's formally undefined in NIR) and use 0, which we know is valid. We
already do this on Valhall for OpenCL, just need to generalize for Midgard and
Bifrost.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Italo Nicola <italonicola@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22223>

19 months agopanfrost: Always upload a workaround sampler
Alyssa Rosenzweig [Thu, 30 Mar 2023 17:55:14 +0000 (13:55 -0400)]
panfrost: Always upload a workaround sampler

The hardware requires a valid sampler even for texelFetch (txf), even though its
contents are ignored. We'd rather not pass on this requirement to the frontends,
so we should handle it by uploading our own workaround sampler in the case when
no sampler is already present. We already do this on Valhall (for rusticl), so
we just need to port the same workaround back to Midgard/Bifrost.

Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Italo Nicola <italonicola@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22223>

19 months agozink: don't try copying multiple results for conditional render copy
Mike Blumenkrantz [Thu, 6 Apr 2023 20:12:46 +0000 (16:12 -0400)]
zink: don't try copying multiple results for conditional render copy

conditional render is only a single result, so multiple results need
to first be aggregated

fixes #8798

cc: mesa-stable

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22345>

19 months agonir/tests: Port almost all loop_analyze tests to new macro-based infastructure
Ian Romanick [Fri, 17 Feb 2023 20:20:09 +0000 (12:20 -0800)]
nir/tests: Port almost all loop_analyze tests to new macro-based infastructure

The one test that remains would have an automatically generated name
that would conflict with another test. This test is also a little
special (per the comment in the test), so it's probably best to leave it
separate anyway.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3445>

19 months agonir/loop_analyze: Determine iteration counts for more kinds of loops
Yevhenii Kolesnikov [Fri, 17 Jan 2020 11:01:01 +0000 (13:01 +0200)]
nir/loop_analyze: Determine iteration counts for more kinds of loops

If loop iterator is incremented with something other than regular
addition, it would be more error prone to calculate the number of
iterations theoretically. What we can do instead, is try to emulate the
loop, and determine the number of iterations empirically.

These operations are covered:
 - imul
 - fmul
 - ishl
 - ishr
 - ushr

Also add unit tests for loop unrollment.

Improves performance of Aztec Ruins (sixonix
gfxbench5.aztec_ruins_vk_high) by -1.28042% +/- 0.498555% (N=5) on Intel
Arc A770.

v2 (idr): Rebase on 3 years. :( Use nir_phi_instr_add_src in the test
cases.

v3 (idr): Use try_eval_const_alu in to evaluate loop termination
condition in get_iteration_empirical. Also restructure the loop
slightly. This fixed off by one iteration errors in "inverted" loop
tests (e.g., nir_loop_analyze_test.ushr_ieq_known_count_invert_31).

v4 (idr): Use try_eval_const_alu in to evaluate induction variable
update in get_iteration_empirical. This fixes non-commutative update
operations (e.g., shifts) when the induction varible is not the first
source. This fixes the unit test
nir_loop_analyze_test.ishl_rev_ieq_infinite_loop_unknown_count.

v5 (idr): Fix _type parameter for fadd and fadd_rev loop unroll
tests. Hopefully that fixes the failure on s390x. Temporarily disable
fmul. This works-around the revealed problem in
glsl-fs-loop-unroll-mul-fp64, and there were no shader-db or fossil-db
changes.

v6 (idr): Plumb max_unroll_iterations into get_iteration_empirical. I
was going to do this, but I forgot. Suggested by Tim.

v7 (idr): Disable fadd tests on s390x. They fail because S390 is weird.

Almost all of the shaders affected (OpenGL or Vulkan) are from gfxbench
or geekbench. A couple shaders in Deus Ex (OpenGL), Dirt Rally (OpenGL),
Octopath Traveler (Vulkan), and Rise of the Tomb Raider (Vulkan) are
helped.

The lost / gained shaders in OpenGL are an Aztec Ruins shader that goes
from SIMD16 to SIMD8. The spills / fills affected are in a single Aztec
Ruins (Vulkan) compute shader.

shader-db results:

Skylake, Ice Lake, and Tiger Lake had similar results. (Tiger Lake shown)
total loops in shared programs: 5514 -> 5470 (-0.80%)
loops in affected programs: 62 -> 18 (-70.97%)
helped: 37 / HURT: 0

LOST:   2
GAINED: 2

Haswell and Broadwell had similar results. (Broadwell shown)
total loops in shared programs: 5346 -> 5298 (-0.90%)
loops in affected programs: 66 -> 18 (-72.73%)
helped: 39 / HURT: 0

fossil-db results:

Skylake, Ice Lake, and Tiger Lake had similar results. (Tiger Lake shown)
Instructions in all programs: 157374679 -> 157397421 (+0.0%)
Instructions hurt: 28

SENDs in all programs: 7463800 -> 7467639 (+0.1%)
SENDs hurt: 28

Loops in all programs: 38980 -> 38950 (-0.1%)
Loops helped: 28

Cycles in all programs: 7559486451 -> 7557455384 (-0.0%)
Cycles helped: 28

Spills in all programs: 11405 -> 11403 (-0.0%)
Spills helped: 1

Fills in all programs: 19578 -> 19588 (+0.1%)
Fills hurt: 1

Lost: 1

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/3445>

19 months agonir/loop_analyze: Track induction variables incremented by more operations
Yevhenii Kolesnikov [Fri, 17 Jan 2020 11:01:01 +0000 (13:01 +0200)]
nir/loop_analyze: Track induction variables incremented by more operations

These operations are covered:

 - imul
 - fmul
 - ishl
 - ishr
 - ushr

The only cases that can be currently affected are those where the
calculated loop-trip count would be zero.

v2 (idr): Split out from original commit. Rebase on lots of other work.

v3 (idr): Move operand size assertion. This code only cares that the
operands have the same size for the iadd and fadd cases. In other
cases, such as shifts, the sizes may not match. Fixes assertion
failures in
tests/spec/arb_gpu_shader_int64/glsl-fs-loop-unroll-ishl-int64.shader_test.

No shader-db or fossil-db changes on any Intel platform.

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/3445>

19 months agonir/loop_analyze: Use try_eval_const_alu and induction variable basis info
Ian Romanick [Tue, 14 Feb 2023 01:33:29 +0000 (17:33 -0800)]
nir/loop_analyze: Use try_eval_const_alu and induction variable basis info

This dramatically simplifies will_break_on_first_iteration, and, much
more importantly, makes it significantly more flexible. It is now
possible to handle loops with more complex exit condition and other
kinds of increment operations.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3445>

19 months agonir/loop_analyze: Change invert_cond instead of changing the condition
Ian Romanick [Wed, 15 Feb 2023 00:12:23 +0000 (16:12 -0800)]
nir/loop_analyze: Change invert_cond instead of changing the condition

This ensures that scenarios like
nir_loop_analyze_test.iadd_inot_ilt_rev_known_count_5 don't regress in
the next commit. It also means we don't change float comparisons. These
are probably fine... but it still made me a little uneasy.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3445>

19 months agonir/loop_analyze: Track induction variable basis information
Ian Romanick [Tue, 14 Feb 2023 01:30:16 +0000 (17:30 -0800)]
nir/loop_analyze: Track induction variable basis information

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3445>

19 months agonir/loop_analyze: Add a function to evaluate an ALU as constant
Ian Romanick [Tue, 14 Feb 2023 01:23:14 +0000 (17:23 -0800)]
nir/loop_analyze: Add a function to evaluate an ALU as constant

...with a substitution. This function is largely a copy-and-paste of
try_fold_alu (nir_opt_constant_folding.c), and an argument could be made
that this function belongs in that file.

v2: Some changes were mistakenly squashed in to "nir/loop_analyze: Use
try_eval_const_alu and induction variable basis info" that should have
been here.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3445>

19 months agonir/tests: Add many loop analysis tests for induction variables modified by imul
Ian Romanick [Thu, 2 Feb 2023 00:18:04 +0000 (16:18 -0800)]
nir/tests: Add many loop analysis tests for induction variables modified by imul

Loop analysis doesn't currently treat values updated by multiplication
as induction variables. Future patches will change this.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3445>

19 months agonir/tests: Add more loop analysis tests for induction vars updated by shifts
Ian Romanick [Thu, 2 Feb 2023 20:44:52 +0000 (12:44 -0800)]
nir/tests: Add more loop analysis tests for induction vars updated by shifts

These reverse the order of the comparison (e.g., -2 >= i vs i >= -2). I
split this into a separate commit because the previous commit was so
large.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3445>

19 months agonir/tests: Add many loop analysis tests for induction vars updated by shifts
Ian Romanick [Wed, 1 Feb 2023 23:18:05 +0000 (15:18 -0800)]
nir/tests: Add many loop analysis tests for induction vars updated by shifts

Loop analysis doesn't currently treat values updated by shifts as
induction variables. Future patches will change this.

v2: Don't use the contradiction ilt(x, INT_MIN).

v3: Delete some errant code in UNKNOWN_COUNT_TEST. Noticed by Tim.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3445>

19 months agoradeonsi/vcn: set bitstream buffer size to encoded bitstream size
Sajeesh Sidharthan [Wed, 15 Mar 2023 04:24:58 +0000 (09:54 +0530)]
radeonsi/vcn: set bitstream buffer size to encoded bitstream size

initial bitstream size was set to width * height * 2 which is
larger than yuv size. set initial bitstream size to encoded
bitstream size approximately to optimize memory consumption.
This is just an initial size setting, it will get resized later
if it's not big enough. As a result of this change, we don't need to
allocate super big size at the every beginning. Only allocate
big size when needed in order to save some memory

Signed-off-by: Sajeesh Sidharthan <sajeesh.sidharthan@amd.com>
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Acked-by: Veerabadhran Gopalakrishnan <Veerabadhran.Gopalakrishnan@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21918>

19 months agodzn: Fix bindless descriptor sets with multiple dynamic buffers that need custom...
Jesse Natalie [Wed, 5 Apr 2023 16:53:47 +0000 (09:53 -0700)]
dzn: Fix bindless descriptor sets with multiple dynamic buffers that need custom descriptors

Fixes: 5d2b4ee4 ("dzn: Allocate descriptor sets in buffers for bindless mode")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agodzn: Batch command lists together
Jesse Natalie [Fri, 31 Mar 2023 17:18:19 +0000 (10:18 -0700)]
dzn: Batch command lists together

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agodzn: Don't do initial-layout barriers for simultaneous-access resources
Jesse Natalie [Fri, 31 Mar 2023 20:43:48 +0000 (13:43 -0700)]
dzn: Don't do initial-layout barriers for simultaneous-access resources

Fixes: 4daeac01 ("dzn: Enhanced barriers fixes/workarounds")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agodzn: Attempt to force depth write states for depth access in LAYOUT_GENERIC
Jesse Natalie [Thu, 30 Mar 2023 20:39:41 +0000 (13:39 -0700)]
dzn: Attempt to force depth write states for depth access in LAYOUT_GENERIC

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agodzn: Ensure buffer offsets are aligned
Jesse Natalie [Thu, 30 Mar 2023 20:38:37 +0000 (13:38 -0700)]
dzn: Ensure buffer offsets are aligned

If the app passes us unaligned buffer offsets, we need to align them
down to the nearest aligned offset, and then put the difference into
the descriptor set buffer.

Fixes: 8bd5fbf8 ("dzn: Bind buffers for bindless descriptor sets")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agodzn: Don't use write-combine memory for cache-coherent UMA
Jesse Natalie [Mon, 27 Mar 2023 18:09:57 +0000 (11:09 -0700)]
dzn: Don't use write-combine memory for cache-coherent UMA

Cache coherent UMA implies that the GPU is reading data through the
CPU caches. Using write-combined CPU pages for such a system would
be bad, since the GPU would then be reading uncached data. One
example of such a system is WARP. This significantly improves WARP's
performance for some apps (including the CTS).

Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agodzn: Ensure pipeline variants are used for dynamic stencil masks
Jesse Natalie [Mon, 27 Mar 2023 18:09:38 +0000 (11:09 -0700)]
dzn: Ensure pipeline variants are used for dynamic stencil masks

Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agodzn: Align descriptor sets in the bindless buffer
Jesse Natalie [Mon, 27 Mar 2023 18:08:50 +0000 (11:08 -0700)]
dzn: Align descriptor sets in the bindless buffer

Fixes: 5d2b4ee4 ("dzn: Allocate descriptor sets in buffers for bindless mode")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agodzn: Report some more caps correctly that are supported
Jesse Natalie [Thu, 23 Mar 2023 15:41:22 +0000 (08:41 -0700)]
dzn: Report some more caps correctly that are supported

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agodzn: Raise max number of descriptor sets to 8
Jesse Natalie [Thu, 23 Mar 2023 15:35:41 +0000 (08:35 -0700)]
dzn: Raise max number of descriptor sets to 8

DOOM Eternal just assumes you support at least 5, which caused corruption
due to overrunning arrays. We can just bump this up. 8 should work with
and without bindless.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agodzn: Fix SRV barrier state on compute command lists
Jesse Natalie [Thu, 23 Mar 2023 15:34:39 +0000 (08:34 -0700)]
dzn: Fix SRV barrier state on compute command lists

Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agodzn: Add a driconf option for enabling subgroup ops in VS/GS
Jesse Natalie [Thu, 30 Mar 2023 21:44:11 +0000 (14:44 -0700)]
dzn: Add a driconf option for enabling subgroup ops in VS/GS

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agodzn: Add a driconf entry for enabling 8bit loads and stores
Jesse Natalie [Fri, 27 Jan 2023 21:05:33 +0000 (13:05 -0800)]
dzn: Add a driconf entry for enabling 8bit loads and stores

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agospirv2dxil: Add some more supported caps
Jesse Natalie [Thu, 23 Mar 2023 15:36:22 +0000 (08:36 -0700)]
spirv2dxil: Add some more supported caps

8-bit loads and stores work via lowering, but they do work

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agomicrosoft/compiler: Fix large shifts
Jesse Natalie [Tue, 4 Apr 2023 20:41:10 +0000 (13:41 -0700)]
microsoft/compiler: Fix large shifts

Unlike DXBC, DXIL's shift instructions don't have the implicit behavior
that they only take the 5 bits. This is observable if you try to have
DXC do a shift of a dynamic value, e.g. a constant buffer value, where
the compiler inserts the appropriate 'and' op. We need to do the same.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agomicrosoft/compiler: Assign 1D wave IDs based on local thread ID
Jesse Natalie [Fri, 31 Mar 2023 17:25:00 +0000 (10:25 -0700)]
microsoft/compiler: Assign 1D wave IDs based on local thread ID

Fixes corruption/flickering seen in DOOM Eternal's decals/lighting.
It seems the shader has an implicit assumption about this property.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agomicrosoft/compiler: Fix barrier for wave ID computation
Jesse Natalie [Thu, 30 Mar 2023 20:24:14 +0000 (13:24 -0700)]
microsoft/compiler: Fix barrier for wave ID computation

Fixes: 2f8a8b59 ("microsoft/compiler: Add lowering passes for basic subgroup vars")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agomicrosoft/compiler: Fix 8-bit loads and stores when supporting 16-bit DXIL
Jesse Natalie [Thu, 23 Mar 2023 15:30:58 +0000 (08:30 -0700)]
microsoft/compiler: Fix 8-bit loads and stores when supporting 16-bit DXIL

Shifts should always use 32bit shift values, and when lowering to
masked, we need to use 32-bit atomics. That means that we should also
treat 24bit stores as a single masked op rather than one 16bit unmasked
and one 8bit masked.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22225>

19 months agoglx: Fix error handling yet again in CreateContextAttribs
Adam Jackson [Wed, 21 Jul 2021 16:02:35 +0000 (12:02 -0400)]
glx: Fix error handling yet again in CreateContextAttribs

Unlike the legacy CreateContext path, we would try to send the
GLXCreateContextAttribs request regardless of whether we'd successfully
created the client context state. And there's not a lot on the server
side to go wrong besides BadAlloc, so if the request succeeded but
the client side didn't we'd need to destroy the server context and
synthesize an X error. Since that itself involves more X protocol it's
tricky to get the request number right in the error, and tests and apps
can notice when you get it wrong.

Since we have now fixed client-side validation to generate the right
errors at the right times, this patch does something simpler, we match
CreateContext and fail early if the client-side setup fails. Now there's
no question of what request number to use, because we haven't sent any
protocol, the error is for the request as if it'd been sent.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4763
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12006>

19 months agoglx: Disable the indirect fallback in CreateContextAttribs
Adam Jackson [Fri, 23 Jul 2021 20:13:45 +0000 (16:13 -0400)]
glx: Disable the indirect fallback in CreateContextAttribs

If your app cares enough to use CreateContextAttribs it's probably not
going to be happy with the pre-GL-1.5 indirect experience.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12006>

19 months agoglx/dri: Fix error generation for invalid GLX_RENDER_TYPE
Adam Jackson [Fri, 23 Jul 2021 22:14:33 +0000 (18:14 -0400)]
glx/dri: Fix error generation for invalid GLX_RENDER_TYPE

This needs to throw BadValue.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12006>

19 months agodri: Validate more of the context version in validate_context_version
Adam Jackson [Thu, 22 Jul 2021 20:28:31 +0000 (16:28 -0400)]
dri: Validate more of the context version in validate_context_version

There's two kinds of "bad version" you might encounter here, either the
combination does not name a defined version (like 1.7) or it names
something the driver can't do (like asking r300 to do 4.0). EGL does not
distinguish these cases, but GLX calls them BadMatch and GLXBadFBConfig
respectively.

Since api_mask is the set of driver supported APIs, and we can only
support defined APIs, don't check it early in driCreateContextAttribs,
just let it fall out from validate_context_version.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12006>

19 months agoglx/dri: Use X/GLX error codes for our create_context_attribs
Adam Jackson [Wed, 21 Jul 2021 22:05:12 +0000 (18:05 -0400)]
glx/dri: Use X/GLX error codes for our create_context_attribs

This has no functional change because everyone calling this is
discarding the error code, because we're relying on the server to
generate the right thing for us. But we create the direct context first
and the server isn't going to enforce everything we want it to
(supported GL versions for example). Convert out from DRI error codes to
X/GLX error codes so we can fail the right way on the client side. We're
still throwing the error away in all of the callers but that'll change
shortly.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12006>

19 months agointel/fs: White space fixes
Ian Romanick [Fri, 31 Mar 2023 22:47:48 +0000 (15:47 -0700)]
intel/fs: White space fixes

Trivial

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22299>

19 months agointel/fs: Preserve meta data more often in brw_nir_move_interpolation_to_top
Ian Romanick [Tue, 21 Mar 2023 03:57:47 +0000 (20:57 -0700)]
intel/fs: Preserve meta data more often in brw_nir_move_interpolation_to_top

This pass rarely makes any changes, so work a little harder to preserve
more meta data.

On my Ice Lake laptop (using a locked CPU speed and other measures to
prevent thermal throttling, etc.) using a debugoptimized build, improves
performance of Vulkan CTS "deqp-vk --deqp-case='dEQP-VK.*spir*'" by
-0.2% ± 0.1% (n = 5, pooled s = 0.431885).

v2: Add some parenthesis. Suggested by Lionel.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22299>

19 months agointel/fs: Linked list micro optimizations in brw_nir_move_interpolation_to_top
Ian Romanick [Tue, 21 Mar 2023 03:57:47 +0000 (20:57 -0700)]
intel/fs: Linked list micro optimizations in brw_nir_move_interpolation_to_top

Two linked list management changes:

- Use the list head sentinel as the initial cursor. It is, after all, a
  proper node in the list.

- Iterate the list of blocks starting with the second block instead of
  skipping the first block in the loop.

On my Ice Lake laptop (using a locked CPU speed and other measures to
prevent thermal throttling, etc.) using a release build, improves
performance of compiling shaders from batman_arkham_city_goty.foz by
-0.24% ± 0.09% (n = 5, pooled s = 0.324106).

v2: Use nir_cursor instead of direct list manipultion. Suggested by
Lionel.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22299>

19 months agointel/compiler: Micro optimize regions_overlap
Ian Romanick [Tue, 14 Mar 2023 02:47:07 +0000 (19:47 -0700)]
intel/compiler: Micro optimize regions_overlap

On my Ice Lake laptop (using a locked CPU speed and other measures to
prevent thermal throttling, etc.) using a release build, improves
performance of compiling shaders from batman_arkham_city_goty.foz by
-1.09% ± 0.084% (n = 5, pooled s = 0.354471)

Reduces the size of a release build by 26k.

   text    data     bss     dec     hex filename
23163641 400720  231360 23795721 16b1809 before/lib64/dri/iris_dri.so
23137264 400720  231360 23769344 16ab100 after/lib64/dri/iris_dri.so

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22299>