AlexIndustrial/mesa
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
164,856 Commits 1 Branch 0 Tags
33e3418cfe2db9983b71eddd97d4d7da44e5ef76
Commit Graph

164856 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Alyssa Rosenzweig 33e3418cfe agx: Consider "stop" a control flow instruction 
...and therefore it needs to be after a "logical end". This means that
"after_block_logical" will do the right thing for the last block.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig f6aa43cf42 agx: Optimize waits locally 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig a01680b979 agx: Remove logical_end later 
So we can use after_block_logical in the wait insertion pass.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig 73ac73308b agx: Validate widths of vectors 
Check the invariant that the widths of vectors in the IR are consistent, by
checking that write registers and read registers match up between the writers
and readers respectively.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig 6685dba75e agx: Add agx_read_registers helper 
To be used for inserting waits post-RA.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig e6631ba5af agx: Compact st_tile argument per mask 
Otherwise the number of read registers won't match the vector we input, which
will trigger validation errors.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig 545a3eb601 agx: Insert waits post-RA 
This is the first step towards reducing stalling.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig 463744e4f9 agx: Pack texture scoreboard slots 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig 01f948ee13 agx: Pack wait instructions 
For different scoreboard slots.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig 640afb33b9 agx: Remove unused idiv const func 
This was used for instancing, but has been unused since 8dcf7648f1 ("agx: Lower VBOs in NIR")

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig 44925a142e agx: Use metadata for VS varying linking 
Rather than variables. This gets rid of all backend nir_variable use.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig 617f2f7a02 agx: Don't use nir_variable when gathering flat varyings 
Walk the IR instead. This happens when preprocessing so it doesn't really
matter, but it complicates the nir_variable audit.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig d00a43f682 agx: Hash agx_instr faster 
Prior to this change, agx_opt_cse is our most expensive backend pass, due to the
time spent hashing instructions. hash_instr was calling into XXH32 a massive
number of times, often to hash only a single bit. It's much faster to hash
entire blocks of memory at a time. Optimize to do just that.

With this change, agx_opt_cse is now cheaper than instruction selection as
it should be.

No shader-db changes (except CPU time decrease).

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig f44afe766f agx: Use texture write mask 
We do need to use undefs instead of zeroes in this internal collect. While this
vector gets copypropped out, it'd cause us to fail compilation if noopt is on.
Fix that.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig 7284e4967c agx: Note that textures clobber even masked 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig ddbec45b6f agx: Plumb in store instruction 
This will be used for compute kernels (and transform feedback) in the (near)
future. For now, let's get the opcode plumbed in the backend to reduce some of
the rebase pain.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig 6b645f12ad asahi/lower_tilebuffer: Avoid loading sample mask 
If a render target isn't written to, we don't use the sample mask. Avoid
generating the intermediate instructions, common with gl_FragColor. It will get
DCE'd, but this means less work for DCE, which should help for shader jank since
this pass gets called per-variant.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig bb677c7fc8 asahi: Remove outdated comment 
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig b612690e38 asahi: Implement linear 2D array textures 
These are useful for layered staging resources. Tested by forcing linear
textures and running dEQP-GLES3.functional.texture.format.sized.2d_array.*

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig 0bc70a0074 asahi: Correct acceleration_buffer for layered images 
The old calculation was wrong if layer != 0.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:49:22 -05:00
Alyssa Rosenzweig 17d4486c6a asahi: Add XML for linear 2D arrays 
These look a bit like compressed images, and elucidate one of the common fields.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:48:13 -05:00
Alyssa Rosenzweig 48c9a9676c asahi: Add XML required for vertex shader side effects 
Basically for rasterizer discard. We'll use these in a moment to implement
transform feedback.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:48:13 -05:00
Alyssa Rosenzweig 6bda0f2a70 asahi: Dump uniforms when decoding 
These often have addresses in them.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:48:13 -05:00
Alyssa Rosenzweig d31a9cf3ae asahi: Check box implement NV_conditional_render 
Use the freedreno lowering. It'll be slow but I don't know of any apps that
actually use this and it's required for GL 3.0.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:48:13 -05:00
Alyssa Rosenzweig 21a35b58ba ail: Add compression_layer_stride_B field 
So we can index into the compression buffer of layered images.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:48:13 -05:00
Alyssa Rosenzweig d703c2887d ail: Support linear 2D arrays 
These are straightforward.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:48:13 -05:00
Alyssa Rosenzweig 35494ea83e docs/features: Add more missed asahi 
Should already be there.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20446>
 2023-01-05 11:48:13 -05:00
Rob Clark 03beb32478 freedreno/decode: Deal with suballocated buffers 
We can end up logging both the buffer that the toplevel cmdstream is
allocated, as well as the sub-allocated part of that buffer.  Possibly
the kernel could do better about this, but to avoid undecodeable
cmdstream dumps and devcores, detect this case and deal with it.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20496>
 2023-01-05 15:44:39 +00:00
Rob Clark 8e437a2203 freedreno: Limit max draws per batch 
Mostly so that piglit drawoverhead doesn't OoM itself with absurd # of
draws.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20496>
 2023-01-05 15:44:39 +00:00
Rob Clark 8d44525feb freedreno/drm: Add nr_suballoc_bos size check 
We also want to ensure we don't hit the limit of max suballoc BOs.
Piglit drawoverhead would manage to hit this.

Fixes: 4861067689 ("freedreno/drm: Add sub-allocator")
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20496>
 2023-01-05 15:44:39 +00:00
Rob Clark 912c91e383 freedreno: Fix FD_MESA_DEBUG=nohw 
We need to construct a dummy fence in this case.

Fixes: 48b5164356 ("freedreno/drm: Return fence from submit flush")
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20496>
 2023-01-05 15:44:39 +00:00
Rob Clark 5c3d5f14ac freedreno/drm: Mark ring heap buffers for dumping 
Ensure that these get FD_RELOC_DUMP flag set so cmdstream dumps and
devcore's capture them.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20496>
 2023-01-05 15:44:39 +00:00
Bas Nieuwenhuizen 0d5570bb47 radv: Always compile accel structure shaders on demand. 
Some CTS tests enable all extensions ... , which combined with having
no shader cache on some platforms results in some CTS tests timing
out (in particular tests recreating the device all the time).

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20422>
 2023-01-05 14:45:47 +00:00
Bas Nieuwenhuizen ef37407971 radv: Support on demand compilation of fmask shaders. 
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20422>
 2023-01-05 14:45:47 +00:00
Pavel Ondračka 32925bf708 r300: lower fsign in nir 
Nothing significant in shader-db on RV530:
total instructions in shared programs: 134963 -> 134957 (<.01%)
instructions in affected programs: 1108 -> 1102 (-0.54%)
helped: 7
HURT: 1
total temps in shared programs: 17153 -> 17154 (<.01%)
temps in affected programs: 38 -> 39 (2.63%)
helped: 2
HURT: 3

Just some fluctuations from pair scheduling due to different code order.

Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Acked-by: Filip Gawin <filip@gawin.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20208>
 2023-01-05 12:01:32 +00:00
Pavel Ondračka 9af7f7d156 r300: move some ftrunc lowering to nir 
There are still some ftruncs left as most of them originates in
nir_lower_int_to_float and that is currently called after nir_opt_algebraic
in ntt.

No change in shader-db.

Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Acked-by: Filip Gawin <filip@gawin.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20208>
 2023-01-05 12:01:32 +00:00
Pavel Ondračka 31f678626f r300: use nir lowering for ffloor 
Negligible amount of instructions saved on RV530:
total instructions in shared programs: 134970 -> 134963 (<.01%)
instructions in affected programs: 2273 -> 2266 (-0.31%)
helped: 9
HURT: 1

The one hurt shader is when we fail to recognize the x - ffract(x)
pattern and skip the don't emit ftrunc optimization as implemented
in the previous patch due to some non-trivial swizzles going on.

Signed-off-by: Pavel Ondračka <pave.ondracka@gmail.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Acked-by: Filip Gawin <filip@gawin.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20208>
 2023-01-05 12:01:32 +00:00
Pavel Ondračka a93bc6afc4 nir: check for x - ffract(x) patterns when lowering f2i32 
We already skip emitting ftrunc in nir_lower_int_to_float when there is
ffloor, fround or any other integer-making opcode preceding f2i32. However
if lower_ffloor is set for driver that doesn't support integers, the lowered
x - ffract(x) patterns would not be recognized and extra ftruct would be
emitted, doing unnecessary rounding.

This optimization only works if there is no non-trivial swizzling used for
the fadd, fneg and ffract involved, which seems to be 99% of the cases according
to my testing.

This is needed to enable nir ffloor lowering on r300 driver without regressions.

I'm not sure if this helps anybody else, the only hardware which sets
lower_ffloor and converts ints to floats (and can't do trunc) are some old
etnaviv cards, so maybe it will help there a bit.

Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20208>
 2023-01-05 12:01:32 +00:00
Pierre-Eric Pelloux-Prayer b111e8aa49 egl: handle NULL loaderPrivate in dri_is_thread_safe 
This can happen when a context is created from
loader_dri3_blit_context_get and glthread is enabled.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7949
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7960
Cc: mesa-stable

Acked-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20479>
 2023-01-05 10:35:58 +00:00
Samuel Pitoiset 5449b6927a radv: use vk_dynamic_graphics_state for the color blend state 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20314>
 2023-01-05 10:07:01 +00:00
Samuel Pitoiset e381ac7aaf radv: rework some color blend state fields for vk_dynamic_graphics_state 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20314>
 2023-01-05 10:07:01 +00:00
Samuel Pitoiset 29ca23dd93 radv: use vk_dynamic_graphics_state for the multisample state 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20314>
 2023-01-05 10:07:01 +00:00
Samuel Pitoiset ed5dc76ea7 radv: use vk_dynamic_graphics_state for the depth stencil state 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20314>
 2023-01-05 10:07:01 +00:00
Samuel Pitoiset a72fcfd40b radv: use vk_dynamic_graphics_state for the rasterization state 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20314>
 2023-01-05 10:07:01 +00:00
Samuel Pitoiset 3bc248e564 radv: use vk_dynamic_graphics_state for the discard rectangles state 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20314>
 2023-01-05 10:07:01 +00:00
Samuel Pitoiset 104a564920 radv: use vk_dynamic_graphic_state for the viewport state 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20314>
 2023-01-05 10:07:01 +00:00
Samuel Pitoiset 251933efa7 radv: use vk_dynamic_graphics_state for the input assembly state 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20314>
 2023-01-05 10:07:01 +00:00
Samuel Pitoiset 3bf256e86d radv: use vk_dynamic_graphics_state for the tessellation state 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20314>
 2023-01-05 10:07:01 +00:00
Samuel Pitoiset 4183002c62 radv: use vk_dynamic_graphics_state for the fragment shading rate state 
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20314>
 2023-01-05 10:07:01 +00:00
Samuel Pitoiset 0241094eb4 radv: move viewport pre-computed xform outside of radv_viewport_state 
I think we should distinguish between dynamic states (applications) and
hardware states, and this will allow us to use vk_viewport_state
instead of our own structs.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20314>
 2023-01-05 10:07:01 +00:00