Mike Blumenkrantz [Mon, 29 Mar 2021 15:04:39 +0000 (11:04 -0400)]
lavapipe: ignore unused clearvalues when beginning renderpass
according to spec, unused clearvalues aren't accessed
> Only elements corresponding to cleared attachments are used. Other elements of pClearValues are ignored.
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9900>
Bas Nieuwenhuizen [Sun, 28 Mar 2021 20:50:02 +0000 (22:50 +0200)]
lavapipe: Free sorted descriptor array.
Fixes:
abc724e440e ("lavapipe: sort bindings before creating descriptor set")
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9880>
Bas Nieuwenhuizen [Sun, 28 Mar 2021 20:46:19 +0000 (22:46 +0200)]
vulkan: Fix descriptor set creation with zero bindings.
MAX2(count * struct size, 1) results in 1 for count=0, not the size of a struct.
Since this MAX only seems to exist so we can keep using NULL for error reporting,
just refactor to return a VkResult.
Fixes:
ad241b15a9e ("vk: consolidate dynamic descriptor binding sorting")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4522
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9880>
Icecream95 [Tue, 22 Dec 2020 10:16:34 +0000 (23:16 +1300)]
panfrost: Implement panfrost_set_global_binding
This is enough OpenCL support to pass the Piglit test
cl-custom-run-simple-kernel with PAN_MESA_DEBUG=deqp.
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9380>
Icecream95 [Tue, 2 Mar 2021 20:11:17 +0000 (09:11 +1300)]
pipe-loader: Stop trying to use kmsro for vgem
Fixes two OpenCL devices being created for a single kmsro-using GPU
when vgem is enabled in the kernel.
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9380>
Icecream95 [Sun, 28 Feb 2021 21:51:45 +0000 (10:51 +1300)]
pipe-loader,gallium/drm: Fix the kmsro pipe_loader target
Include drm_helper.h to define the driver descriptor again, but with a
new define GALLIUM_KMSRO_ONLY to disable defining descriptors for the
drivers that kmsro uses.
Fixes clinfo on Panfrost.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4002
Fixes:
9ec28b8d226 ("gallium/drm: Deduplicate screen creation for the dynamic (clover) pipe loader.")
Acked-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9380>
Lionel Landwerlin [Wed, 24 Mar 2021 07:56:42 +0000 (09:56 +0200)]
intel/fs/copy_prop: check stride constraints with actual final type
In some cases we will change the type of the destination register of
an instruction. This is the type we should use to verify that we're
allow to do the replacement.
Otherwise we can hit restrictions on CHV and upcoming Xe-Hp for
instance where the copy propagation transforms this :
send(16) (mlen: 2) vgrf10:UD, 0u, 0u, vgrf35:D, null:UD
mov(16) vgrf11:UW, vgrf10<2>:UW
mov(16) vgrf12:UW, vgrf10+0.2<2>:UW
mov(16) vgrf15:HF, |vgrf11|:HF
mov(16) vgrf16:HF, |vgrf12|:HF
mov(8) vgrf41<2>:UW, vgrf15+0.0:UW group0
mov(8) vgrf42<2>:UW, vgrf15+0.16:UW group8
mov(8) vgrf45<2>:UW, vgrf16+0.0:UW group0
mov(8) vgrf46<2>:UW, vgrf16+0.16:UW group8
into this :
send(16) (mlen: 2) vgrf10:UD, 0u, 0u, vgrf35:D, null:UD
mov(8) vgrf41<2>:HF, |vgrf10+0.0|<2>:HF group0
mov(8) vgrf42<2>:HF, |vgrf10+1.0|<2>:HF group8
mov(8) vgrf45<2>:HF, |vgrf10+0.2|<2>:HF group0
mov(8) vgrf46<2>:HF, |vgrf10+1.2|<2>:HF group8
Because of the floating point use, stride and offets should be the
same.
v2: Fix final destination type selection (Curro)
v3: constify (Curro)
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9832>
Matt Turner [Mon, 15 Mar 2021 18:54:23 +0000 (18:54 +0000)]
tu: Skip tu_tiling_config_update_tile_layout() if not using gmem
Otherwise pass->tile_align_w will be 0, leading to a divide by zero and
undefined behavior. In practice, I saw this lead to an infinite loop in
tests like
dEQP-VK.draw.instanced.draw_indexed_indirect_vk_primitive_topology_line_list_attrib_divisor_0_multiview
Reviewed-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9606>
Eric Anholt [Mon, 29 Mar 2021 20:54:15 +0000 (13:54 -0700)]
freedreno/a5xx: Use VALIDREG/CONDREG like a6xx do.
Fewer magic numbers and less to diff between the two.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9870>
Eric Anholt [Fri, 26 Mar 2021 23:28:46 +0000 (16:28 -0700)]
freedreno/a5xx: Add support for clip distances and use them for userclip.
A little low-stakes RE effort as I unwind from fighting CI all day. Comes
from diffing dEQP-VK.clipping.user_defined.clip_distance.vert.* on the
blob and comparing to a6xx behavior. (My blob doesn't do tess, so if
there are equivalent tess fields for some of these, I didn't find them)
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9870>
Dave Airlie [Mon, 15 Mar 2021 01:44:17 +0000 (11:44 +1000)]
lavapipe: enable KHR_shader_atomic_int64
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9591>
Dave Airlie [Wed, 17 Mar 2021 00:12:52 +0000 (10:12 +1000)]
gallivm: add 64-bit atomic global support
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9591>
Dave Airlie [Mon, 15 Mar 2021 01:43:43 +0000 (11:43 +1000)]
gallivm: add 64-bit atomic support for ssbo/shared.
This just fixes things up to handle bit-size
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9591>
Samuel Pitoiset [Mon, 15 Mar 2021 21:24:41 +0000 (22:24 +0100)]
ac/surface: do not allocate FMASK or CMASK for stencil-only surfaces on GFX9+
FMASK/CMASK make no sense with depth/stencil or stencil-only surfaces.
gfx9_compute_miptree() was called twice for stencil-only surfaces,
and the first call was allocating FMASK/CMASK.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9612>
Kenneth Graunke [Fri, 26 Mar 2021 15:08:07 +0000 (08:08 -0700)]
ci: Enable iris testing in meson-release
This should catch any build breaks in release mode. iris was already
build tested in debug builds, but not release mode.
Suggested by Michel Dänzer.
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9862>
Kenneth Graunke [Fri, 26 Mar 2021 18:04:49 +0000 (11:04 -0700)]
intel: Mark an otherwise unused variable in intel_dump_gpu as ASSERTED
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9862>
Samuel Pitoiset [Wed, 24 Mar 2021 08:54:28 +0000 (09:54 +0100)]
radv: require DRM 3.35+
Linux Kernel 4.15+ is now required for RADV, this kernel has been
released 3 years ago and should be in most modern distros.
This allows us to remove a lot of legacy code for fence/semaphore.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9800>
Bas Nieuwenhuizen [Sat, 27 Mar 2021 21:38:26 +0000 (22:38 +0100)]
radv: Only set WRITE_COMPRESS_ENABLE on supported HW.
Fixes:
a7c0cf500b3 ("radv: Enable DCC for image stores on GFX10.")
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9876>
Gert Wollny [Sun, 28 Mar 2021 18:03:33 +0000 (20:03 +0200)]
r600/sfn: allocate number of VS outputs based on max driver_location
It seems that when the linker optimized the IO it might leave gaps
in the driver output locations, therefore set the size of the output
array based on the highest driver location occupied by a store_output.
Fixes #4520
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9879>
Gert Wollny [Sat, 27 Mar 2021 18:17:07 +0000 (19:17 +0100)]
r600: Don't advertise support for scaled int16 vertex formats
They are not supported by the hardware.
Fixes:
dEQP-GLES2.functional.vertex_arrays.single_attribute.strides.buffer_0_32_short3_vec4_dynamic_draw_quads_1
dEQP-GLES2.functional.vertex_arrays.single_attribute.strides.buffer_0_32_short3_vec4_dynamic_draw_quads_256
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9877>
Gert Wollny [Sun, 28 Mar 2021 08:27:20 +0000 (10:27 +0200)]
r600: don't set an index_bias for indirect draw calls
The indirect draw call already encodes the index bias so that no
additional encoding in the hardware is needed in this case.
This fixes a regression with a number of tests from
dEQP-GLES31.functional.draw_indirect.random.*
Fixes:
c6c532faa8cf172715b38783ec7175b626a63944
"gallium/u_vbuf: use updated pipe_draw_start_count while using draw_vbo"
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9877>
Gert Wollny [Sun, 28 Mar 2021 09:03:35 +0000 (11:03 +0200)]
r600: Fix texture buffer offset alignment
It seems that this has to be 4 (radeonsi also sets this value to 4)
Fixes deqp-gles31:
functional.texture.texture_buffer.modify.bufferdata.offset_1_alignments
functional.texture.texture_buffer.modify.bufferdata.offset_7_alignments
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9877>
Vinson Lee [Mon, 29 Mar 2021 03:24:27 +0000 (20:24 -0700)]
freedreno: Fix file descriptor leak.
Fix defect reported by Coverity Scan.
Resource leak (RESOURCE_LEAK)
leaked_handle: Handle variable fd going out of scope leaks the handle.
Fixes:
5a13507164a ("freedreno/perfcntrs: add fdperf")
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9889>
Erik Faye-Lund [Thu, 4 Feb 2021 16:27:49 +0000 (17:27 +0100)]
zink: use pre-populated shader-locations
This is the final step of this little journey, which allows us to no
longer deal with the shader-slot mapping inside the NIR to SPIR-V
translation.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9836>
Erik Faye-Lund [Thu, 4 Feb 2021 15:58:52 +0000 (16:58 +0100)]
zink: do not depend on shader_slots_reserved for xfb
This will be helpful in ripping this out later on.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9836>
Erik Faye-Lund [Thu, 4 Feb 2021 15:11:00 +0000 (16:11 +0100)]
zink: pre-populate locations in variables
These locations should match what we get out of the pre-existing pass,
and are currently unused. They will be used in a later commit in this
series, but is left like this because this code is fairly fragile.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9836>
Erik Faye-Lund [Thu, 4 Feb 2021 13:28:46 +0000 (14:28 +0100)]
zink: handle errors in nir_to_spirv
We already have some appropriate error-handling further down here.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9836>
Rhys Perry [Thu, 18 Mar 2021 19:14:24 +0000 (19:14 +0000)]
aco/tests: fix isel.sparse.clause for LLVM 12+
Seems disassembly of this instruction was fixed in LLVM 12.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Tested-by: Vinson Lee <vlee@freedesktop.org>
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4154
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9694>
Erik Faye-Lund [Mon, 29 Mar 2021 13:09:19 +0000 (15:09 +0200)]
frontends/va: correct check for invalid format
We need to check against the correct define, DRM_FORMAT_INVALID.
Checking an uint32_t against DRM_FORMAT_MOD_INVALID will always produce
false anyway.
Caught by Coverity.
Fixes:
b85c6531f6d ("frontends/va: add support for VA_EXPORT_SURFACE_COMPOSED_LAYERS")
Reviewed-by: Simon Ser <contact@emersion.fr>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9897>
Mauro Rossi [Mon, 29 Mar 2021 11:25:19 +0000 (13:25 +0200)]
android: panfrost: Use the blend shader cache attached to the device
pan_blend_shaders.{c,h} files are removed from Makefile.sources
in order to avoid the following building error:
FAILED: out/target/product/x86_64/obj_x86/STATIC_LIBRARIES/libmesa_pipe_panfrost_intermediates/pan_blend_shaders.o
...
clang: error: no such file or directory: 'external/mesa/src/gallium/drivers/panfrost/pan_blend_shaders.c'
clang: error: no input files
Fixes:
7994929e84a3 ("panfrost: Use the blend shader cache attached to the device")
Acked-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9894>
Danylo Piliaiev [Mon, 1 Mar 2021 15:31:56 +0000 (17:31 +0200)]
turnip,ir3: account for dispatch group offsets
Fixes tests:
dEQP-VK.compute.device_group.dispatch_base
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9339>
Lionel Landwerlin [Mon, 29 Mar 2021 09:48:21 +0000 (11:48 +0200)]
etnaviv/drm: only print out fence error on non timeout
While looking at the traces emitted by chromium, I saw a lot of those
errors. But looking at the value of ns, it is 0, so it's probably just
the application checking whether work is done or not. Not much point
in printing out an error.
v2: check ret value (Christian)
check both timeout codes (Lionel)
Signed-off-by: Lionel Landwerlin <llandwerlin@gmail.com>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9875>
Dave Airlie [Thu, 25 Mar 2021 06:59:50 +0000 (16:59 +1000)]
util: rework AMD cpu L3 cache affinity code.
This changes how the L3 cache affinity code works out the affinity
masks. It works better with multi-CPU systems and should also be
capable of handling big/little type situations if they appear in
the future.
It now iterates over all CPU cores, gets the core count for each
CPU, and works out the L3_ID from the physical CPU ID, and
the current cores L3 cache. It then tracks how many L3 caches
it has seen and reallocate the affinity masks for each one.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4496
Fixes:
d8ea5099658 ("util: completely rewrite and do AMD Zen L3 cache pinning correctly")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9782>
Dave Airlie [Thu, 25 Mar 2021 03:40:51 +0000 (13:40 +1000)]
st/glthread: allow for invalid L3 cache id.
If we get 0xffffffff consider L3 cache info invalid and
don't continue.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4496
Fixes:
d8ea5099658 ("util: completely rewrite and do AMD Zen L3 cache pinning correctly")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9782>
Dave Airlie [Mon, 22 Mar 2021 05:31:45 +0000 (15:31 +1000)]
llvmpipe: fix cube image size query
The 3rd coord wasn't being handled properly
Acked-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9746>
Timothee Chabat [Mon, 22 Mar 2021 09:46:26 +0000 (10:46 +0100)]
llvmpipe: increase PIPE_CAP_MAX_TEXTURE_BUFFER_SIZE value
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4237
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9750>
Mike Blumenkrantz [Thu, 25 Mar 2021 19:24:20 +0000 (15:24 -0400)]
llvmpipe: stop flattening multidraws
these are now handled in the draw module
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9844>
Mike Blumenkrantz [Thu, 25 Mar 2021 19:23:33 +0000 (15:23 -0400)]
aux/draw: pass full draw params to draw_pt_arrays()
multidraw depth maximum: loop directly around the llvmpipe call
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9844>
Mike Blumenkrantz [Thu, 25 Mar 2021 19:22:27 +0000 (15:22 -0400)]
aux/draw: track increment_draw_id value from draw info
need this in order to accurately update drawid
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9844>
Mike Blumenkrantz [Thu, 25 Mar 2021 19:08:18 +0000 (15:08 -0400)]
aux/draw: move draw param sanitization to end of function
this triggers more work if sanitizing no-ops the draw, but the rest of the
time it won't matter
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9844>
Mike Blumenkrantz [Thu, 25 Mar 2021 19:04:43 +0000 (15:04 -0400)]
aux/draw: pass the full draw params through to draw_pt_arrays_restart()
multidraw depth 1.5: loop in draw_pt_arrays_restart()
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9844>
Mike Blumenkrantz [Thu, 25 Mar 2021 19:04:43 +0000 (15:04 -0400)]
aux/draw: pass the full draw params through to draw_instances()
multidraw depth 1: loop in draw_instances()
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9844>
Mike Blumenkrantz [Thu, 25 Mar 2021 18:58:22 +0000 (14:58 -0400)]
aux/draw: rewrite PRIM_RESTART_LOOP macro as a function
this was a bit too intense to work with
no functional changes
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9844>
Mike Blumenkrantz [Thu, 25 Mar 2021 18:17:50 +0000 (14:17 -0400)]
aux/draw: stop copying draw params unnecessarily
resolve_draw_info() is only for drawing from streamout, so this check
can be moved forward to avoid overwriting draw params when not necessary
no functional changes at this time
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9844>
Boris Brezillon [Tue, 23 Mar 2021 11:20:35 +0000 (12:20 +0100)]
panfrost: Use the blend shader cache attached to the device
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9831>
Boris Brezillon [Tue, 23 Mar 2021 11:17:03 +0000 (12:17 +0100)]
panfrost: Move the blend shader cache at the device level
So we can re-use it in the Vulkan driver.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9831>
Iago Toral Quiroga [Thu, 25 Mar 2021 13:43:20 +0000 (14:43 +0100)]
broadcom/compiler: flag TMU reads with a read dependency on last TMU config
We were using a write dependency to ensure ordering since LDTMUs sequences
are ordered, but by using a write dependency with TMU config we were also
preserving ordering with TMU config writes that are not a sequence
terminator, which is not required and reduces scheduling flexibility.
Instead, use a write dependency to ensure strict ordering of TMU reads,
but only a read depdency with TMU config.
With this change we also need to update CS barriers to also have a write
dependency with TMU reads to ensure that we don't move TMU reads around
CS barriers.
total instructions in shared programs:
13602500 ->
13597851 (-0.03%)
instructions in affected programs: 2681428 -> 2676779 (-0.17%)
helped: 6567
HURT: 4960
Instructions are helped.
total max-temps in shared programs: 2317927 -> 2317914 (<.01%)
max-temps in affected programs: 13861 -> 13848 (-0.09%)
helped: 355
HURT: 300
Inconclusive result (value mean confidence interval includes 0).
total sfu-stalls in shared programs: 32074 -> 32247 (0.54%)
sfu-stalls in affected programs: 848 -> 1021 (20.40%)
helped: 160
HURT: 327
Inconclusive result (%-change mean confidence interval includes 0).
total inst-and-stalls in shared programs:
13634574 ->
13630098 (-0.03%)
inst-and-stalls in affected programs: 2703041 -> 2698565 (-0.17%)
helped: 6558
HURT: 5020
Inst-and-stalls are helped.
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9856>
Iago Toral Quiroga [Thu, 25 Mar 2021 10:36:03 +0000 (11:36 +0100)]
broadcom/compiler: flag TMU read dependencies against last TMU config
Instead of last TMU write. According to the documentation, the entries
in the output FIFO are pushed with the *final* input write for the
lookup, which is the one terminating the sequence. We flag these
with last_tmu_config.
This will allow us to move all TMU register writes for a lookup except
the last one ahead of the LDTMUs for the previous lookup, possibly
allowing us to pair up these writes the wrtmuc instructions for the
same lookup, turning code like this:
nop ; nop ; wrtmuc (tex[0].p0 | 0x3)
nop ; nop ; wrtmuc (tex[2].p1 | 0x1)
nop ; nop ; ldunif (ubo[2]+0xe0)
fadd r4, rf33, rf51 ; mov unifa, r5 ; ldunif (ubo[2]+0x110)
fmax rf34, 0, r4 ; nop
nop ; mov tmut, rf11
nop ; mov tmus, rf0
into:
nop ; mov tmut, rf11 ; wrtmuc (tex[0].p0 | 0x3)
nop ; nop ; wrtmuc (tex[2].p1 | 0x1)
nop ; nop ; ldunif (ubo[2]+0xe0)
fadd r4, rf33, rf51 ; mov unifa, r5 ; ldunif (ubo[2]+0x110)
fmax rf34, 0, r4 ; nop
nop ; mov tmus, rf0
total instructions in shared programs:
13648140 ->
13602500 (-0.33%)
instructions in affected programs: 3497402 -> 3451762 (-1.30%)
helped: 12044
HURT: 3484
Instructions are helped.
total max-temps in shared programs: 2318687 -> 2317927 (-0.03%)
max-temps in affected programs: 17234 -> 16474 (-4.41%)
helped: 615
HURT: 198
Max-temps are helped.
total sfu-stalls in shared programs: 32354 -> 32074 (-0.87%)
sfu-stalls in affected programs: 1462 -> 1182 (-19.15%)
helped: 461
HURT: 188
Sfu-stalls are helped.
total inst-and-stalls in shared programs:
13680494 ->
13634574 (-0.34%)
inst-and-stalls in affected programs: 3514405 -> 3468485 (-1.31%)
helped: 12062
HURT: 3486
Inst-and-stalls are helped.
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9856>
Mike Blumenkrantz [Thu, 25 Mar 2021 17:54:01 +0000 (13:54 -0400)]
lavapipe: refactor indexed draw dispatch to handle multidraws
each indexed direct draw cmd now contains n draws (though right now n=1)
indexed draws now also have a flag used to avoid recalculating start index if
a cmdbuf is submitted multiple times
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9843>
Mike Blumenkrantz [Thu, 25 Mar 2021 17:54:01 +0000 (13:54 -0400)]
lavapipe: refactor base draw dispatch to handle multidraws
each direct draw cmd now contains n draws (though right now n=1)
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9843>
Mike Blumenkrantz [Wed, 24 Mar 2021 19:24:26 +0000 (15:24 -0400)]
zink: ci updates
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9811>
Mike Blumenkrantz [Fri, 12 Mar 2021 14:00:16 +0000 (09:00 -0500)]
zink: remove explicit fencing for query results
now that the results are automatically being copied to internal buffers,
mapping the buffers will handle any fencing that's required
the only caveat is that the query still needs to be flushed for a result
to be returned
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9811>
Mike Blumenkrantz [Sun, 14 Mar 2021 23:39:24 +0000 (19:39 -0400)]
zink: reorder availability handling for (user) qbos
if the shortcut method without a staging buffer fails, falling back
to the staging buffer is more sane than silently failing
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9811>
Mike Blumenkrantz [Sun, 14 Mar 2021 23:28:52 +0000 (19:28 -0400)]
zink: don't use PARTIAL bit for query results with time queries
this is illegal
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9811>
Mike Blumenkrantz [Wed, 24 Mar 2021 19:34:43 +0000 (15:34 -0400)]
zink: bump pools up to 5k queries each
50 is very small and triggers constant stalls; this should probably be
tuned more at some point in the future since even this value is small in
some scenarios
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9811>
Mike Blumenkrantz [Fri, 18 Dec 2020 17:52:05 +0000 (12:52 -0500)]
zink: rewrite query internals
this adds internal qbos to each query which record the results of queries incrementally,
meaning that any time we want the result of a query we just have to read the
buffer back
in particular this opens up some possibilities for further optimization, specifically with
actual qbos where we can now pass the internal qbo off to a compute shader to
mimic the check_query_results() handling and avoid having to actually sync or
do a cpu read
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9811>
Mike Blumenkrantz [Fri, 18 Dec 2020 23:07:00 +0000 (18:07 -0500)]
zink: remove special casing for occlusion qbos
this seems fine now
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9811>
Mike Blumenkrantz [Fri, 18 Dec 2020 23:05:23 +0000 (18:05 -0500)]
zink: manually handle more bool query types for copying
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9811>
Mike Blumenkrantz [Fri, 18 Dec 2020 19:12:22 +0000 (14:12 -0500)]
zink: remove flush from query buffer copy
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9811>
Mike Blumenkrantz [Fri, 18 Dec 2020 17:44:11 +0000 (12:44 -0500)]
zink: create result buffers for all query streams
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9811>
Mike Blumenkrantz [Fri, 18 Dec 2020 17:40:21 +0000 (12:40 -0500)]
zink: clean up query creation failure paths
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9811>
Mike Blumenkrantz [Mon, 28 Dec 2020 20:08:46 +0000 (15:08 -0500)]
zink: rework border color handling
we should be matching the format type here to the underlying surface type
for int/float
also we can avoid using a custom border color if the color can be reduced to
a non-custom color
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9793>
Mike Blumenkrantz [Mon, 14 Dec 2020 19:26:05 +0000 (14:26 -0500)]
zink: ralloc shader cache and keys
hash table keys (keybox) are allocated, so we can avoid leaking them by
allocating everything on a single ralloc context
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9794>
Mike Blumenkrantz [Mon, 4 Jan 2021 16:55:16 +0000 (11:55 -0500)]
zink: flatten out buffer creation usage flags codepath
GL doesn't know what it's doing, so we just need all the flags all the time
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9788>
Mike Blumenkrantz [Thu, 4 Feb 2021 13:11:13 +0000 (08:11 -0500)]
zink: toggle between linear/optimal tiling during image creation
if we would be creating an unusable image, don't
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9788>
Mike Blumenkrantz [Wed, 3 Feb 2021 20:34:58 +0000 (15:34 -0500)]
zink: check image format props before creating image
ensure that these attrs are viable before failing silently later on
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9788>
Mike Blumenkrantz [Wed, 3 Feb 2021 20:32:52 +0000 (15:32 -0500)]
zink: add color output bit and/or use linear tiling for sampled images
these need to have some way to get data onto them
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9788>
Mike Blumenkrantz [Wed, 3 Feb 2021 20:32:17 +0000 (15:32 -0500)]
zink: incrementally add image usage flags based on device caps
these should be provided, but if not...
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9788>
Mike Blumenkrantz [Sat, 12 Dec 2020 06:45:29 +0000 (01:45 -0500)]
zink: create separate vk image/buffer objects for shader image use
the STORAGE_TEXEL and STORAGE_IMAGE bits can't be accurately applied due
to opengl allowing all resources to be used everywhere, so instead we can
create a separate object on demand which is used only by shaders and gets
extra barriers inferred along with the base object to avoid desync whenever
it is used
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9788>
Mike Blumenkrantz [Mon, 30 Nov 2020 17:36:24 +0000 (12:36 -0500)]
zink: break out barrier struct initializing into helper funcs
make this and the no-op handling reusable for future extension
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9788>
Mike Blumenkrantz [Mon, 30 Nov 2020 17:27:24 +0000 (12:27 -0500)]
zink: reorder barrier util functions to set up barrier struct before batch
no functional changes
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9788>
Mike Blumenkrantz [Thu, 24 Dec 2020 01:15:22 +0000 (20:15 -0500)]
zink: make descriptor state invalidate public
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9788>
Mike Blumenkrantz [Sat, 12 Dec 2020 05:40:19 +0000 (00:40 -0500)]
zink: break out image/buffer create info structs into helper funcs
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9788>
Mike Blumenkrantz [Thu, 10 Dec 2020 21:49:13 +0000 (16:49 -0500)]
zink: fix format support detection for storage texel buffers and shader images
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9788>
Mike Blumenkrantz [Fri, 27 Nov 2020 14:47:10 +0000 (09:47 -0500)]
zink: reset all fences when waiting on batch state
at this point we know all the states are available, so we can shortcut
future state-finding
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9787>
Mike Blumenkrantz [Fri, 27 Nov 2020 14:37:24 +0000 (09:37 -0500)]
zink: stall when we start getting a lot of uncompleted batches
temporary oom handling, though it's unrealistic for this to be hit
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9787>
Mike Blumenkrantz [Mon, 29 Mar 2021 01:27:44 +0000 (21:27 -0400)]
zink: use correct surface ref function for context destroy
the regular surface reference function can never be used in the driver
if it is possible that a surface may be destroyed, as this may reference
a previously-destroyed context
Fixes:
92a5ea13fcd ("zink: implement a global framebuffer cache")
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9787>
Mike Blumenkrantz [Wed, 16 Dec 2020 17:02:51 +0000 (12:02 -0500)]
zink: only do shader updates when relevant stages are dirty
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9792>
Mike Blumenkrantz [Mon, 22 Mar 2021 19:45:37 +0000 (15:45 -0400)]
lavapipe: fix array texture region copies
these need to use different struct members for copying array textures
the buffer2image variants are already doing the right thing
Fixes:
b38879f8c5f ("vallium: initial import of the vulkan frontend")
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9761>
Mike Blumenkrantz [Mon, 22 Mar 2021 19:34:18 +0000 (15:34 -0400)]
zink: also fix image buffer layer copying
this differs based on image type
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9759>
Mike Blumenkrantz [Mon, 22 Mar 2021 19:16:06 +0000 (15:16 -0400)]
zink: fix handling for image types in resource_copy_region hook
different image types are supposed to use different members of this struct
to indicate layer copies, and this can't necessarily be inferred by checking
array_size
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9759>
Mike Blumenkrantz [Fri, 26 Mar 2021 18:04:16 +0000 (14:04 -0400)]
lavapipe: remove lvp_descriptor_update_template::descriptor_set_layout
this wasn't used and caused crashes when running with validation enabled
due to violating spec:
This parameter is ignored if templateType is not VK_DESCRIPTOR_UPDATE_TEMPLATE_TYPE_DESCRIPTOR_SET
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9861>
Mike Blumenkrantz [Fri, 26 Mar 2021 12:09:26 +0000 (08:09 -0400)]
lavapipe: ignore templateType when descriptor template isn't for push descriptors
spec compliance requires this and validation may crash without it
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9853>
Gert Wollny [Sat, 27 Mar 2021 16:42:04 +0000 (17:42 +0100)]
r600/sfn: make allocate_temp_register private to valuepool
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9874>
Gert Wollny [Sat, 27 Mar 2021 16:41:34 +0000 (17:41 +0100)]
r600/sfn: use get_temp_vec4 directly when fetching
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9874>
Gert Wollny [Sat, 27 Mar 2021 16:41:02 +0000 (17:41 +0100)]
r600/sfn: clean up value pool interface usage in emit_stream
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9874>
Gert Wollny [Sat, 27 Mar 2021 16:40:28 +0000 (17:40 +0100)]
r600/sfn: clean up multi-sample texture load
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9874>
Gert Wollny [Sat, 27 Mar 2021 16:15:03 +0000 (17:15 +0100)]
r600/sfn: remove find_msb lowering in driver
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9874>
Gert Wollny [Mon, 22 Mar 2021 07:18:52 +0000 (08:18 +0100)]
r600/sfn: force dual source blend output handling in some cases
If an output has a dual source index > 0 then we need to emit both
outputs, even if the number of outputs is larger than the number
of active outputs.
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9874>
Gert Wollny [Sun, 21 Mar 2021 22:43:23 +0000 (23:43 +0100)]
r600/sfn: remove old deref code
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9874>
Gert Wollny [Sun, 21 Mar 2021 21:45:06 +0000 (22:45 +0100)]
r600/sfn: lowered FS output IO
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9874>
Gert Wollny [Sun, 21 Mar 2021 19:48:19 +0000 (20:48 +0100)]
r600/sfn: lower VS IO and drop old deref code
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9874>
Gert Wollny [Sun, 21 Mar 2021 18:50:02 +0000 (19:50 +0100)]
r600/sfn: drop the local register mep
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9874>
Gert Wollny [Sun, 21 Mar 2021 17:50:36 +0000 (18:50 +0100)]
r600/sfn: switch assembler creation to use visitor
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9874>
Gert Wollny [Mon, 15 Mar 2021 08:22:38 +0000 (09:22 +0100)]
r600/sfn: fix some formatting
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9874>
Gert Wollny [Sun, 14 Mar 2021 16:05:24 +0000 (17:05 +0100)]
r600/sfn: remove extra parameter from alu assemebly emission
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9874>
Gert Wollny [Tue, 2 Feb 2021 12:45:50 +0000 (13:45 +0100)]
r600/sfn: Add skelton for visitor pattern
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9874>
Gert Wollny [Sun, 21 Mar 2021 16:48:12 +0000 (17:48 +0100)]
r600/sfn: Make some value pool functions private
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9874>
Rob Clark [Sun, 21 Mar 2021 18:08:14 +0000 (11:08 -0700)]
util/primconvert: Handle indirect and multi-draw
Indirect handling was completely missing. And even though we have to
emulate multi-draw, this pushes it out of the fast/hot path in the
driver's draw_vbo()
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9742>
Rob Clark [Sun, 21 Mar 2021 17:37:40 +0000 (10:37 -0700)]
freedreno: Don't handle multi-draw in indirect case
num_draws is only applicable to direct draws, so we can simplify things
a bit.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9742>