Mike Blumenkrantz [Mon, 28 Sep 2020 14:51:57 +0000 (10:51 -0400)]
zink: force batch flush if batches are using more than 1/10 total system memory
this is only tracking memory used by resources referenced in the batch, but it
can be adjusted a bit if we see that we're flushing too often
fixes spec@!opengl 1.1@streaming-texture-leak hogging all system memory and ooming
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9274>
Mike Blumenkrantz [Mon, 28 Sep 2020 14:51:06 +0000 (10:51 -0400)]
zink: track resource mem usage per batch
we want to be able to track this so we can check whether a given batch is
going wild with memory usage for resources that might be pending free once
the batch finishes
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9274>
Mike Blumenkrantz [Mon, 28 Sep 2020 14:50:23 +0000 (10:50 -0400)]
zink: store total memory size on zink_screen
need this for oom handling
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9274>
Juan A. Suarez Romero [Mon, 1 Mar 2021 12:44:52 +0000 (13:44 +0100)]
ci/broadcom: retry always when serial log timeout
So far we were retrying the testing (through device rebooting) if we did
not detect the boot sequence.
But found a couple of times that the serial log can also be "lost"
during the testing process. In all those times a manual retry of the job
was enough to complete the test.
Thus, let's apply the retry once automatically in this case.
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9335>
Juan A. Suarez Romero [Mon, 1 Mar 2021 12:20:03 +0000 (13:20 +0100)]
ci/baremetal: highlight message errors
Highlight in red errors from the baremetal run, so user is more aware of
what happened.
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9335>
Samuel Pitoiset [Wed, 24 Feb 2021 17:59:38 +0000 (18:59 +0100)]
radv: remove useless decompression of the DS resolve attachment
The DS resolve attachment is the destination attachment, it doesn't
need to be decompressed before resolving the depth/stencil attachment.
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/9256>
Rhys Perry [Fri, 29 Jan 2021 17:42:36 +0000 (17:42 +0000)]
nir/copy_prop: use nir_{instr,if}_rewrite_{src,condition}_ssa
Compile-time (nir_copy_prop):
Difference at 95.0% confidence
-2470.88 +/- 19.8762
-35.7461% +/- 0.247259%
(Student's t, pooled s = 23.4747)
Compile-time (overall):
Difference at 95.0% confidence
-2175.72 +/- 178.786
-1.73627% +/- 0.140826%
(Student's t, pooled s = 211.155)
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8784>
Rhys Perry [Fri, 29 Jan 2021 17:29:32 +0000 (17:29 +0000)]
nir/copy_prop: visit copies instead of sources
There are less copy instructions than sources, so instead of visiting each
source and rewriting it if it's uses a copy instruction, visit each copy
instruction and rewrite it's users.
Besides improving compile time, this also has a side effect of fixing a
rare situation where copy-propagation does not happen:
loop {
a = phi ..., b
c = vec ...
b = mov c.y
}
It might have been the case that a phi source could not be rewritten until
the copy was visited later.
Compile-time (nir_copy_prop):
Difference at 95.0% confidence
-2613.13 +/- 15.2094
-27.4333% +/- 0.150247%
(Student's t, pooled s = 17.963)
Comple-time (overall):
Difference at 95.0% confidence
-2627.89 +/- 201.557
-2.05404% +/- 0.156221%
(Student's t, pooled s = 238.048)
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8784>
Rhys Perry [Fri, 29 Jan 2021 14:32:50 +0000 (14:32 +0000)]
nir/copy_prop: remove unused copies
These were hurting performance of other passes.
Compile-time (overall):
Difference at 95.0% confidence
-5496.3 +/- 219.752
-4.11912% +/- 0.160285%
(Student's t, pooled s = 259.538)
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8784>
Rhys Perry [Tue, 2 Feb 2021 16:00:53 +0000 (16:00 +0000)]
nir: add nir_ssa_def_is_unused()
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8784>
Italo Nicola [Fri, 26 Feb 2021 10:52:55 +0000 (10:52 +0000)]
pan/mdg: prevent csel_v from being scheduled alongside writeout
Midgard writeout arguments need to be written to in the same bundle the
writeout happens. Both csel, csel_v and their float variants also
require their conditional to be performed on the same bundle.
This patch prevents scheduling csel the same bundle as a writeout,
fixing the scheduling issue.
But... there's still room for optimizations since in some cases it might
be possible to fit all these instructions in the same bundle.
No shader-db changes.
Signed-off-by: Italo Nicola <italonicola@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9340>
Alyssa Rosenzweig [Mon, 1 Mar 2021 16:01:22 +0000 (16:01 +0000)]
panfrost/lcra: Fix constraint counting
We need to iterate the whole row, we can't be clever and only look at
one side, the symmetry doesn't work like that. See the original paper.
total instructions in shared programs: 69392 -> 69322 (-0.10%)
instructions in affected programs: 9002 -> 8932 (-0.78%)
helped: 82
HURT: 28
Instructions are helped.
total bundles in shared programs: 32225 -> 32155 (-0.22%)
bundles in affected programs: 4286 -> 4216 (-1.63%)
helped: 82
HURT: 28
Bundles are helped.
total quadwords in shared programs: 56102 -> 56102 (0.00%)
quadwords in affected programs: 0 -> 0
helped: 0
HURT: 0
total registers in shared programs: 4552 -> 4572 (0.44%)
registers in affected programs: 298 -> 318 (6.71%)
helped: 18
HURT: 38
Registers are HURT.
total threads in shared programs: 3772 -> 3775 (0.08%)
threads in affected programs: 84 -> 87 (3.57%)
helped: 15
HURT: 14
Inconclusive result (value mean confidence interval includes 0).
total spills in shared programs: 0 -> 0
spills in affected programs: 0 -> 0
helped: 0
HURT: 0
total fills in shared programs: 0 -> 0
fills in affected programs: 0 -> 0
helped: 0
HURT: 0
Fixes:
66ad64d73d2 ("pan/midgard: Implement linearly-constrained register allocation")
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9338>
Mike Blumenkrantz [Thu, 25 Feb 2021 21:30:34 +0000 (16:30 -0500)]
zink: track all framebuffers per batch
now that 1 batch != 1 renderpass, this needs to be a set
Fixes:
1cb3015a31c ("zink: just end the current renderpass in zink_batch_no_rp()")
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9298>
Mike Blumenkrantz [Mon, 1 Mar 2021 15:40:05 +0000 (10:40 -0500)]
zink: handle null src for fb refs
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9298>
Mike Blumenkrantz [Wed, 19 Aug 2020 14:10:30 +0000 (10:10 -0400)]
zink: add more usage bits for buffer resource creation
gallium doesn't give us enough info here, so yolo
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9197>
Mike Blumenkrantz [Thu, 25 Feb 2021 15:32:41 +0000 (10:32 -0500)]
zink: don't start renderpasses during descriptor update
this is unnecessary
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9292>
Bas Nieuwenhuizen [Sun, 28 Feb 2021 01:59:25 +0000 (02:59 +0100)]
radv: Add nodisplaydcc option.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9318>
Bas Nieuwenhuizen [Sat, 27 Feb 2021 19:16:21 +0000 (20:16 +0100)]
radv: Add sam option.
So that people without large BAR can try this out.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9318>
Iago Toral Quiroga [Mon, 1 Mar 2021 09:42:42 +0000 (10:42 +0100)]
v3dv: fix branching to large secondaries with more than one BCL buffer.
Fixes:
dEQP-VK.api.command_buffers.record_many_draws_secondary_*
Tested-by: Juan A. Suarez <jasuarez@igalia.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9333>
Bas Nieuwenhuizen [Sun, 28 Feb 2021 02:09:48 +0000 (03:09 +0100)]
radv: Enable linear sampling for depth textures.
Turns out there are CTS tests.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4258
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9319>
Samuel Pitoiset [Fri, 26 Feb 2021 14:19:25 +0000 (15:19 +0100)]
radeonsi,radv: do not overallocate the SQTT buffer size
The number of shader engines isn't always 4.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9307>
Samuel Pitoiset [Fri, 26 Feb 2021 14:23:33 +0000 (15:23 +0100)]
Revert "radv: do not overallocate the SQTT buffer"
This fixes computing the thread trace data offset.
This reverts commit
c7e6f4ff3da39f36fdda8f0ec6eea4925cd3afdc.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9307>
Samuel Pitoiset [Fri, 26 Feb 2021 15:36:39 +0000 (16:36 +0100)]
radv: exclude perf counters for SQTT also on GFX10.3
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/9300>
Samuel Pitoiset [Fri, 26 Feb 2021 07:53:57 +0000 (08:53 +0100)]
radv: fix exporting SQTT pipelines with LLVM
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/9300>
Samuel Pitoiset [Fri, 26 Feb 2021 07:43:48 +0000 (08:43 +0100)]
radv: always select the first active CU when profiling with SQTT
This probably fixes instruction tracing on many chips.
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/9300>
Samuel Pitoiset [Fri, 26 Feb 2021 07:27:16 +0000 (08:27 +0100)]
radv: remove duplicate REG_INCLUDE_CONTEXT setting for SQTT
It was set twice.
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/9300>
Andreas Bergmeier [Sat, 27 Feb 2021 22:28:19 +0000 (22:28 +0000)]
v3dv: Output a message if file open fails in physical_device_init
In the caller, this error simply gets mapped to VK_ERROR_INIT[...].
Especially for users it is very valuable to know what the driver
tried and what kind of failure occured. Thus just straight out log
to stderr.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9317>
Tapani Pälli [Mon, 1 Mar 2021 06:39:16 +0000 (08:39 +0200)]
gitlab-ci: bump piglit commit for windows
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8520>
Tapani Pälli [Tue, 19 Jan 2021 05:43:41 +0000 (07:43 +0200)]
mesa: add check that non base level attachment is mipmap complete
Patch adds a check for mipmap completeness of framebuffer object texture
attachments. Since a glTexImage call might have updated miplevels
meanwhile, we test the completeness before setting framebuffer object
incomplete.
Fixes some upcoming framebuffer completeness CTS tests that explicitly
test case where we have mipmap incomplete non base level texture which
should make also framebuffer object incomplete. After update to the
texture it should make framebuffer object complete again.
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8520>
Jordan Justen [Sun, 28 Feb 2021 21:23:07 +0000 (13:23 -0800)]
i965: Call util_cpu_detect() early in screen creation
This helps to avoid this assertion in debug builds:
src/util/u_cpu_detect.h:116: util_get_cpu_caps: Assertion `util_cpu_caps.nr_cpus >= 1' failed.
Fixes:
a9618e7c421 ("util: Add accessor for util_cpu_caps")
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9325>
Greg V [Sat, 14 Nov 2020 14:53:28 +0000 (17:53 +0300)]
meson: Add missing lavapipe dep
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7620>
Bas Nieuwenhuizen [Tue, 23 Feb 2021 23:00:58 +0000 (00:00 +0100)]
frontends/va: Use correct size for secondary planes.
And initialize the whandle format while at it.
Fixes:
f7a4051b836 ("radeonsi: Check pitch and offset for validity.")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4126
Reviewed-by: Simon Ser <contact@emersion.fr>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9236>
Simon Ser [Fri, 26 Feb 2021 15:49:17 +0000 (16:49 +0100)]
radeonsi/uvd: make format modifiers-aware
When format modifiers are supported, use
resource_create_with_modifiers instead of resource_create. This
allows radeonsi to set the modifier field, and allows VA-API
clients to have a proper modifier instead of
DRM_FORMAT_MOD_INVALID.
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9308>
Pierre Moreau [Thu, 7 May 2020 11:45:11 +0000 (13:45 +0200)]
docs/features: Add OpenCL status
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/142
Signed-off-by: Pierre Moreau <dev@pmoreau.org>
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4979>
Andres Gomez [Wed, 17 Feb 2021 16:23:30 +0000 (18:23 +0200)]
ci: add libdrm to the x86_test-vk container
Fixes:
5d0ba8b1834 (ci: Split building of libdrm to its own script)
Signed-off-by: Andres Gomez <agomez@igalia.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9102>
Andrii Simiklit [Thu, 25 Feb 2021 12:28:24 +0000 (14:28 +0200)]
mesa: ensure parameter list capacity before associating uniform storage
We have to reserve at lease 16 program parameters in storage to
avoid its reallocation.
v2: move allocation to `st_deserialise_ir_program` and add helper for that
( Eric Anholt <eric@anholt.net> )
v3 amend comments a bit
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4352
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Andrii Simiklit <andrii.simiklit@globallogic.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9282>
Lionel Landwerlin [Wed, 9 Dec 2020 11:23:16 +0000 (13:23 +0200)]
anv: group as many command buffers into a single execbuf
When semaphores are not involved, try to batch things up as much as
possible across VkSubmitInfo and also batch command buffers within a
VkSubmitInfo.
v2: Reuse anv_cmd_buffer_is_chainable()
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2371>
Lionel Landwerlin [Wed, 9 Dec 2020 11:22:45 +0000 (13:22 +0200)]
anv: allow multiple command buffers in anv_queue_submit
v2: Fixup crash spotted by Mark about missing alloc vfuncs
v3: Fixup double iteration over device->memory_objects (that ought to
be expensive...) (Ken)
v4: Add more asserts for non-softpin cases (Ken)
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2371>
Lionel Landwerlin [Wed, 9 Dec 2020 11:19:30 +0000 (13:19 +0200)]
anv: end command buffer with a potential jump
We would like to chain multiple primary command buffer to be submitted
together to i915. For prepare this, add end the command buffers with a
MI_BATCH_BUFFER_START and at submit time, replace it with
MI_BATHC_BUFFER_END if needed.
v2: Don't even consider non softpin platforms
v3: Fix inverted condition
v4: Limit is_chainable() to checking device->use_softpin (Jason)
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2371>
Lionel Landwerlin [Wed, 9 Dec 2020 09:56:31 +0000 (11:56 +0200)]
anv: track the end of the command buffers
Where MI_BATCH_BUFFER_END is recorded.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2371>
Lionel Landwerlin [Sun, 6 Dec 2020 21:38:03 +0000 (23:38 +0200)]
anv: make use of new helper function directly in anv_QueueSubmit()
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2371>
Bas Nieuwenhuizen [Thu, 25 Feb 2021 12:29:09 +0000 (13:29 +0100)]
radv: Expose robustBufferAccessUpdateAfterBind correctly.
We do support it.
Cc: mesa-stable
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4351
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9281>
Marek Olšák [Wed, 3 Feb 2021 07:40:59 +0000 (02:40 -0500)]
mesa: skip a subset of _mesa_update_state if no relevant flags are set
We only need to check 12 out of 32 flags.
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Wed, 3 Feb 2021 19:02:39 +0000 (14:02 -0500)]
mesa: don't update tnl spaces on irrelevant _NEW_POINT/TEXTURE_STATE changes
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Wed, 3 Feb 2021 18:13:51 +0000 (13:13 -0500)]
mesa: don't update the fixed-func frag prog on irrelevant _NEW_COLOR changes
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Wed, 3 Feb 2021 18:13:51 +0000 (13:13 -0500)]
mesa: don't update fixed-func programs on irrelevant _NEW_RENDER_MODE changes
The fixed-func vertex program uses it too, which was ignored. This commit
fixes it.
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Wed, 3 Feb 2021 18:13:51 +0000 (13:13 -0500)]
mesa: don't update fixed-func programs on irrelevant _NEW_FOG changes
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Wed, 3 Feb 2021 18:13:51 +0000 (13:13 -0500)]
mesa: don't update fixed-func programs on irrelevant _NEW_POINT changes
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Wed, 3 Feb 2021 18:13:51 +0000 (13:13 -0500)]
mesa: don't update fixed-func vert prog on irrelevant _NEW_TRANSFORM changes
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Wed, 3 Feb 2021 18:03:24 +0000 (13:03 -0500)]
mesa: only update fixed-func programs on texture matrix enablement changes
This should reduce fixed-func program key recomputations.
I also update the fixed-func fragment program, which was incorrectly
ignored because it's clearly used there.
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Wed, 3 Feb 2021 17:56:52 +0000 (12:56 -0500)]
mesa: move _mesa_update_pixel out of _mesa_update_state
This only has to be called in a few places and not in normal draw calls.
egl_image_target_texture doesn't call _mesa_update_pixel because it only
assigns an EGL image to a texture object.
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Wed, 3 Feb 2021 17:41:18 +0000 (12:41 -0500)]
mesa: be precise about when to set _NEW_CURRENT_ATTRIB and _NEW_MATERIAL
This commit fixes _mesa_update_color_material, which allows cleaning up
the unnecessary state flagging.
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Wed, 3 Feb 2021 16:58:49 +0000 (11:58 -0500)]
vbo: don't call update_color_material in copy_to_current if it's a no-op
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Tue, 2 Feb 2021 01:01:13 +0000 (20:01 -0500)]
vbo: optimize copy_to_current functions
- execute vbo_set_vertex_format in a separate skipable conditional block
- replace dmul with dmul_shift
- don't check <= VBO_ATTRIB_MAT_BACK_INDEXES because there is no attrib
above that
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Sun, 13 Dec 2020 04:04:56 +0000 (23:04 -0500)]
mesa: merge STATE_LIGHT_ATTENUATION and STATE_LIGHT_POSITION_* parameters
This decreases the CPU time spent in fetch_state.
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Sat, 5 Dec 2020 21:00:14 +0000 (16:00 -0500)]
mesa: merge STATE_LIGHTPROD parameters
This decreases the CPU time spent in fetch_state.
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Mon, 1 Feb 2021 23:55:42 +0000 (18:55 -0500)]
mesa: sort state parameters for ffvp to enable better parameter merging
_mesa_optimize_state_parameters will benefit from this.
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Mon, 1 Feb 2021 21:57:12 +0000 (16:57 -0500)]
mesa: remove the fixed-func vert prog dependency on all texture states
Just flag _NEW_FF_VERT_PROGRAM where needed. There are only a few places
that must do it.
Also do the same with _NEW_FF_FRAG_PROGRAM, but this is not sufficient
for the ff frag prog to ignore texture states.
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Wed, 3 Feb 2021 18:54:34 +0000 (13:54 -0500)]
mesa: don't push/pop ctx->Texture._* derived states
They are always recomputed by _mesa_update_state, which will need the old
values, so that it can update other dependent states if needed.
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Mon, 1 Feb 2021 21:09:04 +0000 (16:09 -0500)]
mesa: remove _NEW_LIGHT_FF_PROGRAM in favor of _NEW_FF_(VERT|FRAG)_PROGRAM
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Mon, 1 Feb 2021 21:09:04 +0000 (16:09 -0500)]
mesa: remove _NEW_VARYING_VP_INPUTS in favor of _NEW_FF_(VERT|FRAG)_PROGRAM
This adds _NEW_FF_FRAG_PROGRAM.
_mesa_set_varying_vp_inputs flags both fixed-func programs because both use
the state.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Mon, 1 Feb 2021 20:46:25 +0000 (15:46 -0500)]
mesa: don't update derived material values in _mesa_update_state and elsewhere
To my great surprise, many drivers don't use these values at all.
Move the update to the places where they are used.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Mon, 1 Feb 2021 20:07:31 +0000 (15:07 -0500)]
mesa: add _NEW_MATERIAL to reduce the weight of _NEW_LIGHT_CONSTANTS
This reduces no-op compute_light_positions in _mesa_update_state.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Mon, 1 Feb 2021 19:21:29 +0000 (14:21 -0500)]
mesa: don't compute the ModelView * Projection matrix if not used
Only GLSL built-in uniforms use it.
This further reduces the time spent in _mesa_update_state.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Mon, 1 Feb 2021 19:03:46 +0000 (14:03 -0500)]
mesa: don't compute the inverted projection matrix if not used
Only clip planes and GLSL built-in uniforms use it.
update_projection (called by _mesa_update_state) removes
the _math_matrix_analyse call, reducing the time spent
in _mesa_update_state.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Mon, 1 Feb 2021 18:17:13 +0000 (13:17 -0500)]
mesa: rework _MESA_NEW_NEED_EYE_COORDS to reduce fixed-func program updates
This eliminates a lot of the remaining no-op fixed-func program key
recomputations in _mesa_update_state.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Mon, 1 Feb 2021 09:07:18 +0000 (04:07 -0500)]
mesa: split _NEW_LIGHT into 3 flags: _NEW_LIGHT_(FF_PROGRAM|CONSTANTS|STATE)
_NEW_LIGHT_CONSTANTS: state parameters
_NEW_LIGHT_FF_PROGRAM: keys for fixed-func programs
_NEW_LIGHT_STATE: gallium rasterizer state
This reduces:
- the number of no-op fixed-func program key recomputations
in _mesa_update_state
- the number of no-op rasterizer state updates in st/mesa
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Mon, 1 Feb 2021 05:48:03 +0000 (00:48 -0500)]
mesa: move fixed-func-related _mesa_update_state code closer together
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Marek Olšák [Mon, 1 Feb 2021 04:10:26 +0000 (23:10 -0500)]
mesa: remove unnecessary NewState flagging for glPopAttrib(GL_ENABLE_BIT)
pop_enable_group calls _mesa_set_enable for every state it changes,
so we don't need do anything else.
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8850>
Bas Nieuwenhuizen [Fri, 26 Feb 2021 22:59:52 +0000 (23:59 +0100)]
ac/rgp: Only report double the prims per clock on GFX10.
Misinterpreted review comment.
Fixes:
4ded99f99dd ("ac/rgp: report the number of primitives per clock")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9312>
Rhys Perry [Fri, 26 Feb 2021 09:08:56 +0000 (09:08 +0000)]
nir: fix build at -O1
At -O1 with GCC 10.2.1, _nir_visit_dest_indirect (declared ALWAYS_INLINE)
will fail to inline if it's caller (nir_foreach_dest) is not inlined,
because _nir_visit_dest_indirect is passed as a function pointer. This
results in a compilation error.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Reviewed-by: Witold Baryluk <witold.baryluk@gmail.com>
Fixes:
336bcbacd05 ("nir: inline nir_foreach_{src,dest}")
Tested-by: Witold Baryluk <witold.baryluk@gmail.com>
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4353
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9301>
Christian Gmeiner [Fri, 26 Feb 2021 21:05:42 +0000 (22:05 +0100)]
gallium: call util_cpu_detect()
Fix undefined behavior from using util_cpu_caps.
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9311>
Danylo Piliaiev [Fri, 26 Feb 2021 20:24:37 +0000 (22:24 +0200)]
turnip/ir3: check for bindless IBOs in atomic dests fixup
Otherwise destinations may remain unfixed because ir3_shader_nibo
doesn't count bindless IBOs.
Fixes tests:
dEQP-VK.image.atomic_operations.*intermediate_values
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9309>
Tamara Schmitz [Thu, 25 Feb 2021 21:30:18 +0000 (21:30 +0000)]
util: add mesa_glthread for Valheim in OpenGL mode.
Drastically reduces hitching when traversing the landscape.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9296>
Christian Gmeiner [Fri, 26 Feb 2021 08:05:57 +0000 (09:05 +0100)]
etnaviv: extend lower ubo tests
Test a full transformation path (load_uniform -> load_ubo -> load_uniform)
and validate the load_uniform offset.
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Acked-by: Lucas Stach <l.stach@pengutronix.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9305>
Christian Gmeiner [Thu, 25 Feb 2021 13:01:28 +0000 (14:01 +0100)]
etnaviv: fix etna_nir_lower_ubo_to_uniform pass
The restoring of the acutal uniform offset was wrong.
Fixes:
1837135f7ca ("etnaviv: nir: add ubo lowering pass")
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Acked-by: Lucas Stach <l.stach@pengutronix.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9305>
Adam Jackson [Thu, 25 Feb 2021 18:58:37 +0000 (13:58 -0500)]
docs: Document GL_EXT_depth_bounds_test
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9287>
Adam Jackson [Thu, 25 Feb 2021 00:20:20 +0000 (19:20 -0500)]
softpipe: Implement GL_EXT_depth_bounds_test
This is a little bit contorted because the Z storage for the tile is
either float or int depending on the Z format, so we have to be careful
about types when comparing.
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9287>
Adam Jackson [Wed, 24 Feb 2021 22:31:53 +0000 (17:31 -0500)]
softpipe: Fix depth comparison with float Z formats
We just stuff the Z bits into [bq]zzzz literally for floats, but
comparing those like they're integers only works for == and !=.
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9287>
Adam Jackson [Thu, 25 Feb 2021 00:57:50 +0000 (19:57 -0500)]
mesa: Store depth bounds test bounds as GLclampd
... instead of truncating to GLfloat. This seems somewhat silly since
the "clamp" part means only values [0.0, 1.0] are defined, but if the
depth buffer is Z32_UNORM then storing as GLfloat means you lose 8 bits
of depth bounds precision. This happens not to matter, yet, since swrast
classic doesn't support Z32_UNORM for depth, and the software gallium
drivers don't support EXT_depth_bounds_test. But the latter part is
about to change.
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9287>
Rob Clark [Thu, 25 Feb 2021 18:04:50 +0000 (10:04 -0800)]
util: Add accessor for util_cpu_caps
In release builds, there should be no change, but in debug builds the
assert will help us catch undefined behavior resulting from using
util_cpu_caps before it is initialized.
With fix for u_half_test for MSVC from Jesse Natalie squashed in.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9266>
Rob Clark [Wed, 24 Feb 2021 21:59:15 +0000 (13:59 -0800)]
util/u_queue: Ensure num_cpu_mask_bits is valid
I noticed that we were hitting this before st_create_context() called
util_cpu_detect() and so num_cpu_mask_bits was zero. But there is no
harm in calling util_cpu_detect(), so lets just call it here to be safe.
Fixes:
d877451b48a ("util/u_queue: add UTIL_QUEUE_INIT_SET_FULL_THREAD_AFFINITY")
Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9266>
Samuel Pitoiset [Fri, 26 Feb 2021 13:07:56 +0000 (14:07 +0100)]
ac/rgp: report the number of primitives per clock
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/9303>
Samuel Pitoiset [Fri, 26 Feb 2021 13:01:19 +0000 (14:01 +0100)]
ac/rgp: report the number of memory operations per clock
So that RGP reports the memory type and the memory throughput.
Based on AMDVLK.
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/9303>
Samuel Pitoiset [Fri, 26 Feb 2021 12:55:55 +0000 (13:55 +0100)]
ac/rgp: report LDS size in CU mode on GFX10+
RGP expects that.
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/9303>
Samuel Pitoiset [Fri, 26 Feb 2021 09:22:12 +0000 (09:22 +0000)]
ac/rgp: recognize more memory types
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/9303>
Samuel Pitoiset [Fri, 26 Feb 2021 09:18:07 +0000 (09:18 +0000)]
include/drm-uapi: bump AMDGPU headers
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/9303>
Gert Wollny [Mon, 4 Jan 2021 08:55:40 +0000 (09:55 +0100)]
r600/sfn: remove old cube texturing code
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9302>
Gert Wollny [Wed, 23 Dec 2020 21:42:45 +0000 (22:42 +0100)]
r600/sfn: use lowering pass for cube textures
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9302>
Gert Wollny [Thu, 24 Dec 2020 15:48:00 +0000 (16:48 +0100)]
r600/sfn: use lower bool to int32 and lower int_tg4 only on shader clone
These changes should not be visible to shader variants that may go through
the optimization another time.
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9302>
Gert Wollny [Mon, 4 Jan 2021 08:47:38 +0000 (09:47 +0100)]
r600/sfn: fix gather with cube lowering
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9302>
Gert Wollny [Wed, 23 Dec 2020 21:42:26 +0000 (22:42 +0100)]
r600/sfn: add lowering pass for cube textures
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9302>
Gert Wollny [Wed, 23 Dec 2020 21:42:07 +0000 (22:42 +0100)]
r600/sfn: Add support for cube_r600 instruction
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9302>
Rhys Perry [Fri, 26 Feb 2021 09:45:07 +0000 (09:45 +0000)]
aco: use p_as_uniform for get_sampler_desc and convert_pointer_to_64_bit
Since value-numbering no longer works across loops, we no longer need to
use v_readfirstlane_b32.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9288>
Rhys Perry [Thu, 25 Feb 2021 15:37:17 +0000 (15:37 +0000)]
aco: calculate all p_as_uniform and v_readfirstlane_b32 sources in WQM
We should avoid a situation where a v_readfirstlane_b32 is in WQM but it's
source is calculated in Exact.
Fixes hang when running Assassin's Creed: Valhalla benchmark.
fossil-db (GFX10.3):
Totals from 1021 (0.70% of 146267) affected shaders:
CodeSize: 7835228 -> 7842992 (+0.10%); split: -0.00%, +0.10%
Instrs: 1519208 -> 1521149 (+0.13%); split: -0.00%, +0.13%
SClause: 78921 -> 78920 (-0.00%)
Copies: 44456 -> 45421 (+2.17%); split: -0.05%, +2.22%
Branches: 12987 -> 13933 (+7.28%)
PreSGPRs: 47599 -> 47813 (+0.45%)
Cycles:
10037540 ->
10045304 (+0.08%); split: -0.00%, +0.08%
VMEM: 538381 -> 538777 (+0.07%); split: +0.11%, -0.03%
SMEM: 84553 -> 84554 (+0.00%); split: +0.01%, -0.01%
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9288>
Juan A. Suarez Romero [Thu, 25 Feb 2021 11:21:57 +0000 (12:21 +0100)]
ci/piglit: fix parallel piglit results
Commit
bcea453d4a6 removed the summary from the expected piglit
results, but handling of results when using parallel jobs was also doing
the same, which ends up on removing too many lines from results.
Fixes:
bcea453d4a6 ("ci/piglit: Stop including the test counts at the
end of expectations.")
Acked-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9286>
Gert Wollny [Wed, 23 Dec 2020 15:19:31 +0000 (16:19 +0100)]
nir: Add r600 specific CUBE opcode to evaluate cube texture coords and
face
The opcode evaluates tha unnormalized coordinates, the length of the
major axis, and the cube face.
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Acked-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9200>
Gert Wollny [Wed, 13 Jan 2021 09:28:59 +0000 (10:28 +0100)]
nir: Add flag to tex instruction to indicate lowering cube to array
E.g. r600 a cube texture lookup uses a specific cube instruction
to evaluate the sample coordinates and the face ID, so that the cube
texture lookup can be lowered to a array texture lookup, thereby sharing
the code with the 2D array texture lopkup.
However, for TXD the given gradients still need to be three-component
vectors, so add a flag that the NIR validation knows that we deal with
cube texture that was lowered to an array and can validate accordingly.
v2: Handle new flag in serialization (Marek)
v3: Rebase so that the change does not require the patch to deduct the
number of offset and grad components from sampler type
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com> (v2)
Acked-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9200>
Mike Blumenkrantz [Wed, 7 Oct 2020 16:46:38 +0000 (12:46 -0400)]
zink: use pre-fetched format properties everywhere
this is a noticeable perf improvement
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9293>
Mike Blumenkrantz [Wed, 7 Oct 2020 16:45:04 +0000 (12:45 -0400)]
zink: pre-fetch all format properties during screen init
this ends up being a tradeoff where we waste a little startup time and
an extra ~4k memory for the overall screen object in exchange for never having
to fetch format properties again, which is a surprisingly expensive call
to be making as much as we have to make it
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9293>