mesa.git
6 years agoac/radv: cleanup some tcs output values access
Dave Airlie [Mon, 19 Feb 2018 06:19:07 +0000 (06:19 +0000)]
ac/radv: cleanup some tcs output values access

Just consolidates some code to make it easier to change.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
6 years agoac/radv: remove total_vertices variable
Dave Airlie [Mon, 19 Feb 2018 06:53:21 +0000 (06:53 +0000)]
ac/radv: remove total_vertices variable

This just removes an unneeded variable.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
6 years agoac/radv: don't mark tess inner as used if we don't use it.
Dave Airlie [Mon, 19 Feb 2018 20:33:17 +0000 (20:33 +0000)]
ac/radv: don't mark tess inner as used if we don't use it.

This just avoids marking it as a used output if we don't
actually use it.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
6 years agoac/nir: to integer the args to bcsel.
Dave Airlie [Tue, 20 Feb 2018 00:15:18 +0000 (10:15 +1000)]
ac/nir: to integer the args to bcsel.

dEQP-VK.tessellation.invariance.outer_edge_symmetry.triangles_equal_spacing_ccw
was hitting an llvm assert due to one value being an int and the
other a float.

This just casts both values to integer and fixes the test.

Fixes: dEQP-VK.tessellation.invariance.outer_edge_symmetry.triangles_equal_spacing_ccw
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>
6 years agoanv/blorp: Use layout_to_aux_usage when a layout is provided
Jason Ekstrand [Fri, 2 Feb 2018 22:51:56 +0000 (14:51 -0800)]
anv/blorp: Use layout_to_aux_usage when a layout is provided

Instead of having aux usage and ANV_AUX_USAGE_DEFAULT to mean "give me
something reasonable" we now use anv_layout_to_aux_usage whenever a
layout is available.  If a layout is available, we ignore the aux_usage
parameter.  For the cases where we have an explicit aux usage such as
clears and aux ops, we have a new ANV_IMAGE_LAYOUT_EXPLICIT_AUX layout.

Reviewed-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
6 years agoanv/cmd_buffer: Delete some assert-only variables
Jason Ekstrand [Fri, 2 Feb 2018 04:02:48 +0000 (20:02 -0800)]
anv/cmd_buffer: Delete some assert-only variables

Checking the sample count is almost as good as aux usage in this case.

Reviewed-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
6 years agoanv/cmd_buffer: Use layout_to_* helpers in compute_aux_usage
Jason Ekstrand [Fri, 2 Feb 2018 03:36:22 +0000 (19:36 -0800)]
anv/cmd_buffer: Use layout_to_* helpers in compute_aux_usage

Reviewed-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
6 years agoanv/cmd_buffer: Simplify transition_depth_buffer
Jason Ekstrand [Fri, 2 Feb 2018 03:13:12 +0000 (19:13 -0800)]
anv/cmd_buffer: Simplify transition_depth_buffer

If we don't have HiZ, then anv_layout_to_aux_usage will return NONE for
both layouts.  If the two layouts are the same, they will get the aux
usage.  In either case, the code below will give us ISL_AUX_OP_NONE and
we'll return without doing anything.

Reviewed-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
6 years agoanv/cmd_buffer: Do subpass image transitions in begin/end_subpass
Jason Ekstrand [Tue, 21 Nov 2017 23:56:35 +0000 (15:56 -0800)]
anv/cmd_buffer: Do subpass image transitions in begin/end_subpass

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
6 years agoanv/cmd_buffer: Mark depth/stencil surfaces written in begin_subpass
Jason Ekstrand [Sat, 13 Jan 2018 18:59:05 +0000 (10:59 -0800)]
anv/cmd_buffer: Mark depth/stencil surfaces written in begin_subpass

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
6 years agoanv/cmd_buffer: Sync clear values in begin_subpass
Jason Ekstrand [Tue, 21 Nov 2017 23:16:40 +0000 (15:16 -0800)]
anv/cmd_buffer: Sync clear values in begin_subpass

This is quite a bit cleaner because we now sync the clear values at the
same time as we do the fast clear.  For loading the clear values into
the surface state, we now do it once when we handle the LOAD_OP_LOAD
instead of every subpass.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
6 years agoanv/pass: Store usage in each subpass attachment
Jason Ekstrand [Sat, 13 Jan 2018 18:45:55 +0000 (10:45 -0800)]
anv/pass: Store usage in each subpass attachment

This requires us to ditch the VkAttachmentReference struct in favor of
an anv-specific struct.  However, we can now easily identify from just
the subpass attachment what kind of an attachment it is.  This will make
iteration over anv_subpass::attachments a little easier in some case.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
6 years agoanv/cmd_buffer: Add a concept of pending load aspects
Jason Ekstrand [Wed, 22 Nov 2017 04:29:36 +0000 (20:29 -0800)]
anv/cmd_buffer: Add a concept of pending load aspects

These are the same as pending clear aspects only for the "load"
operation.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
6 years agoanv/cmd_buffer: Iterate all subpass attachments when clearing
Jason Ekstrand [Mon, 27 Nov 2017 18:43:03 +0000 (10:43 -0800)]
anv/cmd_buffer: Iterate all subpass attachments when clearing

This unifies things a bit because we now handle depth and stencil at the
same time.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
6 years agoanv/cmd_buffer: Decide whether or not to HiZ clear up-front
Jason Ekstrand [Mon, 27 Nov 2017 18:20:00 +0000 (10:20 -0800)]
anv/cmd_buffer: Decide whether or not to HiZ clear up-front

This moves the decision out of begin_subpass and into BeginRenderPass
like the decision for color clears.  We use a similar name for the
function for depth/stencil as for color even though no aux usage is
really getting computed.

v2 (Jason Ekstrand):
 - Don't always disable HiZ clears by accident
 - Use the initial layout to decide whether to do fast clears

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
6 years agoanv/cmd_buffer: Move the rest of clear_subpass into begin_subpass
Jason Ekstrand [Tue, 21 Nov 2017 22:46:25 +0000 (14:46 -0800)]
anv/cmd_buffer: Move the rest of clear_subpass into begin_subpass

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
6 years agointel/blorp: Add a blorp_hiz_clear_depth_stencil helper
Jason Ekstrand [Tue, 21 Nov 2017 22:00:44 +0000 (14:00 -0800)]
intel/blorp: Add a blorp_hiz_clear_depth_stencil helper

This is similar to blorp_gen8_hiz_clear_attachments except that it takes
actual images instead of trusting in the already set depth state.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
6 years agoanv/cmd_buffer: Move the color portion of clear_subpass into begin_subpass
Jason Ekstrand [Tue, 21 Nov 2017 21:30:49 +0000 (13:30 -0800)]
anv/cmd_buffer: Move the color portion of clear_subpass into begin_subpass

This doesn't really change much now but it will give us more/better
control over clears in the future.  The one interesting functional
change here is that we are now re-emitting 3DSTATE_DEPTH_BUFFERS and
friends for each clear.  However, this only happens at begin_subpass
time so it shouldn't be substantially more expensive.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
6 years agoanv/cmd_buffer: Pass a subpass id into begin_subpass
Jason Ekstrand [Tue, 21 Nov 2017 20:42:45 +0000 (12:42 -0800)]
anv/cmd_buffer: Pass a subpass id into begin_subpass

This is a bit less awkward than passing in the subpass because it means
we don't have to extract the subpass id from the subpass.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
6 years agoanv/cmd_buffer: Add begin/end_subpass helpers
Jason Ekstrand [Tue, 21 Nov 2017 20:41:01 +0000 (12:41 -0800)]
anv/cmd_buffer: Add begin/end_subpass helpers

Having begin/end_subpass is a bit nicer than the begin/next/end hooks
that Vulkan gives us.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
6 years agoanv/cmd_buffer: Apply subpass flushes before set_subpass
Jason Ekstrand [Tue, 21 Nov 2017 20:27:43 +0000 (12:27 -0800)]
anv/cmd_buffer: Apply subpass flushes before set_subpass

This seems slightly more correct because it means that the flushes
happen before any clears or resolves implied by the subpass transition.

Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
6 years agoanv: Use framebuffer layers for implicit subpass transitions
Jason Ekstrand [Fri, 9 Feb 2018 00:44:56 +0000 (16:44 -0800)]
anv: Use framebuffer layers for implicit subpass transitions

Fixes: de3be618016 "anv/cmd_buffer: Rework aux tracking"
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
6 years agoanv: Be more careful about fast-clear colors
Jason Ekstrand [Tue, 13 Feb 2018 00:03:28 +0000 (16:03 -0800)]
anv: Be more careful about fast-clear colors

Previously, we just used all the channels regardless of the format.
This is less than ideal because some channels may have undefined values
and this should be ok from the client's perspective.  Even though the
driver should do the correct thing regardless of what is in the
undefined value, it makes things less deterministic.  In particular, the
driver may choose to fast-clear or not based on undefined values.  This
level of nondeterminism is bad.

Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
6 years agointel/isl: Add an isl_color_value_is_zero helper
Jason Ekstrand [Mon, 12 Feb 2018 23:50:12 +0000 (15:50 -0800)]
intel/isl: Add an isl_color_value_is_zero helper

Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
6 years agoanv/gpu_memcpy: CS Stall before a MI memcpy on gen7
Jason Ekstrand [Sat, 17 Feb 2018 01:35:15 +0000 (17:35 -0800)]
anv/gpu_memcpy: CS Stall before a MI memcpy on gen7

This fixes a pile of hangs caused by the recent shuffling of resolves
and transitions.  The particularly problematic case is when you have at
least three attachments with load ops of CLEAR, LOAD, CLEAR.  In this
case, we execute the first CLEAR followed by a MI memcpy to copy the
clear values over for the LOAD followed by a second CLEAR.  The MI
commands cause the first CLEAR to hang which causes us to get stuck on
the 3DSTATE_MULTISAMPLE in the second CLEAR.

We also add guards for BLORP to fix the same issue.  These shouldn't
actually do anything right now because the only use of indirect clears
in BLORP today is for resolves which are already guarded by a render
cache flush and CS stall.  However, this will guard us against potential
issues in the future.

Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Nanley Chery <nanley.g.chery@intel.com>
6 years agost/mesa: Factorize duplicate code for atomic buffer binding
Guillaume Charifi [Tue, 20 Feb 2018 11:49:28 +0000 (12:49 +0100)]
st/mesa: Factorize duplicate code for atomic buffer binding

Signed-off-by: Guillaume Charifi <guillaume.charifi@sfr.fr>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
6 years agost/mesa: Factorize duplicate code in st_update_framebuffer_state()
Guillaume Charifi [Fri, 5 Jan 2018 16:49:39 +0000 (17:49 +0100)]
st/mesa: Factorize duplicate code in st_update_framebuffer_state()

Signed-off-by: Guillaume Charifi <guillaume.charifi@sfr.fr>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
6 years agofreedreno/ir3: fix use_count refcnt'ing issue
Rob Clark [Tue, 20 Feb 2018 18:40:46 +0000 (13:40 -0500)]
freedreno/ir3: fix use_count refcnt'ing issue

Was hitting an assert with vs-varying-array-mat4-index-col-row-wr.shader_test

When eliminating a copy, we were dropping the use_count of the mov that
is skipped, but not increasing the use_count of it's src instruction.

Fixes: 76440fcca91 freedreno/ir3: clean up dangling false-dep's
Signed-off-by: Rob Clark <robdclark@gmail.com>
6 years agodocs: fix patent url
Eric Engestrom [Tue, 20 Feb 2018 13:35:56 +0000 (13:35 +0000)]
docs: fix patent url

Reported-by: Pierre Moreau <pierre.morrow@free.fr>
Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
6 years agosvga: replaced 'unsigned' with proper enum types in shader code
Brian Paul [Fri, 16 Feb 2018 20:57:51 +0000 (13:57 -0700)]
svga: replaced 'unsigned' with proper enum types in shader code

Reviewed-by: Charmaine Lee <charmainel@vmware.com>
6 years agoconfigure.ac: pthread-stubs not present on OpenBSD
Jonathan Gray [Tue, 20 Feb 2018 06:38:00 +0000 (17:38 +1100)]
configure.ac: pthread-stubs not present on OpenBSD

pthread-stubs is no longer required on OpenBSD and has been removed.
libpthread parts involved moved to libc.

Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Cc: 17.3 18.0 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
6 years agoswr: bump minimum supported LLVM version to 4.0
Andres Gomez [Tue, 13 Feb 2018 22:42:57 +0000 (00:42 +0200)]
swr: bump minimum supported LLVM version to 4.0

Since radv and radeonsi removed support for LLVM 3.9 the distcheck
target got broken because SWR distribution needed 3.9.x.

After checking with George Kyriazis, SWR is OK with moving to LLVM 4.0
and above, which will solve this problem.

Fixes: 3bf1e036e8a ("amd: remove support for LLVM 3.9")
Cc: George Kyriazis <george.kyriazis@intel.com>
Cc: Tim Rowley <timothy.o.rowley@intel.com>
Cc: Emil Velikov <emil.velikov@collabora.com>
Cc: Dylan Baker <dylan@pnwbakers.com>
Cc: Eric Engestrom <eric.engestrom@imgtec.com>
Signed-off-by: Andres Gomez <agomez@igalia.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
Reviewed-by: George Kyriazis <george.kyriazis@intel.com>
6 years agotravis: radeonsi and radv need LLVM 4.0
Andres Gomez [Tue, 6 Feb 2018 15:42:42 +0000 (17:42 +0200)]
travis: radeonsi and radv need LLVM 4.0

Fixes: 3bf1e036e8a ("amd: remove support for LLVM 3.9")
Cc: Marek Olšák <marek.olsak@amd.com>
Cc: Emil Velikov <emil.velikov@collabora.com>
Cc: Jan Vesely <jan.vesely@rutgers.edu>
Signed-off-by: Andres Gomez <agomez@igalia.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
6 years agoac/nir: move ac_declare_lds_as_pointer() outside of the switch
Samuel Pitoiset [Fri, 16 Feb 2018 09:33:10 +0000 (10:33 +0100)]
ac/nir: move ac_declare_lds_as_pointer() outside of the switch

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
6 years agoradv: allow to force family using RADV_FORCE_FAMILY
Samuel Pitoiset [Fri, 16 Feb 2018 10:00:14 +0000 (11:00 +0100)]
radv: allow to force family using RADV_FORCE_FAMILY

Useful for pipeline-db.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
6 years agoloader_dri3/glx/egl: Reinstate the loader_dri3_vtable get_dri_screen callback
Thomas Hellstrom [Fri, 9 Feb 2018 08:37:19 +0000 (09:37 +0100)]
loader_dri3/glx/egl: Reinstate the loader_dri3_vtable get_dri_screen callback

Removing this callback caused rendering corruption in some multi-screen cases,
so it is reinstated but without the drawable argument which was never used
by implementations and was confusing since the drawable could have been
created with another screen.

Cc: "17.3 18.0" mesa-stable@lists.freedesktop.org
Fixes: 5198e48a0d (loader_dri3/glx/egl: Remove the loader_dri3_vtable get_dri_screen callback)
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=105013
Reported-by: Daniel van Vugt <daniel.van.vugt@canonical.com>
Tested-by: Timo Aaltonen <tjaalton@ubuntu.com>
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
6 years agosvga: Fix a leftover debug hack
Thomas Hellstrom [Mon, 15 Jan 2018 11:51:27 +0000 (12:51 +0100)]
svga: Fix a leftover debug hack

Fix what appears to be a leftover debug hack.
The hack would force the driver to take a different blit path; possibly,
although unverified, reverting to software blits.

Tested using piglit tests/quick. No related regressions.

Cc: "17.2 17.3 18.0" <mesa-stable@lists.freedesktop.org>
Fixes: 9d81ab7376 (svga: Relax the format checks for copy_region_vgpu10 somewhat)
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=104625
Reported-by: Grazvydas Ignotas <notasas@gmail.com>
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
6 years agoanv/entrypoints: make vkGetDeviceProcAddr return NULL for instance commands
Iago Toral Quiroga [Wed, 7 Feb 2018 08:21:47 +0000 (09:21 +0100)]
anv/entrypoints: make vkGetDeviceProcAddr return NULL for instance commands

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
6 years agonv50,nvc0: mark ABGR format as displayable instead of ARGB format
Ilia Mirkin [Sun, 31 Dec 2017 07:39:11 +0000 (02:39 -0500)]
nv50,nvc0: mark ABGR format as displayable instead of ARGB format

This matches the hardware's capabilities.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
6 years agost/dri: only expose config formats that are display targets
Ilia Mirkin [Sun, 31 Dec 2017 07:36:39 +0000 (02:36 -0500)]
st/dri: only expose config formats that are display targets

In the case of NVIDIA hardware, ABGR is displayable but ARGB is not.
Only advertise the one set in the visuals list.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Daniel Stone <daniels@collabora.com>
6 years agomesa: add xbgr support adjacent to xrgb
Ilia Mirkin [Sun, 31 Dec 2017 06:05:06 +0000 (01:05 -0500)]
mesa: add xbgr support adjacent to xrgb

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Acked-by: Daniel Stone <daniels@collabora.com>
6 years agost/shader_cache: copy nir pointer to gl_program after deserializing
Timothy Arceri [Fri, 16 Feb 2018 00:41:17 +0000 (11:41 +1100)]
st/shader_cache: copy nir pointer to gl_program after deserializing

This fixes a crash when running the arb_get_program_binary-api-errors
piglit test twice.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
6 years agoradeonsi: add nir shader cache support
Timothy Arceri [Thu, 15 Feb 2018 23:14:05 +0000 (10:14 +1100)]
radeonsi: add nir shader cache support

In future we might want to try avoid calling nir_serialize() but
this works for now.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
6 years agoradeonsi: rename variables tgsi_binary -> ir_binary
Timothy Arceri [Thu, 15 Feb 2018 05:58:07 +0000 (16:58 +1100)]
radeonsi: rename variables tgsi_binary -> ir_binary

This better represents that the ir could be either tgsi or nir.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
6 years agodocs: update calendar, add news and link release notes to 17.3.5
Emil Velikov [Mon, 19 Feb 2018 22:10:18 +0000 (22:10 +0000)]
docs: update calendar, add news and link release notes to 17.3.5

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
6 years agodocs: add sha256 checksums for 17.3.5
Emil Velikov [Mon, 19 Feb 2018 22:07:23 +0000 (22:07 +0000)]
docs: add sha256 checksums for 17.3.5

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
(cherry picked from commit 164a993112cc7278d46b7ec8f7f617eb683b212c)

6 years agodocs: add release notes for 17.3.5
Emil Velikov [Mon, 19 Feb 2018 22:01:35 +0000 (22:01 +0000)]
docs: add release notes for 17.3.5

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
(cherry picked from commit 2529d77179065b983d69c620c7f71281aefe4f98)

6 years agoradeonsi: fix regression from 32-bit pointers on CI
Marek Olšák [Mon, 19 Feb 2018 16:55:34 +0000 (17:55 +0100)]
radeonsi: fix regression from 32-bit pointers on CI

Tested-by: Michel Dänzer <michel.daenzer@amd.com>
6 years agoradv: compact varyings after removing unused ones
Samuel Pitoiset [Fri, 16 Feb 2018 09:28:37 +0000 (10:28 +0100)]
radv: compact varyings after removing unused ones

It makes no sense to compact before, and the description of
nir_compact_varyings() confirms that.

Polaris10:
Totals from affected shaders:
SGPRS: 108528 -> 108128 (-0.37 %)
VGPRS: 74548 -> 74500 (-0.06 %)
Spilled SGPRs: 844 -> 814 (-3.55 %)
Code Size: 3007328 -> 2992932 (-0.48 %) bytes
Max Waves: 16019 -> 16009 (-0.06 %)

Vega10:
Totals from affected shaders:
SGPRS: 106088 -> 106232 (0.14 %)
VGPRS: 74652 -> 74700 (0.06 %)
Spilled SGPRs: 692 -> 658 (-4.91 %)
Code Size: 2967708 -> 2953028 (-0.49 %) bytes
Max Waves: 18178 -> 18162 (-0.09 %)

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
6 years agoradeonsi/nir: fix gl_FragCoord for pixel_center_integer
Timothy Arceri [Fri, 16 Feb 2018 05:14:29 +0000 (16:14 +1100)]
radeonsi/nir: fix gl_FragCoord for pixel_center_integer

Fixes piglit test glsl-arb-fragment-coord-conventions

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
6 years agoglsl/nir: add pixel_center_integer to shader info
Timothy Arceri [Fri, 16 Feb 2018 05:10:58 +0000 (16:10 +1100)]
glsl/nir: add pixel_center_integer to shader info

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
6 years agogm107/ir: avoid using kepler instruction capabilities
Ilia Mirkin [Sat, 10 Feb 2018 18:39:56 +0000 (13:39 -0500)]
gm107/ir: avoid using kepler instruction capabilities

Split up the op properties table into generation-specific bits, and only
use the kepler ones on kepler. Fixes some CTS images tests.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
6 years agonvc0: add support for bindless on maxwell+
Ilia Mirkin [Sat, 13 Jan 2018 17:32:41 +0000 (12:32 -0500)]
nvc0: add support for bindless on maxwell+

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
6 years agogm107/ir: change how SUQ works in preparation for bindless
Ilia Mirkin [Sat, 13 Jan 2018 17:28:16 +0000 (12:28 -0500)]
gm107/ir: change how SUQ works in preparation for bindless

All this information can be retrieved from the TIC directly. Avoid
having to dip into the constbuf information about the image.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
6 years agoi965: Use absolute addressing for constant buffer 0 on Kernel 4.16+.
Kenneth Graunke [Fri, 15 Aug 2014 05:36:45 +0000 (22:36 -0700)]
i965: Use absolute addressing for constant buffer 0 on Kernel 4.16+.

By default, 3DSTATE_CONSTANT_* Constant Buffer 0 is relative to dynamic
state base address.  This makes it unusable for pushing UBOs.

There is a bit in the INSTPM register (or CS_DEBUG_MODE2 on Skylake)
which controls whether buffer 0 is relative to dynamic state base
address, or simply a normal pointer.  Setting that gives us full
flexibility.  This lets us push up to 4 UBO ranges.

We can't currently write this on Haswell and earlier, and will need
to update the kernel command parser, and then do the whole version
checking song and dance.  We also need a brand new kernel that supports
context isolation - on older kernels, newly created contexts inherit
register state from whatever happened to be running.  So, setting this
would have catastrophic impact on other drivers such as libva, Beignet,
or older Mesa.

See commit 8ec5a4e4a4a32f4de351c5fc2bf0eb615b6eef1b where we did this
once before, but had to revert it in commit 013d33122028f2492da90a03a.

Reviewed-by: Francisco Jerez <currojerez@riseup.net>
6 years agoi965: Stop restoring the default L3 configuration on Kernel 4.16+.
Kenneth Graunke [Fri, 16 Feb 2018 19:03:58 +0000 (11:03 -0800)]
i965: Stop restoring the default L3 configuration on Kernel 4.16+.

Kernel 4.16 has proper context isolation, which means we can change
the L3 configuration without worrying about that leaking to other
newly created contexts, breaking the assumptions of other userspace.

So, disable our workaround to reprogram it back to the default.

Reviewed-by: Francisco Jerez <currojerez@riseup.net>
6 years agonvc0: Use GP100_COMPUTE_CLASS on GP10B
Mikko Perttunen [Thu, 15 Feb 2018 18:13:20 +0000 (20:13 +0200)]
nvc0: Use GP100_COMPUTE_CLASS on GP10B

GP10B requires the use of GP100_COMPUTE_CLASS instead of
GP104_COMPUTE_CLASS as is used for other non-GP100 chips.

Signed-off-by: Mikko Perttunen <mperttunen@nvidia.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
6 years agoi965: Fix aux-surface size check
Daniel Stone [Thu, 15 Feb 2018 15:04:51 +0000 (15:04 +0000)]
i965: Fix aux-surface size check

The previous commit reworked the checks intel_from_planar() to check the
right individual cases for regular/planar/aux buffers, and do size
checks in all cases.

Unfortunately, the aux size check was broken, and required the aux
surface to be allocated with the correct aux stride, but full image
height (!).

As the ISL aux surface is not recorded in the DRIimage, we cannot easily
access it to check. Instead, store the aux size from when we do have the
ISL surface to hand, and check against that later when we go to access
the aux surface.

Signed-off-by: Daniel Stone <daniels@collabora.com>
Fixes: c2c4e5bae3ba ("i965: Fix bugs in intel_from_planar")
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
6 years agoradeonsi: implement 32-bit pointers in user data SGPRs (v2)
Marek Olšák [Mon, 1 Jan 2018 20:04:22 +0000 (21:04 +0100)]
radeonsi: implement 32-bit pointers in user data SGPRs (v2)

User SGPRs changes:
    VS:     14 ->  9
    TCS:    14 -> 10
    TES:    10 ->  6
    GS:      8 ->  4
    GSCOPY:  2 ->  1
    PS:      9 ->  5
    Merged VS-TCS: 24 -> 16
    Merged VS-GS:  18 -> 11
    Merged TES-GS: 18 -> 11

SGPRS: 2170102 -> 2158430 (-0.54 %)
VGPRS: 1645656 -> 1641516 (-0.25 %)
Spilled SGPRs: 9078 -> 8810 (-2.95 %)
Spilled VGPRs: 130 -> 114 (-12.31 %)
Scratch size: 1508 -> 1492 (-1.06 %) dwords per thread
Code Size: 52094872 -> 52692540 (1.15 %) bytes
Max Waves: 371848 -> 372723 (0.24 %)

v2: - the shader cache needs to take address32_hi into account
    - set amdgpu-32bit-address-high-bits

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> (v1)
6 years agoradeonsi: disallow constant buffers with a 64-bit address in slot 0
Marek Olšák [Sun, 31 Dec 2017 21:58:57 +0000 (22:58 +0100)]
radeonsi: disallow constant buffers with a 64-bit address in slot 0

State trackers must use a user buffer or const_uploader,
or set pipe_resource::flags same as const_uploader->flags.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
6 years agoradeonsi: move const_uploader allocations to 32-bit address space
Marek Olšák [Sun, 31 Dec 2017 21:51:14 +0000 (22:51 +0100)]
radeonsi: move const_uploader allocations to 32-bit address space

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
6 years agowinsys/radeon: implement and enable 32-bit VM allocations
Marek Olšák [Sun, 31 Dec 2017 21:34:45 +0000 (22:34 +0100)]
winsys/radeon: implement and enable 32-bit VM allocations

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
6 years agowinsys/radeon: add struct radeon_vm_heap
Marek Olšák [Sun, 31 Dec 2017 21:02:35 +0000 (22:02 +0100)]
winsys/radeon: add struct radeon_vm_heap

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
6 years agowinsys/amdgpu: enable 32-bit VM allocations
Marek Olšák [Sun, 31 Dec 2017 20:36:37 +0000 (21:36 +0100)]
winsys/amdgpu: enable 32-bit VM allocations

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
6 years agogallium/radeon: add 32-bit address space heaps
Marek Olšák [Sun, 31 Dec 2017 20:32:36 +0000 (21:32 +0100)]
gallium/radeon: add 32-bit address space heaps

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
6 years agoac: query high bits of 32-bit address space
Marek Olšák [Fri, 2 Feb 2018 17:22:15 +0000 (18:22 +0100)]
ac: query high bits of 32-bit address space

6 years agogallium: use PIPE_CAP_CONSTBUF0_FLAGS
Marek Olšák [Sat, 27 Jan 2018 01:05:23 +0000 (02:05 +0100)]
gallium: use PIPE_CAP_CONSTBUF0_FLAGS

6 years agogallium: allow drivers to impose BO flags restrictions on constant buffer 0
Marek Olšák [Sat, 27 Jan 2018 00:52:08 +0000 (01:52 +0100)]
gallium: allow drivers to impose BO flags restrictions on constant buffer 0

Required by radeonsi for optimal behavior.

6 years agomeson: Add Haiku platform support v4
Alexander von Gluck IV [Fri, 16 Feb 2018 22:56:31 +0000 (16:56 -0600)]
meson: Add Haiku platform support v4

Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
6 years agoanv/icl: Add render target flush after uploading binding table
Anuj Phogat [Thu, 15 Feb 2018 23:35:42 +0000 (15:35 -0800)]
anv/icl: Add render target flush after uploading binding table

The PIPE_CONTROL command description says:

"Whenever a Binding Table Index (BTI) used by a Render Taget Message
points to a different RENDER_SURFACE_STATE, SW must issue a Render
Target Cache Flush by enabling this bit. When render target flush
is set due to new association of BTI, PS Scoreboard Stall bit must
be set in this packet."

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
6 years agoanv/icl: Enable float blend optimization
Anuj Phogat [Tue, 13 Feb 2018 21:48:34 +0000 (13:48 -0800)]
anv/icl: Enable float blend optimization

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
6 years agoanv/icl: Use gen11 functions
Anuj Phogat [Fri, 26 May 2017 19:32:23 +0000 (12:32 -0700)]
anv/icl: Use gen11 functions

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
6 years agoanv/icl: Build anv libs for gen11
Anuj Phogat [Fri, 26 May 2017 18:10:26 +0000 (11:10 -0700)]
anv/icl: Build anv libs for gen11

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
6 years agoanv/icl: Generate gen11 entry point functions
Anuj Phogat [Fri, 26 May 2017 19:31:58 +0000 (12:31 -0700)]
anv/icl: Generate gen11 entry point functions

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
6 years agoanv/icl: Don't use DISPATCH_MODE_SIMD4X2
Anuj Phogat [Fri, 26 May 2017 22:40:55 +0000 (15:40 -0700)]
anv/icl: Don't use DISPATCH_MODE_SIMD4X2

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
6 years agoanv/icl: Don't use SingleVertexDispatch
Anuj Phogat [Fri, 26 May 2017 22:42:02 +0000 (15:42 -0700)]
anv/icl: Don't use SingleVertexDispatch

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
6 years agoanv/icl: Don't set ResetGatewayTimer
Anuj Phogat [Fri, 26 May 2017 22:37:03 +0000 (15:37 -0700)]
anv/icl: Don't set ResetGatewayTimer

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
6 years agoanv/icl: Add #define genX
Anuj Phogat [Fri, 26 May 2017 17:57:45 +0000 (10:57 -0700)]
anv/icl: Add #define genX

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
6 years agoanv/icl: Add gen11 mocs defines
Anuj Phogat [Fri, 26 May 2017 17:55:22 +0000 (10:55 -0700)]
anv/icl: Add gen11 mocs defines

Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
6 years agoi965: Implement GenerateMipmap directly, rather than using Meta.
Kenneth Graunke [Sun, 19 Jun 2016 06:05:51 +0000 (23:05 -0700)]
i965: Implement GenerateMipmap directly, rather than using Meta.

Meta is awful and we'd like to stop using it.  Implementing this using
BLORP allows us to stop trashing a bunch of GL state every time.

This follows the structure of st_generate_mipmap().
compute_num_levels is lifted directly from there.

Improves performance in Gl41HdrBloom by about 11.794% +/- 1.01919% (n=3)
on Kabylake GT2 at 1280x720 (the difference seems much smaller at higher
resolutions).

v2 (idr): Don't try depth or depth-stencil blorp blits on Gen4 or Gen5
because it's not implemented yet.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
6 years agomesa: Move compute_num_levels from st_gen_mipmap.c to mipmap.c.
Kenneth Graunke [Sun, 19 Jun 2016 07:36:48 +0000 (00:36 -0700)]
mesa: Move compute_num_levels from st_gen_mipmap.c to mipmap.c.

I want to use compute_num_levels inside i965.  Rather than duplicating
it, move it from mesa/st to core Mesa, and make it non-static.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
6 years agomeson: freedreno depends on nir
Dylan Baker [Fri, 16 Feb 2018 17:07:25 +0000 (09:07 -0800)]
meson: freedreno depends on nir

This fixes a race condition in building targets that link in freedreno.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=105120
Fixes: 0bbecc5a8548883f76a7 ("meson: define driver dependencies")
Signed-off-by: Dylan Baker <dylan.c.baker@intel.com>
Acked-by: Mark Janes <mark.a.janes@intel.com>
6 years agoswr/rast: blend_epi32() should return Integer, not Float
George Kyriazis [Tue, 13 Feb 2018 16:41:48 +0000 (10:41 -0600)]
swr/rast: blend_epi32() should return Integer, not Float

fix gcc8 compiler error for KNL.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=105029
Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
6 years agoswr/rast: Normalize path for debug metadata
George Kyriazis [Tue, 13 Feb 2018 18:53:39 +0000 (12:53 -0600)]
swr/rast: Normalize path for debug metadata

in template gen_llvm.hpp

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
6 years agoswr/rast: Consolidate archrast Draw events
George Kyriazis [Mon, 12 Feb 2018 23:54:30 +0000 (17:54 -0600)]
swr/rast: Consolidate archrast Draw events

Consolidate archrst draw events into single draw event with an attribute
that represents the type of draw

- Add handlers for new private proto versions of DrawInstancedEvent,
  DrawIndexedInstancedEvent, DrawInstancedSplitEvent, and
  DrawIndexedInstancedSplitEvent
- Convert the draw events to generic DrawInfoEvents
- parse_proto_event_fields() replaces 'AR_DRAW_TYPE' as a field type with
  'uint32_t'. This draw type is actually an enum, but can be represented
  as an unsigned integer.
- is_draw_or_dispatch() recognizes DrawInfoEvent as a draw event

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
6 years agoswr/rast: Add semantics for translating address
George Kyriazis [Fri, 9 Feb 2018 20:04:14 +0000 (14:04 -0600)]
swr/rast: Add semantics for translating address

Added support for another full translation path in fetch jitter.

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
6 years agoswr/rast: Convert C Sampler intrinsics
George Kyriazis [Fri, 9 Feb 2018 17:37:17 +0000 (11:37 -0600)]
swr/rast: Convert C Sampler intrinsics

Convert portions of the C sampler to the rasty SIMD lib.

Also fix SRL call with a non-immediate.  Don't count on the compiler
automagically converting an srli call to srl if the shift count isn't
an immediate.

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
6 years agoswr/rast: Make SIMDLib templated types easier to use
George Kyriazis [Wed, 7 Feb 2018 22:51:41 +0000 (16:51 -0600)]
swr/rast: Make SIMDLib templated types easier to use

"typename SIMD_T::TypeName" --> "TypeName<SIMD_T>"

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
6 years agoswr/rast: Be more explicit when fetching next component
George Kyriazis [Wed, 7 Feb 2018 07:02:00 +0000 (01:02 -0600)]
swr/rast: Be more explicit when fetching next component

Use a new function to denote that we want to get offset to next component
and hide the fact that GEP is used underneath.

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
6 years agoswr/rast: Fix bug related to passing AR handle
George Kyriazis [Wed, 7 Feb 2018 06:39:54 +0000 (00:39 -0600)]
swr/rast: Fix bug related to passing AR handle

We were passing a garbage handle. Let's not do that.

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
6 years agoswr/rast: Fix primitive replication issue in tesselation PA.
George Kyriazis [Tue, 6 Feb 2018 23:28:12 +0000 (17:28 -0600)]
swr/rast: Fix primitive replication issue in tesselation PA.

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
6 years agoswr/rast: Use llvm intrinsic masked gather
George Kyriazis [Fri, 2 Feb 2018 23:03:01 +0000 (17:03 -0600)]
swr/rast: Use llvm intrinsic masked gather

Use llvm intrinsic masked.gather instead of manual unroll for the cases
where we have vector of pointers.  Improves llvm IR debug experience by
reducing a ton of IR to a single intrinsic call. Also seems to reduce
overall stack use considerably.

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
6 years agoswr/rast: Misc cleanup
George Kyriazis [Fri, 2 Feb 2018 07:12:29 +0000 (01:12 -0600)]
swr/rast: Misc cleanup

Together with correct detection of clipDistance NaNs when no cullDistance is set

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
6 years agoswr/rast: Renamed variable in vertexbufferstate
George Kyriazis [Fri, 2 Feb 2018 04:12:22 +0000 (22:12 -0600)]
swr/rast: Renamed variable in vertexbufferstate

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
6 years agoswr/rast: Fix GATHERPS to avoid assertions.
George Kyriazis [Mon, 12 Feb 2018 19:38:45 +0000 (13:38 -0600)]
swr/rast: Fix GATHERPS to avoid assertions.

With the pBase type change, LLVM was asserting because of wrong types.
Cast appropriately.

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
6 years agoswr/rast: More precise user clip distance interpolation
George Kyriazis [Fri, 2 Feb 2018 01:28:58 +0000 (19:28 -0600)]
swr/rast: More precise user clip distance interpolation

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
6 years agoswr/rast: Cull prims when all verts have negative clip distances
George Kyriazis [Thu, 1 Feb 2018 23:43:04 +0000 (17:43 -0600)]
swr/rast: Cull prims when all verts have negative clip distances

Performance optimization, and fixes some clipping issues.

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
6 years agoswr/rast: whitespace and comment cleanup
George Kyriazis [Thu, 1 Feb 2018 21:37:36 +0000 (15:37 -0600)]
swr/rast: whitespace and comment cleanup

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
6 years agoswr/rast: Fix invalid number of attributes
George Kyriazis [Wed, 31 Jan 2018 23:07:59 +0000 (17:07 -0600)]
swr/rast: Fix invalid number of attributes

Fix invalid number of attributes passed into tesselation PA.
Needs to take into account any offsets from the shader.
Innocuous issue, but removes an assert firing in debug.

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
6 years agoswr/rast: Add clipper stats.
George Kyriazis [Wed, 31 Jan 2018 19:22:19 +0000 (13:22 -0600)]
swr/rast: Add clipper stats.

Clipper event is now:

event ClipperEvent
{
    uint32_t drawId;
    uint32_t trivialRejectCount;
    uint32_t trivialAcceptCount;
    uint32_t mustClipCount;
};

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>