Danylo Piliaiev [Wed, 15 Aug 2018 12:46:23 +0000 (15:46 +0300)]
glsl: Add an assert when cloning ir_dereference_record with invalid field
Signed-off-by: Danylo Piliaiev <danylo.piliaiev@globallogic.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Danylo Piliaiev [Wed, 15 Aug 2018 12:46:22 +0000 (15:46 +0300)]
glsl: Avoid propagating incompatible type of initializer
do_assignment validated assigment but when rhs type was not compatible
it proceeded without issues and returned error_emitted = false.
On the other hand process_initializer expected do_assignment to always
return compatible type and never fail.
As a result when variable was initialized with incompatible type
the type of variable changed to the incompatible one.
This manifested in unnecessary error messages and in one case in crash.
Example GLSL:
vec4 tmp = vec2(0.0);
tmp.z -= 1.0;
Past error messages:
initializer of type vec2 cannot be assigned to variable of type vec4
invalid swizzle / mask `z'
type mismatch
operands to arithmetic operators must be numeric
After this patch:
initializer of type vec2 cannot be assigned to variable of type vec4
In the other case when we initialize variable with incompatible struct,
accessing variable's field leaded to a crash. Example:
uniform struct {float field;} data;
...
vec4 tmp = data;
tmp.x -= 1.0;
After the patch there is only error line without a crash:
initializer of type #anon_struct cannot be assigned to variable of
type vec4
Signed-off-by: Danylo Piliaiev <danylo.piliaiev@globallogic.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107547
Michal Srb [Thu, 15 Mar 2018 16:27:57 +0000 (17:27 +0100)]
st/dri: don't set queryDmaBufFormats/queryDmaBufModifiers if the driver does not implement it
This is equivalent to commit
a65db0ad1c3, but for dri_kms_init_screen. Without
this gbm_dri_is_format_supported always returns false.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=104926
Fixes: e14fe41e0bf ("st/dri: implement createImageFromRenderbuffer(2)")
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Tested-by: Adam Williamson <adamwill@fedoraproject.org>
Jason Ekstrand [Mon, 10 Sep 2018 21:36:10 +0000 (16:36 -0500)]
anv/so_memcpy: Don't consider src/dst_offset when computing block size
The only thing that matters is the size since we never specify any
offsets in terms of blocks.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Jakob Bornecrantz [Wed, 19 Sep 2018 14:21:26 +0000 (15:21 +0100)]
Revert "mesa: only update framebuffer-state for clears"
This reverts commit
fb86365148d5b8f3f06c5e42d9c8440fc1f6693f.
Samuel Pitoiset [Tue, 18 Sep 2018 14:18:37 +0000 (16:18 +0200)]
radv: use a 64-bit unsigned integer when allocating a descriptor pool
pool->size is a 64-bit unsigned integer too.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Tue, 18 Sep 2018 13:27:52 +0000 (15:27 +0200)]
radv: enable VK_SUBGROUP_FEATURE_ARITHMETIC_BIT
All CTS pass on Polaris/Vega with LLVM 6, 7 and master, so
I think it's safe to enable the feature.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Tue, 18 Sep 2018 13:06:42 +0000 (15:06 +0200)]
radv: do not support blitting surfaces with depth and stencil
Fixes:
dEQP-VK.api.copy_and_blit.core.blit_image.all_formats.depth_stencil.d32_sfloat_s8_uint_d32_sfloat_s8_uint.optimal_optimal_nearest
And all friends that try to blit a surface with different
depth and stencil formats.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Erik Faye-Lund [Mon, 10 Sep 2018 20:11:16 +0000 (22:11 +0200)]
mesa: only update framebuffer-state for clears
If we update the program-state etc, we risk compiling needless shaders,
which can cost quite a bit of performance.
Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Juan A. Suarez Romero [Wed, 19 Sep 2018 09:27:49 +0000 (11:27 +0200)]
nir: add initializer data to fix MSVC compile error
CC: Jason Ekstrand <jason@jlekstrand.net>
Fixes: 82799a5d1b8 ("nir: Add a small pass to rematerialize derefs
per-block")
Reviewed-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Jason Ekstrand [Tue, 11 Sep 2018 18:06:01 +0000 (13:06 -0500)]
nir: Add some asserts that we don't put derefs in phis
The lcssa and phis_to_regs passes are used by various NIR optimizations
that modify the CFG. Putting a couple of asserts will help ensure that
we don't accidentally put derefs in phis as part of an optimization
pass.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Jason Ekstrand [Tue, 11 Sep 2018 17:55:45 +0000 (12:55 -0500)]
nir/opt_if: Re-materialize derefs in use blocks before peeling loops
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107879
Cc: "18.2" <mesa-stable@lists.freedesktop.org>
Jason Ekstrand [Tue, 11 Sep 2018 17:51:09 +0000 (12:51 -0500)]
nir/loop_unroll: Re-materialize derefs in use blocks before unrolling
When we're about to re-arrange a bunch of blocks, it's a good idea to
make sure that we don't have deref uses crossing block boundaries.
Otherwise we may end up with a deref going through a phi and that would
be bad.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Cc: "18.2" <mesa-stable@lists.freedesktop.org>
Jason Ekstrand [Tue, 11 Sep 2018 17:15:22 +0000 (12:15 -0500)]
nir: Add a small pass to rematerialize derefs per-block
This pass re-materializes deref instructions on a per-block basis to
ensure that every use of a deref occurs in the same block as the
instruction which uses it.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Cc: "18.2" <mesa-stable@lists.freedesktop.org>
Kenneth Feng [Thu, 26 Jul 2018 02:53:33 +0000 (10:53 +0800)]
amd: Add Picasso device id
No changes here compared to Raven.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Kenneth Feng <kenneth.feng@amd.com>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Cc: 18.1 18.2 <mesa-stable@lists.freedesktop.org>
Bas Nieuwenhuizen [Tue, 18 Sep 2018 20:46:43 +0000 (22:46 +0200)]
Revert "radv: fix descriptor pool allocation size"
This reverts commit
90819abb56f6b1a0cd4946b13b6caf24fb46e500.
This logic was wrong, the original code is correct. The direct
impact is that we allocate up to approximately a squared amount
of memory compared to what we should allocate.
Acked-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Samuel Pitoiset [Mon, 17 Sep 2018 14:34:11 +0000 (16:34 +0200)]
radv: implement VK_EXT_conservative_rasterization
Only supported by GFX9+.
The conservativeraster Sascha demo seems to work as expected.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Mon, 17 Sep 2018 12:57:51 +0000 (14:57 +0200)]
radv: do not re-create the sampler for every blits in CmdBlitImage()
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Mon, 17 Sep 2018 20:23:19 +0000 (22:23 +0200)]
radv: allow to force anisotropy via RADV_TEX_ANISO
Ported from RadeonSI.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Timothy Arceri [Sat, 8 Sep 2018 04:20:17 +0000 (14:20 +1000)]
mesa: enable EXT_framebuffer_object in core profile
Since user defined names are not allowed in core profile
we remove the allow_user_names bool and just check if
we have a core profile like all other buffer/texture
object handling code does.
This extension is required by "Wolfenstein: The Old Blood"
and is exposed in core in the Nvidia binary driver.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Timothy Arceri [Thu, 30 Aug 2018 00:19:08 +0000 (10:19 +1000)]
mesa: move legacy dri config option texture_depth
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Timothy Arceri [Thu, 30 Aug 2018 00:19:07 +0000 (10:19 +1000)]
mesa: move legacy dri config option fthrottle_mode
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Timothy Arceri [Thu, 30 Aug 2018 00:19:06 +0000 (10:19 +1000)]
mesa: move legacy dri config option def_max_anisotropy
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Timothy Arceri [Thu, 30 Aug 2018 00:19:05 +0000 (10:19 +1000)]
mesa: move legacy dri config option no_neg_lod_bias
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Timothy Arceri [Thu, 30 Aug 2018 00:19:04 +0000 (10:19 +1000)]
mesa: move legacy dri config option round_mode
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Timothy Arceri [Thu, 30 Aug 2018 00:19:03 +0000 (10:19 +1000)]
mesa: remove unused dri option float_depth
This seems to have only been used by DRI1 drivers which were
removed with
e4344161bde2.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Timothy Arceri [Thu, 30 Aug 2018 00:19:02 +0000 (10:19 +1000)]
mesa: move legacy dri config option dither_mode
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Timothy Arceri [Thu, 30 Aug 2018 00:19:01 +0000 (10:19 +1000)]
mesa: move legacy dri config option color_reduction
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Timothy Arceri [Thu, 30 Aug 2018 00:19:00 +0000 (10:19 +1000)]
mesa: move legacy TCL dri config options
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Timothy Arceri [Wed, 12 Sep 2018 00:52:07 +0000 (10:52 +1000)]
util: use force_compat_profile for Wolfenstein The Old Blood
This game is looking for some odd extension after creating a core
context such as ARB_vertex_program and EXT_framebuffer_object.
Rather then enabling these in core this forces the game to use
compat. This allows the game to run and seems to work without
issues. All other id tech games/engines use a compat profile.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Timothy Arceri [Wed, 12 Sep 2018 00:52:06 +0000 (10:52 +1000)]
mesa/st: add force_compat_profile option to driconfig
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Timothy Arceri [Wed, 12 Sep 2018 10:50:34 +0000 (20:50 +1000)]
Revert "radeonsi: avoid syncing the driver thread in si_fence_finish"
This reverts commit
bc65dcab3bc48673ff6180afb036561a4b8b1119.
This was manually reverted. Reverting stops the menu hanging in
some id tech games such as RAGE and Wolfenstein The New Order.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107891
Eric Anholt [Thu, 13 Sep 2018 19:56:18 +0000 (12:56 -0700)]
v3d: Switch from FLUSH_ALL_STATE to FLUSH for ending our bin CLs.
The HW for FLUSH_ALL_STATE isn't validated, since the closed driver only
uses FLUSH. Now that we don't have any new state at the end of our bin
CLs, follow their lead.
Eric Anholt [Thu, 13 Sep 2018 19:59:13 +0000 (12:59 -0700)]
v3d: Stop clearing the OQ state at the end of the job.
Ever since we added OQ support, we've been clearing OQ state at the start
of the job anyway. We're intentionally breaking old-and-new-driver-mix
systems, because we need to stop using the unvalidated FLUSH_ALL_STATE.
Eric Anholt [Thu, 13 Sep 2018 19:54:26 +0000 (12:54 -0700)]
v3d: Always emit a TF disable at the start of drawing on V3D 4.x.
The HW's FLUSH_ALL_STATE is not validated, so we probably shouldn't use
it, meaning that we need to reset state at the start. By doing this, we
also make ourselves more resilient to another client leaving the TF state
enabled at the end of their batch (as we now do, ourselves).
However, we still need to emit a single TF disable at the end of the
frame, for SWVC5-718.
Dylan Baker [Mon, 17 Sep 2018 17:17:48 +0000 (10:17 -0700)]
build: Don't overlink gallium xlib target
Currently gallium's xlib target will fail to link due to multiple
definitions of all the symbols in libmesautil, this only shows up in
autotools, and not in meson due to differences in the way that meson and
autotools handle linking static archives into static archives. Autotools
uses -Wl,--whole-archive implicitly, meson requires this behavior to be
opted-into. The solution is just to remove libmesautils from the
libgl-xlib target, since it will get all of those symbols form
libmesagallium.
I've dropped the link from meson as well, it doesn't seem to hurt
anything and should make linking just a little faster.
Fixes: 8396043f304bb2a752130230055605c5c966e89f
("Replace uses of _mesa_bitcount with util_bitcount")
bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107923
Tested-by: Brian Paul <brianp@vmware.com>
Tested-by: Vinson Lee <vlee@freedesktop.org>
Cc: Sergii Romantsov<sergii.romantsov@globallogic.com>
Dylan Baker [Thu, 13 Sep 2018 18:06:09 +0000 (11:06 -0700)]
move pthread_setaffinity_np check to the build system
Rather than trying to encode all of the rules in a header, lets just put
them in the build system where they belong. This fixes the build on
FreeBSD, which does have pthraed_setaffinity_np, but it's in a
pthread_np.h, not behind _GNU_SOURCE. FreeBSD also implements cpu_set
slightly differently, so additional changes would be required to get it
working right there anyway.
v2: - fix #define in autotools
Fixes: 9f1bbbdbbd77d346c74c7abbb31f399151a85713
("util: try to fix the Android and MacOS build")
Cc: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
Fritz Koenig [Fri, 14 Sep 2018 18:40:49 +0000 (11:40 -0700)]
mesa: FramebufferParameteri parameter checking
Missing break; causes parameter checking to
never pass GL_FRAMEBUFFER_FLIP_Y_MESA parameters.
Fixes: 318c265160 ("mesa: GL_MESA_framebuffer_flip_y extension [v4]")
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Fritz Koenig [Mon, 10 Sep 2018 19:11:16 +0000 (12:11 -0700)]
mesa: Additional FlipY applications
Instances where direction was determined based on
winsys or user fbo and should be determined based on
FlipY.
Key STATE_FB_WPOS_Y_TRANSFORM for of FlipY instead of
_mesa_is_user_fbo. This corrects gl_FragCoord usage
when applying GL_MESA_framebuffer_flip_y.
Fixes: ab05dd183cc ("i965: implement GL_MESA_framebuffer_flip_y [v3]")
Reviewed-by: Brian Paul <brianp@vmware.com>
Bas Nieuwenhuizen [Sun, 16 Sep 2018 00:50:34 +0000 (02:50 +0200)]
radv: Use build ID if available for cache UUID.
To get an useful UUID for systems that have a non-useful mtime
for the binaries.
I started using SHA1 to ensure we get reasonable mixing in the
various possibilities and the various build id lengths.
CC: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Samuel Pitoiset [Fri, 14 Sep 2018 10:52:40 +0000 (12:52 +0200)]
radv: enable shaderInt16 capability
Not sure if this is all wired up. CTS does pass and the Tangrams
demo works fine on Vega. There are corruption issues on Polaris
but not sure if that related to 16-bit support.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Fri, 14 Sep 2018 10:52:39 +0000 (12:52 +0200)]
ac: add 16-bit support to ac_build_bitfield_reverse()
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Fri, 14 Sep 2018 10:52:38 +0000 (12:52 +0200)]
ac: add 16-bit support to ac_build_bit_count()
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Fri, 14 Sep 2018 10:52:37 +0000 (12:52 +0200)]
ac: add 16-bit support to ac_find_lsb()
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Fri, 14 Sep 2018 10:52:36 +0000 (12:52 +0200)]
ac: add 16-bit support to ac_build_umsb()
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Fri, 14 Sep 2018 10:52:35 +0000 (12:52 +0200)]
ac: add 16-bit support to ac_build_isign()
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Fri, 14 Sep 2018 10:52:34 +0000 (12:52 +0200)]
ac: add 16-bit constant values for zero and one
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Fri, 14 Sep 2018 10:52:33 +0000 (12:52 +0200)]
ac: add ac_build_bifield_reverse() helper
Are we missing 64-bit support?
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Fri, 14 Sep 2018 10:52:32 +0000 (12:52 +0200)]
ac: add ac_build_bit_count() helper
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Mon, 17 Sep 2018 09:18:06 +0000 (11:18 +0200)]
radv: fix use of unreachable() in the meta blit path
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Samuel Pitoiset [Mon, 17 Sep 2018 09:20:57 +0000 (11:20 +0200)]
Revert "radv: Optimize rebinding the same descriptor set."
This introduces random GPU hangs on Vega, at least.
This reverts commit
02a43edf186cb9998741ba765cb948bb238a122d.
Samuel Pitoiset [Fri, 14 Sep 2018 12:56:38 +0000 (14:56 +0200)]
radv: fix descriptor pool allocation size
The size has to be multiplied by the number of sets.
This gets rid of the OUT_OF_POOL_KHR error and fixes
a crash with the Tangrams demo.
CC: 18.1 18.2 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Jason Ekstrand [Fri, 14 Sep 2018 22:21:44 +0000 (17:21 -0500)]
anv/query: Add an emit_srm helper
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Jason Ekstrand [Fri, 14 Sep 2018 22:06:48 +0000 (17:06 -0500)]
anv: Add a mi_memset and use it for zeroing queries
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Jason Ekstrand [Fri, 14 Sep 2018 22:02:08 +0000 (17:02 -0500)]
anv/query: Use anv_address everywhere
Instead of passing around BOs and offsets, use addresses which are anv's
GPU equivalent of pointers.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Jason Ekstrand [Fri, 14 Sep 2018 21:34:22 +0000 (16:34 -0500)]
anv/query: Write both dwords in emit_zero_queries
Each query slot is a uint64_t and we were only zeroing half of it.
Fixes: 7ec6e4e68980 "anv/query: implement multiview interactions"
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Jason Ekstrand [Fri, 14 Sep 2018 20:07:36 +0000 (15:07 -0500)]
anv/query: Increment an index while writing results
Instead of computing an index at the end which we hope maps to the
number of things written, just count the number of things as we go.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Ian Romanick [Thu, 13 Sep 2018 00:16:50 +0000 (17:16 -0700)]
i965/fs: Don't propagate conditional modifiers from integer compares to adds
No shader-db changes on any Intel platform... which probably explains
why no bugs have been bisected to this problem since it landed in Mesa
18.1. :( The commit mentioned below is in 18.2, so 18.1 would need a
slightly different fix (due to code refactoring).
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Fixes: 77f269bb560 "i965/fs: Refactor propagation of conditional modifiers from compares to adds"
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com> (reviewed the original patch)
Cc: Matt Turner <mattst88@gmail.com> (reviewed the original patch)
Bas Nieuwenhuizen [Sun, 16 Sep 2018 10:28:33 +0000 (12:28 +0200)]
radv: Only allow 16 user SGPRs for compute on GFX9+.
Apparently for compute there are only 16 instead of the 32 for the
graphics path.
Fixes dEQP-VK.binding_model.descriptorset_random.sets16.noarray.ubolimitlow.sbolimitlow.imglimitlow.noiub.comp.0
CC: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Bas Nieuwenhuizen [Sun, 16 Sep 2018 10:17:00 +0000 (12:17 +0200)]
radv: Set the user SGPR MSB for Vega.
Otherwise using 32 user SGPRs would be broken.
CC: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Bas Nieuwenhuizen [Sun, 16 Sep 2018 00:17:32 +0000 (02:17 +0200)]
radv: Optimize rebinding the same descriptor set.
This makes it cheaper to just change the dynamic offsets with
the same descriptor sets.
Suggested-by: Philip Rebohle <philip.rebohle@tu-dortmund.de>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Gert Wollny [Fri, 14 Sep 2018 14:56:48 +0000 (16:56 +0200)]
r600/sb: use safe math optimizations when TGSI contains precise operations
Fixes:
dEQP-GLES3.functional.shaders.invariance.highp.common_subexpression_3
dEQP-GLES3.functional.shaders.invariance.mediump.common_subexpression_3
dEQP-GLES3.functional.shaders.invariance.lowp.common_subexpression_3
Signed-off-by: Gert Wollny <gw.fossdev@gmail.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Mauro Rossi [Sun, 26 Aug 2018 21:38:12 +0000 (23:38 +0200)]
android: broadcom/cle: export the broadcom top level path headers
Fixes the following building error in vc4 build:
In file included from external/mesa/src/gallium/drivers/vc4/kernel/vc4_render_cl.c:34:
In file included from external/mesa/src/gallium/drivers/vc4/kernel/vc4_drv.h:27:
In file included from external/mesa/src/gallium/drivers/vc4/vc4_simulator_validate.h:34:
In file included from external/mesa/src/gallium/drivers/vc4/vc4_context.h:39:
In file included from external/mesa/src/gallium/drivers/vc4/vc4_cl.h:56:
gen/STATIC_LIBRARIES/libmesa_broadcom_genxml_intermediates/broadcom/cle/v3d_packet_v21_pack.h:12:10:
fatal error: 'cle/v3d_packet_helpers.h' file not found
^~~~~~~~~~~~~~~~~~~~~~~~~~
1 error generated.
Fixes: 5b102160ae ("broadcom/genxml: Introduce a V3D packet/struct decoder.")
Cc: "18.2" <mesa-stable@lists.freedesktop.org>
Acked-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Signed-off-by: Mauro Rossi <issor.oruam@gmail.com>
Mauro Rossi [Sun, 26 Aug 2018 21:11:02 +0000 (23:11 +0200)]
android: broadcom/cle: add gallium include path
Fixes the following building error:
In file included from external/mesa/src/broadcom/cle/v3d_decoder.c:38:
In file included from external/mesa/src/broadcom/cle/v3d_packet_helpers.h:29:
external/mesa/src/gallium/auxiliary/util/u_math.h:42:10:
fatal error: 'pipe/p_compiler.h' file not found
^~~~~~~~~~~~~~~~~~~
1 error generated.
Fixes: 5b102160ae ("broadcom/genxml: Introduce a V3D packet/struct decoder.")
Cc: "18.2" <mesa-stable@lists.freedesktop.org>
Acked-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Signed-off-by: Mauro Rossi <issor.oruam@gmail.com>
Mauro Rossi [Sat, 25 Aug 2018 16:17:23 +0000 (18:17 +0200)]
android: broadcom/genxml: fix collision with intel/genxml header-gen macro
Fixes the following building error, happening when building both intel and broadcom:
Gen Header: libmesa_broadcom_genxml_32 <= v3d_packet_v21_pack.h
FAILED: gen/STATIC_LIBRARIES/libmesa_broadcom_genxml_intermediates/broadcom/cle/v3d_packet_v21_pack.h
/bin/bash -c "python external/mesa/src/broadcom/cle/gen_pack_header.py \
external/mesa/src/broadcom/cle/v3d_packet_v21.xml \
> gen/STATIC_LIBRARIES/libmesa_broadcom_genxml_intermediates/broadcom/cle/v3d_packet_v21_pack.h"
Traceback (most recent call last):
File "external/mesa/src/broadcom/cle/gen_pack_header.py", line 626, in <module>
p = Parser(sys.argv[2])
IndexError: list index out of range
header-gen macro is already defined by Intel genxml building rules
and the existing header-gen does not have the $(PRIVATE_VER) argument,
infact the bash command line logged in the building error is missing
exactly $(PRIVATE_VER) argument
Renaming the macro as pack-header-gen in src/broadcom/Android.genxml.mk
solves the building error, another possible way is to keep the gen rules
commands expanded and not use the macros.
Fixes: 7f80a9ff13 ("vc4: Introduce XML-based packet header generation like Intel's.")
Cc: "18.2" <mesa-stable@lists.freedesktop.org>
Acked-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Signed-off-by: Mauro Rossi <issor.oruam@gmail.com>
Caio Marcelo de Oliveira Filho [Sat, 15 Sep 2018 03:53:22 +0000 (20:53 -0700)]
anv/memcpy: fix build after starting to use addresses
The offsets now come from the anv_address, these references were not
updated and using the old variable.
Fixes: e1ab8345574 "anv/memcpy: Use addresses instead of bo+offset"
Tested-by: Clayton Craft <clayton.a.craft@intel.com>
Jason Ekstrand [Tue, 11 Sep 2018 19:59:02 +0000 (14:59 -0500)]
anv/cmd_buffer: Take an address in emit_lrm
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
Jason Ekstrand [Mon, 10 Sep 2018 21:43:34 +0000 (16:43 -0500)]
anv/memcpy: Use addresses instead of bo+offset
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
Jason Ekstrand [Mon, 10 Sep 2018 21:37:17 +0000 (16:37 -0500)]
anv/so_memcpy: Use the correct SO_BUFFER size on gen8+
This shouldn't matter as we'll never write OOB anyway but we may as well
get it right. It's supposed to be in dwords - 1.
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Timothy Arceri [Fri, 27 Jul 2018 05:32:36 +0000 (15:32 +1000)]
ac: fix get_image_coords() for radeonsi
Because this was setting image to true we would end up calling
si_load_image_desc() when we sould be calling
si_load_sampler_desc().
This fixes an assert() in Deus Ex: MD
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Marek Olšák [Thu, 13 Sep 2018 00:29:19 +0000 (20:29 -0400)]
gallium/util: don't let child processes inherit our thread affinity
v2: corrected the comment
Marek Olšák [Fri, 7 Sep 2018 19:37:45 +0000 (15:37 -0400)]
gallium/util: start with a random L3 cache index for AMD Zen
Josh Pieper [Mon, 10 Sep 2018 02:03:27 +0000 (22:03 -0400)]
st/mesa: Validate the result of pipe_transfer_map in make_texture (v2)
When using Freecad, I was getting intermittent segfaults inside of
mesa. I traced it down to this path in st_cb_drawpixels.c where the
result of pipe_transfer_map wasn't being checked. In my case, it was
returning NULL because nouveau_bo_new returned ENOENT. I'm by no
means a mesa developer, but this patch solves the problem for me and
seems reasonable enough.
v2: Marek - also unmap the PBO and release the texture, and call
the make_texture function sooner for less cleanup
Cc: 18.1 18.2 <mesa-stable@lists.freedesktop.org>
Samuel Pitoiset [Thu, 13 Sep 2018 10:30:21 +0000 (12:30 +0200)]
radv: emit the initial config only once in the preambles
It shouldn't be needed to emit the initial graphics or compute
state when beginning a new command buffer. Emitting them in
the preamble should be enough and this will reduce IB sizes.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Wed, 12 Sep 2018 13:40:09 +0000 (15:40 +0200)]
radv: fix setting global locations for indirect descriptors
Indirect descriptors only need one entry, we don't have to
emit a location for every descriptors.
Fixes GPU hangs with new CTS:
dEQP-VK.binding_model.descriptorset_random.*
CC: 18.2 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Wed, 12 Sep 2018 13:40:08 +0000 (15:40 +0200)]
radv: fix flushing indirect descriptors
Let say, we first bind a graphics pipeline that needs indirect
descriptors sets. The userdata pointers will be emitted at draw
time. Then if we bind a compute pipeline that doesn't need any
indirect descriptors, the driver will re-emit them for all
grpahics stages.
To avoid this to happen, just check the bind point type.
CC: 18.2 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Wed, 12 Sep 2018 13:40:07 +0000 (15:40 +0200)]
radv: fix GPU hangs with 32-bit indirect descriptors
LLVM 6 isn't affected.
Fixes GPU hangs with new CTS:
dEQP-VK.binding_model.descriptorset_random.*
CC: 18.2 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Wed, 12 Sep 2018 13:40:06 +0000 (15:40 +0200)]
radv: handle loc->indirect correctly for the first descriptor
This was wrong for descriptor #0 when all of them are indirect.
This is because indirect_offset was 0 and we emitted a
"normal" descriptor pointer for nothing.
While we are at it remove
radv_userdata_info::indirect_offset which is useless.
CC: 18.2 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Wed, 12 Sep 2018 13:40:05 +0000 (15:40 +0200)]
radv: bump the maximum number of arguments to 64
Bumping to 64 should be safe enough.
Fixes some crashes with new CTS:
dEQP-VK.binding_model.descriptorset_random.*
CC: 18.2 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Thu, 13 Sep 2018 13:58:02 +0000 (15:58 +0200)]
radv: tidy up ac_setup_rings() for the GSVS rings
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Thu, 13 Sep 2018 13:58:01 +0000 (15:58 +0200)]
radv: fix setting the number of entries for GSVS on VI+
According to RadeonSI, it's unnecessary to multiply by
the stride. That field seems to always be 64.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Thu, 13 Sep 2018 13:58:00 +0000 (15:58 +0200)]
radv: always compute the number of components from the output mask
That removes two special cases for clip/cull distances.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Thu, 13 Sep 2018 13:57:59 +0000 (15:57 +0200)]
radv: emit data contiguously in the GS->VS ring buffer
Instead of having holes. The other ring parameters like
offset and stride can be updated later.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Thu, 13 Sep 2018 13:57:58 +0000 (15:57 +0200)]
radv: make use of the output usage mask in GS copy shader
This is just for consistency because LLVM can detect and
remove unused loads.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Wed, 12 Sep 2018 21:20:39 +0000 (23:20 +0200)]
radv: improve a comment in si_emit_set_predication_state()
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Wed, 12 Sep 2018 21:20:38 +0000 (23:20 +0200)]
radv: fix VK_EXT_conditional_rendering visibility
It's actually just the opposite.
This fixes the new Sascha conditionalrender demo.
CC: 18.2 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Thu, 13 Sep 2018 14:36:45 +0000 (16:36 +0200)]
radv: make use of ac_unpack_param() instead of ac_build_bfe()
Same code is generated because LLVM ends up by using bfe, but
that seems cleaner to me.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Timothy Arceri [Wed, 11 Jul 2018 00:53:41 +0000 (10:53 +1000)]
nir: add loop unroll support for complex wrapper loops
In GLSL IR we cheat with switch statements and simply convert them
into loops with a single iteration. This allowed us to make use of
the existing jump instruction handling provided by the loop handing
code, it also allows dead code to be cleaned up once we have
wrapped the code in a loop.
However using loops in this way created previously unrollable loops
which limits further optimisations. Here we provide a way to unroll
loops that end in a break and have multiple other exits.
All shader-db changes are from the dolphin uber shaders. There is a
small amount of HURT shaders but in general the improvements far
exceed the HURT.
shader-db results IVB:
total instructions in shared programs:
10018187 ->
10016468 (-0.02%)
instructions in affected programs: 104080 -> 102361 (-1.65%)
helped: 36
HURT: 15
total cycles in shared programs:
220065064 ->
154529655 (-29.78%)
cycles in affected programs:
126063017 ->
60527608 (-51.99%)
helped: 51
HURT: 0
total loops in shared programs: 2515 -> 2308 (-8.23%)
loops in affected programs: 903 -> 696 (-22.92%)
helped: 51
HURT: 0
total spills in shared programs: 4370 -> 4124 (-5.63%)
spills in affected programs: 1397 -> 1151 (-17.61%)
helped: 9
HURT: 12
total fills in shared programs: 4581 -> 4419 (-3.54%)
fills in affected programs: 2201 -> 2039 (-7.36%)
helped: 9
HURT: 15
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Timothy Arceri [Mon, 23 Jul 2018 04:38:34 +0000 (14:38 +1000)]
nir: propagates if condition evaluation down some alu chains
v2:
- only allow nir_op_inot or nir_op_b2i when alu input is 1.
- use some helpers as suggested by Jason.
v3:
- evaluate alu op for single input alu ops
- add helper function to decide if to propagate through alu
- make use of nir_before_src in another spot
shader-db IVB results:
total instructions in shared programs:
9993483 ->
9993472 (-0.00%)
instructions in affected programs: 1300 -> 1289 (-0.85%)
helped: 11
HURT: 0
total cycles in shared programs:
219476091 ->
219476059 (-0.00%)
cycles in affected programs: 7675 -> 7643 (-0.42%)
helped: 10
HURT: 1
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Timothy Arceri [Wed, 29 Aug 2018 15:24:43 +0000 (10:24 -0500)]
nir: evaluate if condition uses inside the if branches
Since we know what side of the branch we ended up on we can just
replace the use with a constant.
All the spill changes in shader-db are from Dolphin uber shaders,
despite some small regressions the change is clearly positive.
V2: insert new constant after any phis in the
use->parent_instr->type == nir_instr_type_phi path.
v3:
- use nir_after_block_before_jump() for inserting const
- check dominance of phi uses correctly
v4:
- create some helpers as suggested by Jason.
v5 (Jason Ekstrand):
- Use LIST_ENTRY to get the phi src
shader-db results IVB:
total instructions in shared programs:
9999201 ->
9993483 (-0.06%)
instructions in affected programs: 163235 -> 157517 (-3.50%)
helped: 132
HURT: 2
total cycles in shared programs:
231670754 ->
219476091 (-5.26%)
cycles in affected programs:
143424120 ->
131229457 (-8.50%)
helped: 115
HURT: 24
total spills in shared programs: 4383 -> 4370 (-0.30%)
spills in affected programs: 1656 -> 1643 (-0.79%)
helped: 9
HURT: 18
total fills in shared programs: 4610 -> 4581 (-0.63%)
fills in affected programs: 374 -> 345 (-7.75%)
helped: 6
HURT: 0
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Erik Faye-Lund [Wed, 12 Sep 2018 07:48:41 +0000 (09:48 +0200)]
virgl: adjust strides when mapping temp-resources
When we're mapping temp-resources, we clip the resource to the
transfer-box, which means the stride might not be correct any more.
So let's update the stride from the temp-resource, and recompute the
layer-stride.
This fixes crashes when running dEQP with --deqp-gl-config-name=rgba8888d24s8ms4
Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Fixes: a8987b88ff1 "virgl: add driver for virtio-gpu 3D (v2)"
Reviewed-by: Dave Airlie <airlied@redhat.com>
Pierre Moreau [Mon, 4 Dec 2017 23:51:04 +0000 (00:51 +0100)]
nvir: Always split 64-bit IMAD/IMUL operations
Those operations do not map to actual hardware instructions, therefore
those should always be lowered to 32-bit instructions.
Fixes: 009c54aa7af "nv50/ir: Split 64-bit integer MAD/MUL operations"
Signed-off-by: Pierre Moreau <pierre.morrow@free.fr>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Leo Liu [Fri, 7 Sep 2018 13:26:08 +0000 (09:26 -0400)]
st/vdpau: Use output buffer as back buffer with 24-bit color only
Using output buffer with 8 bits video RGB as back buffer
certainly is not working for 30 bits color depth visual.
Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Leo Liu [Mon, 10 Sep 2018 20:02:29 +0000 (16:02 -0400)]
vl/dri: add color depth to vl winsys
For VDPAU use later
Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Leo Liu [Mon, 10 Sep 2018 19:33:28 +0000 (15:33 -0400)]
vl/dri3: add support for 10 bits format
Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Leo Liu [Fri, 7 Sep 2018 19:17:31 +0000 (15:17 -0400)]
vl/dri: add 10 bits format supports
v2: Tell B10G10R10X2 and R10G10B10X2 formats for different HW.
Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Kristian H. Kristensen [Wed, 12 Sep 2018 18:38:22 +0000 (11:38 -0700)]
egl/android: Declare droid_load_driver() static
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Signed-off-by: Kristian H. Kristensen <hoegsberg@chromium.org>
Samuel Pitoiset [Wed, 12 Sep 2018 20:54:27 +0000 (22:54 +0200)]
radv: fix function names for VK_EXT_conditional_rendering
Otherwise they are not exported.
CC: 18.2 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Dave Airlie <airlied@redhat.com
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Jason Ekstrand [Wed, 12 Sep 2018 22:02:57 +0000 (17:02 -0500)]
anv: Silence a couple compiler warnings
[63/93] Compiling C object 'src/intel/vulkan/...intel@vulkan@@anv_common@sta/anv_device.c.o'.
../src/intel/vulkan/anv_device.c:685:30: warning: passing 'const char *' to parameter of type 'void *' discards qualifiers [-Wincompatible-pointer-types-discards-qualifiers]
vk_free(&instance->alloc, instance->app_info.app_name);
^~~~~~~~~~~~~~~~~~~~~~~~~~~
../src/vulkan/util/vk_alloc.h:62:51: note: passing argument to parameter 'data' here
vk_free(const VkAllocationCallbacks *alloc, void *data)
^
../src/intel/vulkan/anv_device.c:686:30: warning: passing 'const char *' to parameter of type 'void *' discards qualifiers [-Wincompatible-pointer-types-discards-qualifiers]
vk_free(&instance->alloc, instance->app_info.engine_name);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../src/vulkan/util/vk_alloc.h:62:51: note: passing argument to parameter 'data' here
vk_free(const VkAllocationCallbacks *alloc, void *data)
^
[65/93] Compiling C object 'src/intel/vulkan/...ommon@sta/anv_nir_apply_pipeline_layout.c.o'.
../src/intel/vulkan/anv_nir_apply_pipeline_layout.c:519:13: warning: unused variable 'image_uniform' [-Wunused-variable]
unsigned image_uniform;
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Michel Dänzer [Tue, 4 Sep 2018 16:18:57 +0000 (18:18 +0200)]
loader/dri3: Don't wait for fence of old buffer when re-allocating it
We only need to wait for the fence before drawing to a buffer, not
before reading from it.
This might avoid hangs when re-allocating the fake front buffer, similar
to the previous change. But I haven't seen any evidence that this was
actually happening in practice.
Tested-by: Olivier Fourdan <ofourdan@redhat.com>