Kenneth Graunke [Mon, 15 Oct 2018 17:43:49 +0000 (10:43 -0700)]
iris: Enable fb fetch
needed for ES 3.2
Kenneth Graunke [Mon, 15 Oct 2018 16:51:54 +0000 (09:51 -0700)]
iris: advertise GL_ARB_shader_texture_image_samples
Jordan Justen [Tue, 18 Sep 2018 21:22:34 +0000 (14:22 -0700)]
iris: Set num_uniforms in bytes
Ref: brw_nir_lower_uniforms, type_size_scalar_bytes
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Kenneth Graunke [Thu, 11 Oct 2018 04:44:43 +0000 (21:44 -0700)]
iris: move images next to textures in binding table
Kenneth Graunke [Tue, 2 Oct 2018 17:21:57 +0000 (10:21 -0700)]
iris: null for non-existent cbufs
prevents BTs from being shifted down incorrectly
Kenneth Graunke [Tue, 2 Oct 2018 17:09:26 +0000 (10:09 -0700)]
iris: actually set image access
Jason Ekstrand [Tue, 2 Oct 2018 03:18:00 +0000 (22:18 -0500)]
iris: Don't lower image formats for write-only images
Kenneth Graunke [Fri, 14 Sep 2018 07:49:13 +0000 (00:49 -0700)]
iris: set image access correctly
Kenneth Graunke [Thu, 30 Aug 2018 22:49:32 +0000 (15:49 -0700)]
iris: bother with BTIs
Kenneth Graunke [Thu, 30 Aug 2018 22:45:36 +0000 (15:45 -0700)]
iris: implement set_shader_images hook
Kenneth Graunke [Thu, 30 Aug 2018 22:20:12 +0000 (15:20 -0700)]
iris: lower storage image derefs
Kenneth Graunke [Fri, 14 Sep 2018 21:18:13 +0000 (14:18 -0700)]
iris: set the binding table size
we weren't doing mark_surface_used on images (i965 does it while
uploading the unnecessary image uniforms), so our binding tables were
too small...
Kenneth Graunke [Tue, 9 Oct 2018 23:47:54 +0000 (16:47 -0700)]
iris: X32_S8X24 :/
This can happen when faking Z32_S8X24 and setting StencilSampling = true
I guess we'll just turn it into S8_UINT...
Fixes KHR-GL45.texture_swizzle.functional
Kenneth Graunke [Mon, 8 Oct 2018 03:50:19 +0000 (20:50 -0700)]
iris: enable I/L formats
Kenneth Graunke [Mon, 8 Oct 2018 03:47:21 +0000 (20:47 -0700)]
iris: Use R/RG instead of I/L/A when sampling
Kenneth Graunke [Mon, 8 Oct 2018 03:31:09 +0000 (20:31 -0700)]
iris: rework format translation apis
Kenneth Graunke [Mon, 8 Oct 2018 20:44:01 +0000 (13:44 -0700)]
iris: Allow PIPE_CONTROL with Stall at Scoreboard and RT flush
It's nonsensical, but not illegal, and mandatory on Icelake
Kenneth Graunke [Mon, 8 Oct 2018 13:26:15 +0000 (06:26 -0700)]
iris: add gen11 to genX_call
Kenneth Graunke [Sun, 7 Oct 2018 05:54:38 +0000 (22:54 -0700)]
iris: inline stage_from_pipe to avoid unused warnings
Kenneth Graunke [Sun, 7 Oct 2018 05:50:34 +0000 (22:50 -0700)]
iris: pipe to scs -> iris_pipe.h
Kenneth Graunke [Fri, 5 Oct 2018 22:25:53 +0000 (15:25 -0700)]
iris: force persample interp cap
Kenneth Graunke [Fri, 5 Oct 2018 02:49:06 +0000 (19:49 -0700)]
iris: stencil texturing
Kenneth Graunke [Fri, 5 Oct 2018 02:34:34 +0000 (19:34 -0700)]
iris: fix Z32_S8 depth sampling
We were accidentally using the ISL_FORMAT_R32_FLOAT_X8X24_TYPELESS
format, which is NOT what we use. We just store R32_FLOAT depth.
fixes Piglit's texwrap GL_ARB_depth_buffer_float
Kenneth Graunke [Wed, 3 Oct 2018 03:09:01 +0000 (20:09 -0700)]
iris: don't mark contains_draw = false when chaining batches
chaining to a new batch reuses create_batch(), but we don't need to do
the work of pinning BOs we inherit from a previous batch...when that is
actually part of the same execbuf invocation.
instead, just flag it when setting primary_batch_size = 0, in
iris_batch_reset
Kenneth Graunke [Wed, 3 Oct 2018 02:55:29 +0000 (19:55 -0700)]
iris: vma_free bo->size, not bo_size
this is more obviously correct. I think the two end up being the same
in practice, since this is in the alloc_from_cache case, and presumably
bo from the bucket has bo->size == bucket->size, and bo_size also is
bucket->size...
still. better to do the obvious thing.
brw_bufmgr already does it this way.
Kenneth Graunke [Tue, 2 Oct 2018 01:33:17 +0000 (18:33 -0700)]
iris: drop a bunch of pipe_sampler_state stuff we don't need
Kenneth Graunke [Sun, 30 Sep 2018 06:49:02 +0000 (08:49 +0200)]
iris: just mark snapshots_landed from the CPU
otherwise, get results may check q->map->snapshots_landed...before our
commands to initialize it to false have actually executed...so it'd get
some random garbage from the BO...
Kenneth Graunke [Sat, 29 Sep 2018 09:04:48 +0000 (11:04 +0200)]
iris: Enable ARB_shader_vote
The easiest get out the vote campaign ever
Kenneth Graunke [Sat, 29 Sep 2018 08:47:01 +0000 (10:47 +0200)]
iris: magic number 36 -> #define
Kenneth Graunke [Sat, 29 Sep 2018 08:43:00 +0000 (10:43 +0200)]
iris: better query file comment
Kenneth Graunke [Sat, 29 Sep 2018 08:41:44 +0000 (10:41 +0200)]
iris: early return properly
Kenneth Graunke [Sat, 29 Sep 2018 08:39:39 +0000 (10:39 +0200)]
iris: 36-bit overflow fixes
Kenneth Graunke [Fri, 28 Sep 2018 14:45:04 +0000 (16:45 +0200)]
iris: Need to | 1 when asking for timestamps
Kenneth Graunke [Fri, 28 Sep 2018 13:42:51 +0000 (15:42 +0200)]
iris: glGet timestamps, more correct timestamps
Kenneth Graunke [Fri, 28 Sep 2018 10:39:10 +0000 (12:39 +0200)]
iris: ...and SO prims emitted queries
looks like we have queries
some fails still due to races between snapshots_written and start/end
not being garbage...not sure what that's about
Kenneth Graunke [Fri, 28 Sep 2018 10:25:36 +0000 (12:25 +0200)]
iris: timestamps
Kenneth Graunke [Fri, 28 Sep 2018 10:09:19 +0000 (12:09 +0200)]
iris: drop explicit pinning
writes will already rw_bo or ro_bo that
Kenneth Graunke [Fri, 28 Sep 2018 10:07:54 +0000 (12:07 +0200)]
iris: primitives generated query support
Kenneth Graunke [Thu, 27 Sep 2018 12:36:31 +0000 (14:36 +0200)]
iris: pipeline stats
Kenneth Graunke [Thu, 27 Sep 2018 12:27:41 +0000 (14:27 +0200)]
iris: play chicken with timer queries for now
they have been crashy in the past and I don't want to risk tanking my
laptop right before my XDC talk
Kenneth Graunke [Thu, 27 Sep 2018 10:58:52 +0000 (12:58 +0200)]
iris: gpr0 to bool
I think OQ is basically working now.
Kenneth Graunke [Thu, 27 Sep 2018 10:50:26 +0000 (12:50 +0200)]
iris: fix random failures via CS stall...but why?
Kenneth Graunke [Thu, 27 Sep 2018 10:14:26 +0000 (12:14 +0200)]
iris: flush batch when asking for result via QBO
Kenneth Graunke [Thu, 27 Sep 2018 08:39:04 +0000 (10:39 +0200)]
iris: results write
Kenneth Graunke [Wed, 26 Sep 2018 10:51:44 +0000 (12:51 +0200)]
iris: gen10+ workarounds and break fix
Kenneth Graunke [Mon, 17 Sep 2018 23:37:26 +0000 (16:37 -0700)]
iris: initial query code
Kenneth Graunke [Mon, 17 Sep 2018 23:37:26 +0000 (16:37 -0700)]
iris: LRM/SRM/SDI hooks
Kenneth Graunke [Tue, 18 Sep 2018 04:15:48 +0000 (21:15 -0700)]
iris: rw_bo for pipe controls
this is used for WRITE IMMEDIATE...
but maybe we don't want to for the workaround BO?
Kenneth Graunke [Fri, 21 Sep 2018 22:43:49 +0000 (15:43 -0700)]
iris: use 0 for TCS passthrough program string ID
the passthrough shader doesn't need a real program string ID - that's
basically used for ARB programs indicating total program source code
changes, or other pre-baked uniform changes, etc...none of which a
passthrough shader has...so we don't need a unique identifier to
distinguish them. We want to use a consistent value so we find
existing passthrough shaders in the cache.
Caio Marcelo de Oliveira Filho [Fri, 21 Sep 2018 19:22:34 +0000 (12:22 -0700)]
iris: Add support for TCS passthrough
If no TCS is provided, create a "passthrough" TCS that will take the
default values set in the API as constants and pass to the TES, along
with any other inputs it expects. The code to create the NIR shader
is the same as in i965.
Tested with
./piglit run -t 'tess' quick_shader r
and fixed a dozen crashes from that list.
Kenneth Graunke [Fri, 21 Sep 2018 00:27:47 +0000 (17:27 -0700)]
iris: inherit the index buffer properly
Kenneth Graunke [Thu, 20 Sep 2018 23:46:58 +0000 (16:46 -0700)]
iris: delete bogus comment
Caio asked what was wrong. There is nothing wrong. :)
Kenneth Graunke [Thu, 20 Sep 2018 22:35:14 +0000 (15:35 -0700)]
iris: properly re-pin stencil buffers
Kenneth Graunke [Thu, 20 Sep 2018 20:32:07 +0000 (13:32 -0700)]
iris: fix context restore of 3DSTATE_CONSTANT ranges
if clean we want to DO the pinning...not SKIP the pinning.
thanks to Jordan Justen for catching this!
Kenneth Graunke [Thu, 20 Sep 2018 18:12:27 +0000 (11:12 -0700)]
iris: silence const warning
not sure why this is labeled const, I'm pretty sure we are taking the
reference and owning this, so there's no particular reason we can't
change it. it certainly seems to be working for non-compute. and,
freedreno's ir3_shader.c seems to do this as well. still...gross :/
Kenneth Graunke [Fri, 27 Jul 2018 04:59:20 +0000 (21:59 -0700)]
iris: refactor program CSO stuff
Caio Marcelo de Oliveira Filho [Wed, 19 Sep 2018 17:17:23 +0000 (10:17 -0700)]
iris: Fix uses of gl_TessLevel*
The backend compiler expects the gl_TessLevel* variables to be mapped
as inputs instead of system values. Use the new PIPE_CAP to get this
behavior from GLSL compiler.
Tested with:
tests/spec/arb_tessellation_shader/execution/vs-tcs-tes-tessinner-tessouter-inputs-quads.shader_test
Kenneth Graunke [Tue, 18 Sep 2018 18:04:44 +0000 (11:04 -0700)]
iris: totally untested icelake support
Kenneth Graunke [Tue, 18 Sep 2018 18:07:16 +0000 (11:07 -0700)]
iris: initialize "don't suck" bits, as Ben likes to call them
Kenneth Graunke [Tue, 18 Sep 2018 18:04:24 +0000 (11:04 -0700)]
iris: refactor LRIs in context setup
we're going to have more of them, so reduce the boilerplate
Kenneth Graunke [Sat, 15 Sep 2018 22:27:59 +0000 (15:27 -0700)]
iris: enable ARB_enhanced_layouts
Kenneth Graunke [Sat, 15 Sep 2018 21:50:26 +0000 (14:50 -0700)]
iris: re-pin binding table contents if we didn't re-emit them
fixes glsl-vs-loop and other regressions from multibinder.
Kenneth Graunke [Sat, 15 Sep 2018 21:35:47 +0000 (14:35 -0700)]
iris: move binder pinning outside the dirty == 0 check
This might be a new batch with back to back non-dirty calls, if so we
need to inherit the old binder...
Chris Wilson [Fri, 14 Sep 2018 22:12:36 +0000 (15:12 -0700)]
iris: fix memzone_for_address since multibinder changes
Kenneth Graunke [Thu, 13 Sep 2018 06:31:46 +0000 (23:31 -0700)]
iris: update comments for multibinder
Kenneth Graunke [Thu, 13 Sep 2018 05:23:50 +0000 (22:23 -0700)]
iris: fix SO offset writes for multiple streams
Kenneth Graunke [Sun, 9 Sep 2018 02:43:34 +0000 (19:43 -0700)]
iris: Support multiple binder BOs, update Surface State Base Address
Kenneth Graunke [Tue, 11 Sep 2018 08:09:27 +0000 (01:09 -0700)]
iris: fix null FB and unbound tex surface state addresses
Kenneth Graunke [Sat, 8 Sep 2018 06:39:30 +0000 (23:39 -0700)]
iris: set EXEC_OBJECT_CAPTURE on all driver internal buffers
Kenneth Graunke [Fri, 7 Sep 2018 19:26:55 +0000 (12:26 -0700)]
iris: fix constant buffer 0 to be absolute
thanks to Jason for catching this. Fixes some va64 tests. Surprisingly
not much else, as apparently getting to UBO range 4 is uncommon!
Kenneth Graunke [Sat, 1 Sep 2018 01:03:19 +0000 (18:03 -0700)]
iris: don't unconditionally emit 3DSTATE_VF / 3DSTATE_VF_TOPOLOGY
this was just laziness on my part
Kenneth Graunke [Sat, 1 Sep 2018 07:58:29 +0000 (00:58 -0700)]
iris: skip over whole function if dirty == 0
kinda pointless in non-pathological cases, but does boost our score in
the drawarrays case by 50%...
Kenneth Graunke [Sun, 19 Aug 2018 17:31:29 +0000 (10:31 -0700)]
iris: Allow inlining of require/get_command_space
eliminates so many callqs for ptr++
Kenneth Graunke [Fri, 7 Sep 2018 02:42:58 +0000 (19:42 -0700)]
iris: use Eric's new caps helper
this does change a couple caps...PRIMITIVE_RESTART_FOR_PATCHES...
Kenneth Graunke [Sat, 1 Sep 2018 04:22:31 +0000 (21:22 -0700)]
iris: new caps
Kenneth Graunke [Wed, 29 Aug 2018 22:24:45 +0000 (15:24 -0700)]
iris: fix blend state memcpy
thanks to Jason for noticing grumpy valgrind
Kenneth Graunke [Sat, 25 Aug 2018 07:11:24 +0000 (00:11 -0700)]
iris: Skip primitive ID overrides if the shader wrote a custom value
Fixes glsl-1.50/execution/geometry/primitive-id-out
Kenneth Graunke [Fri, 24 Aug 2018 04:21:04 +0000 (21:21 -0700)]
iris: fix crash when binding optional shader for the first time
Kenneth Graunke [Thu, 23 Aug 2018 09:36:18 +0000 (02:36 -0700)]
iris: handle level/layer in direct maps
needed now that we do 1D linear
Kenneth Graunke [Thu, 23 Aug 2018 06:39:38 +0000 (23:39 -0700)]
iris: use linear for 1D textures
This gets us the gen9 compact linear storage
Kenneth Graunke [Thu, 23 Aug 2018 08:49:49 +0000 (01:49 -0700)]
iris: big old hack for tex-miplevel-selection
copied from ilo. I don't understand this at all..
Kenneth Graunke [Thu, 23 Aug 2018 09:15:12 +0000 (02:15 -0700)]
iris: fix sampler state setting
Kenneth Graunke [Tue, 21 Aug 2018 18:50:56 +0000 (11:50 -0700)]
iris: try to hack around binder issue
Kenneth Graunke [Tue, 21 Aug 2018 18:30:09 +0000 (11:30 -0700)]
iris: fix line-aa-width
we should probably move the roundf to st_atom_raster
Kenneth Graunke [Tue, 21 Aug 2018 06:37:13 +0000 (23:37 -0700)]
iris: implement get_sample_position
Fixes arb_sample_shading/builtin-gl-sample-position
Kenneth Graunke [Tue, 21 Aug 2018 04:23:25 +0000 (21:23 -0700)]
iris: z_res -> s_res
fixes crashes introduced a few commits ago
Kenneth Graunke [Tue, 21 Aug 2018 03:55:17 +0000 (20:55 -0700)]
iris: reenable R32G32B32 texture buffers
This dropped us from GL 4.2 to GL 3.3 by mistake. Thanks to Dave for
catching this!
Chris Wilson [Sun, 19 Aug 2018 09:32:43 +0000 (10:32 +0100)]
iris: Record reusability of bo on construction
We know that if the bufmgr->reuse is set to false or if the bo is too
large for a bucket, the same will be true when we come to free the bo.
Kenneth Graunke [Sun, 19 Aug 2018 21:05:53 +0000 (14:05 -0700)]
iris: Reduce binder alignment from 64 to 32
3DSTATE_BINDING_TABLE_POINTER_XS's alignment requirement is only 32B.
Makes us waste less precious binder space.
Kenneth Graunke [Sun, 19 Aug 2018 17:08:05 +0000 (10:08 -0700)]
iris: precompute hashes for cache tracking
saves a touch of cpu overhead in the new resolve tracking
Chris Wilson [Tue, 31 Jul 2018 13:47:02 +0000 (14:47 +0100)]
iris: AMD_pinned_memory
(rebased by Ken, mainly set res->internal_format)
Kenneth Graunke [Sun, 19 Aug 2018 07:21:34 +0000 (00:21 -0700)]
iris: proper cache tracking
this is copied from the i965 aux resolve stuff...minus the aux resolves
Kenneth Graunke [Sun, 19 Aug 2018 07:18:51 +0000 (00:18 -0700)]
iris: Move cache tracking to iris_resolve.c
Kenneth Graunke [Sun, 19 Aug 2018 07:31:46 +0000 (00:31 -0700)]
iris: use consistent copyright formatting
some of them had typos, didn't say 'authors or copyright holders',
or other mistakes. This is now https://opensource.org/licenses/MIT
text, formatted consistently.
Kenneth Graunke [Sun, 19 Aug 2018 06:21:41 +0000 (23:21 -0700)]
iris: track depth/stencil writes enabled
Kenneth Graunke [Sun, 19 Aug 2018 06:58:54 +0000 (23:58 -0700)]
iris: Move iris_sampler_view declaration to iris_resource.h
We'll need this for resolve tracking. There's also no genxml stuff here
Kenneth Graunke [Sun, 19 Aug 2018 06:43:14 +0000 (23:43 -0700)]
iris: Move things to iris_shader_state
We didn't originally have this struct, so we had lots of ad-hoc arrays.
Now that we have it, it makes sense to group things there.
Kenneth Graunke [Sun, 19 Aug 2018 06:39:48 +0000 (23:39 -0700)]
iris: move iris_shader_state from ice->shaders.state to ice->state.shaders
it's more state related...
Kenneth Graunke [Sun, 19 Aug 2018 06:44:11 +0000 (23:44 -0700)]
iris: Drop bogus sampler state saving
We do this in an earlier loop. This was just reading things out of the
array, and saving them back over the same array...but in the wrong slots
Kenneth Graunke [Sun, 19 Aug 2018 06:21:32 +0000 (23:21 -0700)]
iris: rename pipe to base