fb4637797e
iris/compute: Provide binding table entry for gl_NumWorkGroups
...
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com >
2019-02-21 10:26:09 -08:00
fcd0364857
iris/compute: Wait on compute batch when mapping
...
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com >
2019-02-21 10:26:09 -08:00
ea416d0b5d
iris/program: Don't try to push ubo ranges for compute
...
We only can push constants for compute shaders from one range.
Gallium glsl-to-nir (src/mesa/state_tracker/st_glsl_to_nir.cpp) lowers
all uniform accesses to a ubo.
Unfortunately we also load the subgroup-id as a uniform in the
compiler. Since we use the 1 push range for this subgroup-id, we then
lose the ability to actually push the ubo with all the normal user
uniform values.
In other words, there is lots of room for performance improvement, but
at least retrieving the uniforms as pull-constants is functional for
now.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com >
2019-02-21 10:26:09 -08:00
c7cfa4000f
iris/compute: Get group counts from grid->grid
...
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com >
2019-02-21 10:26:09 -08:00
fd9ccd8b5d
iris/compute: Flush compute batches
...
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com >
2019-02-21 10:26:09 -08:00
9b5cda95aa
iris/compute: Add MEDIA_STATE_FLUSH following WALKER
...
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com >
2019-02-21 10:26:09 -08:00
6ebd04ac8f
iris: Add iris_restore_compute_saved_bos
...
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com >
2019-02-21 10:26:09 -08:00
622aaa290f
iris: Add IRIS_DIRTY_CONSTANTS_CS
...
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com >
2019-02-21 10:26:09 -08:00
25f1625edf
iris/compute: Set mask bits on PIPELINE_SELECT
...
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com >
2019-02-21 10:26:09 -08:00
9fc672428d
iris: little bits of compute basics
2019-02-21 10:26:09 -08:00
860ce6af3f
iris: drop XXX's about swizzling
...
pretty sure this is unnecessary on modern HW
2019-02-21 10:26:09 -08:00
12de56f53d
iris: drop dead format //'s
...
these just aren't supported
2019-02-21 10:26:09 -08:00
f6c68066a6
iris: yes
2019-02-21 10:26:09 -08:00
752abeb690
iris: initial compute caps
...
RET macro borrowed from freedreno
2019-02-21 10:26:09 -08:00
4da28c2c22
iris: Enable fb fetch
...
needed for ES 3.2
2019-02-21 10:26:09 -08:00
be905bd461
iris: advertise GL_ARB_shader_texture_image_samples
2019-02-21 10:26:09 -08:00
6441e906e8
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 >
2019-02-21 10:26:09 -08:00
c29fd34259
iris: move images next to textures in binding table
2019-02-21 10:26:09 -08:00
0d9c5b4e7e
iris: null for non-existent cbufs
...
prevents BTs from being shifted down incorrectly
2019-02-21 10:26:09 -08:00
98e8f80e7d
iris: actually set image access
2019-02-21 10:26:09 -08:00
d9aee25a46
iris: Don't lower image formats for write-only images
2019-02-21 10:26:09 -08:00
a06f0fe517
iris: set image access correctly
2019-02-21 10:26:09 -08:00
5d1dadfc38
iris: bother with BTIs
2019-02-21 10:26:09 -08:00
f5b887da6c
iris: implement set_shader_images hook
2019-02-21 10:26:09 -08:00
26a54ae4b2
iris: lower storage image derefs
2019-02-21 10:26:09 -08:00
e97a24da89
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...
2019-02-21 10:26:09 -08:00
28b41992c8
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
2019-02-21 10:26:09 -08:00
6e7957a22d
iris: enable I/L formats
2019-02-21 10:26:09 -08:00
bfbebbaa36
iris: Use R/RG instead of I/L/A when sampling
2019-02-21 10:26:09 -08:00
94569a6458
iris: rework format translation apis
2019-02-21 10:26:09 -08:00
b9eeed3e8f
iris: Allow PIPE_CONTROL with Stall at Scoreboard and RT flush
...
It's nonsensical, but not illegal, and mandatory on Icelake
2019-02-21 10:26:09 -08:00
65d1cda995
iris: add gen11 to genX_call
2019-02-21 10:26:09 -08:00
0fdcb20803
iris: inline stage_from_pipe to avoid unused warnings
2019-02-21 10:26:09 -08:00
6fbb6ba290
iris: pipe to scs -> iris_pipe.h
2019-02-21 10:26:09 -08:00
87351b8dfe
iris: force persample interp cap
2019-02-21 10:26:09 -08:00
90b9efc1f9
iris: stencil texturing
2019-02-21 10:26:09 -08:00
9b229d266d
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
2019-02-21 10:26:09 -08:00
822f91508e
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
2019-02-21 10:26:09 -08:00
294ce58a30
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.
2019-02-21 10:26:09 -08:00
2f24000662
iris: drop a bunch of pipe_sampler_state stuff we don't need
2019-02-21 10:26:09 -08:00
c6016d3761
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...
2019-02-21 10:26:09 -08:00
3c0ef22edb
iris: Enable ARB_shader_vote
...
The easiest get out the vote campaign ever
2019-02-21 10:26:08 -08:00
0395eba20f
iris: magic number 36 -> #define
2019-02-21 10:26:08 -08:00
57f8a623c5
iris: better query file comment
2019-02-21 10:26:08 -08:00
d3a5d87219
iris: early return properly
2019-02-21 10:26:08 -08:00
07ff8c752f
iris: 36-bit overflow fixes
2019-02-21 10:26:08 -08:00
dff174c103
iris: Need to | 1 when asking for timestamps
2019-02-21 10:26:08 -08:00
1d91eba7dc
iris: glGet timestamps, more correct timestamps
2019-02-21 10:26:08 -08:00
36fbcfb06c
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
2019-02-21 10:26:08 -08:00
ec82be57e8
iris: timestamps
2019-02-21 10:26:08 -08:00
Jordan Justen