Marek Olšák
478afbe525
ac: use llvm.amdgcn.kill with LLVM 6.0
...
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-10-24 14:56:34 +02:00
Marek Olšák
1ff9e27cbd
ac: replace ac_build_kill with ac_build_kill_if_false
...
This will be a new LLVM intrinsic and will also work nicely with
llvm.amdgcn.wqm.vote.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-10-24 14:56:34 +02:00
Eric Anholt
34c04c734f
ac: Fix a compiler warning for possibly undefined "name"
...
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-10-23 10:14:40 -07:00
Dave Airlie
1dda214d9c
ac/nir: init full exec mask for merged shaders.
...
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Dave Airlie <airlied@redhat.com>
2017-10-20 01:50:40 +02:00
Marek Olšák
854593b8eb
ac: clean up ac_build_indexed_load function interfaces
...
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-10-17 22:03:03 +02:00
Marek Olšák
bcd3e761a3
ac: properly document a buffer.store LLVM workaround
...
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-10-06 02:56:11 +02:00
Marek Olšák
94d800bfa3
ac: silence a warning
2017-10-04 17:00:05 +02:00
Nicolai Hähnle
052b974fed
amd/common: move ac_build_phi from radeonsi
...
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
2017-10-02 12:17:15 +02:00
Nicolai Hähnle
a6ea4c1b93
amd/common: save an instruction in the build_cube_select sequence
...
Avoid a v_cndmask: the absolute value is free due to input modifiers.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
2017-09-29 11:43:07 +02:00
Nicolai Hähnle
5be5c1e0fa
amd/common: fix build_cube_select
...
Fix the custom cube coord selection sequence to be identical to
the hardware v_cubesc/tc and OpenGL spec. Affects texture sampling
with user-provided derivatives.
Fixes dEQP-GLES3.functional.shaders.texture_functions.texturegrad.*
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
2017-09-29 11:43:04 +02:00
Nicolai Hähnle
94736d31c3
amd/common: add workaround for cube map array layer clamping
...
Fixes dEQP-GLES31.functional.texture.filtering.cube_array.*
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
2017-09-18 11:25:18 +02:00
Nicolai Hähnle
6772452e4c
amd/common: remove has_ds_bpermute argument from ac_build_ddxy
...
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
2017-09-18 11:25:18 +02:00
Nicolai Hähnle
3db86d86ed
amd/common: add chip_class to ac_llvm_context
...
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
2017-09-18 11:25:18 +02:00
Nicolai Hähnle
e0af3bed2c
amd/common: round cube array slice in ac_prepare_cube_coords
...
The NIR-to-LLVM pass already does this; now the same fix covers
radeonsi as well.
Fixes various tests of
dEQP-GLES31.functional.texture.filtering.cube_array.combinations.*
Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
2017-09-18 11:25:18 +02:00
Connor Abbott
b909d278d0
ac: remove bitcast_to_float()
...
ac_to_float() does a superset of what it does.
Signed-off-by: Dave Airlie <airlied@redhat.com>
2017-09-08 04:24:56 +01:00
Connor Abbott
50967cd0b0
ac: move ac_to_integer() and ac_to_float() to ac_llvm_build.c
...
We'll need to use ac_to_integer() for other stuff in ac_llvm_build.c.
Reviewed-by: Dave Airlie <airlied@redhat.com>
2017-09-08 04:24:02 +01:00
Connor Abbott
fafa299511
ac: fix ac_get_type_size() for doubles
...
Reviewed-by: Dave Airlie <airlied@redhat.com>
2017-09-08 04:19:47 +01:00
Connor Abbott
b8a51c8c4b
radeonsi: move the guts of ARB_shader_group_vote emission to ac
...
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2017-09-08 04:12:49 +01:00
Connor Abbott
bd73b89792
radeonsi: move si_emit_ballot() to ac
...
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2017-09-08 04:12:42 +01:00
Connor Abbott
ac27fa7294
radeonsi: move emit_optimization_barrier() to ac
...
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2017-09-08 04:06:47 +01:00
Connor Abbott
c181d4f2b7
radeonsi: move llvm_get_type_size() to ac
...
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2017-09-08 04:04:16 +01:00
Dave Airlie
cb6f16dce9
radeon/ac: use ds_swizzle for derivs on si/cik.
...
This looks like it's supported since llvm 3.9 at least,
so switch over radeonsi and radv to using it, -pro also
uses this. We can now drop creating lds for these operations
as the ds_swizzle operation doesn't actually write to lds at all.
Acked-by: Marek Olšák <marek.olsak@amd.com>
(stable requested due to fixing radv CIK conformance tests)
Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Dave Airlie <airlied@redhat.com>
2017-08-02 00:12:01 +01:00
Nicolai Hähnle
a69afb68c9
radeonsi: use new function ac_build_umin for edgeflag clamping
...
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
2017-07-31 14:55:42 +02:00
Nicolai Hähnle
ac2ab5acad
ac/nir: add always_vector argument to ac_build_gather_values_extended
...
This simplifies a bunch of places that no longer need special treatment
of value_count == 1. We rely on LLVM to optimize away the 1-element vector
types.
This fixes a bunch of bugs where 1-element arrays are indexed indirectly.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
2017-07-31 14:55:42 +02:00
Dave Airlie
ff422500cc
ac/nir: remove last remnants of v16i8
...
llvm doesn't need this workaround anymore.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2017-06-28 20:22:30 +01:00
Nicolai Hähnle
edfd3be77e
ac: add ac_llvm_context::v8i32
...
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2017-06-27 10:28:29 +10:00
Nicolai Hähnle
331a574732
ac: add ac_llvm_context::{i,f}32_{0,1}
...
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2017-06-27 10:28:29 +10:00
Nicolai Hähnle
7bf8c944dc
ac: add ac_llvm_context::{i16, i64, f16, f64}
...
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2017-06-27 10:28:29 +10:00
Marek Olšák
e019ea8f4b
radeonsi: move building llvm.SI.load.const into ac_build_buffer_load
...
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-05-29 01:52:16 +02:00
Marek Olšák
e1942c970f
radeonsi: rename readonly_memory -> can_speculate
...
This is more accurate.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-05-29 01:52:16 +02:00
Marek Olšák
34bc470fa6
ac: fix broken elimination of duplicated VS exports
...
The renumbering code didn't take into account that multiple VS exports
can have the same PARAM index. This also significantly simplifies
the renumbering. Thankfully, we have piglits for this:
spec@arb_gpu_shader5@arb_gpu_shader5-interpolateatcentroid-packing
spec@glsl-1.50@execution@interface-blocks-complex-vs-fs
Reported by Michel Dänzer.
Fixes: b08715499e
("ac: eliminate duplicated VS exports")
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-05-08 19:18:29 +02:00
Marek Olšák
12beef0374
radeonsi: drop support for LLVM 3.8
...
LLVM 3.8:
- had broken indirect resource indexing
- didn't have scratch coalescing
- was the last user of problematic v16i8
- only supported OpenGL 4.1
This leaves us with LLVM 3.9 and LLVM 4.0 support for Mesa 17.2.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-05-05 00:23:44 +02:00
Marek Olšák
4d32b4ac99
radeonsi: stop using v16i8
...
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-05-05 00:23:44 +02:00
Marek Olšák
b08715499e
ac: eliminate duplicated VS exports
...
Only very few shaders have them (from 48486 shaders):
shaders/private/left_4_dead_2/765.shader_test - ac: 1 matches 2
shaders/private/left_4_dead_2/877.shader_test - ac: 1 matches 6
shaders/private/left_4_dead_2/2141.shader_test - ac: 1 matches 6
shaders/private/ue4_effects_cave/11.shader_test - ac: 4 matches 5
shaders/private/ue4_effects_cave/14.shader_test - ac: 5 matches 6
shaders/private/ue4_effects_cave/46.shader_test - ac: 5 matches 6
shaders/private/ue4_effects_cave/42.shader_test - ac: 4 matches 5
shaders/private/ue4_effects_cave/104.shader_test - ac: 4 matches 5
shaders/private/f1-2015/336.shader_test - ac: 3 matches 4
shaders/private/f1-2015/948.shader_test - ac: 6 matches 7
shaders/private/f1-2015/602.shader_test - ac: 0 matches 3
shaders/private/f1-2015/600.shader_test - ac: 0 matches 3
shaders/private/f1-2015/1214.shader_test - ac: 0 matches 1
shaders/private/f1-2015/988.shader_test - ac: 4 matches 5
shaders/private/ue4_elemental/149.shader_test - ac: 3 matches 4
shaders/private/ue4_elemental/346.shader_test - ac: 4 matches 5
shaders/private/ue4_elemental/178.shader_test - ac: 3 matches 4
shaders/private/ue4_elemental/136.shader_test - ac: 4 matches 5
shaders/private/ue4_elemental/168.shader_test - ac: 4 matches 5
shaders/private/ue4_elemental/690.shader_test - ac: 3 matches 4
shaders/private/ue4_elemental/19.shader_test - ac: 5 matches 6
shaders/private/dota2/1901.shader_test - ac: 0 matches 5
shaders/private/dota2/1357.shader_test - ac: 0 matches 5
shaders/private/dota2/1375.shader_test - ac: 0 matches 5
shaders/private/dota2/1369.shader_test - ac: 0 matches 5
shaders/private/dota2/1583.shader_test - ac: 0 matches 5
shaders/private/dota2/1811.shader_test - ac: 0 matches 5
shaders/private/dota2/1893.shader_test - ac: 0 matches 5
shaders/private/dota2/1533.shader_test - ac: 0 matches 5
shaders/private/dota2/1951.shader_test - ac: 0 matches 5
shaders/private/dota2/1361.shader_test - ac: 0 matches 5
shaders/private/mad_max/2792.shader_test - ac: 0 matches 1
shaders/private/mad_max/2794.shader_test - ac: 0 matches 1
shaders/private/mad_max/2780.shader_test - ac: 0 matches 1
shaders/private/mad_max/2902.shader_test - ac: 0 matches 1
shaders/private/bioshock-infinite/3050.shader_test - ac: 3 matches 7
shaders/private/bioshock-infinite/2544.shader_test - ac: 3 matches 6
shaders/private/bioshock-infinite/3062.shader_test - ac: 3 matches 8
shaders/private/bioshock-infinite/2012.shader_test - ac: 3 matches 7
shaders/private/bioshock-infinite/3058.shader_test - ac: 3 matches 7
shaders/private/bioshock-infinite/3270.shader_test - ac: 3 matches 7
shaders/private/bioshock-infinite/732.shader_test - ac: 3 matches 7
shaders/private/bioshock-infinite/3026.shader_test - ac: 3 matches 7
shaders/private/bioshock-infinite/3258.shader_test - ac: 3 matches 6
shaders/private/bioshock-infinite/3198.shader_test - ac: 3 matches 6
shaders/private/bioshock-infinite/3046.shader_test - ac: 3 matches 7
shaders/private/bioshock-infinite/3168.shader_test - ac: 3 matches 6
shaders/private/bioshock-infinite/2550.shader_test - ac: 3 matches 6
shaders/private/bioshock-infinite/3210.shader_test - ac: 3 matches 6
shaders/private/bioshock-infinite/3032.shader_test - ac: 3 matches 6
shaders/private/bioshock-infinite/668.shader_test - ac: 3 matches 7
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-05-03 20:55:00 +02:00
Marek Olšák
7647e90b15
ac: rename ac_eliminate_const_vs_outputs -> ac_optimize_vs_outputs
...
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-05-03 20:55:00 +02:00
Marek Olšák
faa37475e9
ac: first parse VS exports before eliminating constant ones
...
A later commit will make use of this.
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
2017-05-03 20:55:00 +02:00
Dave Airlie
f4743763ce
radeon/ac: remove assert causing regression
...
This assert wasn't in the original radeonsi code but I added
it without totally understanding the original code, it caused
some regressions in variable-indexing tessellation shaders.
Fixes: e2659176
radeonsi/ac: move vertex export remove to common code.
Reported-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2017-04-27 11:38:54 +01:00
Dave Airlie
e2659176ce
radeonsi/ac: move vertex export remove to common code.
...
This code can be shared by radv, we bump the max to
VARYING_SLOT_MAX here, but that shouldn't have too
much fallout.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2017-04-27 05:17:47 +01:00
Samuel Pitoiset
a1c37ff9e4
ac: add unreachable() in ac_build_image_opcode()
...
To silent the following compiler warning:
common/ac_llvm_build.c: In function ‘ac_build_image_opcode’:
common/ac_llvm_build.c:1080:3: warning: ‘name’ may be used uninitialized in this function [-Wmaybe-uninitialized]
snprintf(intr_name, sizeof(intr_name), "%s%s%s%s.v4f32.%s.v8i32",
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
name,
~~~~~
a->compare ? ".c" : "",
~~~~~~~~~~~~~~~~~~~~~~~
a->bias ? ".b" :
~~~~~~~~~~~~~~~~
a->lod ? ".l" :
~~~~~~~~~~~~~~~
a->deriv ? ".d" :
~~~~~~~~~~~~~~~~~
a->level_zero ? ".lz" : "",
~~~~~~~~~~~~~~~~~~~~~~~~~~~
a->offset ? ".o" : "",
~~~~~~~~~~~~~~~~~~~~~~
type);
~~~~~
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
2017-04-10 23:02:12 +02:00
Dave Airlie
10ae83a9c2
radeonsi/ac: move get_image_intr_name to common
...
This code is used in radv, so move to common build code.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
2017-03-07 08:17:03 +10:00
Marek Olšák
7e1faa79d3
radeonsi: drop support for LLVM 3.6 & 3.7
...
They are too old.
Reviewed-by: Dave Airlie <airlied@redhat.com>
2017-03-06 14:13:04 +01:00
Marek Olšák
d5d74fe2b5
radeonsi: set the convergent attribute where needed
...
Reviewed-by: Dave Airlie <airlied@redhat.com>
2017-03-06 14:13:04 +01:00
Marek Olšák
9b08f044be
radeonsi: fix LLVM 3.9 - don't use non-matching attributes on declarations
...
Call site attributes are used since LLVM 4.0.
This also reverts commit b19caecbd6
"radeon/ac: fix intrinsic version check", because this is the correct fix.
Reviewed-by: Dave Airlie <airlied@redhat.com>
2017-03-06 14:13:04 +01:00
Dave Airlie
b19caecbd6
radeon/ac: fix intrinsic version check
...
Reported-by: 375gnu@gmail.com
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100068
Signed-off-by: Dave Airlie <airlied@redhat.com>
2017-03-06 06:05:58 +10:00
Marek Olšák
7f1446a8a1
ac: normalize build helper names
...
s/emit/build/
Reviewed-by: Dave Airlie <airlied@redhat.com>
2017-03-03 17:30:07 +01:00
Marek Olšák
8bde7fb3fc
ac: replace SI.vs.load.input with amdgcn.buffer.load.format
...
Reviewed-by: Dave Airlie <airlied@redhat.com>
2017-03-03 17:30:07 +01:00
Marek Olšák
94811dc66c
radeonsi: move SI.vs.load.input building into amd/common
...
Reviewed-by: Dave Airlie <airlied@redhat.com>
2017-03-03 17:30:07 +01:00
Marek Olšák
97e21cfa25
ac: replace llvm.SI.tbuffer.store with llvm.amdgcn.buffer.store if ADD_TID=0
...
ADD_TID doesn't work. Needs more investigation.
v2: remove leftover dead code
Reviewed-by: Dave Airlie <airlied@redhat.com> (v1)
2017-03-03 15:29:30 +01:00
Marek Olšák
8cfdbba6c7
ac: remove offen parameter from ac_build_buffer_store_dword
...
Reviewed-by: Dave Airlie <airlied@redhat.com>
2017-03-03 15:29:30 +01:00
Marek Olšák
27439dfdae
radeonsi: merge and simplify tbuffer_store functions
...
Reviewed-by: Dave Airlie <airlied@redhat.com>
2017-03-03 15:29:30 +01:00