1459cce7ec
spirv: add some tests for volatile/available/visible
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com >
Acked-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6090 >
2020-09-01 17:15:22 +00:00
fb6b243c11
spirv: Support big-endian strings
...
This should be all that's required for the SPIR-V parser to work in
big-endian systems. SPIR-V requires that everything be in host
byte-order except for strings which are always little-endian.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com >
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6090 >
2020-09-01 17:15:22 +00:00
a8d8fbb9ce
spirv: implement SpvMemoryAccessVolatileMask
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com >
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com >
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6090 >
2020-09-01 17:15:22 +00:00
f81e1d2824
spirv: implement Volatile image operand
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com >
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com >
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6090 >
2020-09-01 17:15:22 +00:00
2ba3ffa76c
spirv: implement Volatile memory semantic
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com >
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com >
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6090 >
2020-09-01 17:15:22 +00:00
23bfba8663
spirv: implement MakePointerAvailable/MakePointerVisible for OpCopyMemory
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com >
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com >
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6090 >
2020-09-01 17:15:22 +00:00
e01d1a9f16
spirv: add vtn_emit_make_{visible,available}_barrier helpers
...
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com >
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com >
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6090 >
2020-09-01 17:15:22 +00:00
b85c38a86f
spirv: make OpLoad/OpStore visibility/availablity barriers acquire/release
...
I think these are needed to order the visibility/availability operation
with the access.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com >
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com >
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6090 >
2020-09-01 17:15:22 +00:00
5b92392c48
spirv: fix Uniform and Output MemoryAccessMakePointer{Visible,Available}
...
The Uniform storage class can be used for SSBOs. This should also fix make
available/visible for the Output storage class.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com >
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com >
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6090 >
2020-09-01 17:15:22 +00:00
857b9c5027
spirv: Add a vtn_get_mem_operands() helper
...
Add a vtn_get_mem_operands() helper to extract memory operand attached
to load/store operations.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com >
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6090 >
2020-09-01 17:15:22 +00:00
bc123c396a
nir/algebraic: mark some optimizations with fsat(NaN) as inexact
...
If a is Nan, fsat(NaN) is expected to be 0 and some optimizations
should be marked as inexact.
Fixes a GPU hang with Death Stranding and RADV/ACO (RADV/LLVM
isn't affected because it lowers fsat).
No fossils-db change.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3368
Cc: mesa-stable
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com >
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com >
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl >
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6519 >
2020-09-01 11:20:03 +02:00
b778e7bd6c
nir/vtn: Convert constant samplers to variables with data
...
Reviewd-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5242 >
2020-08-31 21:17:39 +00:00
397730edd5
nir/vtn: Add intrinsics for CL image format/order queries
...
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5242 >
2020-08-31 21:17:39 +00:00
5ce6821900
nir/vtn: ImageSizeLod op can be applied to images
...
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5242 >
2020-08-31 21:17:39 +00:00
701cb9d60c
nir/vtn: Handle integer sampling coordinates
...
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5242 >
2020-08-31 21:17:39 +00:00
a196f05fc2
nir/vtn: Use return type rather than image type for tex ops
...
Since OpenCL images don't have types, we can't use the image type here.
Rather than special-casing and only using SPIR-V return type for CL images,
we can just always use the return type to fill out the tex info.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5242 >
2020-08-31 21:17:39 +00:00
de36b5b805
nir/vtn: Add support for kernel images to SPIRV-to-NIR.
...
There's a few quirks: kernel images are untyped, whether they're
sampled is unknown, and they're passed as inputs to the kernel even though
SPIR-V declares their address space as UniformConstant.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5242 >
2020-08-31 21:17:39 +00:00
ce6f66242a
nir/vtn: Add type constant to image intrinsics
...
Since OpenCL supports untyped images, backends might need type info
to be able to support the load/store ops.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5242 >
2020-08-31 21:17:38 +00:00
865a2ad086
clover/nir/spirv: Use uniform rather than shader_in for kernel inputs
...
The semantics of inputs for CL are a closer match to the semantics of uniforms for graphics.
Rather than cross-stage data, it's data that every thread sees uniformly.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Reviewed-by: Karol Herbst <kherbst@redhat.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6497 >
2020-08-31 19:58:14 +00:00
73d2c6cdce
nir: Switch the indexing of block->live_in/out arrays.
...
In nir-to-tgsi, I want to free temps storing SSA values when they go dead,
and NIR liveness has most of the information I need. Hoever, when I reach
the end of a block, I need to free whatever temps were in liveout which
are dead at that point. If liveout is indexed by live_index, then I don't
know the maximum live_index for iterating the live_out bitset, and I also
don't have a way to map that index back to the def->index that my temps
are stored under.
We can use the more typical def->index for these bitsets, which resolves
both of those problems. The only cost is that ssa_undefs don't get merged
into a single bit in the bitfield, but there are generally 1-4 of them in
a shader and we don't track liveness for those anyway so splitting them
apart is fine.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6408 >
2020-08-31 18:28:36 +00:00
5d2b2b59c4
nir: Invalidate live SSA def information when making new SSA defs.
...
Noted by @jekstrand in review of the SSA live index changes.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6408 >
2020-08-31 18:28:36 +00:00
ee288f293b
nir: add shared/global atomics to nir_get_io_offset_src()
...
Signed-off-by: Italo Nicola <italonicola@collabora.com >
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com >
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6521 >
2020-08-31 17:36:12 +00:00
e281ee04df
nir: Look up the shader when printing a single instruction.
...
If you've inserted the instruction into a block, then we can get to the
shader. This improves our instruction output, giving you i/o semantics
and variable names in intrinsics.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6496 >
2020-08-31 17:20:43 +00:00
a3a8322dcd
nir: Fix printing of individual instructions with io semantics.
...
The state->shader is missing when used outside of nir_print_shader, just
drop these details in that case. We can fix nir_print_instr() to look up
the shader, but let's also make sure that an instr detached from a shader
(such as one you're constructing but haven't yet inserted) still works.
Fixes: 2b1ef5df4ejason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6496 >
2020-08-31 17:20:43 +00:00
ac6b8e42ce
nir: Take a variable remap parameter in nir_inline_function_impl
...
Acked-by: Daniel Stone <daniels@collabora.com >
Reviewed-by: Jesse Natalie <jenatali@microsoft.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6411 >
2020-08-29 18:32:46 +00:00
b6f31898a2
nir: Use a switch in nir_inline_function_impl
...
Acked-by: Daniel Stone <daniels@collabora.com >
Reviewed-by: Jesse Natalie <jenatali@microsoft.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6411 >
2020-08-29 18:32:46 +00:00
31f75aaeef
nir/lower_discard_to_demote: Use nir_shader_instructions_pass().
...
Cleans up indentation, and clears the metadata tracking flag that would
break if this pass was used in in NIR_PASS().
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6412 >
2020-08-27 23:01:52 +00:00
265dcb3836
nir/nir_lower_wrmasks: Use the nir_lower_instructions_pass() helper.
...
This fixes the invalidation of metadata when we didn't modify the shader
and unindents a bunch of code.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6412 >
2020-08-27 23:01:52 +00:00
c5e64c041f
nir/lower_io_to_scalar: Convert to use nir_shader_instructions_pass().
...
This unindents a whole bunch of code, and fixes the lack of metadata
tracking in the pass (which wasn't called in an opt loop so it hadn't been
caught before).
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6412 >
2020-08-27 23:01:52 +00:00
dfb8465341
nir/opt_undef: Convert to use nir_shader_instructions_pass().
...
We can't use nir_lower_instructions because we operate on stores which
don't have an SSA def.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6412 >
2020-08-27 23:01:52 +00:00
e5d4bbd840
nir/lower_vec_to_movs: Convert to use nir_shader_instructions_pass().
...
Less pass code, less indenting, should be the same perf.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6412 >
2020-08-27 23:01:52 +00:00
a62098fff2
nir: Add a helper for general instruction-modifying passes.
...
The nir_shader_lower_instructions() is really nice, but it's only for SSA
operations, and sometimes you want something more general. I've put it in
nir_builder.h so it can be inlined and retain the same performance
characteristics we're used to in our lowering passes even in the absence
of LTO.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6412 >
2020-08-27 23:01:52 +00:00
a82f664d0a
nir/opt_copy_prop_vars: Quiet valgrind warning about overlapping memcpy.
...
The warning is kind of silly:
Test case 'dEQP-GLES2.functional.shaders.indexing.tmp_array.vec3_const_write_static_read_vertex'..
==1874780== Source and destination overlap in memcpy(0xa261690, 0xa261690, 160)
==1874780== at 0x484D498: __GI_memcpy (vg_replace_strmem.c:1037)
==1874780== by 0x596FC07: copy_entry_remove (nir_opt_copy_prop_vars.c:296)
The "memcpy is undefined if they overlap" thing is surely meant to be
"memcpy with *partial* overlap is undefined", but let's keep anyone else
from having to debug this.
Reviewed-by: Rob Clark <robdclark@chromium.org >
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6178 >
2020-08-27 21:39:50 +00:00
608c87afdd
nir/vtn: SPIR-V bit count opcodes (core and extension) dest size mismatches nir
...
SPIR-V dest sizes match the input, while nir is always int32. Insert
casts from the nir op to the expected SPIR-V dest.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com >
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6313 >
2020-08-27 16:57:42 +00:00
a54695ddcb
nir: Add bit_count to lower_int64 pass
...
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com >
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6313 >
2020-08-27 16:57:42 +00:00
d91f85f16e
nir: Remove 32bit restriction for uadd_carry optimization
...
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6313 >
2020-08-27 16:57:42 +00:00
9232887c69
nir: Implement mul_high lowering for bit sizes other than 32
...
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com >
Acked-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6313 >
2020-08-27 16:57:42 +00:00
ea715741b5
nir_lower_bit_size: Support lowering ops with differing source/dest sizes
...
Specifically the bit-finding routines always return int32. Don't complain
about the dest already being 32 bits when lowering to 32 bits, and
don't bother casting the dest if it's already right.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com >
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6313 >
2020-08-27 16:57:42 +00:00
0d595f72b2
nir: Relax opt_if logic to prevent re-merging 64bit phis for loop headers
...
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6313 >
2020-08-27 16:57:42 +00:00
43d22c8f20
nir: Add a lowering pass to split 64bit phis
...
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6313 >
2020-08-27 16:57:41 +00:00
27e6117ee9
nir: Report progress properly in nir_lower_bool_to_*
...
All three passes have the same bug where, in the mov/vec case they
unconditionally return true even if they don't change anything. Throw
in a bit size check so they return false properly.
Reviewed-by: Karol Herbst <kherbst@redhat.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6435 >
2020-08-25 22:39:08 +00:00
2a9ffc24cc
nir: Move new edgeflag assert into the io_lowered case
...
We only need to assert this in the `io_lowered` case, which actually
uses num_outputs. This assert also doesn't appear to hold on iris,
where num_outputs is showing up as 0 (because it's likely not yet set).
Fixes assertion failures in edgeflag related tests on iris, which
doesn't use the io_lowered path currently.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3456
Fixes: 484a60d547eric@anholt.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6450 >
2020-08-25 16:43:19 +00:00
2fcfcca842
nir: Copy semantics to nir_intrinsic_load_fs_input_interp_deltas
...
When using nir_lower_interpolation, we need to propagate the IO
semantics from the load_interpolated_input to the new
load_fs_input_interp_deltas intrinsics. nir_lower_io assumes
they will be filled out.
This fixes assertions in most tests on iris since commit
01ab308edc01ab308edc502abfce7feric@anholt.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6450 >
2020-08-25 16:43:19 +00:00
a79dad950b
nir,amd: remove trinary_minmax opcodes
...
These consist of the variations nir_op_{i|u|f}{min|max|med}3 which are either
lowered in the backend (LLVM) anyway or can be recombined by the backend (ACO).
Reviewed-by: Marek Olšák <marek.olsak@amd.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6421 >
2020-08-24 20:56:11 +00:00
484a60d547
nir: generate lowered IO in nir_lower_passthrough_edgeflags
...
Reviewed-by: Eric Anholt <eric@anholt.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6442 >
2020-08-24 19:07:18 +00:00
7980f3e519
nir: add interpolation qualifiers for color sysvals into shader_info
...
needed by radeonsi
Reviewed-by: Eric Anholt <eric@anholt.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6442 >
2020-08-24 19:07:18 +00:00
4d36dceeab
nir: add shader_info::io_lowered
...
This will drive decisions in many NIR passes and st/mesa.
Reviewed-by: Eric Anholt <eric@anholt.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6442 >
2020-08-24 19:07:18 +00:00
844167d100
nir: properly identify texcoords for lowered IO in nir_lower_drawpixels
...
Reviewed-by: Eric Anholt <eric@anholt.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6442 >
2020-08-24 19:07:18 +00:00
2b1ef5df4e
nir: print IO semantics (v2)
...
v2: print GS streams readably
Reviewed-by: Eric Anholt <eric@anholt.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6442 >
2020-08-24 19:07:18 +00:00
01ab308edc
nir: update IO semantics in nir_io_add_const_offset_to_base
...
Reviewed-by: Eric Anholt <eric@anholt.net >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6442 >
2020-08-24 19:07:18 +00:00
Rhys Perry