Allows a shader to have multiple ray queries without spilling them to a shadow
stack. Instead, the driver provides the shader with an array of multiple
RTDispatchGlobals structs to give each query its own dedicated stack.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38778>
Since the move to MEMORY_*_LOGICAL the result value was being ignored, so
change to use that.
Since the conversion to use new registers, some issues were introduced:
- Even with `has_64bit_int` ADD with 64-bit immediate value is not supported;
- `dst_high` was not being filled if there was no overflow;
- Only `dst_low` returned.
Found when writing some new code involving large block loads.
Fixes: b79e85a93f ("brw: always use new registers for load address increments")
Fixes: b55f77161d ("intel/brw: Switch to emitting MEMORY_*_LOGICAL opcodes")
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39282>
Each group of 16 lanes inside a SIMD32 shader will load different globals.
In SIMD8/16 shaders, the divergence analysis will turn this load into
nir_load_global_constant_uniform_block_intel.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: mesa-stable
Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36181>
If it wasn't for the workaround, it wouldn't be necessary to track the
whether instructions are exec_all or not. The workaround affects
results when mixing a dep and inst with different exec_all.
Add the predicate so that, when the workaround is disabled, none of
the effects of having different exec_all will kick in, all them will
be considered `exec_all = true`.
This patch don't change any behavior, just adds the predicate.
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36659>
nr_params & params array are gone.
brw_ubo_range is not stored on the prog_data structure anymore (Anv
already stored a copy of that with its own additional information)
The backend now only deals with load_push_data_intel. load_uniform &
load_push_constant have to be lowered by the driver.
Pre Gfx12.5 platforms have to provide a subgroup_id_param to specify
where the subgroup_id value is located in the push constants.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38975>
Anv already manages this itself. This allows removing the logic from
the compiler.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38975>
Drivers can do all the lowering to push constants to find the only
value useful in that array (subgroup_id). Then drivers call into
brw_cs_fill_push_const_info() to get the cross/per thread constant
layout computed in the prog_data.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38975>
The current code walks the instructions, and when needed,
it will scan to find the next "end of scope" and sometimes
the next "end of block". It also has a separate patching
logic for HALTs.
The new code collects the necessary scope information up front,
then walks the instruction backwards, making avoiding the need
to scan for the end of scope. It will also walk only the
relevant instructions that were previously collected. It also
replaces the previous HALT-specific patching logic.
With this new change, many cases that were jumping to
intermediate HALTs, will now jump straight to the end of
scope (or the "end of the program" section). E.g. in
```
if
...
(...) HALT
...
(...) HALT
endif
```
both HALTs now will jump to the end of the scope, instead of the
first HALT jumping into the second one.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38914>
Most of instructions follow the basic formats (1, 2 and 3 src), so
consolidate their emission code in generator.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38878>
Move validation, noting that LRP only supports BRW_TYPE_F -- the
previous assert had DF because it also was used by MAD in the past.
With that change, ALU3F can be replaced by ALU3 for LRP.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38878>
When repctrl is used, the swizzle/chansel is ignored. Instead of setting
a swizzle that has all zeros and encode that, don't encode anything.
For context see e7598c5a62 ("intel/compiler: Set swizzle to BRW_SWIZZLE_XXXX
for scalar region").
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38878>
There is no users for that function, is_volatile is only used in
brw_opt_cse.cpp is_expression() but it access the information using brw_send_inst
struct.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39104>
We don't need one bit per bitsize per instruction if only one actually
matters in the end.
First step towards moving NIR in the direction of full float_controls2
only.
Also rename this from fp_fast_math, because that name implied that 0 is
the no fast math mode, while the opposite was the case.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39026>
Decode logic in Gfx12+ has become complex with the new types, so Caio
suggested that we move to the table like other gens.
Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39007>
These type encodings were first were used in dpas instructions, but
continue to be used in more places.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39007>
Later commits will call DCE after lowering has been performed. Creating
more things that would need lowering is problematic.
No shader-db or fossil-db changes on any Intel platform.
Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38315>
Later commits will call cmod prop after register allocation. At that
time, there is only FIXED_GRF.
No shader-db or fossil-db changes on any Intel platform.
v2: FIXED_GRF uses subnr instead of offset. Add a unit test to
demonstrate the issue.
Reviewed-by: Caio Oliveira <caio.oliveira@intel.com> [v1]
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38315>
PRMs for G35 (Gfx4) through Ivy Bridge (Gfx7) all say that conditional
modifiers are allowed for MACH. Starting with Haswell (Gfx7.5), this
seems to be removed. This function doesn't have any way to know the
platform, so false is returned for all platforms.
No shader-db or fossil-db changes on any Intel platform.
Prevents a failure in "brw: Do cmod prop again after post-RA scheduling"
in piglit's builtin-uint-mad_sat-1.0.generated.cl.
Cc: stable
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38315>
The group implicity selects which flags the instruction can write. This
was discovered while working on another set of changes that could change
some logical operations into predicated MOV instructions.
Prevents regressions later in the series in
dEQP-VK.graphicsfuzz.cov-loop-fragcoord-identical-condition.
No shader-db or fossil-db changes on any Intel platform.
v2: Update the comment in the test case. Suggested by Caio.
Fixes: 95ac3b1dae ("i965/fs: don't propagate cmod when the exec sizes differ")
Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38315>
We leave GS pushed inputs using load_per_vertex_input for now - they're
relatively simple, and using load_attribute_payload doesn't work well
since it's assumed to be convergent (for TES, FS inputs) while GS inputs
are divergent.
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38990>
We're going to be deciding on push vs. pull in the NIR lowering pass
soon, so move the code to limit our register usage from brw's thread
payload code to brw_nir_lower_gs_inputs().
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38990>
Avoid using exec_node::remove() and the initial "main list of
instructions", and instead use the existing helpers like other
passes.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37146>
