There are too many algebraic optimizations to be certain that one of them
couldn't create instructions which need lowering. It also creates better
code for some reason.
fossil-db (parallel-rdp, Navi):
Totals from 217 (31.77% of 683) affected shaders:
VGPRs: 7716 -> 7672 (-0.57%)
CodeSize: 1516152 -> 1510688 (-0.36%); split: -0.38%, +0.02%
MaxWaves: 3964 -> 3982 (+0.45%)
Instrs: 269445 -> 268508 (-0.35%); split: -0.36%, +0.02%
Cycles: 37963416 -> 37912592 (-0.13%); split: -0.15%, +0.01%
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4791>
load/store vectorization can create 8/16-bit alu to do packing/unpacking,
which would make shader_info::bit_sizes_used out of date.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4791>
We rename it to "modes" to make it clear that it may contain more than
one mode and adjust all the uses of nir_deref_instr::modes to attempt to
handle multiple modes.
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6332>
NIR derefs currently have exactly one variable mode. This is about to
change so we can handle OpenCL generic pointers. In order to transition
safely, we need to audit every deref->mode check. This commit adds a
set of helpers that provide more nuanced mode checks and converts most
of NIR to use them.
For simple cases, we add nir_deref_mode_is and nir_deref_mode_is_one_of
helpers. These can be used in passes which don't have to bother with
generic pointers and just want to know what mode a thing is. If the
pass ever encounters generic pointers in a way that this check would be
unsafe, it will assert-fail to alert developers that they need to think
harder about things and fix the pass.
For more complex passes which require a more nuanced understanding of
modes, we add nir_deref_mode_may_be and nir_deref_mode_must_be helpers
which accurately describe the compiler's best knowledge about the given
deref. Unfortunately, we may not be able to exactly identify the mode
in a generic pointers scenario so we have to be very careful when we use
these. Conversion of these passes is left to later commits.
For the case of mass lowering of a particular mode (nir_lower_explicit_io
is one good example), we add nir_deref_mode_is_in_set. This is also
pretty assert-happy like nir_deref_mode_is but is for a set containment
comparison on deref modes where you expect the deref to either be all-in
or all-out.
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6332>
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Fixes: 6f21995f98 ("radv: add new drirc option radv_enable_mrt_output_nan_fixup")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7423>
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Fixes: bdd7587414 ("radv: use nir_lower_discard_to_demote to work around game bugs")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7423>
I don't know why these values were introduced for but it seems like
we can optimize this by just doing:
gl_SampleMaskIn[0] = (SampleCoverage & (1 << gl_SampleID))
AMDGPU-PRO and AMDVLK apply the same formula to compute the
sample mask when per-sample shading is enabled.
No fossils-db changes.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7377>
Fixes the following building error:
external/mesa/src/amd/compiler/aco_interface.cpp:160:
error: undefined reference to 'aco::form_hard_clauses(aco::Program*)'
Fixes: 3dfbed2a8 ("aco: create s_clause on GFX10+")
Signed-off-by: Mauro Rossi <issor.oruam@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7380>
Use auto to avoid mistyping the constness of the pair key, which
triggers implicit conversions rather than compilation errors.
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7346>
Apply SGPRs/modifiers when possible and try not to break when SDWA
instructions are encountered.
No shader-db changes.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7349>
This seems to give no measurable benefit to Strange Brigade or Shadow of
Mordor, but it's simple to do, helps in theory and all other compilers do
it.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5919>
If the workgroup_size variable is lower than the actual workgroup size,
that means it's possible that ACO won't emit some s_barrier instructions
when in fact it should. This can possibly cause a GPU hang.
This is just for the sake of general correctness, currently this
can't cause a real problem because the maximum vertex count is always
greater than (or equal to) the primitive count in GS, and already
takes into account the number of GS invocations.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7232>
This is a workaround for a bug in Navi 1x NGG HW.
Very rarely, the Navi 1x PA can hang when an NGG workgroup exports
0 total primitives. According to AMD, we always need this workaround
when it is possible that the number of primitives is 0.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7232>
This is to make sure we don't compile a shader which doesn't
fit the available LDS space.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7232>
The p_reduce instruction only works if this operand is in a VGPR,
and otherwise gets lowered to incorrect code.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7232>
Example:
It is possible for some NGG GS waves to have 0 ES and/or GS invocations,
and in that case having an s_barrier inside divergent control flow can
very possibly hang the GPU.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7232>
Streams are really stateful and (IMO) difficult to read for non-trivial
usage. This is also more consistent with NIR and the rest of ACO.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7166>
This seems simpler to me. It should also work correctly when repeated
instructions cross blocks.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7166>
For non-arithmetic opcodes such as buffer_load_dword and buffer_load_short,
default to a definition size of 32.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7276>
The pack_* instructions are now lowered via nir_lower_alu_to_scalar()
and unpack_* are not lowered anymore.
These bitcasts are no-ops, and lowering prevents
some optimizations like vectorization.
Note: There are still some *_split variations remaining
from different other NIR passes.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6527>
Partial rollback as GFX9 really requires height = 1 to work.
The two substantial parts of the fix remaining:
1) Deal with views with multiple levels.
2) Limit the expansion to the base mip pitch/height. On GFX9 this
is exactly equal to the surf_pitch that was used before. I've
done some investigation to make sure that on GFX10 this always
results in the right physical layout.
Remaining stupid question is how the actual extents for bounds
checking never end up too low when the size gets clamped, but
this change and the previous change don't change that ...
Fixes: 1fb3e1fb70 "radv: Fix mipmap extent adjustment on GFX9+."
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7245>
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Neha Bhende <bhenden@vmware.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7260>
MSVC cannot perform GCC __typeof__ for C code. (C++ has decltype.)
Add adjacent functions to allow specifying types manually.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7270>
This doesn't look like it would create correct IR for 8/16-bit phis and
doesn't seem to help anything. If we ever want to do this, it's probably
better done in nir_opt_remove_phis().
No fossil-db changes.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7216>
These are unused now that we almost always use p_parallelcopy for simple
copies.
No fossil-db changes.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7216>
