There are blit shader specific optimizations available.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9002>
Now that we lower FAU correctly, we don't need to write the extra move
explicitly, it will be lowered in later.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8973>
Move and reshape bi_lower_fau to bi_schedule.c. This generalizes the
pass for FAU reads, allowing copyprop to work with FAU without problems.
The pass must run immediately before scheduling. Its post-conditions are
directly specified as the scheduler's pre-conditions. It momentarily
will depend on internal scheduler predicates. It is, for all intents and
purposes, part of the scheduler. Keep it all together.
Finally, adjust the 0 handling to avoid a move at the expense of
constrained scheduling of something like `FADD.v2f16.clamp_0_1 u0, #0`
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8973>
Will prevent failures when we start using FAU together with modifiers in
a few commits.
Fixes: fc7770b1dd ("pan/bi: Add trivial rewrite helpers")
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8973>
Ensure we don't fall over if we have an instruction like
FADD.f32 u0, #0
In this case, the tuple's FAU requirement implies the instruction can be
scheduler without lowering to the FMA slot but not the ADD slot.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8973>
Uses "u3, u3[1]" syntax which is close enough to the assembly syntax
"u3.w0, u3.w1".
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8973>
I keep open-coding this, incorrectly... Since bi_index contains both
"position" and "modifier" data, it's common to want to swap the position
while preserving modifiers.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8973>
Will be used to generalize RMU on Midgard and also to support Bifrost's
FAU (which is essentially the same thing).
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8973>
This makes UBO 0 less special, allowing us to generalize uniform
optimization. Note this disables RMU on Midgard as we're about to
rewrite the RMU mechanism.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8973>
LD_UNIFORM supports constant shifts and biases, just like LD, so take
advantage of that. Will avoid a regression in code quality from lowering
uniforms to UBOs.
No shader-db changes.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8973>
Helpful to disable the scheduler when debugging, so the assembly can be
compared against the NIR directly when lost in a big dEQP test.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8973>
These cannot schedule to the last tuple, there needs to be other work in
the last clause, or a tuple of NOP's failing that. Can occur depending
on scheduling of CUBEFACE instructions, and will apply to computational
atomics in the near future.
Fixes: 77933d16d8 ("pan/bi: Switch to new scheduler")
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8962>
This incorrectly assumed the last tuple was nontrivial, an invariant
that holds by accident on the current scheduler but will break
momentarily. Fix this, and document the invariants.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8962>
Forces a trivial schedule to replicate the old behaviour (for debugging
or benchmarking). Actually the new scheduler is still used, just highly
constrained; the net result should still do what's expected.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
Replicate the pattern, greedily select clauses until we run out of
instructions.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
Once we've merged the clauses' constants, we need to....
1. Swap where necessary so non-pcrel constants are correctly encoded.
2. Swap where necessary so pcrel constants are in canonical positions.
3. Force M1 values for pcrel constants and final single constants.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
These work as you would expect: first handling paired constants
(swapping to a canonical form to deduplicate), then handling unpaired
constants (packing together in a canonical form). Most of the added
complexity is from pcrel handling, but we impose strict invariants on
pcrel (no more than one PC-relative constant per clause, only M1=4 mode)
without which the algorithm would be even uglier.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
Based on the tuple state's constants, satisfying the pcrel invariant we
impose.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
Mimicks the one previously done at pack time, but designed for schedule
time instead.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
We need to do certain rewrites during scheduling before RA runs in order
to satsify scheduler post-conditions.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
It's a hardware invariant, and useful for the scheduler (not just
packing).
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
Rather specialized but keeps down obnoxious indentation in scheduler
passes.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
Useful for scheduling decisions as well as packing.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
Passthroughs are _required_ for correct scheduling, so we have to handle
this now. The legitimacy of using passthroughs is justified by the
constraint checks and verified with asserts at pack time.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
Wrapper to select the best legal instruction, pop it off the worklist,
update the clause/tuple states, and return it.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
In the future we'll want a heuristic minimizing register pressure but
for in-order this will suffice.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
Using the previously defined checks for architectural scheduling
constraints, define one top-level predicate to check if an instruction
on the worklist is ready for scheduling.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
ATEST is a bit of a wrinkle in this, so let's keep it in one place.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
Not visible on real shaders yet, but it will be when we schedule
out-of-order (or implement 64-bit multiplication).
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
The number of register writes in a tuple must be bounded by a number
based on the number of register reads in the succeeding tuple, since
writes and reads are interleaved. This helper calculates that number of
reads, noting that not every read actually counts - if the preceding
tuple writes to a read source, that will become a temporary instead of
consuming a register slot.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
To determine the number of register reads in a tuple (which must be
bounded by 3, or 5 if you force writes), we need to count "new" sources:
those that are not already in a partially scheduled tuple.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
This comes in destructive and nondestructive flavours, to be used to
insert an instruction into a tuple and check if an instruction is
insertable respectively. It is responsible for FAU slot matching.
It's annoying this sort of logic is duplicated in 3 places
(bi_lower_fau, here, and packing) but they each work with different sets
of assumptions...
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
Needed to satisfy max constant constaints. These aren't precise but
they should be a good enough approximation for now.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
In the near future we'll schedule out-of-order via a dependendency graph
and worklist. For now, emulate in-order operation.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>
