First, we need to give the parent_instr field a unique name to be able to
replace with a helper. We have parent_instr fields for both nir_src and
nir_def, so let's rename nir_src::parent_instr in preparation for rework.
This was done with a combination of sed and manual fix-ups.
Then we use semantic patches plus manual fixups:
@@
expression s;
@@
-s->renamed_parent_instr
+nir_src_parent_instr(s)
@@
expression s;
@@
-s.renamed_parent_instr
+nir_src_parent_instr(&s)
@@
expression s;
@@
-s->parent_if
+nir_src_parent_if(s)
@@
expression s;
@@
-s.renamed_parent_if
+nir_src_parent_if(&s)
@@
expression s;
@@
-s->is_if
+nir_src_is_if(s)
@@
expression s;
@@
-s.is_if
+nir_src_is_if(&s)
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Acked-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24671>
Undef will be replaced by NaN whenever it leads to elimination of FP
instructions. This implements the elimination part.
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24792>
Instead, we replace it directly with nir_def. We could replace it with
nir_dest but the next commit gets rid of that so this avoids unnecessary
churn. Most of this commit was generated by sed:
sed -i -e 's/dest.dest.ssa/def/g' src/**/*.h src/**/*.c src/**/*.cpp
There were a few manual fixups required in the nir_legacy.c and
nir_from_ssa.c as nir_legacy_reg and nir_parallel_copy_entry both have a
similar pattern.
Acked-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24674>
These are mostly just obvious patterns that somebody will eventually
want to add.
DG2, Tiger Lake, Ice Lake, Skylake, Broadwell, and Haswell had similar
results (Ice Lake shown)
total instructions in shared programs: 20570033 -> 20570026 (<.01%)
instructions in affected programs: 7363 -> 7356 (-0.10%)
helped: 6 / HURT: 0
total cycles in shared programs: 902118781 -> 902118854 (<.01%)
cycles in affected programs: 419132 -> 419205 (0.02%)
helped: 4 / HURT: 2
DG2, Tiger Lake, Ice Lake, and Skylake had similar results (Ice Lake shown)
Totals:
Instrs: 152819500 -> 152819380 (-0.00%)
Cycles: 15014627187 -> 15014624437 (-0.00%)
Totals from 115 (0.02% of 662497) affected shaders:
Instrs: 28963 -> 28843 (-0.41%)
Cycles: 404582 -> 401832 (-0.68%)
Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19968>
v2: Require that the constant value be representable as either uint16_t
or int16_t. Suggested by Matt.
v3: Remove redundant patterns. Noticed by Matt.
shader-db:
DG2
total instructions in shared programs: 23103767 -> 23103577 (<.01%)
instructions in affected programs: 51822 -> 51632 (-0.37%)
helped: 98 / HURT: 15
total cycles in shared programs: 842347714 -> 842380017 (<.01%)
cycles in affected programs: 1942595 -> 1974898 (1.66%)
helped: 97 / HURT: 32
Nearly all of the affected shaders (around 9,900) are shaders in
Cyberpunk 2077. It's about an even split between vertex and fragment
shaders. The majority of the remaining affected shaders (3,600) are
from Strange Brigade. This was also a nearly even split between
fragment and vertex.
All but two of the lost shaders are SIMD32 fragment shaders in
Cyberpunk 2077. The other two are SIMD32 fragment shaders in Dota2.
fossil-db:
DG2
Instructions in all programs: 196379107 -> 196248608 (-0.1%)
helped: 13467 / HURT: 1210
Cycles in all programs: 13931355281 -> 13929955971 (-0.0%)
helped: 11801 / HURT: 2922
Lost: 90
Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23262>
Every nir_ssa_def is part of a chain of uses, implemented with doubly linked
lists. That means each requires 2 * 64-bit = 16 bytes per def, which is
memory intensive. Together they require 32 bytes per def. Not cool.
To cut that memory use in half, we can combine the two linked lists into a
single use list that contains both regular instruction uses and if-uses. To do
this, we augment the nir_src with a boolean "is_if", and reimplement the
abstract if-uses operations on top of that list. That boolean should fit into
the padding already in nir_src so should not actually affect memory use, and in
the future we sneak it into the bottom bit of a pointer.
However, this creates a new inefficiency: now iterating over regular uses
separate from if-uses is (nominally) more expensive. It turns out virtually
every caller of nir_foreach_if_use(_safe) also calls nir_foreach_use(_safe)
immediately before, so we rewrite most of the callers to instead call a new
single `nir_foreach_use_including_if(_safe)` which predicates the logic based on
`src->is_if`. This should mitigate the performance difference.
There's a bit of churn, but this is largely a mechanical set of changes.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22343>
If there is a single use of fmul, and that single use is fadd, it makes
sense to fuse ffma, as we already do. However, if there are multiple
uses, fusing may impede code gen. Consider the source fragment:
a = fmul(x, y)
b = fadd(a, z)
c = fmin(a, t)
d = fmax(b, c)
The fmul has two uses. The current ffma fusing is greedy and will
produce the following "optimized" code.
a = fmul(x, y)
b = ffma(x, y, z)
c = fmin(a, t)
d = fmax(b, c)
Actually, this code is worse! Instead of 1 fmul + 1 fadd, we now have 1
fmul + 1 ffma. In effect, two multiplies (and a fused add) instead of
one multiply and an add. Depending on the ISA, that could impede
scheduling or increase code size. It can also increase register
pressure, extending the live range.
It's tempting to gate on is_used_once, but that would hurt in cases
where we really do fuse everything, e.g.:
a = fmul(x, y)
b = fadd(a, z)
c = fadd(a, t)
For ISAs that fuse ffma, we expect that 2 ffma is faster than 1 fmul + 2
fadd. So what we really want is to fuse ffma iff the fmul will get
deleted. That occurs iff all uses of the fmul are fadd and will
themselves get fused to ffma, leaving fmul to get dead code eliminated.
That's easy to implement with a new NIR search helper, checking that all
uses are fadd.
shader-db results on Mali-G57 [open shader-db + subset of closed]:
total instructions in shared programs: 179491 -> 178991 (-0.28%)
instructions in affected programs: 36862 -> 36362 (-1.36%)
helped: 190
HURT: 27
total cycles in shared programs: 10573.20 -> 10571.75 (-0.01%)
cycles in affected programs: 72.02 -> 70.56 (-2.02%)
helped: 28
HURT: 1
total fma in shared programs: 1590.47 -> 1582.61 (-0.49%)
fma in affected programs: 319.95 -> 312.09 (-2.46%)
helped: 194
HURT: 1
total cvt in shared programs: 812.98 -> 813.03 (<.01%)
cvt in affected programs: 118.53 -> 118.58 (0.04%)
helped: 65
HURT: 81
total quadwords in shared programs: 98968 -> 98840 (-0.13%)
quadwords in affected programs: 2960 -> 2832 (-4.32%)
helped: 20
HURT: 4
total threads in shared programs: 4693 -> 4697 (0.09%)
threads in affected programs: 4 -> 8 (100.00%)
helped: 4
HURT: 0
v2: Update trace checksums for virgl due to numerical differences.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18814>
Optimizes patterns which are created by recent versions of vkd3d-proton,
when constant folding doesn't eliminate it entirely:
- ubitfield_extract(value, offset, umin(bits, 32-(offset&0x1f)))
- ibitfield_extract(value, offset, umin(bits, 32-(offset&0x1f)))
- bitfield_insert(base, insert, offset, umin(bits, 32-(offset&0x1f)))
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/13225>
Two 64-bit shifts and an addition are usually faster than the several
multiplications nir_lower_int64 creates.
No fossil-db changes.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14227>
v2: Fix copy-and-paste bugs in lowering patterns.
v3: Add has_sudot_4x8 flag. Requested by Rhys.
v4: Since the names of the opcodes changed from dp4 to dot_4x8, also
change the names of the lowering helpers. Suggested by Jason.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12142>
ineg(INT_MIN)/iabs(INT_MIN) won't work as expected.
No fossil-db changes.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12039>
It seems worth the small amount of damage to give an extra cushion of
not having to debug problems later.
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
All Intel platforms had similar results. (Tiger Lake shown)
total instructions in shared programs: 21043197 -> 21043359 (<.01%)
instructions in affected programs: 4409 -> 4571 (3.67%)
helped: 0
HURT: 25
HURT stats (abs) min: 1 max: 16 x̄: 6.48 x̃: 5
HURT stats (rel) min: 0.39% max: 15.38% x̄: 4.59% x̃: 4.40%
95% mean confidence interval for instructions value: 4.37 8.59
95% mean confidence interval for instructions %-change: 2.93% 6.26%
Instructions are HURT.
total cycles in shared programs: 856175986 -> 856176921 (<.01%)
cycles in affected programs: 58908 -> 59843 (1.59%)
helped: 0
HURT: 25
HURT stats (abs) min: 7 max: 70 x̄: 37.40 x̃: 38
HURT stats (rel) min: 0.27% max: 5.63% x̄: 1.87% x̃: 1.39%
95% mean confidence interval for cycles value: 31.11 43.69
95% mean confidence interval for cycles %-change: 1.35% 2.39%
Cycles are HURT.
No fossil-db changes on any Intel platform.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10012>
There are only a couple patterns that use is_finite, so the changes
aren't huge. Mostly shaders from Batman Arkham City and a few shaders
from Shadow of the Tomb Raider were affected.
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Tiger Lake
Instructions in all programs: 160902591 -> 160902489 (-0.0%)
SENDs in all programs: 6812270 -> 6812270 (+0.0%)
Loops in all programs: 38225 -> 38225 (+0.0%)
Cycles in all programs: 7429003266 -> 7428992369 (-0.0%)
Spills in all programs: 192582 -> 192582 (+0.0%)
Fills in all programs: 304539 -> 304539 (+0.0%)
Ice Lake
Instructions in all programs: 145301634 -> 145301460 (-0.0%)
SENDs in all programs: 6863890 -> 6863890 (+0.0%)
Loops in all programs: 38219 -> 38219 (+0.0%)
Cycles in all programs: 8798589772 -> 8798575869 (-0.0%)
Spills in all programs: 216880 -> 216880 (+0.0%)
Fills in all programs: 334250 -> 334250 (+0.0%)
Skylake
Instructions in all programs: 135892010 -> 135891836 (-0.0%)
SENDs in all programs: 6802916 -> 6802916 (+0.0%)
Loops in all programs: 38216 -> 38216 (+0.0%)
Cycles in all programs: 8442597324 -> 8442583202 (-0.0%)
Spills in all programs: 194839 -> 194839 (+0.0%)
Fills in all programs: 301116 -> 301116 (+0.0%)
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9108>
This commit is necessary to support "nir/range_analysis: Fix analysis of
fmin and fmax with NaN".
No shader-db or fossil-db changes on any Intel platform.
v2: Pack and unpack is_a_number.
v3: Don't set is_a_number of integer constants. The bit pattern might
be NaN.
v4: Update handling of b2i32. intBitsToFloat(int(true)) is
1.401298464324817e-45. Return a value consistent with that.
Fixes: 405de7ccb6 ("nir/range-analysis: Rudimentary value range analysis pass")
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9108>
The search helps must *never* modify the instruction passed in, so let
the compiler enforce this.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9378>
There are a bunch of cases where we can pretty quickly determine that
the high bits don't matter. In these cases, delete the casts.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8872>
This prevents some fossil-db regressions in "spir-v: Mark floating point
comparisons exact".
v2: Note that the patterns and replacements produce the same value when
isnan(b). Suggested by Caio.
v3: Use C99 isfinite() instead of (obsolete) BSD finite(). Fixes
various Windows builds.
No fossil-db changes on any Inetl platform, Vega, or Polaris10.
All Intel platforms had similar results. (Tiger Lake shown)
total instructions in shared programs: 20908670 -> 20908672 (<.01%)
instructions in affected programs: 69 -> 71 (2.90%)
helped: 0
HURT: 1
total cycles in shared programs: 473515288 -> 473513940 (<.01%)
cycles in affected programs: 4942 -> 3594 (-27.28%)
helped: 2
HURT: 0
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6358>
Otherwise, if the base type is (for example) uint32, we would
incorrectly think that PoT optimizations could not apply.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Jason Ekstsrand <jason@jleksrand.net>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Fixes: 09705747d7 ("nir/algebraic: Reassociate fadd into fmul in DPH-like pattern")
This lets us memoize range analysis work across instructions. Reduces
runtime of shader-db on Intel by -30.0288% +/- 2.1693% (n=3).
Fixes: 405de7ccb6 ("nir/range-analysis: Rudimentary value range analysis pass")
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Most integer operations are omitted because dealing with integer
overflow is hard. There are a few things that could be smarter if there
was a small amount more tracking of ranges of integer types (i.e.,
operands are Boolean, operand values fit in 16 bits, etc.).
The changes to nir_search_helpers.h are included in this patch to
simplify reordering the changes to nir_opt_algebraic.py.
v2: Memoize range analysis results. Without this, some shaders appear
to get stuck in infinite loops.
v3: Rebase on many months of Mesa changes, including 1-bit Boolean
changes.
v4: Rebase on "nir: Drop imov/fmov in favor of one mov instruction".
v5: Use nir_alu_srcs_equal for detecting (a*a). Previously just the SSA
value was compared, and this incorrectly matched (a.x*a.y).
v6: Many code improvements including (but not limited to) better names,
more comments, and better use of helper functions. All suggested by
Caio. Rework the handling of several opcodes to use a table for mapping
source ranges to a result range. This change fixed a bug that caused
fmax(gt_zero, ge_zero) to be incorrectly recognized as ge_zero.
Slightly tighten the range of fmul by recognizing that x*x is gt_zero if
x is gt_zero. Add similar handling for -x*x.
v7: Use _______ in the tables as an alias for unknown. Suggested by
Caio.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
The helpers are needed so we can use the syntax `instr(cond)` in the
algebraic rules. Add simple rule for dropping a pair of mul-div of
the same value when wrapping is guaranteed to not happen.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
For umul_low (al * bl), zero is returned if the low 16-bits word of either
source is zero.
for imadsh_mix16 (ah * bl << 16 + c), c is returned if either 'ah' or 'bl'
is zero.
A couple of nir_search_helpers are added:
is_upper_half_zero() returns true if the highest word of all components of
an integer NIR alu src are zero.
is_lower_half_zero() returns true if the lowest word of all components of
an integer nir alu src are zero.
Reviewed-by: Eric Anholt <eric@anholt.net>
The goal is to avoid having an extra MOV instruction to perform the
saturate. Doing the subtraction first allows the saturate to be applied
to the ADD instruction making the MOV unnecessary. Values generated in
different block and values from non-ALU instructions (e.g., texture
instructions) almost always need the extra MOV.
Multiply instructions are restricted because doing this rearrangement
can interfere with the generation of flrp and ffma instructions.
v2: Now that the final method has been selected, squash three commits
into one.
All Intel platforms has similar results. (Ice Lake shown)
total instructions in shared programs: 17223214 -> 17219386 (-0.02%)
instructions in affected programs: 1524376 -> 1520548 (-0.25%)
helped: 2686
HURT: 26
helped stats (abs) min: 1 max: 32 x̄: 1.44 x̃: 1
helped stats (rel) min: 0.03% max: 16.67% x̄: 0.54% x̃: 0.37%
HURT stats (abs) min: 1 max: 2 x̄: 1.69 x̃: 2
HURT stats (rel) min: 0.33% max: 1.67% x̄: 0.54% x̃: 0.35%
95% mean confidence interval for instructions value: -1.46 -1.36
95% mean confidence interval for instructions %-change: -0.56% -0.50%
Instructions are helped.
total cycles in shared programs: 360811571 -> 360791896 (<.01%)
cycles in affected programs: 103650214 -> 103630539 (-0.02%)
helped: 1557
HURT: 675
helped stats (abs) min: 1 max: 1773 x̄: 41.44 x̃: 16
helped stats (rel) min: <.01% max: 26.77% x̄: 1.37% x̃: 0.64%
HURT stats (abs) min: 1 max: 1513 x̄: 66.44 x̃: 14
HURT stats (rel) min: <.01% max: 46.16% x̄: 2.00% x̃: 0.49%
95% mean confidence interval for cycles value: -14.82 -2.81
95% mean confidence interval for cycles %-change: -0.50% -0.20%
Cycles are helped.
LOST: 2
GAINED: 0
Reviewed-by: Matt Turner <mattst88@gmail.com> [v1]
Reviewed-by: Thomas Helland <thomashelland90@gmail.com>
This pass attempts to dectect code sequences like
if (x < y) {
z = y - x;
...
}
and replace them with sequences like
t = x - y;
if (t < 0) {
z = -t;
...
}
On architectures where the subtract can generate the flags used by the
if-statement, this saves an instruction. It's also possible that moving
an instruction out of the if-statement will allow
nir_opt_peephole_select to convert the whole thing to a bcsel.
Currently only floating point compares and adds are supported. Adding
support for integer will be a challenge due to integer overflow. There
are a couple possible solutions, but they may not apply to all
architectures.
v2: Fix a typo in the commit message and a couple typos in comments.
Fix possible NULL pointer deref from result of push_block(). Add
missing (-A + B) case. Suggested by Caio.
v3: Fix is_not_const_zero to work correctly with types other than
nir_type_float32. Suggested by Ken.
v4: Add some comments explaining how this works. Suggested by Ken.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
This not only makes them safe for more bit sizes but it also fixes a bug
in is_zero_to_one where it would return true for constant NaN.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
V2: mark float opts as inexact
If one of the inputs to an mul/add is the result of another
mul/add there is a chance that we can reuse the result of that
mul/add in other calls if we do the multiplication in the right
order.
Also by attempting to move all constants to the top we increase
the chance of constant folding.
For example it is a fairly common pattern for shaders to do something
similar to this:
const float a = 0.5;
in vec4 b;
in float c;
...
b.x = b.x * c;
b.y = b.y * c;
...
b.x = b.x * a + a;
b.y = b.y * a + a;
So by simply detecting that constant a is part of the multiplication
in ffma and switching it with previous fmul that updates b we end up
with:
...
c = a * c;
...
b.x = b.x * c + a;
b.y = b.y * c + a;
Shader-db results BDW:
total instructions in shared programs: 13011050 -> 12967888 (-0.33%)
instructions in affected programs: 4118366 -> 4075204 (-1.05%)
helped: 17739
HURT: 1343
total cycles in shared programs: 246717952 -> 246410716 (-0.12%)
cycles in affected programs: 166870802 -> 166563566 (-0.18%)
helped: 18493
HURT: 7965
total spills in shared programs: 14937 -> 14560 (-2.52%)
spills in affected programs: 9331 -> 8954 (-4.04%)
helped: 284
HURT: 33
total fills in shared programs: 20211 -> 19671 (-2.67%)
fills in affected programs: 12586 -> 12046 (-4.29%)
helped: 286
HURT: 33
LOST: 39
GAINED: 33
Some of the hurt will go away when we shuffle things back down to the
bottom in the following patch. It's also noteworthy that almost all of the
spill changes are in Deus Ex both hurt and helped.
Reviewed-by: Elie Tournier <elie.tournier@collabora.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Didn't turn out as useful as I'd hoped, but it will help alot more on
i965 by reducing regressions when we drop brw_do_channel_expressions()
and brw_do_vector_splitting().
I'm not sure how much sense 'is_not_used_by_conditional' makes on
platforms other than i965 but since this is a new opt it at least
won't do any harm.
shader-db BDW:
total instructions in shared programs: 13029581 -> 13029415 (-0.00%)
instructions in affected programs: 15268 -> 15102 (-1.09%)
helped: 86
HURT: 0
total cycles in shared programs: 247038346 -> 247036198 (-0.00%)
cycles in affected programs: 692634 -> 690486 (-0.31%)
helped: 183
HURT: 27
Reviewed-by: Elie Tournier <elie.tournier@collabora.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
