nir: shuffle constants to the top

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>

2 files changed, 19 insertions, 1 deletions

diff --git a/src/compiler/nir/nir_opt_algebraic.py b/src/compiler/nir/nir_opt_algebraic.py
index bd65eff8b59..974345a17e5 100644
--- a/src/compiler/nir/nir_opt_algebraic.py
+++ b/src/compiler/nir/nir_opt_algebraic.py
@@ -343,8 +343,14 @@ optimizations = [
    (('fmul', ('fneg', a), b), ('fneg', ('fmul', a, b))),
    (('imul', ('ineg', a), b), ('ineg', ('imul', a, b))),
 
+   # Propagate constants up multiplication chains
+   (('~fmul(is_used_once)', ('fmul(is_used_once)', 'a(is_not_const)', 'b(is_not_const)'), '#c'), ('fmul', ('fmul', a, c), b)),
+   (('imul(is_used_once)', ('imul(is_used_once)', 'a(is_not_const)', 'b(is_not_const)'), '#c'), ('imul', ('imul', a, c), b)),
+   (('~fadd(is_used_once)', ('fadd(is_used_once)', 'a(is_not_const)', 'b(is_not_const)'), '#c'), ('fadd', ('fadd', a, c), b)),
+   (('iadd(is_used_once)', ('iadd(is_used_once)', 'a(is_not_const)', 'b(is_not_const)'), '#c'), ('iadd', ('iadd', a, c), b)),
+
    # Reassociate constants in add/mul chains so they can be folded together.
-   # For now, we only handle cases where the constants are separated by
+   # For now, we mostly only handle cases where the constants are separated by
    # a single non-constant.  We could do better eventually.
    (('~fmul', '#a', ('fmul', b, '#c')), ('fmul', ('fmul', a, c), b)),
    (('imul', '#a', ('imul', b, '#c')), ('imul', ('imul', a, c), b)),
diff --git a/src/compiler/nir/nir_search_helpers.h b/src/compiler/nir/nir_search_helpers.h
index 7decf6582a7..200f2471f84 100644
--- a/src/compiler/nir/nir_search_helpers.h
+++ b/src/compiler/nir/nir_search_helpers.h
@@ -115,6 +115,18 @@ is_zero_to_one(nir_alu_instr *instr, unsigned src, unsigned num_components,
 }
 
 static inline bool
+is_not_const(nir_alu_instr *instr, unsigned src, unsigned num_components,
+             const uint8_t *swizzle)
+{
+   nir_const_value *val = nir_src_as_const_value(instr->src[src].src);
+
+   if (val)
+      return false;
+
+   return true;
+}
+
+static inline bool
 is_used_more_than_once(nir_alu_instr *instr)
 {
    bool zero_if_use = list_empty(&instr->dest.dest.ssa.if_uses);