nir/algebraic: mark float optimizations returning one parameter as inexact

With the arrival of VK_KHR_shader_float_controls algebraic
optimizations for float types of the form (('fop', a, b), a) become
inexact depending on the execution mode.

For example, if we have activated SHADER_DENORM_FLUSH_TO_ZERO, in case
of a denorm value for the "a" parameter, we cannot return it still as
a denorm, it needs to be flushed to zero. Therefore, we mark now all
those operations as inexact.

Suggested-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Signed-off-by: Andres Gomez <agomez@igalia.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>

diff --git a/src/compiler/nir/nir_opt_algebraic.py b/src/compiler/nir/nir_opt_algebraic.py
index a94829d9bfbaea1ce7e8cad64349b0da7f4e5bb0..e3b503f38259fa56856675556385a0cede088818 100644 (file)
--- a/src/compiler/nir/nir_opt_algebraic.py
+++ b/src/compiler/nir/nir_opt_algebraic.py
@@ -100,7 +100,7 @@ optimizations = [
    (('idiv', a, '#b@32(is_neg_power_of_two)'), ('ineg', ('imul', ('isign', a), ('ushr', ('iabs', a), ('find_lsb', ('iabs', b))))), 'options->lower_idiv'),
    (('umod', a, '#b(is_pos_power_of_two)'),    ('iand', a, ('isub', b, 1))),
 
    (('idiv', a, '#b@32(is_neg_power_of_two)'), ('ineg', ('imul', ('isign', a), ('ushr', ('iabs', a), ('find_lsb', ('iabs', b))))), 'options->lower_idiv'),
    (('umod', a, '#b(is_pos_power_of_two)'),    ('iand', a, ('isub', b, 1))),
 

-   (('fneg', ('fneg', a)), a),
+   (('~fneg', ('fneg', a)), a),

    (('ineg', ('ineg', a)), a),
    (('fabs', ('fabs', a)), ('fabs', a)),
    (('fabs', ('fneg', a)), ('fabs', a)),
    (('ineg', ('ineg', a)), a),
    (('fabs', ('fabs', a)), ('fabs', a)),
    (('fabs', ('fneg', a)), ('fabs', a)),


@@ -126,7 +126,7 @@ optimizations = [
    (('imul', a, 0), 0),
    (('umul_unorm_4x8', a, 0), 0),
    (('umul_unorm_4x8', a, ~0), a),
    (('imul', a, 0), 0),
    (('umul_unorm_4x8', a, 0), 0),
    (('umul_unorm_4x8', a, ~0), a),

-   (('fmul', a, 1.0), a),
+   (('~fmul', a, 1.0), a),

    (('imul', a, 1), a),
    (('fmul', a, -1.0), ('fneg', a)),
    (('imul', a, -1), ('ineg', a)),
    (('imul', a, 1), a),
    (('fmul', a, -1.0), ('fneg', a)),
    (('imul', a, -1), ('ineg', a)),


@@ -383,7 +383,7 @@ optimizations.extend([
    (('feq', ('fmin', ('fneg', ('b2f', 'a@1')), 'b@1'), 0.0), ('iand', ('inot', a), ('fge', b, 0.0))),
 
    (('feq', ('b2f', 'a@1'), 0.0), ('inot', a)),
    (('feq', ('fmin', ('fneg', ('b2f', 'a@1')), 'b@1'), 0.0), ('iand', ('inot', a), ('fge', b, 0.0))),
 
    (('feq', ('b2f', 'a@1'), 0.0), ('inot', a)),

-   (('fne', ('b2f', 'a@1'), 0.0), a),
+   (('~fne', ('b2f', 'a@1'), 0.0), a),

    (('ieq', ('b2i', 'a@1'), 0),   ('inot', a)),
    (('ine', ('b2i', 'a@1'), 0),   a),
 
    (('ieq', ('b2i', 'a@1'), 0),   ('inot', a)),
    (('ine', ('b2i', 'a@1'), 0),   a),
 


@@ -462,8 +462,8 @@ optimizations.extend([
    (('bcsel', a, a, b), ('ior', a, b)),
    (('bcsel', a, b, False), ('iand', a, b)),
    (('bcsel', a, b, a), ('iand', a, b)),
    (('bcsel', a, a, b), ('ior', a, b)),
    (('bcsel', a, b, False), ('iand', a, b)),
    (('bcsel', a, b, a), ('iand', a, b)),

-   (('fmin', a, a), a),
-   (('fmax', a, a), a),
+   (('~fmin', a, a), a),
+   (('~fmax', a, a), a),

    (('imin', a, a), a),
    (('imax', a, a), a),
    (('umin', a, a), a),
    (('imin', a, a), a),
    (('imax', a, a), a),
    (('umin', a, a), a),


@@ -480,9 +480,9 @@ optimizations.extend([
    (('imin', a, ('ineg', a)), ('ineg', ('iabs', a))),
    (('fmin', a, ('fneg', ('fabs', a))), ('fneg', ('fabs', a))),
    (('imin', a, ('ineg', ('iabs', a))), ('ineg', ('iabs', a))),
    (('imin', a, ('ineg', a)), ('ineg', ('iabs', a))),
    (('fmin', a, ('fneg', ('fabs', a))), ('fneg', ('fabs', a))),
    (('imin', a, ('ineg', ('iabs', a))), ('ineg', ('iabs', a))),

-   (('fmin', a, ('fabs', a)), a),
+   (('~fmin', a, ('fabs', a)), a),

    (('imin', a, ('iabs', a)), a),
    (('imin', a, ('iabs', a)), a),

-   (('fmax', a, ('fneg', ('fabs', a))), a),
+   (('~fmax', a, ('fneg', ('fabs', a))), a),

    (('imax', a, ('ineg', ('iabs', a))), a),
    (('fmax', a, ('fabs', a)), ('fabs', a)),
    (('imax', a, ('iabs', a)), ('iabs', a)),
    (('imax', a, ('ineg', ('iabs', a))), a),
    (('fmax', a, ('fabs', a)), ('fabs', a)),
    (('imax', a, ('iabs', a)), ('iabs', a)),


@@ -768,7 +768,7 @@ optimizations.extend([
    (('bcsel', a, ('b2f(is_used_once)', 'b@32'), ('b2f', 'c@32')), ('b2f', ('bcsel', a, b, c))),
 
    (('bcsel', a, b, b), b),
    (('bcsel', a, ('b2f(is_used_once)', 'b@32'), ('b2f', 'c@32')), ('b2f', ('bcsel', a, b, c))),
 
    (('bcsel', a, b, b), b),

-   (('fcsel', a, b, b), b),
+   (('~fcsel', a, b, b), b),

 
    # D3D Boolean emulation
    (('bcsel', a, -1, 0), ('ineg', ('b2i', 'a@1'))),
 
    # D3D Boolean emulation
    (('bcsel', a, -1, 0), ('ineg', ('b2i', 'a@1'))),