nir/algebraic: improve irem by power-of-two optimization

Requires one less instruction.

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>

diff --git a/src/compiler/nir/nir_opt_algebraic.py b/src/compiler/nir/nir_opt_algebraic.py
index 225bb3e..d4f2cdc 100644 (file)
--- a/src/compiler/nir/nir_opt_algebraic.py
+++ b/src/compiler/nir/nir_opt_algebraic.py
@@ -116,11 +116,14 @@ optimizations = [
    (('udiv', a, '#b(is_pos_power_of_two)'), ('ushr', a, ('find_lsb', b)), '!options->lower_bitops'),
    (('idiv', a, '#b(is_pos_power_of_two)'), ('imul', ('isign', a), ('ushr', ('iabs', a), ('find_lsb', b))), '!options->lower_bitops'),
    (('idiv', a, '#b(is_neg_power_of_two)'), ('ineg', ('imul', ('isign', a), ('ushr', ('iabs', a), ('find_lsb', ('iabs', b))))), '!options->lower_bitops'),
-   (('umod', a, '#b(is_pos_power_of_two)'),    ('iand', a, ('isub', b, 1)), '!options->lower_bitops'),
-   (('imod', a, '#b(is_pos_power_of_two)'),    ('iand', a, ('isub', b, 1)), '!options->lower_bitops'),
-   (('imod', a, '#b(is_neg_power_of_two)'),    ('bcsel', ('ieq', ('ior', a, b), b), 0, ('ior', a, b)), '!options->lower_bitops'),
-   (('irem', a, '#b(is_pos_power_of_two)'),    ('bcsel', ('ige', a, 0), ('iand', a, ('isub', b, 1)), ('imod', a, ('ineg', b))), '!options->lower_bitops'),
-   (('irem', a, '#b(is_neg_power_of_two)'),    ('bcsel', ('ige', a, 0), ('iand', a, ('inot', b)), ('imod', a, b)), '!options->lower_bitops'),
+   (('umod', a, '#b(is_pos_power_of_two)'), ('iand', a, ('isub', b, 1)), '!options->lower_bitops'),
+   (('imod', a, '#b(is_pos_power_of_two)'), ('iand', a, ('isub', b, 1)), '!options->lower_bitops'),
+   (('imod', a, '#b(is_neg_power_of_two)'), ('bcsel', ('ieq', ('ior', a, b), b), 0, ('ior', a, b)), '!options->lower_bitops'),
+   # 'irem(a, b)' -> 'a - ((a < 0 ? (a + b - 1) : a) & -b)'
+   (('irem', a, '#b(is_pos_power_of_two)'),
+    ('isub', a, ('iand', ('bcsel', ('ilt', a, 0), ('iadd', a, ('isub', b, 1)), a), ('ineg', b))),
+    '!options->lower_bitops'),
+   (('irem', a, '#b(is_neg_power_of_two)'), ('irem', a, ('iabs', b)), '!options->lower_bitops'),
 
    (('~fneg', ('fneg', a)), a),
    (('ineg', ('ineg', a)), a),