nir: Add simplistic lowering for bany_equal/ball_inequal.

It would be nice if we could do swizzling of an expression on the
replacement side so that we could have a single ieq/ine of the vector
after CSE.  However, if you do want vector operations, nir_opt_vectorize()
does just fine.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6567>

diff --git a/src/compiler/nir/nir.h b/src/compiler/nir/nir.h
index 9f95dca20e843023241134f3dc808364c221b35b..cbf5c0746c351fe8e8ad6b2b1765db51bcd45344 100644 (file)
--- a/src/compiler/nir/nir.h
+++ b/src/compiler/nir/nir.h
@@ -3082,7 +3082,7 @@ typedef struct nir_shader_compiler_options {
    /* lower {slt,sge,seq,sne} to {flt,fge,feq,fneu} + b2f: */
    bool lower_scmp;
 
-   /* lower fall_equalN/fany_nequalN (ex:fany_nequal4 to sne+fdot4+fsat) */
+   /* lower b/fall_equalN/b/fany_nequalN (ex:fany_nequal4 to sne+fdot4+fsat) */
    bool lower_vector_cmp;
 
    /** enables rules to lower idiv by power-of-two: */

diff --git a/src/compiler/nir/nir_opt_algebraic.py b/src/compiler/nir/nir_opt_algebraic.py
index d775a23fceec574d8c56ac77095aca36894823e5..0529e31cad05f536252e8db9e0a0013ecddddedd 100644 (file)
--- a/src/compiler/nir/nir_opt_algebraic.py
+++ b/src/compiler/nir/nir_opt_algebraic.py
@@ -753,6 +753,23 @@ optimizations.extend([
    (('fany_nequal2', a, b), ('fmax', ('sne', 'a.x', 'b.x'), ('sne', 'a.y', 'b.y')), 'options->lower_vector_cmp'),
    (('fany_nequal3', a, b), ('fsat', ('fdot3', ('sne', a, b), ('sne', a, b))), 'options->lower_vector_cmp'),
    (('fany_nequal4', a, b), ('fsat', ('fdot4', ('sne', a, b), ('sne', a, b))), 'options->lower_vector_cmp'),
+
+   (('ball_iequal2', a, b), ('iand', ('ieq', 'a.x', 'b.x'), ('ieq', 'a.y', 'b.y')), 'options->lower_vector_cmp'),
+   (('ball_iequal3', a, b), ('iand', ('iand', ('ieq', 'a.x', 'b.x'), ('ieq', 'a.y', 'b.y')), ('ieq', 'a.z', 'b.z')), 'options->lower_vector_cmp'),
+   (('ball_iequal4', a, b), ('iand', ('iand', ('ieq', 'a.x', 'b.x'), ('ieq', 'a.y', 'b.y')), ('iand', ('ieq', 'a.z', 'b.z'), ('ieq', 'a.w', 'b.w'))), 'options->lower_vector_cmp'),
+
+   (('bany_inequal2', a, b), ('ior', ('ine', 'a.x', 'b.x'), ('ine', 'a.y', 'b.y')), 'options->lower_vector_cmp'),
+   (('bany_inequal3', a, b), ('ior', ('ior', ('ine', 'a.x', 'b.x'), ('ine', 'a.y', 'b.y')), ('ine', 'a.z', 'b.z')), 'options->lower_vector_cmp'),
+   (('bany_inequal4', a, b), ('ior', ('ior', ('ine', 'a.x', 'b.x'), ('ine', 'a.y', 'b.y')), ('ior', ('ine', 'a.z', 'b.z'), ('ine', 'a.w', 'b.w'))), 'options->lower_vector_cmp'),
+
+   (('ball_fequal2', a, b), ('iand', ('feq', 'a.x', 'b.x'), ('feq', 'a.y', 'b.y')), 'options->lower_vector_cmp'),
+   (('ball_fequal3', a, b), ('iand', ('iand', ('feq', 'a.x', 'b.x'), ('feq', 'a.y', 'b.y')), ('feq', 'a.z', 'b.z')), 'options->lower_vector_cmp'),
+   (('ball_fequal4', a, b), ('iand', ('iand', ('feq', 'a.x', 'b.x'), ('feq', 'a.y', 'b.y')), ('iand', ('feq', 'a.z', 'b.z'), ('feq', 'a.w', 'b.w'))), 'options->lower_vector_cmp'),
+
+   (('bany_fnequal2', a, b), ('ior', ('fneu', 'a.x', 'b.x'), ('fneu', 'a.y', 'b.y')), 'options->lower_vector_cmp'),
+   (('bany_fnequal3', a, b), ('ior', ('ior', ('fneu', 'a.x', 'b.x'), ('fneu', 'a.y', 'b.y')), ('fneu', 'a.z', 'b.z')), 'options->lower_vector_cmp'),
+   (('bany_fnequal4', a, b), ('ior', ('ior', ('fneu', 'a.x', 'b.x'), ('fneu', 'a.y', 'b.y')), ('ior', ('fneu', 'a.z', 'b.z'), ('fneu', 'a.w', 'b.w'))), 'options->lower_vector_cmp'),
+
    (('fneu', ('fneg', a), a), ('fneu', a, 0.0)),
    (('feq', ('fneg', a), a), ('feq', a, 0.0)),
    # Emulating booleans