AlexIndustrial/mesa
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

nir/algebraic: generalize vector_cmp lowering

Browse Source 
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19150>
This commit is contained in:
Karol Herbst
2022-10-28 13:38:56 +02:00
committed by Marge Bot
parent f27e2234e1
commit 5efbef833a
1 changed files with 20 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/compiler/nir/nir_opt_algebraic.py
+20 -14
View File
@@ -1170,23 +1170,29 @@ 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'),
def vector_cmp(reduce_op, cmp_op, comps):
if len(comps) == 1:
return (cmp_op, 'a.' + comps[0], 'b.' + comps[0])
else:
mid = len(comps) // 2
return (reduce_op, vector_cmp(reduce_op, cmp_op, comps[:mid]),
vector_cmp(reduce_op, cmp_op, comps[mid:]))
(('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'),
for op in [
('ball_iequal', 'ieq', 'iand'),
('ball_fequal', 'feq', 'iand'),
('bany_inequal', 'ine', 'ior'),
('bany_fnequal', 'fneu', 'ior'),
]:
optimizations.extend([
((op[0] + '2', a, b), vector_cmp(op[2], op[1], 'xy'), 'options->lower_vector_cmp'),
((op[0] + '3', a, b), vector_cmp(op[2], op[1], 'xyz'), 'options->lower_vector_cmp'),
((op[0] + '4', a, b), vector_cmp(op[2], op[1], 'xyzw'), 'options->lower_vector_cmp'),
])
optimizations.extend([
(('feq', ('seq', a, b), 1.0), ('feq', a, b)),
(('feq', ('sne', a, b), 1.0), ('fneu', a, b)),
(('feq', ('slt', a, b), 1.0), ('flt', a, b)),