fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-08 13:28:06 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

nir/algebraic: add vec8/16 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 13:55:17 +02:00 • committed by Marge Bot
parent 5efbef833a
commit e58c004870
1 changed files with 6 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

6
src/compiler/nir/nir_opt_algebraic.py
View file

@ -1167,9 +1167,13 @@ optimizations.extend([
(('fall_equal2', a, b), ('fmin', ('seq', 'a.x', 'b.x'), ('seq', 'a.y', 'b.y')), 'options->lower_vector_cmp'),
(('fall_equal3', a, b), ('seq', ('fany_nequal3', a, b), 0.0), 'options->lower_vector_cmp'),
(('fall_equal4', a, b), ('seq', ('fany_nequal4', a, b), 0.0), 'options->lower_vector_cmp'),
(('fall_equal8', a, b), ('seq', ('fany_nequal8', a, b), 0.0), 'options->lower_vector_cmp'),
(('fall_equal16', a, b), ('seq', ('fany_nequal16', a, b), 0.0), 'options->lower_vector_cmp'),
(('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'),
(('fany_nequal8', a, b), ('fsat', ('fdot8', ('sne', a, b), ('sne', a, b))), 'options->lower_vector_cmp'),
(('fany_nequal16', a, b), ('fsat', ('fdot16', ('sne', a, b), ('sne', a, b))), 'options->lower_vector_cmp'),
])
def vector_cmp(reduce_op, cmp_op, comps):
@ -1190,6 +1194,8 @@ for op in [
((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'),
((op[0] + '8', a, b), vector_cmp(op[2], op[1], 'abcdefgh'), 'options->lower_vector_cmp'),
((op[0] + '16', a, b), vector_cmp(op[2], op[1], 'abcdefghijklmnop'), 'options->lower_vector_cmp'),
])
optimizations.extend([