fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-07 00:38:48 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 4

nir: add fmul_rtz optimizations

Browse source 
NVK is only going to use it for `fmul_rtz(frcp(ipa), ipa)` patterns, so
try not too hard to optimize this.

Totals from 10 (0.00% of 1212873) affected shaders:
CodeSize: 34480 -> 34288 (-0.56%); split: -0.60%, +0.05%
Static cycle count: 6225 -> 6132 (-1.49%); split: -1.57%, +0.08%

Reviewed-by: Mel Henning <mhenning@darkrefraction.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/41179>
This commit is contained in:
Karol Herbst 2026-04-25 15:47:52 +02:00 • committed by Marge Bot
parent f5e92e5493
commit 4e67582ddf
1 changed files with 8 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

8
src/compiler/nir/nir_opt_algebraic.py
View file

@ -220,6 +220,8 @@ for op in ['idiv', 'udiv', 'umod', 'umod', 'irem']:
optimizations += [
(('fmul', ('fsign', a), ('ffloor', ('fadd', ('fabs', a), 0.5))), ('ftrunc', ('fadd', a, ('fmul', ('fsign', a), 0.5))), '!options->lower_ftrunc || options->lower_ffloor'),
(('fmul_rtz@32', a, b), ('fmul', a, b), 'nir_is_rounding_mode_rtz(info->float_controls_execution_mode, 32)'),
(('fneg', ('fneg', a)), ('fcanonicalize', a)),
(('ineg', ('ineg', a)), a),
(('fabs', ('fneg', a)), ('fabs', a)),
@ -790,6 +792,8 @@ optimizations.extend([
(('bcsel(is_only_used_as_float_nsz)', ('fneu', a, 0.0), a, -0.0), a),
(('bcsel', ('feq', a, 0), 0, ('fsat', ('fmul', a, 'b(is_a_number)'))), ('fsat(preserve_sz)', ('fmul', a, b))),
(('bcsel', ('fneu', a, 0), ('fsat', ('fmul', a, 'b(is_a_number)')), 0), ('fsat(preserve_sz)', ('fmul', a, b))),
(('bcsel', ('feq', a, 0), 0, ('fsat', ('fmul_rtz', a, 'b(is_a_number)'))), ('fsat(preserve_sz)', ('fmul_rtz', a, b))),
(('bcsel', ('fneu', a, 0), ('fsat', ('fmul_rtz', a, 'b(is_a_number)')), 0), ('fsat(preserve_sz)', ('fmul_rtz', a, b))),
(('bcsel', ('feq', a, 0), b, ('fadd', a, 'b(is_not_zero)')), ('fadd', a, b)),
(('bcsel', ('fneu', a, 0), ('fadd', a, 'b(is_not_zero)'), b), ('fadd', a, b)),
@ -1041,8 +1045,10 @@ optimizations.extend([
(('fsat', ('fsat', a)), ('fsat', a)),
(('fsat', ('fneg(is_used_once)', ('fadd(is_used_once)', a, b))), ('fsat', ('fadd', ('fneg', a), ('fneg', b))), '!options->lower_fsat'),
(('fsat', ('fneg(is_used_once)', ('fmul(is_used_once)', a, b))), ('fsat', ('fmul', ('fneg', a), b)), '!options->lower_fsat'),
(('fsat', ('fneg(is_used_once)', ('fmul_rtz(is_used_once)', a, b))), ('fsat', ('fmul_rtz', ('fneg', a), b)), '!options->lower_fsat'),
(('fsat(nsz)', ('fneg(is_used_once)', ('fmulz(is_used_once)', a, b))), ('fsat', ('fmulz', ('fneg', a), b)), '!options->lower_fsat'),
(('fsat', ('fabs(is_used_once)', ('fmul(is_used_once)', a, b))), ('fsat', ('fmul', ('fabs', a), ('fabs', b))), '!options->lower_fsat'),
(('fsat', ('fabs(is_used_once)', ('fmul_rtz(is_used_once)', a, b))), ('fsat', ('fmul_rtz', ('fabs', a), ('fabs', b))), '!options->lower_fsat'),
(('fmin', ('fmax', ('fmin', ('fmax', a, b), c), b), c), ('fmin', ('fmax', a, b), c)),
(('imin', ('imax', ('imin', ('imax', a, b), c), b), c), ('imin', ('imax', a, b), c)),
(('umin', ('umax', ('umin', ('umax', a, b), c), b), c), ('umin', ('umax', a, b), c)),
@ -4103,7 +4109,9 @@ distribute_src_mods = [
(('fneg(is_only_used_as_float)', ('fneg', a)), a),
(('fneg', ('fmul(is_used_once)', a, b)), ('fmul', ('fneg', a), b)),
(('fneg', ('fmul_rtz(is_used_once)', a, b)), ('fmul_rtz', ('fneg', a), b)),
(('fabs', ('fmul(is_used_once)', a, b)), ('fmul', ('fabs', a), ('fabs', b))),
(('fabs', ('fmul_rtz(is_used_once)', a, b)), ('fmul_rtz', ('fabs', a), ('fabs', b))),
(('fneg', ('ffma(is_used_once,nsz)', a, b, c)), ('ffma', ('fneg', a), b, ('fneg', c))),
(('fneg', ('flrp(is_used_once)', a, b, c)), ('flrp', ('fneg', a), ('fneg', b), c), 'true', TestStatus.XFAIL), # XFAIL is -flrp(0, -1, 0) is 0.0 instead of -0.0