OpenHarmony/third_party_mesa3d
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

nir/algebraic: optimize out exact a+0.0 if it's used only as a float

Browse Source 
fossil-db (GFX10):
Totals from 133 (0.10% of 139391) affected shaders:
SGPRs: 7864 -> 7856 (-0.10%); split: -0.20%, +0.10%
VGPRs: 4884 -> 4836 (-0.98%)
CodeSize: 288932 -> 287084 (-0.64%)
MaxWaves: 1973 -> 1979 (+0.30%)
Instrs: 53899 -> 53550 (-0.65%)

fossil-db (GFX10.3):
Totals from 133 (0.10% of 139391) affected shaders:
SGPRs: 7832 -> 7835 (+0.04%); split: -0.06%, +0.10%
VGPRs: 5144 -> 5088 (-1.09%)
CodeSize: 318912 -> 316696 (-0.69%)
MaxWaves: 1735 -> 1746 (+0.63%)
Instrs: 65367 -> 64853 (-0.79%)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5523>
This commit is contained in:
Rhys Perry
2020-07-15 11:16:33 +01:00
parent 2f0d480c73
commit 614ab26afd
1 changed files with 10 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
src/compiler/nir/nir_opt_algebraic.py
View File

@@ -40,6 +40,9 @@ c = 'c'
d = 'd'
e = 'e'
signed_zero_inf_nan_preserve_16 = 'nir_is_float_control_signed_zero_inf_nan_preserve(info->float_controls_execution_mode, 16)'
signed_zero_inf_nan_preserve_32 = 'nir_is_float_control_signed_zero_inf_nan_preserve(info->float_controls_execution_mode, 32)'
# Written in the form (<search>, <replace>) where <search> is an expression
# and <replace> is either an expression or a value. An expression is
# defined as a tuple of the form ([~]<op>, <src0>, <src1>, <src2>, <src3>)
@@ -124,6 +127,11 @@ optimizations = [
(('f2b', ('fneg', a)), ('f2b', a)),
(('i2b', ('ineg', a)), ('i2b', a)),
(('~fadd', a, 0.0), a),
# a+0.0 is 'a' unless 'a' is denormal or -0.0. If it's only used by a
# floating point instruction, they should flush any input denormals and we
# can replace -0.0 with 0.0 if the float execution mode allows it.
(('fadd(is_only_used_as_float)', 'a@16', 0.0), a, '!'+signed_zero_inf_nan_preserve_16),
(('fadd(is_only_used_as_float)', 'a@32', 0.0), a, '!'+signed_zero_inf_nan_preserve_32),
(('iadd', a, 0), a),
(('usadd_4x8', a, 0), a),
(('usadd_4x8', a, ~0), ~0),
@@ -158,6 +166,8 @@ optimizations = [
(('fmul', ('fmul', ('fsign', a), a), a), ('fmul', ('fabs', a), a)),
(('~ffma', 0.0, a, b), b),
(('~ffma', a, b, 0.0), ('fmul', a, b)),
(('ffma@16', a, b, 0.0), ('fmul', a, b), '!'+signed_zero_inf_nan_preserve_16),
(('ffma@32', a, b, 0.0), ('fmul', a, b), '!'+signed_zero_inf_nan_preserve_32),
(('ffma', 1.0, a, b), ('fadd', a, b)),
(('ffma', -1.0, a, b), ('fadd', ('fneg', a), b)),
(('~flrp', a, b, 0.0), a),