# Authors:
# Jason Ekstrand (jason@jlekstrand.net)
+from __future__ import print_function
+
+from collections import OrderedDict
import nir_algebraic
+import itertools
# Convenience variables
a = 'a'
(('ineg', ('ineg', a)), a),
(('fabs', ('fabs', a)), ('fabs', a)),
(('fabs', ('fneg', a)), ('fabs', a)),
- (('fabs', ('u2f32', a)), ('u2f32', a)),
+ (('fabs', ('u2f', a)), ('u2f', a)),
(('iabs', ('iabs', a)), ('iabs', a)),
(('iabs', ('ineg', a)), ('iabs', a)),
+ (('f2b', ('fneg', a)), ('f2b', a)),
+ (('i2b', ('ineg', a)), ('i2b', a)),
(('~fadd', a, 0.0), a),
(('iadd', a, 0), a),
(('usadd_4x8', a, 0), a),
(('imul', a, 1), a),
(('fmul', a, -1.0), ('fneg', a)),
(('imul', a, -1), ('ineg', a)),
+ # If a < 0: fsign(a)*a*a => -1*a*a => -a*a => abs(a)*a
+ # If a > 0: fsign(a)*a*a => 1*a*a => a*a => abs(a)*a
+ # If a == 0: fsign(a)*a*a => 0*0*0 => abs(0)*0
+ (('fmul', ('fsign', a), ('fmul', a, a)), ('fmul', ('fabs', a), a)),
+ (('fmul', ('fmul', ('fsign', a), a), a), ('fmul', ('fabs', a), a)),
(('~ffma', 0.0, a, b), b),
(('~ffma', a, 0.0, b), b),
(('~ffma', a, b, 0.0), ('fmul', a, b)),
(('~flrp', a, b, 1.0), b),
(('~flrp', a, a, b), a),
(('~flrp', 0.0, a, b), ('fmul', a, b)),
- (('~flrp', a, b, ('b2f', c)), ('bcsel', c, b, a), 'options->lower_flrp32'),
+ (('~flrp', a, b, ('b2f', 'c@32')), ('bcsel', c, b, a), 'options->lower_flrp32'),
(('~flrp', a, 0.0, c), ('fadd', ('fmul', ('fneg', a), c), a)),
(('flrp@32', a, b, c), ('fadd', ('fmul', c, ('fsub', b, a)), a), 'options->lower_flrp32'),
(('flrp@64', a, b, c), ('fadd', ('fmul', c, ('fsub', b, a)), a), 'options->lower_flrp64'),
+ (('ffloor', a), ('fsub', a, ('ffract', a)), 'options->lower_ffloor'),
(('ffract', a), ('fsub', a, ('ffloor', a)), 'options->lower_ffract'),
- (('~fadd', ('fmul', a, ('fadd', 1.0, ('fneg', ('b2f', c)))), ('fmul', b, ('b2f', c))), ('bcsel', c, b, a), 'options->lower_flrp32'),
+ (('fceil', a), ('fneg', ('ffloor', ('fneg', a))), 'options->lower_fceil'),
+ (('~fadd', ('fmul', a, ('fadd', 1.0, ('fneg', ('b2f', 'c@32')))), ('fmul', b, ('b2f', c))), ('bcsel', c, b, a), 'options->lower_flrp32'),
(('~fadd@32', ('fmul', a, ('fadd', 1.0, ('fneg', c ))), ('fmul', b, c )), ('flrp', a, b, c), '!options->lower_flrp32'),
(('~fadd@64', ('fmul', a, ('fadd', 1.0, ('fneg', c ))), ('fmul', b, c )), ('flrp', a, b, c), '!options->lower_flrp64'),
- (('~fadd', a, ('fmul', ('b2f', c), ('fadd', b, ('fneg', a)))), ('bcsel', c, b, a), 'options->lower_flrp32'),
+ (('~fadd', a, ('fmul', ('b2f', 'c@32'), ('fadd', b, ('fneg', a)))), ('bcsel', c, b, a), 'options->lower_flrp32'),
(('~fadd@32', a, ('fmul', c , ('fadd', b, ('fneg', a)))), ('flrp', a, b, c), '!options->lower_flrp32'),
(('~fadd@64', a, ('fmul', c , ('fadd', b, ('fneg', a)))), ('flrp', a, b, c), '!options->lower_flrp64'),
(('ffma', a, b, c), ('fadd', ('fmul', a, b), c), 'options->lower_ffma'),
(('~fadd', ('fmul', a, b), c), ('ffma', a, b, c), 'options->fuse_ffma'),
+ (('fdot4', ('vec4', a, b, c, 1.0), d), ('fdph', ('vec3', a, b, c), d)),
+ (('fdot4', ('vec4', a, 0.0, 0.0, 0.0), b), ('fmul', a, b)),
+ (('fdot4', ('vec4', a, b, 0.0, 0.0), c), ('fdot2', ('vec2', a, b), c)),
+ (('fdot4', ('vec4', a, b, c, 0.0), d), ('fdot3', ('vec3', a, b, c), d)),
+
+ (('fdot3', ('vec3', a, 0.0, 0.0), b), ('fmul', a, b)),
+ (('fdot3', ('vec3', a, b, 0.0), c), ('fdot2', ('vec2', a, b), c)),
+
# (a * #b + #c) << #d
# ((a * #b) << #d) + (#c << #d)
# (a * (#b << #d)) + (#c << #d)
(('~inot', ('feq', a, b)), ('fne', a, b)),
(('~inot', ('fne', a, b)), ('feq', a, b)),
(('inot', ('ilt', a, b)), ('ige', a, b)),
+ (('inot', ('ult', a, b)), ('uge', a, b)),
(('inot', ('ige', a, b)), ('ilt', a, b)),
+ (('inot', ('uge', a, b)), ('ult', a, b)),
(('inot', ('ieq', a, b)), ('ine', a, b)),
(('inot', ('ine', a, b)), ('ieq', a, b)),
# b2f(a) <= 0.0
# b2f(a) == 0.0 because b2f(a) can only be 0 or 1
# inot(a)
- (('fge', 0.0, ('b2f', a)), ('inot', a)),
-
- (('fge', ('fneg', ('b2f', a)), 0.0), ('inot', a)),
+ (('fge', 0.0, ('b2f', 'a@32')), ('inot', a)),
+
+ (('fge', ('fneg', ('b2f', 'a@32')), 0.0), ('inot', a)),
+
+ (('fne', ('fadd', ('b2f', 'a@32'), ('b2f', 'b@32')), 0.0), ('ior', a, b)),
+ (('fne', ('fmax', ('b2f', 'a@32'), ('b2f', 'b@32')), 0.0), ('ior', a, b)),
+ (('fne', ('bcsel', a, 1.0, ('b2f', 'b@32')) , 0.0), ('ior', a, b)),
+ (('fne', ('b2f', 'a@32'), ('fneg', ('b2f', 'b@32'))), ('ior', a, b)),
+ (('fne', ('fmul', ('b2f', 'a@32'), ('b2f', 'b@32')), 0.0), ('iand', a, b)),
+ (('fne', ('fmin', ('b2f', 'a@32'), ('b2f', 'b@32')), 0.0), ('iand', a, b)),
+ (('fne', ('bcsel', a, ('b2f', 'b@32'), 0.0) , 0.0), ('iand', a, b)),
+ (('fne', ('fadd', ('b2f', 'a@32'), ('fneg', ('b2f', 'b@32'))), 0.0), ('ixor', a, b)),
+ (('fne', ('b2f', 'a@32') , ('b2f', 'b@32') ), ('ixor', a, b)),
+ (('fne', ('fneg', ('b2f', 'a@32')), ('fneg', ('b2f', 'b@32'))), ('ixor', a, b)),
+ (('feq', ('fadd', ('b2f', 'a@32'), ('b2f', 'b@32')), 0.0), ('inot', ('ior', a, b))),
+ (('feq', ('fmax', ('b2f', 'a@32'), ('b2f', 'b@32')), 0.0), ('inot', ('ior', a, b))),
+ (('feq', ('bcsel', a, 1.0, ('b2f', 'b@32')) , 0.0), ('inot', ('ior', a, b))),
+ (('feq', ('b2f', 'a@32'), ('fneg', ('b2f', 'b@32'))), ('inot', ('ior', a, b))),
+ (('feq', ('fmul', ('b2f', 'a@32'), ('b2f', 'b@32')), 0.0), ('inot', ('iand', a, b))),
+ (('feq', ('fmin', ('b2f', 'a@32'), ('b2f', 'b@32')), 0.0), ('inot', ('iand', a, b))),
+ (('feq', ('bcsel', a, ('b2f', 'b@32'), 0.0) , 0.0), ('inot', ('iand', a, b))),
+ (('feq', ('fadd', ('b2f', 'a@32'), ('fneg', ('b2f', 'b@32'))), 0.0), ('ieq', a, b)),
+ (('feq', ('b2f', 'a@32') , ('b2f', 'b@32') ), ('ieq', a, b)),
+ (('feq', ('fneg', ('b2f', 'a@32')), ('fneg', ('b2f', 'b@32'))), ('ieq', a, b)),
+
+ # -(b2f(a) + b2f(b)) < 0
+ # 0 < b2f(a) + b2f(b)
+ # 0 != b2f(a) + b2f(b) b2f must be 0 or 1, so the sum is non-negative
+ # a || b
+ (('flt', ('fneg', ('fadd', ('b2f', 'a@32'), ('b2f', 'b@32'))), 0.0), ('ior', a, b)),
+ (('flt', 0.0, ('fadd', ('b2f', 'a@32'), ('b2f', 'b@32'))), ('ior', a, b)),
+
+ # -(b2f(a) + b2f(b)) >= 0
+ # 0 >= b2f(a) + b2f(b)
+ # 0 == b2f(a) + b2f(b) b2f must be 0 or 1, so the sum is non-negative
+ # !(a || b)
+ (('fge', ('fneg', ('fadd', ('b2f', 'a@32'), ('b2f', 'b@32'))), 0.0), ('inot', ('ior', a, b))),
+ (('fge', 0.0, ('fadd', ('b2f', 'a@32'), ('b2f', 'b@32'))), ('inot', ('ior', a, b))),
+
+ # Some optimizations (below) convert things like (a < b || c < b) into
+ # (min(a, c) < b). However, this interfers with the previous optimizations
+ # that try to remove comparisons with negated sums of b2f. This just
+ # breaks that apart.
+ (('flt', ('fmin', c, ('fneg', ('fadd', ('b2f', 'a@32'), ('b2f', 'b@32')))), 0.0),
+ ('ior', ('flt', c, 0.0), ('ior', a, b))),
+
+ (('~flt', ('fadd', a, b), a), ('flt', b, 0.0)),
+ (('~fge', ('fadd', a, b), a), ('fge', b, 0.0)),
+ (('~feq', ('fadd', a, b), a), ('feq', b, 0.0)),
+ (('~fne', ('fadd', a, b), a), ('fne', b, 0.0)),
+
+ # Cannot remove the addition from ilt or ige due to overflow.
+ (('ieq', ('iadd', a, b), a), ('ieq', b, 0)),
+ (('ine', ('iadd', a, b), a), ('ine', b, 0)),
+
+ # fmin(-b2f(a), b) >= 0.0
+ # -b2f(a) >= 0.0 && b >= 0.0
+ # -b2f(a) == 0.0 && b >= 0.0 -b2f can only be 0 or -1, never >0
+ # b2f(a) == 0.0 && b >= 0.0
+ # a == False && b >= 0.0
+ # !a && b >= 0.0
+ #
+ # The fge in the second replacement is not a typo. I leave the proof that
+ # "fmin(-b2f(a), b) >= 0 <=> fmin(-b2f(a), b) == 0" as an exercise for the
+ # reader.
+ (('fge', ('fmin', ('fneg', ('b2f', 'a@32')), 'b@32'), 0.0), ('iand', ('inot', a), ('fge', b, 0.0))),
+ (('feq', ('fmin', ('fneg', ('b2f', 'a@32')), 'b@32'), 0.0), ('iand', ('inot', a), ('fge', b, 0.0))),
+
+ (('feq', ('b2f', 'a@32'), 0.0), ('inot', a)),
+ (('fne', ('b2f', 'a@32'), 0.0), a),
+ (('ieq', ('b2i', 'a@32'), 0), ('inot', a)),
+ (('ine', ('b2i', 'a@32'), 0), a),
+
+ (('fne', ('u2f', a), 0.0), ('ine', a, 0)),
+ (('feq', ('u2f', a), 0.0), ('ieq', a, 0)),
+ (('fge', ('u2f', a), 0.0), True),
+ (('fge', 0.0, ('u2f', a)), ('uge', 0, a)), # ieq instead?
+ (('flt', ('u2f', a), 0.0), False),
+ (('flt', 0.0, ('u2f', a)), ('ult', 0, a)), # ine instead?
+ (('fne', ('i2f', a), 0.0), ('ine', a, 0)),
+ (('feq', ('i2f', a), 0.0), ('ieq', a, 0)),
+ (('fge', ('i2f', a), 0.0), ('ige', a, 0)),
+ (('fge', 0.0, ('i2f', a)), ('ige', 0, a)),
+ (('flt', ('i2f', a), 0.0), ('ilt', a, 0)),
+ (('flt', 0.0, ('i2f', a)), ('ilt', 0, a)),
# 0.0 < fabs(a)
# fabs(a) > 0.0
# fabs(a) != 0.0 because fabs(a) must be >= 0
# a != 0.0
- (('flt', 0.0, ('fabs', a)), ('fne', a, 0.0)),
+ (('~flt', 0.0, ('fabs', a)), ('fne', a, 0.0)),
+
+ # -fabs(a) < 0.0
+ # fabs(a) > 0.0
+ (('~flt', ('fneg', ('fabs', a)), 0.0), ('fne', a, 0.0)),
+
+ # 0.0 >= fabs(a)
+ # 0.0 == fabs(a) because fabs(a) must be >= 0
+ # 0.0 == a
+ (('fge', 0.0, ('fabs', a)), ('feq', a, 0.0)),
+
+ # -fabs(a) >= 0.0
+ # 0.0 >= fabs(a)
+ (('fge', ('fneg', ('fabs', a)), 0.0), ('feq', a, 0.0)),
+
+ (('fmax', ('b2f(is_used_once)', 'a@32'), ('b2f', 'b@32')), ('b2f', ('ior', a, b))),
+ (('fmax', ('fneg(is_used_once)', ('b2f(is_used_once)', 'a@32')), ('fneg', ('b2f', 'b@32'))), ('fneg', ('b2f', ('ior', a, b)))),
+ (('fmin', ('b2f(is_used_once)', 'a@32'), ('b2f', 'b@32')), ('b2f', ('iand', a, b))),
+ (('fmin', ('fneg(is_used_once)', ('b2f(is_used_once)', 'a@32')), ('fneg', ('b2f', 'b@32'))), ('fneg', ('b2f', ('iand', a, b)))),
- (('fmax', ('b2f(is_used_once)', a), ('b2f', b)), ('b2f', ('ior', a, b))),
- (('fmax', ('fneg(is_used_once)', ('b2f(is_used_once)', a)), ('fneg', ('b2f', b))), ('fneg', ('b2f', ('ior', a, b)))),
- (('fmin', ('b2f(is_used_once)', a), ('b2f', b)), ('b2f', ('iand', a, b))),
- (('fmin', ('fneg(is_used_once)', ('b2f(is_used_once)', a)), ('fneg', ('b2f', b))), ('fneg', ('b2f', ('iand', a, b)))),
+ # fmin(b2f(a), b)
+ # bcsel(a, fmin(b2f(a), b), fmin(b2f(a), b))
+ # bcsel(a, fmin(b2f(True), b), fmin(b2f(False), b))
+ # bcsel(a, fmin(1.0, b), fmin(0.0, b))
+ #
+ # Since b is a constant, constant folding will eliminate the fmin and the
+ # fmax. If b is > 1.0, the bcsel will be replaced with a b2f.
+ (('fmin', ('b2f', 'a@32'), '#b'), ('bcsel', a, ('fmin', b, 1.0), ('fmin', b, 0.0))),
- # ignore this opt when the result is used by a bcsel or if so we can make
- # use of conditional modifiers on supported hardware.
- (('flt(is_not_used_by_conditional)', ('fadd(is_used_once)', a, ('fneg', b)), 0.0), ('flt', a, b)),
+ (('flt', ('fadd(is_used_once)', a, ('fneg', b)), 0.0), ('flt', a, b)),
(('fge', ('fneg', ('fabs', a)), 0.0), ('feq', a, 0.0)),
- (('bcsel', ('flt', b, a), b, a), ('fmin', a, b)),
- (('bcsel', ('flt', a, b), b, a), ('fmax', a, b)),
+ (('~bcsel', ('flt', b, a), b, a), ('fmin', a, b)),
+ (('~bcsel', ('flt', a, b), b, a), ('fmax', a, b)),
+ (('~bcsel', ('fge', a, b), b, a), ('fmin', a, b)),
+ (('~bcsel', ('fge', b, a), b, a), ('fmax', a, b)),
(('bcsel', ('inot', a), b, c), ('bcsel', a, c, b)),
(('bcsel', a, ('bcsel', a, b, c), d), ('bcsel', a, b, d)),
+ (('bcsel', a, b, ('bcsel', a, c, d)), ('bcsel', a, b, d)),
+ (('bcsel', a, ('bcsel', b, c, d), ('bcsel(is_used_once)', b, c, 'e')), ('bcsel', b, c, ('bcsel', a, d, 'e'))),
+ (('bcsel', a, ('bcsel(is_used_once)', b, c, d), ('bcsel', b, c, 'e')), ('bcsel', b, c, ('bcsel', a, d, 'e'))),
+ (('bcsel', a, ('bcsel', b, c, d), ('bcsel(is_used_once)', b, 'e', d)), ('bcsel', b, ('bcsel', a, c, 'e'), d)),
+ (('bcsel', a, ('bcsel(is_used_once)', b, c, d), ('bcsel', b, 'e', d)), ('bcsel', b, ('bcsel', a, c, 'e'), d)),
(('bcsel', a, True, 'b@bool'), ('ior', a, b)),
(('fmin', a, a), a),
(('fmax', a, a), a),
(('imax', a, a), a),
(('umin', a, a), a),
(('umax', a, a), a),
+ (('fmax', ('fmax', a, b), b), ('fmax', a, b)),
+ (('umax', ('umax', a, b), b), ('umax', a, b)),
+ (('imax', ('imax', a, b), b), ('imax', a, b)),
+ (('fmin', ('fmin', a, b), b), ('fmin', a, b)),
+ (('umin', ('umin', a, b), b), ('umin', a, b)),
+ (('imin', ('imin', a, b), b), ('imin', a, b)),
+ (('fmax', a, ('fneg', a)), ('fabs', a)),
+ (('imax', a, ('ineg', a)), ('iabs', a)),
(('fmin', a, ('fneg', a)), ('fneg', ('fabs', a))),
(('imin', a, ('ineg', a)), ('ineg', ('iabs', a))),
(('fmin', a, ('fneg', ('fabs', a))), ('fneg', ('fabs', a))),
(('imax', a, ('ineg', a)), ('iabs', a)),
(('~fmin', ('fmax', a, 0.0), 1.0), ('fsat', a), '!options->lower_fsat'),
(('~fmax', ('fmin', a, 1.0), 0.0), ('fsat', a), '!options->lower_fsat'),
+ (('fsat', ('fsign', a)), ('b2f', ('flt', 0.0, a))),
(('fsat', a), ('fmin', ('fmax', a, 0.0), 1.0), 'options->lower_fsat'),
(('fsat', ('fsat', a)), ('fsat', a)),
(('fmin', ('fmax', ('fmin', ('fmax', a, b), c), b), c), ('fmin', ('fmax', a, b), c)),
(('~ior', ('flt', '#a', c), ('flt', '#b', c)), ('flt', ('fmin', a, b), c)),
(('~ior', ('fge', a, '#b'), ('fge', a, '#c')), ('fge', a, ('fmin', b, c))),
(('~ior', ('fge', '#a', c), ('fge', '#b', c)), ('fge', ('fmax', a, b), c)),
+ (('~iand', ('flt(is_used_once)', a, b), ('flt', a, c)), ('flt', a, ('fmin', b, c))),
+ (('~iand', ('flt(is_used_once)', a, c), ('flt', b, c)), ('flt', ('fmax', a, b), c)),
+ (('~iand', ('fge(is_used_once)', a, b), ('fge', a, c)), ('fge', a, ('fmax', b, c))),
+ (('~iand', ('fge(is_used_once)', a, c), ('fge', b, c)), ('fge', ('fmin', a, b), c)),
+ (('~iand', ('flt', a, '#b'), ('flt', a, '#c')), ('flt', a, ('fmin', b, c))),
+ (('~iand', ('flt', '#a', c), ('flt', '#b', c)), ('flt', ('fmax', a, b), c)),
+ (('~iand', ('fge', a, '#b'), ('fge', a, '#c')), ('fge', a, ('fmax', b, c))),
+ (('~iand', ('fge', '#a', c), ('fge', '#b', c)), ('fge', ('fmin', a, b), c)),
+
+ (('ior', ('ilt(is_used_once)', a, b), ('ilt', a, c)), ('ilt', a, ('imax', b, c))),
+ (('ior', ('ilt(is_used_once)', a, c), ('ilt', b, c)), ('ilt', ('imin', a, b), c)),
+ (('ior', ('ige(is_used_once)', a, b), ('ige', a, c)), ('ige', a, ('imin', b, c))),
+ (('ior', ('ige(is_used_once)', a, c), ('ige', b, c)), ('ige', ('imax', a, b), c)),
+ (('ior', ('ult(is_used_once)', a, b), ('ult', a, c)), ('ult', a, ('umax', b, c))),
+ (('ior', ('ult(is_used_once)', a, c), ('ult', b, c)), ('ult', ('umin', a, b), c)),
+ (('ior', ('uge(is_used_once)', a, b), ('uge', a, c)), ('uge', a, ('umin', b, c))),
+ (('ior', ('uge(is_used_once)', a, c), ('uge', b, c)), ('uge', ('umax', a, b), c)),
+ (('iand', ('ilt(is_used_once)', a, b), ('ilt', a, c)), ('ilt', a, ('imin', b, c))),
+ (('iand', ('ilt(is_used_once)', a, c), ('ilt', b, c)), ('ilt', ('imax', a, b), c)),
+ (('iand', ('ige(is_used_once)', a, b), ('ige', a, c)), ('ige', a, ('imax', b, c))),
+ (('iand', ('ige(is_used_once)', a, c), ('ige', b, c)), ('ige', ('imin', a, b), c)),
+ (('iand', ('ult(is_used_once)', a, b), ('ult', a, c)), ('ult', a, ('umin', b, c))),
+ (('iand', ('ult(is_used_once)', a, c), ('ult', b, c)), ('ult', ('umax', a, b), c)),
+ (('iand', ('uge(is_used_once)', a, b), ('uge', a, c)), ('uge', a, ('umax', b, c))),
+ (('iand', ('uge(is_used_once)', a, c), ('uge', b, c)), ('uge', ('umin', a, b), c)),
+
+ (('ior', 'a@bool', ('ieq', a, False)), True),
+ (('ior', a, ('inot', a)), -1),
+
+ (('iand', ('ieq', 'a@32', 0), ('ieq', 'b@32', 0)), ('ieq', ('ior', 'a@32', 'b@32'), 0)),
# These patterns can result when (a < b || a < c) => (a < min(b, c))
# transformations occur before constant propagation and loop-unrolling.
(('~fge', ('fmax', a, b), a), True),
(('~flt', a, ('fmin', b, a)), False),
(('~flt', ('fmax', a, b), a), False),
+ (('~fge', a, ('fmax', b, a)), ('fge', a, b)),
+ (('~fge', ('fmin', a, b), a), ('fge', b, a)),
+
+ (('ilt', a, ('imax', b, a)), ('ilt', a, b)),
+ (('ilt', ('imin', a, b), a), ('ilt', b, a)),
+ (('ige', a, ('imin', b, a)), True),
+ (('ige', ('imax', a, b), a), True),
+ (('ult', a, ('umax', b, a)), ('ult', a, b)),
+ (('ult', ('umin', a, b), a), ('ult', b, a)),
+ (('uge', a, ('umin', b, a)), True),
+ (('uge', ('umax', a, b), a), True),
+ (('ilt', a, ('imin', b, a)), False),
+ (('ilt', ('imax', a, b), a), False),
+ (('ige', a, ('imax', b, a)), ('ige', a, b)),
+ (('ige', ('imin', a, b), a), ('ige', b, a)),
+ (('ult', a, ('umin', b, a)), False),
+ (('ult', ('umax', a, b), a), False),
+ (('uge', a, ('umax', b, a)), ('uge', a, b)),
+ (('uge', ('umin', a, b), a), ('uge', b, a)),
+
+ (('ilt', '#a', ('imax', '#b', c)), ('ior', ('ilt', a, b), ('ilt', a, c))),
+ (('ilt', ('imin', '#a', b), '#c'), ('ior', ('ilt', a, c), ('ilt', b, c))),
+ (('ige', '#a', ('imin', '#b', c)), ('ior', ('ige', a, b), ('ige', a, c))),
+ (('ige', ('imax', '#a', b), '#c'), ('ior', ('ige', a, c), ('ige', b, c))),
+ (('ult', '#a', ('umax', '#b', c)), ('ior', ('ult', a, b), ('ult', a, c))),
+ (('ult', ('umin', '#a', b), '#c'), ('ior', ('ult', a, c), ('ult', b, c))),
+ (('uge', '#a', ('umin', '#b', c)), ('ior', ('uge', a, b), ('uge', a, c))),
+ (('uge', ('umax', '#a', b), '#c'), ('ior', ('uge', a, c), ('uge', b, c))),
+ (('ilt', '#a', ('imin', '#b', c)), ('iand', ('ilt', a, b), ('ilt', a, c))),
+ (('ilt', ('imax', '#a', b), '#c'), ('iand', ('ilt', a, c), ('ilt', b, c))),
+ (('ige', '#a', ('imax', '#b', c)), ('iand', ('ige', a, b), ('ige', a, c))),
+ (('ige', ('imin', '#a', b), '#c'), ('iand', ('ige', a, c), ('ige', b, c))),
+ (('ult', '#a', ('umin', '#b', c)), ('iand', ('ult', a, b), ('ult', a, c))),
+ (('ult', ('umax', '#a', b), '#c'), ('iand', ('ult', a, c), ('ult', b, c))),
+ (('uge', '#a', ('umax', '#b', c)), ('iand', ('uge', a, b), ('uge', a, c))),
+ (('uge', ('umin', '#a', b), '#c'), ('iand', ('uge', a, c), ('uge', b, c))),
(('fabs', ('slt', a, b)), ('slt', a, b)),
(('fabs', ('sge', a, b)), ('sge', a, b)),
(('fne', ('fneg', a), a), ('fne', a, 0.0)),
(('feq', ('fneg', a), a), ('feq', a, 0.0)),
# Emulating booleans
- (('imul', ('b2i', a), ('b2i', b)), ('b2i', ('iand', a, b))),
- (('fmul', ('b2f', a), ('b2f', b)), ('b2f', ('iand', a, b))),
- (('fsat', ('fadd', ('b2f', a), ('b2f', b))), ('b2f', ('ior', a, b))),
- (('iand', 'a@bool', 1.0), ('b2f', a)),
+ (('imul', ('b2i', 'a@32'), ('b2i', 'b@32')), ('b2i', ('iand', a, b))),
+ (('fmul', ('b2f', 'a@32'), ('b2f', 'b@32')), ('b2f', ('iand', a, b))),
+ (('fsat', ('fadd', ('b2f', 'a@32'), ('b2f', 'b@32'))), ('b2f', ('ior', a, b))),
+ (('iand', 'a@bool', 1.0), ('b2f', a), '!options->lower_b2f'),
# True/False are ~0 and 0 in NIR. b2i of True is 1, and -1 is ~0 (True).
- (('ineg', ('b2i@32', a)), a),
- (('flt', ('fneg', ('b2f', a)), 0), a), # Generated by TGSI KILL_IF.
- (('flt', ('fsub', 0.0, ('b2f', a)), 0), a), # Generated by TGSI KILL_IF.
+ (('ineg', ('b2i32', 'a@32')), a),
+ (('flt', ('fneg', ('b2f', 'a@32')), 0), a), # Generated by TGSI KILL_IF.
+ (('flt', ('fsub', 0.0, ('b2f', 'a@32')), 0), a), # Generated by TGSI KILL_IF.
# Comparison with the same args. Note that these are not done for
# the float versions because NaN always returns false on float
# inequalities.
(('ixor', a, a), 0),
(('ixor', a, 0), a),
(('inot', ('inot', a)), a),
+ (('ior', ('iand', a, b), b), b),
+ (('ior', ('ior', a, b), b), ('ior', a, b)),
+ (('iand', ('ior', a, b), b), b),
+ (('iand', ('iand', a, b), b), ('iand', a, b)),
# DeMorgan's Laws
(('iand', ('inot', a), ('inot', b)), ('inot', ('ior', a, b))),
(('ior', ('inot', a), ('inot', b)), ('inot', ('iand', a, b))),
(('~fexp2', ('fmul', ('flog2', a), b)), ('fpow', a, b), '!options->lower_fpow'), # 2^(lg2(a)*b) = a^b
(('~fexp2', ('fadd', ('fmul', ('flog2', a), b), ('fmul', ('flog2', c), d))),
('~fmul', ('fpow', a, b), ('fpow', c, d)), '!options->lower_fpow'), # 2^(lg2(a) * b + lg2(c) + d) = a^b * c^d
+ (('~fexp2', ('fmul', ('flog2', a), 2.0)), ('fmul', a, a)),
+ (('~fexp2', ('fmul', ('flog2', a), 4.0)), ('fmul', ('fmul', a, a), ('fmul', a, a))),
(('~fpow', a, 1.0), a),
(('~fpow', a, 2.0), ('fmul', a, a)),
(('~fpow', a, 4.0), ('fmul', ('fmul', a, a), ('fmul', a, a))),
(('~flog2', ('frcp', a)), ('fneg', ('flog2', a))),
(('~flog2', ('frsq', a)), ('fmul', -0.5, ('flog2', a))),
(('~flog2', ('fpow', a, b)), ('fmul', b, ('flog2', a))),
- (('~fmul', ('fexp2', a), ('fexp2', b)), ('fexp2', ('fadd', a, b))),
+ (('~fmul', ('fexp2(is_used_once)', a), ('fexp2(is_used_once)', b)), ('fexp2', ('fadd', a, b))),
# Division and reciprocal
(('~fdiv', 1.0, a), ('frcp', a)),
(('fdiv', a, b), ('fmul', a, ('frcp', b)), 'options->lower_fdiv'),
(('bcsel@32', a, -0.0, -1.0), ('fneg', ('b2f', ('inot', a)))),
(('bcsel', True, b, c), b),
(('bcsel', False, b, c), c),
+ (('bcsel', a, ('b2f(is_used_once)', 'b@32'), ('b2f', 'c@32')), ('b2f', ('bcsel', a, b, c))),
# The result of this should be hit by constant propagation and, in the
# next round of opt_algebraic, get picked up by one of the above two.
(('bcsel', '#a', b, c), ('bcsel', ('ine', 'a', 0), b, c)),
(('fcsel', a, b, b), b),
# Conversions
- (('i2b', ('b2i', a)), a),
- (('f2i32', ('ftrunc', a)), ('f2i32', a)),
- (('f2u32', ('ftrunc', a)), ('f2u32', a)),
+ (('i2b32', ('b2i', 'a@32')), a),
+ (('i2b32', 'a@bool'), a),
+ (('f2i', ('ftrunc', a)), ('f2i', a)),
+ (('f2u', ('ftrunc', a)), ('f2u', a)),
(('i2b', ('ineg', a)), ('i2b', a)),
(('i2b', ('iabs', a)), ('i2b', a)),
(('fabs', ('b2f', a)), ('b2f', a)),
(('iabs', ('b2i', a)), ('b2i', a)),
+ (('inot', ('f2b32', a)), ('feq', a, 0.0)),
+
+ # Ironically, mark these as imprecise because removing the conversions may
+ # preserve more precision than doing the conversions (e.g.,
+ # uint(float(0x81818181u)) == 0x81818200).
+ (('~f2i32', ('i2f', 'a@32')), a),
+ (('~f2i32', ('u2f', 'a@32')), a),
+ (('~f2u32', ('i2f', 'a@32')), a),
+ (('~f2u32', ('u2f', 'a@32')), a),
# Packing and then unpacking does nothing
(('unpack_64_2x32_split_x', ('pack_64_2x32_split', a, b)), a),
('unpack_64_2x32_split_y', a)), a),
# Byte extraction
- (('ushr', a, 24), ('extract_u8', a, 3), '!options->lower_extract_byte'),
+ (('ushr', ('ishl', 'a@32', 24), 24), ('extract_u8', a, 0), '!options->lower_extract_byte'),
+ (('ushr', ('ishl', 'a@32', 16), 24), ('extract_u8', a, 1), '!options->lower_extract_byte'),
+ (('ushr', ('ishl', 'a@32', 8), 24), ('extract_u8', a, 2), '!options->lower_extract_byte'),
+ (('ushr', 'a@32', 24), ('extract_u8', a, 3), '!options->lower_extract_byte'),
+ (('ishr', ('ishl', 'a@32', 24), 24), ('extract_i8', a, 0), '!options->lower_extract_byte'),
+ (('ishr', ('ishl', 'a@32', 16), 24), ('extract_i8', a, 1), '!options->lower_extract_byte'),
+ (('ishr', ('ishl', 'a@32', 8), 24), ('extract_i8', a, 2), '!options->lower_extract_byte'),
+ (('ishr', 'a@32', 24), ('extract_i8', a, 3), '!options->lower_extract_byte'),
(('iand', 0xff, ('ushr', a, 16)), ('extract_u8', a, 2), '!options->lower_extract_byte'),
(('iand', 0xff, ('ushr', a, 8)), ('extract_u8', a, 1), '!options->lower_extract_byte'),
(('iand', 0xff, a), ('extract_u8', a, 0), '!options->lower_extract_byte'),
# Word extraction
- (('ushr', a, 16), ('extract_u16', a, 1), '!options->lower_extract_word'),
+ (('ushr', ('ishl', 'a@32', 16), 16), ('extract_u16', a, 0), '!options->lower_extract_word'),
+ (('ushr', 'a@32', 16), ('extract_u16', a, 1), '!options->lower_extract_word'),
+ (('ishr', ('ishl', 'a@32', 16), 16), ('extract_i16', a, 0), '!options->lower_extract_word'),
+ (('ishr', 'a@32', 16), ('extract_i16', a, 1), '!options->lower_extract_word'),
(('iand', 0xffff, a), ('extract_u16', a, 0), '!options->lower_extract_word'),
# Subtracts
('bfi', ('bfm', 'bits', 'offset'), 'insert', 'base')),
'options->lower_bitfield_insert'),
+ # Alternative lowering that doesn't rely on bfi.
+ (('bitfield_insert', 'base', 'insert', 'offset', 'bits'),
+ ('bcsel', ('ilt', 31, 'bits'),
+ 'insert',
+ ('ior',
+ ('iand', 'base', ('inot', ('bfm', 'bits', 'offset'))),
+ ('iand', ('ishl', 'insert', 'offset'), ('bfm', 'bits', 'offset')))),
+ 'options->lower_bitfield_insert_to_shifts'),
+
+ # bfm lowering -- note that the NIR opcode is undefined if either arg is 32.
+ (('bfm', 'bits', 'offset'),
+ ('ishl', ('isub', ('ishl', 1, 'bits'), 1), 'offset'),
+ 'options->lower_bfm'),
+
(('ibitfield_extract', 'value', 'offset', 'bits'),
('bcsel', ('ilt', 31, 'bits'), 'value',
('ibfe', 'value', 'offset', 'bits')),
('ubfe', 'value', 'offset', 'bits')),
'options->lower_bitfield_extract'),
+ (('ibitfield_extract', 'value', 'offset', 'bits'),
+ ('bcsel', ('ieq', 0, 'bits'),
+ 0,
+ ('ishr',
+ ('ishl', 'value', ('isub', ('isub', 32, 'bits'), 'offset')),
+ ('isub', 32, 'bits'))),
+ 'options->lower_bitfield_extract_to_shifts'),
+
+ (('ubitfield_extract', 'value', 'offset', 'bits'),
+ ('iand',
+ ('ushr', 'value', 'offset'),
+ ('bcsel', ('ieq', 'bits', 32),
+ 0xffffffff,
+ ('bfm', 'bits', 0))),
+ 'options->lower_bitfield_extract_to_shifts'),
+
+ (('ifind_msb', 'value'),
+ ('ufind_msb', ('bcsel', ('ilt', 'value', 0), ('inot', 'value'), 'value')),
+ 'options->lower_ifind_msb'),
+
+ (('find_lsb', 'value'),
+ ('ufind_msb', ('iand', 'value', ('ineg', 'value'))),
+ 'options->lower_find_lsb'),
+
(('extract_i8', a, 'b@32'),
('ishr', ('ishl', a, ('imul', ('isub', 3, b), 8)), 24),
'options->lower_extract_byte'),
(('unpack_unorm_2x16', 'v'),
('fdiv', ('u2f32', ('vec2', ('extract_u16', 'v', 0),
- ('extract_u16', 'v', 1))),
+ ('extract_u16', 'v', 1))),
65535.0),
'options->lower_unpack_unorm_2x16'),
(('unpack_unorm_4x8', 'v'),
('fdiv', ('u2f32', ('vec4', ('extract_u8', 'v', 0),
- ('extract_u8', 'v', 1),
- ('extract_u8', 'v', 2),
- ('extract_u8', 'v', 3))),
+ ('extract_u8', 'v', 1),
+ ('extract_u8', 'v', 2),
+ ('extract_u8', 'v', 3))),
255.0),
'options->lower_unpack_unorm_4x8'),
(('unpack_snorm_2x16', 'v'),
- ('fmin', 1.0, ('fmax', -1.0, ('fdiv', ('i2f32', ('vec2', ('extract_i16', 'v', 0),
- ('extract_i16', 'v', 1))),
+ ('fmin', 1.0, ('fmax', -1.0, ('fdiv', ('i2f', ('vec2', ('extract_i16', 'v', 0),
+ ('extract_i16', 'v', 1))),
32767.0))),
'options->lower_unpack_snorm_2x16'),
(('unpack_snorm_4x8', 'v'),
- ('fmin', 1.0, ('fmax', -1.0, ('fdiv', ('i2f32', ('vec4', ('extract_i8', 'v', 0),
- ('extract_i8', 'v', 1),
- ('extract_i8', 'v', 2),
- ('extract_i8', 'v', 3))),
+ ('fmin', 1.0, ('fmax', -1.0, ('fdiv', ('i2f', ('vec4', ('extract_i8', 'v', 0),
+ ('extract_i8', 'v', 1),
+ ('extract_i8', 'v', 2),
+ ('extract_i8', 'v', 3))),
127.0))),
'options->lower_unpack_snorm_4x8'),
]
+invert = OrderedDict([('feq', 'fne'), ('fne', 'feq'), ('fge', 'flt'), ('flt', 'fge')])
+
+for left, right in itertools.combinations_with_replacement(invert.keys(), 2):
+ optimizations.append((('inot', ('ior(is_used_once)', (left, a, b), (right, c, d))),
+ ('iand', (invert[left], a, b), (invert[right], c, d))))
+ optimizations.append((('inot', ('iand(is_used_once)', (left, a, b), (right, c, d))),
+ ('ior', (invert[left], a, b), (invert[right], c, d))))
+
+# Optimize x2yN(b2x(x)) -> b2y
+optimizations.append((('f2b32', ('b2f', 'a@32')), a))
+optimizations.append((('i2b32', ('b2i', 'a@32')), a))
+for x, y in itertools.product(['f', 'u', 'i'], ['f', 'u', 'i']):
+ if x != 'f' and y != 'f' and x != y:
+ continue
+
+ b2x = 'b2f' if x == 'f' else 'b2i'
+ b2y = 'b2f' if y == 'f' else 'b2i'
+ x2yN = '{}2{}'.format(x, y)
+ optimizations.append(((x2yN, (b2x, a)), (b2y, a)))
+
def fexp2i(exp, bits):
# We assume that exp is already in the right range.
if bits == 32:
return ('fmul', ('fmul', f, pow2_1), pow2_2)
optimizations += [
- (('ldexp@32', 'x', 'exp'), ldexp('x', 'exp', 32)),
- (('ldexp@64', 'x', 'exp'), ldexp('x', 'exp', 64)),
+ (('ldexp@32', 'x', 'exp'), ldexp('x', 'exp', 32), 'options->lower_ldexp'),
+ (('ldexp@64', 'x', 'exp'), ldexp('x', 'exp', 64), 'options->lower_ldexp'),
]
# Unreal Engine 4 demo applications open-codes bitfieldReverse()
('bcsel', 'a', (op, 'd', 'b'), (op, 'd', 'c'))),
]
+
+# For example, this converts things like
+#
+# 1 + mix(0, a - 1, condition)
+#
+# into
+#
+# mix(1, (a-1)+1, condition)
+#
+# Other optimizations will rearrange the constants.
+for op in ['fadd', 'fmul', 'iadd', 'imul']:
+ optimizations += [
+ ((op, ('bcsel(is_used_once)', a, '#b', c), '#d'), ('bcsel', a, (op, b, d), (op, c, d)))
+ ]
+
# This section contains "late" optimizations that should be run before
# creating ffmas and calling regular optimizations for the final time.
# Optimizations should go here if they help code generation and conflict
(('fdot4', a, b), ('fdot_replicated4', a, b), 'options->fdot_replicates'),
(('fdph', a, b), ('fdph_replicated', a, b), 'options->fdot_replicates'),
- (('b2f(is_used_more_than_once)', ('inot', a)), ('bcsel', a, 0.0, 1.0)),
- (('fneg(is_used_more_than_once)', ('b2f', ('inot', a))), ('bcsel', a, -0.0, -1.0)),
+ (('b2f(is_used_more_than_once)', ('inot', 'a@32')), ('bcsel', a, 0.0, 1.0)),
+ (('fneg(is_used_more_than_once)', ('b2f', ('inot', 'a@32'))), ('bcsel', a, -0.0, -1.0)),
# we do these late so that we don't get in the way of creating ffmas
(('fmin', ('fadd(is_used_once)', '#c', a), ('fadd(is_used_once)', '#c', b)), ('fadd', c, ('fmin', a, b))),
(('fmax', ('fadd(is_used_once)', '#c', a), ('fadd(is_used_once)', '#c', b)), ('fadd', c, ('fmax', a, b))),
+
+ # Lowered for backends without a dedicated b2f instruction
+ (('b2f32', 'a@32'), ('iand', a, 1.0), 'options->lower_b2f'),
]
-print nir_algebraic.AlgebraicPass("nir_opt_algebraic", optimizations).render()
-print nir_algebraic.AlgebraicPass("nir_opt_algebraic_before_ffma",
- before_ffma_optimizations).render()
-print nir_algebraic.AlgebraicPass("nir_opt_algebraic_late",
- late_optimizations).render()
+print(nir_algebraic.AlgebraicPass("nir_opt_algebraic", optimizations).render())
+print(nir_algebraic.AlgebraicPass("nir_opt_algebraic_before_ffma",
+ before_ffma_optimizations).render())
+print(nir_algebraic.AlgebraicPass("nir_opt_algebraic_late",
+ late_optimizations).render())