• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1#
2# Copyright (C) 2020 Collabora, Ltd.
3# Copyright (C) 2018 Alyssa Rosenzweig
4# Copyright (C) 2016 Intel Corporation
5#
6# Permission is hereby granted, free of charge, to any person obtaining a
7# copy of this software and associated documentation files (the "Software"),
8# to deal in the Software without restriction, including without limitation
9# the rights to use, copy, modify, merge, publish, distribute, sublicense,
10# and/or sell copies of the Software, and to permit persons to whom the
11# Software is furnished to do so, subject to the following conditions:
12#
13# The above copyright notice and this permission notice (including the next
14# paragraph) shall be included in all copies or substantial portions of the
15# Software.
16#
17# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
20# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
21# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
22# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
23# IN THE SOFTWARE.
24
25import argparse
26import sys
27import math
28
29a = 'a'
30b = 'b'
31c = 'c'
32
33algebraic_late = [
34    # ineg must be lowered late, but only for integers; floats will try to
35    # have modifiers attached... hence why this has to be here rather than
36    # a more standard lower_negate approach
37
38    (('ineg', a), ('isub', 0, a)),
39]
40
41for isz in ('8', '16', '32'):
42        for osz in ('16', '32', '64'):
43                algebraic_late += [(('b2f' + osz, 'a@' + isz), ('b' + isz + 'csel', a, 1.0, 0.0))]
44
45# There's no native integer min/max instruction, lower those to cmp+bcsel
46for sz in ('8', '16', '32'):
47    for t in ('i', 'u'):
48        algebraic_late += [
49            ((t + 'min', 'a@' + sz, 'b@' + sz), ('b' + sz + 'csel', (t + 'lt' + sz, a, b), a, b)),
50            ((t + 'max', 'a@' + sz, 'b@' + sz), ('b' + sz + 'csel', (t + 'lt' + sz, b, a), a, b))
51        ]
52
53# Midgard is able to type convert down by only one "step" per instruction; if
54# NIR wants more than one step, we need to break up into multiple instructions
55
56converts = []
57
58for op in ('u2u', 'i2i', 'f2f', 'i2f', 'u2f', 'f2i', 'f2u'):
59    srcsz_max = 64
60    dstsz_max = 64
61    # 8 bit float doesn't exist
62    srcsz_min = 8 if op[0] != 'f' else 16
63    dstsz_min = 8 if op[2] != 'f' else 16
64    dstsz = dstsz_min
65    # Iterate over all possible destination and source sizes
66    while dstsz <= dstsz_max:
67        srcsz = srcsz_min
68        while srcsz <= srcsz_max:
69            # Size converter lowering is only needed if src and dst sizes are
70            # spaced by a factor > 2.
71            # Type converter lowering is needed as soon as src_size != dst_size
72            if srcsz != dstsz and ((srcsz * 2 != dstsz and srcsz != dstsz * 2) or op[0] != op[2]):
73                cursz = srcsz
74                rule = a
75                # When converting down we first do the type conversion followed
76                # by one or more size conversions. When converting up, we do
77                # the type conversion at the end. This way we don't have to
78                # deal with the fact that f2f8 doesn't exists.
79                sizeconvop = op[0] + '2' + op[0] if srcsz < dstsz else op[2] + '2' + op[2]
80                if srcsz > dstsz and op[0] != op[2]:
81                    rule = (op + str(int(cursz)), rule)
82                while cursz != dstsz:
83                    cursz = cursz / 2 if dstsz < srcsz else cursz * 2
84                    rule = (sizeconvop + str(int(cursz)), rule)
85                if srcsz < dstsz and op[0] != op[2]:
86                    rule = (op + str(int(cursz)), rule)
87                converts += [((op + str(int(dstsz)), 'a@' + str(int(srcsz))), rule)]
88            srcsz *= 2
89        dstsz *= 2
90
91# Bifrost doesn't have fp16 for a lot of special ops
92SPECIAL = ['fexp2', 'flog2', 'fsin', 'fcos']
93
94for op in SPECIAL:
95        converts += [((op + '@16', a), ('f2f16', (op, ('f2f32', a))))]
96
97converts += [(('f2b32', a), ('fneu32', a, 0.0)),
98             (('i2b32', a), ('ine32', a, 0)),
99             (('b2i32', a), ('iand', 'a@32', 1))]
100
101def main():
102    parser = argparse.ArgumentParser()
103    parser.add_argument('-p', '--import-path', required=True)
104    args = parser.parse_args()
105    sys.path.insert(0, args.import_path)
106    run()
107
108
109def run():
110    import nir_algebraic  # pylint: disable=import-error
111
112    print('#include "bifrost_nir.h"')
113
114    print(nir_algebraic.AlgebraicPass("bifrost_nir_lower_algebraic_late",
115                                      algebraic_late + converts).render())
116
117if __name__ == '__main__':
118    main()
119