1/*
2 * Copyright © 2015 Intel Corporation
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
13 * Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
21 * IN THE SOFTWARE.
22 *
23 */
24
25#include "nir.h"
26#include "nir_builder.h"
27
28/*
29 * lowers:
30 *
31 * packDouble2x32(foo) -> packDouble2x32Split(foo.x, foo.y)
32 * unpackDouble2x32(foo) -> vec2(unpackDouble2x32_x(foo), unpackDouble2x32_y(foo))
33 * packInt2x32(foo) -> packInt2x32Split(foo.x, foo.y)
34 * unpackInt2x32(foo) -> vec2(unpackInt2x32_x(foo), unpackInt2x32_y(foo))
35 */
36
37static nir_ssa_def *
38lower_pack_64_from_32(nir_builder *b, nir_ssa_def *src)
39{
40   return nir_pack_64_2x32_split(b, nir_channel(b, src, 0),
41                                    nir_channel(b, src, 1));
42}
43
44static nir_ssa_def *
45lower_unpack_64_to_32(nir_builder *b, nir_ssa_def *src)
46{
47   return nir_vec2(b, nir_unpack_64_2x32_split_x(b, src),
48                      nir_unpack_64_2x32_split_y(b, src));
49}
50
51static nir_ssa_def *
52lower_pack_32_from_16(nir_builder *b, nir_ssa_def *src)
53{
54   return nir_pack_32_2x16_split(b, nir_channel(b, src, 0),
55                                    nir_channel(b, src, 1));
56}
57
58static nir_ssa_def *
59lower_unpack_32_to_16(nir_builder *b, nir_ssa_def *src)
60{
61   return nir_vec2(b, nir_unpack_32_2x16_split_x(b, src),
62                      nir_unpack_32_2x16_split_y(b, src));
63}
64
65static nir_ssa_def *
66lower_pack_64_from_16(nir_builder *b, nir_ssa_def *src)
67{
68   nir_ssa_def *xy = nir_pack_32_2x16_split(b, nir_channel(b, src, 0),
69                                               nir_channel(b, src, 1));
70
71   nir_ssa_def *zw = nir_pack_32_2x16_split(b, nir_channel(b, src, 2),
72                                               nir_channel(b, src, 3));
73
74   return nir_pack_64_2x32_split(b, xy, zw);
75}
76
77static nir_ssa_def *
78lower_unpack_64_to_16(nir_builder *b, nir_ssa_def *src)
79{
80   nir_ssa_def *xy = nir_unpack_64_2x32_split_x(b, src);
81   nir_ssa_def *zw = nir_unpack_64_2x32_split_y(b, src);
82
83   return nir_vec4(b, nir_unpack_32_2x16_split_x(b, xy),
84                      nir_unpack_32_2x16_split_y(b, xy),
85                      nir_unpack_32_2x16_split_x(b, zw),
86                      nir_unpack_32_2x16_split_y(b, zw));
87}
88
89static bool
90lower_pack_impl(nir_function_impl *impl)
91{
92   nir_builder b;
93   nir_builder_init(&b, impl);
94   bool progress = false;
95
96   nir_foreach_block(block, impl) {
97      nir_foreach_instr_safe(instr, block) {
98         if (instr->type != nir_instr_type_alu)
99            continue;
100
101         nir_alu_instr *alu_instr = (nir_alu_instr *) instr;
102
103         if (alu_instr->op != nir_op_pack_64_2x32 &&
104             alu_instr->op != nir_op_unpack_64_2x32 &&
105             alu_instr->op != nir_op_pack_64_4x16 &&
106             alu_instr->op != nir_op_unpack_64_4x16 &&
107             alu_instr->op != nir_op_pack_32_2x16 &&
108             alu_instr->op != nir_op_unpack_32_2x16)
109            continue;
110
111         b.cursor = nir_before_instr(&alu_instr->instr);
112
113         nir_ssa_def *src = nir_ssa_for_alu_src(&b, alu_instr, 0);
114         nir_ssa_def *dest;
115
116         switch (alu_instr->op) {
117         case nir_op_pack_64_2x32:
118            dest = lower_pack_64_from_32(&b, src);
119            break;
120         case nir_op_unpack_64_2x32:
121            dest = lower_unpack_64_to_32(&b, src);
122            break;
123         case nir_op_pack_64_4x16:
124            dest = lower_pack_64_from_16(&b, src);
125            break;
126         case nir_op_unpack_64_4x16:
127            dest = lower_unpack_64_to_16(&b, src);
128            break;
129         case nir_op_pack_32_2x16:
130            dest = lower_pack_32_from_16(&b, src);
131            break;
132         case nir_op_unpack_32_2x16:
133            dest = lower_unpack_32_to_16(&b, src);
134            break;
135         default:
136            unreachable("Impossible opcode");
137         }
138
139         nir_ssa_def_rewrite_uses(&alu_instr->dest.dest.ssa, nir_src_for_ssa(dest));
140         nir_instr_remove(&alu_instr->instr);
141         nir_metadata_preserve(impl, nir_metadata_block_index |
142                                     nir_metadata_dominance);
143         progress = true;
144      }
145   }
146
147   return progress;
148}
149
150bool
151nir_lower_pack(nir_shader *shader)
152{
153   bool progress = false;
154
155   nir_foreach_function(function, shader) {
156      if (function->impl)
157         progress |= lower_pack_impl(function->impl);
158   }
159
160   return false;
161}
162