pan/midgard: Fix ball/bany opcode table
[mesa.git] / src / panfrost / midgard / midgard_ops.c
1 /* Copyright (c) 2018-2019 Alyssa Rosenzweig (alyssa@rosenzweig.io)
2 *
3 * Permission is hereby granted, free of charge, to any person obtaining a copy
4 * of this software and associated documentation files (the "Software"), to deal
5 * in the Software without restriction, including without limitation the rights
6 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
7 * copies of the Software, and to permit persons to whom the Software is
8 * furnished to do so, subject to the following conditions:
9 *
10 * The above copyright notice and this permission notice shall be included in
11 * all copies or substantial portions of the Software.
12 *
13 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
14 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
15 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
16 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
17 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
18 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
19 * THE SOFTWARE.
20 */
21
22 #include "midgard.h"
23
24 /* Include the definitions of the macros and such */
25
26 #define MIDGARD_OPS_TABLE
27 #include "helpers.h"
28 #undef MIDGARD_OPS_TABLE
29
30 /* Table of mapping opcodes to accompanying properties. This is used for both
31 * the disassembler and the compiler. It is placed in a .c file like this to
32 * avoid duplications in the binary */
33
34 struct mir_op_props alu_opcode_props[256] = {
35 [midgard_alu_op_fadd] = {"fadd", UNITS_ADD | OP_COMMUTES},
36 [midgard_alu_op_fmul] = {"fmul", UNITS_MUL | UNIT_VLUT | OP_COMMUTES},
37 [midgard_alu_op_fmin] = {"fmin", UNITS_MOST | OP_COMMUTES},
38 [midgard_alu_op_fmax] = {"fmax", UNITS_MOST | OP_COMMUTES},
39 [midgard_alu_op_imin] = {"imin", UNITS_MOST | OP_COMMUTES},
40 [midgard_alu_op_imax] = {"imax", UNITS_MOST | OP_COMMUTES},
41 [midgard_alu_op_umin] = {"umin", UNITS_MOST | OP_COMMUTES},
42 [midgard_alu_op_umax] = {"umax", UNITS_MOST | OP_COMMUTES},
43 [midgard_alu_op_ihadd] = {"ihadd", UNITS_ADD | OP_COMMUTES},
44 [midgard_alu_op_uhadd] = {"uhadd", UNITS_ADD | OP_COMMUTES},
45 [midgard_alu_op_irhadd] = {"irhadd", UNITS_ADD | OP_COMMUTES},
46 [midgard_alu_op_urhadd] = {"urhadd", UNITS_ADD | OP_COMMUTES},
47
48 [midgard_alu_op_fmov] = {"fmov", UNITS_ALL | QUIRK_FLIPPED_R24},
49 [midgard_alu_op_fmov_rtz] = {"fmov_rtz", UNITS_ALL | QUIRK_FLIPPED_R24},
50 [midgard_alu_op_fmov_rtn] = {"fmov_rtn", UNITS_ALL | QUIRK_FLIPPED_R24},
51 [midgard_alu_op_fmov_rtp] = {"fmov_rtp", UNITS_ALL | QUIRK_FLIPPED_R24},
52 [midgard_alu_op_fround] = {"fround", UNITS_ADD},
53 [midgard_alu_op_froundeven] = {"froundeven", UNITS_ADD},
54 [midgard_alu_op_ftrunc] = {"ftrunc", UNITS_ADD},
55 [midgard_alu_op_ffloor] = {"ffloor", UNITS_ADD},
56 [midgard_alu_op_fceil] = {"fceil", UNITS_ADD},
57 [midgard_alu_op_ffma] = {"ffma", UNIT_VLUT},
58
59 /* Though they output a scalar, they need to run on a vector unit
60 * since they process vectors */
61 [midgard_alu_op_fdot3] = {"fdot3", UNIT_VMUL | OP_CHANNEL_COUNT(3) | OP_COMMUTES},
62 [midgard_alu_op_fdot3r] = {"fdot3r", UNIT_VMUL | OP_CHANNEL_COUNT(3) | OP_COMMUTES},
63 [midgard_alu_op_fdot4] = {"fdot4", UNIT_VMUL | OP_CHANNEL_COUNT(4) | OP_COMMUTES},
64
65 /* Incredibly, iadd can run on vmul, etc */
66 [midgard_alu_op_iadd] = {"iadd", UNITS_MOST | OP_COMMUTES},
67 [midgard_alu_op_iaddsat] = {"iaddsat", UNITS_ADD | OP_COMMUTES},
68 [midgard_alu_op_uaddsat] = {"uaddsat", UNITS_ADD | OP_COMMUTES},
69 [midgard_alu_op_iabsdiff] = {"iabsdiff", UNITS_ADD},
70 [midgard_alu_op_uabsdiff] = {"uabsdiff", UNITS_ADD},
71 [midgard_alu_op_ichoose] = {"ichoose", UNITS_ADD},
72 [midgard_alu_op_isub] = {"isub", UNITS_MOST},
73 [midgard_alu_op_isubsat] = {"isubsat", UNITS_MOST},
74 [midgard_alu_op_usubsat] = {"usubsat", UNITS_MOST},
75 [midgard_alu_op_imul] = {"imul", UNITS_MUL | OP_COMMUTES},
76 [midgard_alu_op_imov] = {"imov", UNITS_ALL | QUIRK_FLIPPED_R24},
77
78 /* For vector comparisons, use ball etc */
79 [midgard_alu_op_feq] = {"feq", UNITS_MOST | OP_TYPE_CONVERT | OP_COMMUTES},
80 [midgard_alu_op_fne] = {"fne", UNITS_MOST | OP_TYPE_CONVERT | OP_COMMUTES},
81 [midgard_alu_op_fle] = {"fle", UNITS_MOST | OP_TYPE_CONVERT},
82 [midgard_alu_op_flt] = {"flt", UNITS_MOST | OP_TYPE_CONVERT},
83 [midgard_alu_op_ieq] = {"ieq", UNITS_MOST | OP_COMMUTES},
84 [midgard_alu_op_ine] = {"ine", UNITS_MOST | OP_COMMUTES},
85 [midgard_alu_op_ilt] = {"ilt", UNITS_MOST},
86 [midgard_alu_op_ile] = {"ile", UNITS_MOST},
87 [midgard_alu_op_ult] = {"ult", UNITS_MOST},
88 [midgard_alu_op_ule] = {"ule", UNITS_MOST},
89
90 /* csel must run in the second pipeline stage (r31 written in first) */
91 [midgard_alu_op_icsel] = {"icsel", UNIT_VADD | UNIT_SMUL},
92 [midgard_alu_op_icsel_v] = {"icsel_v", UNIT_VADD | UNIT_SMUL}, /* Acts as bitselect() */
93 [midgard_alu_op_fcsel_v] = {"fcsel_v", UNIT_VADD | UNIT_SMUL},
94 [midgard_alu_op_fcsel] = {"fcsel", UNIT_VADD | UNIT_SMUL},
95
96 [midgard_alu_op_frcp] = {"frcp", UNIT_VLUT},
97 [midgard_alu_op_frsqrt] = {"frsqrt", UNIT_VLUT},
98 [midgard_alu_op_fsqrt] = {"fsqrt", UNIT_VLUT},
99 [midgard_alu_op_fpow_pt1] = {"fpow_pt1", UNIT_VLUT},
100 [midgard_alu_op_fpown_pt1] = {"fpown_pt1", UNIT_VLUT},
101 [midgard_alu_op_fpowr_pt1] = {"fpowr_pt1", UNIT_VLUT},
102 [midgard_alu_op_fexp2] = {"fexp2", UNIT_VLUT},
103 [midgard_alu_op_flog2] = {"flog2", UNIT_VLUT},
104
105 [midgard_alu_op_f2i_rte] = {"f2i_rte", UNITS_ADD | OP_TYPE_CONVERT},
106 [midgard_alu_op_f2i_rtz] = {"f2i_rtz", UNITS_ADD | OP_TYPE_CONVERT},
107 [midgard_alu_op_f2i_rtn] = {"f2i_rtn", UNITS_ADD | OP_TYPE_CONVERT},
108 [midgard_alu_op_f2i_rtp] = {"f2i_rtp", UNITS_ADD | OP_TYPE_CONVERT},
109 [midgard_alu_op_f2u_rte] = {"f2i_rte", UNITS_ADD | OP_TYPE_CONVERT},
110 [midgard_alu_op_f2u_rtz] = {"f2i_rtz", UNITS_ADD | OP_TYPE_CONVERT},
111 [midgard_alu_op_f2u_rtn] = {"f2i_rtn", UNITS_ADD | OP_TYPE_CONVERT},
112 [midgard_alu_op_f2u_rtp] = {"f2i_rtp", UNITS_ADD | OP_TYPE_CONVERT},
113 [midgard_alu_op_i2f_rte] = {"i2f", UNITS_ADD | OP_TYPE_CONVERT},
114 [midgard_alu_op_i2f_rtz] = {"i2f_rtz", UNITS_ADD | OP_TYPE_CONVERT},
115 [midgard_alu_op_i2f_rtn] = {"i2f_rtn", UNITS_ADD | OP_TYPE_CONVERT},
116 [midgard_alu_op_i2f_rtp] = {"i2f_rtp", UNITS_ADD | OP_TYPE_CONVERT},
117 [midgard_alu_op_u2f_rte] = {"u2f", UNITS_ADD | OP_TYPE_CONVERT},
118 [midgard_alu_op_u2f_rtz] = {"u2f_rtz", UNITS_ADD | OP_TYPE_CONVERT},
119 [midgard_alu_op_u2f_rtn] = {"u2f_rtn", UNITS_ADD | OP_TYPE_CONVERT},
120 [midgard_alu_op_u2f_rtp] = {"u2f_rtp", UNITS_ADD | OP_TYPE_CONVERT},
121
122 [midgard_alu_op_fsin] = {"fsin", UNIT_VLUT},
123 [midgard_alu_op_fcos] = {"fcos", UNIT_VLUT},
124
125 [midgard_alu_op_iand] = {"iand", UNITS_MOST | OP_COMMUTES},
126 [midgard_alu_op_iandnot] = {"iandnot", UNITS_MOST},
127
128 [midgard_alu_op_ior] = {"ior", UNITS_MOST | OP_COMMUTES},
129 [midgard_alu_op_iornot] = {"iornot", UNITS_MOST | OP_COMMUTES},
130 [midgard_alu_op_inor] = {"inor", UNITS_MOST | OP_COMMUTES},
131 [midgard_alu_op_ixor] = {"ixor", UNITS_MOST | OP_COMMUTES},
132 [midgard_alu_op_inxor] = {"inxor", UNITS_MOST | OP_COMMUTES},
133 [midgard_alu_op_iclz] = {"iclz", UNITS_ADD},
134 [midgard_alu_op_ibitcount8] = {"ibitcount8", UNITS_ADD},
135 [midgard_alu_op_inand] = {"inand", UNITS_MOST},
136 [midgard_alu_op_ishl] = {"ishl", UNITS_ADD},
137 [midgard_alu_op_iasr] = {"iasr", UNITS_ADD},
138 [midgard_alu_op_ilsr] = {"ilsr", UNITS_ADD},
139
140 [midgard_alu_op_fball_eq] = {"fball_eq", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES | OP_TYPE_CONVERT},
141 [midgard_alu_op_fball_neq] = {"fball_neq", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES | OP_TYPE_CONVERT},
142 [midgard_alu_op_fball_lt] = {"fball_lt", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES | OP_TYPE_CONVERT},
143 [midgard_alu_op_fball_lte] = {"fball_lte", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES | OP_TYPE_CONVERT},
144
145 [midgard_alu_op_fbany_eq] = {"fbany_eq", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES | OP_TYPE_CONVERT},
146 [midgard_alu_op_fbany_neq] = {"fbany_neq", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES | OP_TYPE_CONVERT},
147 [midgard_alu_op_fbany_lt] = {"fbany_lt", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES | OP_TYPE_CONVERT},
148 [midgard_alu_op_fbany_lte] = {"fbany_lte", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES | OP_TYPE_CONVERT},
149
150 [midgard_alu_op_iball_eq] = {"iball_eq", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES},
151 [midgard_alu_op_iball_neq] = {"iball_neq", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES},
152 [midgard_alu_op_iball_lt] = {"iball_lt", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES},
153 [midgard_alu_op_iball_lte] = {"iball_lte", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES},
154 [midgard_alu_op_uball_lt] = {"uball_lt", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES},
155 [midgard_alu_op_uball_lte] = {"uball_lte", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES},
156
157 [midgard_alu_op_ibany_eq] = {"ibany_eq", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES},
158 [midgard_alu_op_ibany_neq] = {"ibany_neq", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES},
159 [midgard_alu_op_ibany_lt] = {"ibany_lt", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES},
160 [midgard_alu_op_ibany_lte] = {"ibany_lte", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES},
161 [midgard_alu_op_ubany_lt] = {"ubany_lt", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES},
162 [midgard_alu_op_ubany_lte] = {"ubany_lte", UNITS_VECTOR | OP_CHANNEL_COUNT(4) | OP_COMMUTES},
163
164 /* These instructions are not yet emitted by the compiler, so
165 * don't speculate about units yet */
166 [midgard_alu_op_ishladd] = {"ishladd", 0},
167 [midgard_alu_op_freduce] = {"freduce", 0},
168 [midgard_alu_op_fatan2_pt1] = {"fatan2_pt1", 0},
169 [midgard_alu_op_fatan_pt2] = {"fatan_pt2", 0},
170 };
171
172 const char *load_store_opcode_names[256] = {
173 [midgard_op_st_cubemap_coords] = "st_cubemap_coords",
174 [midgard_op_ld_global_id] = "ld_global_id",
175 [midgard_op_ldst_perspective_division_z] = "ldst_perspective_division_z",
176 [midgard_op_ldst_perspective_division_w] = "ldst_perspective_division_w",
177
178 [midgard_op_atomic_add] = "atomic_add",
179 [midgard_op_atomic_and] = "atomic_and",
180 [midgard_op_atomic_or] = "atomic_or",
181 [midgard_op_atomic_xor] = "atomic_xor",
182 [midgard_op_atomic_imin] = "atomic_imin",
183 [midgard_op_atomic_umin] = "atomic_umin",
184 [midgard_op_atomic_imax] = "atomic_imax",
185 [midgard_op_atomic_umax] = "atomic_umax",
186 [midgard_op_atomic_xchg] = "atomic_xchg",
187
188 [midgard_op_ld_char] = "ld_char",
189 [midgard_op_ld_char2] = "ld_char2",
190 [midgard_op_ld_short] = "ld_short",
191 [midgard_op_ld_char4] = "ld_char4",
192 [midgard_op_ld_short4] = "ld_short4",
193 [midgard_op_ld_int4] = "ld_int4",
194
195 [midgard_op_ld_attr_32] = "ld_attr_32",
196 [midgard_op_ld_attr_16] = "ld_attr_16",
197 [midgard_op_ld_attr_32i] = "ld_attr_32i",
198 [midgard_op_ld_attr_32u] = "ld_attr_32u",
199
200 [midgard_op_ld_vary_32] = "ld_vary_32",
201 [midgard_op_ld_vary_16] = "ld_vary_16",
202 [midgard_op_ld_vary_32i] = "ld_vary_32i",
203 [midgard_op_ld_vary_32u] = "ld_vary_32u",
204
205 [midgard_op_ld_color_buffer_16] = "ld_color_buffer_16",
206
207 [midgard_op_ld_uniform_16] = "ld_uniform_16",
208 [midgard_op_ld_uniform_32] = "ld_uniform_32",
209 [midgard_op_ld_uniform_32i] = "ld_uniform_32i",
210 [midgard_op_ld_color_buffer_8] = "ld_color_buffer_8",
211
212 [midgard_op_st_char] = "st_char",
213 [midgard_op_st_char2] = "st_char2",
214 [midgard_op_st_char4] = "st_char4",
215 [midgard_op_st_short4] = "st_short4",
216 [midgard_op_st_int4] = "st_int4",
217
218 [midgard_op_st_vary_32] = "st_vary_32",
219 [midgard_op_st_vary_16] = "st_vary_16",
220 [midgard_op_st_vary_32i] = "st_vary_32i",
221 [midgard_op_st_vary_32u] = "st_vary_32u",
222
223 [midgard_op_st_image_f] = "st_image_f",
224 [midgard_op_st_image_ui] = "st_image_ui",
225 [midgard_op_st_image_i] = "st_image_i",
226 };