summaryrefslogtreecommitdiffstats
path: root/src/glsl/nir/nir_opcodes.h
blob: c16b7fe25e2430009437dbdff9a6bd7f2dec6d2c (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
/*
 * Copyright © 2014 Intel Corporation
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
 * IN THE SOFTWARE.
 *
 * Authors:
 *    Connor Abbott (cwabbott0@gmail.com)
 *
 */

/**
 * This header file defines all the available opcodes in one place. It expands
 * to a list of macros of the form:
 *
 * OPCODE(name, num_inputs, output_size, output_type,
 *        input_sizes, input_types, algebraic_properties)
 *
 * Which should correspond one-to-one with the nir_op_info structure. It is
 * included in both ir.h to create the nir_op enum (with members of the form
 * nir_op_(name)) and and in opcodes.c to create nir_op_infos, which is a
 * const array of nir_op_info structures for each opcode.
 */

#define ARR(...) { __VA_ARGS__ }

#define UNOP(name, type) OPCODE(name, 1, 0, type, ARR(0), ARR(type), 0)
#define UNOP_CONVERT(name, in_type, out_type) \
   OPCODE(name, 1, 0, out_type, ARR(0), ARR(in_type), 0)
#define UNOP_HORIZ(name, output_size, output_type, input_size, input_type) \
   OPCODE(name, 1, output_size, output_type, ARR(input_size), \
          ARR(input_type), 0)

#define UNOP_REDUCE(name, output_size, output_type, input_type) \
   UNOP_HORIZ(name##2, output_size, output_type, 2, input_type) \
   UNOP_HORIZ(name##3, output_size, output_type, 3, input_type) \
   UNOP_HORIZ(name##4, output_size, output_type, 4, input_type)

/**
 * These two move instructions differ in what modifiers they support and what
 * the negate modifier means. Otherwise, they are identical.
 */
UNOP(fmov, nir_type_float)
UNOP(imov, nir_type_int)

UNOP(ineg, nir_type_int)
UNOP(fneg, nir_type_float)
UNOP(inot, nir_type_int) /* invert every bit of the integer */
UNOP(fnot, nir_type_float) /* (src == 0.0) ? 1.0 : 0.0 */
UNOP(fsign, nir_type_float)
UNOP(isign, nir_type_int)
UNOP(iabs, nir_type_int)
UNOP(fabs, nir_type_float)
UNOP(fsat, nir_type_float)
UNOP(frcp, nir_type_float)
UNOP(frsq, nir_type_float)
UNOP(fsqrt, nir_type_float)
UNOP(fexp, nir_type_float) /* < e^x */
UNOP(flog, nir_type_float) /* log base e */
UNOP(fexp2, nir_type_float)
UNOP(flog2, nir_type_float)
UNOP_CONVERT(f2i, nir_type_float, nir_type_int)       /**< Float-to-integer conversion. */
UNOP_CONVERT(f2u, nir_type_float, nir_type_unsigned)  /**< Float-to-unsigned conversion. */
UNOP_CONVERT(i2f, nir_type_int, nir_type_float)       /**< Integer-to-float conversion. */
UNOP_CONVERT(f2b, nir_type_float, nir_type_bool)      /**< Float-to-boolean conversion */
UNOP_CONVERT(b2f, nir_type_bool, nir_type_float)      /**< Boolean-to-float conversion */
UNOP_CONVERT(i2b, nir_type_int, nir_type_bool)        /**< int-to-boolean conversion */
UNOP_CONVERT(b2i, nir_type_bool, nir_type_int)        /**< Boolean-to-int conversion */
UNOP_CONVERT(u2f, nir_type_unsigned, nir_type_float)  /**< Unsigned-to-float conversion. */

UNOP_REDUCE(bany, 1, nir_type_bool, nir_type_bool) /* returns ~0 if any component of src[0] != 0 */
UNOP_REDUCE(ball, 1, nir_type_bool, nir_type_bool) /* returns ~0 if all components of src[0] != 0 */
UNOP_REDUCE(fany, 1, nir_type_float, nir_type_float) /* returns 1.0 if any component of src[0] != 0 */
UNOP_REDUCE(fall, 1, nir_type_float, nir_type_float) /* returns 1.0 if all components of src[0] != 0 */

/**
 * \name Unary floating-point rounding operations.
 */
/*@{*/
UNOP(ftrunc, nir_type_float)
UNOP(fceil, nir_type_float)
UNOP(ffloor, nir_type_float)
UNOP(ffract, nir_type_float)
UNOP(fround_even, nir_type_float)
/*@}*/

/**
 * \name Trigonometric operations.
 */
/*@{*/
UNOP(fsin, nir_type_float)
UNOP(fcos, nir_type_float)
UNOP(fsin_reduced, nir_type_float)
UNOP(fcos_reduced, nir_type_float)
/*@}*/

/**
 * \name Partial derivatives.
 */
/*@{*/
UNOP(fddx, nir_type_float)
UNOP(fddy, nir_type_float)
UNOP(fddx_fine, nir_type_float)
UNOP(fddy_fine, nir_type_float)
UNOP(fddx_coarse, nir_type_float)
UNOP(fddy_coarse, nir_type_float)
/*@}*/

/**
 * \name Floating point pack and unpack operations.
 */
/*@{*/
UNOP_HORIZ(pack_snorm_2x16, 1, nir_type_unsigned, 2, nir_type_float)
UNOP_HORIZ(pack_snorm_4x8, 1, nir_type_unsigned, 4, nir_type_float)
UNOP_HORIZ(pack_unorm_2x16, 1, nir_type_unsigned, 2, nir_type_float)
UNOP_HORIZ(pack_unorm_4x8, 1, nir_type_unsigned, 4, nir_type_float)
UNOP_HORIZ(pack_half_2x16, 1, nir_type_unsigned, 2, nir_type_float)
UNOP_HORIZ(unpack_snorm_2x16, 2, nir_type_float, 1, nir_type_unsigned)
UNOP_HORIZ(unpack_snorm_4x8, 4, nir_type_float, 1, nir_type_unsigned)
UNOP_HORIZ(unpack_unorm_2x16, 2, nir_type_float, 1, nir_type_unsigned)
UNOP_HORIZ(unpack_unorm_4x8, 4, nir_type_float, 1, nir_type_unsigned)
UNOP_HORIZ(unpack_half_2x16, 2, nir_type_float, 1, nir_type_unsigned)
/*@}*/

/**
 * \name Lowered floating point unpacking operations.
 */
/*@{*/
UNOP_HORIZ(unpack_half_2x16_split_x, 1, nir_type_float, 1, nir_type_unsigned)
UNOP_HORIZ(unpack_half_2x16_split_y, 1, nir_type_float, 1, nir_type_unsigned)
/*@}*/

/**
 * \name Bit operations, part of ARB_gpu_shader5.
 */
/*@{*/
UNOP(bitfield_reverse, nir_type_unsigned)
UNOP(bit_count, nir_type_unsigned)
UNOP_CONVERT(ufind_msb, nir_type_unsigned, nir_type_int)
UNOP(ifind_msb, nir_type_int)
UNOP(find_lsb, nir_type_int)
/*@}*/

UNOP_HORIZ(fnoise1_1, 1, nir_type_float, 1, nir_type_float)
UNOP_HORIZ(fnoise1_2, 1, nir_type_float, 2, nir_type_float)
UNOP_HORIZ(fnoise1_3, 1, nir_type_float, 3, nir_type_float)
UNOP_HORIZ(fnoise1_4, 1, nir_type_float, 4, nir_type_float)
UNOP_HORIZ(fnoise2_1, 2, nir_type_float, 1, nir_type_float)
UNOP_HORIZ(fnoise2_2, 2, nir_type_float, 2, nir_type_float)
UNOP_HORIZ(fnoise2_3, 2, nir_type_float, 3, nir_type_float)
UNOP_HORIZ(fnoise2_4, 2, nir_type_float, 4, nir_type_float)
UNOP_HORIZ(fnoise3_1, 3, nir_type_float, 1, nir_type_float)
UNOP_HORIZ(fnoise3_2, 3, nir_type_float, 2, nir_type_float)
UNOP_HORIZ(fnoise3_3, 3, nir_type_float, 3, nir_type_float)
UNOP_HORIZ(fnoise3_4, 3, nir_type_float, 4, nir_type_float)
UNOP_HORIZ(fnoise4_1, 4, nir_type_float, 1, nir_type_float)
UNOP_HORIZ(fnoise4_2, 4, nir_type_float, 2, nir_type_float)
UNOP_HORIZ(fnoise4_3, 4, nir_type_float, 3, nir_type_float)
UNOP_HORIZ(fnoise4_4, 4, nir_type_float, 4, nir_type_float)

#define BINOP(name, type, alg_props) \
   OPCODE(name, 2, 0, type, ARR(0, 0), ARR(type, type), alg_props)
#define BINOP_CONVERT(name, out_type, in_type, alg_props) \
   OPCODE(name, 2, 0, out_type, ARR(0, 0), ARR(in_type, in_type), alg_props)
#define BINOP_COMPARE(name, type, alg_props) \
   OPCODE(name, 2, 0, nir_type_bool, ARR(0, 0), ARR(type, type), alg_props)
#define BINOP_HORIZ(name, output_size, output_type, src1_size, src1_type, \
                    src2_size, src2_type) \
   OPCODE(name, 2, output_size, output_type, ARR(src1_size, src2_size), \
          ARR(src1_type, src2_type), 0)
#define BINOP_REDUCE(name, output_size, output_type, src_type) \
   OPCODE(name##2, 2, output_size, output_type, \
          ARR(2, 2), ARR(src_type, src_type), NIR_OP_IS_COMMUTATIVE) \
   OPCODE(name##3, 2, output_size, output_type, \
          ARR(3, 3), ARR(src_type, src_type), NIR_OP_IS_COMMUTATIVE) \
   OPCODE(name##4, 2, output_size, output_type, \
          ARR(4, 4), ARR(src_type, src_type), NIR_OP_IS_COMMUTATIVE)

BINOP(fadd, nir_type_float, NIR_OP_IS_COMMUTATIVE | NIR_OP_IS_ASSOCIATIVE)
BINOP(iadd, nir_type_int, NIR_OP_IS_COMMUTATIVE | NIR_OP_IS_ASSOCIATIVE)
BINOP(fsub, nir_type_float, 0)
BINOP(isub, nir_type_int, 0)

BINOP(fmul, nir_type_float, NIR_OP_IS_COMMUTATIVE | NIR_OP_IS_ASSOCIATIVE)
/* low 32-bits of signed/unsigned integer multiply */
BINOP(imul, nir_type_int, NIR_OP_IS_COMMUTATIVE | NIR_OP_IS_ASSOCIATIVE)
/* high 32-bits of signed integer multiply */
BINOP(imul_high, nir_type_int, NIR_OP_IS_COMMUTATIVE)
/* high 32-bits of unsigned integer multiply */
BINOP(umul_high, nir_type_unsigned, NIR_OP_IS_COMMUTATIVE)

BINOP(fdiv, nir_type_float, 0)
BINOP(idiv, nir_type_int, 0)
BINOP(udiv, nir_type_unsigned, 0)

/**
 * returns a boolean representing the carry resulting from the addition of
 * the two unsigned arguments.
 */
BINOP_CONVERT(uadd_carry, nir_type_bool, nir_type_unsigned,
              NIR_OP_IS_COMMUTATIVE)

/**
 * returns a boolean representing the borrow resulting from the subtraction
 * of the two unsigned arguments.
 */
BINOP_CONVERT(usub_borrow, nir_type_bool, nir_type_unsigned, 0)

BINOP(fmod, nir_type_float, 0)
BINOP(umod, nir_type_unsigned, 0)

/**
 * \name comparisons
 */
/*@{*/

/**
 * these integer-aware comparisons return a boolean (0 or ~0)
 */
BINOP_COMPARE(flt, nir_type_float, 0)
BINOP_COMPARE(fge, nir_type_float, 0)
BINOP_COMPARE(feq, nir_type_float, NIR_OP_IS_COMMUTATIVE)
BINOP_COMPARE(fne, nir_type_float, NIR_OP_IS_COMMUTATIVE)
BINOP_COMPARE(ilt, nir_type_int, 0)
BINOP_COMPARE(ige, nir_type_int, 0)
BINOP_COMPARE(ieq, nir_type_int, NIR_OP_IS_COMMUTATIVE)
BINOP_COMPARE(ine, nir_type_int, NIR_OP_IS_COMMUTATIVE)
BINOP_COMPARE(ult, nir_type_unsigned, 0)
BINOP_COMPARE(uge, nir_type_unsigned, 0)

/** integer-aware GLSL-style comparisons that compare floats and ints */
BINOP_REDUCE(ball_fequal,  1, nir_type_bool, nir_type_float)
BINOP_REDUCE(bany_fnequal, 1, nir_type_bool, nir_type_float)
BINOP_REDUCE(ball_iequal,  1, nir_type_bool, nir_type_int)
BINOP_REDUCE(bany_inequal, 1, nir_type_bool, nir_type_int)

/** non-integer-aware GLSL-style comparisons that return 0.0 or 1.0 */
BINOP_REDUCE(fall_equal,  1, nir_type_float, nir_type_float)
BINOP_REDUCE(fany_nequal, 1, nir_type_float, nir_type_float)

/**
 * These comparisons for integer-less hardware return 1.0 and 0.0 for true
 * and false respectively
 */
BINOP(slt, nir_type_float, 0) /* Set on Less Than */
BINOP(sge, nir_type_float, 0) /* Set on Greater Than or Equal */
BINOP(seq, nir_type_float, NIR_OP_IS_COMMUTATIVE) /* Set on Equal */
BINOP(sne, nir_type_float, NIR_OP_IS_COMMUTATIVE) /* Set on Not Equal */

/*@}*/

BINOP(ishl, nir_type_int, 0)
BINOP(ishr, nir_type_int, 0)
BINOP(ushr, nir_type_unsigned, 0)

/**
 * \name bitwise logic operators
 *
 * These are also used as boolean and, or, xor for hardware supporting
 * integers.
 */
/*@{*/
BINOP(iand, nir_type_unsigned, NIR_OP_IS_COMMUTATIVE | NIR_OP_IS_ASSOCIATIVE)
BINOP(ior, nir_type_unsigned, NIR_OP_IS_COMMUTATIVE | NIR_OP_IS_ASSOCIATIVE)
BINOP(ixor, nir_type_unsigned, NIR_OP_IS_COMMUTATIVE | NIR_OP_IS_ASSOCIATIVE)
/*@{*/

/**
 * \name floating point logic operators
 *
 * These use (src != 0.0) for testing the truth of the input, and output 1.0
 * for true and 0.0 for false
 */
BINOP(fand, nir_type_float, NIR_OP_IS_COMMUTATIVE)
BINOP(for, nir_type_float, NIR_OP_IS_COMMUTATIVE)
BINOP(fxor, nir_type_float, NIR_OP_IS_COMMUTATIVE)

BINOP_REDUCE(fdot, 1, nir_type_float, nir_type_float)

BINOP(fmin, nir_type_float, 0)
BINOP(imin, nir_type_int, NIR_OP_IS_COMMUTATIVE | NIR_OP_IS_ASSOCIATIVE)
BINOP(umin, nir_type_unsigned, NIR_OP_IS_COMMUTATIVE | NIR_OP_IS_ASSOCIATIVE)
BINOP(fmax, nir_type_float, 0)
BINOP(imax, nir_type_int, NIR_OP_IS_COMMUTATIVE | NIR_OP_IS_ASSOCIATIVE)
BINOP(umax, nir_type_unsigned, NIR_OP_IS_COMMUTATIVE | NIR_OP_IS_ASSOCIATIVE)

BINOP(fpow, nir_type_float, 0)

BINOP_HORIZ(pack_half_2x16_split, 1, nir_type_unsigned, 1, nir_type_float, 1, nir_type_float)

BINOP(bfm, nir_type_unsigned, 0)

BINOP(ldexp, nir_type_unsigned, 0)

/**
 * Combines the first component of each input to make a 2-component vector.
 */
BINOP_HORIZ(vec2, 2, nir_type_unsigned, 1, nir_type_unsigned, 1, nir_type_unsigned)

#define TRIOP(name, type) \
   OPCODE(name, 3, 0, type, ARR(0, 0, 0), ARR(type, type, type), 0)
#define TRIOP_HORIZ(name, output_size, src1_size, src2_size, src3_size) \
   OPCODE(name, 3, output_size, nir_type_unsigned, \
   ARR(src1_size, src2_size, src3_size), \
   ARR(nir_type_unsigned, nir_type_unsigned, nir_type_unsigned), 0)

/* fma(a, b, c) = (a * b) + c */
TRIOP(ffma, nir_type_float)

TRIOP(flrp, nir_type_float)

/**
 * \name Conditional Select
 *
 * A vector conditional select instruction (like ?:, but operating per-
 * component on vectors). There are two versions, one for floating point
 * bools (0.0 vs 1.0) and one for integer bools (0 vs ~0).
 */

TRIOP(fcsel, nir_type_float)
OPCODE(bcsel, 3, 0, nir_type_unsigned, ARR(0, 0, 0),
       ARR(nir_type_bool, nir_type_unsigned, nir_type_unsigned), 0)

TRIOP(bfi, nir_type_unsigned)

TRIOP(ubitfield_extract, nir_type_unsigned)
OPCODE(ibitfield_extract, 3, 0, nir_type_int, ARR(0, 0, 0),
       ARR(nir_type_int, nir_type_unsigned, nir_type_unsigned), 0)

/**
 * Combines the first component of each input to make a 3-component vector.
 */
TRIOP_HORIZ(vec3, 3, 1, 1, 1)

#define QUADOP(name) \
   OPCODE(name, 4, 0, nir_type_unsigned, ARR(0, 0, 0, 0), \
          ARR(nir_type_unsigned, nir_type_unsigned, nir_type_unsigned, nir_type_unsigned), \
          0)
#define QUADOP_HORIZ(name, output_size, src1_size, src2_size, src3_size, \
                     src4_size) \
   OPCODE(name, 4, output_size, nir_type_unsigned, \
          ARR(src1_size, src2_size, src3_size, src4_size), \
          ARR(nir_type_unsigned, nir_type_unsigned, nir_type_unsigned, nir_type_unsigned), \
          0)

QUADOP(bitfield_insert)

QUADOP_HORIZ(vec4, 4, 1, 1, 1, 1)

LAST_OPCODE(vec4)