test/MC/AMDGPU/vop3-modifiers.s

   1 // RUN: llvm-mc -arch=amdgcn -mcpu=tonga -show-encoding %s | FileCheck %s
   2
   3 //---------------------------------------------------------------------------//
   4 // VOP1/VOP3 F16
   5 //---------------------------------------------------------------------------//
   6
   7 v_ceil_f16 v0, -1
   8 // CHECK: [0xc1,0x8a,0x00,0x7e]
   9
  10 v_ceil_f16 v0, -2
  11 // CHECK: [0xc2,0x8a,0x00,0x7e]
  12
  13 v_ceil_f16 v0, -16
  14 // CHECK: [0xd0,0x8a,0x00,0x7e]
  15
  16 v_ceil_f16 v0, -0.5
  17 // CHECK: [0xf1,0x8a,0x00,0x7e]
  18
  19 v_ceil_f16 v0, -1.0
  20 // CHECK: [0xf3,0x8a,0x00,0x7e]
  21
  22 v_ceil_f16 v0, -2.0
  23 // CHECK: [0xf5,0x8a,0x00,0x7e]
  24
  25 v_ceil_f16 v0, -4.0
  26 // CHECK: [0xf7,0x8a,0x00,0x7e]
  27
  28 // Arbitrary f16 literal in hex
  29 v_ceil_f16 v0, 0xabcd
  30 // CHECK: [0xff,0x8a,0x00,0x7e,0xcd,0xab,0x00,0x00]
  31
  32 // '-' is a part of hex literal (not a 'neg' modifier)
  33 v_ceil_f16 v0, -0x5433
  34 // CHECK: [0xff,0x8a,0x00,0x7e,0xcd,0xab,0x00,0x00]
  35
  36 v_ceil_f16 v0, abs(0xabcd)
  37 // CHECK: [0xff,0x8a,0x00,0x7e,0xcd,0x2b,0x00,0x00]
  38
  39 v_ceil_f16 v0, neg(0xabcd)
  40 // CHECK: [0xff,0x8a,0x00,0x7e,0xcd,0x2b,0x00,0x00]
  41
  42 v_ceil_f16 v0, neg(abs(0xabcd))
  43 // CHECK: [0xff,0x8a,0x00,0x7e,0xcd,0xab,0x00,0x00]
  44
  45 v_ceil_f16 v0, -abs(0xabcd)
  46 // CHECK: [0xff,0x8a,0x00,0x7e,0xcd,0xab,0x00,0x00]
  47
  48 // 1/(2*pi) encoded as inline constant in VOP1
  49 v_ceil_f16 v0, 0x3118
  50 // CHECK: [0xf8,0x8a,0x00,0x7e]
  51
  52 // 1/(2*pi) encoded as inline constant in VOP3
  53 v_ceil_f16_e64 v0, 0x3118
  54 // CHECK: [0x00,0x00,0x85,0xd1,0xf8,0x00,0x00,0x00]
  55
  56 // neg(-1/(2*pi)) = 1/(2*pi)
  57 v_ceil_f16 v0, neg(0xb118)
  58 // CHECK: [0xf8,0x8a,0x00,0x7e]
  59
  60 // -1/(2*pi) cannot be encoded as inline constant in VOP1
  61 v_ceil_f16 v0, 0xb118
  62 // CHECK: [0xff,0x8a,0x00,0x7e,0x18,0xb1,0x00,0x00]
  63
  64 // -1/(2*pi) cannot be encoded as inline constant in VOP1
  65 v_ceil_f16 v0, neg(0x3118)
  66 // CHECK: [0xff,0x8a,0x00,0x7e,0x18,0xb1,0x00,0x00]
  67
  68 // -1/(2*pi) can be encoded as inline constant w/ modifiers in VOP3
  69 v_ceil_f16_e64 v0, neg(0x3118)
  70 // CHECK: [0x00,0x00,0x85,0xd1,0xf8,0x00,0x00,0x20]
  71
  72 v_ceil_f16_e64 v0, abs(0x3118)
  73 // CHECK: 0x00,0x01,0x85,0xd1,0xf8,0x00,0x00,0x00]
  74
  75 v_ceil_f16_e64 v0, neg(abs(0x3118))
  76 // CHECK: [0x00,0x01,0x85,0xd1,0xf8,0x00,0x00,0x20]
  77
  78 v_ceil_f16_e64 v0, neg(|v1|)
  79 // CHECK: [0x00,0x01,0x85,0xd1,0x01,0x01,0x00,0x20]
  80
  81 v_ceil_f16_e64 v0, -|v1|
  82 // CHECK: [0x00,0x01,0x85,0xd1,0x01,0x01,0x00,0x20]
  83
  84 //---------------------------------------------------------------------------//
  85 // VOP1/VOP3 F64
  86 //---------------------------------------------------------------------------//
  87
  88 // Encoded as inline constant 1 with 'neg' modifier
  89 v_ceil_f64 v[0:1], neg(1)
  90 // CHECK: [0x00,0x00,0x58,0xd1,0x81,0x00,0x00,0x20]
  91
  92 // Encoded as inline constant -1 with 'neg' modifier
  93 v_ceil_f64 v[0:1], neg(-1)
  94 // CHECK: [0x00,0x00,0x58,0xd1,0xc1,0x00,0x00,0x20]
  95
  96 v_ceil_f64_e32 v[0:1], 1.0
  97 // CHECK: [0xf2,0x30,0x00,0x7e]
  98
  99 // abs(1.0) = 1.0
 100 v_ceil_f64_e32 v[0:1], abs(1.0)
 101 // CHECK: [0xf2,0x30,0x00,0x7e]
 102
 103 // neg(1.0) = -1.0
 104 v_ceil_f64_e32 v[0:1], neg(1.0)
 105 // CHECK: [0xf3,0x30,0x00,0x7e]
 106
 107 // 1/(2*pi) encoded as inline constant in VOP1
 108 v_ceil_f64 v[0:1], 0x3fc45f306dc9c882
 109 // CHECK: [0xf8,0x30,0x00,0x7e]
 110
 111 // 1/(2*pi) encoded as inline constant in VOP3
 112 v_ceil_f64_e64 v[0:1], 0x3fc45f306dc9c882
 113 // CHECK: [0x00,0x00,0x58,0xd1,0xf8,0x00,0x00,0x00]
 114
 115 // -1/(2*pi) cannot be encoded as inline constant in VOP1.
 116 // It cannot be encoded as literal either due to int literal rules.
 117 // So it is encoded as VOP3
 118 v_ceil_f64 v[0:1], abs(0x3fc45f306dc9c882)
 119 // CHECK: [0x00,0x01,0x58,0xd1,0xf8,0x00,0x00,0x00]
 120
 121 v_ceil_f64 v[0:1], neg(abs(0x3fc45f306dc9c882))
 122 // CHECK: [0x00,0x01,0x58,0xd1,0xf8,0x00,0x00,0x20]
 123
 124
 125 //---------------------------------------------------------------------------//
 126 // VOP2/VOP3 F32
 127 //---------------------------------------------------------------------------//
 128
 129 v_add_f32 v5, -1, v2
 130 // CHECK: [0xc1,0x04,0x0a,0x02]
 131
 132 v_add_f32 v5, -16, v2
 133 // CHECK: [0xd0,0x04,0x0a,0x02]
 134
 135 v_add_f32 v5, 0x3e22f983, v2
 136 // CHECK: [0xf8,0x04,0x0a,0x02]
 137
 138 // abs(1/(2*pi)) = 1/(2*pi)
 139 v_add_f32 v5, abs(0x3e22f983), v2
 140 // CHECK: [0xf8,0x04,0x0a,0x02]
 141
 142 // neg(-1/(2*pi)) = 1/(2*pi)
 143 v_add_f32 v5, neg(0xbe22f983), v2
 144 // CHECK: [0xf8,0x04,0x0a,0x02]
 145
 146 // -1/(2*pi) cannot be encoded as inline constant in VOP1
 147 v_add_f32 v5, neg(0x3e22f983), v2
 148 // CHECK: [0xff,0x04,0x0a,0x02,0x83,0xf9,0x22,0xbe]
 149
 150
 151 v_add_f32_e64 v0, -2, s0
 152 // CHECK: [0x00,0x00,0x01,0xd1,0xc2,0x00,0x00,0x00]
 153
 154 v_add_f32_e64 v0, -16, s0
 155 // CHECK: [0x00,0x00,0x01,0xd1,0xd0,0x00,0x00,0x00]
 156
 157 v_add_f32_e64 v0, -0.5, s0
 158 // CHECK: [0x00,0x00,0x01,0xd1,0xf1,0x00,0x00,0x00]
 159
 160 v_add_f32_e64 v0, -1.0, s0
 161 // CHECK: [0x00,0x00,0x01,0xd1,0xf3,0x00,0x00,0x00]
 162
 163 v_add_f32_e64 v0, -2.0, s0
 164 // CHECK: [0x00,0x00,0x01,0xd1,0xf5,0x00,0x00,0x00]
 165
 166 v_add_f32_e64 v0, -4.0, s0
 167 // CHECK: [0x00,0x00,0x01,0xd1,0xf7,0x00,0x00,0x00]
 168
 169 v_add_f32_e64 v0, 0x3e22f983, s0
 170 // CHECK: [0x00,0x00,0x01,0xd1,0xf8,0x00,0x00,0x00]
 171
 172 v_add_f32_e64 v0, neg(0x3e22f983), s0
 173 // CHECK: [0x00,0x00,0x01,0xd1,0xf8,0x00,0x00,0x20]
 174
 175 //---------------------------------------------------------------------------//
 176 // VOPC/VOP3
 177 //---------------------------------------------------------------------------//
 178
 179 v_cmp_eq_f16 vcc, -1, v0
 180 // CHECK: [0xc1,0x00,0x44,0x7c]
 181
 182 v_cmp_eq_f16_e64 s[0:1], s0, -1
 183 // CHECK: [0x00,0x00,0x22,0xd0,0x00,0x82,0x01,0x00]
 184
 185 v_cmp_eq_f16_e64 s[0:1], s0, 0x3118
 186 // CHECK: [0x00,0x00,0x22,0xd0,0x00,0xf0,0x01,0x00]
 187
 188 v_cmp_eq_f16_e64 s[0:1], s0, neg(0x3118)
 189 // CHECK: [0x00,0x00,0x22,0xd0,0x00,0xf0,0x01,0x40]
 190
 191 v_cmp_eq_f32 vcc, -4.0, v0
 192 // CHECK: [0xf7,0x00,0x84,0x7c]
 193
 194 // 1/(2*pi) can be encoded as inline constant
 195 v_cmp_eq_f32 vcc, 0x3e22f983, v0
 196 // CHECK: [0xf8,0x00,0x84,0x7c]
 197
 198 // -1/(2*pi) cannot be encoded as inline constant in VOPC
 199 v_cmp_eq_f32 vcc, neg(0x3e22f983), v0
 200 // CHECK: [0xff,0x00,0x84,0x7c,0x83,0xf9,0x22,0xbe]
 201
 202 // abs(1/(2*pi)) = 1/(2*pi)
 203 v_cmp_eq_f32 vcc, abs(0x3e22f983), v0
 204 // CHECK: [0xf8,0x00,0x84,0x7c]
 205
 206 // -1/(2*pi) can be encoded as inline constant w/ modifiers in VOP3
 207 v_cmp_eq_f32_e64 vcc, neg(0x3e22f983), v0
 208 // CHECK: [0x6a,0x00,0x42,0xd0,0xf8,0x00,0x02,0x20]
 209
 210 v_cmp_eq_f32_e64 vcc, v0, abs(0x3e22f983)
 211 // CHECK: [0x6a,0x02,0x42,0xd0,0x00,0xf1,0x01,0x00]
 212
 213 v_cmp_eq_f32_e64 vcc, v0, -abs(0x3e22f983)
 214 // CHECK: [0x6a,0x02,0x42,0xd0,0x00,0xf1,0x01,0x40]
 215
 216 //---------------------------------------------------------------------------//
 217 // VOP3
 218 //---------------------------------------------------------------------------//
 219
 220 v_add_f64 v[0:1], s[0:1], -1
 221 // CHECK: [0x00,0x00,0x80,0xd2,0x00,0x82,0x01,0x00]
 222
 223 v_add_f64 v[0:1], s[0:1], -16
 224 // CHECK: [0x00,0x00,0x80,0xd2,0x00,0xa0,0x01,0x00]
 225
 226 v_add_f64 v[0:1], s[0:1], -0.5
 227 // CHECK: [0x00,0x00,0x80,0xd2,0x00,0xe2,0x01,0x00]
 228
 229 v_add_f64 v[0:1], s[0:1], -1.0
 230 // CHECK: [0x00,0x00,0x80,0xd2,0x00,0xe6,0x01,0x00]
 231
 232 v_add_f64 v[0:1], s[0:1], -2.0
 233 // CHECK: [0x00,0x00,0x80,0xd2,0x00,0xea,0x01,0x00]
 234
 235 v_add_f64 v[0:1], s[0:1], -4.0
 236 // CHECK: [0x00,0x00,0x80,0xd2,0x00,0xee,0x01,0x00]
 237
 238 v_add_f64 v[4:5], s[0:1], 0x3fc45f306dc9c882
 239 // CHECK: [0x04,0x00,0x80,0xd2,0x00,0xf0,0x01,0x00]
 240
 241 v_add_f64 v[4:5], s[0:1], neg(0x3fc45f306dc9c882)
 242 // CHECK: [0x04,0x00,0x80,0xd2,0x00,0xf0,0x01,0x40]
 243
 244
 245 v_cubeid_f32 v0, s0, s0, -1
 246 // CHECK: [0x00,0x00,0xc4,0xd1,0x00,0x00,0x04,0x03]
 247
 248 v_cubeid_f32 v0, s0, s0, -4.0
 249 // CHECK: [0x00,0x00,0xc4,0xd1,0x00,0x00,0xdc,0x03]
 250
 251 v_cubeid_f32 v0, s0, s0, 0x3e22f983
 252 // CHECK: [0x00,0x00,0xc4,0xd1,0x00,0x00,0xe0,0x03]
 253
 254 v_cubeid_f32 v0, s0, s0, neg(0x3e22f983)
 255 // CHECK: [0x00,0x00,0xc4,0xd1,0x00,0x00,0xe0,0x83]
 256
 257 v_cubeid_f32 v0, s0, s0, abs(0x3e22f983)
 258 // CHECK: [0x00,0x04,0xc4,0xd1,0x00,0x00,0xe0,0x03]
 259
 260
 261 //---------------------------------------------------------------------------//
 262 // VOP3 Instructions without Input Modifiers but with Output Modifiers
 263 //---------------------------------------------------------------------------//
 264
 265 v_cvt_f64_i32_e64 v[5:6], s1 clamp
 266 // CHECK: [0x05,0x80,0x44,0xd1,0x01,0x00,0x00,0x00]
 267
 268 v_cvt_f64_i32_e64 v[5:6], s1 mul:2
 269 // CHECK: [0x05,0x00,0x44,0xd1,0x01,0x00,0x00,0x08]
 270
 271 v_cvt_f64_i32_e64 v[5:6], s1 mul:4
 272 // CHECK: [0x05,0x00,0x44,0xd1,0x01,0x00,0x00,0x10]
 273
 274 v_cvt_f64_i32_e64 v[5:6], s1 div:2
 275 // CHECK: [0x05,0x00,0x44,0xd1,0x01,0x00,0x00,0x18]
 276
 277
 278 v_cvt_f64_u32_e64 v[5:6], s1 clamp
 279 // CHECK: [0x05,0x80,0x56,0xd1,0x01,0x00,0x00,0x00]
 280
 281 v_cvt_f64_u32_e64 v[5:6], s1 mul:2
 282 // CHECK: [0x05,0x00,0x56,0xd1,0x01,0x00,0x00,0x08]
 283
 284 v_cvt_f64_u32_e64 v[5:6], s1 mul:4
 285 // CHECK: [0x05,0x00,0x56,0xd1,0x01,0x00,0x00,0x10]
 286
 287 v_cvt_f64_u32_e64 v[5:6], s1 div:2
 288 // CHECK: [0x05,0x00,0x56,0xd1,0x01,0x00,0x00,0x18]
 289
 290
 291 v_cvt_f32_i32_e64 v5, s1 clamp
 292 // CHECK: [0x05,0x80,0x45,0xd1,0x01,0x00,0x00,0x00]
 293
 294 v_cvt_f32_i32_e64 v5, s1 mul:2
 295 // CHECK: [0x05,0x00,0x45,0xd1,0x01,0x00,0x00,0x08]
 296
 297 v_cvt_f32_i32_e64 v5, s1 mul:4
 298 // CHECK: [0x05,0x00,0x45,0xd1,0x01,0x00,0x00,0x10]
 299
 300 v_cvt_f32_i32_e64 v5, s1 div:2
 301 // CHECK: [0x05,0x00,0x45,0xd1,0x01,0x00,0x00,0x18]
 302
 303
 304 v_cvt_f32_u32_e64 v5, s1 clamp
 305 // CHECK: [0x05,0x80,0x46,0xd1,0x01,0x00,0x00,0x00]
 306
 307 v_cvt_f32_u32_e64 v5, s1 mul:2
 308 // CHECK: [0x05,0x00,0x46,0xd1,0x01,0x00,0x00,0x08]
 309
 310 v_cvt_f32_u32_e64 v5, s1 mul:4
 311 // CHECK: [0x05,0x00,0x46,0xd1,0x01,0x00,0x00,0x10]
 312
 313 v_cvt_f32_u32_e64 v5, s1 div:2
 314 // CHECK: [0x05,0x00,0x46,0xd1,0x01,0x00,0x00,0x18]
 315
 316
 317 v_cvt_off_f32_i4_e64 v5, s1 clamp
 318 // CHECK: [0x05,0x80,0x4e,0xd1,0x01,0x00,0x00,0x00]
 319
 320 v_cvt_off_f32_i4_e64 v5, s1 mul:2
 321 // CHECK: [0x05,0x00,0x4e,0xd1,0x01,0x00,0x00,0x08]
 322
 323 v_cvt_off_f32_i4_e64 v5, s1 mul:4
 324 // CHECK: [0x05,0x00,0x4e,0xd1,0x01,0x00,0x00,0x10]
 325
 326 v_cvt_off_f32_i4_e64 v5, s1 div:2
 327 // CHECK: [0x05,0x00,0x4e,0xd1,0x01,0x00,0x00,0x18]
 328
 329
 330 v_cvt_f32_ubyte0_e64 v5, s1 clamp
 331 // CHECK: [0x05,0x80,0x51,0xd1,0x01,0x00,0x00,0x00]
 332
 333 v_cvt_f32_ubyte0_e64 v5, s1 mul:2
 334 // CHECK: [0x05,0x00,0x51,0xd1,0x01,0x00,0x00,0x08]
 335
 336 v_cvt_f32_ubyte0_e64 v5, s1 mul:4
 337 // CHECK: [0x05,0x00,0x51,0xd1,0x01,0x00,0x00,0x10]
 338
 339 v_cvt_f32_ubyte0_e64 v5, s1 div:2
 340 // CHECK: [0x05,0x00,0x51,0xd1,0x01,0x00,0x00,0x18]
 341
 342
 343 v_cvt_f32_ubyte1_e64 v5, s1 clamp
 344 // CHECK: [0x05,0x80,0x52,0xd1,0x01,0x00,0x00,0x00]
 345
 346 v_cvt_f32_ubyte1_e64 v5, s1 mul:2
 347 // CHECK: [0x05,0x00,0x52,0xd1,0x01,0x00,0x00,0x08]
 348
 349 v_cvt_f32_ubyte1_e64 v5, s1 mul:4
 350 // CHECK: [0x05,0x00,0x52,0xd1,0x01,0x00,0x00,0x10]
 351
 352 v_cvt_f32_ubyte1_e64 v5, s1 div:2
 353 // CHECK: [0x05,0x00,0x52,0xd1,0x01,0x00,0x00,0x18]
 354
 355
 356 v_cvt_f32_ubyte2_e64 v5, s1 clamp
 357 // CHECK: [0x05,0x80,0x53,0xd1,0x01,0x00,0x00,0x00]
 358
 359 v_cvt_f32_ubyte2_e64 v5, s1 mul:2
 360 // CHECK: [0x05,0x00,0x53,0xd1,0x01,0x00,0x00,0x08]
 361
 362 v_cvt_f32_ubyte2_e64 v5, s1 mul:4
 363 // CHECK: [0x05,0x00,0x53,0xd1,0x01,0x00,0x00,0x10]
 364
 365 v_cvt_f32_ubyte2_e64 v5, s1 div:2
 366 // CHECK: [0x05,0x00,0x53,0xd1,0x01,0x00,0x00,0x18]
 367
 368
 369 v_cvt_f32_ubyte3_e64 v5, s1 clamp
 370 // CHECK: [0x05,0x80,0x54,0xd1,0x01,0x00,0x00,0x00]
 371
 372 v_cvt_f32_ubyte3_e64 v5, s1 mul:2
 373 // CHECK: [0x05,0x00,0x54,0xd1,0x01,0x00,0x00,0x08]
 374
 375 v_cvt_f32_ubyte3_e64 v5, s1 mul:4
 376 // CHECK: [0x05,0x00,0x54,0xd1,0x01,0x00,0x00,0x10]
 377
 378 v_cvt_f32_ubyte3_e64 v5, s1 div:2
 379 // CHECK: [0x05,0x00,0x54,0xd1,0x01,0x00,0x00,0x18]
 380
 381
 382 // NB: output modifiers are not supported for f16
 383 v_cvt_f16_i16_e64 v5, s1 clamp
 384 // CHECK: [0x05,0x80,0x7a,0xd1,0x01,0x00,0x00,0x00]
 385
 386 // NB: output modifiers are not supported for f16
 387 v_cvt_f16_u16_e64 v5, s1 clamp
 388 // CHECK: [0x05,0x80,0x79,0xd1,0x01,0x00,0x00,0x00]