1from __future__ import print_function 2import re 3import sys 4 5from . import common 6 7if sys.version_info[0] > 2: 8 9 class string: 10 expandtabs = str.expandtabs 11 12else: 13 import string 14 15# RegEx: this is where the magic happens. 16 17##### Assembly parser 18# 19# The set of per-arch regular expressions define several groups. 20# The required groups are "func" (function name) and "body" (body of the function). 21# Although some backends require some additional groups like: "directives" 22# and "func_name_separator" 23 24ASM_FUNCTION_X86_RE = re.compile( 25 r'^_?(?P<func>[^:]+):[ \t]*#+[ \t]*(@"?(?P=func)"?| -- Begin function (?P=func))\n(?:\s*\.?Lfunc_begin[^:\n]*:\n)?' 26 r"(?:\.L(?P=func)\$local:\n)?" # drop .L<func>$local: 27 r"(?:\s*\.type\s+\.L(?P=func)\$local,@function\n)?" # drop .type .L<func>$local 28 r"(?:[ \t]*(?:\.cfi_startproc|\.cfi_personality|\.cfi_lsda|\.seh_proc|\.seh_handler)\b[^\n]*\n)*" # drop optional cfi 29 r"(?P<body>^##?[ \t]+[^:]+:.*?)\s*" 30 r"^\s*(?:[^:\n]+?:\s*\n\s*\.size|\.cfi_endproc|\.globl|\.comm|\.(?:sub)?section|#+ -- End function)", 31 flags=(re.M | re.S), 32) 33 34ASM_FUNCTION_ARM_RE = re.compile( 35 r"^(?P<func>[0-9a-zA-Z_$]+):\n" # f: (name of function) 36 r"(?:\.L(?P=func)\$local:\n)?" # drop .L<func>$local: 37 r"(?:\s*\.type\s+\.L(?P=func)\$local,@function\n)?" # drop .type .L<func>$local 38 r"\s+\.fnstart\n" # .fnstart 39 r"(?P<body>.*?)" # (body of the function) 40 r"^.Lfunc_end[0-9]+:", # .Lfunc_end0: or # -- End function 41 flags=(re.M | re.S), 42) 43 44ASM_FUNCTION_AARCH64_RE = re.compile( 45 r'^_?(?P<func>[^:]+):[ \t]*\/\/[ \t]*@"?(?P=func)"?( (Function|Tail Call))?\n' 46 r"(?:[ \t]+.cfi_startproc\n)?" # drop optional cfi noise 47 r"(?P<body>.*?)\n" 48 # This list is incomplete 49 r"^\s*(\.Lfunc_end[0-9]+|// -- End function)", 50 flags=(re.M | re.S), 51) 52 53ASM_FUNCTION_AMDGPU_RE = re.compile( 54 r"\.type\s+_?(?P<func>[^,\n]+),@function\n" 55 r'^_?(?P=func):(?:[ \t]*;+[ \t]*@"?(?P=func)"?)?\n' 56 r"(?P<body>.*?)\n" # (body of the function) 57 # This list is incomplete 58 r"^\s*(\.Lfunc_end[0-9]+:\n|\.section)", 59 flags=(re.M | re.S), 60) 61 62ASM_FUNCTION_BPF_RE = re.compile( 63 r'^_?(?P<func>[^:]+):[ \t]*#+[ \t]*@"?(?P=func)"?\n' 64 r"(?:[ \t]+.cfi_startproc\n|.seh_proc[^\n]+\n)?" # drop optional cfi 65 r"(?P<body>.*?)\s*" 66 r".Lfunc_end[0-9]+:\n", 67 flags=(re.M | re.S), 68) 69 70ASM_FUNCTION_HEXAGON_RE = re.compile( 71 r'^_?(?P<func>[^:]+):[ \t]*//[ \t]*@"?(?P=func)"?\n[^:]*?' 72 r"(?P<body>.*?)\n" # (body of the function) 73 # This list is incomplete 74 r".Lfunc_end[0-9]+:\n", 75 flags=(re.M | re.S), 76) 77 78ASM_FUNCTION_M68K_RE = re.compile( 79 r'^_?(?P<func>[^:]+):[ \t]*;[ \t]*@"?(?P=func)"?\n' 80 r'(?:\.L(?P=func)\$local:\n)?' # drop .L<func>$local: 81 r'(?:[ \t]+\.type[ \t]+\.L(?P=func)\$local,@function\n)?' # drop .type .L<func>$local,@function 82 r'(?P<body>.*?)\s*' # (body of the function) 83 r'.Lfunc_end[0-9]+:\n', 84 flags=(re.M | re.S)) 85 86ASM_FUNCTION_MIPS_RE = re.compile( 87 r'^_?(?P<func>[^:]+):[ \t]*#+[ \t]*@"?(?P=func)"?\n[^:]*?' # f: (name of func) 88 r"(?:\s*\.?Ltmp[^:\n]*:\n)?[^:]*?" # optional .Ltmp<N> for EH 89 r"(?:^[ \t]+\.(frame|f?mask|set).*?\n)+" # Mips+LLVM standard asm prologue 90 r"(?P<body>.*?)\n" # (body of the function) 91 # Mips+LLVM standard asm epilogue 92 r"(?:(^[ \t]+\.set[^\n]*?\n)*^[ \t]+\.end.*?\n)" 93 r"(\$|\.L)func_end[0-9]+:\n", # $func_end0: (mips32 - O32) or 94 # .Lfunc_end0: (mips64 - NewABI) 95 flags=(re.M | re.S), 96) 97 98ASM_FUNCTION_MSP430_RE = re.compile( 99 r'^_?(?P<func>[^:]+):[ \t]*;+[ \t]*@"?(?P=func)"?\n[^:]*?' 100 r"(?P<body>.*?)\n" 101 r"(\$|\.L)func_end[0-9]+:\n", # $func_end0: 102 flags=(re.M | re.S), 103) 104 105ASM_FUNCTION_AVR_RE = re.compile( 106 r'^_?(?P<func>[^:]+):[ \t]*;+[ \t]*@"?(?P=func)"?\n[^:]*?' 107 r"(?P<body>.*?)\n" 108 r".Lfunc_end[0-9]+:\n", 109 flags=(re.M | re.S), 110) 111 112ASM_FUNCTION_PPC_RE = re.compile( 113 r"#[ \-\t]*Begin function (?P<func>[^.:]+)\n" 114 r".*?" 115 r'^[_.]?(?P=func):(?:[ \t]*#+[ \t]*@"?(?P=func)"?)?\n' 116 r"(?:^[^#]*\n)*" 117 r"(?P<body>.*?)\n" 118 # This list is incomplete 119 r"(?:^[ \t]*(?:\.(?:long|quad|v?byte)[ \t]+[^\n]+)\n)*" 120 r"(?:\.Lfunc_end|L\.\.(?P=func))[0-9]+:\n", 121 flags=(re.M | re.S), 122) 123 124ASM_FUNCTION_RISCV_RE = re.compile( 125 r'^_?(?P<func>[^:]+):[ \t]*#+[ \t]*@"?(?P=func)"?\n' 126 r"(?:\s*\.?L(?P=func)\$local:\n)?" # optional .L<func>$local: due to -fno-semantic-interposition 127 r"(?:\s*\.type\s+\.?L(?P=func)\$local,@function\n)?" # optional .type .L<func>$local 128 r"(?:\s*\.?Lfunc_begin[^:\n]*:\n)?[^:]*?" 129 r"(?P<body>^##?[ \t]+[^:]+:.*?)\s*" 130 r".Lfunc_end[0-9]+:\n", 131 flags=(re.M | re.S), 132) 133 134ASM_FUNCTION_LANAI_RE = re.compile( 135 r'^_?(?P<func>[^:]+):[ \t]*!+[ \t]*@"?(?P=func)"?\n' 136 r"(?:[ \t]+.cfi_startproc\n)?" # drop optional cfi noise 137 r"(?P<body>.*?)\s*" 138 r".Lfunc_end[0-9]+:\n", 139 flags=(re.M | re.S), 140) 141 142ASM_FUNCTION_SPARC_RE = re.compile( 143 r'^_?(?P<func>[^:]+):[ \t]*!+[ \t]*@"?(?P=func)"?\n' 144 r"(?P<body>.*?)\s*" 145 r".Lfunc_end[0-9]+:\n", 146 flags=(re.M | re.S), 147) 148 149ASM_FUNCTION_SYSTEMZ_RE = re.compile( 150 r'^_?(?P<func>[^:]+):[ \t]*#+[ \t]*@"?(?P=func)"?\n' 151 r"(?:[ \t]+.cfi_startproc\n)?" 152 r"(?P<body>.*?)\n" 153 r".Lfunc_end[0-9]+:\n", 154 flags=(re.M | re.S), 155) 156 157ASM_FUNCTION_AARCH64_DARWIN_RE = re.compile( 158 r'^_(?P<func>[^:]+):[ \t]*;[ \t]@"?(?P=func)"?\n' 159 r"([ \t]*.cfi_startproc\n[\s]*)?" 160 r"(?P<body>.*?)" 161 r"([ \t]*.cfi_endproc\n[\s]*)?" 162 r"^[ \t]*;[ \t]--[ \t]End[ \t]function", 163 flags=(re.M | re.S), 164) 165 166ASM_FUNCTION_ARM_DARWIN_RE = re.compile( 167 r"@[ \t]--[ \t]Begin[ \t]function[ \t](?P<func>[^ \t]+?)\n" 168 r"^[ \t]*\.globl[ \t]*_(?P=func)[ \t]*" 169 r"(?P<directives>.*?)" 170 r"^_(?P=func):\n[ \t]*" 171 r"(?P<body>.*?)" 172 r"^[ \t]*@[ \t]--[ \t]End[ \t]function", 173 flags=(re.M | re.S), 174) 175 176ASM_FUNCTION_ARM_MACHO_RE = re.compile( 177 r"^_(?P<func>[^:]+):[ \t]*\n" 178 r"([ \t]*.cfi_startproc\n[ \t]*)?" 179 r"(?P<body>.*?)\n" 180 r"[ \t]*\.cfi_endproc\n", 181 flags=(re.M | re.S), 182) 183 184ASM_FUNCTION_THUMBS_DARWIN_RE = re.compile( 185 r"^_(?P<func>[^:]+):\n" r"(?P<body>.*?)\n" r"[ \t]*\.data_region\n", 186 flags=(re.M | re.S), 187) 188 189ASM_FUNCTION_THUMB_DARWIN_RE = re.compile( 190 r"^_(?P<func>[^:]+):\n" r"(?P<body>.*?)\n" r"^[ \t]*@[ \t]--[ \t]End[ \t]function", 191 flags=(re.M | re.S), 192) 193 194ASM_FUNCTION_ARM_IOS_RE = re.compile( 195 r"^_(?P<func>[^:]+):\n" r"(?P<body>.*?)" r"^[ \t]*@[ \t]--[ \t]End[ \t]function", 196 flags=(re.M | re.S), 197) 198 199ASM_FUNCTION_WASM_RE = re.compile( 200 r'^_?(?P<func>[^:]+):[ \t]*#+[ \t]*@"?(?P=func)"?\n' 201 r"(?P<body>.*?)\n" 202 r"^\s*(\.Lfunc_end[0-9]+:\n|end_function)", 203 flags=(re.M | re.S), 204) 205 206# We parse the function name from OpName, and grab the variable name 'var' 207# for this function. Then we match that when the variable is assigned with 208# OpFunction and match its body. 209ASM_FUNCTION_SPIRV_RE = re.compile( 210 r'OpName (?P<var>%[0-9]+) "(?P<func>[^"]+)(?P<func_name_separator>)".*(?P<body>(?P=var) = OpFunction.+?OpFunctionEnd)', 211 flags=(re.M | re.S), 212) 213 214ASM_FUNCTION_VE_RE = re.compile( 215 r"^_?(?P<func>[^:]+):[ \t]*#+[ \t]*@(?P=func)\n" 216 r"(?:\s*\.?L(?P=func)\$local:\n)?" # optional .L<func>$local: due to -fno-semantic-interposition 217 r"(?:\s*\.type\s+\.?L(?P=func)\$local,@function\n)?" # optional .type .L<func>$local 218 r"(?:\s*\.?Lfunc_begin[^:\n]*:\n)?[^:]*?" 219 r"(?P<body>^##?[ \t]+[^:]+:.*?)\s*" 220 r".Lfunc_end[0-9]+:\n", 221 flags=(re.M | re.S), 222) 223 224ASM_FUNCTION_CSKY_RE = re.compile( 225 r"^_?(?P<func>[^:]+):[ \t]*#+[ \t]*@(?P=func)\n(?:\s*\.?Lfunc_begin[^:\n]*:\n)?[^:]*?" 226 r"(?P<body>^##?[ \t]+[^:]+:.*?)\s*" 227 r".Lfunc_end[0-9]+:\n", 228 flags=(re.M | re.S), 229) 230 231ASM_FUNCTION_NVPTX_RE = re.compile( 232 # function attributes and retval 233 # .visible .func (.param .align 16 .b8 func_retval0[32]) 234 # r'^(\.visible\s+)?\.func\s+(\([^\)]*\)\s*)?' 235 r"^(\.(func|visible|weak|entry|noreturn|extern)\s+)+(\([^\)]*\)\s*)?" 236 # function name 237 r"(?P<func>[^\(\n]+)" 238 # function name separator (opening brace) 239 r"(?P<func_name_separator>\()" 240 # function parameters 241 # ( 242 # .param .align 16 .b8 callee_St8x4_param_0[32] 243 # ) // -- Begin function callee_St8x4 244 r"[^\)]*\)(\s*//[^\n]*)?\n" 245 # function body 246 r"(?P<body>.*?)\n" 247 # function body end marker 248 r"\s*// -- End function", 249 flags=(re.M | re.S), 250) 251 252ASM_FUNCTION_LOONGARCH_RE = re.compile( 253 r'^_?(?P<func>[^:]+):[ \t]*#+[ \t]*@"?(?P=func)"?\n' 254 r"(?:\s*\.?Lfunc_begin[^:\n]*:\n)?[^:]*?" 255 r"(?P<body>^##?[ \t]+[^:]+:.*?)\s*" 256 r".Lfunc_end[0-9]+:\n", 257 flags=(re.M | re.S), 258) 259 260SCRUB_X86_SHUFFLES_RE = re.compile( 261 r"^(\s*\w+) [^#\n]+#+ ((?:[xyz]mm\d+|mem)( \{%k\d+\}( \{z\})?)? = .*)$", flags=re.M 262) 263 264SCRUB_X86_SHUFFLES_NO_MEM_RE = re.compile( 265 r"^(\s*\w+) [^#\n]+#+ ((?:[xyz]mm\d+|mem)( \{%k\d+\}( \{z\})?)? = (?!.*(?:mem)).*)$", 266 flags=re.M, 267) 268 269SCRUB_X86_SPILL_RELOAD_RE = re.compile( 270 r"-?\d+\(%([er])[sb]p\)(.*(?:Spill|Reload))$", flags=re.M 271) 272SCRUB_X86_SP_RE = re.compile(r"\d+\(%(esp|rsp)\)") 273SCRUB_X86_RIP_RE = re.compile(r"[.\w]+\(%rip\)") 274SCRUB_X86_LCP_RE = re.compile(r"\.?LCPI[0-9]+_[0-9]+") 275SCRUB_X86_RET_RE = re.compile(r"ret[l|q]") 276 277 278def scrub_asm_x86(asm, args): 279 # Scrub runs of whitespace out of the assembly, but leave the leading 280 # whitespace in place. 281 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 282 # Expand the tabs used for indentation. 283 asm = string.expandtabs(asm, 2) 284 285 # Detect shuffle asm comments and hide the operands in favor of the comments. 286 if getattr(args, "no_x86_scrub_mem_shuffle", True): 287 asm = SCRUB_X86_SHUFFLES_NO_MEM_RE.sub(r"\1 {{.*#+}} \2", asm) 288 else: 289 asm = SCRUB_X86_SHUFFLES_RE.sub(r"\1 {{.*#+}} \2", asm) 290 291 # Detect stack spills and reloads and hide their exact offset and whether 292 # they used the stack pointer or frame pointer. 293 asm = SCRUB_X86_SPILL_RELOAD_RE.sub(r"{{[-0-9]+}}(%\1{{[sb]}}p)\2", asm) 294 if getattr(args, "x86_scrub_sp", True): 295 # Generically match the stack offset of a memory operand. 296 asm = SCRUB_X86_SP_RE.sub(r"{{[0-9]+}}(%\1)", asm) 297 if getattr(args, "x86_scrub_rip", False): 298 # Generically match a RIP-relative memory operand. 299 asm = SCRUB_X86_RIP_RE.sub(r"{{.*}}(%rip)", asm) 300 # Generically match a LCP symbol. 301 asm = SCRUB_X86_LCP_RE.sub(r"{{\.?LCPI[0-9]+_[0-9]+}}", asm) 302 if getattr(args, "extra_scrub", False): 303 # Avoid generating different checks for 32- and 64-bit because of 'retl' vs 'retq'. 304 asm = SCRUB_X86_RET_RE.sub(r"ret{{[l|q]}}", asm) 305 # Strip kill operands inserted into the asm. 306 asm = common.SCRUB_KILL_COMMENT_RE.sub("", asm) 307 # Strip trailing whitespace. 308 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 309 return asm 310 311 312def scrub_asm_amdgpu(asm, args): 313 # Scrub runs of whitespace out of the assembly, but leave the leading 314 # whitespace in place. 315 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 316 # Expand the tabs used for indentation. 317 asm = string.expandtabs(asm, 2) 318 # Strip trailing whitespace. 319 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 320 return asm 321 322 323def scrub_asm_arm_eabi(asm, args): 324 # Scrub runs of whitespace out of the assembly, but leave the leading 325 # whitespace in place. 326 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 327 # Expand the tabs used for indentation. 328 asm = string.expandtabs(asm, 2) 329 # Strip kill operands inserted into the asm. 330 asm = common.SCRUB_KILL_COMMENT_RE.sub("", asm) 331 # Strip trailing whitespace. 332 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 333 return asm 334 335 336def scrub_asm_bpf(asm, args): 337 # Scrub runs of whitespace out of the assembly, but leave the leading 338 # whitespace in place. 339 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 340 # Expand the tabs used for indentation. 341 asm = string.expandtabs(asm, 2) 342 # Strip trailing whitespace. 343 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 344 return asm 345 346 347def scrub_asm_hexagon(asm, args): 348 # Scrub runs of whitespace out of the assembly, but leave the leading 349 # whitespace in place. 350 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 351 # Expand the tabs used for indentation. 352 asm = string.expandtabs(asm, 2) 353 # Strip trailing whitespace. 354 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 355 return asm 356 357 358def scrub_asm_powerpc(asm, args): 359 # Scrub runs of whitespace out of the assembly, but leave the leading 360 # whitespace in place. 361 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 362 # Expand the tabs used for indentation. 363 asm = string.expandtabs(asm, 2) 364 # Strip unimportant comments, but leave the token '#' in place. 365 asm = common.SCRUB_LOOP_COMMENT_RE.sub(r"#", asm) 366 # Strip trailing whitespace. 367 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 368 # Strip the tailing token '#', except the line only has token '#'. 369 asm = common.SCRUB_TAILING_COMMENT_TOKEN_RE.sub(r"", asm) 370 return asm 371 372 373def scrub_asm_m68k(asm, args): 374 # Scrub runs of whitespace out of the assembly, but leave the leading 375 # whitespace in place. 376 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 377 # Expand the tabs used for indentation. 378 asm = string.expandtabs(asm, 2) 379 # Strip trailing whitespace. 380 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 381 return asm 382 383 384def scrub_asm_mips(asm, args): 385 # Scrub runs of whitespace out of the assembly, but leave the leading 386 # whitespace in place. 387 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 388 # Expand the tabs used for indentation. 389 asm = string.expandtabs(asm, 2) 390 # Strip trailing whitespace. 391 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 392 return asm 393 394 395def scrub_asm_msp430(asm, args): 396 # Scrub runs of whitespace out of the assembly, but leave the leading 397 # whitespace in place. 398 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 399 # Expand the tabs used for indentation. 400 asm = string.expandtabs(asm, 2) 401 # Strip trailing whitespace. 402 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 403 return asm 404 405 406def scrub_asm_avr(asm, args): 407 # Scrub runs of whitespace out of the assembly, but leave the leading 408 # whitespace in place. 409 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 410 # Expand the tabs used for indentation. 411 asm = string.expandtabs(asm, 2) 412 # Strip trailing whitespace. 413 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 414 return asm 415 416 417def scrub_asm_riscv(asm, args): 418 # Scrub runs of whitespace out of the assembly, but leave the leading 419 # whitespace in place. 420 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 421 # Expand the tabs used for indentation. 422 asm = string.expandtabs(asm, 2) 423 # Strip trailing whitespace. 424 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 425 return asm 426 427 428def scrub_asm_lanai(asm, args): 429 # Scrub runs of whitespace out of the assembly, but leave the leading 430 # whitespace in place. 431 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 432 # Expand the tabs used for indentation. 433 asm = string.expandtabs(asm, 2) 434 # Strip trailing whitespace. 435 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 436 return asm 437 438 439def scrub_asm_sparc(asm, args): 440 # Scrub runs of whitespace out of the assembly, but leave the leading 441 # whitespace in place. 442 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 443 # Expand the tabs used for indentation. 444 asm = string.expandtabs(asm, 2) 445 # Strip trailing whitespace. 446 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 447 return asm 448 449 450def scrub_asm_spirv(asm, args): 451 # Scrub runs of whitespace out of the assembly, but leave the leading 452 # whitespace in place. 453 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 454 # Expand the tabs used for indentation. 455 asm = string.expandtabs(asm, 2) 456 # Strip trailing whitespace. 457 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 458 return asm 459 460 461def scrub_asm_systemz(asm, args): 462 # Scrub runs of whitespace out of the assembly, but leave the leading 463 # whitespace in place. 464 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 465 # Expand the tabs used for indentation. 466 asm = string.expandtabs(asm, 2) 467 # Strip trailing whitespace. 468 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 469 return asm 470 471 472def scrub_asm_wasm(asm, args): 473 # Scrub runs of whitespace out of the assembly, but leave the leading 474 # whitespace in place. 475 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 476 # Expand the tabs used for indentation. 477 asm = string.expandtabs(asm, 2) 478 # Strip trailing whitespace. 479 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 480 return asm 481 482 483def scrub_asm_ve(asm, args): 484 # Scrub runs of whitespace out of the assembly, but leave the leading 485 # whitespace in place. 486 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 487 # Expand the tabs used for indentation. 488 asm = string.expandtabs(asm, 2) 489 # Strip trailing whitespace. 490 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 491 return asm 492 493 494def scrub_asm_csky(asm, args): 495 # Scrub runs of whitespace out of the assembly, but leave the leading 496 # whitespace in place. 497 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 498 # Expand the tabs used for indentation. 499 asm = string.expandtabs(asm, 2) 500 # Strip kill operands inserted into the asm. 501 asm = common.SCRUB_KILL_COMMENT_RE.sub("", asm) 502 # Strip trailing whitespace. 503 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 504 return asm 505 506 507def scrub_asm_nvptx(asm, args): 508 # Scrub runs of whitespace out of the assembly, but leave the leading 509 # whitespace in place. 510 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 511 # Expand the tabs used for indentation. 512 asm = string.expandtabs(asm, 2) 513 # Strip trailing whitespace. 514 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 515 return asm 516 517 518def scrub_asm_loongarch(asm, args): 519 # Scrub runs of whitespace out of the assembly, but leave the leading 520 # whitespace in place. 521 asm = common.SCRUB_WHITESPACE_RE.sub(r" ", asm) 522 # Expand the tabs used for indentation. 523 asm = string.expandtabs(asm, 2) 524 # Strip trailing whitespace. 525 asm = common.SCRUB_TRAILING_WHITESPACE_RE.sub(r"", asm) 526 return asm 527 528 529# Returns a tuple of a scrub function and a function regex. Scrub function is 530# used to alter function body in some way, for example, remove trailing spaces. 531# Function regex is used to match function name, body, etc. in raw llc output. 532def get_run_handler(triple): 533 target_handlers = { 534 "i686": (scrub_asm_x86, ASM_FUNCTION_X86_RE), 535 "x86": (scrub_asm_x86, ASM_FUNCTION_X86_RE), 536 "i386": (scrub_asm_x86, ASM_FUNCTION_X86_RE), 537 "arm64_32": (scrub_asm_arm_eabi, ASM_FUNCTION_AARCH64_DARWIN_RE), 538 "arm64_32-apple-ios": (scrub_asm_arm_eabi, ASM_FUNCTION_AARCH64_DARWIN_RE), 539 "arm64_32-apple-watchos": ( 540 scrub_asm_arm_eabi, 541 ASM_FUNCTION_AARCH64_DARWIN_RE, 542 ), 543 "aarch64": (scrub_asm_arm_eabi, ASM_FUNCTION_AARCH64_RE), 544 "aarch64-apple-darwin": (scrub_asm_arm_eabi, ASM_FUNCTION_AARCH64_DARWIN_RE), 545 "aarch64-apple-ios": (scrub_asm_arm_eabi, ASM_FUNCTION_AARCH64_DARWIN_RE), 546 "bpf": (scrub_asm_bpf, ASM_FUNCTION_BPF_RE), 547 "bpfel": (scrub_asm_bpf, ASM_FUNCTION_BPF_RE), 548 "bpfeb": (scrub_asm_bpf, ASM_FUNCTION_BPF_RE), 549 "hexagon": (scrub_asm_hexagon, ASM_FUNCTION_HEXAGON_RE), 550 "r600": (scrub_asm_amdgpu, ASM_FUNCTION_AMDGPU_RE), 551 "amdgcn": (scrub_asm_amdgpu, ASM_FUNCTION_AMDGPU_RE), 552 "arm": (scrub_asm_arm_eabi, ASM_FUNCTION_ARM_RE), 553 "arm64": (scrub_asm_arm_eabi, ASM_FUNCTION_AARCH64_RE), 554 "arm64e": (scrub_asm_arm_eabi, ASM_FUNCTION_AARCH64_DARWIN_RE), 555 "arm64ec": (scrub_asm_arm_eabi, ASM_FUNCTION_AARCH64_RE), 556 "arm64-apple-ios": (scrub_asm_arm_eabi, ASM_FUNCTION_AARCH64_DARWIN_RE), 557 "arm64-apple-macosx": (scrub_asm_arm_eabi, ASM_FUNCTION_AARCH64_DARWIN_RE), 558 "armv7-apple-ios": (scrub_asm_arm_eabi, ASM_FUNCTION_ARM_IOS_RE), 559 "armv7-apple-darwin": (scrub_asm_arm_eabi, ASM_FUNCTION_ARM_DARWIN_RE), 560 "thumb": (scrub_asm_arm_eabi, ASM_FUNCTION_ARM_RE), 561 "thumb-macho": (scrub_asm_arm_eabi, ASM_FUNCTION_ARM_MACHO_RE), 562 "thumbv5-macho": (scrub_asm_arm_eabi, ASM_FUNCTION_ARM_MACHO_RE), 563 "thumbv7s-apple-darwin": (scrub_asm_arm_eabi, ASM_FUNCTION_THUMBS_DARWIN_RE), 564 "thumbv7-apple-darwin": (scrub_asm_arm_eabi, ASM_FUNCTION_THUMB_DARWIN_RE), 565 "thumbv7-apple-ios": (scrub_asm_arm_eabi, ASM_FUNCTION_ARM_IOS_RE), 566 "m68k": (scrub_asm_m68k, ASM_FUNCTION_M68K_RE), 567 "mips": (scrub_asm_mips, ASM_FUNCTION_MIPS_RE), 568 "msp430": (scrub_asm_msp430, ASM_FUNCTION_MSP430_RE), 569 "avr": (scrub_asm_avr, ASM_FUNCTION_AVR_RE), 570 "ppc32": (scrub_asm_powerpc, ASM_FUNCTION_PPC_RE), 571 "ppc64": (scrub_asm_powerpc, ASM_FUNCTION_PPC_RE), 572 "powerpc": (scrub_asm_powerpc, ASM_FUNCTION_PPC_RE), 573 "riscv32": (scrub_asm_riscv, ASM_FUNCTION_RISCV_RE), 574 "riscv64": (scrub_asm_riscv, ASM_FUNCTION_RISCV_RE), 575 "lanai": (scrub_asm_lanai, ASM_FUNCTION_LANAI_RE), 576 "sparc": (scrub_asm_sparc, ASM_FUNCTION_SPARC_RE), 577 "spirv32": (scrub_asm_spirv, ASM_FUNCTION_SPIRV_RE), 578 "spirv64": (scrub_asm_spirv, ASM_FUNCTION_SPIRV_RE), 579 "s390x": (scrub_asm_systemz, ASM_FUNCTION_SYSTEMZ_RE), 580 "wasm32": (scrub_asm_wasm, ASM_FUNCTION_WASM_RE), 581 "wasm64": (scrub_asm_wasm, ASM_FUNCTION_WASM_RE), 582 "ve": (scrub_asm_ve, ASM_FUNCTION_VE_RE), 583 "csky": (scrub_asm_csky, ASM_FUNCTION_CSKY_RE), 584 "nvptx": (scrub_asm_nvptx, ASM_FUNCTION_NVPTX_RE), 585 "loongarch32": (scrub_asm_loongarch, ASM_FUNCTION_LOONGARCH_RE), 586 "loongarch64": (scrub_asm_loongarch, ASM_FUNCTION_LOONGARCH_RE), 587 } 588 handler = None 589 best_prefix = "" 590 for prefix, s in target_handlers.items(): 591 if triple.startswith(prefix) and len(prefix) > len(best_prefix): 592 handler = s 593 best_prefix = prefix 594 595 if handler is None: 596 raise KeyError("Triple %r is not supported" % (triple)) 597 598 return handler 599 600 601##### Generator of assembly CHECK lines 602 603 604def add_checks( 605 output_lines, 606 comment_marker, 607 prefix_list, 608 func_dict, 609 func_name, 610 ginfo: common.GeneralizerInfo, 611 global_vars_seen_dict, 612 is_filtered, 613): 614 # Label format is based on ASM string. 615 check_label_format = "{} %s-LABEL: %s%s%s%s".format(comment_marker) 616 return common.add_checks( 617 output_lines, 618 comment_marker, 619 prefix_list, 620 func_dict, 621 func_name, 622 check_label_format, 623 ginfo, 624 global_vars_seen_dict, 625 is_filtered=is_filtered, 626 ) 627