| 1 |
/* memcpy with SSSE3 |
| 2 |
Copyright (C) 2010-2026 Free Software Foundation, Inc. |
| 3 |
This file is part of the GNU C Library. |
| 4 |
|
| 5 |
The GNU C Library is free software; you can redistribute it and/or |
| 6 |
modify it under the terms of the GNU Lesser General Public |
| 7 |
License as published by the Free Software Foundation; either |
| 8 |
version 2.1 of the License, or (at your option) any later version. |
| 9 |
|
| 10 |
The GNU C Library is distributed in the hope that it will be useful, |
| 11 |
but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 12 |
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 13 |
Lesser General Public License for more details. |
| 14 |
|
| 15 |
You should have received a copy of the GNU Lesser General Public |
| 16 |
License along with the GNU C Library; if not, see |
| 17 |
<https://www.gnu.org/licenses/>. */ |
| 18 |
|
| 19 |
#if IS_IN (libc) \ |
| 20 |
&& (defined SHARED \ |
| 21 |
|| defined USE_AS_MEMMOVE \ |
| 22 |
|| !defined USE_MULTIARCH) |
| 23 |
|
| 24 |
# include <sysdep.h> |
| 25 |
# include "asm-syntax.h" |
| 26 |
|
| 27 |
# ifndef MEMCPY |
| 28 |
# define MEMCPY __memcpy_ssse3 |
| 29 |
# define MEMCPY_CHK __memcpy_chk_ssse3 |
| 30 |
# endif |
| 31 |
|
| 32 |
# define DEST PARMS |
| 33 |
# define SRC DEST+4 |
| 34 |
# define LEN SRC+4 |
| 35 |
|
| 36 |
# define CFI_PUSH(REG) \ |
| 37 |
cfi_adjust_cfa_offset (4); \ |
| 38 |
cfi_rel_offset (REG, 0) |
| 39 |
|
| 40 |
# define CFI_POP(REG) \ |
| 41 |
cfi_adjust_cfa_offset (-4); \ |
| 42 |
cfi_restore (REG) |
| 43 |
|
| 44 |
# define PUSH(REG) pushl REG; CFI_PUSH (REG) |
| 45 |
# define POP(REG) popl REG; CFI_POP (REG) |
| 46 |
|
| 47 |
# ifdef PIC |
| 48 |
# define PARMS 8 /* Preserve EBX. */ |
| 49 |
# define ENTRANCE PUSH (%ebx); |
| 50 |
# define RETURN_END POP (%ebx); ret |
| 51 |
# define RETURN RETURN_END; CFI_PUSH (%ebx) |
| 52 |
# define JMPTBL(I, B) I - B |
| 53 |
|
| 54 |
/* Load an entry in a jump table into EBX and branch to it. TABLE is a |
| 55 |
jump table with relative offsets. INDEX is a register contains the |
| 56 |
index into the jump table. SCALE is the scale of INDEX. */ |
| 57 |
|
| 58 |
# define BRANCH_TO_JMPTBL_ENTRY(TABLE, INDEX, SCALE) \ |
| 59 |
/* We first load PC into EBX. */ \ |
| 60 |
SETUP_PIC_REG(bx); \ |
| 61 |
/* Get the address of the jump table. */ \ |
| 62 |
addl $(TABLE - .), %ebx; \ |
| 63 |
/* Get the entry and convert the relative offset to the \ |
| 64 |
absolute address. */ \ |
| 65 |
addl (%ebx, INDEX, SCALE), %ebx; \ |
| 66 |
/* We loaded the jump table. Go. */ \ |
| 67 |
jmp *%ebx |
| 68 |
# else |
| 69 |
|
| 70 |
# define PARMS 4 |
| 71 |
# define ENTRANCE |
| 72 |
# define RETURN_END ret |
| 73 |
# define RETURN RETURN_END |
| 74 |
# define JMPTBL(I, B) I |
| 75 |
|
| 76 |
/* Branch to an entry in a jump table. TABLE is a jump table with |
| 77 |
absolute offsets. INDEX is a register contains the index into the |
| 78 |
jump table. SCALE is the scale of INDEX. */ |
| 79 |
|
| 80 |
# define BRANCH_TO_JMPTBL_ENTRY(TABLE, INDEX, SCALE) \ |
| 81 |
jmp *TABLE(, INDEX, SCALE) |
| 82 |
# endif |
| 83 |
|
| 84 |
.section .text.ssse3,"ax",@progbits |
| 85 |
# ifdef SHARED |
| 86 |
ENTRY (MEMCPY_CHK) |
| 87 |
movl 12(%esp), %eax |
| 88 |
cmpl %eax, 16(%esp) |
| 89 |
jb HIDDEN_JUMPTARGET (__chk_fail) |
| 90 |
END (MEMCPY_CHK) |
| 91 |
# endif |
| 92 |
ENTRY (MEMCPY) |
| 93 |
ENTRANCE |
| 94 |
movl LEN(%esp), %ecx |
| 95 |
movl SRC(%esp), %eax |
| 96 |
movl DEST(%esp), %edx |
| 97 |
|
| 98 |
# ifdef USE_AS_MEMMOVE |
| 99 |
cmp %eax, %edx |
| 100 |
jb L(copy_forward) |
| 101 |
je L(fwd_write_0bytes) |
| 102 |
cmp $32, %ecx |
| 103 |
jae L(memmove_bwd) |
| 104 |
jmp L(bk_write_less32bytes_2) |
| 105 |
|
| 106 |
.p2align 4 |
| 107 |
L(memmove_bwd): |
| 108 |
add %ecx, %eax |
| 109 |
cmp %eax, %edx |
| 110 |
movl SRC(%esp), %eax |
| 111 |
jb L(copy_backward) |
| 112 |
|
| 113 |
L(copy_forward): |
| 114 |
# endif |
| 115 |
cmp $48, %ecx |
| 116 |
jae L(48bytesormore) |
| 117 |
|
| 118 |
L(fwd_write_less32bytes): |
| 119 |
# ifndef USE_AS_MEMMOVE |
| 120 |
cmp %dl, %al |
| 121 |
jb L(bk_write) |
| 122 |
# endif |
| 123 |
add %ecx, %edx |
| 124 |
add %ecx, %eax |
| 125 |
BRANCH_TO_JMPTBL_ENTRY (L(table_48bytes_fwd), %ecx, 4) |
| 126 |
# ifndef USE_AS_MEMMOVE |
| 127 |
.p2align 4 |
| 128 |
L(bk_write): |
| 129 |
BRANCH_TO_JMPTBL_ENTRY (L(table_48_bytes_bwd), %ecx, 4) |
| 130 |
# endif |
| 131 |
|
| 132 |
.p2align 4 |
| 133 |
L(48bytesormore): |
| 134 |
# ifndef USE_AS_MEMMOVE |
| 135 |
movlpd (%eax), %xmm0 |
| 136 |
movlpd 8(%eax), %xmm1 |
| 137 |
movlpd %xmm0, (%edx) |
| 138 |
movlpd %xmm1, 8(%edx) |
| 139 |
# else |
| 140 |
movdqu (%eax), %xmm0 |
| 141 |
# endif |
| 142 |
PUSH (%edi) |
| 143 |
movl %edx, %edi |
| 144 |
and $-16, %edx |
| 145 |
add $16, %edx |
| 146 |
sub %edx, %edi |
| 147 |
add %edi, %ecx |
| 148 |
sub %edi, %eax |
| 149 |
|
| 150 |
# ifdef SHARED_CACHE_SIZE_HALF |
| 151 |
cmp $SHARED_CACHE_SIZE_HALF, %ecx |
| 152 |
# else |
| 153 |
# ifdef PIC |
| 154 |
SETUP_PIC_REG(bx) |
| 155 |
add $_GLOBAL_OFFSET_TABLE_, %ebx |
| 156 |
cmp __x86_shared_cache_size_half@GOTOFF(%ebx), %ecx |
| 157 |
# else |
| 158 |
cmp __x86_shared_cache_size_half, %ecx |
| 159 |
# endif |
| 160 |
# endif |
| 161 |
|
| 162 |
mov %eax, %edi |
| 163 |
jae L(large_page) |
| 164 |
and $0xf, %edi |
| 165 |
jz L(shl_0) |
| 166 |
BRANCH_TO_JMPTBL_ENTRY (L(shl_table), %edi, 4) |
| 167 |
|
| 168 |
.p2align 4 |
| 169 |
L(shl_0): |
| 170 |
# ifdef USE_AS_MEMMOVE |
| 171 |
movl DEST+4(%esp), %edi |
| 172 |
movdqu %xmm0, (%edi) |
| 173 |
# endif |
| 174 |
xor %edi, %edi |
| 175 |
cmp $127, %ecx |
| 176 |
ja L(shl_0_gobble) |
| 177 |
lea -32(%ecx), %ecx |
| 178 |
|
| 179 |
.p2align 4 |
| 180 |
L(shl_0_loop): |
| 181 |
movdqa (%eax, %edi), %xmm0 |
| 182 |
movdqa 16(%eax, %edi), %xmm1 |
| 183 |
sub $32, %ecx |
| 184 |
movdqa %xmm0, (%edx, %edi) |
| 185 |
movdqa %xmm1, 16(%edx, %edi) |
| 186 |
lea 32(%edi), %edi |
| 187 |
jb L(shl_0_end) |
| 188 |
|
| 189 |
movdqa (%eax, %edi), %xmm0 |
| 190 |
movdqa 16(%eax, %edi), %xmm1 |
| 191 |
sub $32, %ecx |
| 192 |
movdqa %xmm0, (%edx, %edi) |
| 193 |
movdqa %xmm1, 16(%edx, %edi) |
| 194 |
lea 32(%edi), %edi |
| 195 |
jb L(shl_0_end) |
| 196 |
|
| 197 |
movdqa (%eax, %edi), %xmm0 |
| 198 |
movdqa 16(%eax, %edi), %xmm1 |
| 199 |
sub $32, %ecx |
| 200 |
movdqa %xmm0, (%edx, %edi) |
| 201 |
movdqa %xmm1, 16(%edx, %edi) |
| 202 |
lea 32(%edi), %edi |
| 203 |
jb L(shl_0_end) |
| 204 |
|
| 205 |
movdqa (%eax, %edi), %xmm0 |
| 206 |
movdqa 16(%eax, %edi), %xmm1 |
| 207 |
sub $32, %ecx |
| 208 |
movdqa %xmm0, (%edx, %edi) |
| 209 |
movdqa %xmm1, 16(%edx, %edi) |
| 210 |
lea 32(%edi), %edi |
| 211 |
|
| 212 |
L(shl_0_end): |
| 213 |
lea 32(%ecx), %ecx |
| 214 |
add %ecx, %edi |
| 215 |
add %edi, %edx |
| 216 |
add %edi, %eax |
| 217 |
POP (%edi) |
| 218 |
BRANCH_TO_JMPTBL_ENTRY (L(table_48bytes_fwd_align), %ecx, 4) |
| 219 |
|
| 220 |
CFI_PUSH (%edi) |
| 221 |
|
| 222 |
.p2align 4 |
| 223 |
L(shl_0_gobble): |
| 224 |
# ifdef DATA_CACHE_SIZE_HALF |
| 225 |
cmp $DATA_CACHE_SIZE_HALF, %ecx |
| 226 |
# else |
| 227 |
# ifdef PIC |
| 228 |
SETUP_PIC_REG(bx) |
| 229 |
add $_GLOBAL_OFFSET_TABLE_, %ebx |
| 230 |
cmp __x86_data_cache_size_half@GOTOFF(%ebx), %ecx |
| 231 |
# else |
| 232 |
cmp __x86_data_cache_size_half, %ecx |
| 233 |
# endif |
| 234 |
# endif |
| 235 |
POP (%edi) |
| 236 |
lea -128(%ecx), %ecx |
| 237 |
jae L(shl_0_gobble_mem_loop) |
| 238 |
|
| 239 |
.p2align 4 |
| 240 |
L(shl_0_gobble_cache_loop): |
| 241 |
movdqa (%eax), %xmm0 |
| 242 |
movdqa 0x10(%eax), %xmm1 |
| 243 |
movdqa 0x20(%eax), %xmm2 |
| 244 |
movdqa 0x30(%eax), %xmm3 |
| 245 |
movdqa 0x40(%eax), %xmm4 |
| 246 |
movdqa 0x50(%eax), %xmm5 |
| 247 |
movdqa 0x60(%eax), %xmm6 |
| 248 |
movdqa 0x70(%eax), %xmm7 |
| 249 |
lea 0x80(%eax), %eax |
| 250 |
sub $128, %ecx |
| 251 |
movdqa %xmm0, (%edx) |
| 252 |
movdqa %xmm1, 0x10(%edx) |
| 253 |
movdqa %xmm2, 0x20(%edx) |
| 254 |
movdqa %xmm3, 0x30(%edx) |
| 255 |
movdqa %xmm4, 0x40(%edx) |
| 256 |
movdqa %xmm5, 0x50(%edx) |
| 257 |
movdqa %xmm6, 0x60(%edx) |
| 258 |
movdqa %xmm7, 0x70(%edx) |
| 259 |
lea 0x80(%edx), %edx |
| 260 |
|
| 261 |
jae L(shl_0_gobble_cache_loop) |
| 262 |
cmp $-0x40, %ecx |
| 263 |
lea 0x80(%ecx), %ecx |
| 264 |
jl L(shl_0_cache_less_64bytes) |
| 265 |
|
| 266 |
movdqa (%eax), %xmm0 |
| 267 |
sub $0x40, %ecx |
| 268 |
movdqa 0x10(%eax), %xmm1 |
| 269 |
movdqa %xmm0, (%edx) |
| 270 |
movdqa %xmm1, 0x10(%edx) |
| 271 |
movdqa 0x20(%eax), %xmm0 |
| 272 |
movdqa 0x30(%eax), %xmm1 |
| 273 |
add $0x40, %eax |
| 274 |
movdqa %xmm0, 0x20(%edx) |
| 275 |
movdqa %xmm1, 0x30(%edx) |
| 276 |
add $0x40, %edx |
| 277 |
|
| 278 |
L(shl_0_cache_less_64bytes): |
| 279 |
cmp $0x20, %ecx |
| 280 |
jb L(shl_0_cache_less_32bytes) |
| 281 |
movdqa (%eax), %xmm0 |
| 282 |
sub $0x20, %ecx |
| 283 |
movdqa 0x10(%eax), %xmm1 |
| 284 |
add $0x20, %eax |
| 285 |
movdqa %xmm0, (%edx) |
| 286 |
movdqa %xmm1, 0x10(%edx) |
| 287 |
add $0x20, %edx |
| 288 |
|
| 289 |
L(shl_0_cache_less_32bytes): |
| 290 |
cmp $0x10, %ecx |
| 291 |
jb L(shl_0_cache_less_16bytes) |
| 292 |
sub $0x10, %ecx |
| 293 |
movdqa (%eax), %xmm0 |
| 294 |
add $0x10, %eax |
| 295 |
movdqa %xmm0, (%edx) |
| 296 |
add $0x10, %edx |
| 297 |
|
| 298 |
L(shl_0_cache_less_16bytes): |
| 299 |
add %ecx, %edx |
| 300 |
add %ecx, %eax |
| 301 |
BRANCH_TO_JMPTBL_ENTRY (L(table_48bytes_fwd), %ecx, 4) |
| 302 |
|
| 303 |
.p2align 4 |
| 304 |
L(shl_0_gobble_mem_loop): |
| 305 |
prefetcht0 0x1c0(%eax) |
| 306 |
prefetcht0 0x280(%eax) |
| 307 |
prefetcht0 0x1c0(%edx) |
| 308 |
|
| 309 |
movdqa (%eax), %xmm0 |
| 310 |
movdqa 0x10(%eax), %xmm1 |
| 311 |
movdqa 0x20(%eax), %xmm2 |
| 312 |
movdqa 0x30(%eax), %xmm3 |
| 313 |
movdqa 0x40(%eax), %xmm4 |
| 314 |
movdqa 0x50(%eax), %xmm5 |
| 315 |
movdqa 0x60(%eax), %xmm6 |
| 316 |
movdqa 0x70(%eax), %xmm7 |
| 317 |
lea 0x80(%eax), %eax |
| 318 |
sub $0x80, %ecx |
| 319 |
movdqa %xmm0, (%edx) |
| 320 |
movdqa %xmm1, 0x10(%edx) |
| 321 |
movdqa %xmm2, 0x20(%edx) |
| 322 |
movdqa %xmm3, 0x30(%edx) |
| 323 |
movdqa %xmm4, 0x40(%edx) |
| 324 |
movdqa %xmm5, 0x50(%edx) |
| 325 |
movdqa %xmm6, 0x60(%edx) |
| 326 |
movdqa %xmm7, 0x70(%edx) |
| 327 |
lea 0x80(%edx), %edx |
| 328 |
|
| 329 |
jae L(shl_0_gobble_mem_loop) |
| 330 |
cmp $-0x40, %ecx |
| 331 |
lea 0x80(%ecx), %ecx |
| 332 |
jl L(shl_0_mem_less_64bytes) |
| 333 |
|
| 334 |
movdqa (%eax), %xmm0 |
| 335 |
sub $0x40, %ecx |
| 336 |
movdqa 0x10(%eax), %xmm1 |
| 337 |
|
| 338 |
movdqa %xmm0, (%edx) |
| 339 |
movdqa %xmm1, 0x10(%edx) |
| 340 |
|
| 341 |
movdqa 0x20(%eax), %xmm0 |
| 342 |
movdqa 0x30(%eax), %xmm1 |
| 343 |
add $0x40, %eax |
| 344 |
|
| 345 |
movdqa %xmm0, 0x20(%edx) |
| 346 |
movdqa %xmm1, 0x30(%edx) |
| 347 |
add $0x40, %edx |
| 348 |
|
| 349 |
L(shl_0_mem_less_64bytes): |
| 350 |
cmp $0x20, %ecx |
| 351 |
jb L(shl_0_mem_less_32bytes) |
| 352 |
movdqa (%eax), %xmm0 |
| 353 |
sub $0x20, %ecx |
| 354 |
movdqa 0x10(%eax), %xmm1 |
| 355 |
add $0x20, %eax |
| 356 |
movdqa %xmm0, (%edx) |
| 357 |
movdqa %xmm1, 0x10(%edx) |
| 358 |
add $0x20, %edx |
| 359 |
|
| 360 |
L(shl_0_mem_less_32bytes): |
| 361 |
cmp $0x10, %ecx |
| 362 |
jb L(shl_0_mem_less_16bytes) |
| 363 |
sub $0x10, %ecx |
| 364 |
movdqa (%eax), %xmm0 |
| 365 |
add $0x10, %eax |
| 366 |
movdqa %xmm0, (%edx) |
| 367 |
add $0x10, %edx |
| 368 |
|
| 369 |
L(shl_0_mem_less_16bytes): |
| 370 |
add %ecx, %edx |
| 371 |
add %ecx, %eax |
| 372 |
BRANCH_TO_JMPTBL_ENTRY (L(table_48bytes_fwd_align), %ecx, 4) |
| 373 |
|
| 374 |
.p2align 4 |
| 375 |
L(shl_1): |
| 376 |
# ifndef USE_AS_MEMMOVE |
| 377 |
movaps -1(%eax), %xmm1 |
| 378 |
# else |
| 379 |
movl DEST+4(%esp), %edi |
| 380 |
movaps -1(%eax), %xmm1 |
| 381 |
movdqu %xmm0, (%edi) |
| 382 |
# endif |
| 383 |
# ifdef DATA_CACHE_SIZE_HALF |
| 384 |
cmp $DATA_CACHE_SIZE_HALF, %ecx |
| 385 |
# else |
| 386 |
# ifdef PIC |
| 387 |
SETUP_PIC_REG(bx) |
| 388 |
add $_GLOBAL_OFFSET_TABLE_, %ebx |
| 389 |
cmp __x86_data_cache_size_half@GOTOFF(%ebx), %ecx |
| 390 |
# else |
| 391 |
cmp __x86_data_cache_size_half, %ecx |
| 392 |
# endif |
| 393 |
# endif |
| 394 |
jb L(sh_1_no_prefetch) |
| 395 |
|
| 396 |
lea -64(%ecx), %ecx |
| 397 |
|
| 398 |
.p2align 4 |
| 399 |
L(Shl1LoopStart): |
| 400 |
prefetcht0 0x1c0(%eax) |
| 401 |
prefetcht0 0x1c0(%edx) |
| 402 |
movaps 15(%eax), %xmm2 |
| 403 |
movaps 31(%eax), %xmm3 |
| 404 |
movaps 47(%eax), %xmm4 |
| 405 |
movaps 63(%eax), %xmm5 |
| 406 |
movaps %xmm5, %xmm7 |
| 407 |
palignr $1, %xmm4, %xmm5 |
| 408 |
palignr $1, %xmm3, %xmm4 |
| 409 |
movaps %xmm5, 48(%edx) |
| 410 |
palignr $1, %xmm2, %xmm3 |
| 411 |
lea 64(%eax), %eax |
| 412 |
palignr $1, %xmm1, %xmm2 |
| 413 |
movaps %xmm4, 32(%edx) |
| 414 |
movaps %xmm3, 16(%edx) |
| 415 |
movaps %xmm7, %xmm1 |
| 416 |
movaps %xmm2, (%edx) |
| 417 |
lea 64(%edx), %edx |
| 418 |
sub $64, %ecx |
| 419 |
ja L(Shl1LoopStart) |
| 420 |
|
| 421 |
L(Shl1LoopLeave): |
| 422 |
add $32, %ecx |
| 423 |
jle L(shl_end_0) |
| 424 |
|
| 425 |
movaps 15(%eax), %xmm2 |
| 426 |
movaps 31(%eax), %xmm3 |
| 427 |
palignr $1, %xmm2, %xmm3 |
| 428 |
palignr $1, %xmm1, %xmm2 |
| 429 |
movaps %xmm2, (%edx) |
| 430 |
movaps %xmm3, 16(%edx) |
| 431 |
lea 32(%edx, %ecx), %edx |
| 432 |
lea 32(%eax, %ecx), %eax |
| 433 |
POP (%edi) |
| 434 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 435 |
|
| 436 |
CFI_PUSH (%edi) |
| 437 |
|
| 438 |
.p2align 4 |
| 439 |
L(sh_1_no_prefetch): |
| 440 |
lea -32(%ecx), %ecx |
| 441 |
lea -1(%eax), %eax |
| 442 |
xor %edi, %edi |
| 443 |
|
| 444 |
.p2align 4 |
| 445 |
L(sh_1_no_prefetch_loop): |
| 446 |
movdqa 16(%eax, %edi), %xmm2 |
| 447 |
sub $32, %ecx |
| 448 |
movdqa 32(%eax, %edi), %xmm3 |
| 449 |
movdqa %xmm3, %xmm4 |
| 450 |
palignr $1, %xmm2, %xmm3 |
| 451 |
palignr $1, %xmm1, %xmm2 |
| 452 |
lea 32(%edi), %edi |
| 453 |
movdqa %xmm2, -32(%edx, %edi) |
| 454 |
movdqa %xmm3, -16(%edx, %edi) |
| 455 |
jb L(sh_1_end_no_prefetch_loop) |
| 456 |
|
| 457 |
movdqa 16(%eax, %edi), %xmm2 |
| 458 |
sub $32, %ecx |
| 459 |
movdqa 32(%eax, %edi), %xmm3 |
| 460 |
movdqa %xmm3, %xmm1 |
| 461 |
palignr $1, %xmm2, %xmm3 |
| 462 |
palignr $1, %xmm4, %xmm2 |
| 463 |
lea 32(%edi), %edi |
| 464 |
movdqa %xmm2, -32(%edx, %edi) |
| 465 |
movdqa %xmm3, -16(%edx, %edi) |
| 466 |
jae L(sh_1_no_prefetch_loop) |
| 467 |
|
| 468 |
L(sh_1_end_no_prefetch_loop): |
| 469 |
lea 32(%ecx), %ecx |
| 470 |
add %ecx, %edi |
| 471 |
add %edi, %edx |
| 472 |
lea 1(%edi, %eax), %eax |
| 473 |
POP (%edi) |
| 474 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 475 |
|
| 476 |
CFI_PUSH (%edi) |
| 477 |
|
| 478 |
.p2align 4 |
| 479 |
L(shl_2): |
| 480 |
# ifndef USE_AS_MEMMOVE |
| 481 |
movaps -2(%eax), %xmm1 |
| 482 |
# else |
| 483 |
movl DEST+4(%esp), %edi |
| 484 |
movaps -2(%eax), %xmm1 |
| 485 |
movdqu %xmm0, (%edi) |
| 486 |
# endif |
| 487 |
# ifdef DATA_CACHE_SIZE_HALF |
| 488 |
cmp $DATA_CACHE_SIZE_HALF, %ecx |
| 489 |
# else |
| 490 |
# ifdef PIC |
| 491 |
SETUP_PIC_REG(bx) |
| 492 |
add $_GLOBAL_OFFSET_TABLE_, %ebx |
| 493 |
cmp __x86_data_cache_size_half@GOTOFF(%ebx), %ecx |
| 494 |
# else |
| 495 |
cmp __x86_data_cache_size_half, %ecx |
| 496 |
# endif |
| 497 |
# endif |
| 498 |
jb L(sh_2_no_prefetch) |
| 499 |
|
| 500 |
lea -64(%ecx), %ecx |
| 501 |
|
| 502 |
.p2align 4 |
| 503 |
L(Shl2LoopStart): |
| 504 |
prefetcht0 0x1c0(%eax) |
| 505 |
prefetcht0 0x1c0(%edx) |
| 506 |
movaps 14(%eax), %xmm2 |
| 507 |
movaps 30(%eax), %xmm3 |
| 508 |
movaps 46(%eax), %xmm4 |
| 509 |
movaps 62(%eax), %xmm5 |
| 510 |
movaps %xmm5, %xmm7 |
| 511 |
palignr $2, %xmm4, %xmm5 |
| 512 |
palignr $2, %xmm3, %xmm4 |
| 513 |
movaps %xmm5, 48(%edx) |
| 514 |
palignr $2, %xmm2, %xmm3 |
| 515 |
lea 64(%eax), %eax |
| 516 |
palignr $2, %xmm1, %xmm2 |
| 517 |
movaps %xmm4, 32(%edx) |
| 518 |
movaps %xmm3, 16(%edx) |
| 519 |
movaps %xmm7, %xmm1 |
| 520 |
movaps %xmm2, (%edx) |
| 521 |
lea 64(%edx), %edx |
| 522 |
sub $64, %ecx |
| 523 |
ja L(Shl2LoopStart) |
| 524 |
|
| 525 |
L(Shl2LoopLeave): |
| 526 |
add $32, %ecx |
| 527 |
jle L(shl_end_0) |
| 528 |
|
| 529 |
movaps 14(%eax), %xmm2 |
| 530 |
movaps 30(%eax), %xmm3 |
| 531 |
palignr $2, %xmm2, %xmm3 |
| 532 |
palignr $2, %xmm1, %xmm2 |
| 533 |
movaps %xmm2, (%edx) |
| 534 |
movaps %xmm3, 16(%edx) |
| 535 |
lea 32(%edx, %ecx), %edx |
| 536 |
lea 32(%eax, %ecx), %eax |
| 537 |
POP (%edi) |
| 538 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 539 |
|
| 540 |
CFI_PUSH (%edi) |
| 541 |
|
| 542 |
.p2align 4 |
| 543 |
L(sh_2_no_prefetch): |
| 544 |
lea -32(%ecx), %ecx |
| 545 |
lea -2(%eax), %eax |
| 546 |
xor %edi, %edi |
| 547 |
|
| 548 |
.p2align 4 |
| 549 |
L(sh_2_no_prefetch_loop): |
| 550 |
movdqa 16(%eax, %edi), %xmm2 |
| 551 |
sub $32, %ecx |
| 552 |
movdqa 32(%eax, %edi), %xmm3 |
| 553 |
movdqa %xmm3, %xmm4 |
| 554 |
palignr $2, %xmm2, %xmm3 |
| 555 |
palignr $2, %xmm1, %xmm2 |
| 556 |
lea 32(%edi), %edi |
| 557 |
movdqa %xmm2, -32(%edx, %edi) |
| 558 |
movdqa %xmm3, -16(%edx, %edi) |
| 559 |
jb L(sh_2_end_no_prefetch_loop) |
| 560 |
|
| 561 |
movdqa 16(%eax, %edi), %xmm2 |
| 562 |
sub $32, %ecx |
| 563 |
movdqa 32(%eax, %edi), %xmm3 |
| 564 |
movdqa %xmm3, %xmm1 |
| 565 |
palignr $2, %xmm2, %xmm3 |
| 566 |
palignr $2, %xmm4, %xmm2 |
| 567 |
lea 32(%edi), %edi |
| 568 |
movdqa %xmm2, -32(%edx, %edi) |
| 569 |
movdqa %xmm3, -16(%edx, %edi) |
| 570 |
jae L(sh_2_no_prefetch_loop) |
| 571 |
|
| 572 |
L(sh_2_end_no_prefetch_loop): |
| 573 |
lea 32(%ecx), %ecx |
| 574 |
add %ecx, %edi |
| 575 |
add %edi, %edx |
| 576 |
lea 2(%edi, %eax), %eax |
| 577 |
POP (%edi) |
| 578 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 579 |
|
| 580 |
CFI_PUSH (%edi) |
| 581 |
|
| 582 |
.p2align 4 |
| 583 |
L(shl_3): |
| 584 |
# ifndef USE_AS_MEMMOVE |
| 585 |
movaps -3(%eax), %xmm1 |
| 586 |
# else |
| 587 |
movl DEST+4(%esp), %edi |
| 588 |
movaps -3(%eax), %xmm1 |
| 589 |
movdqu %xmm0, (%edi) |
| 590 |
# endif |
| 591 |
# ifdef DATA_CACHE_SIZE_HALF |
| 592 |
cmp $DATA_CACHE_SIZE_HALF, %ecx |
| 593 |
# else |
| 594 |
# ifdef PIC |
| 595 |
SETUP_PIC_REG(bx) |
| 596 |
add $_GLOBAL_OFFSET_TABLE_, %ebx |
| 597 |
cmp __x86_data_cache_size_half@GOTOFF(%ebx), %ecx |
| 598 |
# else |
| 599 |
cmp __x86_data_cache_size_half, %ecx |
| 600 |
# endif |
| 601 |
# endif |
| 602 |
jb L(sh_3_no_prefetch) |
| 603 |
|
| 604 |
lea -64(%ecx), %ecx |
| 605 |
|
| 606 |
.p2align 4 |
| 607 |
L(Shl3LoopStart): |
| 608 |
prefetcht0 0x1c0(%eax) |
| 609 |
prefetcht0 0x1c0(%edx) |
| 610 |
movaps 13(%eax), %xmm2 |
| 611 |
movaps 29(%eax), %xmm3 |
| 612 |
movaps 45(%eax), %xmm4 |
| 613 |
movaps 61(%eax), %xmm5 |
| 614 |
movaps %xmm5, %xmm7 |
| 615 |
palignr $3, %xmm4, %xmm5 |
| 616 |
palignr $3, %xmm3, %xmm4 |
| 617 |
movaps %xmm5, 48(%edx) |
| 618 |
palignr $3, %xmm2, %xmm3 |
| 619 |
lea 64(%eax), %eax |
| 620 |
palignr $3, %xmm1, %xmm2 |
| 621 |
movaps %xmm4, 32(%edx) |
| 622 |
movaps %xmm3, 16(%edx) |
| 623 |
movaps %xmm7, %xmm1 |
| 624 |
movaps %xmm2, (%edx) |
| 625 |
lea 64(%edx), %edx |
| 626 |
sub $64, %ecx |
| 627 |
ja L(Shl3LoopStart) |
| 628 |
|
| 629 |
L(Shl3LoopLeave): |
| 630 |
add $32, %ecx |
| 631 |
jle L(shl_end_0) |
| 632 |
|
| 633 |
movaps 13(%eax), %xmm2 |
| 634 |
movaps 29(%eax), %xmm3 |
| 635 |
palignr $3, %xmm2, %xmm3 |
| 636 |
palignr $3, %xmm1, %xmm2 |
| 637 |
movaps %xmm2, (%edx) |
| 638 |
movaps %xmm3, 16(%edx) |
| 639 |
lea 32(%edx, %ecx), %edx |
| 640 |
lea 32(%eax, %ecx), %eax |
| 641 |
POP (%edi) |
| 642 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 643 |
|
| 644 |
CFI_PUSH (%edi) |
| 645 |
|
| 646 |
.p2align 4 |
| 647 |
L(sh_3_no_prefetch): |
| 648 |
lea -32(%ecx), %ecx |
| 649 |
lea -3(%eax), %eax |
| 650 |
xor %edi, %edi |
| 651 |
|
| 652 |
.p2align 4 |
| 653 |
L(sh_3_no_prefetch_loop): |
| 654 |
movdqa 16(%eax, %edi), %xmm2 |
| 655 |
sub $32, %ecx |
| 656 |
movdqa 32(%eax, %edi), %xmm3 |
| 657 |
movdqa %xmm3, %xmm4 |
| 658 |
palignr $3, %xmm2, %xmm3 |
| 659 |
palignr $3, %xmm1, %xmm2 |
| 660 |
lea 32(%edi), %edi |
| 661 |
movdqa %xmm2, -32(%edx, %edi) |
| 662 |
movdqa %xmm3, -16(%edx, %edi) |
| 663 |
|
| 664 |
jb L(sh_3_end_no_prefetch_loop) |
| 665 |
|
| 666 |
movdqa 16(%eax, %edi), %xmm2 |
| 667 |
sub $32, %ecx |
| 668 |
movdqa 32(%eax, %edi), %xmm3 |
| 669 |
movdqa %xmm3, %xmm1 |
| 670 |
palignr $3, %xmm2, %xmm3 |
| 671 |
palignr $3, %xmm4, %xmm2 |
| 672 |
lea 32(%edi), %edi |
| 673 |
movdqa %xmm2, -32(%edx, %edi) |
| 674 |
movdqa %xmm3, -16(%edx, %edi) |
| 675 |
|
| 676 |
jae L(sh_3_no_prefetch_loop) |
| 677 |
|
| 678 |
L(sh_3_end_no_prefetch_loop): |
| 679 |
lea 32(%ecx), %ecx |
| 680 |
add %ecx, %edi |
| 681 |
add %edi, %edx |
| 682 |
lea 3(%edi, %eax), %eax |
| 683 |
POP (%edi) |
| 684 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 685 |
|
| 686 |
CFI_PUSH (%edi) |
| 687 |
|
| 688 |
.p2align 4 |
| 689 |
L(shl_4): |
| 690 |
# ifndef USE_AS_MEMMOVE |
| 691 |
movaps -4(%eax), %xmm1 |
| 692 |
# else |
| 693 |
movl DEST+4(%esp), %edi |
| 694 |
movaps -4(%eax), %xmm1 |
| 695 |
movdqu %xmm0, (%edi) |
| 696 |
# endif |
| 697 |
# ifdef DATA_CACHE_SIZE_HALF |
| 698 |
cmp $DATA_CACHE_SIZE_HALF, %ecx |
| 699 |
# else |
| 700 |
# ifdef PIC |
| 701 |
SETUP_PIC_REG(bx) |
| 702 |
add $_GLOBAL_OFFSET_TABLE_, %ebx |
| 703 |
cmp __x86_data_cache_size_half@GOTOFF(%ebx), %ecx |
| 704 |
# else |
| 705 |
cmp __x86_data_cache_size_half, %ecx |
| 706 |
# endif |
| 707 |
# endif |
| 708 |
jb L(sh_4_no_prefetch) |
| 709 |
|
| 710 |
lea -64(%ecx), %ecx |
| 711 |
|
| 712 |
.p2align 4 |
| 713 |
L(Shl4LoopStart): |
| 714 |
prefetcht0 0x1c0(%eax) |
| 715 |
prefetcht0 0x1c0(%edx) |
| 716 |
movaps 12(%eax), %xmm2 |
| 717 |
movaps 28(%eax), %xmm3 |
| 718 |
movaps 44(%eax), %xmm4 |
| 719 |
movaps 60(%eax), %xmm5 |
| 720 |
movaps %xmm5, %xmm7 |
| 721 |
palignr $4, %xmm4, %xmm5 |
| 722 |
palignr $4, %xmm3, %xmm4 |
| 723 |
movaps %xmm5, 48(%edx) |
| 724 |
palignr $4, %xmm2, %xmm3 |
| 725 |
lea 64(%eax), %eax |
| 726 |
palignr $4, %xmm1, %xmm2 |
| 727 |
movaps %xmm4, 32(%edx) |
| 728 |
movaps %xmm3, 16(%edx) |
| 729 |
movaps %xmm7, %xmm1 |
| 730 |
movaps %xmm2, (%edx) |
| 731 |
lea 64(%edx), %edx |
| 732 |
sub $64, %ecx |
| 733 |
ja L(Shl4LoopStart) |
| 734 |
|
| 735 |
L(Shl4LoopLeave): |
| 736 |
add $32, %ecx |
| 737 |
jle L(shl_end_0) |
| 738 |
|
| 739 |
movaps 12(%eax), %xmm2 |
| 740 |
movaps 28(%eax), %xmm3 |
| 741 |
palignr $4, %xmm2, %xmm3 |
| 742 |
palignr $4, %xmm1, %xmm2 |
| 743 |
movaps %xmm2, (%edx) |
| 744 |
movaps %xmm3, 16(%edx) |
| 745 |
lea 32(%edx, %ecx), %edx |
| 746 |
lea 32(%eax, %ecx), %eax |
| 747 |
POP (%edi) |
| 748 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 749 |
|
| 750 |
CFI_PUSH (%edi) |
| 751 |
|
| 752 |
.p2align 4 |
| 753 |
L(sh_4_no_prefetch): |
| 754 |
lea -32(%ecx), %ecx |
| 755 |
lea -4(%eax), %eax |
| 756 |
xor %edi, %edi |
| 757 |
|
| 758 |
.p2align 4 |
| 759 |
L(sh_4_no_prefetch_loop): |
| 760 |
movdqa 16(%eax, %edi), %xmm2 |
| 761 |
sub $32, %ecx |
| 762 |
movdqa 32(%eax, %edi), %xmm3 |
| 763 |
movdqa %xmm3, %xmm4 |
| 764 |
palignr $4, %xmm2, %xmm3 |
| 765 |
palignr $4, %xmm1, %xmm2 |
| 766 |
lea 32(%edi), %edi |
| 767 |
movdqa %xmm2, -32(%edx, %edi) |
| 768 |
movdqa %xmm3, -16(%edx, %edi) |
| 769 |
|
| 770 |
jb L(sh_4_end_no_prefetch_loop) |
| 771 |
|
| 772 |
movdqa 16(%eax, %edi), %xmm2 |
| 773 |
sub $32, %ecx |
| 774 |
movdqa 32(%eax, %edi), %xmm3 |
| 775 |
movdqa %xmm3, %xmm1 |
| 776 |
palignr $4, %xmm2, %xmm3 |
| 777 |
palignr $4, %xmm4, %xmm2 |
| 778 |
lea 32(%edi), %edi |
| 779 |
movdqa %xmm2, -32(%edx, %edi) |
| 780 |
movdqa %xmm3, -16(%edx, %edi) |
| 781 |
|
| 782 |
jae L(sh_4_no_prefetch_loop) |
| 783 |
|
| 784 |
L(sh_4_end_no_prefetch_loop): |
| 785 |
lea 32(%ecx), %ecx |
| 786 |
add %ecx, %edi |
| 787 |
add %edi, %edx |
| 788 |
lea 4(%edi, %eax), %eax |
| 789 |
POP (%edi) |
| 790 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 791 |
|
| 792 |
CFI_PUSH (%edi) |
| 793 |
|
| 794 |
.p2align 4 |
| 795 |
L(shl_5): |
| 796 |
# ifndef USE_AS_MEMMOVE |
| 797 |
movaps -5(%eax), %xmm1 |
| 798 |
# else |
| 799 |
movl DEST+4(%esp), %edi |
| 800 |
movaps -5(%eax), %xmm1 |
| 801 |
movdqu %xmm0, (%edi) |
| 802 |
# endif |
| 803 |
# ifdef DATA_CACHE_SIZE_HALF |
| 804 |
cmp $DATA_CACHE_SIZE_HALF, %ecx |
| 805 |
# else |
| 806 |
# ifdef PIC |
| 807 |
SETUP_PIC_REG(bx) |
| 808 |
add $_GLOBAL_OFFSET_TABLE_, %ebx |
| 809 |
cmp __x86_data_cache_size_half@GOTOFF(%ebx), %ecx |
| 810 |
# else |
| 811 |
cmp __x86_data_cache_size_half, %ecx |
| 812 |
# endif |
| 813 |
# endif |
| 814 |
jb L(sh_5_no_prefetch) |
| 815 |
|
| 816 |
lea -64(%ecx), %ecx |
| 817 |
|
| 818 |
.p2align 4 |
| 819 |
L(Shl5LoopStart): |
| 820 |
prefetcht0 0x1c0(%eax) |
| 821 |
prefetcht0 0x1c0(%edx) |
| 822 |
movaps 11(%eax), %xmm2 |
| 823 |
movaps 27(%eax), %xmm3 |
| 824 |
movaps 43(%eax), %xmm4 |
| 825 |
movaps 59(%eax), %xmm5 |
| 826 |
movaps %xmm5, %xmm7 |
| 827 |
palignr $5, %xmm4, %xmm5 |
| 828 |
palignr $5, %xmm3, %xmm4 |
| 829 |
movaps %xmm5, 48(%edx) |
| 830 |
palignr $5, %xmm2, %xmm3 |
| 831 |
lea 64(%eax), %eax |
| 832 |
palignr $5, %xmm1, %xmm2 |
| 833 |
movaps %xmm4, 32(%edx) |
| 834 |
movaps %xmm3, 16(%edx) |
| 835 |
movaps %xmm7, %xmm1 |
| 836 |
movaps %xmm2, (%edx) |
| 837 |
lea 64(%edx), %edx |
| 838 |
sub $64, %ecx |
| 839 |
ja L(Shl5LoopStart) |
| 840 |
|
| 841 |
L(Shl5LoopLeave): |
| 842 |
add $32, %ecx |
| 843 |
jle L(shl_end_0) |
| 844 |
|
| 845 |
movaps 11(%eax), %xmm2 |
| 846 |
movaps 27(%eax), %xmm3 |
| 847 |
palignr $5, %xmm2, %xmm3 |
| 848 |
palignr $5, %xmm1, %xmm2 |
| 849 |
movaps %xmm2, (%edx) |
| 850 |
movaps %xmm3, 16(%edx) |
| 851 |
lea 32(%edx, %ecx), %edx |
| 852 |
lea 32(%eax, %ecx), %eax |
| 853 |
POP (%edi) |
| 854 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 855 |
|
| 856 |
CFI_PUSH (%edi) |
| 857 |
|
| 858 |
.p2align 4 |
| 859 |
L(sh_5_no_prefetch): |
| 860 |
lea -32(%ecx), %ecx |
| 861 |
lea -5(%eax), %eax |
| 862 |
xor %edi, %edi |
| 863 |
|
| 864 |
.p2align 4 |
| 865 |
L(sh_5_no_prefetch_loop): |
| 866 |
movdqa 16(%eax, %edi), %xmm2 |
| 867 |
sub $32, %ecx |
| 868 |
movdqa 32(%eax, %edi), %xmm3 |
| 869 |
movdqa %xmm3, %xmm4 |
| 870 |
palignr $5, %xmm2, %xmm3 |
| 871 |
palignr $5, %xmm1, %xmm2 |
| 872 |
lea 32(%edi), %edi |
| 873 |
movdqa %xmm2, -32(%edx, %edi) |
| 874 |
movdqa %xmm3, -16(%edx, %edi) |
| 875 |
|
| 876 |
jb L(sh_5_end_no_prefetch_loop) |
| 877 |
|
| 878 |
movdqa 16(%eax, %edi), %xmm2 |
| 879 |
sub $32, %ecx |
| 880 |
movdqa 32(%eax, %edi), %xmm3 |
| 881 |
movdqa %xmm3, %xmm1 |
| 882 |
palignr $5, %xmm2, %xmm3 |
| 883 |
palignr $5, %xmm4, %xmm2 |
| 884 |
lea 32(%edi), %edi |
| 885 |
movdqa %xmm2, -32(%edx, %edi) |
| 886 |
movdqa %xmm3, -16(%edx, %edi) |
| 887 |
|
| 888 |
jae L(sh_5_no_prefetch_loop) |
| 889 |
|
| 890 |
L(sh_5_end_no_prefetch_loop): |
| 891 |
lea 32(%ecx), %ecx |
| 892 |
add %ecx, %edi |
| 893 |
add %edi, %edx |
| 894 |
lea 5(%edi, %eax), %eax |
| 895 |
POP (%edi) |
| 896 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 897 |
|
| 898 |
CFI_PUSH (%edi) |
| 899 |
|
| 900 |
.p2align 4 |
| 901 |
L(shl_6): |
| 902 |
# ifndef USE_AS_MEMMOVE |
| 903 |
movaps -6(%eax), %xmm1 |
| 904 |
# else |
| 905 |
movl DEST+4(%esp), %edi |
| 906 |
movaps -6(%eax), %xmm1 |
| 907 |
movdqu %xmm0, (%edi) |
| 908 |
# endif |
| 909 |
# ifdef DATA_CACHE_SIZE_HALF |
| 910 |
cmp $DATA_CACHE_SIZE_HALF, %ecx |
| 911 |
# else |
| 912 |
# ifdef PIC |
| 913 |
SETUP_PIC_REG(bx) |
| 914 |
add $_GLOBAL_OFFSET_TABLE_, %ebx |
| 915 |
cmp __x86_data_cache_size_half@GOTOFF(%ebx), %ecx |
| 916 |
# else |
| 917 |
cmp __x86_data_cache_size_half, %ecx |
| 918 |
# endif |
| 919 |
# endif |
| 920 |
jb L(sh_6_no_prefetch) |
| 921 |
|
| 922 |
lea -64(%ecx), %ecx |
| 923 |
|
| 924 |
.p2align 4 |
| 925 |
L(Shl6LoopStart): |
| 926 |
prefetcht0 0x1c0(%eax) |
| 927 |
prefetcht0 0x1c0(%edx) |
| 928 |
movaps 10(%eax), %xmm2 |
| 929 |
movaps 26(%eax), %xmm3 |
| 930 |
movaps 42(%eax), %xmm4 |
| 931 |
movaps 58(%eax), %xmm5 |
| 932 |
movaps %xmm5, %xmm7 |
| 933 |
palignr $6, %xmm4, %xmm5 |
| 934 |
palignr $6, %xmm3, %xmm4 |
| 935 |
movaps %xmm5, 48(%edx) |
| 936 |
palignr $6, %xmm2, %xmm3 |
| 937 |
lea 64(%eax), %eax |
| 938 |
palignr $6, %xmm1, %xmm2 |
| 939 |
movaps %xmm4, 32(%edx) |
| 940 |
movaps %xmm3, 16(%edx) |
| 941 |
movaps %xmm7, %xmm1 |
| 942 |
movaps %xmm2, (%edx) |
| 943 |
lea 64(%edx), %edx |
| 944 |
sub $64, %ecx |
| 945 |
ja L(Shl6LoopStart) |
| 946 |
|
| 947 |
L(Shl6LoopLeave): |
| 948 |
add $32, %ecx |
| 949 |
jle L(shl_end_0) |
| 950 |
|
| 951 |
movaps 10(%eax), %xmm2 |
| 952 |
movaps 26(%eax), %xmm3 |
| 953 |
palignr $6, %xmm2, %xmm3 |
| 954 |
palignr $6, %xmm1, %xmm2 |
| 955 |
movaps %xmm2, (%edx) |
| 956 |
movaps %xmm3, 16(%edx) |
| 957 |
lea 32(%edx, %ecx), %edx |
| 958 |
lea 32(%eax, %ecx), %eax |
| 959 |
POP (%edi) |
| 960 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 961 |
|
| 962 |
CFI_PUSH (%edi) |
| 963 |
|
| 964 |
.p2align 4 |
| 965 |
L(sh_6_no_prefetch): |
| 966 |
lea -32(%ecx), %ecx |
| 967 |
lea -6(%eax), %eax |
| 968 |
xor %edi, %edi |
| 969 |
|
| 970 |
.p2align 4 |
| 971 |
L(sh_6_no_prefetch_loop): |
| 972 |
movdqa 16(%eax, %edi), %xmm2 |
| 973 |
sub $32, %ecx |
| 974 |
movdqa 32(%eax, %edi), %xmm3 |
| 975 |
movdqa %xmm3, %xmm4 |
| 976 |
palignr $6, %xmm2, %xmm3 |
| 977 |
palignr $6, %xmm1, %xmm2 |
| 978 |
lea 32(%edi), %edi |
| 979 |
movdqa %xmm2, -32(%edx, %edi) |
| 980 |
movdqa %xmm3, -16(%edx, %edi) |
| 981 |
|
| 982 |
jb L(sh_6_end_no_prefetch_loop) |
| 983 |
|
| 984 |
movdqa 16(%eax, %edi), %xmm2 |
| 985 |
sub $32, %ecx |
| 986 |
movdqa 32(%eax, %edi), %xmm3 |
| 987 |
movdqa %xmm3, %xmm1 |
| 988 |
palignr $6, %xmm2, %xmm3 |
| 989 |
palignr $6, %xmm4, %xmm2 |
| 990 |
lea 32(%edi), %edi |
| 991 |
movdqa %xmm2, -32(%edx, %edi) |
| 992 |
movdqa %xmm3, -16(%edx, %edi) |
| 993 |
|
| 994 |
jae L(sh_6_no_prefetch_loop) |
| 995 |
|
| 996 |
L(sh_6_end_no_prefetch_loop): |
| 997 |
lea 32(%ecx), %ecx |
| 998 |
add %ecx, %edi |
| 999 |
add %edi, %edx |
| 1000 |
lea 6(%edi, %eax), %eax |
| 1001 |
POP (%edi) |
| 1002 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1003 |
|
| 1004 |
CFI_PUSH (%edi) |
| 1005 |
|
| 1006 |
.p2align 4 |
| 1007 |
L(shl_7): |
| 1008 |
# ifndef USE_AS_MEMMOVE |
| 1009 |
movaps -7(%eax), %xmm1 |
| 1010 |
# else |
| 1011 |
movl DEST+4(%esp), %edi |
| 1012 |
movaps -7(%eax), %xmm1 |
| 1013 |
movdqu %xmm0, (%edi) |
| 1014 |
# endif |
| 1015 |
# ifdef DATA_CACHE_SIZE_HALF |
| 1016 |
cmp $DATA_CACHE_SIZE_HALF, %ecx |
| 1017 |
# else |
| 1018 |
# ifdef PIC |
| 1019 |
SETUP_PIC_REG(bx) |
| 1020 |
add $_GLOBAL_OFFSET_TABLE_, %ebx |
| 1021 |
cmp __x86_data_cache_size_half@GOTOFF(%ebx), %ecx |
| 1022 |
# else |
| 1023 |
cmp __x86_data_cache_size_half, %ecx |
| 1024 |
# endif |
| 1025 |
# endif |
| 1026 |
jb L(sh_7_no_prefetch) |
| 1027 |
|
| 1028 |
lea -64(%ecx), %ecx |
| 1029 |
|
| 1030 |
.p2align 4 |
| 1031 |
L(Shl7LoopStart): |
| 1032 |
prefetcht0 0x1c0(%eax) |
| 1033 |
prefetcht0 0x1c0(%edx) |
| 1034 |
movaps 9(%eax), %xmm2 |
| 1035 |
movaps 25(%eax), %xmm3 |
| 1036 |
movaps 41(%eax), %xmm4 |
| 1037 |
movaps 57(%eax), %xmm5 |
| 1038 |
movaps %xmm5, %xmm7 |
| 1039 |
palignr $7, %xmm4, %xmm5 |
| 1040 |
palignr $7, %xmm3, %xmm4 |
| 1041 |
movaps %xmm5, 48(%edx) |
| 1042 |
palignr $7, %xmm2, %xmm3 |
| 1043 |
lea 64(%eax), %eax |
| 1044 |
palignr $7, %xmm1, %xmm2 |
| 1045 |
movaps %xmm4, 32(%edx) |
| 1046 |
movaps %xmm3, 16(%edx) |
| 1047 |
movaps %xmm7, %xmm1 |
| 1048 |
movaps %xmm2, (%edx) |
| 1049 |
lea 64(%edx), %edx |
| 1050 |
sub $64, %ecx |
| 1051 |
ja L(Shl7LoopStart) |
| 1052 |
|
| 1053 |
L(Shl7LoopLeave): |
| 1054 |
add $32, %ecx |
| 1055 |
jle L(shl_end_0) |
| 1056 |
|
| 1057 |
movaps 9(%eax), %xmm2 |
| 1058 |
movaps 25(%eax), %xmm3 |
| 1059 |
palignr $7, %xmm2, %xmm3 |
| 1060 |
palignr $7, %xmm1, %xmm2 |
| 1061 |
movaps %xmm2, (%edx) |
| 1062 |
movaps %xmm3, 16(%edx) |
| 1063 |
lea 32(%edx, %ecx), %edx |
| 1064 |
lea 32(%eax, %ecx), %eax |
| 1065 |
POP (%edi) |
| 1066 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1067 |
|
| 1068 |
CFI_PUSH (%edi) |
| 1069 |
|
| 1070 |
.p2align 4 |
| 1071 |
L(sh_7_no_prefetch): |
| 1072 |
lea -32(%ecx), %ecx |
| 1073 |
lea -7(%eax), %eax |
| 1074 |
xor %edi, %edi |
| 1075 |
|
| 1076 |
.p2align 4 |
| 1077 |
L(sh_7_no_prefetch_loop): |
| 1078 |
movdqa 16(%eax, %edi), %xmm2 |
| 1079 |
sub $32, %ecx |
| 1080 |
movdqa 32(%eax, %edi), %xmm3 |
| 1081 |
movdqa %xmm3, %xmm4 |
| 1082 |
palignr $7, %xmm2, %xmm3 |
| 1083 |
palignr $7, %xmm1, %xmm2 |
| 1084 |
lea 32(%edi), %edi |
| 1085 |
movdqa %xmm2, -32(%edx, %edi) |
| 1086 |
movdqa %xmm3, -16(%edx, %edi) |
| 1087 |
jb L(sh_7_end_no_prefetch_loop) |
| 1088 |
|
| 1089 |
movdqa 16(%eax, %edi), %xmm2 |
| 1090 |
sub $32, %ecx |
| 1091 |
movdqa 32(%eax, %edi), %xmm3 |
| 1092 |
movdqa %xmm3, %xmm1 |
| 1093 |
palignr $7, %xmm2, %xmm3 |
| 1094 |
palignr $7, %xmm4, %xmm2 |
| 1095 |
lea 32(%edi), %edi |
| 1096 |
movdqa %xmm2, -32(%edx, %edi) |
| 1097 |
movdqa %xmm3, -16(%edx, %edi) |
| 1098 |
jae L(sh_7_no_prefetch_loop) |
| 1099 |
|
| 1100 |
L(sh_7_end_no_prefetch_loop): |
| 1101 |
lea 32(%ecx), %ecx |
| 1102 |
add %ecx, %edi |
| 1103 |
add %edi, %edx |
| 1104 |
lea 7(%edi, %eax), %eax |
| 1105 |
POP (%edi) |
| 1106 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1107 |
|
| 1108 |
CFI_PUSH (%edi) |
| 1109 |
|
| 1110 |
.p2align 4 |
| 1111 |
L(shl_8): |
| 1112 |
# ifndef USE_AS_MEMMOVE |
| 1113 |
movaps -8(%eax), %xmm1 |
| 1114 |
# else |
| 1115 |
movl DEST+4(%esp), %edi |
| 1116 |
movaps -8(%eax), %xmm1 |
| 1117 |
movdqu %xmm0, (%edi) |
| 1118 |
# endif |
| 1119 |
# ifdef DATA_CACHE_SIZE_HALF |
| 1120 |
cmp $DATA_CACHE_SIZE_HALF, %ecx |
| 1121 |
# else |
| 1122 |
# ifdef PIC |
| 1123 |
SETUP_PIC_REG(bx) |
| 1124 |
add $_GLOBAL_OFFSET_TABLE_, %ebx |
| 1125 |
cmp __x86_data_cache_size_half@GOTOFF(%ebx), %ecx |
| 1126 |
# else |
| 1127 |
cmp __x86_data_cache_size_half, %ecx |
| 1128 |
# endif |
| 1129 |
# endif |
| 1130 |
jb L(sh_8_no_prefetch) |
| 1131 |
|
| 1132 |
lea -64(%ecx), %ecx |
| 1133 |
|
| 1134 |
.p2align 4 |
| 1135 |
L(Shl8LoopStart): |
| 1136 |
prefetcht0 0x1c0(%eax) |
| 1137 |
prefetcht0 0x1c0(%edx) |
| 1138 |
movaps 8(%eax), %xmm2 |
| 1139 |
movaps 24(%eax), %xmm3 |
| 1140 |
movaps 40(%eax), %xmm4 |
| 1141 |
movaps 56(%eax), %xmm5 |
| 1142 |
movaps %xmm5, %xmm7 |
| 1143 |
palignr $8, %xmm4, %xmm5 |
| 1144 |
palignr $8, %xmm3, %xmm4 |
| 1145 |
movaps %xmm5, 48(%edx) |
| 1146 |
palignr $8, %xmm2, %xmm3 |
| 1147 |
lea 64(%eax), %eax |
| 1148 |
palignr $8, %xmm1, %xmm2 |
| 1149 |
movaps %xmm4, 32(%edx) |
| 1150 |
movaps %xmm3, 16(%edx) |
| 1151 |
movaps %xmm7, %xmm1 |
| 1152 |
movaps %xmm2, (%edx) |
| 1153 |
lea 64(%edx), %edx |
| 1154 |
sub $64, %ecx |
| 1155 |
ja L(Shl8LoopStart) |
| 1156 |
|
| 1157 |
L(LoopLeave8): |
| 1158 |
add $32, %ecx |
| 1159 |
jle L(shl_end_0) |
| 1160 |
|
| 1161 |
movaps 8(%eax), %xmm2 |
| 1162 |
movaps 24(%eax), %xmm3 |
| 1163 |
palignr $8, %xmm2, %xmm3 |
| 1164 |
palignr $8, %xmm1, %xmm2 |
| 1165 |
movaps %xmm2, (%edx) |
| 1166 |
movaps %xmm3, 16(%edx) |
| 1167 |
lea 32(%edx, %ecx), %edx |
| 1168 |
lea 32(%eax, %ecx), %eax |
| 1169 |
POP (%edi) |
| 1170 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1171 |
|
| 1172 |
CFI_PUSH (%edi) |
| 1173 |
|
| 1174 |
.p2align 4 |
| 1175 |
L(sh_8_no_prefetch): |
| 1176 |
lea -32(%ecx), %ecx |
| 1177 |
lea -8(%eax), %eax |
| 1178 |
xor %edi, %edi |
| 1179 |
|
| 1180 |
.p2align 4 |
| 1181 |
L(sh_8_no_prefetch_loop): |
| 1182 |
movdqa 16(%eax, %edi), %xmm2 |
| 1183 |
sub $32, %ecx |
| 1184 |
movdqa 32(%eax, %edi), %xmm3 |
| 1185 |
movdqa %xmm3, %xmm4 |
| 1186 |
palignr $8, %xmm2, %xmm3 |
| 1187 |
palignr $8, %xmm1, %xmm2 |
| 1188 |
lea 32(%edi), %edi |
| 1189 |
movdqa %xmm2, -32(%edx, %edi) |
| 1190 |
movdqa %xmm3, -16(%edx, %edi) |
| 1191 |
jb L(sh_8_end_no_prefetch_loop) |
| 1192 |
|
| 1193 |
movdqa 16(%eax, %edi), %xmm2 |
| 1194 |
sub $32, %ecx |
| 1195 |
movdqa 32(%eax, %edi), %xmm3 |
| 1196 |
movdqa %xmm3, %xmm1 |
| 1197 |
palignr $8, %xmm2, %xmm3 |
| 1198 |
palignr $8, %xmm4, %xmm2 |
| 1199 |
lea 32(%edi), %edi |
| 1200 |
movdqa %xmm2, -32(%edx, %edi) |
| 1201 |
movdqa %xmm3, -16(%edx, %edi) |
| 1202 |
jae L(sh_8_no_prefetch_loop) |
| 1203 |
|
| 1204 |
L(sh_8_end_no_prefetch_loop): |
| 1205 |
lea 32(%ecx), %ecx |
| 1206 |
add %ecx, %edi |
| 1207 |
add %edi, %edx |
| 1208 |
lea 8(%edi, %eax), %eax |
| 1209 |
POP (%edi) |
| 1210 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1211 |
|
| 1212 |
CFI_PUSH (%edi) |
| 1213 |
|
| 1214 |
.p2align 4 |
| 1215 |
L(shl_9): |
| 1216 |
# ifndef USE_AS_MEMMOVE |
| 1217 |
movaps -9(%eax), %xmm1 |
| 1218 |
# else |
| 1219 |
movl DEST+4(%esp), %edi |
| 1220 |
movaps -9(%eax), %xmm1 |
| 1221 |
movdqu %xmm0, (%edi) |
| 1222 |
# endif |
| 1223 |
# ifdef DATA_CACHE_SIZE_HALF |
| 1224 |
cmp $DATA_CACHE_SIZE_HALF, %ecx |
| 1225 |
# else |
| 1226 |
# ifdef PIC |
| 1227 |
SETUP_PIC_REG(bx) |
| 1228 |
add $_GLOBAL_OFFSET_TABLE_, %ebx |
| 1229 |
cmp __x86_data_cache_size_half@GOTOFF(%ebx), %ecx |
| 1230 |
# else |
| 1231 |
cmp __x86_data_cache_size_half, %ecx |
| 1232 |
# endif |
| 1233 |
# endif |
| 1234 |
jb L(sh_9_no_prefetch) |
| 1235 |
|
| 1236 |
lea -64(%ecx), %ecx |
| 1237 |
|
| 1238 |
.p2align 4 |
| 1239 |
L(Shl9LoopStart): |
| 1240 |
prefetcht0 0x1c0(%eax) |
| 1241 |
prefetcht0 0x1c0(%edx) |
| 1242 |
movaps 7(%eax), %xmm2 |
| 1243 |
movaps 23(%eax), %xmm3 |
| 1244 |
movaps 39(%eax), %xmm4 |
| 1245 |
movaps 55(%eax), %xmm5 |
| 1246 |
movaps %xmm5, %xmm7 |
| 1247 |
palignr $9, %xmm4, %xmm5 |
| 1248 |
palignr $9, %xmm3, %xmm4 |
| 1249 |
movaps %xmm5, 48(%edx) |
| 1250 |
palignr $9, %xmm2, %xmm3 |
| 1251 |
lea 64(%eax), %eax |
| 1252 |
palignr $9, %xmm1, %xmm2 |
| 1253 |
movaps %xmm4, 32(%edx) |
| 1254 |
movaps %xmm3, 16(%edx) |
| 1255 |
movaps %xmm7, %xmm1 |
| 1256 |
movaps %xmm2, (%edx) |
| 1257 |
lea 64(%edx), %edx |
| 1258 |
sub $64, %ecx |
| 1259 |
ja L(Shl9LoopStart) |
| 1260 |
|
| 1261 |
L(Shl9LoopLeave): |
| 1262 |
add $32, %ecx |
| 1263 |
jle L(shl_end_0) |
| 1264 |
|
| 1265 |
movaps 7(%eax), %xmm2 |
| 1266 |
movaps 23(%eax), %xmm3 |
| 1267 |
palignr $9, %xmm2, %xmm3 |
| 1268 |
palignr $9, %xmm1, %xmm2 |
| 1269 |
|
| 1270 |
movaps %xmm2, (%edx) |
| 1271 |
movaps %xmm3, 16(%edx) |
| 1272 |
lea 32(%edx, %ecx), %edx |
| 1273 |
lea 32(%eax, %ecx), %eax |
| 1274 |
POP (%edi) |
| 1275 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1276 |
|
| 1277 |
CFI_PUSH (%edi) |
| 1278 |
|
| 1279 |
.p2align 4 |
| 1280 |
L(sh_9_no_prefetch): |
| 1281 |
lea -32(%ecx), %ecx |
| 1282 |
lea -9(%eax), %eax |
| 1283 |
xor %edi, %edi |
| 1284 |
|
| 1285 |
.p2align 4 |
| 1286 |
L(sh_9_no_prefetch_loop): |
| 1287 |
movdqa 16(%eax, %edi), %xmm2 |
| 1288 |
sub $32, %ecx |
| 1289 |
movdqa 32(%eax, %edi), %xmm3 |
| 1290 |
movdqa %xmm3, %xmm4 |
| 1291 |
palignr $9, %xmm2, %xmm3 |
| 1292 |
palignr $9, %xmm1, %xmm2 |
| 1293 |
lea 32(%edi), %edi |
| 1294 |
movdqa %xmm2, -32(%edx, %edi) |
| 1295 |
movdqa %xmm3, -16(%edx, %edi) |
| 1296 |
jb L(sh_9_end_no_prefetch_loop) |
| 1297 |
|
| 1298 |
movdqa 16(%eax, %edi), %xmm2 |
| 1299 |
sub $32, %ecx |
| 1300 |
movdqa 32(%eax, %edi), %xmm3 |
| 1301 |
movdqa %xmm3, %xmm1 |
| 1302 |
palignr $9, %xmm2, %xmm3 |
| 1303 |
palignr $9, %xmm4, %xmm2 |
| 1304 |
lea 32(%edi), %edi |
| 1305 |
movdqa %xmm2, -32(%edx, %edi) |
| 1306 |
movdqa %xmm3, -16(%edx, %edi) |
| 1307 |
jae L(sh_9_no_prefetch_loop) |
| 1308 |
|
| 1309 |
L(sh_9_end_no_prefetch_loop): |
| 1310 |
lea 32(%ecx), %ecx |
| 1311 |
add %ecx, %edi |
| 1312 |
add %edi, %edx |
| 1313 |
lea 9(%edi, %eax), %eax |
| 1314 |
POP (%edi) |
| 1315 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1316 |
|
| 1317 |
CFI_PUSH (%edi) |
| 1318 |
|
| 1319 |
.p2align 4 |
| 1320 |
L(shl_10): |
| 1321 |
# ifndef USE_AS_MEMMOVE |
| 1322 |
movaps -10(%eax), %xmm1 |
| 1323 |
# else |
| 1324 |
movl DEST+4(%esp), %edi |
| 1325 |
movaps -10(%eax), %xmm1 |
| 1326 |
movdqu %xmm0, (%edi) |
| 1327 |
# endif |
| 1328 |
# ifdef DATA_CACHE_SIZE_HALF |
| 1329 |
cmp $DATA_CACHE_SIZE_HALF, %ecx |
| 1330 |
# else |
| 1331 |
# ifdef PIC |
| 1332 |
SETUP_PIC_REG(bx) |
| 1333 |
add $_GLOBAL_OFFSET_TABLE_, %ebx |
| 1334 |
cmp __x86_data_cache_size_half@GOTOFF(%ebx), %ecx |
| 1335 |
# else |
| 1336 |
cmp __x86_data_cache_size_half, %ecx |
| 1337 |
# endif |
| 1338 |
# endif |
| 1339 |
jb L(sh_10_no_prefetch) |
| 1340 |
|
| 1341 |
lea -64(%ecx), %ecx |
| 1342 |
|
| 1343 |
.p2align 4 |
| 1344 |
L(Shl10LoopStart): |
| 1345 |
prefetcht0 0x1c0(%eax) |
| 1346 |
prefetcht0 0x1c0(%edx) |
| 1347 |
movaps 6(%eax), %xmm2 |
| 1348 |
movaps 22(%eax), %xmm3 |
| 1349 |
movaps 38(%eax), %xmm4 |
| 1350 |
movaps 54(%eax), %xmm5 |
| 1351 |
movaps %xmm5, %xmm7 |
| 1352 |
palignr $10, %xmm4, %xmm5 |
| 1353 |
palignr $10, %xmm3, %xmm4 |
| 1354 |
movaps %xmm5, 48(%edx) |
| 1355 |
palignr $10, %xmm2, %xmm3 |
| 1356 |
lea 64(%eax), %eax |
| 1357 |
palignr $10, %xmm1, %xmm2 |
| 1358 |
movaps %xmm4, 32(%edx) |
| 1359 |
movaps %xmm3, 16(%edx) |
| 1360 |
movaps %xmm7, %xmm1 |
| 1361 |
movaps %xmm2, (%edx) |
| 1362 |
lea 64(%edx), %edx |
| 1363 |
sub $64, %ecx |
| 1364 |
ja L(Shl10LoopStart) |
| 1365 |
|
| 1366 |
L(Shl10LoopLeave): |
| 1367 |
add $32, %ecx |
| 1368 |
jle L(shl_end_0) |
| 1369 |
|
| 1370 |
movaps 6(%eax), %xmm2 |
| 1371 |
movaps 22(%eax), %xmm3 |
| 1372 |
palignr $10, %xmm2, %xmm3 |
| 1373 |
palignr $10, %xmm1, %xmm2 |
| 1374 |
|
| 1375 |
movaps %xmm2, (%edx) |
| 1376 |
movaps %xmm3, 16(%edx) |
| 1377 |
lea 32(%edx, %ecx), %edx |
| 1378 |
lea 32(%eax, %ecx), %eax |
| 1379 |
POP (%edi) |
| 1380 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1381 |
|
| 1382 |
CFI_PUSH (%edi) |
| 1383 |
|
| 1384 |
.p2align 4 |
| 1385 |
L(sh_10_no_prefetch): |
| 1386 |
lea -32(%ecx), %ecx |
| 1387 |
lea -10(%eax), %eax |
| 1388 |
xor %edi, %edi |
| 1389 |
|
| 1390 |
.p2align 4 |
| 1391 |
L(sh_10_no_prefetch_loop): |
| 1392 |
movdqa 16(%eax, %edi), %xmm2 |
| 1393 |
sub $32, %ecx |
| 1394 |
movdqa 32(%eax, %edi), %xmm3 |
| 1395 |
movdqa %xmm3, %xmm4 |
| 1396 |
palignr $10, %xmm2, %xmm3 |
| 1397 |
palignr $10, %xmm1, %xmm2 |
| 1398 |
lea 32(%edi), %edi |
| 1399 |
movdqa %xmm2, -32(%edx, %edi) |
| 1400 |
movdqa %xmm3, -16(%edx, %edi) |
| 1401 |
jb L(sh_10_end_no_prefetch_loop) |
| 1402 |
|
| 1403 |
movdqa 16(%eax, %edi), %xmm2 |
| 1404 |
sub $32, %ecx |
| 1405 |
movdqa 32(%eax, %edi), %xmm3 |
| 1406 |
movdqa %xmm3, %xmm1 |
| 1407 |
palignr $10, %xmm2, %xmm3 |
| 1408 |
palignr $10, %xmm4, %xmm2 |
| 1409 |
lea 32(%edi), %edi |
| 1410 |
movdqa %xmm2, -32(%edx, %edi) |
| 1411 |
movdqa %xmm3, -16(%edx, %edi) |
| 1412 |
jae L(sh_10_no_prefetch_loop) |
| 1413 |
|
| 1414 |
L(sh_10_end_no_prefetch_loop): |
| 1415 |
lea 32(%ecx), %ecx |
| 1416 |
add %ecx, %edi |
| 1417 |
add %edi, %edx |
| 1418 |
lea 10(%edi, %eax), %eax |
| 1419 |
POP (%edi) |
| 1420 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1421 |
|
| 1422 |
CFI_PUSH (%edi) |
| 1423 |
|
| 1424 |
.p2align 4 |
| 1425 |
L(shl_11): |
| 1426 |
# ifndef USE_AS_MEMMOVE |
| 1427 |
movaps -11(%eax), %xmm1 |
| 1428 |
# else |
| 1429 |
movl DEST+4(%esp), %edi |
| 1430 |
movaps -11(%eax), %xmm1 |
| 1431 |
movdqu %xmm0, (%edi) |
| 1432 |
# endif |
| 1433 |
# ifdef DATA_CACHE_SIZE_HALF |
| 1434 |
cmp $DATA_CACHE_SIZE_HALF, %ecx |
| 1435 |
# else |
| 1436 |
# ifdef PIC |
| 1437 |
SETUP_PIC_REG(bx) |
| 1438 |
add $_GLOBAL_OFFSET_TABLE_, %ebx |
| 1439 |
cmp __x86_data_cache_size_half@GOTOFF(%ebx), %ecx |
| 1440 |
# else |
| 1441 |
cmp __x86_data_cache_size_half, %ecx |
| 1442 |
# endif |
| 1443 |
# endif |
| 1444 |
jb L(sh_11_no_prefetch) |
| 1445 |
|
| 1446 |
lea -64(%ecx), %ecx |
| 1447 |
|
| 1448 |
.p2align 4 |
| 1449 |
L(Shl11LoopStart): |
| 1450 |
prefetcht0 0x1c0(%eax) |
| 1451 |
prefetcht0 0x1c0(%edx) |
| 1452 |
movaps 5(%eax), %xmm2 |
| 1453 |
movaps 21(%eax), %xmm3 |
| 1454 |
movaps 37(%eax), %xmm4 |
| 1455 |
movaps 53(%eax), %xmm5 |
| 1456 |
movaps %xmm5, %xmm7 |
| 1457 |
palignr $11, %xmm4, %xmm5 |
| 1458 |
palignr $11, %xmm3, %xmm4 |
| 1459 |
movaps %xmm5, 48(%edx) |
| 1460 |
palignr $11, %xmm2, %xmm3 |
| 1461 |
lea 64(%eax), %eax |
| 1462 |
palignr $11, %xmm1, %xmm2 |
| 1463 |
movaps %xmm4, 32(%edx) |
| 1464 |
movaps %xmm3, 16(%edx) |
| 1465 |
movaps %xmm7, %xmm1 |
| 1466 |
movaps %xmm2, (%edx) |
| 1467 |
lea 64(%edx), %edx |
| 1468 |
sub $64, %ecx |
| 1469 |
ja L(Shl11LoopStart) |
| 1470 |
|
| 1471 |
L(Shl11LoopLeave): |
| 1472 |
add $32, %ecx |
| 1473 |
jle L(shl_end_0) |
| 1474 |
|
| 1475 |
movaps 5(%eax), %xmm2 |
| 1476 |
movaps 21(%eax), %xmm3 |
| 1477 |
palignr $11, %xmm2, %xmm3 |
| 1478 |
palignr $11, %xmm1, %xmm2 |
| 1479 |
|
| 1480 |
movaps %xmm2, (%edx) |
| 1481 |
movaps %xmm3, 16(%edx) |
| 1482 |
lea 32(%edx, %ecx), %edx |
| 1483 |
lea 32(%eax, %ecx), %eax |
| 1484 |
POP (%edi) |
| 1485 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1486 |
|
| 1487 |
CFI_PUSH (%edi) |
| 1488 |
|
| 1489 |
.p2align 4 |
| 1490 |
L(sh_11_no_prefetch): |
| 1491 |
lea -32(%ecx), %ecx |
| 1492 |
lea -11(%eax), %eax |
| 1493 |
xor %edi, %edi |
| 1494 |
|
| 1495 |
.p2align 4 |
| 1496 |
L(sh_11_no_prefetch_loop): |
| 1497 |
movdqa 16(%eax, %edi), %xmm2 |
| 1498 |
sub $32, %ecx |
| 1499 |
movdqa 32(%eax, %edi), %xmm3 |
| 1500 |
movdqa %xmm3, %xmm4 |
| 1501 |
palignr $11, %xmm2, %xmm3 |
| 1502 |
palignr $11, %xmm1, %xmm2 |
| 1503 |
lea 32(%edi), %edi |
| 1504 |
movdqa %xmm2, -32(%edx, %edi) |
| 1505 |
movdqa %xmm3, -16(%edx, %edi) |
| 1506 |
jb L(sh_11_end_no_prefetch_loop) |
| 1507 |
|
| 1508 |
movdqa 16(%eax, %edi), %xmm2 |
| 1509 |
sub $32, %ecx |
| 1510 |
movdqa 32(%eax, %edi), %xmm3 |
| 1511 |
movdqa %xmm3, %xmm1 |
| 1512 |
palignr $11, %xmm2, %xmm3 |
| 1513 |
palignr $11, %xmm4, %xmm2 |
| 1514 |
lea 32(%edi), %edi |
| 1515 |
movdqa %xmm2, -32(%edx, %edi) |
| 1516 |
movdqa %xmm3, -16(%edx, %edi) |
| 1517 |
jae L(sh_11_no_prefetch_loop) |
| 1518 |
|
| 1519 |
L(sh_11_end_no_prefetch_loop): |
| 1520 |
lea 32(%ecx), %ecx |
| 1521 |
add %ecx, %edi |
| 1522 |
add %edi, %edx |
| 1523 |
lea 11(%edi, %eax), %eax |
| 1524 |
POP (%edi) |
| 1525 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1526 |
|
| 1527 |
CFI_PUSH (%edi) |
| 1528 |
|
| 1529 |
.p2align 4 |
| 1530 |
L(shl_12): |
| 1531 |
# ifndef USE_AS_MEMMOVE |
| 1532 |
movaps -12(%eax), %xmm1 |
| 1533 |
# else |
| 1534 |
movl DEST+4(%esp), %edi |
| 1535 |
movaps -12(%eax), %xmm1 |
| 1536 |
movdqu %xmm0, (%edi) |
| 1537 |
# endif |
| 1538 |
# ifdef DATA_CACHE_SIZE_HALF |
| 1539 |
cmp $DATA_CACHE_SIZE_HALF, %ecx |
| 1540 |
# else |
| 1541 |
# ifdef PIC |
| 1542 |
SETUP_PIC_REG(bx) |
| 1543 |
add $_GLOBAL_OFFSET_TABLE_, %ebx |
| 1544 |
cmp __x86_data_cache_size_half@GOTOFF(%ebx), %ecx |
| 1545 |
# else |
| 1546 |
cmp __x86_data_cache_size_half, %ecx |
| 1547 |
# endif |
| 1548 |
# endif |
| 1549 |
jb L(sh_12_no_prefetch) |
| 1550 |
|
| 1551 |
lea -64(%ecx), %ecx |
| 1552 |
|
| 1553 |
.p2align 4 |
| 1554 |
L(Shl12LoopStart): |
| 1555 |
prefetcht0 0x1c0(%eax) |
| 1556 |
prefetcht0 0x1c0(%edx) |
| 1557 |
movaps 4(%eax), %xmm2 |
| 1558 |
movaps 20(%eax), %xmm3 |
| 1559 |
movaps 36(%eax), %xmm4 |
| 1560 |
movaps 52(%eax), %xmm5 |
| 1561 |
movaps %xmm5, %xmm7 |
| 1562 |
palignr $12, %xmm4, %xmm5 |
| 1563 |
palignr $12, %xmm3, %xmm4 |
| 1564 |
movaps %xmm5, 48(%edx) |
| 1565 |
palignr $12, %xmm2, %xmm3 |
| 1566 |
lea 64(%eax), %eax |
| 1567 |
palignr $12, %xmm1, %xmm2 |
| 1568 |
movaps %xmm4, 32(%edx) |
| 1569 |
movaps %xmm3, 16(%edx) |
| 1570 |
movaps %xmm7, %xmm1 |
| 1571 |
movaps %xmm2, (%edx) |
| 1572 |
lea 64(%edx), %edx |
| 1573 |
sub $64, %ecx |
| 1574 |
ja L(Shl12LoopStart) |
| 1575 |
|
| 1576 |
L(Shl12LoopLeave): |
| 1577 |
add $32, %ecx |
| 1578 |
jle L(shl_end_0) |
| 1579 |
|
| 1580 |
movaps 4(%eax), %xmm2 |
| 1581 |
movaps 20(%eax), %xmm3 |
| 1582 |
palignr $12, %xmm2, %xmm3 |
| 1583 |
palignr $12, %xmm1, %xmm2 |
| 1584 |
|
| 1585 |
movaps %xmm2, (%edx) |
| 1586 |
movaps %xmm3, 16(%edx) |
| 1587 |
lea 32(%edx, %ecx), %edx |
| 1588 |
lea 32(%eax, %ecx), %eax |
| 1589 |
POP (%edi) |
| 1590 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1591 |
|
| 1592 |
CFI_PUSH (%edi) |
| 1593 |
|
| 1594 |
.p2align 4 |
| 1595 |
L(sh_12_no_prefetch): |
| 1596 |
lea -32(%ecx), %ecx |
| 1597 |
lea -12(%eax), %eax |
| 1598 |
xor %edi, %edi |
| 1599 |
|
| 1600 |
.p2align 4 |
| 1601 |
L(sh_12_no_prefetch_loop): |
| 1602 |
movdqa 16(%eax, %edi), %xmm2 |
| 1603 |
sub $32, %ecx |
| 1604 |
movdqa 32(%eax, %edi), %xmm3 |
| 1605 |
movdqa %xmm3, %xmm4 |
| 1606 |
palignr $12, %xmm2, %xmm3 |
| 1607 |
palignr $12, %xmm1, %xmm2 |
| 1608 |
lea 32(%edi), %edi |
| 1609 |
movdqa %xmm2, -32(%edx, %edi) |
| 1610 |
movdqa %xmm3, -16(%edx, %edi) |
| 1611 |
jb L(sh_12_end_no_prefetch_loop) |
| 1612 |
|
| 1613 |
movdqa 16(%eax, %edi), %xmm2 |
| 1614 |
sub $32, %ecx |
| 1615 |
movdqa 32(%eax, %edi), %xmm3 |
| 1616 |
movdqa %xmm3, %xmm1 |
| 1617 |
palignr $12, %xmm2, %xmm3 |
| 1618 |
palignr $12, %xmm4, %xmm2 |
| 1619 |
lea 32(%edi), %edi |
| 1620 |
movdqa %xmm2, -32(%edx, %edi) |
| 1621 |
movdqa %xmm3, -16(%edx, %edi) |
| 1622 |
jae L(sh_12_no_prefetch_loop) |
| 1623 |
|
| 1624 |
L(sh_12_end_no_prefetch_loop): |
| 1625 |
lea 32(%ecx), %ecx |
| 1626 |
add %ecx, %edi |
| 1627 |
add %edi, %edx |
| 1628 |
lea 12(%edi, %eax), %eax |
| 1629 |
POP (%edi) |
| 1630 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1631 |
|
| 1632 |
CFI_PUSH (%edi) |
| 1633 |
|
| 1634 |
.p2align 4 |
| 1635 |
L(shl_13): |
| 1636 |
# ifndef USE_AS_MEMMOVE |
| 1637 |
movaps -13(%eax), %xmm1 |
| 1638 |
# else |
| 1639 |
movl DEST+4(%esp), %edi |
| 1640 |
movaps -13(%eax), %xmm1 |
| 1641 |
movdqu %xmm0, (%edi) |
| 1642 |
# endif |
| 1643 |
# ifdef DATA_CACHE_SIZE_HALF |
| 1644 |
cmp $DATA_CACHE_SIZE_HALF, %ecx |
| 1645 |
# else |
| 1646 |
# ifdef PIC |
| 1647 |
SETUP_PIC_REG(bx) |
| 1648 |
add $_GLOBAL_OFFSET_TABLE_, %ebx |
| 1649 |
cmp __x86_data_cache_size_half@GOTOFF(%ebx), %ecx |
| 1650 |
# else |
| 1651 |
cmp __x86_data_cache_size_half, %ecx |
| 1652 |
# endif |
| 1653 |
# endif |
| 1654 |
jb L(sh_13_no_prefetch) |
| 1655 |
|
| 1656 |
lea -64(%ecx), %ecx |
| 1657 |
|
| 1658 |
.p2align 4 |
| 1659 |
L(Shl13LoopStart): |
| 1660 |
prefetcht0 0x1c0(%eax) |
| 1661 |
prefetcht0 0x1c0(%edx) |
| 1662 |
movaps 3(%eax), %xmm2 |
| 1663 |
movaps 19(%eax), %xmm3 |
| 1664 |
movaps 35(%eax), %xmm4 |
| 1665 |
movaps 51(%eax), %xmm5 |
| 1666 |
movaps %xmm5, %xmm7 |
| 1667 |
palignr $13, %xmm4, %xmm5 |
| 1668 |
palignr $13, %xmm3, %xmm4 |
| 1669 |
movaps %xmm5, 48(%edx) |
| 1670 |
palignr $13, %xmm2, %xmm3 |
| 1671 |
lea 64(%eax), %eax |
| 1672 |
palignr $13, %xmm1, %xmm2 |
| 1673 |
movaps %xmm4, 32(%edx) |
| 1674 |
movaps %xmm3, 16(%edx) |
| 1675 |
movaps %xmm7, %xmm1 |
| 1676 |
movaps %xmm2, (%edx) |
| 1677 |
lea 64(%edx), %edx |
| 1678 |
sub $64, %ecx |
| 1679 |
ja L(Shl13LoopStart) |
| 1680 |
|
| 1681 |
L(Shl13LoopLeave): |
| 1682 |
add $32, %ecx |
| 1683 |
jle L(shl_end_0) |
| 1684 |
|
| 1685 |
movaps 3(%eax), %xmm2 |
| 1686 |
movaps 19(%eax), %xmm3 |
| 1687 |
palignr $13, %xmm2, %xmm3 |
| 1688 |
palignr $13, %xmm1, %xmm2 |
| 1689 |
|
| 1690 |
movaps %xmm2, (%edx) |
| 1691 |
movaps %xmm3, 16(%edx) |
| 1692 |
lea 32(%edx, %ecx), %edx |
| 1693 |
lea 32(%eax, %ecx), %eax |
| 1694 |
POP (%edi) |
| 1695 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1696 |
|
| 1697 |
CFI_PUSH (%edi) |
| 1698 |
|
| 1699 |
.p2align 4 |
| 1700 |
L(sh_13_no_prefetch): |
| 1701 |
lea -32(%ecx), %ecx |
| 1702 |
lea -13(%eax), %eax |
| 1703 |
xor %edi, %edi |
| 1704 |
|
| 1705 |
.p2align 4 |
| 1706 |
L(sh_13_no_prefetch_loop): |
| 1707 |
movdqa 16(%eax, %edi), %xmm2 |
| 1708 |
sub $32, %ecx |
| 1709 |
movdqa 32(%eax, %edi), %xmm3 |
| 1710 |
movdqa %xmm3, %xmm4 |
| 1711 |
palignr $13, %xmm2, %xmm3 |
| 1712 |
palignr $13, %xmm1, %xmm2 |
| 1713 |
lea 32(%edi), %edi |
| 1714 |
movdqa %xmm2, -32(%edx, %edi) |
| 1715 |
movdqa %xmm3, -16(%edx, %edi) |
| 1716 |
jb L(sh_13_end_no_prefetch_loop) |
| 1717 |
|
| 1718 |
movdqa 16(%eax, %edi), %xmm2 |
| 1719 |
sub $32, %ecx |
| 1720 |
movdqa 32(%eax, %edi), %xmm3 |
| 1721 |
movdqa %xmm3, %xmm1 |
| 1722 |
palignr $13, %xmm2, %xmm3 |
| 1723 |
palignr $13, %xmm4, %xmm2 |
| 1724 |
lea 32(%edi), %edi |
| 1725 |
movdqa %xmm2, -32(%edx, %edi) |
| 1726 |
movdqa %xmm3, -16(%edx, %edi) |
| 1727 |
jae L(sh_13_no_prefetch_loop) |
| 1728 |
|
| 1729 |
L(sh_13_end_no_prefetch_loop): |
| 1730 |
lea 32(%ecx), %ecx |
| 1731 |
add %ecx, %edi |
| 1732 |
add %edi, %edx |
| 1733 |
lea 13(%edi, %eax), %eax |
| 1734 |
POP (%edi) |
| 1735 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1736 |
|
| 1737 |
CFI_PUSH (%edi) |
| 1738 |
|
| 1739 |
.p2align 4 |
| 1740 |
L(shl_14): |
| 1741 |
# ifndef USE_AS_MEMMOVE |
| 1742 |
movaps -14(%eax), %xmm1 |
| 1743 |
# else |
| 1744 |
movl DEST+4(%esp), %edi |
| 1745 |
movaps -14(%eax), %xmm1 |
| 1746 |
movdqu %xmm0, (%edi) |
| 1747 |
# endif |
| 1748 |
# ifdef DATA_CACHE_SIZE_HALF |
| 1749 |
cmp $DATA_CACHE_SIZE_HALF, %ecx |
| 1750 |
# else |
| 1751 |
# ifdef PIC |
| 1752 |
SETUP_PIC_REG(bx) |
| 1753 |
add $_GLOBAL_OFFSET_TABLE_, %ebx |
| 1754 |
cmp __x86_data_cache_size_half@GOTOFF(%ebx), %ecx |
| 1755 |
# else |
| 1756 |
cmp __x86_data_cache_size_half, %ecx |
| 1757 |
# endif |
| 1758 |
# endif |
| 1759 |
jb L(sh_14_no_prefetch) |
| 1760 |
|
| 1761 |
lea -64(%ecx), %ecx |
| 1762 |
|
| 1763 |
.p2align 4 |
| 1764 |
L(Shl14LoopStart): |
| 1765 |
prefetcht0 0x1c0(%eax) |
| 1766 |
prefetcht0 0x1c0(%edx) |
| 1767 |
movaps 2(%eax), %xmm2 |
| 1768 |
movaps 18(%eax), %xmm3 |
| 1769 |
movaps 34(%eax), %xmm4 |
| 1770 |
movaps 50(%eax), %xmm5 |
| 1771 |
movaps %xmm5, %xmm7 |
| 1772 |
palignr $14, %xmm4, %xmm5 |
| 1773 |
palignr $14, %xmm3, %xmm4 |
| 1774 |
movaps %xmm5, 48(%edx) |
| 1775 |
palignr $14, %xmm2, %xmm3 |
| 1776 |
lea 64(%eax), %eax |
| 1777 |
palignr $14, %xmm1, %xmm2 |
| 1778 |
movaps %xmm4, 32(%edx) |
| 1779 |
movaps %xmm3, 16(%edx) |
| 1780 |
movaps %xmm7, %xmm1 |
| 1781 |
movaps %xmm2, (%edx) |
| 1782 |
lea 64(%edx), %edx |
| 1783 |
sub $64, %ecx |
| 1784 |
ja L(Shl14LoopStart) |
| 1785 |
|
| 1786 |
L(Shl14LoopLeave): |
| 1787 |
add $32, %ecx |
| 1788 |
jle L(shl_end_0) |
| 1789 |
|
| 1790 |
movaps 2(%eax), %xmm2 |
| 1791 |
movaps 18(%eax), %xmm3 |
| 1792 |
palignr $14, %xmm2, %xmm3 |
| 1793 |
palignr $14, %xmm1, %xmm2 |
| 1794 |
|
| 1795 |
movaps %xmm2, (%edx) |
| 1796 |
movaps %xmm3, 16(%edx) |
| 1797 |
lea 32(%edx, %ecx), %edx |
| 1798 |
lea 32(%eax, %ecx), %eax |
| 1799 |
POP (%edi) |
| 1800 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1801 |
|
| 1802 |
CFI_PUSH (%edi) |
| 1803 |
|
| 1804 |
.p2align 4 |
| 1805 |
L(sh_14_no_prefetch): |
| 1806 |
lea -32(%ecx), %ecx |
| 1807 |
lea -14(%eax), %eax |
| 1808 |
xor %edi, %edi |
| 1809 |
|
| 1810 |
.p2align 4 |
| 1811 |
L(sh_14_no_prefetch_loop): |
| 1812 |
movdqa 16(%eax, %edi), %xmm2 |
| 1813 |
sub $32, %ecx |
| 1814 |
movdqa 32(%eax, %edi), %xmm3 |
| 1815 |
movdqa %xmm3, %xmm4 |
| 1816 |
palignr $14, %xmm2, %xmm3 |
| 1817 |
palignr $14, %xmm1, %xmm2 |
| 1818 |
lea 32(%edi), %edi |
| 1819 |
movdqa %xmm2, -32(%edx, %edi) |
| 1820 |
movdqa %xmm3, -16(%edx, %edi) |
| 1821 |
jb L(sh_14_end_no_prefetch_loop) |
| 1822 |
|
| 1823 |
movdqa 16(%eax, %edi), %xmm2 |
| 1824 |
sub $32, %ecx |
| 1825 |
movdqa 32(%eax, %edi), %xmm3 |
| 1826 |
movdqa %xmm3, %xmm1 |
| 1827 |
palignr $14, %xmm2, %xmm3 |
| 1828 |
palignr $14, %xmm4, %xmm2 |
| 1829 |
lea 32(%edi), %edi |
| 1830 |
movdqa %xmm2, -32(%edx, %edi) |
| 1831 |
movdqa %xmm3, -16(%edx, %edi) |
| 1832 |
jae L(sh_14_no_prefetch_loop) |
| 1833 |
|
| 1834 |
L(sh_14_end_no_prefetch_loop): |
| 1835 |
lea 32(%ecx), %ecx |
| 1836 |
add %ecx, %edi |
| 1837 |
add %edi, %edx |
| 1838 |
lea 14(%edi, %eax), %eax |
| 1839 |
POP (%edi) |
| 1840 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1841 |
|
| 1842 |
CFI_PUSH (%edi) |
| 1843 |
|
| 1844 |
.p2align 4 |
| 1845 |
L(shl_15): |
| 1846 |
# ifndef USE_AS_MEMMOVE |
| 1847 |
movaps -15(%eax), %xmm1 |
| 1848 |
# else |
| 1849 |
movl DEST+4(%esp), %edi |
| 1850 |
movaps -15(%eax), %xmm1 |
| 1851 |
movdqu %xmm0, (%edi) |
| 1852 |
# endif |
| 1853 |
# ifdef DATA_CACHE_SIZE_HALF |
| 1854 |
cmp $DATA_CACHE_SIZE_HALF, %ecx |
| 1855 |
# else |
| 1856 |
# ifdef PIC |
| 1857 |
SETUP_PIC_REG(bx) |
| 1858 |
add $_GLOBAL_OFFSET_TABLE_, %ebx |
| 1859 |
cmp __x86_data_cache_size_half@GOTOFF(%ebx), %ecx |
| 1860 |
# else |
| 1861 |
cmp __x86_data_cache_size_half, %ecx |
| 1862 |
# endif |
| 1863 |
# endif |
| 1864 |
jb L(sh_15_no_prefetch) |
| 1865 |
|
| 1866 |
lea -64(%ecx), %ecx |
| 1867 |
|
| 1868 |
.p2align 4 |
| 1869 |
L(Shl15LoopStart): |
| 1870 |
prefetcht0 0x1c0(%eax) |
| 1871 |
prefetcht0 0x1c0(%edx) |
| 1872 |
movaps 1(%eax), %xmm2 |
| 1873 |
movaps 17(%eax), %xmm3 |
| 1874 |
movaps 33(%eax), %xmm4 |
| 1875 |
movaps 49(%eax), %xmm5 |
| 1876 |
movaps %xmm5, %xmm7 |
| 1877 |
palignr $15, %xmm4, %xmm5 |
| 1878 |
palignr $15, %xmm3, %xmm4 |
| 1879 |
movaps %xmm5, 48(%edx) |
| 1880 |
palignr $15, %xmm2, %xmm3 |
| 1881 |
lea 64(%eax), %eax |
| 1882 |
palignr $15, %xmm1, %xmm2 |
| 1883 |
movaps %xmm4, 32(%edx) |
| 1884 |
movaps %xmm3, 16(%edx) |
| 1885 |
movaps %xmm7, %xmm1 |
| 1886 |
movaps %xmm2, (%edx) |
| 1887 |
lea 64(%edx), %edx |
| 1888 |
sub $64, %ecx |
| 1889 |
ja L(Shl15LoopStart) |
| 1890 |
|
| 1891 |
L(Shl15LoopLeave): |
| 1892 |
add $32, %ecx |
| 1893 |
jle L(shl_end_0) |
| 1894 |
|
| 1895 |
movaps 1(%eax), %xmm2 |
| 1896 |
movaps 17(%eax), %xmm3 |
| 1897 |
palignr $15, %xmm2, %xmm3 |
| 1898 |
palignr $15, %xmm1, %xmm2 |
| 1899 |
|
| 1900 |
movaps %xmm2, (%edx) |
| 1901 |
movaps %xmm3, 16(%edx) |
| 1902 |
lea 32(%edx, %ecx), %edx |
| 1903 |
lea 32(%eax, %ecx), %eax |
| 1904 |
POP (%edi) |
| 1905 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1906 |
|
| 1907 |
CFI_PUSH (%edi) |
| 1908 |
|
| 1909 |
.p2align 4 |
| 1910 |
L(sh_15_no_prefetch): |
| 1911 |
lea -32(%ecx), %ecx |
| 1912 |
lea -15(%eax), %eax |
| 1913 |
xor %edi, %edi |
| 1914 |
|
| 1915 |
.p2align 4 |
| 1916 |
L(sh_15_no_prefetch_loop): |
| 1917 |
movdqa 16(%eax, %edi), %xmm2 |
| 1918 |
sub $32, %ecx |
| 1919 |
movdqa 32(%eax, %edi), %xmm3 |
| 1920 |
movdqa %xmm3, %xmm4 |
| 1921 |
palignr $15, %xmm2, %xmm3 |
| 1922 |
palignr $15, %xmm1, %xmm2 |
| 1923 |
lea 32(%edi), %edi |
| 1924 |
movdqa %xmm2, -32(%edx, %edi) |
| 1925 |
movdqa %xmm3, -16(%edx, %edi) |
| 1926 |
jb L(sh_15_end_no_prefetch_loop) |
| 1927 |
|
| 1928 |
movdqa 16(%eax, %edi), %xmm2 |
| 1929 |
sub $32, %ecx |
| 1930 |
movdqa 32(%eax, %edi), %xmm3 |
| 1931 |
movdqa %xmm3, %xmm1 |
| 1932 |
palignr $15, %xmm2, %xmm3 |
| 1933 |
palignr $15, %xmm4, %xmm2 |
| 1934 |
lea 32(%edi), %edi |
| 1935 |
movdqa %xmm2, -32(%edx, %edi) |
| 1936 |
movdqa %xmm3, -16(%edx, %edi) |
| 1937 |
jae L(sh_15_no_prefetch_loop) |
| 1938 |
|
| 1939 |
L(sh_15_end_no_prefetch_loop): |
| 1940 |
lea 32(%ecx), %ecx |
| 1941 |
add %ecx, %edi |
| 1942 |
add %edi, %edx |
| 1943 |
lea 15(%edi, %eax), %eax |
| 1944 |
POP (%edi) |
| 1945 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1946 |
|
| 1947 |
CFI_PUSH (%edi) |
| 1948 |
|
| 1949 |
.p2align 4 |
| 1950 |
L(shl_end_0): |
| 1951 |
lea 32(%ecx), %ecx |
| 1952 |
lea (%edx, %ecx), %edx |
| 1953 |
lea (%eax, %ecx), %eax |
| 1954 |
POP (%edi) |
| 1955 |
BRANCH_TO_JMPTBL_ENTRY(L(table_48bytes_fwd), %ecx, 4) |
| 1956 |
|
| 1957 |
.p2align 4 |
| 1958 |
L(fwd_write_44bytes): |
| 1959 |
movq -44(%eax), %xmm0 |
| 1960 |
movq %xmm0, -44(%edx) |
| 1961 |
L(fwd_write_36bytes): |
| 1962 |
movq -36(%eax), %xmm0 |
| 1963 |
movq %xmm0, -36(%edx) |
| 1964 |
L(fwd_write_28bytes): |
| 1965 |
movq -28(%eax), %xmm0 |
| 1966 |
movq %xmm0, -28(%edx) |
| 1967 |
L(fwd_write_20bytes): |
| 1968 |
movq -20(%eax), %xmm0 |
| 1969 |
movq %xmm0, -20(%edx) |
| 1970 |
L(fwd_write_12bytes): |
| 1971 |
movq -12(%eax), %xmm0 |
| 1972 |
movq %xmm0, -12(%edx) |
| 1973 |
L(fwd_write_4bytes): |
| 1974 |
movl -4(%eax), %ecx |
| 1975 |
movl %ecx, -4(%edx) |
| 1976 |
# ifdef USE_AS_MEMPCPY |
| 1977 |
movl %edx, %eax |
| 1978 |
# else |
| 1979 |
movl DEST(%esp), %eax |
| 1980 |
# endif |
| 1981 |
RETURN |
| 1982 |
|
| 1983 |
.p2align 4 |
| 1984 |
L(fwd_write_40bytes): |
| 1985 |
movq -40(%eax), %xmm0 |
| 1986 |
movq %xmm0, -40(%edx) |
| 1987 |
L(fwd_write_32bytes): |
| 1988 |
movq -32(%eax), %xmm0 |
| 1989 |
movq %xmm0, -32(%edx) |
| 1990 |
L(fwd_write_24bytes): |
| 1991 |
movq -24(%eax), %xmm0 |
| 1992 |
movq %xmm0, -24(%edx) |
| 1993 |
L(fwd_write_16bytes): |
| 1994 |
movq -16(%eax), %xmm0 |
| 1995 |
movq %xmm0, -16(%edx) |
| 1996 |
L(fwd_write_8bytes): |
| 1997 |
movq -8(%eax), %xmm0 |
| 1998 |
movq %xmm0, -8(%edx) |
| 1999 |
L(fwd_write_0bytes): |
| 2000 |
# ifdef USE_AS_MEMPCPY |
| 2001 |
movl %edx, %eax |
| 2002 |
# else |
| 2003 |
movl DEST(%esp), %eax |
| 2004 |
# endif |
| 2005 |
RETURN |
| 2006 |
|
| 2007 |
.p2align 4 |
| 2008 |
L(fwd_write_5bytes): |
| 2009 |
movl -5(%eax), %ecx |
| 2010 |
movl -4(%eax), %eax |
| 2011 |
movl %ecx, -5(%edx) |
| 2012 |
movl %eax, -4(%edx) |
| 2013 |
# ifdef USE_AS_MEMPCPY |
| 2014 |
movl %edx, %eax |
| 2015 |
# else |
| 2016 |
movl DEST(%esp), %eax |
| 2017 |
# endif |
| 2018 |
RETURN |
| 2019 |
|
| 2020 |
.p2align 4 |
| 2021 |
L(fwd_write_45bytes): |
| 2022 |
movq -45(%eax), %xmm0 |
| 2023 |
movq %xmm0, -45(%edx) |
| 2024 |
L(fwd_write_37bytes): |
| 2025 |
movq -37(%eax), %xmm0 |
| 2026 |
movq %xmm0, -37(%edx) |
| 2027 |
L(fwd_write_29bytes): |
| 2028 |
movq -29(%eax), %xmm0 |
| 2029 |
movq %xmm0, -29(%edx) |
| 2030 |
L(fwd_write_21bytes): |
| 2031 |
movq -21(%eax), %xmm0 |
| 2032 |
movq %xmm0, -21(%edx) |
| 2033 |
L(fwd_write_13bytes): |
| 2034 |
movq -13(%eax), %xmm0 |
| 2035 |
movq %xmm0, -13(%edx) |
| 2036 |
movl -5(%eax), %ecx |
| 2037 |
movl %ecx, -5(%edx) |
| 2038 |
movzbl -1(%eax), %ecx |
| 2039 |
movb %cl, -1(%edx) |
| 2040 |
# ifdef USE_AS_MEMPCPY |
| 2041 |
movl %edx, %eax |
| 2042 |
# else |
| 2043 |
movl DEST(%esp), %eax |
| 2044 |
# endif |
| 2045 |
RETURN |
| 2046 |
|
| 2047 |
.p2align 4 |
| 2048 |
L(fwd_write_41bytes): |
| 2049 |
movq -41(%eax), %xmm0 |
| 2050 |
movq %xmm0, -41(%edx) |
| 2051 |
L(fwd_write_33bytes): |
| 2052 |
movq -33(%eax), %xmm0 |
| 2053 |
movq %xmm0, -33(%edx) |
| 2054 |
L(fwd_write_25bytes): |
| 2055 |
movq -25(%eax), %xmm0 |
| 2056 |
movq %xmm0, -25(%edx) |
| 2057 |
L(fwd_write_17bytes): |
| 2058 |
movq -17(%eax), %xmm0 |
| 2059 |
movq %xmm0, -17(%edx) |
| 2060 |
L(fwd_write_9bytes): |
| 2061 |
movq -9(%eax), %xmm0 |
| 2062 |
movq %xmm0, -9(%edx) |
| 2063 |
L(fwd_write_1bytes): |
| 2064 |
movzbl -1(%eax), %ecx |
| 2065 |
movb %cl, -1(%edx) |
| 2066 |
# ifdef USE_AS_MEMPCPY |
| 2067 |
movl %edx, %eax |
| 2068 |
# else |
| 2069 |
movl DEST(%esp), %eax |
| 2070 |
# endif |
| 2071 |
RETURN |
| 2072 |
|
| 2073 |
.p2align 4 |
| 2074 |
L(fwd_write_46bytes): |
| 2075 |
movq -46(%eax), %xmm0 |
| 2076 |
movq %xmm0, -46(%edx) |
| 2077 |
L(fwd_write_38bytes): |
| 2078 |
movq -38(%eax), %xmm0 |
| 2079 |
movq %xmm0, -38(%edx) |
| 2080 |
L(fwd_write_30bytes): |
| 2081 |
movq -30(%eax), %xmm0 |
| 2082 |
movq %xmm0, -30(%edx) |
| 2083 |
L(fwd_write_22bytes): |
| 2084 |
movq -22(%eax), %xmm0 |
| 2085 |
movq %xmm0, -22(%edx) |
| 2086 |
L(fwd_write_14bytes): |
| 2087 |
movq -14(%eax), %xmm0 |
| 2088 |
movq %xmm0, -14(%edx) |
| 2089 |
L(fwd_write_6bytes): |
| 2090 |
movl -6(%eax), %ecx |
| 2091 |
movl %ecx, -6(%edx) |
| 2092 |
movzwl -2(%eax), %ecx |
| 2093 |
movw %cx, -2(%edx) |
| 2094 |
# ifdef USE_AS_MEMPCPY |
| 2095 |
movl %edx, %eax |
| 2096 |
# else |
| 2097 |
movl DEST(%esp), %eax |
| 2098 |
# endif |
| 2099 |
RETURN |
| 2100 |
|
| 2101 |
.p2align 4 |
| 2102 |
L(fwd_write_42bytes): |
| 2103 |
movq -42(%eax), %xmm0 |
| 2104 |
movq %xmm0, -42(%edx) |
| 2105 |
L(fwd_write_34bytes): |
| 2106 |
movq -34(%eax), %xmm0 |
| 2107 |
movq %xmm0, -34(%edx) |
| 2108 |
L(fwd_write_26bytes): |
| 2109 |
movq -26(%eax), %xmm0 |
| 2110 |
movq %xmm0, -26(%edx) |
| 2111 |
L(fwd_write_18bytes): |
| 2112 |
movq -18(%eax), %xmm0 |
| 2113 |
movq %xmm0, -18(%edx) |
| 2114 |
L(fwd_write_10bytes): |
| 2115 |
movq -10(%eax), %xmm0 |
| 2116 |
movq %xmm0, -10(%edx) |
| 2117 |
L(fwd_write_2bytes): |
| 2118 |
movzwl -2(%eax), %ecx |
| 2119 |
movw %cx, -2(%edx) |
| 2120 |
# ifdef USE_AS_MEMPCPY |
| 2121 |
movl %edx, %eax |
| 2122 |
# else |
| 2123 |
movl DEST(%esp), %eax |
| 2124 |
# endif |
| 2125 |
RETURN |
| 2126 |
|
| 2127 |
.p2align 4 |
| 2128 |
L(fwd_write_47bytes): |
| 2129 |
movq -47(%eax), %xmm0 |
| 2130 |
movq %xmm0, -47(%edx) |
| 2131 |
L(fwd_write_39bytes): |
| 2132 |
movq -39(%eax), %xmm0 |
| 2133 |
movq %xmm0, -39(%edx) |
| 2134 |
L(fwd_write_31bytes): |
| 2135 |
movq -31(%eax), %xmm0 |
| 2136 |
movq %xmm0, -31(%edx) |
| 2137 |
L(fwd_write_23bytes): |
| 2138 |
movq -23(%eax), %xmm0 |
| 2139 |
movq %xmm0, -23(%edx) |
| 2140 |
L(fwd_write_15bytes): |
| 2141 |
movq -15(%eax), %xmm0 |
| 2142 |
movq %xmm0, -15(%edx) |
| 2143 |
L(fwd_write_7bytes): |
| 2144 |
movl -7(%eax), %ecx |
| 2145 |
movl %ecx, -7(%edx) |
| 2146 |
movzwl -3(%eax), %ecx |
| 2147 |
movzbl -1(%eax), %eax |
| 2148 |
movw %cx, -3(%edx) |
| 2149 |
movb %al, -1(%edx) |
| 2150 |
# ifdef USE_AS_MEMPCPY |
| 2151 |
movl %edx, %eax |
| 2152 |
# else |
| 2153 |
movl DEST(%esp), %eax |
| 2154 |
# endif |
| 2155 |
RETURN |
| 2156 |
|
| 2157 |
.p2align 4 |
| 2158 |
L(fwd_write_43bytes): |
| 2159 |
movq -43(%eax), %xmm0 |
| 2160 |
movq %xmm0, -43(%edx) |
| 2161 |
L(fwd_write_35bytes): |
| 2162 |
movq -35(%eax), %xmm0 |
| 2163 |
movq %xmm0, -35(%edx) |
| 2164 |
L(fwd_write_27bytes): |
| 2165 |
movq -27(%eax), %xmm0 |
| 2166 |
movq %xmm0, -27(%edx) |
| 2167 |
L(fwd_write_19bytes): |
| 2168 |
movq -19(%eax), %xmm0 |
| 2169 |
movq %xmm0, -19(%edx) |
| 2170 |
L(fwd_write_11bytes): |
| 2171 |
movq -11(%eax), %xmm0 |
| 2172 |
movq %xmm0, -11(%edx) |
| 2173 |
L(fwd_write_3bytes): |
| 2174 |
movzwl -3(%eax), %ecx |
| 2175 |
movzbl -1(%eax), %eax |
| 2176 |
movw %cx, -3(%edx) |
| 2177 |
movb %al, -1(%edx) |
| 2178 |
# ifdef USE_AS_MEMPCPY |
| 2179 |
movl %edx, %eax |
| 2180 |
# else |
| 2181 |
movl DEST(%esp), %eax |
| 2182 |
# endif |
| 2183 |
RETURN |
| 2184 |
|
| 2185 |
.p2align 4 |
| 2186 |
L(fwd_write_40bytes_align): |
| 2187 |
movdqa -40(%eax), %xmm0 |
| 2188 |
movdqa %xmm0, -40(%edx) |
| 2189 |
L(fwd_write_24bytes_align): |
| 2190 |
movdqa -24(%eax), %xmm0 |
| 2191 |
movdqa %xmm0, -24(%edx) |
| 2192 |
L(fwd_write_8bytes_align): |
| 2193 |
movq -8(%eax), %xmm0 |
| 2194 |
movq %xmm0, -8(%edx) |
| 2195 |
L(fwd_write_0bytes_align): |
| 2196 |
# ifdef USE_AS_MEMPCPY |
| 2197 |
movl %edx, %eax |
| 2198 |
# else |
| 2199 |
movl DEST(%esp), %eax |
| 2200 |
# endif |
| 2201 |
RETURN |
| 2202 |
|
| 2203 |
.p2align 4 |
| 2204 |
L(fwd_write_32bytes_align): |
| 2205 |
movdqa -32(%eax), %xmm0 |
| 2206 |
movdqa %xmm0, -32(%edx) |
| 2207 |
L(fwd_write_16bytes_align): |
| 2208 |
movdqa -16(%eax), %xmm0 |
| 2209 |
movdqa %xmm0, -16(%edx) |
| 2210 |
# ifdef USE_AS_MEMPCPY |
| 2211 |
movl %edx, %eax |
| 2212 |
# else |
| 2213 |
movl DEST(%esp), %eax |
| 2214 |
# endif |
| 2215 |
RETURN |
| 2216 |
|
| 2217 |
.p2align 4 |
| 2218 |
L(fwd_write_5bytes_align): |
| 2219 |
movl -5(%eax), %ecx |
| 2220 |
movl -4(%eax), %eax |
| 2221 |
movl %ecx, -5(%edx) |
| 2222 |
movl %eax, -4(%edx) |
| 2223 |
# ifdef USE_AS_MEMPCPY |
| 2224 |
movl %edx, %eax |
| 2225 |
# else |
| 2226 |
movl DEST(%esp), %eax |
| 2227 |
# endif |
| 2228 |
RETURN |
| 2229 |
|
| 2230 |
.p2align 4 |
| 2231 |
L(fwd_write_45bytes_align): |
| 2232 |
movdqa -45(%eax), %xmm0 |
| 2233 |
movdqa %xmm0, -45(%edx) |
| 2234 |
L(fwd_write_29bytes_align): |
| 2235 |
movdqa -29(%eax), %xmm0 |
| 2236 |
movdqa %xmm0, -29(%edx) |
| 2237 |
L(fwd_write_13bytes_align): |
| 2238 |
movq -13(%eax), %xmm0 |
| 2239 |
movq %xmm0, -13(%edx) |
| 2240 |
movl -5(%eax), %ecx |
| 2241 |
movl %ecx, -5(%edx) |
| 2242 |
movzbl -1(%eax), %ecx |
| 2243 |
movb %cl, -1(%edx) |
| 2244 |
# ifdef USE_AS_MEMPCPY |
| 2245 |
movl %edx, %eax |
| 2246 |
# else |
| 2247 |
movl DEST(%esp), %eax |
| 2248 |
# endif |
| 2249 |
RETURN |
| 2250 |
|
| 2251 |
.p2align 4 |
| 2252 |
L(fwd_write_37bytes_align): |
| 2253 |
movdqa -37(%eax), %xmm0 |
| 2254 |
movdqa %xmm0, -37(%edx) |
| 2255 |
L(fwd_write_21bytes_align): |
| 2256 |
movdqa -21(%eax), %xmm0 |
| 2257 |
movdqa %xmm0, -21(%edx) |
| 2258 |
movl -5(%eax), %ecx |
| 2259 |
movl %ecx, -5(%edx) |
| 2260 |
movzbl -1(%eax), %ecx |
| 2261 |
movb %cl, -1(%edx) |
| 2262 |
# ifdef USE_AS_MEMPCPY |
| 2263 |
movl %edx, %eax |
| 2264 |
# else |
| 2265 |
movl DEST(%esp), %eax |
| 2266 |
# endif |
| 2267 |
RETURN |
| 2268 |
|
| 2269 |
.p2align 4 |
| 2270 |
L(fwd_write_41bytes_align): |
| 2271 |
movdqa -41(%eax), %xmm0 |
| 2272 |
movdqa %xmm0, -41(%edx) |
| 2273 |
L(fwd_write_25bytes_align): |
| 2274 |
movdqa -25(%eax), %xmm0 |
| 2275 |
movdqa %xmm0, -25(%edx) |
| 2276 |
L(fwd_write_9bytes_align): |
| 2277 |
movq -9(%eax), %xmm0 |
| 2278 |
movq %xmm0, -9(%edx) |
| 2279 |
L(fwd_write_1bytes_align): |
| 2280 |
movzbl -1(%eax), %ecx |
| 2281 |
movb %cl, -1(%edx) |
| 2282 |
# ifdef USE_AS_MEMPCPY |
| 2283 |
movl %edx, %eax |
| 2284 |
# else |
| 2285 |
movl DEST(%esp), %eax |
| 2286 |
# endif |
| 2287 |
RETURN |
| 2288 |
|
| 2289 |
.p2align 4 |
| 2290 |
L(fwd_write_33bytes_align): |
| 2291 |
movdqa -33(%eax), %xmm0 |
| 2292 |
movdqa %xmm0, -33(%edx) |
| 2293 |
L(fwd_write_17bytes_align): |
| 2294 |
movdqa -17(%eax), %xmm0 |
| 2295 |
movdqa %xmm0, -17(%edx) |
| 2296 |
movzbl -1(%eax), %ecx |
| 2297 |
movb %cl, -1(%edx) |
| 2298 |
# ifdef USE_AS_MEMPCPY |
| 2299 |
movl %edx, %eax |
| 2300 |
# else |
| 2301 |
movl DEST(%esp), %eax |
| 2302 |
# endif |
| 2303 |
RETURN |
| 2304 |
|
| 2305 |
.p2align 4 |
| 2306 |
L(fwd_write_46bytes_align): |
| 2307 |
movdqa -46(%eax), %xmm0 |
| 2308 |
movdqa %xmm0, -46(%edx) |
| 2309 |
L(fwd_write_30bytes_align): |
| 2310 |
movdqa -30(%eax), %xmm0 |
| 2311 |
movdqa %xmm0, -30(%edx) |
| 2312 |
L(fwd_write_14bytes_align): |
| 2313 |
movq -14(%eax), %xmm0 |
| 2314 |
movq %xmm0, -14(%edx) |
| 2315 |
L(fwd_write_6bytes_align): |
| 2316 |
movl -6(%eax), %ecx |
| 2317 |
movl %ecx, -6(%edx) |
| 2318 |
movzwl -2(%eax), %ecx |
| 2319 |
movw %cx, -2(%edx) |
| 2320 |
# ifdef USE_AS_MEMPCPY |
| 2321 |
movl %edx, %eax |
| 2322 |
# else |
| 2323 |
movl DEST(%esp), %eax |
| 2324 |
# endif |
| 2325 |
RETURN |
| 2326 |
|
| 2327 |
.p2align 4 |
| 2328 |
L(fwd_write_38bytes_align): |
| 2329 |
movdqa -38(%eax), %xmm0 |
| 2330 |
movdqa %xmm0, -38(%edx) |
| 2331 |
L(fwd_write_22bytes_align): |
| 2332 |
movdqa -22(%eax), %xmm0 |
| 2333 |
movdqa %xmm0, -22(%edx) |
| 2334 |
movl -6(%eax), %ecx |
| 2335 |
movl %ecx, -6(%edx) |
| 2336 |
movzwl -2(%eax), %ecx |
| 2337 |
movw %cx, -2(%edx) |
| 2338 |
# ifdef USE_AS_MEMPCPY |
| 2339 |
movl %edx, %eax |
| 2340 |
# else |
| 2341 |
movl DEST(%esp), %eax |
| 2342 |
# endif |
| 2343 |
RETURN |
| 2344 |
|
| 2345 |
.p2align 4 |
| 2346 |
L(fwd_write_42bytes_align): |
| 2347 |
movdqa -42(%eax), %xmm0 |
| 2348 |
movdqa %xmm0, -42(%edx) |
| 2349 |
L(fwd_write_26bytes_align): |
| 2350 |
movdqa -26(%eax), %xmm0 |
| 2351 |
movdqa %xmm0, -26(%edx) |
| 2352 |
L(fwd_write_10bytes_align): |
| 2353 |
movq -10(%eax), %xmm0 |
| 2354 |
movq %xmm0, -10(%edx) |
| 2355 |
L(fwd_write_2bytes_align): |
| 2356 |
movzwl -2(%eax), %ecx |
| 2357 |
movw %cx, -2(%edx) |
| 2358 |
# ifdef USE_AS_MEMPCPY |
| 2359 |
movl %edx, %eax |
| 2360 |
# else |
| 2361 |
movl DEST(%esp), %eax |
| 2362 |
# endif |
| 2363 |
RETURN |
| 2364 |
|
| 2365 |
.p2align 4 |
| 2366 |
L(fwd_write_34bytes_align): |
| 2367 |
movdqa -34(%eax), %xmm0 |
| 2368 |
movdqa %xmm0, -34(%edx) |
| 2369 |
L(fwd_write_18bytes_align): |
| 2370 |
movdqa -18(%eax), %xmm0 |
| 2371 |
movdqa %xmm0, -18(%edx) |
| 2372 |
movzwl -2(%eax), %ecx |
| 2373 |
movw %cx, -2(%edx) |
| 2374 |
# ifdef USE_AS_MEMPCPY |
| 2375 |
movl %edx, %eax |
| 2376 |
# else |
| 2377 |
movl DEST(%esp), %eax |
| 2378 |
# endif |
| 2379 |
RETURN |
| 2380 |
|
| 2381 |
.p2align 4 |
| 2382 |
L(fwd_write_47bytes_align): |
| 2383 |
movdqa -47(%eax), %xmm0 |
| 2384 |
movdqa %xmm0, -47(%edx) |
| 2385 |
L(fwd_write_31bytes_align): |
| 2386 |
movdqa -31(%eax), %xmm0 |
| 2387 |
movdqa %xmm0, -31(%edx) |
| 2388 |
L(fwd_write_15bytes_align): |
| 2389 |
movq -15(%eax), %xmm0 |
| 2390 |
movq %xmm0, -15(%edx) |
| 2391 |
L(fwd_write_7bytes_align): |
| 2392 |
movl -7(%eax), %ecx |
| 2393 |
movl %ecx, -7(%edx) |
| 2394 |
movzwl -3(%eax), %ecx |
| 2395 |
movzbl -1(%eax), %eax |
| 2396 |
movw %cx, -3(%edx) |
| 2397 |
movb %al, -1(%edx) |
| 2398 |
# ifdef USE_AS_MEMPCPY |
| 2399 |
movl %edx, %eax |
| 2400 |
# else |
| 2401 |
movl DEST(%esp), %eax |
| 2402 |
# endif |
| 2403 |
RETURN |
| 2404 |
|
| 2405 |
.p2align 4 |
| 2406 |
L(fwd_write_39bytes_align): |
| 2407 |
movdqa -39(%eax), %xmm0 |
| 2408 |
movdqa %xmm0, -39(%edx) |
| 2409 |
L(fwd_write_23bytes_align): |
| 2410 |
movdqa -23(%eax), %xmm0 |
| 2411 |
movdqa %xmm0, -23(%edx) |
| 2412 |
movl -7(%eax), %ecx |
| 2413 |
movl %ecx, -7(%edx) |
| 2414 |
movzwl -3(%eax), %ecx |
| 2415 |
movzbl -1(%eax), %eax |
| 2416 |
movw %cx, -3(%edx) |
| 2417 |
movb %al, -1(%edx) |
| 2418 |
# ifdef USE_AS_MEMPCPY |
| 2419 |
movl %edx, %eax |
| 2420 |
# else |
| 2421 |
movl DEST(%esp), %eax |
| 2422 |
# endif |
| 2423 |
RETURN |
| 2424 |
|
| 2425 |
.p2align 4 |
| 2426 |
L(fwd_write_43bytes_align): |
| 2427 |
movdqa -43(%eax), %xmm0 |
| 2428 |
movdqa %xmm0, -43(%edx) |
| 2429 |
L(fwd_write_27bytes_align): |
| 2430 |
movdqa -27(%eax), %xmm0 |
| 2431 |
movdqa %xmm0, -27(%edx) |
| 2432 |
L(fwd_write_11bytes_align): |
| 2433 |
movq -11(%eax), %xmm0 |
| 2434 |
movq %xmm0, -11(%edx) |
| 2435 |
L(fwd_write_3bytes_align): |
| 2436 |
movzwl -3(%eax), %ecx |
| 2437 |
movzbl -1(%eax), %eax |
| 2438 |
movw %cx, -3(%edx) |
| 2439 |
movb %al, -1(%edx) |
| 2440 |
# ifdef USE_AS_MEMPCPY |
| 2441 |
movl %edx, %eax |
| 2442 |
# else |
| 2443 |
movl DEST(%esp), %eax |
| 2444 |
# endif |
| 2445 |
RETURN |
| 2446 |
|
| 2447 |
.p2align 4 |
| 2448 |
L(fwd_write_35bytes_align): |
| 2449 |
movdqa -35(%eax), %xmm0 |
| 2450 |
movdqa %xmm0, -35(%edx) |
| 2451 |
L(fwd_write_19bytes_align): |
| 2452 |
movdqa -19(%eax), %xmm0 |
| 2453 |
movdqa %xmm0, -19(%edx) |
| 2454 |
movzwl -3(%eax), %ecx |
| 2455 |
movzbl -1(%eax), %eax |
| 2456 |
movw %cx, -3(%edx) |
| 2457 |
movb %al, -1(%edx) |
| 2458 |
# ifdef USE_AS_MEMPCPY |
| 2459 |
movl %edx, %eax |
| 2460 |
# else |
| 2461 |
movl DEST(%esp), %eax |
| 2462 |
# endif |
| 2463 |
RETURN |
| 2464 |
|
| 2465 |
.p2align 4 |
| 2466 |
L(fwd_write_44bytes_align): |
| 2467 |
movdqa -44(%eax), %xmm0 |
| 2468 |
movdqa %xmm0, -44(%edx) |
| 2469 |
L(fwd_write_28bytes_align): |
| 2470 |
movdqa -28(%eax), %xmm0 |
| 2471 |
movdqa %xmm0, -28(%edx) |
| 2472 |
L(fwd_write_12bytes_align): |
| 2473 |
movq -12(%eax), %xmm0 |
| 2474 |
movq %xmm0, -12(%edx) |
| 2475 |
L(fwd_write_4bytes_align): |
| 2476 |
movl -4(%eax), %ecx |
| 2477 |
movl %ecx, -4(%edx) |
| 2478 |
# ifdef USE_AS_MEMPCPY |
| 2479 |
movl %edx, %eax |
| 2480 |
# else |
| 2481 |
movl DEST(%esp), %eax |
| 2482 |
# endif |
| 2483 |
RETURN |
| 2484 |
|
| 2485 |
.p2align 4 |
| 2486 |
L(fwd_write_36bytes_align): |
| 2487 |
movdqa -36(%eax), %xmm0 |
| 2488 |
movdqa %xmm0, -36(%edx) |
| 2489 |
L(fwd_write_20bytes_align): |
| 2490 |
movdqa -20(%eax), %xmm0 |
| 2491 |
movdqa %xmm0, -20(%edx) |
| 2492 |
movl -4(%eax), %ecx |
| 2493 |
movl %ecx, -4(%edx) |
| 2494 |
# ifdef USE_AS_MEMPCPY |
| 2495 |
movl %edx, %eax |
| 2496 |
# else |
| 2497 |
movl DEST(%esp), %eax |
| 2498 |
# endif |
| 2499 |
RETURN_END |
| 2500 |
|
| 2501 |
CFI_PUSH (%edi) |
| 2502 |
|
| 2503 |
.p2align 4 |
| 2504 |
L(large_page): |
| 2505 |
movdqu (%eax), %xmm1 |
| 2506 |
# ifdef USE_AS_MEMMOVE |
| 2507 |
movl DEST+4(%esp), %edi |
| 2508 |
movdqu %xmm0, (%edi) |
| 2509 |
# endif |
| 2510 |
lea 16(%eax), %eax |
| 2511 |
movntdq %xmm1, (%edx) |
| 2512 |
lea 16(%edx), %edx |
| 2513 |
lea -0x90(%ecx), %ecx |
| 2514 |
POP (%edi) |
| 2515 |
|
| 2516 |
.p2align 4 |
| 2517 |
L(large_page_loop): |
| 2518 |
movdqu (%eax), %xmm0 |
| 2519 |
movdqu 0x10(%eax), %xmm1 |
| 2520 |
movdqu 0x20(%eax), %xmm2 |
| 2521 |
movdqu 0x30(%eax), %xmm3 |
| 2522 |
movdqu 0x40(%eax), %xmm4 |
| 2523 |
movdqu 0x50(%eax), %xmm5 |
| 2524 |
movdqu 0x60(%eax), %xmm6 |
| 2525 |
movdqu 0x70(%eax), %xmm7 |
| 2526 |
lea 0x80(%eax), %eax |
| 2527 |
|
| 2528 |
sub $0x80, %ecx |
| 2529 |
movntdq %xmm0, (%edx) |
| 2530 |
movntdq %xmm1, 0x10(%edx) |
| 2531 |
movntdq %xmm2, 0x20(%edx) |
| 2532 |
movntdq %xmm3, 0x30(%edx) |
| 2533 |
movntdq %xmm4, 0x40(%edx) |
| 2534 |
movntdq %xmm5, 0x50(%edx) |
| 2535 |
movntdq %xmm6, 0x60(%edx) |
| 2536 |
movntdq %xmm7, 0x70(%edx) |
| 2537 |
lea 0x80(%edx), %edx |
| 2538 |
jae L(large_page_loop) |
| 2539 |
cmp $-0x40, %ecx |
| 2540 |
lea 0x80(%ecx), %ecx |
| 2541 |
jl L(large_page_less_64bytes) |
| 2542 |
|
| 2543 |
movdqu (%eax), %xmm0 |
| 2544 |
movdqu 0x10(%eax), %xmm1 |
| 2545 |
movdqu 0x20(%eax), %xmm2 |
| 2546 |
movdqu 0x30(%eax), %xmm3 |
| 2547 |
lea 0x40(%eax), %eax |
| 2548 |
|
| 2549 |
movntdq %xmm0, (%edx) |
| 2550 |
movntdq %xmm1, 0x10(%edx) |
| 2551 |
movntdq %xmm2, 0x20(%edx) |
| 2552 |
movntdq %xmm3, 0x30(%edx) |
| 2553 |
lea 0x40(%edx), %edx |
| 2554 |
sub $0x40, %ecx |
| 2555 |
L(large_page_less_64bytes): |
| 2556 |
cmp $32, %ecx |
| 2557 |
jb L(large_page_less_32bytes) |
| 2558 |
movdqu (%eax), %xmm0 |
| 2559 |
movdqu 0x10(%eax), %xmm1 |
| 2560 |
lea 0x20(%eax), %eax |
| 2561 |
movntdq %xmm0, (%edx) |
| 2562 |
movntdq %xmm1, 0x10(%edx) |
| 2563 |
lea 0x20(%edx), %edx |
| 2564 |
sub $0x20, %ecx |
| 2565 |
L(large_page_less_32bytes): |
| 2566 |
add %ecx, %edx |
| 2567 |
add %ecx, %eax |
| 2568 |
sfence |
| 2569 |
BRANCH_TO_JMPTBL_ENTRY (L(table_48bytes_fwd), %ecx, 4) |
| 2570 |
|
| 2571 |
.p2align 4 |
| 2572 |
L(bk_write_44bytes): |
| 2573 |
movq 36(%eax), %xmm0 |
| 2574 |
movq %xmm0, 36(%edx) |
| 2575 |
L(bk_write_36bytes): |
| 2576 |
movq 28(%eax), %xmm0 |
| 2577 |
movq %xmm0, 28(%edx) |
| 2578 |
L(bk_write_28bytes): |
| 2579 |
movq 20(%eax), %xmm0 |
| 2580 |
movq %xmm0, 20(%edx) |
| 2581 |
L(bk_write_20bytes): |
| 2582 |
movq 12(%eax), %xmm0 |
| 2583 |
movq %xmm0, 12(%edx) |
| 2584 |
L(bk_write_12bytes): |
| 2585 |
movq 4(%eax), %xmm0 |
| 2586 |
movq %xmm0, 4(%edx) |
| 2587 |
L(bk_write_4bytes): |
| 2588 |
movl (%eax), %ecx |
| 2589 |
movl %ecx, (%edx) |
| 2590 |
L(bk_write_0bytes): |
| 2591 |
movl DEST(%esp), %eax |
| 2592 |
# ifdef USE_AS_MEMPCPY |
| 2593 |
movl LEN(%esp), %ecx |
| 2594 |
add %ecx, %eax |
| 2595 |
# endif |
| 2596 |
RETURN |
| 2597 |
|
| 2598 |
.p2align 4 |
| 2599 |
L(bk_write_40bytes): |
| 2600 |
movq 32(%eax), %xmm0 |
| 2601 |
movq %xmm0, 32(%edx) |
| 2602 |
L(bk_write_32bytes): |
| 2603 |
movq 24(%eax), %xmm0 |
| 2604 |
movq %xmm0, 24(%edx) |
| 2605 |
L(bk_write_24bytes): |
| 2606 |
movq 16(%eax), %xmm0 |
| 2607 |
movq %xmm0, 16(%edx) |
| 2608 |
L(bk_write_16bytes): |
| 2609 |
movq 8(%eax), %xmm0 |
| 2610 |
movq %xmm0, 8(%edx) |
| 2611 |
L(bk_write_8bytes): |
| 2612 |
movq (%eax), %xmm0 |
| 2613 |
movq %xmm0, (%edx) |
| 2614 |
movl DEST(%esp), %eax |
| 2615 |
# ifdef USE_AS_MEMPCPY |
| 2616 |
movl LEN(%esp), %ecx |
| 2617 |
add %ecx, %eax |
| 2618 |
# endif |
| 2619 |
RETURN |
| 2620 |
|
| 2621 |
.p2align 4 |
| 2622 |
L(bk_write_45bytes): |
| 2623 |
movq 37(%eax), %xmm0 |
| 2624 |
movq %xmm0, 37(%edx) |
| 2625 |
L(bk_write_37bytes): |
| 2626 |
movq 29(%eax), %xmm0 |
| 2627 |
movq %xmm0, 29(%edx) |
| 2628 |
L(bk_write_29bytes): |
| 2629 |
movq 21(%eax), %xmm0 |
| 2630 |
movq %xmm0, 21(%edx) |
| 2631 |
L(bk_write_21bytes): |
| 2632 |
movq 13(%eax), %xmm0 |
| 2633 |
movq %xmm0, 13(%edx) |
| 2634 |
L(bk_write_13bytes): |
| 2635 |
movq 5(%eax), %xmm0 |
| 2636 |
movq %xmm0, 5(%edx) |
| 2637 |
L(bk_write_5bytes): |
| 2638 |
movl 1(%eax), %ecx |
| 2639 |
movl %ecx, 1(%edx) |
| 2640 |
L(bk_write_1bytes): |
| 2641 |
movzbl (%eax), %ecx |
| 2642 |
movb %cl, (%edx) |
| 2643 |
movl DEST(%esp), %eax |
| 2644 |
# ifdef USE_AS_MEMPCPY |
| 2645 |
movl LEN(%esp), %ecx |
| 2646 |
add %ecx, %eax |
| 2647 |
# endif |
| 2648 |
RETURN |
| 2649 |
|
| 2650 |
.p2align 4 |
| 2651 |
L(bk_write_41bytes): |
| 2652 |
movq 33(%eax), %xmm0 |
| 2653 |
movq %xmm0, 33(%edx) |
| 2654 |
L(bk_write_33bytes): |
| 2655 |
movq 25(%eax), %xmm0 |
| 2656 |
movq %xmm0, 25(%edx) |
| 2657 |
L(bk_write_25bytes): |
| 2658 |
movq 17(%eax), %xmm0 |
| 2659 |
movq %xmm0, 17(%edx) |
| 2660 |
L(bk_write_17bytes): |
| 2661 |
movq 9(%eax), %xmm0 |
| 2662 |
movq %xmm0, 9(%edx) |
| 2663 |
L(bk_write_9bytes): |
| 2664 |
movq 1(%eax), %xmm0 |
| 2665 |
movq %xmm0, 1(%edx) |
| 2666 |
movzbl (%eax), %ecx |
| 2667 |
movb %cl, (%edx) |
| 2668 |
movl DEST(%esp), %eax |
| 2669 |
# ifdef USE_AS_MEMPCPY |
| 2670 |
movl LEN(%esp), %ecx |
| 2671 |
add %ecx, %eax |
| 2672 |
# endif |
| 2673 |
RETURN |
| 2674 |
|
| 2675 |
.p2align 4 |
| 2676 |
L(bk_write_46bytes): |
| 2677 |
movq 38(%eax), %xmm0 |
| 2678 |
movq %xmm0, 38(%edx) |
| 2679 |
L(bk_write_38bytes): |
| 2680 |
movq 30(%eax), %xmm0 |
| 2681 |
movq %xmm0, 30(%edx) |
| 2682 |
L(bk_write_30bytes): |
| 2683 |
movq 22(%eax), %xmm0 |
| 2684 |
movq %xmm0, 22(%edx) |
| 2685 |
L(bk_write_22bytes): |
| 2686 |
movq 14(%eax), %xmm0 |
| 2687 |
movq %xmm0, 14(%edx) |
| 2688 |
L(bk_write_14bytes): |
| 2689 |
movq 6(%eax), %xmm0 |
| 2690 |
movq %xmm0, 6(%edx) |
| 2691 |
L(bk_write_6bytes): |
| 2692 |
movl 2(%eax), %ecx |
| 2693 |
movl %ecx, 2(%edx) |
| 2694 |
movzwl (%eax), %ecx |
| 2695 |
movw %cx, (%edx) |
| 2696 |
movl DEST(%esp), %eax |
| 2697 |
# ifdef USE_AS_MEMPCPY |
| 2698 |
movl LEN(%esp), %ecx |
| 2699 |
add %ecx, %eax |
| 2700 |
# endif |
| 2701 |
RETURN |
| 2702 |
|
| 2703 |
.p2align 4 |
| 2704 |
L(bk_write_42bytes): |
| 2705 |
movq 34(%eax), %xmm0 |
| 2706 |
movq %xmm0, 34(%edx) |
| 2707 |
L(bk_write_34bytes): |
| 2708 |
movq 26(%eax), %xmm0 |
| 2709 |
movq %xmm0, 26(%edx) |
| 2710 |
L(bk_write_26bytes): |
| 2711 |
movq 18(%eax), %xmm0 |
| 2712 |
movq %xmm0, 18(%edx) |
| 2713 |
L(bk_write_18bytes): |
| 2714 |
movq 10(%eax), %xmm0 |
| 2715 |
movq %xmm0, 10(%edx) |
| 2716 |
L(bk_write_10bytes): |
| 2717 |
movq 2(%eax), %xmm0 |
| 2718 |
movq %xmm0, 2(%edx) |
| 2719 |
L(bk_write_2bytes): |
| 2720 |
movzwl (%eax), %ecx |
| 2721 |
movw %cx, (%edx) |
| 2722 |
movl DEST(%esp), %eax |
| 2723 |
# ifdef USE_AS_MEMPCPY |
| 2724 |
movl LEN(%esp), %ecx |
| 2725 |
add %ecx, %eax |
| 2726 |
# endif |
| 2727 |
RETURN |
| 2728 |
|
| 2729 |
.p2align 4 |
| 2730 |
L(bk_write_47bytes): |
| 2731 |
movq 39(%eax), %xmm0 |
| 2732 |
movq %xmm0, 39(%edx) |
| 2733 |
L(bk_write_39bytes): |
| 2734 |
movq 31(%eax), %xmm0 |
| 2735 |
movq %xmm0, 31(%edx) |
| 2736 |
L(bk_write_31bytes): |
| 2737 |
movq 23(%eax), %xmm0 |
| 2738 |
movq %xmm0, 23(%edx) |
| 2739 |
L(bk_write_23bytes): |
| 2740 |
movq 15(%eax), %xmm0 |
| 2741 |
movq %xmm0, 15(%edx) |
| 2742 |
L(bk_write_15bytes): |
| 2743 |
movq 7(%eax), %xmm0 |
| 2744 |
movq %xmm0, 7(%edx) |
| 2745 |
L(bk_write_7bytes): |
| 2746 |
movl 3(%eax), %ecx |
| 2747 |
movl %ecx, 3(%edx) |
| 2748 |
movzwl 1(%eax), %ecx |
| 2749 |
movw %cx, 1(%edx) |
| 2750 |
movzbl (%eax), %eax |
| 2751 |
movb %al, (%edx) |
| 2752 |
movl DEST(%esp), %eax |
| 2753 |
# ifdef USE_AS_MEMPCPY |
| 2754 |
movl LEN(%esp), %ecx |
| 2755 |
add %ecx, %eax |
| 2756 |
# endif |
| 2757 |
RETURN |
| 2758 |
|
| 2759 |
.p2align 4 |
| 2760 |
L(bk_write_43bytes): |
| 2761 |
movq 35(%eax), %xmm0 |
| 2762 |
movq %xmm0, 35(%edx) |
| 2763 |
L(bk_write_35bytes): |
| 2764 |
movq 27(%eax), %xmm0 |
| 2765 |
movq %xmm0, 27(%edx) |
| 2766 |
L(bk_write_27bytes): |
| 2767 |
movq 19(%eax), %xmm0 |
| 2768 |
movq %xmm0, 19(%edx) |
| 2769 |
L(bk_write_19bytes): |
| 2770 |
movq 11(%eax), %xmm0 |
| 2771 |
movq %xmm0, 11(%edx) |
| 2772 |
L(bk_write_11bytes): |
| 2773 |
movq 3(%eax), %xmm0 |
| 2774 |
movq %xmm0, 3(%edx) |
| 2775 |
L(bk_write_3bytes): |
| 2776 |
movzwl 1(%eax), %ecx |
| 2777 |
movw %cx, 1(%edx) |
| 2778 |
movzbl (%eax), %eax |
| 2779 |
movb %al, (%edx) |
| 2780 |
movl DEST(%esp), %eax |
| 2781 |
# ifdef USE_AS_MEMPCPY |
| 2782 |
movl LEN(%esp), %ecx |
| 2783 |
add %ecx, %eax |
| 2784 |
# endif |
| 2785 |
RETURN_END |
| 2786 |
|
| 2787 |
|
| 2788 |
.pushsection .rodata.ssse3,"a",@progbits |
| 2789 |
.p2align 2 |
| 2790 |
L(table_48bytes_fwd): |
| 2791 |
.int JMPTBL (L(fwd_write_0bytes), L(table_48bytes_fwd)) |
| 2792 |
.int JMPTBL (L(fwd_write_1bytes), L(table_48bytes_fwd)) |
| 2793 |
.int JMPTBL (L(fwd_write_2bytes), L(table_48bytes_fwd)) |
| 2794 |
.int JMPTBL (L(fwd_write_3bytes), L(table_48bytes_fwd)) |
| 2795 |
.int JMPTBL (L(fwd_write_4bytes), L(table_48bytes_fwd)) |
| 2796 |
.int JMPTBL (L(fwd_write_5bytes), L(table_48bytes_fwd)) |
| 2797 |
.int JMPTBL (L(fwd_write_6bytes), L(table_48bytes_fwd)) |
| 2798 |
.int JMPTBL (L(fwd_write_7bytes), L(table_48bytes_fwd)) |
| 2799 |
.int JMPTBL (L(fwd_write_8bytes), L(table_48bytes_fwd)) |
| 2800 |
.int JMPTBL (L(fwd_write_9bytes), L(table_48bytes_fwd)) |
| 2801 |
.int JMPTBL (L(fwd_write_10bytes), L(table_48bytes_fwd)) |
| 2802 |
.int JMPTBL (L(fwd_write_11bytes), L(table_48bytes_fwd)) |
| 2803 |
.int JMPTBL (L(fwd_write_12bytes), L(table_48bytes_fwd)) |
| 2804 |
.int JMPTBL (L(fwd_write_13bytes), L(table_48bytes_fwd)) |
| 2805 |
.int JMPTBL (L(fwd_write_14bytes), L(table_48bytes_fwd)) |
| 2806 |
.int JMPTBL (L(fwd_write_15bytes), L(table_48bytes_fwd)) |
| 2807 |
.int JMPTBL (L(fwd_write_16bytes), L(table_48bytes_fwd)) |
| 2808 |
.int JMPTBL (L(fwd_write_17bytes), L(table_48bytes_fwd)) |
| 2809 |
.int JMPTBL (L(fwd_write_18bytes), L(table_48bytes_fwd)) |
| 2810 |
.int JMPTBL (L(fwd_write_19bytes), L(table_48bytes_fwd)) |
| 2811 |
.int JMPTBL (L(fwd_write_20bytes), L(table_48bytes_fwd)) |
| 2812 |
.int JMPTBL (L(fwd_write_21bytes), L(table_48bytes_fwd)) |
| 2813 |
.int JMPTBL (L(fwd_write_22bytes), L(table_48bytes_fwd)) |
| 2814 |
.int JMPTBL (L(fwd_write_23bytes), L(table_48bytes_fwd)) |
| 2815 |
.int JMPTBL (L(fwd_write_24bytes), L(table_48bytes_fwd)) |
| 2816 |
.int JMPTBL (L(fwd_write_25bytes), L(table_48bytes_fwd)) |
| 2817 |
.int JMPTBL (L(fwd_write_26bytes), L(table_48bytes_fwd)) |
| 2818 |
.int JMPTBL (L(fwd_write_27bytes), L(table_48bytes_fwd)) |
| 2819 |
.int JMPTBL (L(fwd_write_28bytes), L(table_48bytes_fwd)) |
| 2820 |
.int JMPTBL (L(fwd_write_29bytes), L(table_48bytes_fwd)) |
| 2821 |
.int JMPTBL (L(fwd_write_30bytes), L(table_48bytes_fwd)) |
| 2822 |
.int JMPTBL (L(fwd_write_31bytes), L(table_48bytes_fwd)) |
| 2823 |
.int JMPTBL (L(fwd_write_32bytes), L(table_48bytes_fwd)) |
| 2824 |
.int JMPTBL (L(fwd_write_33bytes), L(table_48bytes_fwd)) |
| 2825 |
.int JMPTBL (L(fwd_write_34bytes), L(table_48bytes_fwd)) |
| 2826 |
.int JMPTBL (L(fwd_write_35bytes), L(table_48bytes_fwd)) |
| 2827 |
.int JMPTBL (L(fwd_write_36bytes), L(table_48bytes_fwd)) |
| 2828 |
.int JMPTBL (L(fwd_write_37bytes), L(table_48bytes_fwd)) |
| 2829 |
.int JMPTBL (L(fwd_write_38bytes), L(table_48bytes_fwd)) |
| 2830 |
.int JMPTBL (L(fwd_write_39bytes), L(table_48bytes_fwd)) |
| 2831 |
.int JMPTBL (L(fwd_write_40bytes), L(table_48bytes_fwd)) |
| 2832 |
.int JMPTBL (L(fwd_write_41bytes), L(table_48bytes_fwd)) |
| 2833 |
.int JMPTBL (L(fwd_write_42bytes), L(table_48bytes_fwd)) |
| 2834 |
.int JMPTBL (L(fwd_write_43bytes), L(table_48bytes_fwd)) |
| 2835 |
.int JMPTBL (L(fwd_write_44bytes), L(table_48bytes_fwd)) |
| 2836 |
.int JMPTBL (L(fwd_write_45bytes), L(table_48bytes_fwd)) |
| 2837 |
.int JMPTBL (L(fwd_write_46bytes), L(table_48bytes_fwd)) |
| 2838 |
.int JMPTBL (L(fwd_write_47bytes), L(table_48bytes_fwd)) |
| 2839 |
|
| 2840 |
.p2align 2 |
| 2841 |
L(table_48bytes_fwd_align): |
| 2842 |
.int JMPTBL (L(fwd_write_0bytes_align), L(table_48bytes_fwd_align)) |
| 2843 |
.int JMPTBL (L(fwd_write_1bytes_align), L(table_48bytes_fwd_align)) |
| 2844 |
.int JMPTBL (L(fwd_write_2bytes_align), L(table_48bytes_fwd_align)) |
| 2845 |
.int JMPTBL (L(fwd_write_3bytes_align), L(table_48bytes_fwd_align)) |
| 2846 |
.int JMPTBL (L(fwd_write_4bytes_align), L(table_48bytes_fwd_align)) |
| 2847 |
.int JMPTBL (L(fwd_write_5bytes_align), L(table_48bytes_fwd_align)) |
| 2848 |
.int JMPTBL (L(fwd_write_6bytes_align), L(table_48bytes_fwd_align)) |
| 2849 |
.int JMPTBL (L(fwd_write_7bytes_align), L(table_48bytes_fwd_align)) |
| 2850 |
.int JMPTBL (L(fwd_write_8bytes_align), L(table_48bytes_fwd_align)) |
| 2851 |
.int JMPTBL (L(fwd_write_9bytes_align), L(table_48bytes_fwd_align)) |
| 2852 |
.int JMPTBL (L(fwd_write_10bytes_align), L(table_48bytes_fwd_align)) |
| 2853 |
.int JMPTBL (L(fwd_write_11bytes_align), L(table_48bytes_fwd_align)) |
| 2854 |
.int JMPTBL (L(fwd_write_12bytes_align), L(table_48bytes_fwd_align)) |
| 2855 |
.int JMPTBL (L(fwd_write_13bytes_align), L(table_48bytes_fwd_align)) |
| 2856 |
.int JMPTBL (L(fwd_write_14bytes_align), L(table_48bytes_fwd_align)) |
| 2857 |
.int JMPTBL (L(fwd_write_15bytes_align), L(table_48bytes_fwd_align)) |
| 2858 |
.int JMPTBL (L(fwd_write_16bytes_align), L(table_48bytes_fwd_align)) |
| 2859 |
.int JMPTBL (L(fwd_write_17bytes_align), L(table_48bytes_fwd_align)) |
| 2860 |
.int JMPTBL (L(fwd_write_18bytes_align), L(table_48bytes_fwd_align)) |
| 2861 |
.int JMPTBL (L(fwd_write_19bytes_align), L(table_48bytes_fwd_align)) |
| 2862 |
.int JMPTBL (L(fwd_write_20bytes_align), L(table_48bytes_fwd_align)) |
| 2863 |
.int JMPTBL (L(fwd_write_21bytes_align), L(table_48bytes_fwd_align)) |
| 2864 |
.int JMPTBL (L(fwd_write_22bytes_align), L(table_48bytes_fwd_align)) |
| 2865 |
.int JMPTBL (L(fwd_write_23bytes_align), L(table_48bytes_fwd_align)) |
| 2866 |
.int JMPTBL (L(fwd_write_24bytes_align), L(table_48bytes_fwd_align)) |
| 2867 |
.int JMPTBL (L(fwd_write_25bytes_align), L(table_48bytes_fwd_align)) |
| 2868 |
.int JMPTBL (L(fwd_write_26bytes_align), L(table_48bytes_fwd_align)) |
| 2869 |
.int JMPTBL (L(fwd_write_27bytes_align), L(table_48bytes_fwd_align)) |
| 2870 |
.int JMPTBL (L(fwd_write_28bytes_align), L(table_48bytes_fwd_align)) |
| 2871 |
.int JMPTBL (L(fwd_write_29bytes_align), L(table_48bytes_fwd_align)) |
| 2872 |
.int JMPTBL (L(fwd_write_30bytes_align), L(table_48bytes_fwd_align)) |
| 2873 |
.int JMPTBL (L(fwd_write_31bytes_align), L(table_48bytes_fwd_align)) |
| 2874 |
.int JMPTBL (L(fwd_write_32bytes_align), L(table_48bytes_fwd_align)) |
| 2875 |
.int JMPTBL (L(fwd_write_33bytes_align), L(table_48bytes_fwd_align)) |
| 2876 |
.int JMPTBL (L(fwd_write_34bytes_align), L(table_48bytes_fwd_align)) |
| 2877 |
.int JMPTBL (L(fwd_write_35bytes_align), L(table_48bytes_fwd_align)) |
| 2878 |
.int JMPTBL (L(fwd_write_36bytes_align), L(table_48bytes_fwd_align)) |
| 2879 |
.int JMPTBL (L(fwd_write_37bytes_align), L(table_48bytes_fwd_align)) |
| 2880 |
.int JMPTBL (L(fwd_write_38bytes_align), L(table_48bytes_fwd_align)) |
| 2881 |
.int JMPTBL (L(fwd_write_39bytes_align), L(table_48bytes_fwd_align)) |
| 2882 |
.int JMPTBL (L(fwd_write_40bytes_align), L(table_48bytes_fwd_align)) |
| 2883 |
.int JMPTBL (L(fwd_write_41bytes_align), L(table_48bytes_fwd_align)) |
| 2884 |
.int JMPTBL (L(fwd_write_42bytes_align), L(table_48bytes_fwd_align)) |
| 2885 |
.int JMPTBL (L(fwd_write_43bytes_align), L(table_48bytes_fwd_align)) |
| 2886 |
.int JMPTBL (L(fwd_write_44bytes_align), L(table_48bytes_fwd_align)) |
| 2887 |
.int JMPTBL (L(fwd_write_45bytes_align), L(table_48bytes_fwd_align)) |
| 2888 |
.int JMPTBL (L(fwd_write_46bytes_align), L(table_48bytes_fwd_align)) |
| 2889 |
.int JMPTBL (L(fwd_write_47bytes_align), L(table_48bytes_fwd_align)) |
| 2890 |
|
| 2891 |
.p2align 2 |
| 2892 |
L(shl_table): |
| 2893 |
.int JMPTBL (L(shl_0), L(shl_table)) |
| 2894 |
.int JMPTBL (L(shl_1), L(shl_table)) |
| 2895 |
.int JMPTBL (L(shl_2), L(shl_table)) |
| 2896 |
.int JMPTBL (L(shl_3), L(shl_table)) |
| 2897 |
.int JMPTBL (L(shl_4), L(shl_table)) |
| 2898 |
.int JMPTBL (L(shl_5), L(shl_table)) |
| 2899 |
.int JMPTBL (L(shl_6), L(shl_table)) |
| 2900 |
.int JMPTBL (L(shl_7), L(shl_table)) |
| 2901 |
.int JMPTBL (L(shl_8), L(shl_table)) |
| 2902 |
.int JMPTBL (L(shl_9), L(shl_table)) |
| 2903 |
.int JMPTBL (L(shl_10), L(shl_table)) |
| 2904 |
.int JMPTBL (L(shl_11), L(shl_table)) |
| 2905 |
.int JMPTBL (L(shl_12), L(shl_table)) |
| 2906 |
.int JMPTBL (L(shl_13), L(shl_table)) |
| 2907 |
.int JMPTBL (L(shl_14), L(shl_table)) |
| 2908 |
.int JMPTBL (L(shl_15), L(shl_table)) |
| 2909 |
|
| 2910 |
.p2align 2 |
| 2911 |
L(table_48_bytes_bwd): |
| 2912 |
.int JMPTBL (L(bk_write_0bytes), L(table_48_bytes_bwd)) |
| 2913 |
.int JMPTBL (L(bk_write_1bytes), L(table_48_bytes_bwd)) |
| 2914 |
.int JMPTBL (L(bk_write_2bytes), L(table_48_bytes_bwd)) |
| 2915 |
.int JMPTBL (L(bk_write_3bytes), L(table_48_bytes_bwd)) |
| 2916 |
.int JMPTBL (L(bk_write_4bytes), L(table_48_bytes_bwd)) |
| 2917 |
.int JMPTBL (L(bk_write_5bytes), L(table_48_bytes_bwd)) |
| 2918 |
.int JMPTBL (L(bk_write_6bytes), L(table_48_bytes_bwd)) |
| 2919 |
.int JMPTBL (L(bk_write_7bytes), L(table_48_bytes_bwd)) |
| 2920 |
.int JMPTBL (L(bk_write_8bytes), L(table_48_bytes_bwd)) |
| 2921 |
.int JMPTBL (L(bk_write_9bytes), L(table_48_bytes_bwd)) |
| 2922 |
.int JMPTBL (L(bk_write_10bytes), L(table_48_bytes_bwd)) |
| 2923 |
.int JMPTBL (L(bk_write_11bytes), L(table_48_bytes_bwd)) |
| 2924 |
.int JMPTBL (L(bk_write_12bytes), L(table_48_bytes_bwd)) |
| 2925 |
.int JMPTBL (L(bk_write_13bytes), L(table_48_bytes_bwd)) |
| 2926 |
.int JMPTBL (L(bk_write_14bytes), L(table_48_bytes_bwd)) |
| 2927 |
.int JMPTBL (L(bk_write_15bytes), L(table_48_bytes_bwd)) |
| 2928 |
.int JMPTBL (L(bk_write_16bytes), L(table_48_bytes_bwd)) |
| 2929 |
.int JMPTBL (L(bk_write_17bytes), L(table_48_bytes_bwd)) |
| 2930 |
.int JMPTBL (L(bk_write_18bytes), L(table_48_bytes_bwd)) |
| 2931 |
.int JMPTBL (L(bk_write_19bytes), L(table_48_bytes_bwd)) |
| 2932 |
.int JMPTBL (L(bk_write_20bytes), L(table_48_bytes_bwd)) |
| 2933 |
.int JMPTBL (L(bk_write_21bytes), L(table_48_bytes_bwd)) |
| 2934 |
.int JMPTBL (L(bk_write_22bytes), L(table_48_bytes_bwd)) |
| 2935 |
.int JMPTBL (L(bk_write_23bytes), L(table_48_bytes_bwd)) |
| 2936 |
.int JMPTBL (L(bk_write_24bytes), L(table_48_bytes_bwd)) |
| 2937 |
.int JMPTBL (L(bk_write_25bytes), L(table_48_bytes_bwd)) |
| 2938 |
.int JMPTBL (L(bk_write_26bytes), L(table_48_bytes_bwd)) |
| 2939 |
.int JMPTBL (L(bk_write_27bytes), L(table_48_bytes_bwd)) |
| 2940 |
.int JMPTBL (L(bk_write_28bytes), L(table_48_bytes_bwd)) |
| 2941 |
.int JMPTBL (L(bk_write_29bytes), L(table_48_bytes_bwd)) |
| 2942 |
.int JMPTBL (L(bk_write_30bytes), L(table_48_bytes_bwd)) |
| 2943 |
.int JMPTBL (L(bk_write_31bytes), L(table_48_bytes_bwd)) |
| 2944 |
.int JMPTBL (L(bk_write_32bytes), L(table_48_bytes_bwd)) |
| 2945 |
.int JMPTBL (L(bk_write_33bytes), L(table_48_bytes_bwd)) |
| 2946 |
.int JMPTBL (L(bk_write_34bytes), L(table_48_bytes_bwd)) |
| 2947 |
.int JMPTBL (L(bk_write_35bytes), L(table_48_bytes_bwd)) |
| 2948 |
.int JMPTBL (L(bk_write_36bytes), L(table_48_bytes_bwd)) |
| 2949 |
.int JMPTBL (L(bk_write_37bytes), L(table_48_bytes_bwd)) |
| 2950 |
.int JMPTBL (L(bk_write_38bytes), L(table_48_bytes_bwd)) |
| 2951 |
.int JMPTBL (L(bk_write_39bytes), L(table_48_bytes_bwd)) |
| 2952 |
.int JMPTBL (L(bk_write_40bytes), L(table_48_bytes_bwd)) |
| 2953 |
.int JMPTBL (L(bk_write_41bytes), L(table_48_bytes_bwd)) |
| 2954 |
.int JMPTBL (L(bk_write_42bytes), L(table_48_bytes_bwd)) |
| 2955 |
.int JMPTBL (L(bk_write_43bytes), L(table_48_bytes_bwd)) |
| 2956 |
.int JMPTBL (L(bk_write_44bytes), L(table_48_bytes_bwd)) |
| 2957 |
.int JMPTBL (L(bk_write_45bytes), L(table_48_bytes_bwd)) |
| 2958 |
.int JMPTBL (L(bk_write_46bytes), L(table_48_bytes_bwd)) |
| 2959 |
.int JMPTBL (L(bk_write_47bytes), L(table_48_bytes_bwd)) |
| 2960 |
|
| 2961 |
.popsection |
| 2962 |
|
| 2963 |
# ifdef USE_AS_MEMMOVE |
| 2964 |
.p2align 4 |
| 2965 |
L(copy_backward): |
| 2966 |
PUSH (%edi) |
| 2967 |
movl %eax, %edi |
| 2968 |
lea (%ecx,%edx,1),%edx |
| 2969 |
lea (%ecx,%edi,1),%edi |
| 2970 |
testl $0x3, %edx |
| 2971 |
jnz L(bk_align) |
| 2972 |
|
| 2973 |
L(bk_aligned_4): |
| 2974 |
cmp $64, %ecx |
| 2975 |
jae L(bk_write_more64bytes) |
| 2976 |
|
| 2977 |
L(bk_write_64bytesless): |
| 2978 |
cmp $32, %ecx |
| 2979 |
jb L(bk_write_less32bytes) |
| 2980 |
|
| 2981 |
L(bk_write_more32bytes): |
| 2982 |
/* Copy 32 bytes at a time. */ |
| 2983 |
sub $32, %ecx |
| 2984 |
movq -8(%edi), %xmm0 |
| 2985 |
movq %xmm0, -8(%edx) |
| 2986 |
movq -16(%edi), %xmm0 |
| 2987 |
movq %xmm0, -16(%edx) |
| 2988 |
movq -24(%edi), %xmm0 |
| 2989 |
movq %xmm0, -24(%edx) |
| 2990 |
movq -32(%edi), %xmm0 |
| 2991 |
movq %xmm0, -32(%edx) |
| 2992 |
sub $32, %edx |
| 2993 |
sub $32, %edi |
| 2994 |
|
| 2995 |
L(bk_write_less32bytes): |
| 2996 |
movl %edi, %eax |
| 2997 |
sub %ecx, %edx |
| 2998 |
sub %ecx, %eax |
| 2999 |
POP (%edi) |
| 3000 |
L(bk_write_less32bytes_2): |
| 3001 |
BRANCH_TO_JMPTBL_ENTRY (L(table_48_bytes_bwd), %ecx, 4) |
| 3002 |
|
| 3003 |
CFI_PUSH (%edi) |
| 3004 |
|
| 3005 |
.p2align 4 |
| 3006 |
L(bk_align): |
| 3007 |
cmp $8, %ecx |
| 3008 |
jbe L(bk_write_less32bytes) |
| 3009 |
testl $1, %edx |
| 3010 |
/* We get here only if (EDX & 3 ) != 0 so if (EDX & 1) ==0, |
| 3011 |
then (EDX & 2) must be != 0. */ |
| 3012 |
jz L(bk_got2) |
| 3013 |
sub $1, %edi |
| 3014 |
sub $1, %ecx |
| 3015 |
sub $1, %edx |
| 3016 |
movzbl (%edi), %eax |
| 3017 |
movb %al, (%edx) |
| 3018 |
|
| 3019 |
testl $2, %edx |
| 3020 |
jz L(bk_aligned_4) |
| 3021 |
|
| 3022 |
L(bk_got2): |
| 3023 |
sub $2, %edi |
| 3024 |
sub $2, %ecx |
| 3025 |
sub $2, %edx |
| 3026 |
movzwl (%edi), %eax |
| 3027 |
movw %ax, (%edx) |
| 3028 |
jmp L(bk_aligned_4) |
| 3029 |
|
| 3030 |
.p2align 4 |
| 3031 |
L(bk_write_more64bytes): |
| 3032 |
/* Check alignment of last byte. */ |
| 3033 |
testl $15, %edx |
| 3034 |
jz L(bk_ssse3_cpy_pre) |
| 3035 |
|
| 3036 |
/* EDX is aligned 4 bytes, but not 16 bytes. */ |
| 3037 |
L(bk_ssse3_align): |
| 3038 |
sub $4, %edi |
| 3039 |
sub $4, %ecx |
| 3040 |
sub $4, %edx |
| 3041 |
movl (%edi), %eax |
| 3042 |
movl %eax, (%edx) |
| 3043 |
|
| 3044 |
testl $15, %edx |
| 3045 |
jz L(bk_ssse3_cpy_pre) |
| 3046 |
|
| 3047 |
sub $4, %edi |
| 3048 |
sub $4, %ecx |
| 3049 |
sub $4, %edx |
| 3050 |
movl (%edi), %eax |
| 3051 |
movl %eax, (%edx) |
| 3052 |
|
| 3053 |
testl $15, %edx |
| 3054 |
jz L(bk_ssse3_cpy_pre) |
| 3055 |
|
| 3056 |
sub $4, %edi |
| 3057 |
sub $4, %ecx |
| 3058 |
sub $4, %edx |
| 3059 |
movl (%edi), %eax |
| 3060 |
movl %eax, (%edx) |
| 3061 |
|
| 3062 |
L(bk_ssse3_cpy_pre): |
| 3063 |
cmp $64, %ecx |
| 3064 |
jb L(bk_write_more32bytes) |
| 3065 |
|
| 3066 |
.p2align 4 |
| 3067 |
L(bk_ssse3_cpy): |
| 3068 |
sub $64, %edi |
| 3069 |
sub $64, %ecx |
| 3070 |
sub $64, %edx |
| 3071 |
movdqu 0x30(%edi), %xmm3 |
| 3072 |
movdqa %xmm3, 0x30(%edx) |
| 3073 |
movdqu 0x20(%edi), %xmm2 |
| 3074 |
movdqa %xmm2, 0x20(%edx) |
| 3075 |
movdqu 0x10(%edi), %xmm1 |
| 3076 |
movdqa %xmm1, 0x10(%edx) |
| 3077 |
movdqu (%edi), %xmm0 |
| 3078 |
movdqa %xmm0, (%edx) |
| 3079 |
cmp $64, %ecx |
| 3080 |
jae L(bk_ssse3_cpy) |
| 3081 |
jmp L(bk_write_64bytesless) |
| 3082 |
|
| 3083 |
# endif |
| 3084 |
|
| 3085 |
END (MEMCPY) |
| 3086 |
|
| 3087 |
#endif |