projects / thirdparty / glibc.git / blame
| Commit | Line | Data |
|---|---|---|
| 693fb948 | 1 | /* strrchr with SSE2 with bsf and bsr |
| 04277e02 | 2 | Copyright (C) 2011-2019 Free Software Foundation, Inc. |
| 693fb948 LD |
3 | Contributed by Intel Corporation. |
| 4 | This file is part of the GNU C Library. | |
| 5 | ||
| 6 | The GNU C Library is free software; you can redistribute it and/or | |
| 7 | modify it under the terms of the GNU Lesser General Public | |
| 8 | License as published by the Free Software Foundation; either | |
| 9 | version 2.1 of the License, or (at your option) any later version. | |
| 10 | ||
| 11 | The GNU C Library is distributed in the hope that it will be useful, | |
| 12 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
| 13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
| 14 | Lesser General Public License for more details. | |
| 15 | ||
| 16 | You should have received a copy of the GNU Lesser General Public | |
| 59ba27a6 | 17 | License along with the GNU C Library; if not, see |
| 5a82c748 | 18 | <https://www.gnu.org/licenses/>. */ |
| 693fb948 | 19 | |
| 4f41c682 | 20 | #if IS_IN (libc) |
| 693fb948 LD |
21 | |
| 22 | # include <sysdep.h> | |
| 23 | ||
| 24 | # define CFI_PUSH(REG) \ | |
| 25 | cfi_adjust_cfa_offset (4); \ | |
| 26 | cfi_rel_offset (REG, 0) | |
| 27 | ||
| 28 | # define CFI_POP(REG) \ | |
| 29 | cfi_adjust_cfa_offset (-4); \ | |
| 30 | cfi_restore (REG) | |
| 31 | ||
| 32 | # define PUSH(REG) pushl REG; CFI_PUSH (REG) | |
| 33 | # define POP(REG) popl REG; CFI_POP (REG) | |
| 34 | ||
| 35 | # define PARMS 4 | |
| 36 | # define STR1 PARMS | |
| 37 | # define STR2 STR1+4 | |
| 38 | ||
| 48882a1a | 39 | .text |
| 693fb948 LD |
40 | ENTRY (__strrchr_sse2_bsf) |
| 41 | ||
| 42 | mov STR1(%esp), %ecx | |
| 43 | movd STR2(%esp), %xmm1 | |
| 44 | ||
| 45 | PUSH (%edi) | |
| 46 | pxor %xmm2, %xmm2 | |
| 47 | mov %ecx, %edi | |
| 48 | punpcklbw %xmm1, %xmm1 | |
| 49 | punpcklbw %xmm1, %xmm1 | |
| 50 | /* ECX has OFFSET. */ | |
| 51 | and $63, %ecx | |
| 52 | cmp $48, %ecx | |
| 53 | pshufd $0, %xmm1, %xmm1 | |
| 54 | ja L(crosscashe) | |
| 55 | ||
| 56 | /* unaligned string. */ | |
| 57 | movdqu (%edi), %xmm0 | |
| 58 | pcmpeqb %xmm0, %xmm2 | |
| 59 | pcmpeqb %xmm1, %xmm0 | |
| 60 | /* Find where NULL is. */ | |
| 61 | pmovmskb %xmm2, %edx | |
| 62 | /* Check if there is a match. */ | |
| 63 | pmovmskb %xmm0, %eax | |
| 64 | ||
| 65 | test %eax, %eax | |
| 66 | jnz L(unaligned_match1) | |
| 67 | ||
| 68 | test %edx, %edx | |
| 69 | jnz L(return_null) | |
| 70 | ||
| 71 | and $-16, %edi | |
| 72 | add $16, %edi | |
| 73 | ||
| 74 | PUSH (%esi) | |
| 75 | PUSH (%ebx) | |
| 76 | ||
| 77 | xor %ebx, %ebx | |
| 78 | jmp L(loop) | |
| 79 | ||
| 80 | CFI_POP (%esi) | |
| 81 | CFI_POP (%ebx) | |
| 82 | ||
| 83 | .p2align 4 | |
| 84 | L(unaligned_return_value1): | |
| 85 | bsf %edx, %ecx | |
| 86 | mov $2, %edx | |
| 87 | shl %cl, %edx | |
| 88 | sub $1, %edx | |
| 89 | and %edx, %eax | |
| 90 | jz L(return_null) | |
| 91 | bsr %eax, %eax | |
| 92 | add %edi, %eax | |
| 93 | POP (%edi) | |
| 94 | ret | |
| 95 | CFI_PUSH (%edi) | |
| 96 | ||
| 97 | .p2align 4 | |
| 98 | L(unaligned_match1): | |
| 99 | test %edx, %edx | |
| 100 | jnz L(unaligned_return_value1) | |
| 101 | ||
| 102 | PUSH (%esi) | |
| 103 | PUSH (%ebx) | |
| 104 | ||
| 105 | mov %eax, %ebx | |
| 106 | lea 16(%edi), %esi | |
| 107 | and $-16, %edi | |
| 108 | add $16, %edi | |
| 109 | jmp L(loop) | |
| 110 | ||
| 111 | CFI_POP (%esi) | |
| 112 | CFI_POP (%ebx) | |
| 113 | ||
| 114 | .p2align 4 | |
| 115 | L(crosscashe): | |
| 116 | /* Hancle unaligned string. */ | |
| 117 | and $15, %ecx | |
| 118 | and $-16, %edi | |
| 119 | pxor %xmm3, %xmm3 | |
| 120 | movdqa (%edi), %xmm0 | |
| 121 | pcmpeqb %xmm0, %xmm3 | |
| 122 | pcmpeqb %xmm1, %xmm0 | |
| 123 | /* Find where NULL is. */ | |
| 124 | pmovmskb %xmm3, %edx | |
| 125 | /* Check if there is a match. */ | |
| 126 | pmovmskb %xmm0, %eax | |
| 127 | /* Remove the leading bytes. */ | |
| 128 | shr %cl, %edx | |
| 129 | shr %cl, %eax | |
| 130 | ||
| 131 | test %eax, %eax | |
| 132 | jnz L(unaligned_match) | |
| 133 | ||
| 134 | test %edx, %edx | |
| 135 | jnz L(return_null) | |
| 136 | ||
| 137 | add $16, %edi | |
| 138 | ||
| 139 | PUSH (%esi) | |
| 140 | PUSH (%ebx) | |
| 141 | ||
| 142 | xor %ebx, %ebx | |
| 143 | jmp L(loop) | |
| 144 | ||
| 145 | CFI_POP (%esi) | |
| 146 | CFI_POP (%ebx) | |
| 147 | ||
| 148 | .p2align 4 | |
| 149 | L(unaligned_return_value): | |
| 150 | add %ecx, %edi | |
| 151 | bsf %edx, %ecx | |
| 152 | mov $2, %edx | |
| 153 | shl %cl, %edx | |
| 154 | sub $1, %edx | |
| 155 | and %edx, %eax | |
| 156 | jz L(return_null) | |
| 157 | bsr %eax, %eax | |
| 158 | add %edi, %eax | |
| 159 | POP (%edi) | |
| 160 | ret | |
| 161 | CFI_PUSH (%edi) | |
| 162 | ||
| 163 | .p2align 4 | |
| 164 | L(unaligned_match): | |
| 165 | test %edx, %edx | |
| 166 | jnz L(unaligned_return_value) | |
| 167 | ||
| 168 | PUSH (%esi) | |
| 169 | PUSH (%ebx) | |
| 170 | ||
| 171 | mov %eax, %ebx | |
| 172 | add $16, %edi | |
| 173 | lea (%edi, %ecx), %esi | |
| 174 | ||
| 175 | /* Loop start on aligned string. */ | |
| 176 | .p2align 4 | |
| 177 | L(loop): | |
| 178 | movdqa (%edi), %xmm0 | |
| 179 | pcmpeqb %xmm0, %xmm2 | |
| 180 | add $16, %edi | |
| 181 | pcmpeqb %xmm1, %xmm0 | |
| 182 | pmovmskb %xmm2, %ecx | |
| 183 | pmovmskb %xmm0, %eax | |
| 184 | or %eax, %ecx | |
| 185 | jnz L(matches) | |
| 186 | ||
| 187 | movdqa (%edi), %xmm0 | |
| 188 | pcmpeqb %xmm0, %xmm2 | |
| 189 | add $16, %edi | |
| 190 | pcmpeqb %xmm1, %xmm0 | |
| 191 | pmovmskb %xmm2, %ecx | |
| 192 | pmovmskb %xmm0, %eax | |
| 193 | or %eax, %ecx | |
| 194 | jnz L(matches) | |
| 195 | ||
| 196 | movdqa (%edi), %xmm0 | |
| 197 | pcmpeqb %xmm0, %xmm2 | |
| 198 | add $16, %edi | |
| 199 | pcmpeqb %xmm1, %xmm0 | |
| 200 | pmovmskb %xmm2, %ecx | |
| 201 | pmovmskb %xmm0, %eax | |
| 202 | or %eax, %ecx | |
| 203 | jnz L(matches) | |
| 204 | ||
| 205 | movdqa (%edi), %xmm0 | |
| 206 | pcmpeqb %xmm0, %xmm2 | |
| 207 | add $16, %edi | |
| 208 | pcmpeqb %xmm1, %xmm0 | |
| 209 | pmovmskb %xmm2, %ecx | |
| 210 | pmovmskb %xmm0, %eax | |
| 211 | or %eax, %ecx | |
| 212 | jz L(loop) | |
| 213 | ||
| 214 | L(matches): | |
| 215 | test %eax, %eax | |
| 216 | jnz L(match) | |
| 217 | L(return_value): | |
| 218 | test %ebx, %ebx | |
| 219 | jz L(return_null_1) | |
| 220 | bsr %ebx, %eax | |
| 221 | add %esi, %eax | |
| 222 | ||
| 223 | POP (%ebx) | |
| 224 | POP (%esi) | |
| 225 | ||
| 226 | sub $16, %eax | |
| 227 | POP (%edi) | |
| 228 | ret | |
| 229 | ||
| 230 | CFI_PUSH (%edi) | |
| 231 | CFI_PUSH (%ebx) | |
| 232 | CFI_PUSH (%esi) | |
| 233 | ||
| 234 | .p2align 4 | |
| 235 | L(match): | |
| 236 | pmovmskb %xmm2, %ecx | |
| 237 | test %ecx, %ecx | |
| 238 | jnz L(return_value_1) | |
| 239 | mov %eax, %ebx | |
| 240 | mov %edi, %esi | |
| 241 | jmp L(loop) | |
| 242 | ||
| 243 | .p2align 4 | |
| 244 | L(return_value_1): | |
| 245 | bsf %ecx, %ecx | |
| 246 | mov $2, %edx | |
| 247 | shl %cl, %edx | |
| 248 | sub $1, %edx | |
| 249 | and %edx, %eax | |
| 250 | jz L(return_value) | |
| 251 | ||
| 252 | POP (%ebx) | |
| 253 | POP (%esi) | |
| 254 | ||
| 255 | bsr %eax, %eax | |
| 256 | add %edi, %eax | |
| 257 | sub $16, %eax | |
| 258 | POP (%edi) | |
| 259 | ret | |
| 260 | ||
| 261 | CFI_PUSH (%edi) | |
| 262 | /* Return NULL. */ | |
| 263 | .p2align 4 | |
| 264 | L(return_null): | |
| 265 | xor %eax, %eax | |
| 266 | POP (%edi) | |
| 267 | ret | |
| 268 | ||
| 269 | CFI_PUSH (%edi) | |
| 270 | CFI_PUSH (%ebx) | |
| 271 | CFI_PUSH (%esi) | |
| 272 | /* Return NULL. */ | |
| 273 | .p2align 4 | |
| 274 | L(return_null_1): | |
| 275 | POP (%ebx) | |
| 276 | POP (%esi) | |
| 277 | POP (%edi) | |
| 278 | xor %eax, %eax | |
| 279 | ret | |
| 280 | ||
| 281 | END (__strrchr_sse2_bsf) | |
| 282 | #endif |