projects / thirdparty / glibc.git / blame
| Commit | Line | Data |
|---|---|---|
| 04f496d6 AS |
1 | /* Wrapper implementations of vector math functions. |
| 2 | Copyright (C) 2014-2015 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 | <http://www.gnu.org/licenses/>. */ | |
| 18 | ||
| 19 | /* SSE2 ISA version as wrapper to scalar. */ | |
| 20 | .macro WRAPPER_IMPL_SSE2 callee | |
| 21 | subq $40, %rsp | |
| 22 | cfi_adjust_cfa_offset(40) | |
| 23 | movaps %xmm0, (%rsp) | |
| 24 | call \callee@PLT | |
| 25 | movss %xmm0, 16(%rsp) | |
| 26 | movss 4(%rsp), %xmm0 | |
| 27 | call \callee@PLT | |
| 28 | movss %xmm0, 20(%rsp) | |
| 29 | movss 8(%rsp), %xmm0 | |
| 30 | call \callee@PLT | |
| 31 | movss %xmm0, 24(%rsp) | |
| 32 | movss 12(%rsp), %xmm0 | |
| 33 | call \callee@PLT | |
| 34 | movss 16(%rsp), %xmm3 | |
| 35 | movss 20(%rsp), %xmm2 | |
| 36 | movss 24(%rsp), %xmm1 | |
| 37 | movss %xmm0, 28(%rsp) | |
| 38 | unpcklps %xmm1, %xmm3 | |
| 39 | unpcklps %xmm0, %xmm2 | |
| 40 | unpcklps %xmm2, %xmm3 | |
| 41 | movaps %xmm3, %xmm0 | |
| 42 | addq $40, %rsp | |
| 43 | cfi_adjust_cfa_offset(-40) | |
| 44 | ret | |
| 45 | .endm | |
| 46 | ||
| 8aa92022 AS |
47 | /* 2 argument SSE2 ISA version as wrapper to scalar. */ |
| 48 | .macro WRAPPER_IMPL_SSE2_ff callee | |
| 49 | subq $56, %rsp | |
| 50 | cfi_adjust_cfa_offset(56) | |
| 51 | movaps %xmm0, (%rsp) | |
| 52 | movaps %xmm1, 16(%rsp) | |
| 53 | call \callee@PLT | |
| 54 | movss %xmm0, 32(%rsp) | |
| 55 | movss 4(%rsp), %xmm0 | |
| 56 | movss 20(%rsp), %xmm1 | |
| 57 | call \callee@PLT | |
| 58 | movss %xmm0, 36(%rsp) | |
| 59 | movss 8(%rsp), %xmm0 | |
| 60 | movss 24(%rsp), %xmm1 | |
| 61 | call \callee@PLT | |
| 62 | movss %xmm0, 40(%rsp) | |
| 63 | movss 12(%rsp), %xmm0 | |
| 64 | movss 28(%rsp), %xmm1 | |
| 65 | call \callee@PLT | |
| 66 | movss 32(%rsp), %xmm3 | |
| 67 | movss 36(%rsp), %xmm2 | |
| 68 | movss 40(%rsp), %xmm1 | |
| 69 | movss %xmm0, 44(%rsp) | |
| 70 | unpcklps %xmm1, %xmm3 | |
| 71 | unpcklps %xmm0, %xmm2 | |
| 72 | unpcklps %xmm2, %xmm3 | |
| 73 | movaps %xmm3, %xmm0 | |
| 74 | addq $56, %rsp | |
| 75 | cfi_adjust_cfa_offset(-56) | |
| 76 | ret | |
| 77 | .endm | |
| 78 | ||
| 04f496d6 AS |
79 | /* AVX/AVX2 ISA version as wrapper to SSE ISA version. */ |
| 80 | .macro WRAPPER_IMPL_AVX callee | |
| 81 | pushq %rbp | |
| 82 | cfi_adjust_cfa_offset (8) | |
| 83 | cfi_rel_offset (%rbp, 0) | |
| 84 | movq %rsp, %rbp | |
| 85 | cfi_def_cfa_register (%rbp) | |
| 86 | andq $-32, %rsp | |
| 87 | subq $32, %rsp | |
| 88 | vextractf128 $1, %ymm0, (%rsp) | |
| 89 | vzeroupper | |
| 90 | call HIDDEN_JUMPTARGET(\callee) | |
| 91 | vmovaps %xmm0, 16(%rsp) | |
| 92 | vmovaps (%rsp), %xmm0 | |
| 93 | call HIDDEN_JUMPTARGET(\callee) | |
| 94 | vmovaps %xmm0, %xmm1 | |
| 95 | vmovaps 16(%rsp), %xmm0 | |
| 96 | vinsertf128 $1, %xmm1, %ymm0, %ymm0 | |
| 97 | movq %rbp, %rsp | |
| 98 | cfi_def_cfa_register (%rsp) | |
| 99 | popq %rbp | |
| 100 | cfi_adjust_cfa_offset (-8) | |
| 101 | cfi_restore (%rbp) | |
| 102 | ret | |
| 103 | .endm | |
| 104 | ||
| 8aa92022 AS |
105 | /* 2 argument AVX/AVX2 ISA version as wrapper to SSE ISA version. */ |
| 106 | .macro WRAPPER_IMPL_AVX_ff callee | |
| 107 | pushq %rbp | |
| 108 | cfi_adjust_cfa_offset (8) | |
| 109 | cfi_rel_offset (%rbp, 0) | |
| 110 | movq %rsp, %rbp | |
| 111 | cfi_def_cfa_register (%rbp) | |
| 112 | andq $-32, %rsp | |
| 113 | subq $64, %rsp | |
| 114 | vextractf128 $1, %ymm0, 16(%rsp) | |
| 115 | vextractf128 $1, %ymm1, (%rsp) | |
| 116 | vzeroupper | |
| 117 | call HIDDEN_JUMPTARGET(\callee) | |
| 118 | vmovaps %xmm0, 32(%rsp) | |
| 119 | vmovaps 16(%rsp), %xmm0 | |
| 120 | vmovaps (%rsp), %xmm1 | |
| 121 | call HIDDEN_JUMPTARGET(\callee) | |
| 122 | vmovaps %xmm0, %xmm1 | |
| 123 | vmovaps 32(%rsp), %xmm0 | |
| 124 | vinsertf128 $1, %xmm1, %ymm0, %ymm0 | |
| 125 | movq %rbp, %rsp | |
| 126 | cfi_def_cfa_register (%rsp) | |
| 127 | popq %rbp | |
| 128 | cfi_adjust_cfa_offset (-8) | |
| 129 | cfi_restore (%rbp) | |
| 130 | ret | |
| 131 | .endm | |
| 132 | ||
| 04f496d6 AS |
133 | /* AVX512 ISA version as wrapper to AVX2 ISA version. */ |
| 134 | .macro WRAPPER_IMPL_AVX512 callee | |
| 135 | pushq %rbp | |
| 136 | cfi_adjust_cfa_offset (8) | |
| 137 | cfi_rel_offset (%rbp, 0) | |
| 138 | movq %rsp, %rbp | |
| 139 | cfi_def_cfa_register (%rbp) | |
| 140 | andq $-64, %rsp | |
| 141 | subq $64, %rsp | |
| 142 | /* Below is encoding for vmovaps %zmm0, (%rsp). */ | |
| 143 | .byte 0x62 | |
| 144 | .byte 0xf1 | |
| 145 | .byte 0x7c | |
| 146 | .byte 0x48 | |
| 147 | .byte 0x29 | |
| 148 | .byte 0x04 | |
| 149 | .byte 0x24 | |
| 150 | /* Below is encoding for vmovaps (%rsp), %ymm0. */ | |
| 151 | .byte 0xc5 | |
| 152 | .byte 0xfc | |
| 153 | .byte 0x28 | |
| 154 | .byte 0x04 | |
| 155 | .byte 0x24 | |
| 156 | call HIDDEN_JUMPTARGET(\callee) | |
| 157 | /* Below is encoding for vmovaps 32(%rsp), %ymm0. */ | |
| 158 | .byte 0xc5 | |
| 159 | .byte 0xfc | |
| 160 | .byte 0x28 | |
| 161 | .byte 0x44 | |
| 162 | .byte 0x24 | |
| 163 | .byte 0x20 | |
| 164 | call HIDDEN_JUMPTARGET(\callee) | |
| 165 | movq %rbp, %rsp | |
| 166 | cfi_def_cfa_register (%rsp) | |
| 167 | popq %rbp | |
| 168 | cfi_adjust_cfa_offset (-8) | |
| 169 | cfi_restore (%rbp) | |
| 170 | ret | |
| 171 | .endm | |
| 8aa92022 AS |
172 | |
| 173 | /* 2 argument AVX512 ISA version as wrapper to AVX2 ISA version. */ | |
| 174 | .macro WRAPPER_IMPL_AVX512_ff callee | |
| 175 | pushq %rbp | |
| 176 | cfi_adjust_cfa_offset (8) | |
| 177 | cfi_rel_offset (%rbp, 0) | |
| 178 | movq %rsp, %rbp | |
| 179 | cfi_def_cfa_register (%rbp) | |
| 180 | andq $-64, %rsp | |
| 181 | subq $128, %rsp | |
| 182 | /* Below is encoding for vmovaps %zmm0, (%rsp). */ | |
| 183 | .byte 0x62 | |
| 184 | .byte 0xf1 | |
| 185 | .byte 0x7c | |
| 186 | .byte 0x48 | |
| 187 | .byte 0x29 | |
| 188 | .byte 0x04 | |
| 189 | .byte 0x24 | |
| 190 | /* Below is encoding for vmovaps %zmm1, 64(%rsp). */ | |
| 191 | .byte 0x62 | |
| 192 | .byte 0xf1 | |
| 193 | .byte 0x7c | |
| 194 | .byte 0x48 | |
| 195 | .byte 0x29 | |
| 196 | .byte 0x4c | |
| 197 | .byte 0x24 | |
| 198 | /* Below is encoding for vmovaps (%rsp), %ymm0. */ | |
| 199 | .byte 0xc5 | |
| 200 | .byte 0xfc | |
| 201 | .byte 0x28 | |
| 202 | .byte 0x04 | |
| 203 | .byte 0x24 | |
| 204 | /* Below is encoding for vmovaps 64(%rsp), %ymm1. */ | |
| 205 | .byte 0xc5 | |
| 206 | .byte 0xfc | |
| 207 | .byte 0x28 | |
| 208 | .byte 0x4c | |
| 209 | .byte 0x24 | |
| 210 | .byte 0x40 | |
| 211 | call HIDDEN_JUMPTARGET(\callee) | |
| 212 | /* Below is encoding for vmovaps 32(%rsp), %ymm0. */ | |
| 213 | .byte 0xc5 | |
| 214 | .byte 0xfc | |
| 215 | .byte 0x28 | |
| 216 | .byte 0x44 | |
| 217 | .byte 0x24 | |
| 218 | .byte 0x20 | |
| 219 | /* Below is encoding for vmovaps 96(%rsp), %ymm1. */ | |
| 220 | .byte 0xc5 | |
| 221 | .byte 0xfc | |
| 222 | .byte 0x28 | |
| 223 | .byte 0x4c | |
| 224 | .byte 0x24 | |
| 225 | .byte 0x60 | |
| 226 | call HIDDEN_JUMPTARGET(\callee) | |
| 227 | movq %rbp, %rsp | |
| 228 | cfi_def_cfa_register (%rsp) | |
| 229 | popq %rbp | |
| 230 | cfi_adjust_cfa_offset (-8) | |
| 231 | cfi_restore (%rbp) | |
| 232 | ret | |
| 233 | .endm |