about summary refs log tree commit diff
path: root/sysdeps/x86_64/fpu/svml_d_wrapper_impl.h
blob: b71c8636df115a99e2b9fd2d22df80819c55ddbe (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
/* Wrapper implementations of vector math functions.
   Copyright (C) 2014-2021 Free Software Foundation, Inc.
   This file is part of the GNU C Library.

   The GNU C Library is free software; you can redistribute it and/or
   modify it under the terms of the GNU Lesser General Public
   License as published by the Free Software Foundation; either
   version 2.1 of the License, or (at your option) any later version.

   The GNU C Library is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
   Lesser General Public License for more details.

   You should have received a copy of the GNU Lesser General Public
   License along with the GNU C Library; if not, see
   <https://www.gnu.org/licenses/>.  */

/* SSE2 ISA version as wrapper to scalar.  */
.macro WRAPPER_IMPL_SSE2 callee
        subq      $40, %rsp
        cfi_adjust_cfa_offset(40)
        movaps    %xmm0, (%rsp)
        call      JUMPTARGET(\callee)
        movsd     %xmm0, 16(%rsp)
        movsd     8(%rsp), %xmm0
        call      JUMPTARGET(\callee)
        movsd     16(%rsp), %xmm1
        movsd     %xmm0, 24(%rsp)
        unpcklpd  %xmm0, %xmm1
        movaps    %xmm1, %xmm0
        addq      $40, %rsp
        cfi_adjust_cfa_offset(-40)
        ret
.endm

/* 2 argument SSE2 ISA version as wrapper to scalar.  */
.macro WRAPPER_IMPL_SSE2_ff callee
        subq      $56, %rsp
        cfi_adjust_cfa_offset(56)
        movaps    %xmm0, (%rsp)
        movaps    %xmm1, 16(%rsp)
        call      JUMPTARGET(\callee)
        movsd     %xmm0, 32(%rsp)
        movsd     8(%rsp), %xmm0
        movsd     24(%rsp), %xmm1
        call      JUMPTARGET(\callee)
        movsd     32(%rsp), %xmm1
        movsd     %xmm0, 40(%rsp)
        unpcklpd  %xmm0, %xmm1
        movaps    %xmm1, %xmm0
        addq      $56, %rsp
        cfi_adjust_cfa_offset(-56)
        ret
.endm

/* 3 argument SSE2 ISA version as wrapper to scalar.  */
.macro WRAPPER_IMPL_SSE2_fFF callee
        pushq   %rbp
        cfi_adjust_cfa_offset (8)
        cfi_rel_offset (%rbp, 0)
        pushq   %rbx
        cfi_adjust_cfa_offset (8)
        cfi_rel_offset (%rbx, 0)
        movq    %rdi, %rbp
        movq    %rsi, %rbx
        subq    $40, %rsp
        cfi_adjust_cfa_offset(40)
        leaq    16(%rsp), %rsi
        leaq    24(%rsp), %rdi
        movaps  %xmm0, (%rsp)
        call    JUMPTARGET(\callee)
        leaq    16(%rsp), %rsi
        leaq    24(%rsp), %rdi
        movsd   24(%rsp), %xmm0
        movapd  (%rsp), %xmm1
        movsd   %xmm0, 0(%rbp)
        unpckhpd        %xmm1, %xmm1
        movsd   16(%rsp), %xmm0
        movsd   %xmm0, (%rbx)
        movapd  %xmm1, %xmm0
        call    JUMPTARGET(\callee)
        movsd   24(%rsp), %xmm0
        movsd   %xmm0, 8(%rbp)
        movsd   16(%rsp), %xmm0
        movsd   %xmm0, 8(%rbx)
        addq    $40, %rsp
        cfi_adjust_cfa_offset(-40)
        popq    %rbx
        cfi_adjust_cfa_offset (-8)
        cfi_restore (%rbx)
        popq    %rbp
        cfi_adjust_cfa_offset (-8)
        cfi_restore (%rbp)
        ret
.endm

/* AVX/AVX2 ISA version as wrapper to SSE ISA version.  */
.macro WRAPPER_IMPL_AVX callee
        pushq		%rbp
        cfi_adjust_cfa_offset (8)
        cfi_rel_offset (%rbp, 0)
        movq		%rsp, %rbp
        cfi_def_cfa_register (%rbp)
        andq		$-32, %rsp
        subq		$32, %rsp
        vextractf128	$1, %ymm0, (%rsp)
        vzeroupper
        call		HIDDEN_JUMPTARGET(\callee)
        vmovapd		%xmm0, 16(%rsp)
        vmovaps		(%rsp), %xmm0
        call		HIDDEN_JUMPTARGET(\callee)
        vmovapd		%xmm0, %xmm1
        vmovapd		16(%rsp), %xmm0
        vinsertf128	$1, %xmm1, %ymm0, %ymm0
        movq		%rbp, %rsp
        cfi_def_cfa_register (%rsp)
        popq		%rbp
        cfi_adjust_cfa_offset (-8)
        cfi_restore (%rbp)
        ret
.endm

/* 2 argument AVX/AVX2 ISA version as wrapper to SSE ISA version.  */
.macro WRAPPER_IMPL_AVX_ff callee
        pushq     %rbp
        cfi_adjust_cfa_offset (8)
        cfi_rel_offset (%rbp, 0)
        movq      %rsp, %rbp
        cfi_def_cfa_register (%rbp)
        andq      $-32, %rsp
        subq      $64, %rsp
        vextractf128 $1, %ymm0, 16(%rsp)
        vextractf128 $1, %ymm1, (%rsp)
        vzeroupper
        call      HIDDEN_JUMPTARGET(\callee)
        vmovaps   %xmm0, 32(%rsp)
        vmovaps   16(%rsp), %xmm0
        vmovaps   (%rsp), %xmm1
        call      HIDDEN_JUMPTARGET(\callee)
        vmovaps   %xmm0, %xmm1
        vmovaps   32(%rsp), %xmm0
        vinsertf128 $1, %xmm1, %ymm0, %ymm0
        movq      %rbp, %rsp
        cfi_def_cfa_register (%rsp)
        popq      %rbp
        cfi_adjust_cfa_offset (-8)
        cfi_restore (%rbp)
        ret
.endm

/* 3 argument AVX/AVX2 ISA version as wrapper to SSE ISA version.  */
.macro WRAPPER_IMPL_AVX_fFF callee
        pushq     %rbp
        cfi_adjust_cfa_offset (8)
        cfi_rel_offset (%rbp, 0)
        movq      %rsp, %rbp
        cfi_def_cfa_register (%rbp)
        andq      $-32, %rsp
        pushq     %r13
        cfi_adjust_cfa_offset (8)
        cfi_rel_offset (%r13, 0)
        pushq     %r14
        cfi_adjust_cfa_offset (8)
        cfi_rel_offset (%r14, 0)
        subq      $48, %rsp
        movq      %rsi, %r14
        movq      %rdi, %r13
        vextractf128 $1, %ymm0, 32(%rsp)
        vzeroupper
        call      HIDDEN_JUMPTARGET(\callee)
        vmovaps   32(%rsp), %xmm0
        lea       (%rsp), %rdi
        lea       16(%rsp), %rsi
        call      HIDDEN_JUMPTARGET(\callee)
        vmovapd   (%rsp), %xmm0
        vmovapd   16(%rsp), %xmm1
        vmovapd   %xmm0, 16(%r13)
        vmovapd   %xmm1, 16(%r14)
        addq      $48, %rsp
        popq      %r14
        cfi_adjust_cfa_offset (-8)
        cfi_restore (%r14)
        popq      %r13
        cfi_adjust_cfa_offset (-8)
        cfi_restore (%r13)
        movq      %rbp, %rsp
        cfi_def_cfa_register (%rsp)
        popq      %rbp
        cfi_adjust_cfa_offset (-8)
        cfi_restore (%rbp)
        ret
.endm

/* AVX512 ISA version as wrapper to AVX2 ISA version.  */
.macro WRAPPER_IMPL_AVX512 callee
        pushq     %rbp
        cfi_adjust_cfa_offset (8)
        cfi_rel_offset (%rbp, 0)
        movq      %rsp, %rbp
        cfi_def_cfa_register (%rbp)
        andq      $-64, %rsp
        subq      $128, %rsp
        vmovups   %zmm0, (%rsp)
        vmovupd   (%rsp), %ymm0
        call      HIDDEN_JUMPTARGET(\callee)
        vmovupd   %ymm0, 64(%rsp)
        vmovupd   32(%rsp), %ymm0
        call      HIDDEN_JUMPTARGET(\callee)
        vmovupd   %ymm0, 96(%rsp)
        vmovups   64(%rsp), %zmm0
        movq      %rbp, %rsp
        cfi_def_cfa_register (%rsp)
        popq      %rbp
        cfi_adjust_cfa_offset (-8)
        cfi_restore (%rbp)
        ret
.endm

/* 2 argument AVX512 ISA version as wrapper to AVX2 ISA version.  */
.macro WRAPPER_IMPL_AVX512_ff callee
        pushq     %rbp
        cfi_adjust_cfa_offset (8)
        cfi_rel_offset (%rbp, 0)
        movq      %rsp, %rbp
        cfi_def_cfa_register (%rbp)
        andq      $-64, %rsp
        subq      $192, %rsp
        vmovups   %zmm0, (%rsp)
        vmovups   %zmm1, 64(%rsp)
        vmovupd   (%rsp), %ymm0
        vmovupd   64(%rsp), %ymm1
        call      HIDDEN_JUMPTARGET(\callee)
        vmovupd   %ymm0, 128(%rsp)
        vmovupd   32(%rsp), %ymm0
        vmovupd   96(%rsp), %ymm1
        call      HIDDEN_JUMPTARGET(\callee)
        vmovupd   %ymm0, 160(%rsp)
        vmovups   128(%rsp), %zmm0
        movq      %rbp, %rsp
        cfi_def_cfa_register (%rsp)
        popq      %rbp
        cfi_adjust_cfa_offset (-8)
        cfi_restore (%rbp)
        ret
.endm

/* 3 argument AVX512 ISA version as wrapper to AVX2 ISA version.  */
.macro WRAPPER_IMPL_AVX512_fFF callee
        pushq     %rbp
        cfi_adjust_cfa_offset (8)
        cfi_rel_offset (%rbp, 0)
        movq      %rsp, %rbp
        cfi_def_cfa_register (%rbp)
        andq      $-64, %rsp
        pushq     %r12
        cfi_adjust_cfa_offset (8)
        cfi_rel_offset (%r12, 0)
        pushq     %r13
        cfi_adjust_cfa_offset (8)
        cfi_rel_offset (%r13, 0)
        subq      $176, %rsp
        movq      %rsi, %r13
        vmovups   %zmm0, (%rsp)
        movq    %rdi, %r12
        vmovupd (%rsp), %ymm0
        call      HIDDEN_JUMPTARGET(\callee)
        vmovupd   32(%rsp), %ymm0
        lea       64(%rsp), %rdi
        lea       96(%rsp), %rsi
        call      HIDDEN_JUMPTARGET(\callee)
        vmovupd   64(%rsp), %ymm0
        vmovupd   96(%rsp), %ymm1
        vmovupd   %ymm0, 32(%r12)
        vmovupd   %ymm1, 32(%r13)
        vzeroupper
        addq      $176, %rsp
        popq      %r13
        cfi_adjust_cfa_offset (-8)
        cfi_restore (%r13)
        popq      %r12
        cfi_adjust_cfa_offset (-8)
        cfi_restore (%r12)
        movq      %rbp, %rsp
        cfi_def_cfa_register (%rsp)
        popq      %rbp
        cfi_adjust_cfa_offset (-8)
        cfi_restore (%rbp)
        ret
.endm