| 1 | /* -----------------------------------------------------------------------
|
|---|
| 2 | unix64.S - Copyright (c) 2002 Bo Thorsen <bo@suse.de>
|
|---|
| 3 |
|
|---|
| 4 | x86-64 Foreign Function Interface
|
|---|
| 5 |
|
|---|
| 6 | Permission is hereby granted, free of charge, to any person obtaining
|
|---|
| 7 | a copy of this software and associated documentation files (the
|
|---|
| 8 | ``Software''), to deal in the Software without restriction, including
|
|---|
| 9 | without limitation the rights to use, copy, modify, merge, publish,
|
|---|
| 10 | distribute, sublicense, and/or sell copies of the Software, and to
|
|---|
| 11 | permit persons to whom the Software is furnished to do so, subject to
|
|---|
| 12 | the following conditions:
|
|---|
| 13 |
|
|---|
| 14 | The above copyright notice and this permission notice shall be included
|
|---|
| 15 | in all copies or substantial portions of the Software.
|
|---|
| 16 |
|
|---|
| 17 | THE SOFTWARE IS PROVIDED ``AS IS'', WITHOUT WARRANTY OF ANY KIND, EXPRESS
|
|---|
| 18 | OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
|
|---|
| 19 | MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
|
|---|
| 20 | IN NO EVENT SHALL CYGNUS SOLUTIONS BE LIABLE FOR ANY CLAIM, DAMAGES OR
|
|---|
| 21 | OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
|
|---|
| 22 | ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
|
|---|
| 23 | OTHER DEALINGS IN THE SOFTWARE.
|
|---|
| 24 | ----------------------------------------------------------------------- */
|
|---|
| 25 |
|
|---|
| 26 | #ifdef __x86_64__
|
|---|
| 27 | #define LIBFFI_ASM
|
|---|
| 28 | #include <ffi.h>
|
|---|
| 29 |
|
|---|
| 30 | .section .rodata
|
|---|
| 31 | .LC0:
|
|---|
| 32 | .string "asm in progress %lld\n"
|
|---|
| 33 | .LC1:
|
|---|
| 34 | .string "asm in progress\n"
|
|---|
| 35 | .text
|
|---|
| 36 | .align 2
|
|---|
| 37 | .globl ffi_call_UNIX64
|
|---|
| 38 | .type ffi_call_UNIX64,@function
|
|---|
| 39 |
|
|---|
| 40 | ffi_call_UNIX64:
|
|---|
| 41 | .LFB1:
|
|---|
| 42 | pushq %rbp
|
|---|
| 43 | .LCFI0:
|
|---|
| 44 | movq %rsp, %rbp
|
|---|
| 45 | .LCFI1:
|
|---|
| 46 | /* Save all arguments */
|
|---|
| 47 | subq $48, %rsp
|
|---|
| 48 | .LCFI2:
|
|---|
| 49 | movq %rdi, -8(%rbp) /* ffi_prep_args */
|
|---|
| 50 | movq %rsi, -16(%rbp) /* ffi_fill_return_value */
|
|---|
| 51 | movq %rdx, -24(%rbp) /* ecif */
|
|---|
| 52 | movq %rcx, -32(%rbp) /* cif->bytes */
|
|---|
| 53 | movq %r8, -40(%rbp) /* ecif.rvalue */
|
|---|
| 54 | movq %r9, -48(%rbp) /* fn */
|
|---|
| 55 |
|
|---|
| 56 | /* Make room for all of the new args and the register args */
|
|---|
| 57 | addl $176, %ecx
|
|---|
| 58 | .LCFI3:
|
|---|
| 59 | subq %rcx, %rsp
|
|---|
| 60 | .LCFI4:
|
|---|
| 61 | /* Setup the call to ffi_prep_args. */
|
|---|
| 62 | movq %rdi, %rax /* &ffi_prep_args */
|
|---|
| 63 | movq %rsp, %rdi /* stackLayout */
|
|---|
| 64 | movq %rdx, %rsi /* ecif */
|
|---|
| 65 | call *%rax /* ffi_prep_args(stackLayout, ecif);*/
|
|---|
| 66 |
|
|---|
| 67 | /* ffi_prep_args have put all the register contents into the */
|
|---|
| 68 | /* stackLayout struct. Now put the register values in place. */
|
|---|
| 69 | movq (%rsp), %rdi
|
|---|
| 70 | movq 8(%rsp), %rsi
|
|---|
| 71 | movq 16(%rsp), %rdx
|
|---|
| 72 | movq 24(%rsp), %rcx
|
|---|
| 73 | movq 32(%rsp), %r8
|
|---|
| 74 | movq 40(%rsp), %r9
|
|---|
| 75 | movaps 48(%rsp), %xmm0
|
|---|
| 76 | movaps 64(%rsp), %xmm1
|
|---|
| 77 | movaps 80(%rsp), %xmm2
|
|---|
| 78 | movaps 96(%rsp), %xmm3
|
|---|
| 79 | movaps 112(%rsp), %xmm4
|
|---|
| 80 | movaps 128(%rsp), %xmm5
|
|---|
| 81 | movaps 144(%rsp), %xmm6
|
|---|
| 82 | movaps 160(%rsp), %xmm7
|
|---|
| 83 |
|
|---|
| 84 | /* Remove space for stackLayout so stack arguments are placed
|
|---|
| 85 | correctly for the call. */
|
|---|
| 86 | .LCFI5:
|
|---|
| 87 | addq $176, %rsp
|
|---|
| 88 | .LCFI6:
|
|---|
| 89 | /* Call the user function. */
|
|---|
| 90 | call *-48(%rbp)
|
|---|
| 91 |
|
|---|
| 92 | /* Make stack space for the return_value struct. */
|
|---|
| 93 | subq $64, %rsp
|
|---|
| 94 |
|
|---|
| 95 | /* Fill in all potential return values to this struct. */
|
|---|
| 96 | movq %rax, (%rsp)
|
|---|
| 97 | movq %rdx, 8(%rsp)
|
|---|
| 98 | movaps %xmm0, 16(%rsp)
|
|---|
| 99 | movaps %xmm1, 32(%rsp)
|
|---|
| 100 | fstpt 48(%rsp)
|
|---|
| 101 |
|
|---|
| 102 | /* Now call ffi_fill_return_value. */
|
|---|
| 103 | movq %rsp, %rdi /* struct return_value */
|
|---|
| 104 | movq -24(%rbp), %rsi /* ecif */
|
|---|
| 105 | movq -16(%rbp), %rax /* &ffi_fill_return_value */
|
|---|
| 106 | call *%rax /* call it */
|
|---|
| 107 |
|
|---|
| 108 | /* And the work is done. */
|
|---|
| 109 | leave
|
|---|
| 110 | ret
|
|---|
| 111 | .LFE1:
|
|---|
| 112 | .ffi_call_UNIX64_end:
|
|---|
| 113 | .size ffi_call_UNIX64,.ffi_call_UNIX64_end-ffi_call_UNIX64
|
|---|
| 114 |
|
|---|
| 115 | .text
|
|---|
| 116 | .align 2
|
|---|
| 117 | .globl float2sse
|
|---|
| 118 | .type float2sse,@function
|
|---|
| 119 | float2sse:
|
|---|
| 120 | /* Save the contents of this sse-float in a pointer. */
|
|---|
| 121 | movaps %xmm0, (%rdi)
|
|---|
| 122 | ret
|
|---|
| 123 |
|
|---|
| 124 | .align 2
|
|---|
| 125 | .globl floatfloat2sse
|
|---|
| 126 | .type floatfloat2sse,@function
|
|---|
| 127 | floatfloat2sse:
|
|---|
| 128 | /* Save the contents of these two sse-floats in a pointer. */
|
|---|
| 129 | movq (%rdi), %xmm0
|
|---|
| 130 | movaps %xmm0, (%rsi)
|
|---|
| 131 | ret
|
|---|
| 132 |
|
|---|
| 133 | .align 2
|
|---|
| 134 | .globl double2sse
|
|---|
| 135 | .type double2sse,@function
|
|---|
| 136 | double2sse:
|
|---|
| 137 | /* Save the contents of this sse-double in a pointer. */
|
|---|
| 138 | movaps %xmm0, (%rdi)
|
|---|
| 139 | ret
|
|---|
| 140 |
|
|---|
| 141 | .align 2
|
|---|
| 142 | .globl sse2float
|
|---|
| 143 | .type sse2float,@function
|
|---|
| 144 | sse2float:
|
|---|
| 145 | /* Save the contents of this sse-float in a pointer. */
|
|---|
| 146 | movaps (%rdi), %xmm0
|
|---|
| 147 | ret
|
|---|
| 148 |
|
|---|
| 149 | .align 2
|
|---|
| 150 | .globl sse2double
|
|---|
| 151 | .type sse2double,@function
|
|---|
| 152 | sse2double:
|
|---|
| 153 | /* Save the contents of this pointer in a sse-double. */
|
|---|
| 154 | movaps (%rdi), %xmm0
|
|---|
| 155 | ret
|
|---|
| 156 |
|
|---|
| 157 | .align 2
|
|---|
| 158 | .globl sse2floatfloat
|
|---|
| 159 | .type sse2floatfloat,@function
|
|---|
| 160 | sse2floatfloat:
|
|---|
| 161 | /* Save the contents of this pointer in two sse-floats. */
|
|---|
| 162 | movaps (%rdi), %xmm0
|
|---|
| 163 | movq %xmm0, (%rsi)
|
|---|
| 164 | ret
|
|---|
| 165 |
|
|---|
| 166 | .align 2
|
|---|
| 167 | .globl ffi_closure_UNIX64
|
|---|
| 168 | .type ffi_closure_UNIX64,@function
|
|---|
| 169 |
|
|---|
| 170 | ffi_closure_UNIX64:
|
|---|
| 171 | .LFB2:
|
|---|
| 172 | pushq %rbp
|
|---|
| 173 | .LCFI10:
|
|---|
| 174 | movq %rsp, %rbp
|
|---|
| 175 | .LCFI11:
|
|---|
| 176 | subq $240, %rsp
|
|---|
| 177 | .LCFI12:
|
|---|
| 178 | movq %rdi, -176(%rbp)
|
|---|
| 179 | movq %rsi, -168(%rbp)
|
|---|
| 180 | movq %rdx, -160(%rbp)
|
|---|
| 181 | movq %rcx, -152(%rbp)
|
|---|
| 182 | movq %r8, -144(%rbp)
|
|---|
| 183 | movq %r9, -136(%rbp)
|
|---|
| 184 | /* FIXME: We can avoid all this stashing of XMM registers by
|
|---|
| 185 | (in ffi_prep_closure) computing the number of
|
|---|
| 186 | floating-point args and moving it into %rax before calling
|
|---|
| 187 | this function. Once this is done, uncomment the next few
|
|---|
| 188 | lines and only the essential XMM registers will be written
|
|---|
| 189 | to memory. This is a significant saving. */
|
|---|
| 190 | /* movzbl %al, %eax */
|
|---|
| 191 | /* movq %rax, %rdx */
|
|---|
| 192 | /* leaq 0(,%rdx,4), %rax */
|
|---|
| 193 | /* leaq 2f(%rip), %rdx */
|
|---|
| 194 | /* subq %rax, %rdx */
|
|---|
| 195 | leaq -1(%rbp), %rax
|
|---|
| 196 | /* jmp *%rdx */
|
|---|
| 197 | movaps %xmm7, -15(%rax)
|
|---|
| 198 | movaps %xmm6, -31(%rax)
|
|---|
| 199 | movaps %xmm5, -47(%rax)
|
|---|
| 200 | movaps %xmm4, -63(%rax)
|
|---|
| 201 | movaps %xmm3, -79(%rax)
|
|---|
| 202 | movaps %xmm2, -95(%rax)
|
|---|
| 203 | movaps %xmm1, -111(%rax)
|
|---|
| 204 | movaps %xmm0, -127(%rax)
|
|---|
| 205 | 2:
|
|---|
| 206 | movl %edi, -180(%rbp)
|
|---|
| 207 | movl $0, -224(%rbp)
|
|---|
| 208 | movl $48, -220(%rbp)
|
|---|
| 209 | leaq 16(%rbp), %rax
|
|---|
| 210 | movq %rax, -216(%rbp)
|
|---|
| 211 | leaq -176(%rbp), %rdx
|
|---|
| 212 | movq %rdx, -208(%rbp)
|
|---|
| 213 | leaq -224(%rbp), %rsi
|
|---|
| 214 | movq %r10, %rdi
|
|---|
| 215 | movq %rsp, %rdx
|
|---|
| 216 | call ffi_closure_UNIX64_inner@PLT
|
|---|
| 217 |
|
|---|
| 218 | cmpl $FFI_TYPE_FLOAT, %eax
|
|---|
| 219 | je 1f
|
|---|
| 220 | cmpl $FFI_TYPE_DOUBLE, %eax
|
|---|
| 221 | je 2f
|
|---|
| 222 | cmpl $FFI_TYPE_LONGDOUBLE, %eax
|
|---|
| 223 | je 3f
|
|---|
| 224 | cmpl $FFI_TYPE_STRUCT, %eax
|
|---|
| 225 | je 4f
|
|---|
| 226 | popq %rax
|
|---|
| 227 | leave
|
|---|
| 228 | ret
|
|---|
| 229 | 1:
|
|---|
| 230 | 2:
|
|---|
| 231 | 3:
|
|---|
| 232 | movaps -240(%rbp), %xmm0
|
|---|
| 233 | leave
|
|---|
| 234 | ret
|
|---|
| 235 | 4:
|
|---|
| 236 | leave
|
|---|
| 237 | ret
|
|---|
| 238 | .LFE2:
|
|---|
| 239 |
|
|---|
| 240 | .section .eh_frame,"a",@progbits
|
|---|
| 241 | .Lframe0:
|
|---|
| 242 | .long .LECIE1-.LSCIE1
|
|---|
| 243 | .LSCIE1:
|
|---|
| 244 | .long 0x0
|
|---|
| 245 | .byte 0x1
|
|---|
| 246 | .string "zR"
|
|---|
| 247 | .uleb128 0x1
|
|---|
| 248 | .sleb128 -8
|
|---|
| 249 | .byte 0x10
|
|---|
| 250 | .uleb128 0x1
|
|---|
| 251 | .byte 0x1b
|
|---|
| 252 | .byte 0xc
|
|---|
| 253 | .uleb128 0x7
|
|---|
| 254 | .uleb128 0x8
|
|---|
| 255 | .byte 0x90
|
|---|
| 256 | .uleb128 0x1
|
|---|
| 257 | .align 8
|
|---|
| 258 | .LECIE1:
|
|---|
| 259 | .LSFDE1:
|
|---|
| 260 | .long .LEFDE1-.LASFDE1
|
|---|
| 261 | .LASFDE1:
|
|---|
| 262 | .long .LASFDE1-.Lframe0
|
|---|
| 263 |
|
|---|
| 264 | .long .LFB1-.
|
|---|
| 265 | .long .LFE1-.LFB1
|
|---|
| 266 | .uleb128 0x0
|
|---|
| 267 | .byte 0x4 # DW_CFA_advance_loc4
|
|---|
| 268 | .long .LCFI0-.LFB1
|
|---|
| 269 | .byte 0xe # DW_CFA_def_cfa_offset
|
|---|
| 270 | .uleb128 0x10
|
|---|
| 271 | .byte 0x86 # DW_CFA_offset: r6 at cfa-16
|
|---|
| 272 | .uleb128 0x2
|
|---|
| 273 | .byte 0x4 # DW_CFA_advance_loc4
|
|---|
| 274 | .long .LCFI1-.LCFI0
|
|---|
| 275 | .byte 0x86 # DW_CFA_offset: r6 at cfa-16
|
|---|
| 276 | .uleb128 0x2
|
|---|
| 277 | .byte 0xd # DW_CFA_def_cfa_reg: r6
|
|---|
| 278 | .uleb128 0x6
|
|---|
| 279 | .align 8
|
|---|
| 280 | .LEFDE1:
|
|---|
| 281 | .LSFDE3:
|
|---|
| 282 | .long .LEFDE3-.LASFDE3 # FDE Length
|
|---|
| 283 | .LASFDE3:
|
|---|
| 284 | .long .LASFDE3-.Lframe0 # FDE CIE offset
|
|---|
| 285 |
|
|---|
| 286 | .long .LFB2-. # FDE initial location
|
|---|
| 287 | .long .LFE2-.LFB2 # FDE address range
|
|---|
| 288 | .uleb128 0x0 # Augmentation size
|
|---|
| 289 | .byte 0x4 # DW_CFA_advance_loc4
|
|---|
| 290 | .long .LCFI10-.LFB2
|
|---|
| 291 | .byte 0xe # DW_CFA_def_cfa_offset
|
|---|
| 292 | .uleb128 0x10
|
|---|
| 293 | .byte 0x86 # DW_CFA_offset, column 0x6
|
|---|
| 294 | .uleb128 0x2
|
|---|
| 295 | .byte 0x4 # DW_CFA_advance_loc4
|
|---|
| 296 | .long .LCFI11-.LCFI10
|
|---|
| 297 | .byte 0xd # DW_CFA_def_cfa_register
|
|---|
| 298 | .uleb128 0x6
|
|---|
| 299 | .align 8
|
|---|
| 300 | .LEFDE3:
|
|---|
| 301 |
|
|---|
| 302 | #endif /* __x86_64__ */
|
|---|
