/* ----------------------------------------------------------------------- unix64.S - Copyright (c) 2002 Bo Thorsen x86-64 Foreign Function Interface Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the ``Software''), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED ``AS IS'', WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL CYGNUS SOLUTIONS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. ----------------------------------------------------------------------- */ #ifdef __x86_64__ #define LIBFFI_ASM #include #include .section .rodata .LC0: .string "asm in progress %lld\n" .LC1: .string "asm in progress\n" .text .align 2 .globl ffi_call_UNIX64 .type ffi_call_UNIX64,@function ffi_call_UNIX64: .LFB1: pushq %rbp .LCFI0: movq %rsp, %rbp .LCFI1: /* Save all arguments */ subq $48, %rsp .LCFI2: movq %rdi, -8(%rbp) /* ffi_prep_args */ movq %rsi, -16(%rbp) /* ffi_fill_return_value */ movq %rdx, -24(%rbp) /* ecif */ movq %rcx, -32(%rbp) /* cif->bytes */ movq %r8, -40(%rbp) /* ecif.rvalue */ movq %r9, -48(%rbp) /* fn */ /* Make room for all of the new args and the register args */ addl $176, %ecx .LCFI3: subq %rcx, %rsp .LCFI4: /* Setup the call to ffi_prep_args. */ movq %rdi, %rax /* &ffi_prep_args */ movq %rsp, %rdi /* stackLayout */ movq %rdx, %rsi /* ecif */ call *%rax /* ffi_prep_args(stackLayout, ecif);*/ /* ffi_prep_args have put all the register contents into the */ /* stackLayout struct. Now put the register values in place. */ movq (%rsp), %rdi movq 8(%rsp), %rsi movq 16(%rsp), %rdx movq 24(%rsp), %rcx movq 32(%rsp), %r8 movq 40(%rsp), %r9 movaps 48(%rsp), %xmm0 movaps 64(%rsp), %xmm1 movaps 80(%rsp), %xmm2 movaps 96(%rsp), %xmm3 movaps 112(%rsp), %xmm4 movaps 128(%rsp), %xmm5 movaps 144(%rsp), %xmm6 movaps 160(%rsp), %xmm7 /* Remove space for stackLayout so stack arguments are placed correctly for the call. */ .LCFI5: addq $176, %rsp .LCFI6: /* Call the user function. */ call *-48(%rbp) /* Make stack space for the return_value struct. */ subq $64, %rsp /* Fill in all potential return values to this struct. */ movq %rax, (%rsp) movq %rdx, 8(%rsp) movaps %xmm0, 16(%rsp) movaps %xmm1, 32(%rsp) fstpt 48(%rsp) /* Now call ffi_fill_return_value. */ movq %rsp, %rdi /* struct return_value */ movq -24(%rbp), %rsi /* ecif */ movq -16(%rbp), %rax /* &ffi_fill_return_value */ call *%rax /* call it */ /* And the work is done. */ leave ret .LFE1: .ffi_call_UNIX64_end: .size ffi_call_UNIX64,.ffi_call_UNIX64_end-ffi_call_UNIX64 .text .align 2 .globl float2sse .type float2sse,@function float2sse: /* Save the contents of this sse-float in a pointer. */ movaps %xmm0, (%rdi) ret .align 2 .globl floatfloat2sse .type floatfloat2sse,@function floatfloat2sse: /* Save the contents of these two sse-floats in a pointer. */ movq (%rdi), %xmm0 movaps %xmm0, (%rsi) ret .align 2 .globl double2sse .type double2sse,@function double2sse: /* Save the contents of this sse-double in a pointer. */ movaps %xmm0, (%rdi) ret .align 2 .globl sse2float .type sse2float,@function sse2float: /* Save the contents of this sse-float in a pointer. */ movaps (%rdi), %xmm0 ret .align 2 .globl sse2double .type sse2double,@function sse2double: /* Save the contents of this pointer in a sse-double. */ movaps (%rdi), %xmm0 ret .align 2 .globl sse2floatfloat .type sse2floatfloat,@function sse2floatfloat: /* Save the contents of this pointer in two sse-floats. */ movaps (%rdi), %xmm0 movq %xmm0, (%rsi) ret .align 2 .globl ffi_closure_UNIX64 .type ffi_closure_UNIX64,@function ffi_closure_UNIX64: .LFB2: pushq %rbp .LCFI10: movq %rsp, %rbp .LCFI11: subq $240, %rsp .LCFI12: movq %rdi, -176(%rbp) movq %rsi, -168(%rbp) movq %rdx, -160(%rbp) movq %rcx, -152(%rbp) movq %r8, -144(%rbp) movq %r9, -136(%rbp) /* FIXME: We can avoid all this stashing of XMM registers by (in ffi_prep_closure) computing the number of floating-point args and moving it into %rax before calling this function. Once this is done, uncomment the next few lines and only the essential XMM registers will be written to memory. This is a significant saving. */ /* movzbl %al, %eax */ /* movq %rax, %rdx */ /* leaq 0(,%rdx,4), %rax */ /* leaq 2f(%rip), %rdx */ /* subq %rax, %rdx */ leaq -1(%rbp), %rax /* jmp *%rdx */ movaps %xmm7, -15(%rax) movaps %xmm6, -31(%rax) movaps %xmm5, -47(%rax) movaps %xmm4, -63(%rax) movaps %xmm3, -79(%rax) movaps %xmm2, -95(%rax) movaps %xmm1, -111(%rax) movaps %xmm0, -127(%rax) 2: movl %edi, -180(%rbp) movl $0, -224(%rbp) movl $48, -220(%rbp) leaq 16(%rbp), %rax movq %rax, -216(%rbp) leaq -176(%rbp), %rdx movq %rdx, -208(%rbp) leaq -224(%rbp), %rsi movq %r10, %rdi movq %rsp, %rdx call ffi_closure_UNIX64_inner@PLT cmpl $FFI_TYPE_FLOAT, %eax je 1f cmpl $FFI_TYPE_DOUBLE, %eax je 2f cmpl $FFI_TYPE_LONGDOUBLE, %eax je 3f cmpl $FFI_TYPE_STRUCT, %eax je 4f popq %rax leave ret 1: 2: 3: movaps -240(%rbp), %xmm0 leave ret 4: leave ret .LFE2: .section .eh_frame,EH_FRAME_FLAGS,@progbits .Lframe0: .long .LECIE1-.LSCIE1 .LSCIE1: .long 0x0 .byte 0x1 .string "zR" .uleb128 0x1 .sleb128 -8 .byte 0x10 .uleb128 0x1 .byte 0x1b .byte 0xc .uleb128 0x7 .uleb128 0x8 .byte 0x90 .uleb128 0x1 .align 8 .LECIE1: .LSFDE1: .long .LEFDE1-.LASFDE1 .LASFDE1: .long .LASFDE1-.Lframe0 .long .LFB1-. .long .LFE1-.LFB1 .uleb128 0x0 .byte 0x4 # DW_CFA_advance_loc4 .long .LCFI0-.LFB1 .byte 0xe # DW_CFA_def_cfa_offset .uleb128 0x10 .byte 0x86 # DW_CFA_offset: r6 at cfa-16 .uleb128 0x2 .byte 0x4 # DW_CFA_advance_loc4 .long .LCFI1-.LCFI0 .byte 0x86 # DW_CFA_offset: r6 at cfa-16 .uleb128 0x2 .byte 0xd # DW_CFA_def_cfa_reg: r6 .uleb128 0x6 .align 8 .LEFDE1: .LSFDE3: .long .LEFDE3-.LASFDE3 # FDE Length .LASFDE3: .long .LASFDE3-.Lframe0 # FDE CIE offset .long .LFB2-. # FDE initial location .long .LFE2-.LFB2 # FDE address range .uleb128 0x0 # Augmentation size .byte 0x4 # DW_CFA_advance_loc4 .long .LCFI10-.LFB2 .byte 0xe # DW_CFA_def_cfa_offset .uleb128 0x10 .byte 0x86 # DW_CFA_offset, column 0x6 .uleb128 0x2 .byte 0x4 # DW_CFA_advance_loc4 .long .LCFI11-.LCFI10 .byte 0xd # DW_CFA_def_cfa_register .uleb128 0x6 .align 8 .LEFDE3: #endif /* __x86_64__ */