Restoring authorship annotation for Mikhail Borisov <borisov.mikhail@gmail.com>. Commit 1 of 2.

1 files changed, 442 insertions, 442 deletions

diff --git a/contrib/restricted/libffi/src/x86/unix64.S b/contrib/restricted/libffi/src/x86/unix64.S
index 41563f5c60..ae81f77d09 100644
--- a/contrib/restricted/libffi/src/x86/unix64.S
+++ b/contrib/restricted/libffi/src/x86/unix64.S
@@ -30,21 +30,21 @@
 #define LIBFFI_ASM	
 #include <fficonfig.h>
 #include <ffi.h>
-#include "internal64.h"
-#include "asmnames.h"
-
-	.text
-
-/* This macro allows the safe creation of jump tables without an
-   actual table.  The entry points into the table are all 8 bytes.
-   The use of ORG asserts that we're at the correct location.  */
-/* ??? The clang assembler doesn't handle .org with symbolic expressions.  */
-#if defined(__clang__) || defined(__APPLE__) || (defined (__sun__) && defined(__svr4__))
-# define E(BASE, X)	.balign 8
-#else
-# define E(BASE, X)	.balign 8; .org BASE + X * 8
-#endif
-
+#include "internal64.h" 
+#include "asmnames.h" 
+
+	.text 
+
+/* This macro allows the safe creation of jump tables without an 
+   actual table.  The entry points into the table are all 8 bytes. 
+   The use of ORG asserts that we're at the correct location.  */ 
+/* ??? The clang assembler doesn't handle .org with symbolic expressions.  */ 
+#if defined(__clang__) || defined(__APPLE__) || (defined (__sun__) && defined(__svr4__)) 
+# define E(BASE, X)	.balign 8 
+#else 
+# define E(BASE, X)	.balign 8; .org BASE + X * 8 
+#endif 
+ 
 /* ffi_call_unix64 (void *args, unsigned long bytes, unsigned flags,
 	            void *raddr, void (*fnaddr)(void));
 
@@ -52,12 +52,12 @@
    for this function.  This has been allocated by ffi_call.  We also
    deallocate some of the stack that has been alloca'd.  */
 
-	.balign	8
-	.globl	C(ffi_call_unix64)
-	FFI_HIDDEN(C(ffi_call_unix64))
+	.balign	8 
+	.globl	C(ffi_call_unix64) 
+	FFI_HIDDEN(C(ffi_call_unix64)) 
 
-C(ffi_call_unix64):
-L(UW0):
+C(ffi_call_unix64): 
+L(UW0): 
 	movq	(%rsp), %r10		/* Load return address.  */
 	leaq	(%rdi, %rsi), %rax	/* Find local stack base.  */
 	movq	%rdx, (%rax)		/* Save flags.  */
@@ -65,37 +65,37 @@ L(UW0):
 	movq	%rbp, 16(%rax)		/* Save old frame pointer.  */
 	movq	%r10, 24(%rax)		/* Relocate return address.  */
 	movq	%rax, %rbp		/* Finalize local stack frame.  */
-
-	/* New stack frame based off rbp.  This is a itty bit of unwind
-	   trickery in that the CFA *has* changed.  There is no easy way
-	   to describe it correctly on entry to the function.  Fortunately,
-	   it doesn't matter too much since at all points we can correctly
-	   unwind back to ffi_call.  Note that the location to which we
-	   moved the return address is (the new) CFA-8, so from the
-	   perspective of the unwind info, it hasn't moved.  */
-L(UW1):
-	/* cfi_def_cfa(%rbp, 32) */
-	/* cfi_rel_offset(%rbp, 16) */
-
+ 
+	/* New stack frame based off rbp.  This is a itty bit of unwind 
+	   trickery in that the CFA *has* changed.  There is no easy way 
+	   to describe it correctly on entry to the function.  Fortunately, 
+	   it doesn't matter too much since at all points we can correctly 
+	   unwind back to ffi_call.  Note that the location to which we 
+	   moved the return address is (the new) CFA-8, so from the 
+	   perspective of the unwind info, it hasn't moved.  */ 
+L(UW1): 
+	/* cfi_def_cfa(%rbp, 32) */ 
+	/* cfi_rel_offset(%rbp, 16) */ 
+ 
 	movq	%rdi, %r10		/* Save a copy of the register area. */
 	movq	%r8, %r11		/* Save a copy of the target fn.  */
 	movl	%r9d, %eax		/* Set number of SSE registers.  */
 
 	/* Load up all argument registers.  */
 	movq	(%r10), %rdi
-	movq	0x08(%r10), %rsi
-	movq	0x10(%r10), %rdx
-	movq	0x18(%r10), %rcx
-	movq	0x20(%r10), %r8
-	movq	0x28(%r10), %r9
-	movl	0xb0(%r10), %eax
+	movq	0x08(%r10), %rsi 
+	movq	0x10(%r10), %rdx 
+	movq	0x18(%r10), %rcx 
+	movq	0x20(%r10), %r8 
+	movq	0x28(%r10), %r9 
+	movl	0xb0(%r10), %eax 
 	testl	%eax, %eax
-	jnz	L(load_sse)
-L(ret_from_load_sse):
+	jnz	L(load_sse) 
+L(ret_from_load_sse): 
 
-	/* Deallocate the reg arg area, except for r10, then load via pop.  */
-	leaq	0xb8(%r10), %rsp
-	popq	%r10
+	/* Deallocate the reg arg area, except for r10, then load via pop.  */ 
+	leaq	0xb8(%r10), %rsp 
+	popq	%r10 
 
 	/* Call the user function.  */
 	call	*%r11
@@ -106,460 +106,460 @@ L(ret_from_load_sse):
 	movq	0(%rbp), %rcx		/* Reload flags.  */
 	movq	8(%rbp), %rdi		/* Reload raddr.  */
 	movq	16(%rbp), %rbp		/* Reload old frame pointer.  */
-L(UW2):
-	/* cfi_remember_state */
-	/* cfi_def_cfa(%rsp, 8) */
-	/* cfi_restore(%rbp) */
+L(UW2): 
+	/* cfi_remember_state */ 
+	/* cfi_def_cfa(%rsp, 8) */ 
+	/* cfi_restore(%rbp) */ 
 
 	/* The first byte of the flags contains the FFI_TYPE.  */
-	cmpb	$UNIX64_RET_LAST, %cl
+	cmpb	$UNIX64_RET_LAST, %cl 
 	movzbl	%cl, %r10d
-	leaq	L(store_table)(%rip), %r11
-	ja	L(sa)
-	leaq	(%r11, %r10, 8), %r10
-
-	/* Prep for the structure cases: scratch area in redzone.  */
-	leaq	-20(%rsp), %rsi
+	leaq	L(store_table)(%rip), %r11 
+	ja	L(sa) 
+	leaq	(%r11, %r10, 8), %r10 
+ 
+	/* Prep for the structure cases: scratch area in redzone.  */ 
+	leaq	-20(%rsp), %rsi 
 	jmp	*%r10
 
-	.balign	8
-L(store_table):
-E(L(store_table), UNIX64_RET_VOID)
+	.balign	8 
+L(store_table): 
+E(L(store_table), UNIX64_RET_VOID) 
 	ret
-E(L(store_table), UNIX64_RET_UINT8)
-	movzbl	%al, %eax
+E(L(store_table), UNIX64_RET_UINT8) 
+	movzbl	%al, %eax 
 	movq	%rax, (%rdi)
 	ret
-E(L(store_table), UNIX64_RET_UINT16)
-	movzwl	%ax, %eax
-	movq	%rax, (%rdi)
-	ret
-E(L(store_table), UNIX64_RET_UINT32)
-	movl	%eax, %eax
-	movq	%rax, (%rdi)
-	ret
-E(L(store_table), UNIX64_RET_SINT8)
+E(L(store_table), UNIX64_RET_UINT16) 
+	movzwl	%ax, %eax 
+	movq	%rax, (%rdi) 
+	ret 
+E(L(store_table), UNIX64_RET_UINT32) 
+	movl	%eax, %eax 
+	movq	%rax, (%rdi) 
+	ret 
+E(L(store_table), UNIX64_RET_SINT8) 
 	movsbq	%al, %rax
 	movq	%rax, (%rdi)
 	ret
-E(L(store_table), UNIX64_RET_SINT16)
+E(L(store_table), UNIX64_RET_SINT16) 
 	movswq	%ax, %rax
 	movq	%rax, (%rdi)
 	ret
-E(L(store_table), UNIX64_RET_SINT32)
+E(L(store_table), UNIX64_RET_SINT32) 
 	cltq
 	movq	%rax, (%rdi)
 	ret
-E(L(store_table), UNIX64_RET_INT64)
+E(L(store_table), UNIX64_RET_INT64) 
 	movq	%rax, (%rdi)
 	ret
-E(L(store_table), UNIX64_RET_XMM32)
-	movd	%xmm0, (%rdi)
-	ret
-E(L(store_table), UNIX64_RET_XMM64)
-	movq	%xmm0, (%rdi)
+E(L(store_table), UNIX64_RET_XMM32) 
+	movd	%xmm0, (%rdi) 
 	ret
-E(L(store_table), UNIX64_RET_X87)
-	fstpt	(%rdi)
+E(L(store_table), UNIX64_RET_XMM64) 
+	movq	%xmm0, (%rdi) 
 	ret
-E(L(store_table), UNIX64_RET_X87_2)
+E(L(store_table), UNIX64_RET_X87) 
 	fstpt	(%rdi)
-	fstpt	16(%rdi)
 	ret
-E(L(store_table), UNIX64_RET_ST_XMM0_RAX)
-	movq	%rax, 8(%rsi)
-	jmp	L(s3)
-E(L(store_table), UNIX64_RET_ST_RAX_XMM0)
-	movq	%xmm0, 8(%rsi)
-	jmp	L(s2)
-E(L(store_table), UNIX64_RET_ST_XMM0_XMM1)
-	movq	%xmm1, 8(%rsi)
-	jmp	L(s3)
-E(L(store_table), UNIX64_RET_ST_RAX_RDX)
-	movq	%rdx, 8(%rsi)
-L(s2):
+E(L(store_table), UNIX64_RET_X87_2) 
+	fstpt	(%rdi) 
+	fstpt	16(%rdi) 
+	ret 
+E(L(store_table), UNIX64_RET_ST_XMM0_RAX) 
+	movq	%rax, 8(%rsi) 
+	jmp	L(s3) 
+E(L(store_table), UNIX64_RET_ST_RAX_XMM0) 
+	movq	%xmm0, 8(%rsi) 
+	jmp	L(s2) 
+E(L(store_table), UNIX64_RET_ST_XMM0_XMM1) 
+	movq	%xmm1, 8(%rsi) 
+	jmp	L(s3) 
+E(L(store_table), UNIX64_RET_ST_RAX_RDX) 
+	movq	%rdx, 8(%rsi) 
+L(s2): 
 	movq	%rax, (%rsi)
-	shrl	$UNIX64_SIZE_SHIFT, %ecx
-	rep movsb
-	ret
-	.balign 8
-L(s3):
-	movq	%xmm0, (%rsi)
-	shrl	$UNIX64_SIZE_SHIFT, %ecx
+	shrl	$UNIX64_SIZE_SHIFT, %ecx 
 	rep movsb
 	ret
-
-L(sa):	call	PLT(C(abort))
-
+	.balign 8 
+L(s3): 
+	movq	%xmm0, (%rsi) 
+	shrl	$UNIX64_SIZE_SHIFT, %ecx 
+	rep movsb 
+	ret 
+
+L(sa):	call	PLT(C(abort)) 
+ 
 	/* Many times we can avoid loading any SSE registers at all.
 	   It's not worth an indirect jump to load the exact set of
 	   SSE registers needed; zero or all is a good compromise.  */
-	.balign 2
-L(UW3):
-	/* cfi_restore_state */
-L(load_sse):
-	movdqa	0x30(%r10), %xmm0
-	movdqa	0x40(%r10), %xmm1
-	movdqa	0x50(%r10), %xmm2
-	movdqa	0x60(%r10), %xmm3
-	movdqa	0x70(%r10), %xmm4
-	movdqa	0x80(%r10), %xmm5
-	movdqa	0x90(%r10), %xmm6
-	movdqa	0xa0(%r10), %xmm7
-	jmp	L(ret_from_load_sse)
-
-L(UW4):
-ENDF(C(ffi_call_unix64))
-
-/* 6 general registers, 8 vector registers,
-   32 bytes of rvalue, 8 bytes of alignment.  */
-#define ffi_closure_OFS_G	0
-#define ffi_closure_OFS_V	(6*8)
-#define ffi_closure_OFS_RVALUE	(ffi_closure_OFS_V + 8*16)
-#define ffi_closure_FS		(ffi_closure_OFS_RVALUE + 32 + 8)
-
-/* The location of rvalue within the red zone after deallocating the frame.  */
-#define ffi_closure_RED_RVALUE	(ffi_closure_OFS_RVALUE - ffi_closure_FS)
-
-	.balign	2
-	.globl	C(ffi_closure_unix64_sse)
-	FFI_HIDDEN(C(ffi_closure_unix64_sse))
-
-C(ffi_closure_unix64_sse):
-L(UW5):
-	subq	$ffi_closure_FS, %rsp
-L(UW6):
-	/* cfi_adjust_cfa_offset(ffi_closure_FS) */
-
-	movdqa	%xmm0, ffi_closure_OFS_V+0x00(%rsp)
-	movdqa	%xmm1, ffi_closure_OFS_V+0x10(%rsp)
-	movdqa	%xmm2, ffi_closure_OFS_V+0x20(%rsp)
-	movdqa	%xmm3, ffi_closure_OFS_V+0x30(%rsp)
-	movdqa	%xmm4, ffi_closure_OFS_V+0x40(%rsp)
-	movdqa	%xmm5, ffi_closure_OFS_V+0x50(%rsp)
-	movdqa	%xmm6, ffi_closure_OFS_V+0x60(%rsp)
-	movdqa	%xmm7, ffi_closure_OFS_V+0x70(%rsp)
-	jmp	L(sse_entry1)
-
-L(UW7):
-ENDF(C(ffi_closure_unix64_sse))
-
-	.balign	2
-	.globl	C(ffi_closure_unix64)
-	FFI_HIDDEN(C(ffi_closure_unix64))
-
-C(ffi_closure_unix64):
-L(UW8):
-	subq	$ffi_closure_FS, %rsp
-L(UW9):
-	/* cfi_adjust_cfa_offset(ffi_closure_FS) */
-L(sse_entry1):
-	movq	%rdi, ffi_closure_OFS_G+0x00(%rsp)
-	movq    %rsi, ffi_closure_OFS_G+0x08(%rsp)
-	movq    %rdx, ffi_closure_OFS_G+0x10(%rsp)
-	movq    %rcx, ffi_closure_OFS_G+0x18(%rsp)
-	movq    %r8,  ffi_closure_OFS_G+0x20(%rsp)
-	movq    %r9,  ffi_closure_OFS_G+0x28(%rsp)
-
-#ifdef __ILP32__
-	movl	FFI_TRAMPOLINE_SIZE(%r10), %edi		/* Load cif */
-	movl	FFI_TRAMPOLINE_SIZE+4(%r10), %esi	/* Load fun */
-	movl	FFI_TRAMPOLINE_SIZE+8(%r10), %edx	/* Load user_data */
-#else
-	movq	FFI_TRAMPOLINE_SIZE(%r10), %rdi		/* Load cif */
-	movq	FFI_TRAMPOLINE_SIZE+8(%r10), %rsi	/* Load fun */
-	movq	FFI_TRAMPOLINE_SIZE+16(%r10), %rdx	/* Load user_data */
-#endif
-L(do_closure):
-	leaq	ffi_closure_OFS_RVALUE(%rsp), %rcx	/* Load rvalue */
-	movq	%rsp, %r8				/* Load reg_args */
-	leaq	ffi_closure_FS+8(%rsp), %r9		/* Load argp */
-	call	PLT(C(ffi_closure_unix64_inner))
-
+	.balign 2 
+L(UW3): 
+	/* cfi_restore_state */ 
+L(load_sse): 
+	movdqa	0x30(%r10), %xmm0 
+	movdqa	0x40(%r10), %xmm1 
+	movdqa	0x50(%r10), %xmm2 
+	movdqa	0x60(%r10), %xmm3 
+	movdqa	0x70(%r10), %xmm4 
+	movdqa	0x80(%r10), %xmm5 
+	movdqa	0x90(%r10), %xmm6 
+	movdqa	0xa0(%r10), %xmm7 
+	jmp	L(ret_from_load_sse) 
+
+L(UW4): 
+ENDF(C(ffi_call_unix64)) 
+
+/* 6 general registers, 8 vector registers, 
+   32 bytes of rvalue, 8 bytes of alignment.  */ 
+#define ffi_closure_OFS_G	0 
+#define ffi_closure_OFS_V	(6*8) 
+#define ffi_closure_OFS_RVALUE	(ffi_closure_OFS_V + 8*16) 
+#define ffi_closure_FS		(ffi_closure_OFS_RVALUE + 32 + 8) 
+
+/* The location of rvalue within the red zone after deallocating the frame.  */ 
+#define ffi_closure_RED_RVALUE	(ffi_closure_OFS_RVALUE - ffi_closure_FS) 
+
+	.balign	2 
+	.globl	C(ffi_closure_unix64_sse) 
+	FFI_HIDDEN(C(ffi_closure_unix64_sse)) 
+
+C(ffi_closure_unix64_sse): 
+L(UW5): 
+	subq	$ffi_closure_FS, %rsp 
+L(UW6): 
+	/* cfi_adjust_cfa_offset(ffi_closure_FS) */ 
+ 
+	movdqa	%xmm0, ffi_closure_OFS_V+0x00(%rsp) 
+	movdqa	%xmm1, ffi_closure_OFS_V+0x10(%rsp) 
+	movdqa	%xmm2, ffi_closure_OFS_V+0x20(%rsp) 
+	movdqa	%xmm3, ffi_closure_OFS_V+0x30(%rsp) 
+	movdqa	%xmm4, ffi_closure_OFS_V+0x40(%rsp) 
+	movdqa	%xmm5, ffi_closure_OFS_V+0x50(%rsp) 
+	movdqa	%xmm6, ffi_closure_OFS_V+0x60(%rsp) 
+	movdqa	%xmm7, ffi_closure_OFS_V+0x70(%rsp) 
+	jmp	L(sse_entry1) 
+ 
+L(UW7): 
+ENDF(C(ffi_closure_unix64_sse)) 
+ 
+	.balign	2 
+	.globl	C(ffi_closure_unix64) 
+	FFI_HIDDEN(C(ffi_closure_unix64)) 
+ 
+C(ffi_closure_unix64): 
+L(UW8): 
+	subq	$ffi_closure_FS, %rsp 
+L(UW9): 
+	/* cfi_adjust_cfa_offset(ffi_closure_FS) */ 
+L(sse_entry1): 
+	movq	%rdi, ffi_closure_OFS_G+0x00(%rsp) 
+	movq    %rsi, ffi_closure_OFS_G+0x08(%rsp) 
+	movq    %rdx, ffi_closure_OFS_G+0x10(%rsp) 
+	movq    %rcx, ffi_closure_OFS_G+0x18(%rsp) 
+	movq    %r8,  ffi_closure_OFS_G+0x20(%rsp) 
+	movq    %r9,  ffi_closure_OFS_G+0x28(%rsp) 
+ 
+#ifdef __ILP32__ 
+	movl	FFI_TRAMPOLINE_SIZE(%r10), %edi		/* Load cif */ 
+	movl	FFI_TRAMPOLINE_SIZE+4(%r10), %esi	/* Load fun */ 
+	movl	FFI_TRAMPOLINE_SIZE+8(%r10), %edx	/* Load user_data */ 
+#else 
+	movq	FFI_TRAMPOLINE_SIZE(%r10), %rdi		/* Load cif */ 
+	movq	FFI_TRAMPOLINE_SIZE+8(%r10), %rsi	/* Load fun */ 
+	movq	FFI_TRAMPOLINE_SIZE+16(%r10), %rdx	/* Load user_data */ 
+#endif 
+L(do_closure): 
+	leaq	ffi_closure_OFS_RVALUE(%rsp), %rcx	/* Load rvalue */ 
+	movq	%rsp, %r8				/* Load reg_args */ 
+	leaq	ffi_closure_FS+8(%rsp), %r9		/* Load argp */ 
+	call	PLT(C(ffi_closure_unix64_inner)) 
+ 
 	/* Deallocate stack frame early; return value is now in redzone.  */
-	addq	$ffi_closure_FS, %rsp
-L(UW10):
-	/* cfi_adjust_cfa_offset(-ffi_closure_FS) */
+	addq	$ffi_closure_FS, %rsp 
+L(UW10): 
+	/* cfi_adjust_cfa_offset(-ffi_closure_FS) */ 
 
 	/* The first byte of the return value contains the FFI_TYPE.  */
-	cmpb	$UNIX64_RET_LAST, %al
+	cmpb	$UNIX64_RET_LAST, %al 
 	movzbl	%al, %r10d
-	leaq	L(load_table)(%rip), %r11
-	ja	L(la)
-	leaq	(%r11, %r10, 8), %r10
-	leaq	ffi_closure_RED_RVALUE(%rsp), %rsi
+	leaq	L(load_table)(%rip), %r11 
+	ja	L(la) 
+	leaq	(%r11, %r10, 8), %r10 
+	leaq	ffi_closure_RED_RVALUE(%rsp), %rsi 
 	jmp	*%r10
 
-	.balign	8
-L(load_table):
-E(L(load_table), UNIX64_RET_VOID)
+	.balign	8 
+L(load_table): 
+E(L(load_table), UNIX64_RET_VOID) 
 	ret
-E(L(load_table), UNIX64_RET_UINT8)
-	movzbl	(%rsi), %eax
+E(L(load_table), UNIX64_RET_UINT8) 
+	movzbl	(%rsi), %eax 
 	ret
-E(L(load_table), UNIX64_RET_UINT16)
-	movzwl	(%rsi), %eax
+E(L(load_table), UNIX64_RET_UINT16) 
+	movzwl	(%rsi), %eax 
 	ret
-E(L(load_table), UNIX64_RET_UINT32)
-	movl	(%rsi), %eax
+E(L(load_table), UNIX64_RET_UINT32) 
+	movl	(%rsi), %eax 
 	ret
-E(L(load_table), UNIX64_RET_SINT8)
-	movsbl	(%rsi), %eax
+E(L(load_table), UNIX64_RET_SINT8) 
+	movsbl	(%rsi), %eax 
 	ret
-E(L(load_table), UNIX64_RET_SINT16)
-	movswl	(%rsi), %eax
+E(L(load_table), UNIX64_RET_SINT16) 
+	movswl	(%rsi), %eax 
 	ret
-E(L(load_table), UNIX64_RET_SINT32)
-	movl	(%rsi), %eax
+E(L(load_table), UNIX64_RET_SINT32) 
+	movl	(%rsi), %eax 
 	ret
-E(L(load_table), UNIX64_RET_INT64)
-	movq	(%rsi), %rax
+E(L(load_table), UNIX64_RET_INT64) 
+	movq	(%rsi), %rax 
 	ret
-E(L(load_table), UNIX64_RET_XMM32)
-	movd	(%rsi), %xmm0
+E(L(load_table), UNIX64_RET_XMM32) 
+	movd	(%rsi), %xmm0 
 	ret
-E(L(load_table), UNIX64_RET_XMM64)
-	movq	(%rsi), %xmm0
-	ret
-E(L(load_table), UNIX64_RET_X87)
-	fldt	(%rsi)
-	ret
-E(L(load_table), UNIX64_RET_X87_2)
-	fldt	16(%rsi)
-	fldt	(%rsi)
-	ret
-E(L(load_table), UNIX64_RET_ST_XMM0_RAX)
-	movq	8(%rsi), %rax
-	jmp	L(l3)
-E(L(load_table), UNIX64_RET_ST_RAX_XMM0)
-	movq	8(%rsi), %xmm0
-	jmp	L(l2)
-E(L(load_table), UNIX64_RET_ST_XMM0_XMM1)
-	movq	8(%rsi), %xmm1
-	jmp	L(l3)
-E(L(load_table), UNIX64_RET_ST_RAX_RDX)
-	movq	8(%rsi), %rdx
-L(l2):
-	movq	(%rsi), %rax
-	ret
-	.balign	8
-L(l3):
-	movq	(%rsi), %xmm0
-	ret
-
-L(la):	call	PLT(C(abort))
-
-L(UW11):
-ENDF(C(ffi_closure_unix64))
-
-	.balign	2
-	.globl	C(ffi_go_closure_unix64_sse)
-	FFI_HIDDEN(C(ffi_go_closure_unix64_sse))
-
-C(ffi_go_closure_unix64_sse):
-L(UW12):
-	subq	$ffi_closure_FS, %rsp
-L(UW13):
-	/* cfi_adjust_cfa_offset(ffi_closure_FS) */
-
-	movdqa	%xmm0, ffi_closure_OFS_V+0x00(%rsp)
-	movdqa	%xmm1, ffi_closure_OFS_V+0x10(%rsp)
-	movdqa	%xmm2, ffi_closure_OFS_V+0x20(%rsp)
-	movdqa	%xmm3, ffi_closure_OFS_V+0x30(%rsp)
-	movdqa	%xmm4, ffi_closure_OFS_V+0x40(%rsp)
-	movdqa	%xmm5, ffi_closure_OFS_V+0x50(%rsp)
-	movdqa	%xmm6, ffi_closure_OFS_V+0x60(%rsp)
-	movdqa	%xmm7, ffi_closure_OFS_V+0x70(%rsp)
-	jmp	L(sse_entry2)
-
-L(UW14):
-ENDF(C(ffi_go_closure_unix64_sse))
-
-	.balign	2
-	.globl	C(ffi_go_closure_unix64)
-	FFI_HIDDEN(C(ffi_go_closure_unix64))
-
-C(ffi_go_closure_unix64):
-L(UW15):
-	subq	$ffi_closure_FS, %rsp
-L(UW16):
-	/* cfi_adjust_cfa_offset(ffi_closure_FS) */
-L(sse_entry2):
-	movq	%rdi, ffi_closure_OFS_G+0x00(%rsp)
-	movq    %rsi, ffi_closure_OFS_G+0x08(%rsp)
-	movq    %rdx, ffi_closure_OFS_G+0x10(%rsp)
-	movq    %rcx, ffi_closure_OFS_G+0x18(%rsp)
-	movq    %r8,  ffi_closure_OFS_G+0x20(%rsp)
-	movq    %r9,  ffi_closure_OFS_G+0x28(%rsp)
-
-#ifdef __ILP32__
-	movl	4(%r10), %edi		/* Load cif */
-	movl	8(%r10), %esi		/* Load fun */
-	movl	%r10d, %edx		/* Load closure (user_data) */
-#else
-	movq	8(%r10), %rdi		/* Load cif */
-	movq	16(%r10), %rsi		/* Load fun */
-	movq	%r10, %rdx		/* Load closure (user_data) */
-#endif
-	jmp	L(do_closure)
-
-L(UW17):
-ENDF(C(ffi_go_closure_unix64))
-
-/* Sadly, OSX cctools-as doesn't understand .cfi directives at all.  */
-
-#ifdef __APPLE__
-.section __TEXT,__eh_frame,coalesced,no_toc+strip_static_syms+live_support
-EHFrame0:
-#elif defined(HAVE_AS_X86_64_UNWIND_SECTION_TYPE)
-.section .eh_frame,"a",@unwind
+E(L(load_table), UNIX64_RET_XMM64) 
+	movq	(%rsi), %xmm0 
+	ret 
+E(L(load_table), UNIX64_RET_X87) 
+	fldt	(%rsi) 
+	ret 
+E(L(load_table), UNIX64_RET_X87_2) 
+	fldt	16(%rsi) 
+	fldt	(%rsi) 
+	ret 
+E(L(load_table), UNIX64_RET_ST_XMM0_RAX) 
+	movq	8(%rsi), %rax 
+	jmp	L(l3) 
+E(L(load_table), UNIX64_RET_ST_RAX_XMM0) 
+	movq	8(%rsi), %xmm0 
+	jmp	L(l2) 
+E(L(load_table), UNIX64_RET_ST_XMM0_XMM1) 
+	movq	8(%rsi), %xmm1 
+	jmp	L(l3) 
+E(L(load_table), UNIX64_RET_ST_RAX_RDX) 
+	movq	8(%rsi), %rdx 
+L(l2): 
+	movq	(%rsi), %rax 
+	ret 
+	.balign	8 
+L(l3): 
+	movq	(%rsi), %xmm0 
+	ret 
+
+L(la):	call	PLT(C(abort)) 
+
+L(UW11): 
+ENDF(C(ffi_closure_unix64)) 
+
+	.balign	2 
+	.globl	C(ffi_go_closure_unix64_sse) 
+	FFI_HIDDEN(C(ffi_go_closure_unix64_sse)) 
+
+C(ffi_go_closure_unix64_sse): 
+L(UW12): 
+	subq	$ffi_closure_FS, %rsp 
+L(UW13): 
+	/* cfi_adjust_cfa_offset(ffi_closure_FS) */ 
+ 
+	movdqa	%xmm0, ffi_closure_OFS_V+0x00(%rsp) 
+	movdqa	%xmm1, ffi_closure_OFS_V+0x10(%rsp) 
+	movdqa	%xmm2, ffi_closure_OFS_V+0x20(%rsp) 
+	movdqa	%xmm3, ffi_closure_OFS_V+0x30(%rsp) 
+	movdqa	%xmm4, ffi_closure_OFS_V+0x40(%rsp) 
+	movdqa	%xmm5, ffi_closure_OFS_V+0x50(%rsp) 
+	movdqa	%xmm6, ffi_closure_OFS_V+0x60(%rsp) 
+	movdqa	%xmm7, ffi_closure_OFS_V+0x70(%rsp) 
+	jmp	L(sse_entry2) 
+ 
+L(UW14): 
+ENDF(C(ffi_go_closure_unix64_sse)) 
+ 
+	.balign	2 
+	.globl	C(ffi_go_closure_unix64) 
+	FFI_HIDDEN(C(ffi_go_closure_unix64)) 
+ 
+C(ffi_go_closure_unix64): 
+L(UW15): 
+	subq	$ffi_closure_FS, %rsp 
+L(UW16): 
+	/* cfi_adjust_cfa_offset(ffi_closure_FS) */ 
+L(sse_entry2): 
+	movq	%rdi, ffi_closure_OFS_G+0x00(%rsp) 
+	movq    %rsi, ffi_closure_OFS_G+0x08(%rsp) 
+	movq    %rdx, ffi_closure_OFS_G+0x10(%rsp) 
+	movq    %rcx, ffi_closure_OFS_G+0x18(%rsp) 
+	movq    %r8,  ffi_closure_OFS_G+0x20(%rsp) 
+	movq    %r9,  ffi_closure_OFS_G+0x28(%rsp) 
+ 
+#ifdef __ILP32__ 
+	movl	4(%r10), %edi		/* Load cif */ 
+	movl	8(%r10), %esi		/* Load fun */ 
+	movl	%r10d, %edx		/* Load closure (user_data) */ 
 #else
-.section .eh_frame,"a",@progbits
+	movq	8(%r10), %rdi		/* Load cif */ 
+	movq	16(%r10), %rsi		/* Load fun */ 
+	movq	%r10, %rdx		/* Load closure (user_data) */ 
 #endif
-
-#ifdef HAVE_AS_X86_PCREL
-# define PCREL(X)	X - .
-#else
-# define PCREL(X)	X@rel
-#endif
-
-/* Simplify advancing between labels.  Assume DW_CFA_advance_loc1 fits.  */
-#define ADV(N, P)	.byte 2, L(N)-L(P)
-
-	.balign 8
-L(CIE):
-	.set	L(set0),L(ECIE)-L(SCIE)
-	.long	L(set0)			/* CIE Length */
-L(SCIE):
+	jmp	L(do_closure) 
+ 
+L(UW17): 
+ENDF(C(ffi_go_closure_unix64)) 
+ 
+/* Sadly, OSX cctools-as doesn't understand .cfi directives at all.  */ 
+ 
+#ifdef __APPLE__ 
+.section __TEXT,__eh_frame,coalesced,no_toc+strip_static_syms+live_support 
+EHFrame0: 
+#elif defined(HAVE_AS_X86_64_UNWIND_SECTION_TYPE) 
+.section .eh_frame,"a",@unwind 
+#else 
+.section .eh_frame,"a",@progbits 
+#endif 
+ 
+#ifdef HAVE_AS_X86_PCREL 
+# define PCREL(X)	X - . 
+#else 
+# define PCREL(X)	X@rel 
+#endif 
+ 
+/* Simplify advancing between labels.  Assume DW_CFA_advance_loc1 fits.  */ 
+#define ADV(N, P)	.byte 2, L(N)-L(P) 
+ 
+	.balign 8 
+L(CIE): 
+	.set	L(set0),L(ECIE)-L(SCIE) 
+	.long	L(set0)			/* CIE Length */ 
+L(SCIE): 
 	.long	0			/* CIE Identifier Tag */
 	.byte	1			/* CIE Version */
-	.ascii	"zR\0"			/* CIE Augmentation */
-	.byte	1			/* CIE Code Alignment Factor */
-	.byte	0x78			/* CIE Data Alignment Factor */
+	.ascii	"zR\0"			/* CIE Augmentation */ 
+	.byte	1			/* CIE Code Alignment Factor */ 
+	.byte	0x78			/* CIE Data Alignment Factor */ 
 	.byte	0x10			/* CIE RA Column */
-	.byte	1			/* Augmentation size */
+	.byte	1			/* Augmentation size */ 
 	.byte	0x1b			/* FDE Encoding (pcrel sdata4) */
-	.byte	0xc, 7, 8		/* DW_CFA_def_cfa, %rsp offset 8 */
-	.byte	0x80+16, 1		/* DW_CFA_offset, %rip offset 1*-8 */
-	.balign 8
-L(ECIE):
-
-	.set	L(set1),L(EFDE1)-L(SFDE1)
-	.long	L(set1)			/* FDE Length */
-L(SFDE1):
-	.long	L(SFDE1)-L(CIE)		/* FDE CIE offset */
-	.long	PCREL(L(UW0))		/* Initial location */
-	.long	L(UW4)-L(UW0)		/* Address range */
-	.byte	0			/* Augmentation size */
-	ADV(UW1, UW0)
-	.byte	0xc, 6, 32		/* DW_CFA_def_cfa, %rbp 32 */
-	.byte	0x80+6, 2		/* DW_CFA_offset, %rbp 2*-8 */
-	ADV(UW2, UW1)
+	.byte	0xc, 7, 8		/* DW_CFA_def_cfa, %rsp offset 8 */ 
+	.byte	0x80+16, 1		/* DW_CFA_offset, %rip offset 1*-8 */ 
+	.balign 8 
+L(ECIE): 
+
+	.set	L(set1),L(EFDE1)-L(SFDE1) 
+	.long	L(set1)			/* FDE Length */ 
+L(SFDE1): 
+	.long	L(SFDE1)-L(CIE)		/* FDE CIE offset */ 
+	.long	PCREL(L(UW0))		/* Initial location */ 
+	.long	L(UW4)-L(UW0)		/* Address range */ 
+	.byte	0			/* Augmentation size */ 
+	ADV(UW1, UW0) 
+	.byte	0xc, 6, 32		/* DW_CFA_def_cfa, %rbp 32 */ 
+	.byte	0x80+6, 2		/* DW_CFA_offset, %rbp 2*-8 */ 
+	ADV(UW2, UW1) 
 	.byte	0xa			/* DW_CFA_remember_state */
-	.byte	0xc, 7, 8		/* DW_CFA_def_cfa, %rsp 8 */
+	.byte	0xc, 7, 8		/* DW_CFA_def_cfa, %rsp 8 */ 
 	.byte	0xc0+6			/* DW_CFA_restore, %rbp */
-	ADV(UW3, UW2)
+	ADV(UW3, UW2) 
 	.byte	0xb			/* DW_CFA_restore_state */
-	.balign	8
-L(EFDE1):
-
-	.set	L(set2),L(EFDE2)-L(SFDE2)
-	.long	L(set2)			/* FDE Length */
-L(SFDE2):
-	.long	L(SFDE2)-L(CIE)		/* FDE CIE offset */
-	.long	PCREL(L(UW5))		/* Initial location */
-	.long	L(UW7)-L(UW5)		/* Address range */
-	.byte	0			/* Augmentation size */
-	ADV(UW6, UW5)
-	.byte	0xe			/* DW_CFA_def_cfa_offset */
-	.byte	ffi_closure_FS + 8, 1	/* uleb128, assuming 128 <= FS < 255 */
-	.balign	8
-L(EFDE2):
-
-	.set	L(set3),L(EFDE3)-L(SFDE3)
-	.long	L(set3)			/* FDE Length */
-L(SFDE3):
-	.long	L(SFDE3)-L(CIE)		/* FDE CIE offset */
-	.long	PCREL(L(UW8))		/* Initial location */
-	.long	L(UW11)-L(UW8)		/* Address range */
-	.byte	0			/* Augmentation size */
-	ADV(UW9, UW8)
-	.byte	0xe			/* DW_CFA_def_cfa_offset */
-	.byte	ffi_closure_FS + 8, 1	/* uleb128, assuming 128 <= FS < 255 */
-	ADV(UW10, UW9)
-	.byte	0xe, 8			/* DW_CFA_def_cfa_offset 8 */
-L(EFDE3):
-
-	.set	L(set4),L(EFDE4)-L(SFDE4)
-	.long	L(set4)			/* FDE Length */
-L(SFDE4):
-	.long	L(SFDE4)-L(CIE)		/* FDE CIE offset */
-	.long	PCREL(L(UW12))		/* Initial location */
-	.long	L(UW14)-L(UW12)		/* Address range */
-	.byte	0			/* Augmentation size */
-	ADV(UW13, UW12)
+	.balign	8 
+L(EFDE1): 
+
+	.set	L(set2),L(EFDE2)-L(SFDE2) 
+	.long	L(set2)			/* FDE Length */ 
+L(SFDE2): 
+	.long	L(SFDE2)-L(CIE)		/* FDE CIE offset */ 
+	.long	PCREL(L(UW5))		/* Initial location */ 
+	.long	L(UW7)-L(UW5)		/* Address range */ 
+	.byte	0			/* Augmentation size */ 
+	ADV(UW6, UW5) 
+	.byte	0xe			/* DW_CFA_def_cfa_offset */ 
+	.byte	ffi_closure_FS + 8, 1	/* uleb128, assuming 128 <= FS < 255 */ 
+	.balign	8 
+L(EFDE2): 
+
+	.set	L(set3),L(EFDE3)-L(SFDE3) 
+	.long	L(set3)			/* FDE Length */ 
+L(SFDE3): 
+	.long	L(SFDE3)-L(CIE)		/* FDE CIE offset */ 
+	.long	PCREL(L(UW8))		/* Initial location */ 
+	.long	L(UW11)-L(UW8)		/* Address range */ 
+	.byte	0			/* Augmentation size */ 
+	ADV(UW9, UW8) 
 	.byte	0xe			/* DW_CFA_def_cfa_offset */
-	.byte	ffi_closure_FS + 8, 1	/* uleb128, assuming 128 <= FS < 255 */
-	.balign	8
-L(EFDE4):
-
-	.set	L(set5),L(EFDE5)-L(SFDE5)
-	.long	L(set5)			/* FDE Length */
-L(SFDE5):
-	.long	L(SFDE5)-L(CIE)		/* FDE CIE offset */
-	.long	PCREL(L(UW15))		/* Initial location */
-	.long	L(UW17)-L(UW15)		/* Address range */
-	.byte	0			/* Augmentation size */
-	ADV(UW16, UW15)
+	.byte	ffi_closure_FS + 8, 1	/* uleb128, assuming 128 <= FS < 255 */ 
+	ADV(UW10, UW9) 
+	.byte	0xe, 8			/* DW_CFA_def_cfa_offset 8 */ 
+L(EFDE3): 
+
+	.set	L(set4),L(EFDE4)-L(SFDE4) 
+	.long	L(set4)			/* FDE Length */ 
+L(SFDE4): 
+	.long	L(SFDE4)-L(CIE)		/* FDE CIE offset */ 
+	.long	PCREL(L(UW12))		/* Initial location */ 
+	.long	L(UW14)-L(UW12)		/* Address range */ 
+	.byte	0			/* Augmentation size */ 
+	ADV(UW13, UW12) 
 	.byte	0xe			/* DW_CFA_def_cfa_offset */
-	.byte	ffi_closure_FS + 8, 1	/* uleb128, assuming 128 <= FS < 255 */
-	.balign	8
-L(EFDE5):
-#ifdef __APPLE__
-	.subsections_via_symbols
-	.section __LD,__compact_unwind,regular,debug
-
-	/* compact unwind for ffi_call_unix64 */
-	.quad    C(ffi_call_unix64)
-	.set     L1,L(UW4)-L(UW0)
-	.long    L1
-	.long    0x04000000 /* use dwarf unwind info */
-	.quad    0
-	.quad    0
-
-	/* compact unwind for ffi_closure_unix64_sse */
-	.quad    C(ffi_closure_unix64_sse)
-	.set     L2,L(UW7)-L(UW5)
-	.long    L2
-	.long    0x04000000 /* use dwarf unwind info */
-	.quad    0
-	.quad    0
-
-	/* compact unwind for ffi_closure_unix64 */
-	.quad    C(ffi_closure_unix64)
-	.set     L3,L(UW11)-L(UW8)
-	.long    L3
-	.long    0x04000000 /* use dwarf unwind info */
-	.quad    0
-	.quad    0
-
-	/* compact unwind for ffi_go_closure_unix64_sse */
-	.quad    C(ffi_go_closure_unix64_sse)
-	.set     L4,L(UW14)-L(UW12)
-	.long    L4
-	.long    0x04000000 /* use dwarf unwind info */
-	.quad    0
-	.quad    0
-
-	/* compact unwind for ffi_go_closure_unix64 */
-	.quad    C(ffi_go_closure_unix64)
-	.set     L5,L(UW17)-L(UW15)
-	.long    L5
-	.long    0x04000000 /* use dwarf unwind info */
-	.quad    0
-	.quad    0
-#endif
-
+	.byte	ffi_closure_FS + 8, 1	/* uleb128, assuming 128 <= FS < 255 */ 
+	.balign	8 
+L(EFDE4): 
+
+	.set	L(set5),L(EFDE5)-L(SFDE5) 
+	.long	L(set5)			/* FDE Length */ 
+L(SFDE5): 
+	.long	L(SFDE5)-L(CIE)		/* FDE CIE offset */ 
+	.long	PCREL(L(UW15))		/* Initial location */ 
+	.long	L(UW17)-L(UW15)		/* Address range */ 
+	.byte	0			/* Augmentation size */ 
+	ADV(UW16, UW15) 
+	.byte	0xe			/* DW_CFA_def_cfa_offset */ 
+	.byte	ffi_closure_FS + 8, 1	/* uleb128, assuming 128 <= FS < 255 */ 
+	.balign	8 
+L(EFDE5): 
+#ifdef __APPLE__ 
+	.subsections_via_symbols 
+	.section __LD,__compact_unwind,regular,debug 
+
+	/* compact unwind for ffi_call_unix64 */ 
+	.quad    C(ffi_call_unix64) 
+	.set     L1,L(UW4)-L(UW0) 
+	.long    L1 
+	.long    0x04000000 /* use dwarf unwind info */ 
+	.quad    0 
+	.quad    0 
+
+	/* compact unwind for ffi_closure_unix64_sse */ 
+	.quad    C(ffi_closure_unix64_sse) 
+	.set     L2,L(UW7)-L(UW5) 
+	.long    L2 
+	.long    0x04000000 /* use dwarf unwind info */ 
+	.quad    0 
+	.quad    0 
+ 
+	/* compact unwind for ffi_closure_unix64 */ 
+	.quad    C(ffi_closure_unix64) 
+	.set     L3,L(UW11)-L(UW8) 
+	.long    L3 
+	.long    0x04000000 /* use dwarf unwind info */ 
+	.quad    0 
+	.quad    0 
+ 
+	/* compact unwind for ffi_go_closure_unix64_sse */ 
+	.quad    C(ffi_go_closure_unix64_sse) 
+	.set     L4,L(UW14)-L(UW12) 
+	.long    L4 
+	.long    0x04000000 /* use dwarf unwind info */ 
+	.quad    0 
+	.quad    0 
+ 
+	/* compact unwind for ffi_go_closure_unix64 */ 
+	.quad    C(ffi_go_closure_unix64) 
+	.set     L5,L(UW17)-L(UW15) 
+	.long    L5 
+	.long    0x04000000 /* use dwarf unwind info */ 
+	.quad    0 
+	.quad    0 
+#endif 
+ 
 #endif /* __x86_64__ */
 #if defined __ELF__ && defined __linux__
 	.section	.note.GNU-stack,"",@progbits