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

5 files changed, 2114 insertions, 2114 deletions

diff --git a/contrib/restricted/libffi/src/aarch64/ffi.c b/contrib/restricted/libffi/src/aarch64/ffi.c
index 1ebf43c192..84d44ab74a 100644
--- a/contrib/restricted/libffi/src/aarch64/ffi.c
+++ b/contrib/restricted/libffi/src/aarch64/ffi.c
@@ -1,1009 +1,1009 @@
-/* Copyright (c) 2009, 2010, 2011, 2012 ARM Ltd.
-
-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 THE AUTHORS OR COPYRIGHT HOLDERS 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.  */
-
-#if defined(__aarch64__) || defined(__arm64__)|| defined (_M_ARM64)
-#include <stdio.h>
-#include <stdlib.h>
-#include <stdint.h>
-#include <fficonfig.h>
-#include <ffi.h>
-#include <ffi_common.h>
-#include "internal.h"
-#ifdef _M_ARM64
-#include <windows.h> /* FlushInstructionCache */
-#endif
-
-/* Force FFI_TYPE_LONGDOUBLE to be different than FFI_TYPE_DOUBLE;
-   all further uses in this file will refer to the 128-bit type.  */
-#if FFI_TYPE_DOUBLE != FFI_TYPE_LONGDOUBLE
-# if FFI_TYPE_LONGDOUBLE != 4
-#  error FFI_TYPE_LONGDOUBLE out of date
-# endif
-#else
-# undef FFI_TYPE_LONGDOUBLE
-# define FFI_TYPE_LONGDOUBLE 4
-#endif
-
-union _d
-{
-  UINT64 d;
-  UINT32 s[2];
-};
-
-struct _v
-{
-  union _d d[2] __attribute__((aligned(16)));
-};
-
-struct call_context
-{
-  struct _v v[N_V_ARG_REG];
-  UINT64 x[N_X_ARG_REG];
-};
-
-#if FFI_EXEC_TRAMPOLINE_TABLE
-
-#ifdef __MACH__
-#include <mach/vm_param.h>
-#endif
-
-#else
-
-#if defined (__clang__) && defined (__APPLE__)
-extern void sys_icache_invalidate (void *start, size_t len);
-#endif
-
-static inline void
-ffi_clear_cache (void *start, void *end)
-{
-#if defined (__clang__) && defined (__APPLE__)
-  sys_icache_invalidate (start, (char *)end - (char *)start);
-#elif defined (__GNUC__)
-  __builtin___clear_cache (start, end);
-#elif defined (_M_ARM64)
-  FlushInstructionCache(GetCurrentProcess(), start, (char*)end - (char*)start);
-#else
-#error "Missing builtin to flush instruction cache"
-#endif
-}
-
-#endif
-
-/* A subroutine of is_vfp_type.  Given a structure type, return the type code
-   of the first non-structure element.  Recurse for structure elements.
-   Return -1 if the structure is in fact empty, i.e. no nested elements.  */
-
-static int
-is_hfa0 (const ffi_type *ty)
-{
-  ffi_type **elements = ty->elements;
-  int i, ret = -1;
-
-  if (elements != NULL)
-    for (i = 0; elements[i]; ++i)
-      {
-        ret = elements[i]->type;
-        if (ret == FFI_TYPE_STRUCT || ret == FFI_TYPE_COMPLEX)
-          {
-            ret = is_hfa0 (elements[i]);
-            if (ret < 0)
-              continue;
-          }
-        break;
-      }
-
-  return ret;
-}
-
-/* A subroutine of is_vfp_type.  Given a structure type, return true if all
-   of the non-structure elements are the same as CANDIDATE.  */
-
-static int
-is_hfa1 (const ffi_type *ty, int candidate)
-{
-  ffi_type **elements = ty->elements;
-  int i;
-
-  if (elements != NULL)
-    for (i = 0; elements[i]; ++i)
-      {
-        int t = elements[i]->type;
-        if (t == FFI_TYPE_STRUCT || t == FFI_TYPE_COMPLEX)
-          {
-            if (!is_hfa1 (elements[i], candidate))
-              return 0;
-          }
-        else if (t != candidate)
-          return 0;
-      }
-
-  return 1;
-}
-
-/* Determine if TY may be allocated to the FP registers.  This is both an
-   fp scalar type as well as an homogenous floating point aggregate (HFA).
-   That is, a structure consisting of 1 to 4 members of all the same type,
-   where that type is an fp scalar.
-
-   Returns non-zero iff TY is an HFA.  The result is the AARCH64_RET_*
-   constant for the type.  */
-
-static int
-is_vfp_type (const ffi_type *ty)
-{
-  ffi_type **elements;
-  int candidate, i;
-  size_t size, ele_count;
-
-  /* Quickest tests first.  */
-  candidate = ty->type;
-  switch (candidate)
-    {
-    default:
-      return 0;
-    case FFI_TYPE_FLOAT:
-    case FFI_TYPE_DOUBLE:
-    case FFI_TYPE_LONGDOUBLE:
-      ele_count = 1;
-      goto done;
-    case FFI_TYPE_COMPLEX:
-      candidate = ty->elements[0]->type;
-      switch (candidate)
-	{
-	case FFI_TYPE_FLOAT:
-	case FFI_TYPE_DOUBLE:
-	case FFI_TYPE_LONGDOUBLE:
-	  ele_count = 2;
-	  goto done;
-	}
-      return 0;
-    case FFI_TYPE_STRUCT:
-      break;
-    }
-
-  /* No HFA types are smaller than 4 bytes, or larger than 64 bytes.  */
-  size = ty->size;
-  if (size < 4 || size > 64)
-    return 0;
-
-  /* Find the type of the first non-structure member.  */
-  elements = ty->elements;
-  candidate = elements[0]->type;
-  if (candidate == FFI_TYPE_STRUCT || candidate == FFI_TYPE_COMPLEX)
-    {
-      for (i = 0; ; ++i)
-        {
-          candidate = is_hfa0 (elements[i]);
-          if (candidate >= 0)
-            break;
-        }
-    }
-
-  /* If the first member is not a floating point type, it's not an HFA.
-     Also quickly re-check the size of the structure.  */
-  switch (candidate)
-    {
-    case FFI_TYPE_FLOAT:
-      ele_count = size / sizeof(float);
-      if (size != ele_count * sizeof(float))
-        return 0;
-      break;
-    case FFI_TYPE_DOUBLE:
-      ele_count = size / sizeof(double);
-      if (size != ele_count * sizeof(double))
-        return 0;
-      break;
-    case FFI_TYPE_LONGDOUBLE:
-      ele_count = size / sizeof(long double);
-      if (size != ele_count * sizeof(long double))
-        return 0;
-      break;
-    default:
-      return 0;
-    }
-  if (ele_count > 4)
-    return 0;
-
-  /* Finally, make sure that all scalar elements are the same type.  */
-  for (i = 0; elements[i]; ++i)
-    {
-      int t = elements[i]->type;
-      if (t == FFI_TYPE_STRUCT || t == FFI_TYPE_COMPLEX)
-        {
-          if (!is_hfa1 (elements[i], candidate))
-            return 0;
-        }
-      else if (t != candidate)
-        return 0;
-    }
-
-  /* All tests succeeded.  Encode the result.  */
- done:
-  return candidate * 4 + (4 - (int)ele_count);
-}
-
-/* Representation of the procedure call argument marshalling
-   state.
-
-   The terse state variable names match the names used in the AARCH64
-   PCS. */
-
-struct arg_state
-{
-  unsigned ngrn;                /* Next general-purpose register number. */
-  unsigned nsrn;                /* Next vector register number. */
-  size_t nsaa;                  /* Next stack offset. */
-
-#if defined (__APPLE__)
-  unsigned allocating_variadic;
-#endif
-};
-
-/* Initialize a procedure call argument marshalling state.  */
-static void
-arg_init (struct arg_state *state)
-{
-  state->ngrn = 0;
-  state->nsrn = 0;
-  state->nsaa = 0;
-#if defined (__APPLE__)
-  state->allocating_variadic = 0;
-#endif
-}
-
-/* Allocate an aligned slot on the stack and return a pointer to it.  */
-static void *
-allocate_to_stack (struct arg_state *state, void *stack,
-		   size_t alignment, size_t size)
-{
-  size_t nsaa = state->nsaa;
-
-  /* Round up the NSAA to the larger of 8 or the natural
-     alignment of the argument's type.  */
-#if defined (__APPLE__)
-  if (state->allocating_variadic && alignment < 8)
-    alignment = 8;
-#else
-  if (alignment < 8)
-    alignment = 8;
-#endif
-    
-  nsaa = FFI_ALIGN (nsaa, alignment);
-  state->nsaa = nsaa + size;
-
-  return (char *)stack + nsaa;
-}
-
-static ffi_arg
-extend_integer_type (void *source, int type)
-{
-  switch (type)
-    {
-    case FFI_TYPE_UINT8:
-      return *(UINT8 *) source;
-    case FFI_TYPE_SINT8:
-      return *(SINT8 *) source;
-    case FFI_TYPE_UINT16:
-      return *(UINT16 *) source;
-    case FFI_TYPE_SINT16:
-      return *(SINT16 *) source;
-    case FFI_TYPE_UINT32:
-      return *(UINT32 *) source;
-    case FFI_TYPE_INT:
-    case FFI_TYPE_SINT32:
-      return *(SINT32 *) source;
-    case FFI_TYPE_UINT64:
-    case FFI_TYPE_SINT64:
-      return *(UINT64 *) source;
-      break;
-    case FFI_TYPE_POINTER:
-      return *(uintptr_t *) source;
-    default:
-      abort();
-    }
-}
-
-#if defined(_MSC_VER)
-void extend_hfa_type (void *dest, void *src, int h);
-#else
-static void
-extend_hfa_type (void *dest, void *src, int h)
-{
-  ssize_t f = h - AARCH64_RET_S4;
-  void *x0;
-
-  asm volatile (
-	"adr	%0, 0f\n"
-"	add	%0, %0, %1\n"
-"	br	%0\n"
-"0:	ldp	s16, s17, [%3]\n"	/* S4 */
-"	ldp	s18, s19, [%3, #8]\n"
-"	b	4f\n"
-"	ldp	s16, s17, [%3]\n"	/* S3 */
-"	ldr	s18, [%3, #8]\n"
-"	b	3f\n"
-"	ldp	s16, s17, [%3]\n"	/* S2 */
-"	b	2f\n"
-"	nop\n"
-"	ldr	s16, [%3]\n"		/* S1 */
-"	b	1f\n"
-"	nop\n"
-"	ldp	d16, d17, [%3]\n"	/* D4 */
-"	ldp	d18, d19, [%3, #16]\n"
-"	b	4f\n"
-"	ldp	d16, d17, [%3]\n"	/* D3 */
-"	ldr	d18, [%3, #16]\n"
-"	b	3f\n"
-"	ldp	d16, d17, [%3]\n"	/* D2 */
-"	b	2f\n"
-"	nop\n"
-"	ldr	d16, [%3]\n"		/* D1 */
-"	b	1f\n"
-"	nop\n"
-"	ldp	q16, q17, [%3]\n"	/* Q4 */
-"	ldp	q18, q19, [%3, #32]\n"
-"	b	4f\n"
-"	ldp	q16, q17, [%3]\n"	/* Q3 */
-"	ldr	q18, [%3, #32]\n"
-"	b	3f\n"
-"	ldp	q16, q17, [%3]\n"	/* Q2 */
-"	b	2f\n"
-"	nop\n"
-"	ldr	q16, [%3]\n"		/* Q1 */
-"	b	1f\n"
-"4:	str	q19, [%2, #48]\n"
-"3:	str	q18, [%2, #32]\n"
-"2:	str	q17, [%2, #16]\n"
-"1:	str	q16, [%2]"
-    : "=&r"(x0)
-    : "r"(f * 12), "r"(dest), "r"(src)
-    : "memory", "v16", "v17", "v18", "v19");
-}
-#endif
-
-#if defined(_MSC_VER)
-void* compress_hfa_type (void *dest, void *src, int h);
-#else
-static void *
-compress_hfa_type (void *dest, void *reg, int h)
-{
-  switch (h)
-    {
-    case AARCH64_RET_S1:
-      if (dest == reg)
-	{
-#ifdef __AARCH64EB__
-	  dest += 12;
-#endif
-	}
-      else
-	*(float *)dest = *(float *)reg;
-      break;
-    case AARCH64_RET_S2:
-      asm ("ldp q16, q17, [%1]\n\t"
-	   "st2 { v16.s, v17.s }[0], [%0]"
-	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17");
-      break;
-    case AARCH64_RET_S3:
-      asm ("ldp q16, q17, [%1]\n\t"
-	   "ldr q18, [%1, #32]\n\t"
-	   "st3 { v16.s, v17.s, v18.s }[0], [%0]"
-	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17", "v18");
-      break;
-    case AARCH64_RET_S4:
-      asm ("ldp q16, q17, [%1]\n\t"
-	   "ldp q18, q19, [%1, #32]\n\t"
-	   "st4 { v16.s, v17.s, v18.s, v19.s }[0], [%0]"
-	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17", "v18", "v19");
-      break;
-
-    case AARCH64_RET_D1:
-      if (dest == reg)
-	{
-#ifdef __AARCH64EB__
-	  dest += 8;
-#endif
-	}
-      else
-	*(double *)dest = *(double *)reg;
-      break;
-    case AARCH64_RET_D2:
-      asm ("ldp q16, q17, [%1]\n\t"
-	   "st2 { v16.d, v17.d }[0], [%0]"
-	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17");
-      break;
-    case AARCH64_RET_D3:
-      asm ("ldp q16, q17, [%1]\n\t"
-	   "ldr q18, [%1, #32]\n\t"
-	   "st3 { v16.d, v17.d, v18.d }[0], [%0]"
-	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17", "v18");
-      break;
-    case AARCH64_RET_D4:
-      asm ("ldp q16, q17, [%1]\n\t"
-	   "ldp q18, q19, [%1, #32]\n\t"
-	   "st4 { v16.d, v17.d, v18.d, v19.d }[0], [%0]"
-	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17", "v18", "v19");
-      break;
-
-    default:
-      if (dest != reg)
-	return memcpy (dest, reg, 16 * (4 - (h & 3)));
-      break;
-    }
-  return dest;
-}
-#endif
-
-/* Either allocate an appropriate register for the argument type, or if
-   none are available, allocate a stack slot and return a pointer
-   to the allocated space.  */
-
-static void *
-allocate_int_to_reg_or_stack (struct call_context *context,
-			      struct arg_state *state,
-			      void *stack, size_t size)
-{
-  if (state->ngrn < N_X_ARG_REG)
-    return &context->x[state->ngrn++];
-
-  state->ngrn = N_X_ARG_REG;
-  return allocate_to_stack (state, stack, size, size);
-}
-
-ffi_status FFI_HIDDEN
-ffi_prep_cif_machdep (ffi_cif *cif)
-{
-  ffi_type *rtype = cif->rtype;
-  size_t bytes = cif->bytes;
-  int flags, i, n;
-
-  switch (rtype->type)
-    {
-    case FFI_TYPE_VOID:
-      flags = AARCH64_RET_VOID;
-      break;
-    case FFI_TYPE_UINT8:
-      flags = AARCH64_RET_UINT8;
-      break;
-    case FFI_TYPE_UINT16:
-      flags = AARCH64_RET_UINT16;
-      break;
-    case FFI_TYPE_UINT32:
-      flags = AARCH64_RET_UINT32;
-      break;
-    case FFI_TYPE_SINT8:
-      flags = AARCH64_RET_SINT8;
-      break;
-    case FFI_TYPE_SINT16:
-      flags = AARCH64_RET_SINT16;
-      break;
-    case FFI_TYPE_INT:
-    case FFI_TYPE_SINT32:
-      flags = AARCH64_RET_SINT32;
-      break;
-    case FFI_TYPE_SINT64:
-    case FFI_TYPE_UINT64:
-      flags = AARCH64_RET_INT64;
-      break;
-    case FFI_TYPE_POINTER:
-      flags = (sizeof(void *) == 4 ? AARCH64_RET_UINT32 : AARCH64_RET_INT64);
-      break;
-
-    case FFI_TYPE_FLOAT:
-    case FFI_TYPE_DOUBLE:
-    case FFI_TYPE_LONGDOUBLE:
-    case FFI_TYPE_STRUCT:
-    case FFI_TYPE_COMPLEX:
-      flags = is_vfp_type (rtype);
-      if (flags == 0)
-	{
-	  size_t s = rtype->size;
-	  if (s > 16)
-	    {
-	      flags = AARCH64_RET_VOID | AARCH64_RET_IN_MEM;
-	      bytes += 8;
-	    }
-	  else if (s == 16)
-	    flags = AARCH64_RET_INT128;
-	  else if (s == 8)
-	    flags = AARCH64_RET_INT64;
-	  else
-	    flags = AARCH64_RET_INT128 | AARCH64_RET_NEED_COPY;
-	}
-      break;
-
-    default:
-      abort();
-    }
-
-  for (i = 0, n = cif->nargs; i < n; i++)
-    if (is_vfp_type (cif->arg_types[i]))
-      {
-	flags |= AARCH64_FLAG_ARG_V;
-	break;
-      }
-
-  /* Round the stack up to a multiple of the stack alignment requirement. */
-  cif->bytes = (unsigned) FFI_ALIGN(bytes, 16);
-  cif->flags = flags;
-#if defined (__APPLE__)
-  cif->aarch64_nfixedargs = 0;
-#endif
-
-  return FFI_OK;
-}
-
-#if defined (__APPLE__)
-/* Perform Apple-specific cif processing for variadic calls */
-ffi_status FFI_HIDDEN
-ffi_prep_cif_machdep_var(ffi_cif *cif, unsigned int nfixedargs,
-			 unsigned int ntotalargs)
-{
-  ffi_status status = ffi_prep_cif_machdep (cif);
-  cif->aarch64_nfixedargs = nfixedargs;
-  return status;
-}
-#endif /* __APPLE__ */
-
-extern void ffi_call_SYSV (struct call_context *context, void *frame,
-			   void (*fn)(void), void *rvalue, int flags,
-			   void *closure) FFI_HIDDEN;
-
-/* Call a function with the provided arguments and capture the return
-   value.  */
-static void
-ffi_call_int (ffi_cif *cif, void (*fn)(void), void *orig_rvalue,
-	      void **avalue, void *closure)
-{
-  struct call_context *context;
-  void *stack, *frame, *rvalue;
-  struct arg_state state;
-  size_t stack_bytes, rtype_size, rsize;
-  int i, nargs, flags;
-  ffi_type *rtype;
-
-  flags = cif->flags;
-  rtype = cif->rtype;
-  rtype_size = rtype->size;
-  stack_bytes = cif->bytes;
-
-  /* If the target function returns a structure via hidden pointer,
-     then we cannot allow a null rvalue.  Otherwise, mash a null
-     rvalue to void return type.  */
-  rsize = 0;
-  if (flags & AARCH64_RET_IN_MEM)
-    {
-      if (orig_rvalue == NULL)
-	rsize = rtype_size;
-    }
-  else if (orig_rvalue == NULL)
-    flags &= AARCH64_FLAG_ARG_V;
-  else if (flags & AARCH64_RET_NEED_COPY)
-    rsize = 16;
-
-  /* Allocate consectutive stack for everything we'll need.  */
-  context = alloca (sizeof(struct call_context) + stack_bytes + 32 + rsize);
-  stack = context + 1;
-  frame = (void*)((uintptr_t)stack + (uintptr_t)stack_bytes);
-  rvalue = (rsize ? (void*)((uintptr_t)frame + 32) : orig_rvalue);
-
-  arg_init (&state);
-  for (i = 0, nargs = cif->nargs; i < nargs; i++)
-    {
-      ffi_type *ty = cif->arg_types[i];
-      size_t s = ty->size;
-      void *a = avalue[i];
-      int h, t;
-
-      t = ty->type;
-      switch (t)
-	{
-	case FFI_TYPE_VOID:
-	  FFI_ASSERT (0);
-	  break;
-
-	/* If the argument is a basic type the argument is allocated to an
-	   appropriate register, or if none are available, to the stack.  */
-	case FFI_TYPE_INT:
-	case FFI_TYPE_UINT8:
-	case FFI_TYPE_SINT8:
-	case FFI_TYPE_UINT16:
-	case FFI_TYPE_SINT16:
-	case FFI_TYPE_UINT32:
-	case FFI_TYPE_SINT32:
-	case FFI_TYPE_UINT64:
-	case FFI_TYPE_SINT64:
-	case FFI_TYPE_POINTER:
-	do_pointer:
-	  {
-	    ffi_arg ext = extend_integer_type (a, t);
-	    if (state.ngrn < N_X_ARG_REG)
-	      context->x[state.ngrn++] = ext;
-	    else
-	      {
-		void *d = allocate_to_stack (&state, stack, ty->alignment, s);
-		state.ngrn = N_X_ARG_REG;
-		/* Note that the default abi extends each argument
-		   to a full 64-bit slot, while the iOS abi allocates
-		   only enough space. */
-#ifdef __APPLE__
-		memcpy(d, a, s);
-#else
-		*(ffi_arg *)d = ext;
-#endif
-	      }
-	  }
-	  break;
-
-	case FFI_TYPE_FLOAT:
-	case FFI_TYPE_DOUBLE:
-	case FFI_TYPE_LONGDOUBLE:
-	case FFI_TYPE_STRUCT:
-	case FFI_TYPE_COMPLEX:
-	  {
-	    void *dest;
-
-	    h = is_vfp_type (ty);
-	    if (h)
-	      {
-		int elems = 4 - (h & 3);
-#ifdef _M_ARM64 /* for handling armasm calling convention */
-                if (cif->is_variadic)
-                  {
-                    if (state.ngrn + elems <= N_X_ARG_REG)
-                      {
-                        dest = &context->x[state.ngrn];
-                        state.ngrn += elems;
-                        extend_hfa_type(dest, a, h);
-                        break;
-                      }
-                    state.nsrn = N_X_ARG_REG;
-                    dest = allocate_to_stack(&state, stack, ty->alignment, s);
-                  }
-                else
-                  {
-#endif /* for handling armasm calling convention */
-	        if (state.nsrn + elems <= N_V_ARG_REG)
-		  {
-		    dest = &context->v[state.nsrn];
-		    state.nsrn += elems;
-		    extend_hfa_type (dest, a, h);
-		    break;
-		  }
-		state.nsrn = N_V_ARG_REG;
-		dest = allocate_to_stack (&state, stack, ty->alignment, s);
-#ifdef _M_ARM64 /* for handling armasm calling convention */
-	      }
-#endif /* for handling armasm calling convention */
-	      }
-	    else if (s > 16)
-	      {
-		/* If the argument is a composite type that is larger than 16
-		   bytes, then the argument has been copied to memory, and
-		   the argument is replaced by a pointer to the copy.  */
-		a = &avalue[i];
-		t = FFI_TYPE_POINTER;
-		s = sizeof (void *);
-		goto do_pointer;
-	      }
-	    else
-	      {
-		size_t n = (s + 7) / 8;
-		if (state.ngrn + n <= N_X_ARG_REG)
-		  {
-		    /* If the argument is a composite type and the size in
-		       double-words is not more than the number of available
-		       X registers, then the argument is copied into
-		       consecutive X registers.  */
-		    dest = &context->x[state.ngrn];
-                    state.ngrn += (unsigned int)n;
-		  }
-		else
-		  {
-		    /* Otherwise, there are insufficient X registers. Further
-		       X register allocations are prevented, the NSAA is
-		       adjusted and the argument is copied to memory at the
-		       adjusted NSAA.  */
-		    state.ngrn = N_X_ARG_REG;
-		    dest = allocate_to_stack (&state, stack, ty->alignment, s);
-		  }
-		}
-	      memcpy (dest, a, s);
-	    }
-	  break;
-
-	default:
-	  abort();
-	}
-
-#if defined (__APPLE__)
-      if (i + 1 == cif->aarch64_nfixedargs)
-	{
-	  state.ngrn = N_X_ARG_REG;
-	  state.nsrn = N_V_ARG_REG;
-	  state.allocating_variadic = 1;
-	}
-#endif
-    }
-
-  ffi_call_SYSV (context, frame, fn, rvalue, flags, closure);
-
-  if (flags & AARCH64_RET_NEED_COPY)
-    memcpy (orig_rvalue, rvalue, rtype_size);
-}
-
-void
-ffi_call (ffi_cif *cif, void (*fn) (void), void *rvalue, void **avalue)
-{
-  ffi_call_int (cif, fn, rvalue, avalue, NULL);
-}
-
-#ifdef FFI_GO_CLOSURES
-void
-ffi_call_go (ffi_cif *cif, void (*fn) (void), void *rvalue,
-	     void **avalue, void *closure)
-{
-  ffi_call_int (cif, fn, rvalue, avalue, closure);
-}
-#endif /* FFI_GO_CLOSURES */
-
-/* Build a trampoline.  */
-
-extern void ffi_closure_SYSV (void) FFI_HIDDEN;
-extern void ffi_closure_SYSV_V (void) FFI_HIDDEN;
-
-ffi_status
-ffi_prep_closure_loc (ffi_closure *closure,
-                      ffi_cif* cif,
-                      void (*fun)(ffi_cif*,void*,void**,void*),
-                      void *user_data,
-                      void *codeloc)
-{
-  if (cif->abi != FFI_SYSV)
-    return FFI_BAD_ABI;
-
-  void (*start)(void);
-  
-  if (cif->flags & AARCH64_FLAG_ARG_V)
-    start = ffi_closure_SYSV_V;
-  else
-    start = ffi_closure_SYSV;
-
-#if FFI_EXEC_TRAMPOLINE_TABLE
-#ifdef __MACH__
-  void **config = (void **)((uint8_t *)codeloc - PAGE_MAX_SIZE);
-  config[0] = closure;
-  config[1] = start;
-#endif
-#else
-  static const unsigned char trampoline[16] = {
-    0x90, 0x00, 0x00, 0x58,	/* ldr	x16, tramp+16	*/
-    0xf1, 0xff, 0xff, 0x10,	/* adr	x17, tramp+0	*/
-    0x00, 0x02, 0x1f, 0xd6	/* br	x16		*/
-  };
-  char *tramp = closure->tramp;
-  
-  memcpy (tramp, trampoline, sizeof(trampoline));
-  
-  *(UINT64 *)(tramp + 16) = (uintptr_t)start;
-
-  ffi_clear_cache(tramp, tramp + FFI_TRAMPOLINE_SIZE);
-
-  /* Also flush the cache for code mapping.  */
-#ifdef _M_ARM64
-  // Not using dlmalloc.c for Windows ARM64 builds
-  // so calling ffi_data_to_code_pointer() isn't necessary
-  unsigned char *tramp_code = tramp;
-  #else
-  unsigned char *tramp_code = ffi_data_to_code_pointer (tramp);
-  #endif
-  ffi_clear_cache (tramp_code, tramp_code + FFI_TRAMPOLINE_SIZE);
-#endif
-
-  closure->cif = cif;
-  closure->fun = fun;
-  closure->user_data = user_data;
-
-  return FFI_OK;
-}
-
-#ifdef FFI_GO_CLOSURES
-extern void ffi_go_closure_SYSV (void) FFI_HIDDEN;
-extern void ffi_go_closure_SYSV_V (void) FFI_HIDDEN;
-
-ffi_status
-ffi_prep_go_closure (ffi_go_closure *closure, ffi_cif* cif,
-                     void (*fun)(ffi_cif*,void*,void**,void*))
-{
-  void (*start)(void);
-
-  if (cif->abi != FFI_SYSV)
-    return FFI_BAD_ABI;
-
-  if (cif->flags & AARCH64_FLAG_ARG_V)
-    start = ffi_go_closure_SYSV_V;
-  else
-    start = ffi_go_closure_SYSV;
-
-  closure->tramp = start;
-  closure->cif = cif;
-  closure->fun = fun;
-
-  return FFI_OK;
-}
-#endif /* FFI_GO_CLOSURES */
-
-/* Primary handler to setup and invoke a function within a closure.
-
-   A closure when invoked enters via the assembler wrapper
-   ffi_closure_SYSV(). The wrapper allocates a call context on the
-   stack, saves the interesting registers (from the perspective of
-   the calling convention) into the context then passes control to
-   ffi_closure_SYSV_inner() passing the saved context and a pointer to
-   the stack at the point ffi_closure_SYSV() was invoked.
-
-   On the return path the assembler wrapper will reload call context
-   registers.
-
-   ffi_closure_SYSV_inner() marshalls the call context into ffi value
-   descriptors, invokes the wrapped function, then marshalls the return
-   value back into the call context.  */
-
-int FFI_HIDDEN
-ffi_closure_SYSV_inner (ffi_cif *cif,
-			void (*fun)(ffi_cif*,void*,void**,void*),
-			void *user_data,
-			struct call_context *context,
-			void *stack, void *rvalue, void *struct_rvalue)
-{
-  void **avalue = (void**) alloca (cif->nargs * sizeof (void*));
-  int i, h, nargs, flags;
-  struct arg_state state;
-
-  arg_init (&state);
-
-  for (i = 0, nargs = cif->nargs; i < nargs; i++)
-    {
-      ffi_type *ty = cif->arg_types[i];
-      int t = ty->type;
-      size_t n, s = ty->size;
-
-      switch (t)
-	{
-	case FFI_TYPE_VOID:
-	  FFI_ASSERT (0);
-	  break;
-
-	case FFI_TYPE_INT:
-	case FFI_TYPE_UINT8:
-	case FFI_TYPE_SINT8:
-	case FFI_TYPE_UINT16:
-	case FFI_TYPE_SINT16:
-	case FFI_TYPE_UINT32:
-	case FFI_TYPE_SINT32:
-	case FFI_TYPE_UINT64:
-	case FFI_TYPE_SINT64:
-	case FFI_TYPE_POINTER:
-	  avalue[i] = allocate_int_to_reg_or_stack (context, &state, stack, s);
-	  break;
-
-	case FFI_TYPE_FLOAT:
-	case FFI_TYPE_DOUBLE:
-	case FFI_TYPE_LONGDOUBLE:
-	case FFI_TYPE_STRUCT:
-	case FFI_TYPE_COMPLEX:
-	  h = is_vfp_type (ty);
-	  if (h)
-	    {
-	      n = 4 - (h & 3);
-#ifdef _M_ARM64  /* for handling armasm calling convention */
-              if (cif->is_variadic)
-                {
-                  if (state.ngrn + n <= N_X_ARG_REG)
-                    {
-                      void *reg = &context->x[state.ngrn];
-                      state.ngrn += (unsigned int)n;
-    
-                      /* Eeek! We need a pointer to the structure, however the
-                       homogeneous float elements are being passed in individual
-                       registers, therefore for float and double the structure
-                       is not represented as a contiguous sequence of bytes in
-                       our saved register context.  We don't need the original
-                       contents of the register storage, so we reformat the
-                       structure into the same memory.  */
-                      avalue[i] = compress_hfa_type(reg, reg, h);
-                    }
-                  else
-                    {
-                      state.ngrn = N_X_ARG_REG;
-                      state.nsrn = N_V_ARG_REG;
-                      avalue[i] = allocate_to_stack(&state, stack,
-                             ty->alignment, s);
-                    }
-                }
-              else
-                {
-#endif  /* for handling armasm calling convention */
-                  if (state.nsrn + n <= N_V_ARG_REG)
-                    {
-                      void *reg = &context->v[state.nsrn];
-                      state.nsrn += (unsigned int)n;
-                      avalue[i] = compress_hfa_type(reg, reg, h);
-                    }
-                  else
-                    {
-                      state.nsrn = N_V_ARG_REG;
-                      avalue[i] = allocate_to_stack(&state, stack,
-                                                   ty->alignment, s);
-                    }
-#ifdef _M_ARM64  /* for handling armasm calling convention */
-                }
-#endif  /* for handling armasm calling convention */
-            }
-          else if (s > 16)
-            {
-              /* Replace Composite type of size greater than 16 with a
-                  pointer.  */
-              avalue[i] = *(void **)
-              allocate_int_to_reg_or_stack (context, &state, stack,
-                                         sizeof (void *));
-            }
-          else
-            {
-              n = (s + 7) / 8;
-              if (state.ngrn + n <= N_X_ARG_REG)
-                {
-                  avalue[i] = &context->x[state.ngrn];
-                  state.ngrn += (unsigned int)n;
-                }
-              else
-                {
-                  state.ngrn = N_X_ARG_REG;
-                  avalue[i] = allocate_to_stack(&state, stack,
-                                           ty->alignment, s);
-                }
-            }
-          break;
-
-        default:
-          abort();
-      }
-
-#if defined (__APPLE__)
-      if (i + 1 == cif->aarch64_nfixedargs)
-	{
-	  state.ngrn = N_X_ARG_REG;
-	  state.nsrn = N_V_ARG_REG;
-	  state.allocating_variadic = 1;
-	}
-#endif
-    }
-
-  flags = cif->flags;
-  if (flags & AARCH64_RET_IN_MEM)
-    rvalue = struct_rvalue;
-
-  fun (cif, rvalue, avalue, user_data);
-
-  return flags;
-}
-
-#endif /* (__aarch64__) || defined(__arm64__)|| defined (_M_ARM64)*/
+/* Copyright (c) 2009, 2010, 2011, 2012 ARM Ltd. 
+ 
+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 THE AUTHORS OR COPYRIGHT HOLDERS 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.  */ 
+ 
+#if defined(__aarch64__) || defined(__arm64__)|| defined (_M_ARM64) 
+#include <stdio.h> 
+#include <stdlib.h> 
+#include <stdint.h> 
+#include <fficonfig.h> 
+#include <ffi.h> 
+#include <ffi_common.h> 
+#include "internal.h" 
+#ifdef _M_ARM64 
+#include <windows.h> /* FlushInstructionCache */ 
+#endif 
+ 
+/* Force FFI_TYPE_LONGDOUBLE to be different than FFI_TYPE_DOUBLE; 
+   all further uses in this file will refer to the 128-bit type.  */ 
+#if FFI_TYPE_DOUBLE != FFI_TYPE_LONGDOUBLE 
+# if FFI_TYPE_LONGDOUBLE != 4 
+#  error FFI_TYPE_LONGDOUBLE out of date 
+# endif 
+#else 
+# undef FFI_TYPE_LONGDOUBLE 
+# define FFI_TYPE_LONGDOUBLE 4 
+#endif 
+ 
+union _d 
+{ 
+  UINT64 d; 
+  UINT32 s[2]; 
+}; 
+ 
+struct _v 
+{ 
+  union _d d[2] __attribute__((aligned(16))); 
+}; 
+ 
+struct call_context 
+{ 
+  struct _v v[N_V_ARG_REG]; 
+  UINT64 x[N_X_ARG_REG]; 
+}; 
+ 
+#if FFI_EXEC_TRAMPOLINE_TABLE 
+ 
+#ifdef __MACH__ 
+#include <mach/vm_param.h> 
+#endif 
+ 
+#else 
+ 
+#if defined (__clang__) && defined (__APPLE__) 
+extern void sys_icache_invalidate (void *start, size_t len); 
+#endif 
+ 
+static inline void 
+ffi_clear_cache (void *start, void *end) 
+{ 
+#if defined (__clang__) && defined (__APPLE__) 
+  sys_icache_invalidate (start, (char *)end - (char *)start); 
+#elif defined (__GNUC__) 
+  __builtin___clear_cache (start, end); 
+#elif defined (_M_ARM64) 
+  FlushInstructionCache(GetCurrentProcess(), start, (char*)end - (char*)start); 
+#else 
+#error "Missing builtin to flush instruction cache" 
+#endif 
+} 
+ 
+#endif 
+ 
+/* A subroutine of is_vfp_type.  Given a structure type, return the type code 
+   of the first non-structure element.  Recurse for structure elements. 
+   Return -1 if the structure is in fact empty, i.e. no nested elements.  */ 
+ 
+static int 
+is_hfa0 (const ffi_type *ty) 
+{ 
+  ffi_type **elements = ty->elements; 
+  int i, ret = -1; 
+ 
+  if (elements != NULL) 
+    for (i = 0; elements[i]; ++i) 
+      { 
+        ret = elements[i]->type; 
+        if (ret == FFI_TYPE_STRUCT || ret == FFI_TYPE_COMPLEX) 
+          { 
+            ret = is_hfa0 (elements[i]); 
+            if (ret < 0) 
+              continue; 
+          } 
+        break; 
+      } 
+ 
+  return ret; 
+} 
+ 
+/* A subroutine of is_vfp_type.  Given a structure type, return true if all 
+   of the non-structure elements are the same as CANDIDATE.  */ 
+ 
+static int 
+is_hfa1 (const ffi_type *ty, int candidate) 
+{ 
+  ffi_type **elements = ty->elements; 
+  int i; 
+ 
+  if (elements != NULL) 
+    for (i = 0; elements[i]; ++i) 
+      { 
+        int t = elements[i]->type; 
+        if (t == FFI_TYPE_STRUCT || t == FFI_TYPE_COMPLEX) 
+          { 
+            if (!is_hfa1 (elements[i], candidate)) 
+              return 0; 
+          } 
+        else if (t != candidate) 
+          return 0; 
+      } 
+ 
+  return 1; 
+} 
+ 
+/* Determine if TY may be allocated to the FP registers.  This is both an 
+   fp scalar type as well as an homogenous floating point aggregate (HFA). 
+   That is, a structure consisting of 1 to 4 members of all the same type, 
+   where that type is an fp scalar. 
+ 
+   Returns non-zero iff TY is an HFA.  The result is the AARCH64_RET_* 
+   constant for the type.  */ 
+ 
+static int 
+is_vfp_type (const ffi_type *ty) 
+{ 
+  ffi_type **elements; 
+  int candidate, i; 
+  size_t size, ele_count; 
+ 
+  /* Quickest tests first.  */ 
+  candidate = ty->type; 
+  switch (candidate) 
+    { 
+    default: 
+      return 0; 
+    case FFI_TYPE_FLOAT: 
+    case FFI_TYPE_DOUBLE: 
+    case FFI_TYPE_LONGDOUBLE: 
+      ele_count = 1; 
+      goto done; 
+    case FFI_TYPE_COMPLEX: 
+      candidate = ty->elements[0]->type; 
+      switch (candidate) 
+	{ 
+	case FFI_TYPE_FLOAT: 
+	case FFI_TYPE_DOUBLE: 
+	case FFI_TYPE_LONGDOUBLE: 
+	  ele_count = 2; 
+	  goto done; 
+	} 
+      return 0; 
+    case FFI_TYPE_STRUCT: 
+      break; 
+    } 
+ 
+  /* No HFA types are smaller than 4 bytes, or larger than 64 bytes.  */ 
+  size = ty->size; 
+  if (size < 4 || size > 64) 
+    return 0; 
+ 
+  /* Find the type of the first non-structure member.  */ 
+  elements = ty->elements; 
+  candidate = elements[0]->type; 
+  if (candidate == FFI_TYPE_STRUCT || candidate == FFI_TYPE_COMPLEX) 
+    { 
+      for (i = 0; ; ++i) 
+        { 
+          candidate = is_hfa0 (elements[i]); 
+          if (candidate >= 0) 
+            break; 
+        } 
+    } 
+ 
+  /* If the first member is not a floating point type, it's not an HFA. 
+     Also quickly re-check the size of the structure.  */ 
+  switch (candidate) 
+    { 
+    case FFI_TYPE_FLOAT: 
+      ele_count = size / sizeof(float); 
+      if (size != ele_count * sizeof(float)) 
+        return 0; 
+      break; 
+    case FFI_TYPE_DOUBLE: 
+      ele_count = size / sizeof(double); 
+      if (size != ele_count * sizeof(double)) 
+        return 0; 
+      break; 
+    case FFI_TYPE_LONGDOUBLE: 
+      ele_count = size / sizeof(long double); 
+      if (size != ele_count * sizeof(long double)) 
+        return 0; 
+      break; 
+    default: 
+      return 0; 
+    } 
+  if (ele_count > 4) 
+    return 0; 
+ 
+  /* Finally, make sure that all scalar elements are the same type.  */ 
+  for (i = 0; elements[i]; ++i) 
+    { 
+      int t = elements[i]->type; 
+      if (t == FFI_TYPE_STRUCT || t == FFI_TYPE_COMPLEX) 
+        { 
+          if (!is_hfa1 (elements[i], candidate)) 
+            return 0; 
+        } 
+      else if (t != candidate) 
+        return 0; 
+    } 
+ 
+  /* All tests succeeded.  Encode the result.  */ 
+ done: 
+  return candidate * 4 + (4 - (int)ele_count); 
+} 
+ 
+/* Representation of the procedure call argument marshalling 
+   state. 
+ 
+   The terse state variable names match the names used in the AARCH64 
+   PCS. */ 
+ 
+struct arg_state 
+{ 
+  unsigned ngrn;                /* Next general-purpose register number. */ 
+  unsigned nsrn;                /* Next vector register number. */ 
+  size_t nsaa;                  /* Next stack offset. */ 
+ 
+#if defined (__APPLE__) 
+  unsigned allocating_variadic; 
+#endif 
+}; 
+ 
+/* Initialize a procedure call argument marshalling state.  */ 
+static void 
+arg_init (struct arg_state *state) 
+{ 
+  state->ngrn = 0; 
+  state->nsrn = 0; 
+  state->nsaa = 0; 
+#if defined (__APPLE__) 
+  state->allocating_variadic = 0; 
+#endif 
+} 
+ 
+/* Allocate an aligned slot on the stack and return a pointer to it.  */ 
+static void * 
+allocate_to_stack (struct arg_state *state, void *stack, 
+		   size_t alignment, size_t size) 
+{ 
+  size_t nsaa = state->nsaa; 
+ 
+  /* Round up the NSAA to the larger of 8 or the natural 
+     alignment of the argument's type.  */ 
+#if defined (__APPLE__) 
+  if (state->allocating_variadic && alignment < 8) 
+    alignment = 8; 
+#else 
+  if (alignment < 8) 
+    alignment = 8; 
+#endif 
+     
+  nsaa = FFI_ALIGN (nsaa, alignment); 
+  state->nsaa = nsaa + size; 
+ 
+  return (char *)stack + nsaa; 
+} 
+ 
+static ffi_arg 
+extend_integer_type (void *source, int type) 
+{ 
+  switch (type) 
+    { 
+    case FFI_TYPE_UINT8: 
+      return *(UINT8 *) source; 
+    case FFI_TYPE_SINT8: 
+      return *(SINT8 *) source; 
+    case FFI_TYPE_UINT16: 
+      return *(UINT16 *) source; 
+    case FFI_TYPE_SINT16: 
+      return *(SINT16 *) source; 
+    case FFI_TYPE_UINT32: 
+      return *(UINT32 *) source; 
+    case FFI_TYPE_INT: 
+    case FFI_TYPE_SINT32: 
+      return *(SINT32 *) source; 
+    case FFI_TYPE_UINT64: 
+    case FFI_TYPE_SINT64: 
+      return *(UINT64 *) source; 
+      break; 
+    case FFI_TYPE_POINTER: 
+      return *(uintptr_t *) source; 
+    default: 
+      abort(); 
+    } 
+} 
+ 
+#if defined(_MSC_VER) 
+void extend_hfa_type (void *dest, void *src, int h); 
+#else 
+static void 
+extend_hfa_type (void *dest, void *src, int h) 
+{ 
+  ssize_t f = h - AARCH64_RET_S4; 
+  void *x0; 
+ 
+  asm volatile ( 
+	"adr	%0, 0f\n" 
+"	add	%0, %0, %1\n" 
+"	br	%0\n" 
+"0:	ldp	s16, s17, [%3]\n"	/* S4 */ 
+"	ldp	s18, s19, [%3, #8]\n" 
+"	b	4f\n" 
+"	ldp	s16, s17, [%3]\n"	/* S3 */ 
+"	ldr	s18, [%3, #8]\n" 
+"	b	3f\n" 
+"	ldp	s16, s17, [%3]\n"	/* S2 */ 
+"	b	2f\n" 
+"	nop\n" 
+"	ldr	s16, [%3]\n"		/* S1 */ 
+"	b	1f\n" 
+"	nop\n" 
+"	ldp	d16, d17, [%3]\n"	/* D4 */ 
+"	ldp	d18, d19, [%3, #16]\n" 
+"	b	4f\n" 
+"	ldp	d16, d17, [%3]\n"	/* D3 */ 
+"	ldr	d18, [%3, #16]\n" 
+"	b	3f\n" 
+"	ldp	d16, d17, [%3]\n"	/* D2 */ 
+"	b	2f\n" 
+"	nop\n" 
+"	ldr	d16, [%3]\n"		/* D1 */ 
+"	b	1f\n" 
+"	nop\n" 
+"	ldp	q16, q17, [%3]\n"	/* Q4 */ 
+"	ldp	q18, q19, [%3, #32]\n" 
+"	b	4f\n" 
+"	ldp	q16, q17, [%3]\n"	/* Q3 */ 
+"	ldr	q18, [%3, #32]\n" 
+"	b	3f\n" 
+"	ldp	q16, q17, [%3]\n"	/* Q2 */ 
+"	b	2f\n" 
+"	nop\n" 
+"	ldr	q16, [%3]\n"		/* Q1 */ 
+"	b	1f\n" 
+"4:	str	q19, [%2, #48]\n" 
+"3:	str	q18, [%2, #32]\n" 
+"2:	str	q17, [%2, #16]\n" 
+"1:	str	q16, [%2]" 
+    : "=&r"(x0) 
+    : "r"(f * 12), "r"(dest), "r"(src) 
+    : "memory", "v16", "v17", "v18", "v19"); 
+} 
+#endif 
+ 
+#if defined(_MSC_VER) 
+void* compress_hfa_type (void *dest, void *src, int h); 
+#else 
+static void * 
+compress_hfa_type (void *dest, void *reg, int h) 
+{ 
+  switch (h) 
+    { 
+    case AARCH64_RET_S1: 
+      if (dest == reg) 
+	{ 
+#ifdef __AARCH64EB__ 
+	  dest += 12; 
+#endif 
+	} 
+      else 
+	*(float *)dest = *(float *)reg; 
+      break; 
+    case AARCH64_RET_S2: 
+      asm ("ldp q16, q17, [%1]\n\t" 
+	   "st2 { v16.s, v17.s }[0], [%0]" 
+	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17"); 
+      break; 
+    case AARCH64_RET_S3: 
+      asm ("ldp q16, q17, [%1]\n\t" 
+	   "ldr q18, [%1, #32]\n\t" 
+	   "st3 { v16.s, v17.s, v18.s }[0], [%0]" 
+	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17", "v18"); 
+      break; 
+    case AARCH64_RET_S4: 
+      asm ("ldp q16, q17, [%1]\n\t" 
+	   "ldp q18, q19, [%1, #32]\n\t" 
+	   "st4 { v16.s, v17.s, v18.s, v19.s }[0], [%0]" 
+	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17", "v18", "v19"); 
+      break; 
+ 
+    case AARCH64_RET_D1: 
+      if (dest == reg) 
+	{ 
+#ifdef __AARCH64EB__ 
+	  dest += 8; 
+#endif 
+	} 
+      else 
+	*(double *)dest = *(double *)reg; 
+      break; 
+    case AARCH64_RET_D2: 
+      asm ("ldp q16, q17, [%1]\n\t" 
+	   "st2 { v16.d, v17.d }[0], [%0]" 
+	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17"); 
+      break; 
+    case AARCH64_RET_D3: 
+      asm ("ldp q16, q17, [%1]\n\t" 
+	   "ldr q18, [%1, #32]\n\t" 
+	   "st3 { v16.d, v17.d, v18.d }[0], [%0]" 
+	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17", "v18"); 
+      break; 
+    case AARCH64_RET_D4: 
+      asm ("ldp q16, q17, [%1]\n\t" 
+	   "ldp q18, q19, [%1, #32]\n\t" 
+	   "st4 { v16.d, v17.d, v18.d, v19.d }[0], [%0]" 
+	   : : "r"(dest), "r"(reg) : "memory", "v16", "v17", "v18", "v19"); 
+      break; 
+ 
+    default: 
+      if (dest != reg) 
+	return memcpy (dest, reg, 16 * (4 - (h & 3))); 
+      break; 
+    } 
+  return dest; 
+} 
+#endif 
+ 
+/* Either allocate an appropriate register for the argument type, or if 
+   none are available, allocate a stack slot and return a pointer 
+   to the allocated space.  */ 
+ 
+static void * 
+allocate_int_to_reg_or_stack (struct call_context *context, 
+			      struct arg_state *state, 
+			      void *stack, size_t size) 
+{ 
+  if (state->ngrn < N_X_ARG_REG) 
+    return &context->x[state->ngrn++]; 
+ 
+  state->ngrn = N_X_ARG_REG; 
+  return allocate_to_stack (state, stack, size, size); 
+} 
+ 
+ffi_status FFI_HIDDEN 
+ffi_prep_cif_machdep (ffi_cif *cif) 
+{ 
+  ffi_type *rtype = cif->rtype; 
+  size_t bytes = cif->bytes; 
+  int flags, i, n; 
+ 
+  switch (rtype->type) 
+    { 
+    case FFI_TYPE_VOID: 
+      flags = AARCH64_RET_VOID; 
+      break; 
+    case FFI_TYPE_UINT8: 
+      flags = AARCH64_RET_UINT8; 
+      break; 
+    case FFI_TYPE_UINT16: 
+      flags = AARCH64_RET_UINT16; 
+      break; 
+    case FFI_TYPE_UINT32: 
+      flags = AARCH64_RET_UINT32; 
+      break; 
+    case FFI_TYPE_SINT8: 
+      flags = AARCH64_RET_SINT8; 
+      break; 
+    case FFI_TYPE_SINT16: 
+      flags = AARCH64_RET_SINT16; 
+      break; 
+    case FFI_TYPE_INT: 
+    case FFI_TYPE_SINT32: 
+      flags = AARCH64_RET_SINT32; 
+      break; 
+    case FFI_TYPE_SINT64: 
+    case FFI_TYPE_UINT64: 
+      flags = AARCH64_RET_INT64; 
+      break; 
+    case FFI_TYPE_POINTER: 
+      flags = (sizeof(void *) == 4 ? AARCH64_RET_UINT32 : AARCH64_RET_INT64); 
+      break; 
+ 
+    case FFI_TYPE_FLOAT: 
+    case FFI_TYPE_DOUBLE: 
+    case FFI_TYPE_LONGDOUBLE: 
+    case FFI_TYPE_STRUCT: 
+    case FFI_TYPE_COMPLEX: 
+      flags = is_vfp_type (rtype); 
+      if (flags == 0) 
+	{ 
+	  size_t s = rtype->size; 
+	  if (s > 16) 
+	    { 
+	      flags = AARCH64_RET_VOID | AARCH64_RET_IN_MEM; 
+	      bytes += 8; 
+	    } 
+	  else if (s == 16) 
+	    flags = AARCH64_RET_INT128; 
+	  else if (s == 8) 
+	    flags = AARCH64_RET_INT64; 
+	  else 
+	    flags = AARCH64_RET_INT128 | AARCH64_RET_NEED_COPY; 
+	} 
+      break; 
+ 
+    default: 
+      abort(); 
+    } 
+ 
+  for (i = 0, n = cif->nargs; i < n; i++) 
+    if (is_vfp_type (cif->arg_types[i])) 
+      { 
+	flags |= AARCH64_FLAG_ARG_V; 
+	break; 
+      } 
+ 
+  /* Round the stack up to a multiple of the stack alignment requirement. */ 
+  cif->bytes = (unsigned) FFI_ALIGN(bytes, 16); 
+  cif->flags = flags; 
+#if defined (__APPLE__) 
+  cif->aarch64_nfixedargs = 0; 
+#endif 
+ 
+  return FFI_OK; 
+} 
+ 
+#if defined (__APPLE__) 
+/* Perform Apple-specific cif processing for variadic calls */ 
+ffi_status FFI_HIDDEN 
+ffi_prep_cif_machdep_var(ffi_cif *cif, unsigned int nfixedargs, 
+			 unsigned int ntotalargs) 
+{ 
+  ffi_status status = ffi_prep_cif_machdep (cif); 
+  cif->aarch64_nfixedargs = nfixedargs; 
+  return status; 
+} 
+#endif /* __APPLE__ */ 
+ 
+extern void ffi_call_SYSV (struct call_context *context, void *frame, 
+			   void (*fn)(void), void *rvalue, int flags, 
+			   void *closure) FFI_HIDDEN; 
+ 
+/* Call a function with the provided arguments and capture the return 
+   value.  */ 
+static void 
+ffi_call_int (ffi_cif *cif, void (*fn)(void), void *orig_rvalue, 
+	      void **avalue, void *closure) 
+{ 
+  struct call_context *context; 
+  void *stack, *frame, *rvalue; 
+  struct arg_state state; 
+  size_t stack_bytes, rtype_size, rsize; 
+  int i, nargs, flags; 
+  ffi_type *rtype; 
+ 
+  flags = cif->flags; 
+  rtype = cif->rtype; 
+  rtype_size = rtype->size; 
+  stack_bytes = cif->bytes; 
+ 
+  /* If the target function returns a structure via hidden pointer, 
+     then we cannot allow a null rvalue.  Otherwise, mash a null 
+     rvalue to void return type.  */ 
+  rsize = 0; 
+  if (flags & AARCH64_RET_IN_MEM) 
+    { 
+      if (orig_rvalue == NULL) 
+	rsize = rtype_size; 
+    } 
+  else if (orig_rvalue == NULL) 
+    flags &= AARCH64_FLAG_ARG_V; 
+  else if (flags & AARCH64_RET_NEED_COPY) 
+    rsize = 16; 
+ 
+  /* Allocate consectutive stack for everything we'll need.  */ 
+  context = alloca (sizeof(struct call_context) + stack_bytes + 32 + rsize); 
+  stack = context + 1; 
+  frame = (void*)((uintptr_t)stack + (uintptr_t)stack_bytes); 
+  rvalue = (rsize ? (void*)((uintptr_t)frame + 32) : orig_rvalue); 
+ 
+  arg_init (&state); 
+  for (i = 0, nargs = cif->nargs; i < nargs; i++) 
+    { 
+      ffi_type *ty = cif->arg_types[i]; 
+      size_t s = ty->size; 
+      void *a = avalue[i]; 
+      int h, t; 
+ 
+      t = ty->type; 
+      switch (t) 
+	{ 
+	case FFI_TYPE_VOID: 
+	  FFI_ASSERT (0); 
+	  break; 
+ 
+	/* If the argument is a basic type the argument is allocated to an 
+	   appropriate register, or if none are available, to the stack.  */ 
+	case FFI_TYPE_INT: 
+	case FFI_TYPE_UINT8: 
+	case FFI_TYPE_SINT8: 
+	case FFI_TYPE_UINT16: 
+	case FFI_TYPE_SINT16: 
+	case FFI_TYPE_UINT32: 
+	case FFI_TYPE_SINT32: 
+	case FFI_TYPE_UINT64: 
+	case FFI_TYPE_SINT64: 
+	case FFI_TYPE_POINTER: 
+	do_pointer: 
+	  { 
+	    ffi_arg ext = extend_integer_type (a, t); 
+	    if (state.ngrn < N_X_ARG_REG) 
+	      context->x[state.ngrn++] = ext; 
+	    else 
+	      { 
+		void *d = allocate_to_stack (&state, stack, ty->alignment, s); 
+		state.ngrn = N_X_ARG_REG; 
+		/* Note that the default abi extends each argument 
+		   to a full 64-bit slot, while the iOS abi allocates 
+		   only enough space. */ 
+#ifdef __APPLE__ 
+		memcpy(d, a, s); 
+#else 
+		*(ffi_arg *)d = ext; 
+#endif 
+	      } 
+	  } 
+	  break; 
+ 
+	case FFI_TYPE_FLOAT: 
+	case FFI_TYPE_DOUBLE: 
+	case FFI_TYPE_LONGDOUBLE: 
+	case FFI_TYPE_STRUCT: 
+	case FFI_TYPE_COMPLEX: 
+	  { 
+	    void *dest; 
+ 
+	    h = is_vfp_type (ty); 
+	    if (h) 
+	      { 
+		int elems = 4 - (h & 3); 
+#ifdef _M_ARM64 /* for handling armasm calling convention */ 
+                if (cif->is_variadic) 
+                  { 
+                    if (state.ngrn + elems <= N_X_ARG_REG) 
+                      { 
+                        dest = &context->x[state.ngrn]; 
+                        state.ngrn += elems; 
+                        extend_hfa_type(dest, a, h); 
+                        break; 
+                      } 
+                    state.nsrn = N_X_ARG_REG; 
+                    dest = allocate_to_stack(&state, stack, ty->alignment, s); 
+                  } 
+                else 
+                  { 
+#endif /* for handling armasm calling convention */ 
+	        if (state.nsrn + elems <= N_V_ARG_REG) 
+		  { 
+		    dest = &context->v[state.nsrn]; 
+		    state.nsrn += elems; 
+		    extend_hfa_type (dest, a, h); 
+		    break; 
+		  } 
+		state.nsrn = N_V_ARG_REG; 
+		dest = allocate_to_stack (&state, stack, ty->alignment, s); 
+#ifdef _M_ARM64 /* for handling armasm calling convention */ 
+	      } 
+#endif /* for handling armasm calling convention */ 
+	      } 
+	    else if (s > 16) 
+	      { 
+		/* If the argument is a composite type that is larger than 16 
+		   bytes, then the argument has been copied to memory, and 
+		   the argument is replaced by a pointer to the copy.  */ 
+		a = &avalue[i]; 
+		t = FFI_TYPE_POINTER; 
+		s = sizeof (void *); 
+		goto do_pointer; 
+	      } 
+	    else 
+	      { 
+		size_t n = (s + 7) / 8; 
+		if (state.ngrn + n <= N_X_ARG_REG) 
+		  { 
+		    /* If the argument is a composite type and the size in 
+		       double-words is not more than the number of available 
+		       X registers, then the argument is copied into 
+		       consecutive X registers.  */ 
+		    dest = &context->x[state.ngrn]; 
+                    state.ngrn += (unsigned int)n; 
+		  } 
+		else 
+		  { 
+		    /* Otherwise, there are insufficient X registers. Further 
+		       X register allocations are prevented, the NSAA is 
+		       adjusted and the argument is copied to memory at the 
+		       adjusted NSAA.  */ 
+		    state.ngrn = N_X_ARG_REG; 
+		    dest = allocate_to_stack (&state, stack, ty->alignment, s); 
+		  } 
+		} 
+	      memcpy (dest, a, s); 
+	    } 
+	  break; 
+ 
+	default: 
+	  abort(); 
+	} 
+ 
+#if defined (__APPLE__) 
+      if (i + 1 == cif->aarch64_nfixedargs) 
+	{ 
+	  state.ngrn = N_X_ARG_REG; 
+	  state.nsrn = N_V_ARG_REG; 
+	  state.allocating_variadic = 1; 
+	} 
+#endif 
+    } 
+ 
+  ffi_call_SYSV (context, frame, fn, rvalue, flags, closure); 
+ 
+  if (flags & AARCH64_RET_NEED_COPY) 
+    memcpy (orig_rvalue, rvalue, rtype_size); 
+} 
+ 
+void 
+ffi_call (ffi_cif *cif, void (*fn) (void), void *rvalue, void **avalue) 
+{ 
+  ffi_call_int (cif, fn, rvalue, avalue, NULL); 
+} 
+ 
+#ifdef FFI_GO_CLOSURES 
+void 
+ffi_call_go (ffi_cif *cif, void (*fn) (void), void *rvalue, 
+	     void **avalue, void *closure) 
+{ 
+  ffi_call_int (cif, fn, rvalue, avalue, closure); 
+} 
+#endif /* FFI_GO_CLOSURES */ 
+ 
+/* Build a trampoline.  */ 
+ 
+extern void ffi_closure_SYSV (void) FFI_HIDDEN; 
+extern void ffi_closure_SYSV_V (void) FFI_HIDDEN; 
+ 
+ffi_status 
+ffi_prep_closure_loc (ffi_closure *closure, 
+                      ffi_cif* cif, 
+                      void (*fun)(ffi_cif*,void*,void**,void*), 
+                      void *user_data, 
+                      void *codeloc) 
+{ 
+  if (cif->abi != FFI_SYSV) 
+    return FFI_BAD_ABI; 
+ 
+  void (*start)(void); 
+   
+  if (cif->flags & AARCH64_FLAG_ARG_V) 
+    start = ffi_closure_SYSV_V; 
+  else 
+    start = ffi_closure_SYSV; 
+ 
+#if FFI_EXEC_TRAMPOLINE_TABLE 
+#ifdef __MACH__ 
+  void **config = (void **)((uint8_t *)codeloc - PAGE_MAX_SIZE); 
+  config[0] = closure; 
+  config[1] = start; 
+#endif 
+#else 
+  static const unsigned char trampoline[16] = { 
+    0x90, 0x00, 0x00, 0x58,	/* ldr	x16, tramp+16	*/ 
+    0xf1, 0xff, 0xff, 0x10,	/* adr	x17, tramp+0	*/ 
+    0x00, 0x02, 0x1f, 0xd6	/* br	x16		*/ 
+  }; 
+  char *tramp = closure->tramp; 
+   
+  memcpy (tramp, trampoline, sizeof(trampoline)); 
+   
+  *(UINT64 *)(tramp + 16) = (uintptr_t)start; 
+ 
+  ffi_clear_cache(tramp, tramp + FFI_TRAMPOLINE_SIZE); 
+ 
+  /* Also flush the cache for code mapping.  */ 
+#ifdef _M_ARM64 
+  // Not using dlmalloc.c for Windows ARM64 builds 
+  // so calling ffi_data_to_code_pointer() isn't necessary 
+  unsigned char *tramp_code = tramp; 
+  #else 
+  unsigned char *tramp_code = ffi_data_to_code_pointer (tramp); 
+  #endif 
+  ffi_clear_cache (tramp_code, tramp_code + FFI_TRAMPOLINE_SIZE); 
+#endif 
+ 
+  closure->cif = cif; 
+  closure->fun = fun; 
+  closure->user_data = user_data; 
+ 
+  return FFI_OK; 
+} 
+ 
+#ifdef FFI_GO_CLOSURES 
+extern void ffi_go_closure_SYSV (void) FFI_HIDDEN; 
+extern void ffi_go_closure_SYSV_V (void) FFI_HIDDEN; 
+ 
+ffi_status 
+ffi_prep_go_closure (ffi_go_closure *closure, ffi_cif* cif, 
+                     void (*fun)(ffi_cif*,void*,void**,void*)) 
+{ 
+  void (*start)(void); 
+ 
+  if (cif->abi != FFI_SYSV) 
+    return FFI_BAD_ABI; 
+ 
+  if (cif->flags & AARCH64_FLAG_ARG_V) 
+    start = ffi_go_closure_SYSV_V; 
+  else 
+    start = ffi_go_closure_SYSV; 
+ 
+  closure->tramp = start; 
+  closure->cif = cif; 
+  closure->fun = fun; 
+ 
+  return FFI_OK; 
+} 
+#endif /* FFI_GO_CLOSURES */ 
+ 
+/* Primary handler to setup and invoke a function within a closure. 
+ 
+   A closure when invoked enters via the assembler wrapper 
+   ffi_closure_SYSV(). The wrapper allocates a call context on the 
+   stack, saves the interesting registers (from the perspective of 
+   the calling convention) into the context then passes control to 
+   ffi_closure_SYSV_inner() passing the saved context and a pointer to 
+   the stack at the point ffi_closure_SYSV() was invoked. 
+ 
+   On the return path the assembler wrapper will reload call context 
+   registers. 
+ 
+   ffi_closure_SYSV_inner() marshalls the call context into ffi value 
+   descriptors, invokes the wrapped function, then marshalls the return 
+   value back into the call context.  */ 
+ 
+int FFI_HIDDEN 
+ffi_closure_SYSV_inner (ffi_cif *cif, 
+			void (*fun)(ffi_cif*,void*,void**,void*), 
+			void *user_data, 
+			struct call_context *context, 
+			void *stack, void *rvalue, void *struct_rvalue) 
+{ 
+  void **avalue = (void**) alloca (cif->nargs * sizeof (void*)); 
+  int i, h, nargs, flags; 
+  struct arg_state state; 
+ 
+  arg_init (&state); 
+ 
+  for (i = 0, nargs = cif->nargs; i < nargs; i++) 
+    { 
+      ffi_type *ty = cif->arg_types[i]; 
+      int t = ty->type; 
+      size_t n, s = ty->size; 
+ 
+      switch (t) 
+	{ 
+	case FFI_TYPE_VOID: 
+	  FFI_ASSERT (0); 
+	  break; 
+ 
+	case FFI_TYPE_INT: 
+	case FFI_TYPE_UINT8: 
+	case FFI_TYPE_SINT8: 
+	case FFI_TYPE_UINT16: 
+	case FFI_TYPE_SINT16: 
+	case FFI_TYPE_UINT32: 
+	case FFI_TYPE_SINT32: 
+	case FFI_TYPE_UINT64: 
+	case FFI_TYPE_SINT64: 
+	case FFI_TYPE_POINTER: 
+	  avalue[i] = allocate_int_to_reg_or_stack (context, &state, stack, s); 
+	  break; 
+ 
+	case FFI_TYPE_FLOAT: 
+	case FFI_TYPE_DOUBLE: 
+	case FFI_TYPE_LONGDOUBLE: 
+	case FFI_TYPE_STRUCT: 
+	case FFI_TYPE_COMPLEX: 
+	  h = is_vfp_type (ty); 
+	  if (h) 
+	    { 
+	      n = 4 - (h & 3); 
+#ifdef _M_ARM64  /* for handling armasm calling convention */ 
+              if (cif->is_variadic) 
+                { 
+                  if (state.ngrn + n <= N_X_ARG_REG) 
+                    { 
+                      void *reg = &context->x[state.ngrn]; 
+                      state.ngrn += (unsigned int)n; 
+     
+                      /* Eeek! We need a pointer to the structure, however the 
+                       homogeneous float elements are being passed in individual 
+                       registers, therefore for float and double the structure 
+                       is not represented as a contiguous sequence of bytes in 
+                       our saved register context.  We don't need the original 
+                       contents of the register storage, so we reformat the 
+                       structure into the same memory.  */ 
+                      avalue[i] = compress_hfa_type(reg, reg, h); 
+                    } 
+                  else 
+                    { 
+                      state.ngrn = N_X_ARG_REG; 
+                      state.nsrn = N_V_ARG_REG; 
+                      avalue[i] = allocate_to_stack(&state, stack, 
+                             ty->alignment, s); 
+                    } 
+                } 
+              else 
+                { 
+#endif  /* for handling armasm calling convention */ 
+                  if (state.nsrn + n <= N_V_ARG_REG) 
+                    { 
+                      void *reg = &context->v[state.nsrn]; 
+                      state.nsrn += (unsigned int)n; 
+                      avalue[i] = compress_hfa_type(reg, reg, h); 
+                    } 
+                  else 
+                    { 
+                      state.nsrn = N_V_ARG_REG; 
+                      avalue[i] = allocate_to_stack(&state, stack, 
+                                                   ty->alignment, s); 
+                    } 
+#ifdef _M_ARM64  /* for handling armasm calling convention */ 
+                } 
+#endif  /* for handling armasm calling convention */ 
+            } 
+          else if (s > 16) 
+            { 
+              /* Replace Composite type of size greater than 16 with a 
+                  pointer.  */ 
+              avalue[i] = *(void **) 
+              allocate_int_to_reg_or_stack (context, &state, stack, 
+                                         sizeof (void *)); 
+            } 
+          else 
+            { 
+              n = (s + 7) / 8; 
+              if (state.ngrn + n <= N_X_ARG_REG) 
+                { 
+                  avalue[i] = &context->x[state.ngrn]; 
+                  state.ngrn += (unsigned int)n; 
+                } 
+              else 
+                { 
+                  state.ngrn = N_X_ARG_REG; 
+                  avalue[i] = allocate_to_stack(&state, stack, 
+                                           ty->alignment, s); 
+                } 
+            } 
+          break; 
+ 
+        default: 
+          abort(); 
+      } 
+ 
+#if defined (__APPLE__) 
+      if (i + 1 == cif->aarch64_nfixedargs) 
+	{ 
+	  state.ngrn = N_X_ARG_REG; 
+	  state.nsrn = N_V_ARG_REG; 
+	  state.allocating_variadic = 1; 
+	} 
+#endif 
+    } 
+ 
+  flags = cif->flags; 
+  if (flags & AARCH64_RET_IN_MEM) 
+    rvalue = struct_rvalue; 
+ 
+  fun (cif, rvalue, avalue, user_data); 
+ 
+  return flags; 
+} 
+ 
+#endif /* (__aarch64__) || defined(__arm64__)|| defined (_M_ARM64)*/ 
diff --git a/contrib/restricted/libffi/src/aarch64/ffitarget.h b/contrib/restricted/libffi/src/aarch64/ffitarget.h
index ecb6d2deae..ddce8f21eb 100644
--- a/contrib/restricted/libffi/src/aarch64/ffitarget.h
+++ b/contrib/restricted/libffi/src/aarch64/ffitarget.h
@@ -1,92 +1,92 @@
-/* Copyright (c) 2009, 2010, 2011, 2012 ARM Ltd.
-
-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 THE AUTHORS OR COPYRIGHT HOLDERS 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.  */
-
-#ifndef LIBFFI_TARGET_H
-#define LIBFFI_TARGET_H
-
-#ifndef LIBFFI_H
-#error "Please do not include ffitarget.h directly into your source.  Use ffi.h instead."
-#endif
-
-#ifndef LIBFFI_ASM
-#ifdef __ILP32__
-#define FFI_SIZEOF_ARG 8
-#define FFI_SIZEOF_JAVA_RAW  4
-typedef unsigned long long ffi_arg;
-typedef signed long long ffi_sarg;
-#elif defined(_M_ARM64)
-#define FFI_SIZEOF_ARG 8
-typedef unsigned long long ffi_arg;
-typedef signed long long ffi_sarg;
-#else
-typedef unsigned long ffi_arg;
-typedef signed long ffi_sarg;
-#endif
-
-typedef enum ffi_abi
-  {
-    FFI_FIRST_ABI = 0,
-    FFI_SYSV,
-    FFI_LAST_ABI,
-    FFI_DEFAULT_ABI = FFI_SYSV
-  } ffi_abi;
-#endif
-
-/* ---- Definitions for closures ----------------------------------------- */
-
-#define FFI_CLOSURES 1
-#define FFI_NATIVE_RAW_API 0
-
-#if defined (FFI_EXEC_TRAMPOLINE_TABLE) && FFI_EXEC_TRAMPOLINE_TABLE
-
-#ifdef __MACH__
-#define FFI_TRAMPOLINE_SIZE 16
-#define FFI_TRAMPOLINE_CLOSURE_OFFSET 16
-#else
-#error "No trampoline table implementation"
-#endif
-
-#else
-#define FFI_TRAMPOLINE_SIZE 24
-#define FFI_TRAMPOLINE_CLOSURE_OFFSET FFI_TRAMPOLINE_SIZE
-#endif
-
-#ifdef _M_ARM64
-#define FFI_EXTRA_CIF_FIELDS unsigned is_variadic
-#endif
-
-/* ---- Internal ---- */
-
-#if defined (__APPLE__)
-#define FFI_TARGET_SPECIFIC_VARIADIC
-#define FFI_EXTRA_CIF_FIELDS unsigned aarch64_nfixedargs
-#elif !defined(_M_ARM64)
-/* iOS and Windows reserve x18 for the system.  Disable Go closures until
-   a new static chain is chosen.  */
-#define FFI_GO_CLOSURES 1
-#endif
-
-#ifndef _M_ARM64
-/* No complex type on Windows */
-#define FFI_TARGET_HAS_COMPLEX_TYPE
-#endif
-
-#endif
+/* Copyright (c) 2009, 2010, 2011, 2012 ARM Ltd. 
+ 
+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 THE AUTHORS OR COPYRIGHT HOLDERS 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.  */ 
+ 
+#ifndef LIBFFI_TARGET_H 
+#define LIBFFI_TARGET_H 
+ 
+#ifndef LIBFFI_H 
+#error "Please do not include ffitarget.h directly into your source.  Use ffi.h instead." 
+#endif 
+ 
+#ifndef LIBFFI_ASM 
+#ifdef __ILP32__ 
+#define FFI_SIZEOF_ARG 8 
+#define FFI_SIZEOF_JAVA_RAW  4 
+typedef unsigned long long ffi_arg; 
+typedef signed long long ffi_sarg; 
+#elif defined(_M_ARM64) 
+#define FFI_SIZEOF_ARG 8 
+typedef unsigned long long ffi_arg; 
+typedef signed long long ffi_sarg; 
+#else 
+typedef unsigned long ffi_arg; 
+typedef signed long ffi_sarg; 
+#endif 
+ 
+typedef enum ffi_abi 
+  { 
+    FFI_FIRST_ABI = 0, 
+    FFI_SYSV, 
+    FFI_LAST_ABI, 
+    FFI_DEFAULT_ABI = FFI_SYSV 
+  } ffi_abi; 
+#endif 
+ 
+/* ---- Definitions for closures ----------------------------------------- */ 
+ 
+#define FFI_CLOSURES 1 
+#define FFI_NATIVE_RAW_API 0 
+ 
+#if defined (FFI_EXEC_TRAMPOLINE_TABLE) && FFI_EXEC_TRAMPOLINE_TABLE 
+ 
+#ifdef __MACH__ 
+#define FFI_TRAMPOLINE_SIZE 16 
+#define FFI_TRAMPOLINE_CLOSURE_OFFSET 16 
+#else 
+#error "No trampoline table implementation" 
+#endif 
+ 
+#else 
+#define FFI_TRAMPOLINE_SIZE 24 
+#define FFI_TRAMPOLINE_CLOSURE_OFFSET FFI_TRAMPOLINE_SIZE 
+#endif 
+ 
+#ifdef _M_ARM64 
+#define FFI_EXTRA_CIF_FIELDS unsigned is_variadic 
+#endif 
+ 
+/* ---- Internal ---- */ 
+ 
+#if defined (__APPLE__) 
+#define FFI_TARGET_SPECIFIC_VARIADIC 
+#define FFI_EXTRA_CIF_FIELDS unsigned aarch64_nfixedargs 
+#elif !defined(_M_ARM64) 
+/* iOS and Windows reserve x18 for the system.  Disable Go closures until 
+   a new static chain is chosen.  */ 
+#define FFI_GO_CLOSURES 1 
+#endif 
+ 
+#ifndef _M_ARM64 
+/* No complex type on Windows */ 
+#define FFI_TARGET_HAS_COMPLEX_TYPE 
+#endif 
+ 
+#endif 
diff --git a/contrib/restricted/libffi/src/aarch64/internal.h b/contrib/restricted/libffi/src/aarch64/internal.h
index 9c3e07725a..2691dafa98 100644
--- a/contrib/restricted/libffi/src/aarch64/internal.h
+++ b/contrib/restricted/libffi/src/aarch64/internal.h
@@ -1,67 +1,67 @@
-/* 
-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 THE AUTHORS OR COPYRIGHT HOLDERS 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.  */
-
-#define AARCH64_RET_VOID	0
-#define AARCH64_RET_INT64	1
-#define AARCH64_RET_INT128	2
-
-#define AARCH64_RET_UNUSED3	3
-#define AARCH64_RET_UNUSED4	4
-#define AARCH64_RET_UNUSED5	5
-#define AARCH64_RET_UNUSED6	6
-#define AARCH64_RET_UNUSED7	7
-
-/* Note that FFI_TYPE_FLOAT == 2, _DOUBLE == 3, _LONGDOUBLE == 4,
-   so _S4 through _Q1 are layed out as (TYPE * 4) + (4 - COUNT).  */
-#define AARCH64_RET_S4		8
-#define AARCH64_RET_S3		9
-#define AARCH64_RET_S2		10
-#define AARCH64_RET_S1		11
-
-#define AARCH64_RET_D4		12
-#define AARCH64_RET_D3		13
-#define AARCH64_RET_D2		14
-#define AARCH64_RET_D1		15
-
-#define AARCH64_RET_Q4		16
-#define AARCH64_RET_Q3		17
-#define AARCH64_RET_Q2		18
-#define AARCH64_RET_Q1		19
-
-/* Note that each of the sub-64-bit integers gets two entries.  */
-#define AARCH64_RET_UINT8	20
-#define AARCH64_RET_UINT16	22
-#define AARCH64_RET_UINT32	24
-
-#define AARCH64_RET_SINT8	26
-#define AARCH64_RET_SINT16	28
-#define AARCH64_RET_SINT32	30
-
-#define AARCH64_RET_MASK	31
-
-#define AARCH64_RET_IN_MEM	(1 << 5)
-#define AARCH64_RET_NEED_COPY	(1 << 6)
-
-#define AARCH64_FLAG_ARG_V_BIT	7
-#define AARCH64_FLAG_ARG_V	(1 << AARCH64_FLAG_ARG_V_BIT)
-
-#define N_X_ARG_REG		8
-#define N_V_ARG_REG		8
-#define CALL_CONTEXT_SIZE	(N_V_ARG_REG * 16 + N_X_ARG_REG * 8)
+/*  
+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 THE AUTHORS OR COPYRIGHT HOLDERS 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.  */ 
+ 
+#define AARCH64_RET_VOID	0 
+#define AARCH64_RET_INT64	1 
+#define AARCH64_RET_INT128	2 
+ 
+#define AARCH64_RET_UNUSED3	3 
+#define AARCH64_RET_UNUSED4	4 
+#define AARCH64_RET_UNUSED5	5 
+#define AARCH64_RET_UNUSED6	6 
+#define AARCH64_RET_UNUSED7	7 
+ 
+/* Note that FFI_TYPE_FLOAT == 2, _DOUBLE == 3, _LONGDOUBLE == 4, 
+   so _S4 through _Q1 are layed out as (TYPE * 4) + (4 - COUNT).  */ 
+#define AARCH64_RET_S4		8 
+#define AARCH64_RET_S3		9 
+#define AARCH64_RET_S2		10 
+#define AARCH64_RET_S1		11 
+ 
+#define AARCH64_RET_D4		12 
+#define AARCH64_RET_D3		13 
+#define AARCH64_RET_D2		14 
+#define AARCH64_RET_D1		15 
+ 
+#define AARCH64_RET_Q4		16 
+#define AARCH64_RET_Q3		17 
+#define AARCH64_RET_Q2		18 
+#define AARCH64_RET_Q1		19 
+ 
+/* Note that each of the sub-64-bit integers gets two entries.  */ 
+#define AARCH64_RET_UINT8	20 
+#define AARCH64_RET_UINT16	22 
+#define AARCH64_RET_UINT32	24 
+ 
+#define AARCH64_RET_SINT8	26 
+#define AARCH64_RET_SINT16	28 
+#define AARCH64_RET_SINT32	30 
+ 
+#define AARCH64_RET_MASK	31 
+ 
+#define AARCH64_RET_IN_MEM	(1 << 5) 
+#define AARCH64_RET_NEED_COPY	(1 << 6) 
+ 
+#define AARCH64_FLAG_ARG_V_BIT	7 
+#define AARCH64_FLAG_ARG_V	(1 << AARCH64_FLAG_ARG_V_BIT) 
+ 
+#define N_X_ARG_REG		8 
+#define N_V_ARG_REG		8 
+#define CALL_CONTEXT_SIZE	(N_V_ARG_REG * 16 + N_X_ARG_REG * 8) 
diff --git a/contrib/restricted/libffi/src/aarch64/sysv.S b/contrib/restricted/libffi/src/aarch64/sysv.S
index 6761ee1ea9..4d8d85139a 100644
--- a/contrib/restricted/libffi/src/aarch64/sysv.S
+++ b/contrib/restricted/libffi/src/aarch64/sysv.S
@@ -1,440 +1,440 @@
-/* Copyright (c) 2009, 2010, 2011, 2012 ARM Ltd.
-
-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 THE AUTHORS OR COPYRIGHT HOLDERS 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.  */
-
-#if defined(__aarch64__) || defined(__arm64__)
-#define LIBFFI_ASM
-#include <fficonfig.h>
-#include <ffi.h>
-#include <ffi_cfi.h>
-#include "internal.h"
-
-#ifdef HAVE_MACHINE_ASM_H
-#include <machine/asm.h>
-#else
-#ifdef __USER_LABEL_PREFIX__
-#define CONCAT1(a, b) CONCAT2(a, b)
-#define CONCAT2(a, b) a ## b
-
-/* Use the right prefix for global labels.  */
-#define CNAME(x) CONCAT1 (__USER_LABEL_PREFIX__, x)
-#else
-#define CNAME(x) x
-#endif
-#endif
-
-#ifdef __AARCH64EB__
-# define BE(X)	X
-#else
-# define BE(X)	0
-#endif
-
-#ifdef __ILP32__
-#define PTR_REG(n)      w##n
-#else
-#define PTR_REG(n)      x##n
-#endif
-
-#ifdef __ILP32__
-#define PTR_SIZE	4
-#else
-#define PTR_SIZE	8
-#endif
-
-	.text
-	.align 4
-
-/* ffi_call_SYSV
-   extern void ffi_call_SYSV (void *stack, void *frame,
-			      void (*fn)(void), void *rvalue,
-			      int flags, void *closure);
-
-   Therefore on entry we have:
-
-   x0 stack
-   x1 frame
-   x2 fn
-   x3 rvalue
-   x4 flags
-   x5 closure
-*/
-
-	cfi_startproc
-CNAME(ffi_call_SYSV):
-	/* Use a stack frame allocated by our caller.  */
-	cfi_def_cfa(x1, 32);
-	stp	x29, x30, [x1]
-	mov	x29, x1
-	mov	sp, x0
-	cfi_def_cfa_register(x29)
-	cfi_rel_offset (x29, 0)
-	cfi_rel_offset (x30, 8)
-
-	mov	x9, x2			/* save fn */
-	mov	x8, x3			/* install structure return */
-#ifdef FFI_GO_CLOSURES
-	mov	x18, x5			/* install static chain */
-#endif
-	stp	x3, x4, [x29, #16]	/* save rvalue and flags */
-
-	/* Load the vector argument passing registers, if necessary.  */
-	tbz	w4, #AARCH64_FLAG_ARG_V_BIT, 1f
-	ldp     q0, q1, [sp, #0]
-	ldp     q2, q3, [sp, #32]
-	ldp     q4, q5, [sp, #64]
-	ldp     q6, q7, [sp, #96]
-1:
-	/* Load the core argument passing registers, including
-	   the structure return pointer.  */
-	ldp     x0, x1, [sp, #16*N_V_ARG_REG + 0]
-	ldp     x2, x3, [sp, #16*N_V_ARG_REG + 16]
-	ldp     x4, x5, [sp, #16*N_V_ARG_REG + 32]
-	ldp     x6, x7, [sp, #16*N_V_ARG_REG + 48]
-
-	/* Deallocate the context, leaving the stacked arguments.  */
-	add	sp, sp, #CALL_CONTEXT_SIZE
-
-	blr     x9			/* call fn */
-
-	ldp	x3, x4, [x29, #16]	/* reload rvalue and flags */
-
-	/* Partially deconstruct the stack frame.  */
-	mov     sp, x29
-	cfi_def_cfa_register (sp)
-	ldp     x29, x30, [x29]
-
-	/* Save the return value as directed.  */
-	adr	x5, 0f
-	and	w4, w4, #AARCH64_RET_MASK
-	add	x5, x5, x4, lsl #3
-	br	x5
-
-	/* Note that each table entry is 2 insns, and thus 8 bytes.
-	   For integer data, note that we're storing into ffi_arg
-	   and therefore we want to extend to 64 bits; these types
-	   have two consecutive entries allocated for them.  */
-	.align	4
-0:	ret				/* VOID */
-	nop
-1:	str	x0, [x3]		/* INT64 */
-	ret
-2:	stp	x0, x1, [x3]		/* INT128 */
-	ret
-3:	brk	#1000			/* UNUSED */
-	ret
-4:	brk	#1000			/* UNUSED */
-	ret
-5:	brk	#1000			/* UNUSED */
-	ret
-6:	brk	#1000			/* UNUSED */
-	ret
-7:	brk	#1000			/* UNUSED */
-	ret
-8:	st4	{ v0.s, v1.s, v2.s, v3.s }[0], [x3]	/* S4 */
-	ret
-9:	st3	{ v0.s, v1.s, v2.s }[0], [x3]	/* S3 */
-	ret
-10:	stp	s0, s1, [x3]		/* S2 */
-	ret
-11:	str	s0, [x3]		/* S1 */
-	ret
-12:	st4	{ v0.d, v1.d, v2.d, v3.d }[0], [x3]	/* D4 */
-	ret
-13:	st3	{ v0.d, v1.d, v2.d }[0], [x3]	/* D3 */
-	ret
-14:	stp	d0, d1, [x3]		/* D2 */
-	ret
-15:	str	d0, [x3]		/* D1 */
-	ret
-16:	str	q3, [x3, #48]		/* Q4 */
-	nop
-17:	str	q2, [x3, #32]		/* Q3 */
-	nop
-18:	stp	q0, q1, [x3]		/* Q2 */
-	ret
-19:	str	q0, [x3]		/* Q1 */
-	ret
-20:	uxtb	w0, w0			/* UINT8 */
-	str	x0, [x3]
-21:	ret				/* reserved */
-	nop
-22:	uxth	w0, w0			/* UINT16 */
-	str	x0, [x3]
-23:	ret				/* reserved */
-	nop
-24:	mov	w0, w0			/* UINT32 */
-	str	x0, [x3]
-25:	ret				/* reserved */
-	nop
-26:	sxtb	x0, w0			/* SINT8 */
-	str	x0, [x3]
-27:	ret				/* reserved */
-	nop
-28:	sxth	x0, w0			/* SINT16 */
-	str	x0, [x3]
-29:	ret				/* reserved */
-	nop
-30:	sxtw	x0, w0			/* SINT32 */
-	str	x0, [x3]
-31:	ret				/* reserved */
-	nop
-
-	cfi_endproc
-
-	.globl	CNAME(ffi_call_SYSV)
-	FFI_HIDDEN(CNAME(ffi_call_SYSV))
-#ifdef __ELF__
-	.type	CNAME(ffi_call_SYSV), #function
-	.size CNAME(ffi_call_SYSV), .-CNAME(ffi_call_SYSV)
-#endif
-
-/* ffi_closure_SYSV
-
-   Closure invocation glue. This is the low level code invoked directly by
-   the closure trampoline to setup and call a closure.
-
-   On entry x17 points to a struct ffi_closure, x16 has been clobbered
-   all other registers are preserved.
-
-   We allocate a call context and save the argument passing registers,
-   then invoked the generic C ffi_closure_SYSV_inner() function to do all
-   the real work, on return we load the result passing registers back from
-   the call context.
-*/
-
-#define ffi_closure_SYSV_FS (8*2 + CALL_CONTEXT_SIZE + 64)
-
-	.align 4
-CNAME(ffi_closure_SYSV_V):
-	cfi_startproc
-	stp     x29, x30, [sp, #-ffi_closure_SYSV_FS]!
-	cfi_adjust_cfa_offset (ffi_closure_SYSV_FS)
-	cfi_rel_offset (x29, 0)
-	cfi_rel_offset (x30, 8)
-
-	/* Save the argument passing vector registers.  */
-	stp     q0, q1, [sp, #16 + 0]
-	stp     q2, q3, [sp, #16 + 32]
-	stp     q4, q5, [sp, #16 + 64]
-	stp     q6, q7, [sp, #16 + 96]
-	b	0f
-	cfi_endproc
-
-	.globl	CNAME(ffi_closure_SYSV_V)
-	FFI_HIDDEN(CNAME(ffi_closure_SYSV_V))
-#ifdef __ELF__
-	.type	CNAME(ffi_closure_SYSV_V), #function
-	.size	CNAME(ffi_closure_SYSV_V), . - CNAME(ffi_closure_SYSV_V)
-#endif
-
-	.align	4
-	cfi_startproc
-CNAME(ffi_closure_SYSV):
-	stp     x29, x30, [sp, #-ffi_closure_SYSV_FS]!
-	cfi_adjust_cfa_offset (ffi_closure_SYSV_FS)
-	cfi_rel_offset (x29, 0)
-	cfi_rel_offset (x30, 8)
-0:
-	mov     x29, sp
-
-	/* Save the argument passing core registers.  */
-	stp     x0, x1, [sp, #16 + 16*N_V_ARG_REG + 0]
-	stp     x2, x3, [sp, #16 + 16*N_V_ARG_REG + 16]
-	stp     x4, x5, [sp, #16 + 16*N_V_ARG_REG + 32]
-	stp     x6, x7, [sp, #16 + 16*N_V_ARG_REG + 48]
-
-	/* Load ffi_closure_inner arguments.  */
-	ldp	PTR_REG(0), PTR_REG(1), [x17, #FFI_TRAMPOLINE_CLOSURE_OFFSET]	/* load cif, fn */
-	ldr	PTR_REG(2), [x17, #FFI_TRAMPOLINE_CLOSURE_OFFSET+PTR_SIZE*2]	/* load user_data */
-.Ldo_closure:
-	add	x3, sp, #16				/* load context */
-	add	x4, sp, #ffi_closure_SYSV_FS		/* load stack */
-	add	x5, sp, #16+CALL_CONTEXT_SIZE		/* load rvalue */
-	mov	x6, x8					/* load struct_rval */
-	bl      CNAME(ffi_closure_SYSV_inner)
-
-	/* Load the return value as directed.  */
-	adr	x1, 0f
-	and	w0, w0, #AARCH64_RET_MASK
-	add	x1, x1, x0, lsl #3
-	add	x3, sp, #16+CALL_CONTEXT_SIZE
-	br	x1
-
-	/* Note that each table entry is 2 insns, and thus 8 bytes.  */
-	.align	4
-0:	b	99f			/* VOID */
-	nop
-1:	ldr	x0, [x3]		/* INT64 */
-	b	99f
-2:	ldp	x0, x1, [x3]		/* INT128 */
-	b	99f
-3:	brk	#1000			/* UNUSED */
-	nop
-4:	brk	#1000			/* UNUSED */
-	nop
-5:	brk	#1000			/* UNUSED */
-	nop
-6:	brk	#1000			/* UNUSED */
-	nop
-7:	brk	#1000			/* UNUSED */
-	nop
-8:	ldr	s3, [x3, #12]		/* S4 */
-	nop
-9:	ldr	s2, [x3, #8]		/* S3 */
-	nop
-10:	ldp	s0, s1, [x3]		/* S2 */
-	b	99f
-11:	ldr	s0, [x3]		/* S1 */
-	b	99f
-12:	ldr	d3, [x3, #24]		/* D4 */
-	nop
-13:	ldr	d2, [x3, #16]		/* D3 */
-	nop
-14:	ldp	d0, d1, [x3]		/* D2 */
-	b	99f
-15:	ldr	d0, [x3]		/* D1 */
-	b	99f
-16:	ldr	q3, [x3, #48]		/* Q4 */
-	nop
-17:	ldr	q2, [x3, #32]		/* Q3 */
-	nop
-18:	ldp	q0, q1, [x3]		/* Q2 */
-	b	99f
-19:	ldr	q0, [x3]		/* Q1 */
-	b	99f
-20:	ldrb	w0, [x3, #BE(7)]	/* UINT8 */
-	b	99f
-21:	brk	#1000			/* reserved */
-	nop
-22:	ldrh	w0, [x3, #BE(6)]	/* UINT16 */
-	b	99f
-23:	brk	#1000			/* reserved */
-	nop
-24:	ldr	w0, [x3, #BE(4)]	/* UINT32 */
-	b	99f
-25:	brk	#1000			/* reserved */
-	nop
-26:	ldrsb	x0, [x3, #BE(7)]	/* SINT8 */
-	b	99f
-27:	brk	#1000			/* reserved */
-	nop
-28:	ldrsh	x0, [x3, #BE(6)]	/* SINT16 */
-	b	99f
-29:	brk	#1000			/* reserved */
-	nop
-30:	ldrsw	x0, [x3, #BE(4)]	/* SINT32 */
-	nop
-31:					/* reserved */
-99:	ldp     x29, x30, [sp], #ffi_closure_SYSV_FS
-	cfi_adjust_cfa_offset (-ffi_closure_SYSV_FS)
-	cfi_restore (x29)
-	cfi_restore (x30)
-	ret
-	cfi_endproc
-
-	.globl	CNAME(ffi_closure_SYSV)
-	FFI_HIDDEN(CNAME(ffi_closure_SYSV))
-#ifdef __ELF__
-	.type	CNAME(ffi_closure_SYSV), #function
-	.size	CNAME(ffi_closure_SYSV), . - CNAME(ffi_closure_SYSV)
-#endif
-
-#if FFI_EXEC_TRAMPOLINE_TABLE
-
-#ifdef __MACH__
-#include <mach/machine/vm_param.h>
-    .align PAGE_MAX_SHIFT
-CNAME(ffi_closure_trampoline_table_page):
-    .rept PAGE_MAX_SIZE / FFI_TRAMPOLINE_SIZE
-    adr x16, -PAGE_MAX_SIZE
-    ldp x17, x16, [x16]
-    br x16
-	nop		/* each entry in the trampoline config page is 2*sizeof(void*) so the trampoline itself cannot be smaller that 16 bytes */
-    .endr
-
-    .globl CNAME(ffi_closure_trampoline_table_page)
-    FFI_HIDDEN(CNAME(ffi_closure_trampoline_table_page))
-    #ifdef __ELF__
-    	.type	CNAME(ffi_closure_trampoline_table_page), #function
-    	.size	CNAME(ffi_closure_trampoline_table_page), . - CNAME(ffi_closure_trampoline_table_page)
-    #endif
-#endif
-
-#endif /* FFI_EXEC_TRAMPOLINE_TABLE */
-
-#ifdef FFI_GO_CLOSURES
-	.align 4
-CNAME(ffi_go_closure_SYSV_V):
-	cfi_startproc
-	stp     x29, x30, [sp, #-ffi_closure_SYSV_FS]!
-	cfi_adjust_cfa_offset (ffi_closure_SYSV_FS)
-	cfi_rel_offset (x29, 0)
-	cfi_rel_offset (x30, 8)
-
-	/* Save the argument passing vector registers.  */
-	stp     q0, q1, [sp, #16 + 0]
-	stp     q2, q3, [sp, #16 + 32]
-	stp     q4, q5, [sp, #16 + 64]
-	stp     q6, q7, [sp, #16 + 96]
-	b	0f
-	cfi_endproc
-
-	.globl	CNAME(ffi_go_closure_SYSV_V)
-	FFI_HIDDEN(CNAME(ffi_go_closure_SYSV_V))
-#ifdef __ELF__
-	.type	CNAME(ffi_go_closure_SYSV_V), #function
-	.size	CNAME(ffi_go_closure_SYSV_V), . - CNAME(ffi_go_closure_SYSV_V)
-#endif
-
-	.align	4
-	cfi_startproc
-CNAME(ffi_go_closure_SYSV):
-	stp     x29, x30, [sp, #-ffi_closure_SYSV_FS]!
-	cfi_adjust_cfa_offset (ffi_closure_SYSV_FS)
-	cfi_rel_offset (x29, 0)
-	cfi_rel_offset (x30, 8)
-0:
-	mov     x29, sp
-
-	/* Save the argument passing core registers.  */
-	stp     x0, x1, [sp, #16 + 16*N_V_ARG_REG + 0]
-	stp     x2, x3, [sp, #16 + 16*N_V_ARG_REG + 16]
-	stp     x4, x5, [sp, #16 + 16*N_V_ARG_REG + 32]
-	stp     x6, x7, [sp, #16 + 16*N_V_ARG_REG + 48]
-
-	/* Load ffi_closure_inner arguments.  */
-	ldp	PTR_REG(0), PTR_REG(1), [x18, #PTR_SIZE]/* load cif, fn */
-	mov	x2, x18					/* load user_data */
-	b	.Ldo_closure
-	cfi_endproc
-
-	.globl	CNAME(ffi_go_closure_SYSV)
-	FFI_HIDDEN(CNAME(ffi_go_closure_SYSV))
-#ifdef __ELF__
-	.type	CNAME(ffi_go_closure_SYSV), #function
-	.size	CNAME(ffi_go_closure_SYSV), . - CNAME(ffi_go_closure_SYSV)
-#endif
-#endif /* FFI_GO_CLOSURES */
-#endif /* __arm64__ */
-
-#if defined __ELF__ && defined __linux__
-	.section .note.GNU-stack,"",%progbits
-#endif
-
+/* Copyright (c) 2009, 2010, 2011, 2012 ARM Ltd. 
+ 
+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 THE AUTHORS OR COPYRIGHT HOLDERS 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.  */ 
+ 
+#if defined(__aarch64__) || defined(__arm64__) 
+#define LIBFFI_ASM 
+#include <fficonfig.h> 
+#include <ffi.h> 
+#include <ffi_cfi.h> 
+#include "internal.h" 
+ 
+#ifdef HAVE_MACHINE_ASM_H 
+#include <machine/asm.h> 
+#else 
+#ifdef __USER_LABEL_PREFIX__ 
+#define CONCAT1(a, b) CONCAT2(a, b) 
+#define CONCAT2(a, b) a ## b 
+ 
+/* Use the right prefix for global labels.  */ 
+#define CNAME(x) CONCAT1 (__USER_LABEL_PREFIX__, x) 
+#else 
+#define CNAME(x) x 
+#endif 
+#endif 
+ 
+#ifdef __AARCH64EB__ 
+# define BE(X)	X 
+#else 
+# define BE(X)	0 
+#endif 
+ 
+#ifdef __ILP32__ 
+#define PTR_REG(n)      w##n 
+#else 
+#define PTR_REG(n)      x##n 
+#endif 
+ 
+#ifdef __ILP32__ 
+#define PTR_SIZE	4 
+#else 
+#define PTR_SIZE	8 
+#endif 
+ 
+	.text 
+	.align 4 
+ 
+/* ffi_call_SYSV 
+   extern void ffi_call_SYSV (void *stack, void *frame, 
+			      void (*fn)(void), void *rvalue, 
+			      int flags, void *closure); 
+ 
+   Therefore on entry we have: 
+ 
+   x0 stack 
+   x1 frame 
+   x2 fn 
+   x3 rvalue 
+   x4 flags 
+   x5 closure 
+*/ 
+ 
+	cfi_startproc 
+CNAME(ffi_call_SYSV): 
+	/* Use a stack frame allocated by our caller.  */ 
+	cfi_def_cfa(x1, 32); 
+	stp	x29, x30, [x1] 
+	mov	x29, x1 
+	mov	sp, x0 
+	cfi_def_cfa_register(x29) 
+	cfi_rel_offset (x29, 0) 
+	cfi_rel_offset (x30, 8) 
+ 
+	mov	x9, x2			/* save fn */ 
+	mov	x8, x3			/* install structure return */ 
+#ifdef FFI_GO_CLOSURES 
+	mov	x18, x5			/* install static chain */ 
+#endif 
+	stp	x3, x4, [x29, #16]	/* save rvalue and flags */ 
+ 
+	/* Load the vector argument passing registers, if necessary.  */ 
+	tbz	w4, #AARCH64_FLAG_ARG_V_BIT, 1f 
+	ldp     q0, q1, [sp, #0] 
+	ldp     q2, q3, [sp, #32] 
+	ldp     q4, q5, [sp, #64] 
+	ldp     q6, q7, [sp, #96] 
+1: 
+	/* Load the core argument passing registers, including 
+	   the structure return pointer.  */ 
+	ldp     x0, x1, [sp, #16*N_V_ARG_REG + 0] 
+	ldp     x2, x3, [sp, #16*N_V_ARG_REG + 16] 
+	ldp     x4, x5, [sp, #16*N_V_ARG_REG + 32] 
+	ldp     x6, x7, [sp, #16*N_V_ARG_REG + 48] 
+ 
+	/* Deallocate the context, leaving the stacked arguments.  */ 
+	add	sp, sp, #CALL_CONTEXT_SIZE 
+ 
+	blr     x9			/* call fn */ 
+ 
+	ldp	x3, x4, [x29, #16]	/* reload rvalue and flags */ 
+ 
+	/* Partially deconstruct the stack frame.  */ 
+	mov     sp, x29 
+	cfi_def_cfa_register (sp) 
+	ldp     x29, x30, [x29] 
+ 
+	/* Save the return value as directed.  */ 
+	adr	x5, 0f 
+	and	w4, w4, #AARCH64_RET_MASK 
+	add	x5, x5, x4, lsl #3 
+	br	x5 
+ 
+	/* Note that each table entry is 2 insns, and thus 8 bytes. 
+	   For integer data, note that we're storing into ffi_arg 
+	   and therefore we want to extend to 64 bits; these types 
+	   have two consecutive entries allocated for them.  */ 
+	.align	4 
+0:	ret				/* VOID */ 
+	nop 
+1:	str	x0, [x3]		/* INT64 */ 
+	ret 
+2:	stp	x0, x1, [x3]		/* INT128 */ 
+	ret 
+3:	brk	#1000			/* UNUSED */ 
+	ret 
+4:	brk	#1000			/* UNUSED */ 
+	ret 
+5:	brk	#1000			/* UNUSED */ 
+	ret 
+6:	brk	#1000			/* UNUSED */ 
+	ret 
+7:	brk	#1000			/* UNUSED */ 
+	ret 
+8:	st4	{ v0.s, v1.s, v2.s, v3.s }[0], [x3]	/* S4 */ 
+	ret 
+9:	st3	{ v0.s, v1.s, v2.s }[0], [x3]	/* S3 */ 
+	ret 
+10:	stp	s0, s1, [x3]		/* S2 */ 
+	ret 
+11:	str	s0, [x3]		/* S1 */ 
+	ret 
+12:	st4	{ v0.d, v1.d, v2.d, v3.d }[0], [x3]	/* D4 */ 
+	ret 
+13:	st3	{ v0.d, v1.d, v2.d }[0], [x3]	/* D3 */ 
+	ret 
+14:	stp	d0, d1, [x3]		/* D2 */ 
+	ret 
+15:	str	d0, [x3]		/* D1 */ 
+	ret 
+16:	str	q3, [x3, #48]		/* Q4 */ 
+	nop 
+17:	str	q2, [x3, #32]		/* Q3 */ 
+	nop 
+18:	stp	q0, q1, [x3]		/* Q2 */ 
+	ret 
+19:	str	q0, [x3]		/* Q1 */ 
+	ret 
+20:	uxtb	w0, w0			/* UINT8 */ 
+	str	x0, [x3] 
+21:	ret				/* reserved */ 
+	nop 
+22:	uxth	w0, w0			/* UINT16 */ 
+	str	x0, [x3] 
+23:	ret				/* reserved */ 
+	nop 
+24:	mov	w0, w0			/* UINT32 */ 
+	str	x0, [x3] 
+25:	ret				/* reserved */ 
+	nop 
+26:	sxtb	x0, w0			/* SINT8 */ 
+	str	x0, [x3] 
+27:	ret				/* reserved */ 
+	nop 
+28:	sxth	x0, w0			/* SINT16 */ 
+	str	x0, [x3] 
+29:	ret				/* reserved */ 
+	nop 
+30:	sxtw	x0, w0			/* SINT32 */ 
+	str	x0, [x3] 
+31:	ret				/* reserved */ 
+	nop 
+ 
+	cfi_endproc 
+ 
+	.globl	CNAME(ffi_call_SYSV) 
+	FFI_HIDDEN(CNAME(ffi_call_SYSV)) 
+#ifdef __ELF__ 
+	.type	CNAME(ffi_call_SYSV), #function 
+	.size CNAME(ffi_call_SYSV), .-CNAME(ffi_call_SYSV) 
+#endif 
+ 
+/* ffi_closure_SYSV 
+ 
+   Closure invocation glue. This is the low level code invoked directly by 
+   the closure trampoline to setup and call a closure. 
+ 
+   On entry x17 points to a struct ffi_closure, x16 has been clobbered 
+   all other registers are preserved. 
+ 
+   We allocate a call context and save the argument passing registers, 
+   then invoked the generic C ffi_closure_SYSV_inner() function to do all 
+   the real work, on return we load the result passing registers back from 
+   the call context. 
+*/ 
+ 
+#define ffi_closure_SYSV_FS (8*2 + CALL_CONTEXT_SIZE + 64) 
+ 
+	.align 4 
+CNAME(ffi_closure_SYSV_V): 
+	cfi_startproc 
+	stp     x29, x30, [sp, #-ffi_closure_SYSV_FS]! 
+	cfi_adjust_cfa_offset (ffi_closure_SYSV_FS) 
+	cfi_rel_offset (x29, 0) 
+	cfi_rel_offset (x30, 8) 
+ 
+	/* Save the argument passing vector registers.  */ 
+	stp     q0, q1, [sp, #16 + 0] 
+	stp     q2, q3, [sp, #16 + 32] 
+	stp     q4, q5, [sp, #16 + 64] 
+	stp     q6, q7, [sp, #16 + 96] 
+	b	0f 
+	cfi_endproc 
+ 
+	.globl	CNAME(ffi_closure_SYSV_V) 
+	FFI_HIDDEN(CNAME(ffi_closure_SYSV_V)) 
+#ifdef __ELF__ 
+	.type	CNAME(ffi_closure_SYSV_V), #function 
+	.size	CNAME(ffi_closure_SYSV_V), . - CNAME(ffi_closure_SYSV_V) 
+#endif 
+ 
+	.align	4 
+	cfi_startproc 
+CNAME(ffi_closure_SYSV): 
+	stp     x29, x30, [sp, #-ffi_closure_SYSV_FS]! 
+	cfi_adjust_cfa_offset (ffi_closure_SYSV_FS) 
+	cfi_rel_offset (x29, 0) 
+	cfi_rel_offset (x30, 8) 
+0: 
+	mov     x29, sp 
+ 
+	/* Save the argument passing core registers.  */ 
+	stp     x0, x1, [sp, #16 + 16*N_V_ARG_REG + 0] 
+	stp     x2, x3, [sp, #16 + 16*N_V_ARG_REG + 16] 
+	stp     x4, x5, [sp, #16 + 16*N_V_ARG_REG + 32] 
+	stp     x6, x7, [sp, #16 + 16*N_V_ARG_REG + 48] 
+ 
+	/* Load ffi_closure_inner arguments.  */ 
+	ldp	PTR_REG(0), PTR_REG(1), [x17, #FFI_TRAMPOLINE_CLOSURE_OFFSET]	/* load cif, fn */ 
+	ldr	PTR_REG(2), [x17, #FFI_TRAMPOLINE_CLOSURE_OFFSET+PTR_SIZE*2]	/* load user_data */ 
+.Ldo_closure: 
+	add	x3, sp, #16				/* load context */ 
+	add	x4, sp, #ffi_closure_SYSV_FS		/* load stack */ 
+	add	x5, sp, #16+CALL_CONTEXT_SIZE		/* load rvalue */ 
+	mov	x6, x8					/* load struct_rval */ 
+	bl      CNAME(ffi_closure_SYSV_inner) 
+ 
+	/* Load the return value as directed.  */ 
+	adr	x1, 0f 
+	and	w0, w0, #AARCH64_RET_MASK 
+	add	x1, x1, x0, lsl #3 
+	add	x3, sp, #16+CALL_CONTEXT_SIZE 
+	br	x1 
+ 
+	/* Note that each table entry is 2 insns, and thus 8 bytes.  */ 
+	.align	4 
+0:	b	99f			/* VOID */ 
+	nop 
+1:	ldr	x0, [x3]		/* INT64 */ 
+	b	99f 
+2:	ldp	x0, x1, [x3]		/* INT128 */ 
+	b	99f 
+3:	brk	#1000			/* UNUSED */ 
+	nop 
+4:	brk	#1000			/* UNUSED */ 
+	nop 
+5:	brk	#1000			/* UNUSED */ 
+	nop 
+6:	brk	#1000			/* UNUSED */ 
+	nop 
+7:	brk	#1000			/* UNUSED */ 
+	nop 
+8:	ldr	s3, [x3, #12]		/* S4 */ 
+	nop 
+9:	ldr	s2, [x3, #8]		/* S3 */ 
+	nop 
+10:	ldp	s0, s1, [x3]		/* S2 */ 
+	b	99f 
+11:	ldr	s0, [x3]		/* S1 */ 
+	b	99f 
+12:	ldr	d3, [x3, #24]		/* D4 */ 
+	nop 
+13:	ldr	d2, [x3, #16]		/* D3 */ 
+	nop 
+14:	ldp	d0, d1, [x3]		/* D2 */ 
+	b	99f 
+15:	ldr	d0, [x3]		/* D1 */ 
+	b	99f 
+16:	ldr	q3, [x3, #48]		/* Q4 */ 
+	nop 
+17:	ldr	q2, [x3, #32]		/* Q3 */ 
+	nop 
+18:	ldp	q0, q1, [x3]		/* Q2 */ 
+	b	99f 
+19:	ldr	q0, [x3]		/* Q1 */ 
+	b	99f 
+20:	ldrb	w0, [x3, #BE(7)]	/* UINT8 */ 
+	b	99f 
+21:	brk	#1000			/* reserved */ 
+	nop 
+22:	ldrh	w0, [x3, #BE(6)]	/* UINT16 */ 
+	b	99f 
+23:	brk	#1000			/* reserved */ 
+	nop 
+24:	ldr	w0, [x3, #BE(4)]	/* UINT32 */ 
+	b	99f 
+25:	brk	#1000			/* reserved */ 
+	nop 
+26:	ldrsb	x0, [x3, #BE(7)]	/* SINT8 */ 
+	b	99f 
+27:	brk	#1000			/* reserved */ 
+	nop 
+28:	ldrsh	x0, [x3, #BE(6)]	/* SINT16 */ 
+	b	99f 
+29:	brk	#1000			/* reserved */ 
+	nop 
+30:	ldrsw	x0, [x3, #BE(4)]	/* SINT32 */ 
+	nop 
+31:					/* reserved */ 
+99:	ldp     x29, x30, [sp], #ffi_closure_SYSV_FS 
+	cfi_adjust_cfa_offset (-ffi_closure_SYSV_FS) 
+	cfi_restore (x29) 
+	cfi_restore (x30) 
+	ret 
+	cfi_endproc 
+ 
+	.globl	CNAME(ffi_closure_SYSV) 
+	FFI_HIDDEN(CNAME(ffi_closure_SYSV)) 
+#ifdef __ELF__ 
+	.type	CNAME(ffi_closure_SYSV), #function 
+	.size	CNAME(ffi_closure_SYSV), . - CNAME(ffi_closure_SYSV) 
+#endif 
+ 
+#if FFI_EXEC_TRAMPOLINE_TABLE 
+ 
+#ifdef __MACH__ 
+#include <mach/machine/vm_param.h> 
+    .align PAGE_MAX_SHIFT 
+CNAME(ffi_closure_trampoline_table_page): 
+    .rept PAGE_MAX_SIZE / FFI_TRAMPOLINE_SIZE 
+    adr x16, -PAGE_MAX_SIZE 
+    ldp x17, x16, [x16] 
+    br x16 
+	nop		/* each entry in the trampoline config page is 2*sizeof(void*) so the trampoline itself cannot be smaller that 16 bytes */ 
+    .endr 
+ 
+    .globl CNAME(ffi_closure_trampoline_table_page) 
+    FFI_HIDDEN(CNAME(ffi_closure_trampoline_table_page)) 
+    #ifdef __ELF__ 
+    	.type	CNAME(ffi_closure_trampoline_table_page), #function 
+    	.size	CNAME(ffi_closure_trampoline_table_page), . - CNAME(ffi_closure_trampoline_table_page) 
+    #endif 
+#endif 
+ 
+#endif /* FFI_EXEC_TRAMPOLINE_TABLE */ 
+ 
+#ifdef FFI_GO_CLOSURES 
+	.align 4 
+CNAME(ffi_go_closure_SYSV_V): 
+	cfi_startproc 
+	stp     x29, x30, [sp, #-ffi_closure_SYSV_FS]! 
+	cfi_adjust_cfa_offset (ffi_closure_SYSV_FS) 
+	cfi_rel_offset (x29, 0) 
+	cfi_rel_offset (x30, 8) 
+ 
+	/* Save the argument passing vector registers.  */ 
+	stp     q0, q1, [sp, #16 + 0] 
+	stp     q2, q3, [sp, #16 + 32] 
+	stp     q4, q5, [sp, #16 + 64] 
+	stp     q6, q7, [sp, #16 + 96] 
+	b	0f 
+	cfi_endproc 
+ 
+	.globl	CNAME(ffi_go_closure_SYSV_V) 
+	FFI_HIDDEN(CNAME(ffi_go_closure_SYSV_V)) 
+#ifdef __ELF__ 
+	.type	CNAME(ffi_go_closure_SYSV_V), #function 
+	.size	CNAME(ffi_go_closure_SYSV_V), . - CNAME(ffi_go_closure_SYSV_V) 
+#endif 
+ 
+	.align	4 
+	cfi_startproc 
+CNAME(ffi_go_closure_SYSV): 
+	stp     x29, x30, [sp, #-ffi_closure_SYSV_FS]! 
+	cfi_adjust_cfa_offset (ffi_closure_SYSV_FS) 
+	cfi_rel_offset (x29, 0) 
+	cfi_rel_offset (x30, 8) 
+0: 
+	mov     x29, sp 
+ 
+	/* Save the argument passing core registers.  */ 
+	stp     x0, x1, [sp, #16 + 16*N_V_ARG_REG + 0] 
+	stp     x2, x3, [sp, #16 + 16*N_V_ARG_REG + 16] 
+	stp     x4, x5, [sp, #16 + 16*N_V_ARG_REG + 32] 
+	stp     x6, x7, [sp, #16 + 16*N_V_ARG_REG + 48] 
+ 
+	/* Load ffi_closure_inner arguments.  */ 
+	ldp	PTR_REG(0), PTR_REG(1), [x18, #PTR_SIZE]/* load cif, fn */ 
+	mov	x2, x18					/* load user_data */ 
+	b	.Ldo_closure 
+	cfi_endproc 
+ 
+	.globl	CNAME(ffi_go_closure_SYSV) 
+	FFI_HIDDEN(CNAME(ffi_go_closure_SYSV)) 
+#ifdef __ELF__ 
+	.type	CNAME(ffi_go_closure_SYSV), #function 
+	.size	CNAME(ffi_go_closure_SYSV), . - CNAME(ffi_go_closure_SYSV) 
+#endif 
+#endif /* FFI_GO_CLOSURES */ 
+#endif /* __arm64__ */ 
+ 
+#if defined __ELF__ && defined __linux__ 
+	.section .note.GNU-stack,"",%progbits 
+#endif 
+ 
diff --git a/contrib/restricted/libffi/src/aarch64/win64_armasm.S b/contrib/restricted/libffi/src/aarch64/win64_armasm.S
index a79f8a8aa9..90b95def5c 100644
--- a/contrib/restricted/libffi/src/aarch64/win64_armasm.S
+++ b/contrib/restricted/libffi/src/aarch64/win64_armasm.S
@@ -1,506 +1,506 @@
-/* Copyright (c) 2009, 2010, 2011, 2012 ARM Ltd.
-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 THE AUTHORS OR COPYRIGHT HOLDERS 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.  */
-
-#define LIBFFI_ASM
-#include <fficonfig.h>
-#include <ffi.h>
-#include <ffi_cfi.h>
-#include "internal.h"
-
-	OPT	2 /*disable listing */
-/* For some macros to add unwind information */
-#include "ksarm64.h"
-	OPT	1 /*re-enable listing */
-
-#define BE(X)	0
-#define PTR_REG(n)      x##n
-#define PTR_SIZE	8
-
-	IMPORT ffi_closure_SYSV_inner
-	EXPORT	ffi_call_SYSV
-	EXPORT	ffi_closure_SYSV_V
-	EXPORT	ffi_closure_SYSV
-	EXPORT	extend_hfa_type
-	EXPORT	compress_hfa_type
-#ifdef FFI_GO_CLOSURES
-	EXPORT	ffi_go_closure_SYSV_V
-	EXPORT	ffi_go_closure_SYSV
-#endif
-
-	TEXTAREA, ALLIGN=8
-
-/* ffi_call_SYSV
-   extern void ffi_call_SYSV (void *stack, void *frame,
-			      void (*fn)(void), void *rvalue,
-			      int flags, void *closure);
-   Therefore on entry we have:
-   x0 stack
-   x1 frame
-   x2 fn
-   x3 rvalue
-   x4 flags
-   x5 closure
-*/
-
-	NESTED_ENTRY ffi_call_SYSV_fake
-
-	/* For unwind information, Windows has to store fp and lr  */
-	PROLOG_SAVE_REG_PAIR	x29, x30, #-32!
-
-	ALTERNATE_ENTRY ffi_call_SYSV
-	/* Use a stack frame allocated by our caller. */
-	stp	x29, x30, [x1]
-	mov	x29, x1
-	mov	sp, x0
-
-	mov	x9, x2			/* save fn */
-	mov	x8, x3			/* install structure return */
-#ifdef FFI_GO_CLOSURES
-	/*mov	x18, x5			install static chain */
-#endif
-	stp	x3, x4, [x29, #16]	/* save rvalue and flags */
-	
-	/* Load the vector argument passing registers, if necessary.  */
-	tbz	x4, #AARCH64_FLAG_ARG_V_BIT, ffi_call_SYSV_L1
-	ldp	q0, q1, [sp, #0]
-	ldp	q2, q3, [sp, #32]
-	ldp	q4, q5, [sp, #64]
-	ldp	q6, q7, [sp, #96]
-
-ffi_call_SYSV_L1
-	/* Load the core argument passing registers, including
-	   the structure return pointer.  */
-	ldp     x0, x1, [sp, #16*N_V_ARG_REG + 0]
-	ldp     x2, x3, [sp, #16*N_V_ARG_REG + 16]
-	ldp     x4, x5, [sp, #16*N_V_ARG_REG + 32]
-	ldp     x6, x7, [sp, #16*N_V_ARG_REG + 48]
-
-	/* Deallocate the context, leaving the stacked arguments.  */
-	add	sp, sp, #CALL_CONTEXT_SIZE	
-
-	blr     x9			/* call fn */
-
-	ldp	x3, x4, [x29, #16]	/* reload rvalue and flags */
-
-	/* Partially deconstruct the stack frame. */
-	mov     sp, x29 
-	ldp     x29, x30, [x29]
-
-	/* Save the return value as directed.  */
-	adr	x5, ffi_call_SYSV_return
-	and	w4, w4, #AARCH64_RET_MASK
-	add	x5, x5, x4, lsl #3
-	br	x5
-	
-	/* Note that each table entry is 2 insns, and thus 8 bytes.
-	   For integer data, note that we're storing into ffi_arg
-	   and therefore we want to extend to 64 bits; these types
-	   have two consecutive entries allocated for them.  */
-	ALIGN 4
-ffi_call_SYSV_return
-	ret				/* VOID */
-	nop
-	str	x0, [x3]		/* INT64 */
-	ret
-	stp	x0, x1, [x3]		/* INT128 */
-	ret
-	brk	#1000			/* UNUSED */
-	ret
-	brk	#1000			/* UNUSED */
-	ret
-	brk	#1000			/* UNUSED */
-	ret
-	brk	#1000			/* UNUSED */
-	ret
-	brk	#1000			/* UNUSED */
-	ret
-	st4	{ v0.s, v1.s, v2.s, v3.s }[0], [x3]	/* S4 */
-	ret
-	st3	{ v0.s, v1.s, v2.s }[0], [x3]	/* S3 */
-	ret
-	stp	s0, s1, [x3]		/* S2 */
-	ret
-	str	s0, [x3]		/* S1 */
-	ret
-	st4	{ v0.d, v1.d, v2.d, v3.d }[0], [x3]	/* D4 */
-	ret
-	st3	{ v0.d, v1.d, v2.d }[0], [x3]	/* D3 */
-	ret
-	stp	d0, d1, [x3]		/* D2 */
-	ret
-	str	d0, [x3]		/* D1 */
-	ret
-	str	q3, [x3, #48]		/* Q4 */
-	nop
-	str	q2, [x3, #32]		/* Q3 */
-	nop
-	stp	q0, q1, [x3]		/* Q2 */
-	ret
-	str	q0, [x3]		/* Q1 */
-	ret
-	uxtb	w0, w0			/* UINT8 */
-	str	x0, [x3]
-	ret				/* reserved */
-	nop
-	uxth	w0, w0			/* UINT16 */
-	str	x0, [x3]
-	ret				/* reserved */
-	nop
-	mov	w0, w0			/* UINT32 */
-	str	x0, [x3]
-	ret				/* reserved */
-	nop
-	sxtb	x0, w0			/* SINT8 */
-	str	x0, [x3]
-	ret				/* reserved */
-	nop
-	sxth	x0, w0			/* SINT16 */
-	str	x0, [x3]
-	ret				/* reserved */
-	nop
-	sxtw	x0, w0			/* SINT32 */
-	str	x0, [x3]
-	ret				/* reserved */
-	nop
-	
-	
-	NESTED_END ffi_call_SYSV_fake
-	
-
-/* ffi_closure_SYSV
-   Closure invocation glue. This is the low level code invoked directly by
-   the closure trampoline to setup and call a closure.
-   On entry x17 points to a struct ffi_closure, x16 has been clobbered
-   all other registers are preserved.
-   We allocate a call context and save the argument passing registers,
-   then invoked the generic C ffi_closure_SYSV_inner() function to do all
-   the real work, on return we load the result passing registers back from
-   the call context.
-*/
-
-#define ffi_closure_SYSV_FS (8*2 + CALL_CONTEXT_SIZE + 64)
-
-	NESTED_ENTRY	ffi_closure_SYSV_V
-	PROLOG_SAVE_REG_PAIR	x29, x30, #-ffi_closure_SYSV_FS!
-
-	/* Save the argument passing vector registers.  */
-	stp	q0, q1, [sp, #16 + 0]
-	stp	q2, q3, [sp, #16 + 32]
-	stp	q4, q5, [sp, #16 + 64]
-	stp	q6, q7, [sp, #16 + 96]
-
-	b	ffi_closure_SYSV_save_argument
-	NESTED_END	ffi_closure_SYSV_V
-
-	NESTED_ENTRY	ffi_closure_SYSV
-	PROLOG_SAVE_REG_PAIR	x29, x30, #-ffi_closure_SYSV_FS!
-
-ffi_closure_SYSV_save_argument
-	/* Save the argument passing core registers.  */
-	stp     x0, x1, [sp, #16 + 16*N_V_ARG_REG + 0]
-	stp     x2, x3, [sp, #16 + 16*N_V_ARG_REG + 16]
-	stp     x4, x5, [sp, #16 + 16*N_V_ARG_REG + 32]
-	stp     x6, x7, [sp, #16 + 16*N_V_ARG_REG + 48]
-
-	/* Load ffi_closure_inner arguments.  */
-	ldp	PTR_REG(0), PTR_REG(1), [x17, #FFI_TRAMPOLINE_CLOSURE_OFFSET]	/* load cif, fn */
-	ldr	PTR_REG(2), [x17, #FFI_TRAMPOLINE_CLOSURE_OFFSET+PTR_SIZE*2]	/* load user_data */
-
-do_closure
-	add	x3, sp, #16							/* load context */
-	add	x4, sp, #ffi_closure_SYSV_FS		/* load stack */
-	add	x5, sp, #16+CALL_CONTEXT_SIZE		/* load rvalue */
-	mov	x6, x8					/* load struct_rval */
-
-	bl	ffi_closure_SYSV_inner
-
-	/* Load the return value as directed.  */
-	adr	x1, ffi_closure_SYSV_return_base
-	and	w0, w0, #AARCH64_RET_MASK
-	add	x1, x1, x0, lsl #3
-	add	x3, sp, #16+CALL_CONTEXT_SIZE
-	br	x1
-
-	/* Note that each table entry is 2 insns, and thus 8 bytes.  */
-	ALIGN	8
-ffi_closure_SYSV_return_base
-	b	ffi_closure_SYSV_epilog			/* VOID */
-	nop
-	ldr	x0, [x3]		/* INT64 */
-	b	ffi_closure_SYSV_epilog
-	ldp	x0, x1, [x3]		/* INT128 */
-	b	ffi_closure_SYSV_epilog
-	brk	#1000			/* UNUSED */
-	nop
-	brk	#1000			/* UNUSED */
-	nop
-	brk	#1000			/* UNUSED */
-	nop
-	brk	#1000			/* UNUSED */
-	nop
-	brk	#1000			/* UNUSED */
-	nop
-	ldr	s3, [x3, #12]		/* S4 */
-	nop
-	ldr	s2, [x3, #8]		/* S3 */
-	nop
-	ldp	s0, s1, [x3]		/* S2 */
-	b	ffi_closure_SYSV_epilog
-	ldr	s0, [x3]		/* S1 */
-	b	ffi_closure_SYSV_epilog
-	ldr	d3, [x3, #24]		/* D4 */
-	nop
-	ldr	d2, [x3, #16]		/* D3 */
-	nop
-	ldp	d0, d1, [x3]		/* D2 */
-	b	ffi_closure_SYSV_epilog
-	ldr	d0, [x3]		/* D1 */
-	b	ffi_closure_SYSV_epilog
-	ldr	q3, [x3, #48]		/* Q4 */
-	nop
-	ldr	q2, [x3, #32]		/* Q3 */
-	nop
-	ldp	q0, q1, [x3]		/* Q2 */
-	b	ffi_closure_SYSV_epilog
-	ldr	q0, [x3]		/* Q1 */
-	b	ffi_closure_SYSV_epilog
-	ldrb	w0, [x3, #BE(7)]	/* UINT8 */
-	b	ffi_closure_SYSV_epilog
-	brk	#1000			/* reserved */
-	nop
-	ldrh	w0, [x3, #BE(6)]	/* UINT16 */
-	b	ffi_closure_SYSV_epilog
-	brk	#1000			/* reserved */
-	nop
-	ldr	w0, [x3, #BE(4)]	/* UINT32 */
-	b	ffi_closure_SYSV_epilog
-	brk	#1000			/* reserved */
-	nop
-	ldrsb	x0, [x3, #BE(7)]	/* SINT8 */
-	b	ffi_closure_SYSV_epilog
-	brk	#1000			/* reserved */
-	nop
-	ldrsh	x0, [x3, #BE(6)]	/* SINT16 */
-	b	ffi_closure_SYSV_epilog
-	brk	#1000			/* reserved */
-	nop
-	ldrsw	x0, [x3, #BE(4)]	/* SINT32 */
-	nop
-					/* reserved */
-
-ffi_closure_SYSV_epilog
-	EPILOG_RESTORE_REG_PAIR	x29, x30, #ffi_closure_SYSV_FS!
-	EPILOG_RETURN
-	NESTED_END	ffi_closure_SYSV
-
-
-#ifdef FFI_GO_CLOSURES
-	NESTED_ENTRY	ffi_go_closure_SYSV_V
-	PROLOG_SAVE_REG_PAIR	x29, x30, #-ffi_closure_SYSV_FS!
-
-	/* Save the argument passing vector registers.  */
-	stp	q0, q1, [sp, #16 + 0]
-	stp	q2, q3, [sp, #16 + 32]
-	stp	q4, q5, [sp, #16 + 64]
-	stp	q6, q7, [sp, #16 + 96]
-	b	ffi_go_closure_SYSV_save_argument
-	NESTED_END	ffi_go_closure_SYSV_V
-
-	NESTED_ENTRY	ffi_go_closure_SYSV
-	PROLOG_SAVE_REG_PAIR	x29, x30, #-ffi_closure_SYSV_FS!
-
-ffi_go_closure_SYSV_save_argument
-	/* Save the argument passing core registers.  */
-	stp     x0, x1, [sp, #16 + 16*N_V_ARG_REG + 0]
-	stp     x2, x3, [sp, #16 + 16*N_V_ARG_REG + 16]
-	stp     x4, x5, [sp, #16 + 16*N_V_ARG_REG + 32]
-	stp     x6, x7, [sp, #16 + 16*N_V_ARG_REG + 48]
-
-	/* Load ffi_closure_inner arguments.  */
-	ldp	PTR_REG(0), PTR_REG(1), [x18, #PTR_SIZE]/* load cif, fn */
-	mov	x2, x18					/* load user_data */
-	b	do_closure
-	NESTED_END	ffi_go_closure_SYSV
-
-#endif /* FFI_GO_CLOSURES */
-
-
-/* void extend_hfa_type (void *dest, void *src, int h) */
-
-	LEAF_ENTRY	extend_hfa_type
-
-	adr	x3, extend_hfa_type_jump_base
-	and	w2, w2, #AARCH64_RET_MASK
-	sub	x2, x2, #AARCH64_RET_S4
-	add	x3, x3, x2, lsl #4
-	br	x3
-
-	ALIGN	4
-extend_hfa_type_jump_base
-	ldp	s16, s17, [x1]		/* S4 */
-	ldp	s18, s19, [x1, #8]
-	b	extend_hfa_type_store_4
-	nop
-
-	ldp	s16, s17, [x1]		/* S3 */
-	ldr	s18, [x1, #8]
-	b	extend_hfa_type_store_3
-	nop
-
-	ldp	s16, s17, [x1]		/* S2 */
-	b	extend_hfa_type_store_2
-	nop
-	nop
-
-	ldr	s16, [x1]		/* S1 */
-	b	extend_hfa_type_store_1
-	nop
-	nop
-
-	ldp	d16, d17, [x1]		/* D4 */
-	ldp	d18, d19, [x1, #16]
-	b       extend_hfa_type_store_4
-	nop
-
-	ldp     d16, d17, [x1]		/* D3 */
-	ldr     d18, [x1, #16]
-	b	extend_hfa_type_store_3
-	nop
-
-	ldp	d16, d17, [x1]		/* D2 */
-	b	extend_hfa_type_store_2
-	nop
-	nop
-
-	ldr	d16, [x1]		/* D1 */
-	b	extend_hfa_type_store_1
-	nop
-	nop
-
-	ldp	q16, q17, [x1]		/* Q4 */
-	ldp	q18, q19, [x1, #16]
-	b	extend_hfa_type_store_4
-	nop
-
-	ldp	q16, q17, [x1]		/* Q3 */
-	ldr	q18, [x1, #16]
-	b	extend_hfa_type_store_3
-	nop
-
-	ldp	q16, q17, [x1]		/* Q2 */
-	b	extend_hfa_type_store_2
-	nop
-	nop
-
-	ldr	q16, [x1]		/* Q1 */
-	b	extend_hfa_type_store_1
-
-extend_hfa_type_store_4
-	str	q19, [x0, #48]
-extend_hfa_type_store_3
-	str	q18, [x0, #32]
-extend_hfa_type_store_2
-	str	q17, [x0, #16]
-extend_hfa_type_store_1
-	str	q16, [x0]
-	ret
-
-	LEAF_END	extend_hfa_type
-
-
-/* void compress_hfa_type (void *dest, void *reg, int h) */
-
-	LEAF_ENTRY	compress_hfa_type
-
-	adr	x3, compress_hfa_type_jump_base
-	and	w2, w2, #AARCH64_RET_MASK
-	sub	x2, x2, #AARCH64_RET_S4
-	add	x3, x3, x2, lsl #4
-	br	x3
-
-	ALIGN	4
-compress_hfa_type_jump_base
-	ldp	q16, q17, [x1]		/* S4 */
-	ldp	q18, q19, [x1, #32]
-	st4	{ v16.s, v17.s, v18.s, v19.s }[0], [x0]
-	ret
-
-	ldp	q16, q17, [x1]		/* S3 */
-	ldr	q18, [x1, #32]
-	st3	{ v16.s, v17.s, v18.s }[0], [x0]
-	ret
-
-	ldp	q16, q17, [x1]		/* S2 */
-	st2	{ v16.s, v17.s }[0], [x0]
-	ret
-	nop
-
-	ldr	q16, [x1]		/* S1 */
-	st1	{ v16.s }[0], [x0]
-	ret
-	nop
-
-	ldp	q16, q17, [x1]		/* D4 */
-	ldp	q18, q19, [x1, #32]
-	st4	{ v16.d, v17.d, v18.d, v19.d }[0], [x0]
-	ret
-
-	ldp	q16, q17, [x1]		/* D3 */
-	ldr	q18, [x1, #32]
-	st3	{ v16.d, v17.d, v18.d }[0], [x0]
-	ret
-
-	ldp	q16, q17, [x1]		/* D2 */
-	st2	{ v16.d, v17.d }[0], [x0]
-	ret
-	nop
-
-	ldr	q16, [x1]		/* D1 */
-	st1	{ v16.d }[0], [x0]
-	ret
-	nop
-
-	ldp	q16, q17, [x1]		/* Q4 */
-	ldp	q18, q19, [x1, #32]
-	b	compress_hfa_type_store_q4
-	nop
-
-	ldp	q16, q17, [x1]		/* Q3 */
-	ldr	q18, [x1, #32]
-	b	compress_hfa_type_store_q3
-	nop
-
-	ldp	q16, q17, [x1]		/* Q2 */
-	stp	q16, q17, [x0]
-	ret
-	nop
-
-	ldr	q16, [x1]		/* Q1 */
-	str	q16, [x0]
-	ret
-
-compress_hfa_type_store_q4
-	str	q19, [x0, #48]
-compress_hfa_type_store_q3
-	str	q18, [x0, #32]
-	stp	q16, q17, [x0]
-	ret
-
-	LEAF_END	compress_hfa_type
-
-	END
\ No newline at end of file
+/* Copyright (c) 2009, 2010, 2011, 2012 ARM Ltd. 
+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 THE AUTHORS OR COPYRIGHT HOLDERS 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.  */ 
+ 
+#define LIBFFI_ASM 
+#include <fficonfig.h> 
+#include <ffi.h> 
+#include <ffi_cfi.h> 
+#include "internal.h" 
+ 
+	OPT	2 /*disable listing */ 
+/* For some macros to add unwind information */ 
+#include "ksarm64.h" 
+	OPT	1 /*re-enable listing */ 
+ 
+#define BE(X)	0 
+#define PTR_REG(n)      x##n 
+#define PTR_SIZE	8 
+ 
+	IMPORT ffi_closure_SYSV_inner 
+	EXPORT	ffi_call_SYSV 
+	EXPORT	ffi_closure_SYSV_V 
+	EXPORT	ffi_closure_SYSV 
+	EXPORT	extend_hfa_type 
+	EXPORT	compress_hfa_type 
+#ifdef FFI_GO_CLOSURES 
+	EXPORT	ffi_go_closure_SYSV_V 
+	EXPORT	ffi_go_closure_SYSV 
+#endif 
+ 
+	TEXTAREA, ALLIGN=8 
+ 
+/* ffi_call_SYSV 
+   extern void ffi_call_SYSV (void *stack, void *frame, 
+			      void (*fn)(void), void *rvalue, 
+			      int flags, void *closure); 
+   Therefore on entry we have: 
+   x0 stack 
+   x1 frame 
+   x2 fn 
+   x3 rvalue 
+   x4 flags 
+   x5 closure 
+*/ 
+ 
+	NESTED_ENTRY ffi_call_SYSV_fake 
+ 
+	/* For unwind information, Windows has to store fp and lr  */ 
+	PROLOG_SAVE_REG_PAIR	x29, x30, #-32! 
+ 
+	ALTERNATE_ENTRY ffi_call_SYSV 
+	/* Use a stack frame allocated by our caller. */ 
+	stp	x29, x30, [x1] 
+	mov	x29, x1 
+	mov	sp, x0 
+ 
+	mov	x9, x2			/* save fn */ 
+	mov	x8, x3			/* install structure return */ 
+#ifdef FFI_GO_CLOSURES 
+	/*mov	x18, x5			install static chain */ 
+#endif 
+	stp	x3, x4, [x29, #16]	/* save rvalue and flags */ 
+	 
+	/* Load the vector argument passing registers, if necessary.  */ 
+	tbz	x4, #AARCH64_FLAG_ARG_V_BIT, ffi_call_SYSV_L1 
+	ldp	q0, q1, [sp, #0] 
+	ldp	q2, q3, [sp, #32] 
+	ldp	q4, q5, [sp, #64] 
+	ldp	q6, q7, [sp, #96] 
+ 
+ffi_call_SYSV_L1 
+	/* Load the core argument passing registers, including 
+	   the structure return pointer.  */ 
+	ldp     x0, x1, [sp, #16*N_V_ARG_REG + 0] 
+	ldp     x2, x3, [sp, #16*N_V_ARG_REG + 16] 
+	ldp     x4, x5, [sp, #16*N_V_ARG_REG + 32] 
+	ldp     x6, x7, [sp, #16*N_V_ARG_REG + 48] 
+ 
+	/* Deallocate the context, leaving the stacked arguments.  */ 
+	add	sp, sp, #CALL_CONTEXT_SIZE	 
+ 
+	blr     x9			/* call fn */ 
+ 
+	ldp	x3, x4, [x29, #16]	/* reload rvalue and flags */ 
+ 
+	/* Partially deconstruct the stack frame. */ 
+	mov     sp, x29  
+	ldp     x29, x30, [x29] 
+ 
+	/* Save the return value as directed.  */ 
+	adr	x5, ffi_call_SYSV_return 
+	and	w4, w4, #AARCH64_RET_MASK 
+	add	x5, x5, x4, lsl #3 
+	br	x5 
+	 
+	/* Note that each table entry is 2 insns, and thus 8 bytes. 
+	   For integer data, note that we're storing into ffi_arg 
+	   and therefore we want to extend to 64 bits; these types 
+	   have two consecutive entries allocated for them.  */ 
+	ALIGN 4 
+ffi_call_SYSV_return 
+	ret				/* VOID */ 
+	nop 
+	str	x0, [x3]		/* INT64 */ 
+	ret 
+	stp	x0, x1, [x3]		/* INT128 */ 
+	ret 
+	brk	#1000			/* UNUSED */ 
+	ret 
+	brk	#1000			/* UNUSED */ 
+	ret 
+	brk	#1000			/* UNUSED */ 
+	ret 
+	brk	#1000			/* UNUSED */ 
+	ret 
+	brk	#1000			/* UNUSED */ 
+	ret 
+	st4	{ v0.s, v1.s, v2.s, v3.s }[0], [x3]	/* S4 */ 
+	ret 
+	st3	{ v0.s, v1.s, v2.s }[0], [x3]	/* S3 */ 
+	ret 
+	stp	s0, s1, [x3]		/* S2 */ 
+	ret 
+	str	s0, [x3]		/* S1 */ 
+	ret 
+	st4	{ v0.d, v1.d, v2.d, v3.d }[0], [x3]	/* D4 */ 
+	ret 
+	st3	{ v0.d, v1.d, v2.d }[0], [x3]	/* D3 */ 
+	ret 
+	stp	d0, d1, [x3]		/* D2 */ 
+	ret 
+	str	d0, [x3]		/* D1 */ 
+	ret 
+	str	q3, [x3, #48]		/* Q4 */ 
+	nop 
+	str	q2, [x3, #32]		/* Q3 */ 
+	nop 
+	stp	q0, q1, [x3]		/* Q2 */ 
+	ret 
+	str	q0, [x3]		/* Q1 */ 
+	ret 
+	uxtb	w0, w0			/* UINT8 */ 
+	str	x0, [x3] 
+	ret				/* reserved */ 
+	nop 
+	uxth	w0, w0			/* UINT16 */ 
+	str	x0, [x3] 
+	ret				/* reserved */ 
+	nop 
+	mov	w0, w0			/* UINT32 */ 
+	str	x0, [x3] 
+	ret				/* reserved */ 
+	nop 
+	sxtb	x0, w0			/* SINT8 */ 
+	str	x0, [x3] 
+	ret				/* reserved */ 
+	nop 
+	sxth	x0, w0			/* SINT16 */ 
+	str	x0, [x3] 
+	ret				/* reserved */ 
+	nop 
+	sxtw	x0, w0			/* SINT32 */ 
+	str	x0, [x3] 
+	ret				/* reserved */ 
+	nop 
+	 
+	 
+	NESTED_END ffi_call_SYSV_fake 
+	 
+ 
+/* ffi_closure_SYSV 
+   Closure invocation glue. This is the low level code invoked directly by 
+   the closure trampoline to setup and call a closure. 
+   On entry x17 points to a struct ffi_closure, x16 has been clobbered 
+   all other registers are preserved. 
+   We allocate a call context and save the argument passing registers, 
+   then invoked the generic C ffi_closure_SYSV_inner() function to do all 
+   the real work, on return we load the result passing registers back from 
+   the call context. 
+*/ 
+ 
+#define ffi_closure_SYSV_FS (8*2 + CALL_CONTEXT_SIZE + 64) 
+ 
+	NESTED_ENTRY	ffi_closure_SYSV_V 
+	PROLOG_SAVE_REG_PAIR	x29, x30, #-ffi_closure_SYSV_FS! 
+ 
+	/* Save the argument passing vector registers.  */ 
+	stp	q0, q1, [sp, #16 + 0] 
+	stp	q2, q3, [sp, #16 + 32] 
+	stp	q4, q5, [sp, #16 + 64] 
+	stp	q6, q7, [sp, #16 + 96] 
+ 
+	b	ffi_closure_SYSV_save_argument 
+	NESTED_END	ffi_closure_SYSV_V 
+ 
+	NESTED_ENTRY	ffi_closure_SYSV 
+	PROLOG_SAVE_REG_PAIR	x29, x30, #-ffi_closure_SYSV_FS! 
+ 
+ffi_closure_SYSV_save_argument 
+	/* Save the argument passing core registers.  */ 
+	stp     x0, x1, [sp, #16 + 16*N_V_ARG_REG + 0] 
+	stp     x2, x3, [sp, #16 + 16*N_V_ARG_REG + 16] 
+	stp     x4, x5, [sp, #16 + 16*N_V_ARG_REG + 32] 
+	stp     x6, x7, [sp, #16 + 16*N_V_ARG_REG + 48] 
+ 
+	/* Load ffi_closure_inner arguments.  */ 
+	ldp	PTR_REG(0), PTR_REG(1), [x17, #FFI_TRAMPOLINE_CLOSURE_OFFSET]	/* load cif, fn */ 
+	ldr	PTR_REG(2), [x17, #FFI_TRAMPOLINE_CLOSURE_OFFSET+PTR_SIZE*2]	/* load user_data */ 
+ 
+do_closure 
+	add	x3, sp, #16							/* load context */ 
+	add	x4, sp, #ffi_closure_SYSV_FS		/* load stack */ 
+	add	x5, sp, #16+CALL_CONTEXT_SIZE		/* load rvalue */ 
+	mov	x6, x8					/* load struct_rval */ 
+ 
+	bl	ffi_closure_SYSV_inner 
+ 
+	/* Load the return value as directed.  */ 
+	adr	x1, ffi_closure_SYSV_return_base 
+	and	w0, w0, #AARCH64_RET_MASK 
+	add	x1, x1, x0, lsl #3 
+	add	x3, sp, #16+CALL_CONTEXT_SIZE 
+	br	x1 
+ 
+	/* Note that each table entry is 2 insns, and thus 8 bytes.  */ 
+	ALIGN	8 
+ffi_closure_SYSV_return_base 
+	b	ffi_closure_SYSV_epilog			/* VOID */ 
+	nop 
+	ldr	x0, [x3]		/* INT64 */ 
+	b	ffi_closure_SYSV_epilog 
+	ldp	x0, x1, [x3]		/* INT128 */ 
+	b	ffi_closure_SYSV_epilog 
+	brk	#1000			/* UNUSED */ 
+	nop 
+	brk	#1000			/* UNUSED */ 
+	nop 
+	brk	#1000			/* UNUSED */ 
+	nop 
+	brk	#1000			/* UNUSED */ 
+	nop 
+	brk	#1000			/* UNUSED */ 
+	nop 
+	ldr	s3, [x3, #12]		/* S4 */ 
+	nop 
+	ldr	s2, [x3, #8]		/* S3 */ 
+	nop 
+	ldp	s0, s1, [x3]		/* S2 */ 
+	b	ffi_closure_SYSV_epilog 
+	ldr	s0, [x3]		/* S1 */ 
+	b	ffi_closure_SYSV_epilog 
+	ldr	d3, [x3, #24]		/* D4 */ 
+	nop 
+	ldr	d2, [x3, #16]		/* D3 */ 
+	nop 
+	ldp	d0, d1, [x3]		/* D2 */ 
+	b	ffi_closure_SYSV_epilog 
+	ldr	d0, [x3]		/* D1 */ 
+	b	ffi_closure_SYSV_epilog 
+	ldr	q3, [x3, #48]		/* Q4 */ 
+	nop 
+	ldr	q2, [x3, #32]		/* Q3 */ 
+	nop 
+	ldp	q0, q1, [x3]		/* Q2 */ 
+	b	ffi_closure_SYSV_epilog 
+	ldr	q0, [x3]		/* Q1 */ 
+	b	ffi_closure_SYSV_epilog 
+	ldrb	w0, [x3, #BE(7)]	/* UINT8 */ 
+	b	ffi_closure_SYSV_epilog 
+	brk	#1000			/* reserved */ 
+	nop 
+	ldrh	w0, [x3, #BE(6)]	/* UINT16 */ 
+	b	ffi_closure_SYSV_epilog 
+	brk	#1000			/* reserved */ 
+	nop 
+	ldr	w0, [x3, #BE(4)]	/* UINT32 */ 
+	b	ffi_closure_SYSV_epilog 
+	brk	#1000			/* reserved */ 
+	nop 
+	ldrsb	x0, [x3, #BE(7)]	/* SINT8 */ 
+	b	ffi_closure_SYSV_epilog 
+	brk	#1000			/* reserved */ 
+	nop 
+	ldrsh	x0, [x3, #BE(6)]	/* SINT16 */ 
+	b	ffi_closure_SYSV_epilog 
+	brk	#1000			/* reserved */ 
+	nop 
+	ldrsw	x0, [x3, #BE(4)]	/* SINT32 */ 
+	nop 
+					/* reserved */ 
+ 
+ffi_closure_SYSV_epilog 
+	EPILOG_RESTORE_REG_PAIR	x29, x30, #ffi_closure_SYSV_FS! 
+	EPILOG_RETURN 
+	NESTED_END	ffi_closure_SYSV 
+ 
+ 
+#ifdef FFI_GO_CLOSURES 
+	NESTED_ENTRY	ffi_go_closure_SYSV_V 
+	PROLOG_SAVE_REG_PAIR	x29, x30, #-ffi_closure_SYSV_FS! 
+ 
+	/* Save the argument passing vector registers.  */ 
+	stp	q0, q1, [sp, #16 + 0] 
+	stp	q2, q3, [sp, #16 + 32] 
+	stp	q4, q5, [sp, #16 + 64] 
+	stp	q6, q7, [sp, #16 + 96] 
+	b	ffi_go_closure_SYSV_save_argument 
+	NESTED_END	ffi_go_closure_SYSV_V 
+ 
+	NESTED_ENTRY	ffi_go_closure_SYSV 
+	PROLOG_SAVE_REG_PAIR	x29, x30, #-ffi_closure_SYSV_FS! 
+ 
+ffi_go_closure_SYSV_save_argument 
+	/* Save the argument passing core registers.  */ 
+	stp     x0, x1, [sp, #16 + 16*N_V_ARG_REG + 0] 
+	stp     x2, x3, [sp, #16 + 16*N_V_ARG_REG + 16] 
+	stp     x4, x5, [sp, #16 + 16*N_V_ARG_REG + 32] 
+	stp     x6, x7, [sp, #16 + 16*N_V_ARG_REG + 48] 
+ 
+	/* Load ffi_closure_inner arguments.  */ 
+	ldp	PTR_REG(0), PTR_REG(1), [x18, #PTR_SIZE]/* load cif, fn */ 
+	mov	x2, x18					/* load user_data */ 
+	b	do_closure 
+	NESTED_END	ffi_go_closure_SYSV 
+ 
+#endif /* FFI_GO_CLOSURES */ 
+ 
+ 
+/* void extend_hfa_type (void *dest, void *src, int h) */ 
+ 
+	LEAF_ENTRY	extend_hfa_type 
+ 
+	adr	x3, extend_hfa_type_jump_base 
+	and	w2, w2, #AARCH64_RET_MASK 
+	sub	x2, x2, #AARCH64_RET_S4 
+	add	x3, x3, x2, lsl #4 
+	br	x3 
+ 
+	ALIGN	4 
+extend_hfa_type_jump_base 
+	ldp	s16, s17, [x1]		/* S4 */ 
+	ldp	s18, s19, [x1, #8] 
+	b	extend_hfa_type_store_4 
+	nop 
+ 
+	ldp	s16, s17, [x1]		/* S3 */ 
+	ldr	s18, [x1, #8] 
+	b	extend_hfa_type_store_3 
+	nop 
+ 
+	ldp	s16, s17, [x1]		/* S2 */ 
+	b	extend_hfa_type_store_2 
+	nop 
+	nop 
+ 
+	ldr	s16, [x1]		/* S1 */ 
+	b	extend_hfa_type_store_1 
+	nop 
+	nop 
+ 
+	ldp	d16, d17, [x1]		/* D4 */ 
+	ldp	d18, d19, [x1, #16] 
+	b       extend_hfa_type_store_4 
+	nop 
+ 
+	ldp     d16, d17, [x1]		/* D3 */ 
+	ldr     d18, [x1, #16] 
+	b	extend_hfa_type_store_3 
+	nop 
+ 
+	ldp	d16, d17, [x1]		/* D2 */ 
+	b	extend_hfa_type_store_2 
+	nop 
+	nop 
+ 
+	ldr	d16, [x1]		/* D1 */ 
+	b	extend_hfa_type_store_1 
+	nop 
+	nop 
+ 
+	ldp	q16, q17, [x1]		/* Q4 */ 
+	ldp	q18, q19, [x1, #16] 
+	b	extend_hfa_type_store_4 
+	nop 
+ 
+	ldp	q16, q17, [x1]		/* Q3 */ 
+	ldr	q18, [x1, #16] 
+	b	extend_hfa_type_store_3 
+	nop 
+ 
+	ldp	q16, q17, [x1]		/* Q2 */ 
+	b	extend_hfa_type_store_2 
+	nop 
+	nop 
+ 
+	ldr	q16, [x1]		/* Q1 */ 
+	b	extend_hfa_type_store_1 
+ 
+extend_hfa_type_store_4 
+	str	q19, [x0, #48] 
+extend_hfa_type_store_3 
+	str	q18, [x0, #32] 
+extend_hfa_type_store_2 
+	str	q17, [x0, #16] 
+extend_hfa_type_store_1 
+	str	q16, [x0] 
+	ret 
+ 
+	LEAF_END	extend_hfa_type 
+ 
+ 
+/* void compress_hfa_type (void *dest, void *reg, int h) */ 
+ 
+	LEAF_ENTRY	compress_hfa_type 
+ 
+	adr	x3, compress_hfa_type_jump_base 
+	and	w2, w2, #AARCH64_RET_MASK 
+	sub	x2, x2, #AARCH64_RET_S4 
+	add	x3, x3, x2, lsl #4 
+	br	x3 
+ 
+	ALIGN	4 
+compress_hfa_type_jump_base 
+	ldp	q16, q17, [x1]		/* S4 */ 
+	ldp	q18, q19, [x1, #32] 
+	st4	{ v16.s, v17.s, v18.s, v19.s }[0], [x0] 
+	ret 
+ 
+	ldp	q16, q17, [x1]		/* S3 */ 
+	ldr	q18, [x1, #32] 
+	st3	{ v16.s, v17.s, v18.s }[0], [x0] 
+	ret 
+ 
+	ldp	q16, q17, [x1]		/* S2 */ 
+	st2	{ v16.s, v17.s }[0], [x0] 
+	ret 
+	nop 
+ 
+	ldr	q16, [x1]		/* S1 */ 
+	st1	{ v16.s }[0], [x0] 
+	ret 
+	nop 
+ 
+	ldp	q16, q17, [x1]		/* D4 */ 
+	ldp	q18, q19, [x1, #32] 
+	st4	{ v16.d, v17.d, v18.d, v19.d }[0], [x0] 
+	ret 
+ 
+	ldp	q16, q17, [x1]		/* D3 */ 
+	ldr	q18, [x1, #32] 
+	st3	{ v16.d, v17.d, v18.d }[0], [x0] 
+	ret 
+ 
+	ldp	q16, q17, [x1]		/* D2 */ 
+	st2	{ v16.d, v17.d }[0], [x0] 
+	ret 
+	nop 
+ 
+	ldr	q16, [x1]		/* D1 */ 
+	st1	{ v16.d }[0], [x0] 
+	ret 
+	nop 
+ 
+	ldp	q16, q17, [x1]		/* Q4 */ 
+	ldp	q18, q19, [x1, #32] 
+	b	compress_hfa_type_store_q4 
+	nop 
+ 
+	ldp	q16, q17, [x1]		/* Q3 */ 
+	ldr	q18, [x1, #32] 
+	b	compress_hfa_type_store_q3 
+	nop 
+ 
+	ldp	q16, q17, [x1]		/* Q2 */ 
+	stp	q16, q17, [x0] 
+	ret 
+	nop 
+ 
+	ldr	q16, [x1]		/* Q1 */ 
+	str	q16, [x0] 
+	ret 
+ 
+compress_hfa_type_store_q4 
+	str	q19, [x0, #48] 
+compress_hfa_type_store_q3 
+	str	q18, [x0, #32] 
+	stp	q16, q17, [x0] 
+	ret 
+ 
+	LEAF_END	compress_hfa_type 
+ 
+	END 
\ No newline at end of file