/**************************************************************************/
/*                                                                        */
/*                                 OCaml                                  */
/*                                                                        */
/*          Manuel Serrano and Xavier Leroy, INRIA Rocquencourt           */
/*                                                                        */
/*   Copyright 2000 Institut National de Recherche en Informatique et     */
/*     en Automatique.                                                    */
/*                                                                        */
/*   All rights reserved.  This file is distributed under the terms of    */
/*   the GNU Lesser General Public License version 2.1, with the          */
/*   special exception on linking described in the file LICENSE.          */
/*                                                                        */
/**************************************************************************/

#define CAML_INTERNALS

#include <stddef.h>
#include <stdarg.h>
#include <string.h>
#include <assert.h>
#include "caml/alloc.h"
#include "caml/bigarray.h"
#include "caml/custom.h"
#include "caml/fail.h"
#include "caml/intext.h"
#include "caml/hash.h"
#include "caml/memory.h"
#include "caml/mlvalues.h"
#include "caml/signals.h"
#include "caml/atomic_refcount.h"

#define int8 caml_ba_int8
#define uint8 caml_ba_uint8
#define int16 caml_ba_int16
#define uint16 caml_ba_uint16

/* Half-precision floating point numbers */

#if defined(__GNUC__) && defined(__aarch64__)

union float16_bits { uint16_t i; _Float16 f; };

Caml_inline float caml_float16_to_float(uint16 d)
{
  union float16_bits u;
  u.i = d; return u.f;
}

Caml_inline uint16 caml_float_to_float16(float d)
{
  union float16_bits u;
  u.f = d; return u.i;
}

#elif defined(__GNUC__) && defined(__F16C__)

#include <immintrin.h>

Caml_inline float caml_float16_to_float(uint16 d)
{ return _cvtsh_ss(d); }

Caml_inline uint16 caml_float_to_float16(float d)
{ return _cvtss_sh(d, (_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC)); }

#else

union float_bits {
  uint32_t i;
  float f;
};

/*
 * half_to_float_fast5
 * https://gist.github.com/rygorous/2144712
 */
static float caml_float16_to_float(uint16 d)
{
  static const union float_bits magic = { (254 - 15) << 23 };
  static const union float_bits was_infnan = { (127 + 16) << 23 };

  union float_bits o;

  o.i = (d & 0x7fff) << 13;     /* exponent/mantissa bits */
  o.f *= magic.f;               /* exponent adjust */
  if (o.f >= was_infnan.f)      /* make sure Inf/NaN survive */
    o.i |= 255 << 23;
  o.i |= (d & 0x8000) << 16;    /* sign bit */
  return o.f;
}

/*
 * float_to_half_fast3_rtne
 * https://gist.github.com/rygorous/2156668
 */
static uint16 caml_float_to_float16(float d)
{
  static const union float_bits f32infty = { 255 << 23 };
  static const union float_bits f16max = { (127 + 16) << 23 };
  static const union float_bits denorm_magic =
    { ((127 - 15) + (23 - 10) + 1) << 23 };
  static const uint32_t sign_mask = 0x80000000u;

  union float_bits f;
  uint16 o = 0;
  uint32_t sign;

  f.f = d;
  sign = f.i & sign_mask;
  f.i ^= sign;

  // NOTE all the integer compares in this function can be safely
  // compiled into signed compares since all operands are below
  // 0x80000000. Important if you want fast straight SSE2 code
  // (since there's no unsigned PCMPGTD).

  if (f.i >= f16max.i) // result is Inf or NaN (all exponent bits set)
    o = (f.i > f32infty.i) ? 0x7e00 : 0x7c00; // NaN->qNaN and Inf->Inf
  else // (De)normalized number or zero
  {
    if (f.i < (113 << 23)) // resulting FP16 is subnormal or zero
    {
      // use a magic value to align our 10 mantissa bits at the bottom of
      // the float. as long as FP addition is round-to-nearest-even this
      // just works.
      f.f += denorm_magic.f;

      // and one integer subtract of the bias later, we have our final float!
      o = f.i - denorm_magic.i;
    }
    else
    {
      uint32_t mant_odd = (f.i >> 13) & 1; // resulting mantissa is odd

      // update exponent, rounding bias part 1
      f.i += ((uint32_t)(15 - 127) << 23) + 0xfff;
      // rounding bias part 2
      f.i += mant_odd;
      // take the bits!
      o = f.i >> 13;
    }
  }

  o |= sign >> 16;
  return o;
}

#endif  /* defined(__GNUC__) && defined(__F16C__) */

CAMLexport double caml_double_of_float16(intnat x)
{
  return (double) caml_float16_to_float((uint16) x);
}

CAMLexport intnat caml_float16_of_double(double x)
{
  return (intnat) caml_float_to_float16((float) x);
}

Caml_inline uint32_t caml_hash_mix_float16(uint32_t hash, uint16 d)
{
  /* Normalize NaNs */
  if ((d & 0x7c00) == 0x7c00 && (d & 0x03ff) != 0) {
    d = 0x7c01;
  }
  /* Normalize -0 into +0 */
  else if (d == 0x8000) {
    d = 0;
  }
  return caml_hash_mix_uint32(hash, d);
}

/* Compute the number of elements of a big array */

CAMLexport uintnat caml_ba_num_elts(struct caml_ba_array * b)
{
  uintnat num_elts;
  num_elts = 1;
  for (int i = 0; i < b->num_dims; i++) num_elts = num_elts * b->dim[i];
  return num_elts;
}

/* Size in bytes of a bigarray element, indexed by bigarray kind */

CAMLexport int caml_ba_element_size[] =
{ 4 /*FLOAT32*/, 8 /*FLOAT64*/,
  1 /*SINT8*/, 1 /*UINT8*/,
  2 /*SINT16*/, 2 /*UINT16*/,
  4 /*INT32*/, 8 /*INT64*/,
  sizeof(value) /*CAML_INT*/, sizeof(value) /*NATIVE_INT*/,
  8 /*COMPLEX32*/, 16 /*COMPLEX64*/,
  1 /*CHAR*/, 2 /*FLOAT16*/
};

/* Compute the number of bytes for the elements of a big array */

CAMLexport uintnat caml_ba_byte_size(struct caml_ba_array * b)
{
  return caml_ba_num_elts(b)
         * caml_ba_element_size[b->flags & CAML_BA_KIND_MASK];
}

/* Operation table for bigarrays */

CAMLexport const struct custom_operations caml_ba_ops = {
  "_bigarr02",
  caml_ba_finalize,
  caml_ba_compare,
  caml_ba_hash,
  caml_ba_serialize,
  caml_ba_deserialize,
  custom_compare_ext_default,
  custom_fixed_length_default
};

/* Allocation of a big array */

/* [caml_ba_alloc] will allocate a new bigarray object in the heap.
   If [data] is NULL, the memory for the contents is also allocated
   (with [malloc]) by [caml_ba_alloc].
   [data] cannot point into the OCaml heap.
   [dim] may point into an object in the OCaml heap.
*/
CAMLexport value
caml_ba_alloc(int flags, int num_dims, void * data, intnat * dim)
{
  uintnat num_elts, asize, size;
  int uses_resources;
  value res;
  struct caml_ba_array * b;
  intnat dimcopy[CAML_BA_MAX_NUM_DIMS];

  CAMLassert(0 <= num_dims);
  CAMLassert(num_dims <= CAML_BA_MAX_NUM_DIMS);
  CAMLassert((flags & CAML_BA_KIND_MASK) < CAML_BA_FIRST_UNIMPLEMENTED_KIND);
  for (int i = 0; i < num_dims; i++) dimcopy[i] = dim[i];
  num_elts = 1;
  for (int i = 0; i < num_dims; i++) {
    if (caml_umul_overflow(num_elts, dimcopy[i], &num_elts))
      caml_raise_out_of_memory();
  }
  if (caml_umul_overflow(num_elts,
                         caml_ba_element_size[flags & CAML_BA_KIND_MASK],
                         &size))
    caml_raise_out_of_memory();
  if (data == NULL) {
    data = malloc(size);
    if (data == NULL && size != 0) caml_raise_out_of_memory();
    flags |= CAML_BA_MANAGED;
  }
  asize = SIZEOF_BA_ARRAY + num_dims * sizeof(intnat);
  uses_resources =
    ((flags & CAML_BA_MANAGED_MASK) == CAML_BA_MANAGED)
    && !(flags & CAML_BA_SUBARRAY);
  res = caml_alloc_custom_mem(&caml_ba_ops, asize, uses_resources ? size : 0);
  b = Caml_ba_array_val(res);
  b->data = data;
  b->num_dims = num_dims;
  b->flags = flags;
  b->proxy = NULL;
  for (int i = 0; i < num_dims; i++) b->dim[i] = dimcopy[i];
  return res;
}

/* Same as caml_ba_alloc, but dimensions are passed as a list of
   arguments */

CAMLexport value caml_ba_alloc_dims(int flags, int num_dims, void * data, ...)
{
  va_list ap;
  intnat dim[CAML_BA_MAX_NUM_DIMS];
  value res;

  CAMLassert(num_dims <= CAML_BA_MAX_NUM_DIMS);
  va_start(ap, data);
  for (int i = 0; i < num_dims; i++) dim[i] = va_arg(ap, intnat);
  va_end(ap);
  res = caml_ba_alloc(flags, num_dims, data, dim);
  return res;
}

/* Finalization of a big array */

CAMLexport void caml_ba_finalize(value v)
{
  struct caml_ba_array * b = Caml_ba_array_val(v);

  switch ((enum caml_ba_managed)(b->flags & CAML_BA_MANAGED_MASK)) {
  case CAML_BA_EXTERNAL:
    break;
  case CAML_BA_MANAGED:
    if (b->proxy == NULL) {
      free(b->data);
    } else {
      if (caml_atomic_refcount_decr(&b->proxy->refcount) == 1) {
        free(b->proxy->data);
        free(b->proxy);
      }
    }
    break;
  case CAML_BA_MAPPED_FILE:
    /* Bigarrays for mapped files use a different finalization method */
    CAMLunreachable();
  }
}

/* Comparison of two big arrays */

CAMLexport int caml_ba_compare(value v1, value v2)
{
  struct caml_ba_array * b1 = Caml_ba_array_val(v1);
  struct caml_ba_array * b2 = Caml_ba_array_val(v2);
  uintnat num_elts;
  intnat flags1, flags2;

  /* Compare kind & layout in case the arguments are of different types */
  flags1 = b1->flags & (CAML_BA_KIND_MASK | CAML_BA_LAYOUT_MASK);
  flags2 = b2->flags & (CAML_BA_KIND_MASK | CAML_BA_LAYOUT_MASK);
  if (flags1 != flags2) return flags2 - flags1;
  /* Compare number of dimensions */
  if (b1->num_dims != b2->num_dims) return b2->num_dims - b1->num_dims;
  /* Same number of dimensions: compare dimensions lexicographically */
  for (int i = 0; i < b1->num_dims; i++) {
    intnat d1 = b1->dim[i];
    intnat d2 = b2->dim[i];
    if (d1 != d2) return d1 < d2 ? -1 : 1;
  }
  /* Same dimensions: compare contents lexicographically */
  num_elts = caml_ba_num_elts(b1);

#define DO_INTEGER_COMPARISON(type) \
  { type * p1 = b1->data; type * p2 = b2->data; \
    for (uintnat n = 0; n < num_elts; n++) { \
      type e1 = *p1++; type e2 = *p2++; \
      if (e1 < e2) return -1; \
      if (e1 > e2) return 1; \
    } \
    return 0; \
  }
#define DO_GENERIC_UNORDERED_COMPARISON(ptype, etype, conv) \
  { ptype * p1 = b1->data; ptype * p2 = b2->data; \
    for (uintnat n = 0; n < num_elts; n++) { \
      etype e1 = conv(*p1++); etype e2 = conv(*p2++); \
      if (e1 < e2) return -1; \
      if (e1 > e2) return 1; \
      if (e1 != e2) { \
        Caml_state->compare_unordered = 1; \
        if (e1 == e1) return 1; \
        if (e2 == e2) return -1; \
      } \
    } \
    return 0; \
  }
#define DO_FLOAT_COMPARISON(type) \
  DO_GENERIC_UNORDERED_COMPARISON(type, type, )

  switch ((enum caml_ba_kind)(b1->flags & CAML_BA_KIND_MASK)) {
  case CAML_BA_FLOAT16:
    DO_GENERIC_UNORDERED_COMPARISON(uint16, float, caml_float16_to_float);
  case CAML_BA_COMPLEX32:
    num_elts *= 2; fallthrough;
  case CAML_BA_FLOAT32:
    DO_FLOAT_COMPARISON(float);
  case CAML_BA_COMPLEX64:
    num_elts *= 2; fallthrough;
  case CAML_BA_FLOAT64:
    DO_FLOAT_COMPARISON(double);
  case CAML_BA_CHAR:
    DO_INTEGER_COMPARISON(caml_ba_uint8);
  case CAML_BA_SINT8:
    DO_INTEGER_COMPARISON(caml_ba_int8);
  case CAML_BA_UINT8:
    DO_INTEGER_COMPARISON(caml_ba_uint8);
  case CAML_BA_SINT16:
    DO_INTEGER_COMPARISON(caml_ba_int16);
  case CAML_BA_UINT16:
    DO_INTEGER_COMPARISON(caml_ba_uint16);
  case CAML_BA_INT32:
    DO_INTEGER_COMPARISON(int32_t);
  case CAML_BA_INT64:
    DO_INTEGER_COMPARISON(int64_t);
  case CAML_BA_CAML_INT:
  case CAML_BA_NATIVE_INT:
    DO_INTEGER_COMPARISON(intnat);
  default: CAMLunreachable();
  }
#undef DO_INTEGER_COMPARISON
#undef DO_FLOAT_COMPARISON
}

/* Hashing of a bigarray */

CAMLexport intnat caml_ba_hash(value v)
{
  struct caml_ba_array * b = Caml_ba_array_val(v);
  intnat num_elts;
  uint32_t h, w;

  num_elts = 1;
  for (int i = 0; i < b->num_dims; i++) num_elts = num_elts * b->dim[i];
  h = 0;

  switch ((enum caml_ba_kind)(b->flags & CAML_BA_KIND_MASK)) {
  case CAML_BA_CHAR:
  case CAML_BA_SINT8:
  case CAML_BA_UINT8: {
    caml_ba_uint8 * p = b->data;
    if (num_elts > 256) num_elts = 256;
    for (intnat n = 0; n + 4 <= num_elts; n += 4, p += 4) {
      w = p[0] | (p[1] << 8) | (p[2] << 16) | (p[3] << 24);
      h = caml_hash_mix_uint32(h, w);
    }
    w = 0;
    switch (num_elts & 3) {
    case 3: w  = p[2] << 16; fallthrough;
    case 2: w |= p[1] << 8;  fallthrough;
    case 1: w |= p[0];
            h = caml_hash_mix_uint32(h, w);
    }
    break;
  }
  case CAML_BA_SINT16:
  case CAML_BA_UINT16: {
    caml_ba_uint16 * p = b->data;
    if (num_elts > 128) num_elts = 128;
    for (intnat n = 0; n + 2 <= num_elts; n += 2, p += 2) {
      w = p[0] | (p[1] << 16);
      h = caml_hash_mix_uint32(h, w);
    }
    if ((num_elts & 1) != 0)
      h = caml_hash_mix_uint32(h, p[0]);
    break;
  }
  case CAML_BA_INT32:
  {
    uint32_t * p = b->data;
    if (num_elts > 64) num_elts = 64;
    for (intnat n = 0; n < num_elts; n++, p++) h = caml_hash_mix_uint32(h, *p);
    break;
  }
  case CAML_BA_CAML_INT:
  case CAML_BA_NATIVE_INT:
  {
    intnat * p = b->data;
    if (num_elts > 64) num_elts = 64;
    for (intnat n = 0; n < num_elts; n++, p++) h = caml_hash_mix_intnat(h, *p);
    break;
  }
  case CAML_BA_INT64:
  {
    int64_t * p = b->data;
    if (num_elts > 32) num_elts = 32;
    for (intnat n = 0; n < num_elts; n++, p++) h = caml_hash_mix_int64(h, *p);
    break;
  }
  case CAML_BA_FLOAT16:
  {
    uint16 * p = b->data;
    if (num_elts > 128) num_elts = 128;
    for (intnat n = 0; n < num_elts; n++, p++) h = caml_hash_mix_float16(h, *p);
    break;
  }
  case CAML_BA_COMPLEX32:
    num_elts *= 2;
    fallthrough;
  case CAML_BA_FLOAT32:
  {
    float * p = b->data;
    if (num_elts > 64) num_elts = 64;
    for (intnat n = 0; n < num_elts; n++, p++) h = caml_hash_mix_float(h, *p);
    break;
  }
  case CAML_BA_COMPLEX64:
    num_elts *= 2;
    fallthrough;
  case CAML_BA_FLOAT64:
  {
    double * p = b->data;
    if (num_elts > 32) num_elts = 32;
    for (intnat n = 0; n < num_elts; n++, p++) h = caml_hash_mix_double(h, *p);
    break;
  }
  default: CAMLunreachable();
  }
  return h;
}

static void caml_ba_serialize_longarray(void * data,
                                        intnat num_elts,
                                        intnat min_val, intnat max_val)
{
#ifdef ARCH_SIXTYFOUR
  int overflow_32 = 0;
  for (intnat n = 0, *p = data; n < num_elts; n++, p++) {
    if (*p < min_val || *p > max_val) { overflow_32 = 1; break; }
  }
  if (overflow_32) {
    caml_serialize_int_1(1);
    caml_serialize_block_8(data, num_elts);
  } else {
    caml_serialize_int_1(0);
    for (intnat n = 0, *p = data; n < num_elts; n++, p++)
      caml_serialize_int_4((int32_t) *p);
  }
#else
  caml_serialize_int_1(0);
  caml_serialize_block_4(data, num_elts);
#endif
}

CAMLexport void caml_ba_serialize(value v,
                              uintnat * wsize_32,
                              uintnat * wsize_64)
{
  struct caml_ba_array * b = Caml_ba_array_val(v);
  intnat num_elts;

  /* Serialize header information */
  caml_serialize_int_4(b->num_dims);
  caml_serialize_int_4(b->flags & (CAML_BA_KIND_MASK | CAML_BA_LAYOUT_MASK));
  for (int i = 0; i < b->num_dims; i++) {
    intnat len = b->dim[i];
    if (len < 0xffff) {
      caml_serialize_int_2(len);
    } else {
      caml_serialize_int_2(0xffff);
      caml_serialize_int_8(len);
    }
  }
  /* Compute total number of elements */
  num_elts = 1;
  for (int i = 0; i < b->num_dims; i++) num_elts = num_elts * b->dim[i];
  /* Serialize elements */
  switch ((enum caml_ba_kind)(b->flags & CAML_BA_KIND_MASK)) {
  case CAML_BA_CHAR:
  case CAML_BA_SINT8:
  case CAML_BA_UINT8:
    caml_serialize_block_1(b->data, num_elts); break;
  case CAML_BA_SINT16:
  case CAML_BA_UINT16:
  case CAML_BA_FLOAT16:
    caml_serialize_block_2(b->data, num_elts); break;
  case CAML_BA_FLOAT32:
  case CAML_BA_INT32:
    caml_serialize_block_4(b->data, num_elts); break;
  case CAML_BA_COMPLEX32:
    caml_serialize_block_4(b->data, num_elts * 2); break;
  case CAML_BA_FLOAT64:
  case CAML_BA_INT64:
    caml_serialize_block_8(b->data, num_elts); break;
  case CAML_BA_COMPLEX64:
    caml_serialize_block_8(b->data, num_elts * 2); break;
  case CAML_BA_CAML_INT:
    caml_ba_serialize_longarray(b->data, num_elts, INT32_MIN/2, INT32_MAX/2);
    break;
  case CAML_BA_NATIVE_INT:
    caml_ba_serialize_longarray(b->data, num_elts, INT32_MIN, INT32_MAX);
    break;
  default: CAMLunreachable();
  }
  /* Compute required size in OCaml heap.  Assumes struct caml_ba_array
     is exactly 4 + num_dims words */
  static_assert(SIZEOF_BA_ARRAY == 4 * sizeof(value), "");
  *wsize_32 = (4 + b->num_dims) * 4;
  *wsize_64 = (4 + b->num_dims) * 8;
}

static void caml_ba_deserialize_longarray(void * dest, intnat num_elts)
{
  int sixty = caml_deserialize_uint_1();
#ifdef ARCH_SIXTYFOUR
  if (sixty) {
    caml_deserialize_block_8(dest, num_elts);
  } else {
    for (intnat n = 0, *p = dest; n < num_elts; n++, p++)
      *p = caml_deserialize_sint_4();
  }
#else
  if (sixty)
    caml_deserialize_error("input_value: cannot read bigarray "
                      "with 64-bit OCaml ints");
  caml_deserialize_block_4(dest, num_elts);
#endif
}

CAMLexport uintnat caml_ba_deserialize(void * dst)
{
  struct caml_ba_array * b = dst;
  uintnat num_elts, size;

  /* Read back header information */
  b->num_dims = caml_deserialize_uint_4();
  if (b->num_dims < 0 || b->num_dims > CAML_BA_MAX_NUM_DIMS)
    caml_deserialize_error("input_value: wrong number of bigarray dimensions");
  b->flags = caml_deserialize_uint_4() | CAML_BA_MANAGED;
  b->proxy = NULL;
  for (int i = 0; i < b->num_dims; i++) {
    intnat len = caml_deserialize_uint_2();
    if (len == 0xffff) len = caml_deserialize_uint_8();
    b->dim[i] = len;
  }
  /* Compute total number of elements.  Watch out for overflows (MPR#7765). */
  num_elts = 1;
  for (int i = 0; i < b->num_dims; i++) {
    if (caml_umul_overflow(num_elts, b->dim[i], &num_elts))
      caml_deserialize_error("input_value: size overflow for bigarray");
  }
  /* Determine array size in bytes.  Watch out for overflows (MPR#7765). */
  if ((b->flags & CAML_BA_KIND_MASK) >= CAML_BA_FIRST_UNIMPLEMENTED_KIND)
    caml_deserialize_error("input_value: bad bigarray kind");
  if (caml_umul_overflow(num_elts,
                         caml_ba_element_size[b->flags & CAML_BA_KIND_MASK],
                         &size))
    caml_deserialize_error("input_value: size overflow for bigarray");
  /* Allocate room for data */
  b->data = malloc(size);
  if (b->data == NULL)
    caml_deserialize_error("input_value: out of memory for bigarray");
  /* Read data */
  switch ((enum caml_ba_kind)(b->flags & CAML_BA_KIND_MASK)) {
  case CAML_BA_CHAR:
  case CAML_BA_SINT8:
  case CAML_BA_UINT8:
    caml_deserialize_block_1(b->data, num_elts); break;
  case CAML_BA_SINT16:
  case CAML_BA_UINT16:
  case CAML_BA_FLOAT16:
    caml_deserialize_block_2(b->data, num_elts); break;
  case CAML_BA_FLOAT32:
  case CAML_BA_INT32:
    caml_deserialize_block_4(b->data, num_elts); break;
  case CAML_BA_COMPLEX32:
    caml_deserialize_block_4(b->data, num_elts * 2); break;
  case CAML_BA_FLOAT64:
  case CAML_BA_INT64:
    caml_deserialize_block_8(b->data, num_elts); break;
  case CAML_BA_COMPLEX64:
    caml_deserialize_block_8(b->data, num_elts * 2); break;
  case CAML_BA_CAML_INT:
  case CAML_BA_NATIVE_INT:
    caml_ba_deserialize_longarray(b->data, num_elts); break;
  default: CAMLunreachable();
  }
  /* PR#5516: use C99's flexible array types if possible */
  return SIZEOF_BA_ARRAY + b->num_dims * sizeof(intnat);
}

/* Allocate a bigarray from OCaml */

CAMLprim value caml_ba_create(value vkind, value vlayout, value vdim)
{
  intnat dim[CAML_BA_MAX_NUM_DIMS];
  mlsize_t num_dims;
  int flags;

  num_dims = Wosize_val(vdim);
  /* here num_dims is unsigned (mlsize_t) so no need to check (num_dims >= 0) */
  if (num_dims > CAML_BA_MAX_NUM_DIMS)
    caml_invalid_argument("Bigarray.create: bad number of dimensions");
  for (int i = 0; i < num_dims; i++) {
    dim[i] = Long_val(Field(vdim, i));
    if (dim[i] < 0)
      caml_invalid_argument("Bigarray.create: negative dimension");
  }
  flags = Caml_ba_kind_val(vkind) | Caml_ba_layout_val(vlayout);
  return caml_ba_alloc(flags, num_dims, NULL, dim);
}

/* Given a big array and a vector of indices, check that the indices
   are within the bounds and return the offset of the corresponding
   array element in the data part of the array. */

static intnat caml_ba_offset(struct caml_ba_array * b, intnat * index)
{
  intnat offset;

  offset = 0;
  switch ((enum caml_ba_layout)(b->flags & CAML_BA_LAYOUT_MASK)) {
  case CAML_BA_C_LAYOUT:
    /* C-style layout: row major, indices start at 0 */
    for (int i = 0; i < b->num_dims; i++) {
      if ((uintnat) index[i] >= (uintnat) b->dim[i])
        caml_array_bound_error();
      offset = offset * b->dim[i] + index[i];
    }
    break;
  case CAML_BA_FORTRAN_LAYOUT:
    /* Fortran-style layout: column major, indices start at 1 */
    for (int i = b->num_dims - 1; i >= 0; i--) {
      if ((uintnat) (index[i] - 1) >= (uintnat) b->dim[i])
        caml_array_bound_error();
      offset = offset * b->dim[i] + (index[i] - 1);
    }
    break;
  }
  return offset;
}

/* Helper function to allocate a record of two double floats */

static value copy_two_doubles(double d0, double d1)
{
  value res = caml_alloc_small(2 * Double_wosize, Double_array_tag);
  Store_double_flat_field(res, 0, d0);
  Store_double_flat_field(res, 1, d1);
  return res;
}

/* Generic code to read from a big array */

value caml_ba_get_N(value vb, volatile value * vind, int nind)
{
  struct caml_ba_array * b = Caml_ba_array_val(vb);
  intnat index[CAML_BA_MAX_NUM_DIMS];
  intnat offset;

  /* Check number of indices = number of dimensions of array
     (maybe not necessary if ML typing guarantees this) */
  if (nind != b->num_dims)
    caml_invalid_argument("Bigarray.get: wrong number of indices");
  /* Compute offset and check bounds */
  for (int i = 0; i < b->num_dims; i++) index[i] = Long_val(vind[i]);
  offset = caml_ba_offset(b, index);
  /* Perform read */
  switch ((enum caml_ba_kind)((b->flags) & CAML_BA_KIND_MASK)) {
  case CAML_BA_FLOAT16:
    return caml_copy_double(
      (double) caml_float16_to_float(((uint16 *) b->data)[offset]));
  case CAML_BA_FLOAT32:
    return caml_copy_double((double) ((float *) b->data)[offset]);
  case CAML_BA_FLOAT64:
    return caml_copy_double(((double *) b->data)[offset]);
  case CAML_BA_SINT8:
    return Val_int(((int8 *) b->data)[offset]);
  case CAML_BA_UINT8:
    return Val_int(((uint8 *) b->data)[offset]);
  case CAML_BA_SINT16:
    return Val_int(((int16 *) b->data)[offset]);
  case CAML_BA_UINT16:
    return Val_int(((uint16 *) b->data)[offset]);
  case CAML_BA_INT32:
    return caml_copy_int32(((int32_t *) b->data)[offset]);
  case CAML_BA_INT64:
    return caml_copy_int64(((int64_t *) b->data)[offset]);
  case CAML_BA_NATIVE_INT:
    return caml_copy_nativeint(((intnat *) b->data)[offset]);
  case CAML_BA_CAML_INT:
    return Val_long(((intnat *) b->data)[offset]);
  case CAML_BA_COMPLEX32:
    { float * p = ((float *) b->data) + offset * 2;
      return copy_two_doubles((double) p[0], (double) p[1]); }
  case CAML_BA_COMPLEX64:
    { double * p = ((double *) b->data) + offset * 2;
      return copy_two_doubles(p[0], p[1]); }
  case CAML_BA_CHAR:
    return Val_int(((unsigned char *) b->data)[offset]);
  default: CAMLunreachable();
  }
}

CAMLprim value caml_ba_get_1(value vb, value vind1)
{
  return caml_ba_get_N(vb, &vind1, 1);
}

CAMLprim value caml_ba_get_2(value vb, value vind1, value vind2)
{
  value vind[2];
  vind[0] = vind1; vind[1] = vind2;
  return caml_ba_get_N(vb, vind, 2);
}

CAMLprim value caml_ba_get_3(value vb, value vind1, value vind2, value vind3)
{
  value vind[3];
  vind[0] = vind1; vind[1] = vind2; vind[2] = vind3;
  return caml_ba_get_N(vb, vind, 3);
}

CAMLprim value caml_ba_get_generic(value vb, value vind)
{
  return caml_ba_get_N(vb, &Field(vind, 0), Wosize_val(vind));
}


CAMLprim value caml_ba_uint8_get16(value vb, value vind)
{
  intnat res;
  unsigned char b1, b2;
  intnat idx = Long_val(vind);
  struct caml_ba_array * b = Caml_ba_array_val(vb);
  if (idx < 0 || idx >= b->dim[0] - 1) caml_array_bound_error();
  b1 = ((unsigned char*) b->data)[idx];
  b2 = ((unsigned char*) b->data)[idx+1];
#ifdef ARCH_BIG_ENDIAN
  res = b1 << 8 | b2;
#else
  res = b2 << 8 | b1;
#endif
  return Val_int(res);
}

CAMLprim value caml_ba_uint8_get32(value vb, value vind)
{
  uint32_t res;
  unsigned char b1, b2, b3, b4;
  intnat idx = Long_val(vind);
  struct caml_ba_array * b = Caml_ba_array_val(vb);
  if (idx < 0 || idx >= b->dim[0] - 3) caml_array_bound_error();
  b1 = ((unsigned char*) b->data)[idx];
  b2 = ((unsigned char*) b->data)[idx+1];
  b3 = ((unsigned char*) b->data)[idx+2];
  b4 = ((unsigned char*) b->data)[idx+3];
#ifdef ARCH_BIG_ENDIAN
  res = b1 << 24 | b2 << 16 | b3 << 8 | b4;
#else
  res = b4 << 24 | b3 << 16 | b2 << 8 | b1;
#endif
  return caml_copy_int32(res);
}

CAMLprim value caml_ba_uint8_get64(value vb, value vind)
{
  uint64_t res;
  unsigned char b1, b2, b3, b4, b5, b6, b7, b8;
  intnat idx = Long_val(vind);
  struct caml_ba_array * b = Caml_ba_array_val(vb);
  if (idx < 0 || idx >= b->dim[0] - 7) caml_array_bound_error();
  b1 = ((unsigned char*) b->data)[idx];
  b2 = ((unsigned char*) b->data)[idx+1];
  b3 = ((unsigned char*) b->data)[idx+2];
  b4 = ((unsigned char*) b->data)[idx+3];
  b5 = ((unsigned char*) b->data)[idx+4];
  b6 = ((unsigned char*) b->data)[idx+5];
  b7 = ((unsigned char*) b->data)[idx+6];
  b8 = ((unsigned char*) b->data)[idx+7];
#ifdef ARCH_BIG_ENDIAN
  res = (uint64_t) b1 << 56 | (uint64_t) b2 << 48
        | (uint64_t) b3 << 40 | (uint64_t) b4 << 32
        | (uint64_t) b5 << 24 | (uint64_t) b6 << 16
        | (uint64_t) b7 << 8 | (uint64_t) b8;
#else
  res = (uint64_t) b8 << 56 | (uint64_t) b7 << 48
        | (uint64_t) b6 << 40 | (uint64_t) b5 << 32
        | (uint64_t) b4 << 24 | (uint64_t) b3 << 16
        | (uint64_t) b2 << 8 | (uint64_t) b1;
#endif
  return caml_copy_int64(res);
}

/* Generic write to a big array */

static value caml_ba_set_aux(value vb, volatile value * vind,
                             intnat nind, value newval)
{
  struct caml_ba_array * b = Caml_ba_array_val(vb);
  intnat index[CAML_BA_MAX_NUM_DIMS];
  intnat offset;

  /* Check number of indices = number of dimensions of array
     (maybe not necessary if ML typing guarantees this) */
  if (nind != b->num_dims)
    caml_invalid_argument("Bigarray.set: wrong number of indices");
  /* Compute offset and check bounds */
  for (int i = 0; i < b->num_dims; i++) index[i] = Long_val(vind[i]);
  offset = caml_ba_offset(b, index);
  /* Perform write */
  switch ((enum caml_ba_kind)(b->flags & CAML_BA_KIND_MASK)) {
  case CAML_BA_FLOAT16:
    ((uint16 *) b->data)[offset] =
      caml_float_to_float16(Double_val(newval)); break;
  case CAML_BA_FLOAT32:
    ((float *) b->data)[offset] = Double_val(newval); break;
  case CAML_BA_FLOAT64:
    ((double *) b->data)[offset] = Double_val(newval); break;
  case CAML_BA_CHAR:
  case CAML_BA_SINT8:
  case CAML_BA_UINT8:
    ((int8 *) b->data)[offset] = Int_val(newval); break;
  case CAML_BA_SINT16:
  case CAML_BA_UINT16:
    ((int16 *) b->data)[offset] = Int_val(newval); break;
  case CAML_BA_INT32:
    ((int32_t *) b->data)[offset] = Int32_val(newval); break;
  case CAML_BA_INT64:
    ((int64_t *) b->data)[offset] = Int64_val(newval); break;
  case CAML_BA_NATIVE_INT:
    ((intnat *) b->data)[offset] = Nativeint_val(newval); break;
  case CAML_BA_CAML_INT:
    ((intnat *) b->data)[offset] = Long_val(newval); break;
  case CAML_BA_COMPLEX32:
    { float * p = ((float *) b->data) + offset * 2;
      p[0] = Double_flat_field(newval, 0);
      p[1] = Double_flat_field(newval, 1);
      break; }
  case CAML_BA_COMPLEX64:
    { double * p = ((double *) b->data) + offset * 2;
      p[0] = Double_flat_field(newval, 0);
      p[1] = Double_flat_field(newval, 1);
      break; }
  default: CAMLunreachable();
  }
  return Val_unit;
}

CAMLprim value caml_ba_set_1(value vb, value vind1, value newval)
{
  return caml_ba_set_aux(vb, &vind1, 1, newval);
}

CAMLprim value caml_ba_set_2(value vb, value vind1, value vind2, value newval)
{
  value vind[2];
  vind[0] = vind1; vind[1] = vind2;
  return caml_ba_set_aux(vb, vind, 2, newval);
}

CAMLprim value caml_ba_set_3(value vb, value vind1, value vind2, value vind3,
                     value newval)
{
  value vind[3];
  vind[0] = vind1; vind[1] = vind2; vind[2] = vind3;
  return caml_ba_set_aux(vb, vind, 3, newval);
}

value caml_ba_set_N(value vb, value * vind, int nargs)
{
  return caml_ba_set_aux(vb, vind, nargs - 1, vind[nargs - 1]);
}

CAMLprim value caml_ba_set_generic(value vb, value vind, value newval)
{
  return caml_ba_set_aux(vb, &Field(vind, 0), Wosize_val(vind), newval);
}

CAMLprim value caml_ba_uint8_set16(value vb, value vind, value newval)
{
  unsigned char b1, b2;
  intnat val;
  intnat idx = Long_val(vind);
  struct caml_ba_array * b = Caml_ba_array_val(vb);
  if (idx < 0 || idx >= b->dim[0] - 1) caml_array_bound_error();
  val = Long_val(newval);
#ifdef ARCH_BIG_ENDIAN
  b1 = 0xFF & val >> 8;
  b2 = 0xFF & val;
#else
  b2 = 0xFF & val >> 8;
  b1 = 0xFF & val;
#endif
  ((unsigned char*) b->data)[idx] = b1;
  ((unsigned char*) b->data)[idx+1] = b2;
  return Val_unit;
}

CAMLprim value caml_ba_uint8_set32(value vb, value vind, value newval)
{
  unsigned char b1, b2, b3, b4;
  intnat idx = Long_val(vind);
  intnat val;
  struct caml_ba_array * b = Caml_ba_array_val(vb);
  if (idx < 0 || idx >= b->dim[0] - 3) caml_array_bound_error();
  val = Int32_val(newval);
#ifdef ARCH_BIG_ENDIAN
  b1 = 0xFF & val >> 24;
  b2 = 0xFF & val >> 16;
  b3 = 0xFF & val >> 8;
  b4 = 0xFF & val;
#else
  b4 = 0xFF & val >> 24;
  b3 = 0xFF & val >> 16;
  b2 = 0xFF & val >> 8;
  b1 = 0xFF & val;
#endif
  ((unsigned char*) b->data)[idx] = b1;
  ((unsigned char*) b->data)[idx+1] = b2;
  ((unsigned char*) b->data)[idx+2] = b3;
  ((unsigned char*) b->data)[idx+3] = b4;
  return Val_unit;
}

CAMLprim value caml_ba_uint8_set64(value vb, value vind, value newval)
{
  unsigned char b1, b2, b3, b4, b5, b6, b7, b8;
  intnat idx = Long_val(vind);
  int64_t val;
  struct caml_ba_array * b = Caml_ba_array_val(vb);
  if (idx < 0 || idx >= b->dim[0] - 7) caml_array_bound_error();
  val = Int64_val(newval);
#ifdef ARCH_BIG_ENDIAN
  b1 = 0xFF & val >> 56;
  b2 = 0xFF & val >> 48;
  b3 = 0xFF & val >> 40;
  b4 = 0xFF & val >> 32;
  b5 = 0xFF & val >> 24;
  b6 = 0xFF & val >> 16;
  b7 = 0xFF & val >> 8;
  b8 = 0xFF & val;
#else
  b8 = 0xFF & val >> 56;
  b7 = 0xFF & val >> 48;
  b6 = 0xFF & val >> 40;
  b5 = 0xFF & val >> 32;
  b4 = 0xFF & val >> 24;
  b3 = 0xFF & val >> 16;
  b2 = 0xFF & val >> 8;
  b1 = 0xFF & val;
#endif
  ((unsigned char*) b->data)[idx] = b1;
  ((unsigned char*) b->data)[idx+1] = b2;
  ((unsigned char*) b->data)[idx+2] = b3;
  ((unsigned char*) b->data)[idx+3] = b4;
  ((unsigned char*) b->data)[idx+4] = b5;
  ((unsigned char*) b->data)[idx+5] = b6;
  ((unsigned char*) b->data)[idx+6] = b7;
  ((unsigned char*) b->data)[idx+7] = b8;
  return Val_unit;
}

/* Return the number of dimensions of a big array */

CAMLprim value caml_ba_num_dims(value vb)
{
  struct caml_ba_array * b = Caml_ba_array_val(vb);
  return Val_long(b->num_dims);
}

/* Return the n-th dimension of a big array */

CAMLprim value caml_ba_dim(value vb, value vn)
{
  struct caml_ba_array * b = Caml_ba_array_val(vb);
  intnat n = Long_val(vn);
  if (n < 0 || n >= b->num_dims) caml_invalid_argument("Bigarray.dim");
  return Val_long(b->dim[n]);
}

CAMLprim value caml_ba_dim_1(value vb)
{
  return caml_ba_dim(vb, Val_int(0));
}

CAMLprim value caml_ba_dim_2(value vb)
{
  return caml_ba_dim(vb, Val_int(1));
}

CAMLprim value caml_ba_dim_3(value vb)
{
  return caml_ba_dim(vb, Val_int(2));
}

/* Return the kind of a big array */

CAMLprim value caml_ba_kind(value vb)
{
  return Val_caml_ba_kind(Caml_ba_array_val(vb)->flags & CAML_BA_KIND_MASK);
}

/* Return the layout of a big array */

CAMLprim value caml_ba_layout(value vb)
{
  int layout = Caml_ba_array_val(vb)->flags & CAML_BA_LAYOUT_MASK;
  return Val_caml_ba_layout(layout);
}

/* Create / update proxy to indicate that b2 is a sub-array of b1 */

static void caml_ba_update_proxy(struct caml_ba_array * b1,
                                 struct caml_ba_array * b2)
{
  struct caml_ba_proxy * proxy;
  /* Nothing to do for un-managed arrays */
  if ((b1->flags & CAML_BA_MANAGED_MASK) == CAML_BA_EXTERNAL) return;
  if (b1->proxy != NULL) {
    /* If b1 is already a proxy for a larger array, increment refcount of
       proxy */
    b2->proxy = b1->proxy;
    caml_atomic_refcount_incr(&b1->proxy->refcount);
  } else {
    /* Otherwise, create proxy and attach it to both b1 and b2 */
    proxy = malloc(sizeof(struct caml_ba_proxy));
    if (proxy == NULL) caml_raise_out_of_memory();
    caml_atomic_refcount_init(&proxy->refcount, 2);
    /* initial refcount: 2 = original array + sub array */
    proxy->data = b1->data;
    proxy->size =
      b1->flags & CAML_BA_MAPPED_FILE ? caml_ba_byte_size(b1) : 0;
    b1->proxy = proxy;
    b2->proxy = proxy;
  }
}

/* Slicing */

CAMLprim value caml_ba_slice(value vb, value vind)
{
  CAMLparam2 (vb, vind);
  #define b (Caml_ba_array_val(vb))
  CAMLlocal1 (res);
  intnat index[CAML_BA_MAX_NUM_DIMS];
  int num_inds;
  intnat offset;
  intnat * sub_dims;
  char * sub_data;

  /* Check number of indices <= number of dimensions of array */
  num_inds = Wosize_val(vind);
  if (num_inds > b->num_dims)
    caml_invalid_argument("Bigarray.slice: too many indices");
  /* Compute offset and check bounds */
  switch ((enum caml_ba_layout)(b->flags & CAML_BA_LAYOUT_MASK)) {
  case CAML_BA_C_LAYOUT: {
    int i;
    /* We slice from the left */
    for (i = 0; i < num_inds; i++) index[i] = Long_val(Field(vind, i));
    for (/*nothing*/; i < b->num_dims; i++) index[i] = 0;
    offset = caml_ba_offset(b, index);
    sub_dims = b->dim + num_inds;
    break;
  }
  case CAML_BA_FORTRAN_LAYOUT:
    /* We slice from the right */
    for (int i = 0; i < num_inds; i++)
      index[b->num_dims - num_inds + i] = Long_val(Field(vind, i));
    for (int i = 0; i < b->num_dims - num_inds; i++) index[i] = 1;
    offset = caml_ba_offset(b, index);
    sub_dims = b->dim;
    break;
  default: CAMLunreachable();
  }
  sub_data =
    (char *) b->data +
    offset * caml_ba_element_size[b->flags & CAML_BA_KIND_MASK];
  /* Allocate an OCaml bigarray to hold the result */
  res = caml_ba_alloc(b->flags | CAML_BA_SUBARRAY,
                      b->num_dims - num_inds, sub_data, sub_dims);
  /* Copy the finalization function from the original array (PR#8568) */
  Custom_ops_val(res) = Custom_ops_val(vb);
  /* Create or update proxy in case of managed bigarray */
  caml_ba_update_proxy(b, Caml_ba_array_val(res));
  /* Return result */
  CAMLreturn (res);

  #undef b
}

/* Changing the layout of an array (memory is shared) */

CAMLprim value caml_ba_change_layout(value vb, value vlayout)
{
  CAMLparam2 (vb, vlayout);
  CAMLlocal1 (res);
  #define b (Caml_ba_array_val(vb))
  /* if the layout is different, change the flags and reverse the dimensions */
  if (Caml_ba_layout_val(vlayout) != (b->flags & CAML_BA_LAYOUT_MASK)) {
    /* change the flags to reflect the new layout */
    int flags = (b->flags & (CAML_BA_KIND_MASK | CAML_BA_MANAGED_MASK))
                 | Caml_ba_layout_val(vlayout);
    /* reverse the dimensions */
    intnat new_dim[CAML_BA_MAX_NUM_DIMS];
    for (unsigned int i = 0; i < b->num_dims; i++)
      new_dim[i] = b->dim[b->num_dims - i - 1];
    res = caml_ba_alloc(flags | CAML_BA_SUBARRAY,
                        b->num_dims, b->data, new_dim);
    /* Copy the finalization function from the original array (PR#8568) */
    Custom_ops_val(res) = Custom_ops_val(vb);
    caml_ba_update_proxy(b, Caml_ba_array_val(res));
    CAMLreturn(res);
  } else {
    /* otherwise, do nothing */
    CAMLreturn(vb);
  }
  #undef b
}


/* Extracting a sub-array of same number of dimensions */

CAMLprim value caml_ba_sub(value vb, value vofs, value vlen)
{
  CAMLparam3 (vb, vofs, vlen);
  CAMLlocal1 (res);
  #define b (Caml_ba_array_val(vb))
  intnat ofs = Long_val(vofs);
  intnat len = Long_val(vlen);
  int changed_dim;
  intnat mul;
  char * sub_data;

  /* Compute offset and check bounds */
  switch ((enum caml_ba_layout)(b->flags & CAML_BA_LAYOUT_MASK)) {
  case CAML_BA_C_LAYOUT:
    /* We reduce the first dimension */
    mul = 1;
    for (int i = 1; i < b->num_dims; i++) mul *= b->dim[i];
    changed_dim = 0;
    break;
  case CAML_BA_FORTRAN_LAYOUT:
    /* We reduce the last dimension */
    mul = 1;
    for (int i = 0; i < b->num_dims - 1; i++) mul *= b->dim[i];
    changed_dim = b->num_dims - 1;
    ofs--;                      /* Fortran arrays start at 1 */
    break;
  default: CAMLunreachable();
  }
  if (ofs < 0 || len < 0 || ofs + len > b->dim[changed_dim])
    caml_invalid_argument("Bigarray.sub: bad sub-array");
  sub_data =
    (char *) b->data +
    ofs * mul * caml_ba_element_size[b->flags & CAML_BA_KIND_MASK];
  /* Allocate an OCaml bigarray to hold the result */
  res = caml_ba_alloc(b->flags | CAML_BA_SUBARRAY,
                      b->num_dims, sub_data, b->dim);
  /* Copy the finalization function from the original array (PR#8568) */
  Custom_ops_val(res) = Custom_ops_val(vb);
  /* Doctor the changed dimension */
  Caml_ba_array_val(res)->dim[changed_dim] = len;
  /* Create or update proxy in case of managed bigarray */
  caml_ba_update_proxy(b, Caml_ba_array_val(res));
  /* Return result */
  CAMLreturn (res);

  #undef b
}

/* Copying a big array into another one */

#define LEAVE_RUNTIME_OP_CUTOFF 4096
#define is_mmapped(ba) ((ba)->flags & CAML_BA_MAPPED_FILE)

CAMLprim value caml_ba_blit(value vsrc, value vdst)
{
  CAMLparam2(vsrc, vdst);
  struct caml_ba_array * src = Caml_ba_array_val(vsrc);
  struct caml_ba_array * dst = Caml_ba_array_val(vdst);
  void *src_data = src->data;
  void *dst_data = dst->data;
  intnat num_bytes;
  int leave_runtime;

  /* Check same numbers of dimensions and same dimensions */
  if (src->num_dims != dst->num_dims) goto blit_error;
  for (int i = 0; i < src->num_dims; i++)
    if (src->dim[i] != dst->dim[i]) goto blit_error;
  /* Compute number of bytes in array data */
  num_bytes =
    caml_ba_num_elts(src)
    * caml_ba_element_size[src->flags & CAML_BA_KIND_MASK];
  leave_runtime =
    (
      (num_bytes >= LEAVE_RUNTIME_OP_CUTOFF*sizeof(long))
      || is_mmapped(src)
      || is_mmapped(dst)
    );
  /* Do the copying */
  if (leave_runtime) caml_enter_blocking_section();
  memmove (dst_data, src_data, num_bytes);
  if (leave_runtime) caml_leave_blocking_section();
  CAMLreturn (Val_unit);
 blit_error:
  caml_invalid_argument("Bigarray.blit: dimension mismatch");
  CAMLreturn (Val_unit);              /* not reached */
}

/* Filling a big array with a given value */

#define FILL_GEN_LOOP(n_ops, loop) do{                                       \
  int leave_runtime = ((n_ops >= LEAVE_RUNTIME_OP_CUTOFF) || is_mmapped(b)); \
  if (leave_runtime) caml_enter_blocking_section();                          \
  loop;                                                                      \
  if (leave_runtime) caml_leave_blocking_section();                          \
}while(0)

#define FILL_SCALAR_LOOP(T)                                     \
  FILL_GEN_LOOP(num_elts,                                       \
    for (T p = data; num_elts > 0; p++, num_elts--) *p = init)

#define FILL_COMPLEX_LOOP(T)                                                 \
  FILL_GEN_LOOP(num_elts + num_elts,                                         \
    for (T p = data; num_elts > 0; num_elts--) { *p++ = init0; *p++ = init1; })

CAMLprim value caml_ba_fill(value vb, value vinit)
{
  CAMLparam1(vb);
  struct caml_ba_array * b = Caml_ba_array_val(vb);
  void *data = b->data;
  intnat num_elts = caml_ba_num_elts(b);

  switch ((enum caml_ba_kind)(b->flags & CAML_BA_KIND_MASK)) {
  case CAML_BA_FLOAT16: {
    uint16 init = caml_float_to_float16(Double_val(vinit));
    FILL_SCALAR_LOOP(uint16 *);
    break;
  }
  case CAML_BA_FLOAT32: {
    float init = Double_val(vinit);
    FILL_SCALAR_LOOP(float *);
    break;
  }
  case CAML_BA_FLOAT64: {
    double init = Double_val(vinit);
    FILL_SCALAR_LOOP(double *);
    break;
  }
  case CAML_BA_CHAR:
  case CAML_BA_SINT8:
  case CAML_BA_UINT8: {
    int init = Int_val(vinit);
    FILL_SCALAR_LOOP(unsigned char *);
    break;
  }
  case CAML_BA_SINT16:
  case CAML_BA_UINT16: {
    int init = Int_val(vinit);
    FILL_SCALAR_LOOP(int16 *);
    break;
  }
  case CAML_BA_INT32: {
    int32_t init = Int32_val(vinit);
    FILL_SCALAR_LOOP(int32_t *);
    break;
  }
  case CAML_BA_INT64: {
    int64_t init = Int64_val(vinit);
    FILL_SCALAR_LOOP(int64_t *);
    break;
  }
  case CAML_BA_NATIVE_INT: {
    intnat init = Nativeint_val(vinit);
    FILL_SCALAR_LOOP(intnat *);
    break;
  }
  case CAML_BA_CAML_INT: {
    intnat init = Long_val(vinit);
    FILL_SCALAR_LOOP(intnat *);
    break;
  }
  case CAML_BA_COMPLEX32: {
    float init0 = Double_flat_field(vinit, 0);
    float init1 = Double_flat_field(vinit, 1);
    FILL_COMPLEX_LOOP(float *);
    break;
  }
  case CAML_BA_COMPLEX64: {
    double init0 = Double_flat_field(vinit, 0);
    double init1 = Double_flat_field(vinit, 1);
    FILL_COMPLEX_LOOP(double *);
    break;
  }
  default: CAMLunreachable();
  }
  CAMLreturn (Val_unit);
}

/* Reshape an array: change dimensions and number of dimensions, preserving
   array contents */

CAMLprim value caml_ba_reshape(value vb, value vdim)
{
  CAMLparam2 (vb, vdim);
  CAMLlocal1 (res);
#define b (Caml_ba_array_val(vb))
  intnat dim[CAML_BA_MAX_NUM_DIMS];
  mlsize_t num_dims;
  uintnat num_elts;

  num_dims = Wosize_val(vdim);
  /* here num_dims is unsigned (mlsize_t) so no need to check (num_dims >= 0) */
  if (num_dims > CAML_BA_MAX_NUM_DIMS)
    caml_invalid_argument("Bigarray.reshape: bad number of dimensions");
  num_elts = 1;
  for (int i = 0; i < num_dims; i++) {
    dim[i] = Long_val(Field(vdim, i));
    if (dim[i] < 0)
      caml_invalid_argument("Bigarray.reshape: negative dimension");
    num_elts *= dim[i];
  }
  /* Check that sizes agree */
  if (num_elts != caml_ba_num_elts(b))
    caml_invalid_argument("Bigarray.reshape: size mismatch");
  /* Create bigarray with same data and new dimensions */
  res = caml_ba_alloc(b->flags | CAML_BA_SUBARRAY, num_dims, b->data, dim);
  /* Copy the finalization function from the original array (PR#8568) */
  Custom_ops_val(res) = Custom_ops_val(vb);
  /* Create or update proxy in case of managed bigarray */
  caml_ba_update_proxy(b, Caml_ba_array_val(res));
  /* Return result */
  CAMLreturn (res);

#undef b
}