lib/libc/emscripten_memcpy.c

/*
 * A simple memcpy optimized for wasm.
 */

#include <stdint.h>
#include <string.h>

// An external JS implementation that is efficient for very large copies, using
// HEAPU8.set()
extern void *emscripten_memcpy_big(void *restrict dest, const void *restrict src, size_t n);

// XXX EMSCRIPTEN ASAN: build an uninstrumented version of memcpy
#if defined(__EMSCRIPTEN__) && defined(__has_feature)
#if __has_feature(address_sanitizer)
#define memcpy __attribute__((no_sanitize("address"))) emscripten_builtin_memcpy
#endif
#endif

void *memcpy(void *restrict dest, const void *restrict src, size_t n)
{
  unsigned char *d = dest;
  const unsigned char *s = src;

  unsigned char *aligned_d_end;
  unsigned char *block_aligned_d_end;
  unsigned char *d_end;

  if (n >= 8192) {
    emscripten_memcpy_big(dest, src, n);
    return dest;
  }

  d_end = d + n;
  if ((((uintptr_t)d) & 3) == (((uintptr_t)s) & 3)) {
    // The initial unaligned < 4-byte front.
    while ((((uintptr_t)d) & 3) && d < d_end) {
      *d++ = *s++;
    }
    aligned_d_end = (unsigned char *)(((uintptr_t)d_end) & -4);
    if (((uintptr_t)aligned_d_end) >= 64) {
      block_aligned_d_end = aligned_d_end - 64;
      while (d <= block_aligned_d_end) {
        // TODO: we could use 64-bit ops here, but we'd need to make sure the
        //       alignment is 64-bit, which might cost us
        *(((uint32_t*)d)) = *(((uint32_t*)s));
        *(((uint32_t*)d) + 1) = *(((uint32_t*)s) + 1);
        *(((uint32_t*)d) + 2) = *(((uint32_t*)s) + 2);
        *(((uint32_t*)d) + 3) = *(((uint32_t*)s) + 3);
        *(((uint32_t*)d) + 4) = *(((uint32_t*)s) + 4);
        *(((uint32_t*)d) + 5) = *(((uint32_t*)s) + 5);
        *(((uint32_t*)d) + 6) = *(((uint32_t*)s) + 6);
        *(((uint32_t*)d) + 7) = *(((uint32_t*)s) + 7);
        *(((uint32_t*)d) + 8) = *(((uint32_t*)s) + 8);
        *(((uint32_t*)d) + 9) = *(((uint32_t*)s) + 9);
        *(((uint32_t*)d) + 10) = *(((uint32_t*)s) + 10);
        *(((uint32_t*)d) + 11) = *(((uint32_t*)s) + 11);
        *(((uint32_t*)d) + 12) = *(((uint32_t*)s) + 12);
        *(((uint32_t*)d) + 13) = *(((uint32_t*)s) + 13);
        *(((uint32_t*)d) + 14) = *(((uint32_t*)s) + 14);
        *(((uint32_t*)d) + 15) = *(((uint32_t*)s) + 15);
        d += 64;
        s += 64;
      }
    }
    while (d < aligned_d_end) {
      *((uint32_t *)d) = *((uint32_t *)s);
      d += 4;
      s += 4;
    }
  } else {
    // In the unaligned copy case, unroll a bit as well.
    if (((uintptr_t)d_end) >= 4) {
      aligned_d_end = d_end - 4;
      while (d <= aligned_d_end) {
        *d = *s;
        *(d + 1) = *(s + 1);
        *(d + 2) = *(s + 2);
        *(d + 3) = *(s + 3);
        d += 4;
        s += 4;
      }
    }
  }
  // The remaining unaligned < 4 byte tail.
  while (d < d_end) {
    *d++ = *s++;
  }
  return dest;


  for (; n; n--) *d++ = *s++;
  return dest;
}