memcpy(), memcpy_isr()

Updated: April 19, 2023

Copy bytes from one buffer to another


#include <string.h>

void* memcpy( void* dst,
              const void* src,
              size_t length );

void* memcpy_isr( void* dst,
                  const void* src,
                  size_t length );


A pointer to where you want the function to copy the data.
A pointer to the buffer that you want to copy data from.
The number of bytes to copy.



Use the -l c option to qcc to link against this library. This library is usually included automatically.


The memcpy() function copies length bytes from the buffer pointed to by src into the buffer pointed to by dst.

The memcpy_isr() function does the same thing, but avoids using special-purpose registers (e.g., sse2) that incur additional cost when their state is saved during context switches. The relative performance of a process using this first or second function can be better or worse depending on the frequency and sizes of memory-copying operations, the specific target, and what other processes are doing. If memory copying does not play a major role in the process's overall performance, which function is faster isn't so important. Otherwise, developers are strongly encouraged to do their own testing and select the correct function.

In this release, both memcpy() and memcpy_isr() are safe to call from an interrupt handler.

Note: Copying overlapping buffers isn't guaranteed to work; use memmove() to copy buffers that overlap.


A pointer to the destination buffer (that is, the value of dst).


#include <stdio.h>
#include <string.h>
#include <stdlib.h>

int main( void )
    char buffer[80];

    memcpy( buffer, "Hello", 5 );
    buffer[5] = '\0';
    printf( "%s\n", buffer );
    return EXIT_SUCCESS;

Environment variables:

On certain targets, you can use this environment variable to select the implementation of memcpy(). It doesn't affect the implementation of memcpy_isr(). The value is one of the strings given below.
  • for AArch64 targets:
    • aarch64_neon — optimized for AARCH64 targets using NEON
    • generic — the default
  • for ARMv7 targets:
    • cortex_a9 — optimized for the ARM Cortex-A9 processor; assumes that no unaligned access is supported
    • cortex_a9_aligned — optimized for ARM Cortex-A9; requires that unaligned memory access be enabled on the platform. If memory access is misaligned, this implementation falls back to the NEON version.
    • cortex_a9_neon — optimized for ARM Cortex-A9 using NEON
    • generic — the default
    • krait — optimized for the Qualcomm Krait CPU
    • krait_neon — optimized for Qualcomm Krait using NEON

Processes that register ISRs shouldn't use the NEON versions.


memcpy() is ANSI, POSIX 1003.1; memcpy_isr() is QNX Neutrino.

Cancellation point No
Interrupt handler Yes
Signal handler Yes
Thread Yes