Using memcpy on device causes an error : Signal: caught in PU_adjust_addr_flags phase.

Accelerated Computing CUDA CUDA Programming and Performance
1

Hello,

I have found no mention of possibility or impossibility of using standard memcpy() on device code.

So, when I try to use it such way:

[codebox]#include <stdio.h>

#include <cuda.h>

global void decodeOnDevice(char *a) {

char b[MAX_SIZE];

int i;

memcpy (b, a, 16);

for (i = 0; i < 16; i++)

  b[i] = b[i] ^ 0xFF;

memcpy (a,b, 16);

}

int main(void) {

char *temp_host; // pointers to host memory

char *temp_device; // pointers to device memory

char commands[2][MAX_SIZE];

int i;

// allocate arrays on host

temp_host = (char *)malloc(MAX_SIZE);

// allocate arrays on device

cudaMalloc((void **) &temp_device, MAX_SIZE);

// initialize host data

memset(commands[0], 0, MAX_SIZE);

// these are the encoded commands

memcpy(commands[0], “\xB7\x9A\x93\x93\x90\xDF\xBC\xAA\xBB\xBE\xDF\xA8\x90\x8D\x93\x9B”, strlen(“Hello CUDA World!”));

i = 0;

memset(temp_host, 0, MAX_SIZE);

memcpy(temp_host, commands[i], strlen(commands[i]));

// send data from host to device

cudaMemcpy(temp_device, temp_host, MAX_SIZE, cudaMemcpyHostToDevice);

// data copied on device, invoking kernel

decodeOnDevice <<< 1, 1 >>> (temp_device);

// retrieve data from device

cudaMemcpy(temp_host, temp_device, MAX_SIZE, cudaMemcpyDeviceToHost);

// execute the decoded command

 printf (temp_host);

}[/codebox]

I’ve got the error:

tmpxft_00000c48_00000000-8_mal.cudafe2.gpu

Signal: caught in PU_adjust_addr_flags phase.

(0): Error: Signal caught in phase PU_adjust_addr_flags – processing aborted

nvopencc ERROR: D:\Visual\CUDA\bin/…/open64/lib//be.exe returned non-zero status 3

If I try to write _memcpy emulate function, it’s ok:

[codebox]device void * _memcpy (char *to, char *from, size_t len);

global void decodeOnDevice(char *a) {

char b[MAX_SIZE];

int i;

_memcpy (b, a, 16);

for (i = 0; i < 16; i++)

  b[i] = b[i] ^ 0xFF;

_memcpy (a,b, 16);

}

device void * _memcpy (char *to, char *from, size_t len)

{

size_t i;

for (i=0; i< len; i++) to[i] = from[i];

}

[/codebox]

It’s a compiler bug? I can’t use memcpy() on device, can I?

CUDA 2.1, WinXP 32 bit, SDK 2.1