cudaError_t cudaMemcpy (void *p1, const void *p2, size_t p3, enum cudaMemcpyKind p4)
{
cudaError_t res;
#if defined(DEBUG)
fprintf (stderr, PACKAGE_NAME": THREAD %d cudaMemcpy is at %p\n", THREADID, real_cudaMemcpy);
fprintf (stderr, PACKAGE_NAME": THREAD %d cudaMemcpy params %p %p %d %d\n", THREADID, p1, p2, p3, p4);
#endif
if (real_cudaMemcpy != NULL && mpitrace_on && Extrae_get_trace_CUDA())
{
Extrae_cudaMemcpy_Enter (p1, p2, p3, p4);
res = real_cudaMemcpy (p1, p2, p3, p4);
Extrae_cudaMemcpy_Exit ();
}
else if (real_cudaMemcpy != NULL && !(mpitrace_on && Extrae_get_trace_CUDA()))
{
res = real_cudaMemcpy (p1, p2, p3, p4);
}
else
{
fprintf (stderr, "Unable to find cudaMemcpy in DSOs!! Dying...\n");
exit (0);
}
return res;
}
cudaError_t cudaMemcpy (void *dst, const void *src,
size_t count, enum cudaMemcpyKind kind)
{
cudaError_t retcode = real_cudaMemcpy(dst, src, count, kind);
handle_cudaMemcpy(dst, src, count, kind, (void *) -1);
return retcode;
}
cudaError_t cudaMemcpy(
void *dst,
void *src,
size_t size,
enum cudaMemcpyKind kind)
{
CUDATRACE_INTERPOSE(cudaMemcpy);
fprintf(stderr, "TRACE: cudaMemcpy %p %p %lu %d\n",
dst, src, size, kind);
return real_cudaMemcpy(
dst, src, size, kind);
}
int cuda_getBufferContent (struct ld_mem_s *ldBuffer, void *buffer,
size_t offset, size_t size)
{
cudaError_t err = cudaSuccess;
if (offset != 0) {
/* need to use the type to multiply correctly the buffer handle. */
warning("offset == %zu != 0, not handled yet in Cuda's ldChecker\n",
offset);
}
err = real_cudaMemcpy (buffer,
ldBuffer->handle,
size,
cudaMemcpyDeviceToHost);
return err == cudaSuccess;
}
