cudaError_t cudaMemcpyAsync (void *p1, const void *p2, size_t p3, enum cudaMemcpyKind p4, cudaStream_t p5)
{
cudaError_t res;
#if defined(DEBUG)
fprintf (stderr, PACKAGE_NAME": THREAD %d cudaMemcpyAsync is at %p\n", THREADID, real_cudaMemcpyAsync);
fprintf (stderr, PACKAGE_NAME": THREAD %d cudaMemcpyAsync params %p %p %d %d %d\n", THREADID, p1, p2, p3, p4, p5);
#endif
if (real_cudaMemcpyAsync != NULL && mpitrace_on && Extrae_get_trace_CUDA())
{
Extrae_cudaMemcpyAsync_Enter (p1, p2, p3, p4, p5);
res = real_cudaMemcpyAsync (p1, p2, p3, p4, p5);
Extrae_cudaMemcpyAsync_Exit ();
}
else if (real_cudaMemcpyAsync != NULL && !(mpitrace_on && Extrae_get_trace_CUDA()))
{
res = real_cudaMemcpyAsync (p1, p2, p3, p4, p5);
}
else
{
fprintf (stderr, "Unable to find cudaMemcpyAsync in DSOs!! Dying...\n");
exit (0);
}
return res;
}
cudaError_t cudaMemcpyAsync(void *dst, const void * src, size_t count,
enum cudaMemcpyKind kind, cudaStream_t stream)
{
cudaError_t retcode = real_cudaMemcpyAsync(dst, src, count, kind, stream);
handle_cudaMemcpy(dst, src, count, kind, stream);
return retcode;
}
cudaError_t cudaMemcpyAsync(
void *dst,
void *src,
size_t size,
enum cudaMemcpyKind kind,
cudaStream_t stream)
{
CUDATRACE_INTERPOSE(cudaMemcpyAsync);
fprintf(stderr, "TRACE: cudaMemcpyAsync %p %p %lu %d %p\n",
dst, src, size, kind, stream);
return real_cudaMemcpyAsync(
dst, src, size, kind, stream);
}
