Loading...
cudaErrorStreamCaptureUnjoined (904)cudaErrorStreamCaptureUnjoined occurs when forked streams not joined before EndCapture.
CUDA error: stream capture unjoined cudaErrorStreamCaptureUnjoined
Wait on all.
cudaEventRecord(e, s2);
cudaStreamWaitEvent(s1, e); // Join
cudaStreamEndCapture(s1, &graph);Unjoined.
cudaStreamWaitEvent(s2, e); // Fork
kernel<<<g,b,0,s2>>>();
cudaStreamEndCapture(s1, &g); // s2 not joined!All joined.
cudaEventRecord(e, s2); cudaStreamWaitEvent(s1, e); // Join
cudaStreamEndCapture(s1, &g);Yes, fork and join creates parallel nodes.
Need help debugging CUDA errors? Download RightNow AI for intelligent error analysis and optimization suggestions.