[XLA:GPU] Emit MOF tuples using just one thread.
Previously all threads in a multi-output fusion would emit the tuple pointers, now just one thread does it. PiperOrigin-RevId: 224043124
Loading
Please sign in to comment
Previously all threads in a multi-output fusion would emit the tuple pointers, now just one thread does it. PiperOrigin-RevId: 224043124