Use LocalAliasAnalysis to improve handling of side effects in nested scf.parallel. If the written memory outside nested scf.parallel is not alias to the memory accessed inside the nested loop, we can convert it to gpu.launch.
Use LocalAliasAnalysis to improve handling of side effects in nested scf.parallel. If the written memory outside nested scf.parallel is not alias to the memory accessed inside the nested loop, we can convert it to gpu.launch.