explicit threads_stack_limit (for recent Poly/ML SVN versions), which leads to soft interrupt instead of exhaustion of virtual memory, which is particularly relevant for the bigger address space of x86_64;
(* Title: Pure/ML-Systems/maximum_ml_stack_polyml-5.5.3.ML
Maximum stack size (in words) for ML threads -- Poly/ML 5.5.3, or later.
*)
fun maximum_ml_stack limit = [Thread.MaximumMLStack limit];