nx_parexec.nx

nx_parexec.nx -- ACTUAL PARALLEL multi-core execution (closes distributed-compute's last blocked half). The "1-core threading gap" was inherited, not measured: nx_syscalls has had sys_fork/sys_wait4/ sys_mmap_shared all along (hundreds of organs fork; nx_arbiter_race_gate proves shared-sink races). So the distexec partition->merge algebra runs for REAL in parallel: the parent allocates a SHARED result region, forks one child PER PARTITION (the kernel schedules them across cores), each child computes its partition's partial into its OWN shared slot (disjoint writes => no locks needed), the parent wait4()s all and merges. From the first byte up: raw fork, no pthreads, no libc, no shell. Correctness is byte- identical to the sequential run (same algebra); the win is wall-clock, MEASURED live on the box's cores. license_tier: ORIGINAL No hardware writes (Rule 26).