STEPS_BIT
p->steps[p->nsteps / STEPS_BIT] |=
(uintptr_t)1 << (p->nsteps % STEPS_BIT);
p->steps[p->nsteps / STEPS_BIT] &=
~((uintptr_t)1 << (p->nsteps % STEPS_BIT));
p->steps[0] = (p->steps[0] >> 1) | (p->steps[1] << (STEPS_BIT - 1));