SiFive P870 (Lanxin LX500, Epic Semi Contrail AIx) uarch-tool benchmarks

VLEN: 128

Detect all1s tail/mask policy with simple code snippet:
Tail agnostic policy: all1s
Mask agnostic policy: all1s
Is vl always set to min(AVL,VLMAX): yes
    Note: spec allows ceil(AVL/2)<=vl<=VLMAX for VLMAX<AVL<2*VLMAX

Measures overhead of reinterpreting a mask as a vector:
A) reinterpret:       7.0018000 cycles/iter
B) don't reinterpret: 6.9998000 cycles/iter

• There seems to be no overhead in reinterpreting a mask register.