[hardware] Shuffle suppress fix

[antellopeTweet]

fix: Shuffle suppressed when required for whole reg read-write

Shuffle suppression previously occured when: (full vd overwrite) and (vd is a source register) and (vd needs a re-shuffle)

Previous implementation identified that source was also vd so suppressed source reshuffle, relying on destination reshuffle, but if vd was fully overwritten it also suppressed destination reshuffle, resulting in no shuffle

For example in the following assembly sequence:

vsetvli zero, zero, e64, m8, ta, ma
vl8re64.v v8, (lhs) // whole register loads always use ew8
vl8re64.v v16, (rhs)
vfsub.vv v8, v8, v16 // v8 is used as source, totally overwritten destination, and needs a shuffle
vs8r.v v8, (dst)

Changelog

Fixed

• Fix shuffle suppression to not suppress reshuffle if destination is 1) totally overwritten 2) used as a source 3) needs to be reshuffled

Checklist

• Automated tests pass
• Changelog updated
• Code style guideline is observed