On 09/26/2012 11:33 PM, Paolo Bonzini wrote: > Should you make tcg/optimize.c prefer "movcond a, a, b" to "movcond a, > b, a", similar to commit c2b0e2f (tcg/optimize: prefer the "op a, a, b" > form for commutative ops, 2012-09-19)?
I explicitly went with "a, b, a" because I find it is easier to reason with positive logic (move-if-true) than negative logic. r~