HBO043%DJUKFA11.BITNET@cunyvm.cuny.edu (Christoph van Wuellen) (03/11/91)
I learned that the 68000 reads the operand which it overwrites with zeroes later upon a clr instruction. This may have unwanted effects when programming interface registers -- so I will withdraw the clr instruction from the peephole optimizer of c68. There is only a single case where clr is faster than move #0. this is clr.b or clr.w Dn. There are cases where clr is slower: clr.w -(An) and clr.l -(An) The advantage of the clr instruction is that it is shorter. The next c68 patch will generate only clr.b and clr.w to data registers. C.v.W.