firth@sei.cmu.edu (Robert Firth) (02/16/88)
In article <2303@umd5.umd.edu> chris@trantor.umd.edu (Chris Torek) writes: >Without further supporting evidence, I will claim that any optimiser >could convert > > a = 0 & f(); > >into > > (void) f(), a = 0 You now have supporting evidence. I just compiled a := 0 & f() on my old familiar BCPL/Vax compiler, and got out BSBW F CLRL A Done by a fairly trivial peephole optimiser, in two steps: (a) if the operation is commutative, put any constant operand on the right (b) if the operation is logical-and, and the right operand is zero, the result is zero so flush the left operand (naturally that is only one of several possible patterns this step will look for) Your second example needed to be changed, but a := f() | <all-ones> indeed reduces to BSBW F MCOML #0, A Hope this helps ---- PS: the stupid program at my site for some reason won't let me post followups. They get bounced back with mailer messages that are (to me) incomprehensible. Is anyone able to offer advice? It all worked in 1987.