search for: newsema2

...;> poison if >> bit i of 'A op undef' may yield 0 and 1 (same for 'undef op B'). >> This one unfortunately breaks some algebraic rewrites like: x * -1 >> -> 0 >> - x >> >> >> I've implemented both semantics in Alive (branches: newsema2 and >> bitwise-poison2, respectively) and then we run Alive over a large set >> of >> optimizations to see which were ok and which became wrong. >> >> Many optimizations became wrong with either of the bitwise semantics. > > To be clear, "wrong" here me...

...2) (thanks to Sanjoy): for an operation R = A op B, bit R[i] is poison if bit i of 'A op undef' may yield 0 and 1 (same for 'undef op B'). This one unfortunately breaks some algebraic rewrites like: x * -1 -> 0 - x I've implemented both semantics in Alive (branches: newsema2 and bitwise-poison2, respectively) and then we run Alive over a large set of optimizations to see which were ok and which became wrong. Many optimizations became wrong with either of the bitwise semantics. But the important ones are: - We keep the problem of not being able to increase number of r...