On Mon, 1 Dec 2014, Richard Biener wrote: > +/* Combine two successive divisions. */ > +(for div (trunc_div ceil_div floor_div round_div exact_div)
This doesn't seem correct for all kinds of division and signedness of arguments. TRUNC_DIV_EXPR (C division) and EXACT_DIV_EXPR should be OK regardless. But for CEIL_DIV_EXPR and FLOOR_DIV_EXPR, in ((a / b) / c) you need c to be positive so that both roundings are in the same direction (consider e.g. 3 FLOOR_DIV -2 FLOOR_DIV -2 == 1, but 3 FLOOR_DIV 4 == 0). And for ROUND_DIV_EXPR, it can be incorrect whatever the signs (e.g. 2 ROUND_DIV 3 ROUND_DIV 2 == 1, but 2 ROUND_DIV 6 == 0). I'm not sure if these forms of division actually occur in places where this could cause a problem, but it does look like Ada may enable you to generate ROUND_DIV_EXPR. -- Joseph S. Myers jos...@codesourcery.com
