------- Additional Comments From sebastian dot pop at cri dot ensmp dot fr 2005-01-27 15:12 ------- Subject: Re: [4.0 Regression] IV-OPTS is O(N^3)
rakdver at atrey dot karlin dot mff dot cuni dot cz wrote: > the patch is below (in stronger form -- only removing entries that > contain invalidated symbols), and indeed helps with this testcase. Many thanks. > It however causes measurable slow down on normal code (see analysis > in one of the previous mails). > I see. Another idea: would it be possible to insert the invalidated names during the optimization pass instead of invalidating all the symbols? This would avoid the first pass over the scev database that search for symbols. > Note that your patch is not entirely correct -- in case loop is > unrolled, its number of iterations is changed, so in this case > scev_reset should clear its number of iterations unconditionally or the unroller should do that work on the unrolled loops? -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=18595
