[LLVMbugs] [Bug 4895] New: Missed phi/phi/select folding opportunity

bugzilla-daemon at cs.uiuc.edu bugzilla-daemon at cs.uiuc.edu
Fri Sep 4 22:14:47 PDT 2009


http://llvm.org/bugs/show_bug.cgi?id=4895

           Summary: Missed phi/phi/select folding opportunity
           Product: new-bugs
           Version: unspecified
          Platform: Other
        OS/Version: other
            Status: NEW
          Severity: minor
          Priority: P2
         Component: new bugs
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: me22.ca at gmail.com
                CC: llvmbugs at cs.uiuc.edu


I compiled the following C++ code with bitter melon on 2009-09-05T05:00Z with
LTO enabled:

#include <set>
int main() {
std::set<int> s;
if (s.find(0) != s.end()) return *s.find(0);
}

This resulted in this basic block (among others):

_ZNSt8_Rb_treeIiiSt9_IdentityIiESt4lessIiESaIiEED1Ev.exit:              ; preds
= %invcont2, %invcont1
        %.0 = phi i32 [ %29, %invcont2 ], [ undef, %invcont1 ]          ; <i32>
[#uses=1]
        %finally_tmp.74.0 = phi i1 [ true, %invcont2 ], [ false, %invcont1 ]   
        ; <i1> [#uses=1]
        %.1 = select i1 %finally_tmp.74.0, i32 %.0, i32 0               ; <i32>
[#uses=1]
        ret i32 %.1

Which, which names shortened, looks like this:

%.0 = phi i32 [%29, %A], [undef, %B]
%t = phi i1 [true, %A], [false, %B]
%.1 = select i1 %t, i32 %.0, i32 0
ret i32 %.1

Selecting on the result of a phi[true][false] seems strange to me, as AFAIK it
could have simply been written like this:

%.0 = phi i32 [%29, %A], [0, %B]
ret i32 %.0


-- 
Configure bugmail: http://llvm.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.



More information about the llvm-bugs mailing list