[LLVMbugs] [Bug 8882] New: Poor scalar optimization loses trip count info

[bugzilla-daemon at llvm.org]

http://llvm.org/bugs/show_bug.cgi?id=8882

           Summary: Poor scalar optimization loses trip count info
           Product: libraries
           Version: 1.0
          Platform: PC
        OS/Version: All
            Status: NEW
          Severity: normal
          Priority: P
         Component: Scalar Optimizations
        AssignedTo: unassignedbugs at nondot.org
        ReportedBy: clattner at apple.com
                CC: llvmbugs at cs.uiuc.edu


We compile this:


template <typename Iterator, typename T>
void fill(Iterator first, Iterator last, T value) {
        while (first != last) *first++ = value;
}

int X[1000];

void foo() {
  fill(X, X+1000, 0);
}

with: clang t.cpp -S -o - -emit-llvm -O3 
into:

while.body.i:                                     ; preds = %while.body.i,
%entry
  %indvar.i = phi i64 [ %tmp.i, %while.body.i ], [ 0, %entry ]
  %tmp.i = add i64 %indvar.i, 1
  %first.addr.05.i = getelementptr [1000 x i32]* @X, i64 0, i64 %indvar.i
  store i32 0, i32* %first.addr.05.i, align 4, !tbaa !0
  %ptrincdec.i.idx.mask = and i64 %tmp.i, 4611686018427387903
  %cmp.i = icmp eq i64 %ptrincdec.i.idx.mask, 1000
  br i1 %cmp.i, label %_Z4fillIPiiEvT_S1_T0_.exit, label %while.body.i

This loop executes exactly 1000 times.  This appears to be being pessimized by
the -indvars pass, as can be seen by:

clang t.cpp -S -o - -emit-llvm -O0 | opt -mem2reg -S -inline -loop-rotate
-instcombine -indvars

which introduces the 'and'.  The and is not needed because the GEP can't wrap
around.

-- 
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.