Home | History | Annotate | Download | only in LoopUnroll
      1 ; RUN: opt < %s -S -loop-unroll -unroll-max-iteration-count-to-analyze=100 -unroll-dynamic-cost-savings-discount=1000 -unroll-threshold=10 -unroll-percent-dynamic-cost-saved-threshold=40 | FileCheck %s
      2 target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
      3 
      4 @known_constant = internal unnamed_addr constant [10 x i32] [i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1], align 16
      5 
      6 ; If we can figure out result of comparison on each iteration, we can resolve
      7 ; the depending branch. That means, that the unrolled version of the loop would
      8 ; have less code, because we don't need not-taken basic blocks there.
      9 ; This test checks that this is taken into consideration.
     10 ; We expect this loop to be unrolled, because the most complicated part of its
     11 ; body (if.then block) is never actually executed.
     12 ; CHECK-LABEL: @branch_folded
     13 ; CHECK-NOT: br i1 %
     14 ; CHECK: ret i32
     15 define i32 @branch_folded(i32* noalias nocapture readonly %b) {
     16 entry:
     17   br label %for.body
     18 
     19 for.body:                                         ; preds = %for.inc, %entry
     20   %iv.0 = phi i64 [ 0, %entry ], [ %iv.1, %for.inc ]
     21   %r.0 = phi i32 [ 0, %entry ], [ %r.1, %for.inc ]
     22   %arrayidx1 = getelementptr inbounds [10 x i32], [10 x i32]* @known_constant, i64 0, i64 %iv.0
     23   %x1 = load i32, i32* %arrayidx1, align 4
     24   %cmp = icmp eq i32 %x1, 0
     25   %iv.1 = add nuw nsw i64 %iv.0, 1
     26   br i1 %cmp, label %if.then, label %for.inc
     27 
     28 if.then:                                          ; preds = %for.body
     29   %arrayidx2 = getelementptr inbounds i32, i32* %b, i64 %iv.0
     30   %x2 = load i32, i32* %arrayidx2, align 4
     31   %add = add nsw i32 %x2, %r.0
     32   br label %for.inc
     33 
     34 for.inc:                                          ; preds = %for.body, %if.then
     35   %r.1 = phi i32 [ %add, %if.then ], [ %x1, %for.body ]
     36   %exitcond = icmp eq i64 %iv.1, 10
     37   br i1 %exitcond, label %for.end, label %for.body
     38 
     39 for.end:                                          ; preds = %for.inc
     40   ret i32 %r.1
     41 }
     42 
     43 ; Check that we don't crash when we analyze icmp with pointer-typed IV and a
     44 ; pointer.
     45 ; CHECK-LABEL: @ptr_cmp_crash
     46 ; CHECK:   ret void
     47 define void @ptr_cmp_crash() {
     48 entry:
     49   br label %while.body
     50 
     51 while.body:
     52   %iv.0 = phi i32* [ getelementptr inbounds ([10 x i32], [10 x i32]* @known_constant, i64 0, i64 0), %entry ], [ %iv.1, %while.body ]
     53   %iv.1 = getelementptr inbounds i32, i32* %iv.0, i64 1
     54   %exitcond = icmp eq i32* %iv.1, getelementptr inbounds ([10 x i32], [10 x i32]* @known_constant, i64 0, i64 9)
     55   br i1 %exitcond, label %loop.exit, label %while.body
     56 
     57 loop.exit:
     58   ret void
     59 }
     60 
     61 ; Check that we don't crash when we analyze ptrtoint cast.
     62 ; CHECK-LABEL: @ptrtoint_cast_crash
     63 ; CHECK:   ret void
     64 define void @ptrtoint_cast_crash(i8 * %a) {
     65 entry:
     66   %limit = getelementptr i8, i8* %a, i64 512
     67   br label %loop.body
     68 
     69 loop.body:
     70   %iv.0 = phi i8* [ %a, %entry ], [ %iv.1, %loop.body ]
     71   %cast = ptrtoint i8* %iv.0 to i64
     72   %iv.1 = getelementptr inbounds i8, i8* %iv.0, i64 1
     73   %exitcond = icmp ne i8* %iv.1, %limit
     74   br i1 %exitcond, label %loop.body, label %loop.exit
     75 
     76 loop.exit:
     77   ret void
     78 }
     79