1 ; RUN: opt < %s -loop-vectorize -force-vector-unroll=1 -force-vector-width=4 -dce -instcombine -S | FileCheck %s 2 3 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128" 4 target triple = "x86_64-apple-macosx10.8.0" 5 6 ; We are vectorizing with 6 runtime checks. 7 ;CHECK-LABEL: func1x6( 8 ;CHECK: <4 x i32> 9 ;CHECK: ret 10 define i32 @func1x6(i32* nocapture %out, i32* nocapture %A, i32* nocapture %B, i32* nocapture %C, i32* nocapture %D, i32* nocapture %E, i32* nocapture %F) { 11 entry: 12 br label %for.body 13 14 for.body: ; preds = %for.body, %entry 15 %i.016 = phi i64 [ 0, %entry ], [ %inc, %for.body ] 16 %arrayidx = getelementptr inbounds i32* %A, i64 %i.016 17 %0 = load i32* %arrayidx, align 4 18 %arrayidx1 = getelementptr inbounds i32* %B, i64 %i.016 19 %1 = load i32* %arrayidx1, align 4 20 %add = add nsw i32 %1, %0 21 %arrayidx2 = getelementptr inbounds i32* %C, i64 %i.016 22 %2 = load i32* %arrayidx2, align 4 23 %add3 = add nsw i32 %add, %2 24 %arrayidx4 = getelementptr inbounds i32* %E, i64 %i.016 25 %3 = load i32* %arrayidx4, align 4 26 %add5 = add nsw i32 %add3, %3 27 %arrayidx6 = getelementptr inbounds i32* %F, i64 %i.016 28 %4 = load i32* %arrayidx6, align 4 29 %add7 = add nsw i32 %add5, %4 30 %arrayidx8 = getelementptr inbounds i32* %out, i64 %i.016 31 store i32 %add7, i32* %arrayidx8, align 4 32 %inc = add i64 %i.016, 1 33 %exitcond = icmp eq i64 %inc, 256 34 br i1 %exitcond, label %for.end, label %for.body 35 36 for.end: ; preds = %for.body 37 ret i32 undef 38 } 39 40 ; We are not vectorizing with 12 runtime checks. 41 ;CHECK-LABEL: func2x6( 42 ;CHECK-NOT: <4 x i32> 43 ;CHECK: ret 44 define i32 @func2x6(i32* nocapture %out, i32* nocapture %out2, i32* nocapture %A, i32* nocapture %B, i32* nocapture %C, i32* nocapture %D, i32* nocapture %E, i32* nocapture %F) { 45 entry: 46 br label %for.body 47 48 for.body: ; preds = %for.body, %entry 49 %i.037 = phi i64 [ 0, %entry ], [ %inc, %for.body ] 50 %arrayidx = getelementptr inbounds i32* %A, i64 %i.037 51 %0 = load i32* %arrayidx, align 4 52 %arrayidx1 = getelementptr inbounds i32* %B, i64 %i.037 53 %1 = load i32* %arrayidx1, align 4 54 %add = add nsw i32 %1, %0 55 %arrayidx2 = getelementptr inbounds i32* %C, i64 %i.037 56 %2 = load i32* %arrayidx2, align 4 57 %add3 = add nsw i32 %add, %2 58 %arrayidx4 = getelementptr inbounds i32* %E, i64 %i.037 59 %3 = load i32* %arrayidx4, align 4 60 %add5 = add nsw i32 %add3, %3 61 %arrayidx6 = getelementptr inbounds i32* %F, i64 %i.037 62 %4 = load i32* %arrayidx6, align 4 63 %add7 = add nsw i32 %add5, %4 64 %arrayidx8 = getelementptr inbounds i32* %out, i64 %i.037 65 store i32 %add7, i32* %arrayidx8, align 4 66 %5 = load i32* %arrayidx, align 4 67 %6 = load i32* %arrayidx1, align 4 68 %add11 = add nsw i32 %6, %5 69 %7 = load i32* %arrayidx2, align 4 70 %add13 = add nsw i32 %add11, %7 71 %8 = load i32* %arrayidx4, align 4 72 %add15 = add nsw i32 %add13, %8 73 %9 = load i32* %arrayidx6, align 4 74 %add17 = add nsw i32 %add15, %9 75 %arrayidx18 = getelementptr inbounds i32* %out2, i64 %i.037 76 store i32 %add17, i32* %arrayidx18, align 4 77 %inc = add i64 %i.037, 1 78 %exitcond = icmp eq i64 %inc, 256 79 br i1 %exitcond, label %for.end, label %for.body 80 81 for.end: ; preds = %for.body 82 ret i32 undef 83 } 84 85
