1 ; RUN: opt < %s -loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -dce -instcombine -pass-remarks=loop-vectorize -pass-remarks-missed=loop-vectorize -S 2>&1 | FileCheck %s -check-prefix=OVERRIDE 2 ; RUN: opt < %s -loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -pragma-vectorize-memory-check-threshold=6 -dce -instcombine -pass-remarks=loop-vectorize -pass-remarks-missed=loop-vectorize -S 2>&1 | FileCheck %s 3 4 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" 5 6 ; First loop produced diagnostic pass remark. 7 ;CHECK: remark: {{.*}}:0:0: vectorized loop (vectorization width: 4, interleaved count: 1) 8 ; Second loop produces diagnostic analysis remark. 9 ;CHECK: remark: {{.*}}:0:0: loop not vectorized: cannot prove it is safe to reorder memory operations 10 11 ; First loop produced diagnostic pass remark. 12 ;OVERRIDE: remark: {{.*}}:0:0: vectorized loop (vectorization width: 4, interleaved count: 1) 13 ; Second loop produces diagnostic pass remark. 14 ;OVERRIDE: remark: {{.*}}:0:0: vectorized loop (vectorization width: 4, interleaved count: 1) 15 16 ; We are vectorizing with 6 runtime checks. 17 ;CHECK-LABEL: func1x6( 18 ;CHECK: <4 x i32> 19 ;CHECK: ret 20 ;OVERRIDE-LABEL: func1x6( 21 ;OVERRIDE: <4 x i32> 22 ;OVERRIDE: ret 23 define i32 @func1x6(i32* nocapture %out, i32* nocapture %A, i32* nocapture %B, i32* nocapture %C, i32* nocapture %D, i32* nocapture %E, i32* nocapture %F) { 24 entry: 25 br label %for.body 26 27 for.body: ; preds = %for.body, %entry 28 %i.016 = phi i64 [ 0, %entry ], [ %inc, %for.body ] 29 %arrayidx = getelementptr inbounds i32, i32* %A, i64 %i.016 30 %0 = load i32, i32* %arrayidx, align 4 31 %arrayidx1 = getelementptr inbounds i32, i32* %B, i64 %i.016 32 %1 = load i32, i32* %arrayidx1, align 4 33 %add = add nsw i32 %1, %0 34 %arrayidx2 = getelementptr inbounds i32, i32* %C, i64 %i.016 35 %2 = load i32, i32* %arrayidx2, align 4 36 %add3 = add nsw i32 %add, %2 37 %arrayidx4 = getelementptr inbounds i32, i32* %E, i64 %i.016 38 %3 = load i32, i32* %arrayidx4, align 4 39 %add5 = add nsw i32 %add3, %3 40 %arrayidx6 = getelementptr inbounds i32, i32* %F, i64 %i.016 41 %4 = load i32, i32* %arrayidx6, align 4 42 %add7 = add nsw i32 %add5, %4 43 %arrayidx8 = getelementptr inbounds i32, i32* %out, i64 %i.016 44 store i32 %add7, i32* %arrayidx8, align 4 45 %inc = add i64 %i.016, 1 46 %exitcond = icmp eq i64 %inc, 256 47 br i1 %exitcond, label %for.end, label %for.body 48 49 for.end: ; preds = %for.body 50 ret i32 undef 51 } 52 53 ; We are not vectorizing with 12 runtime checks. 54 ;CHECK-LABEL: func2x6( 55 ;CHECK-NOT: <4 x i32> 56 ;CHECK: ret 57 ; We vectorize with 12 checks if a vectorization hint is provided. 58 ;OVERRIDE-LABEL: func2x6( 59 ;OVERRIDE: <4 x i32> 60 ;OVERRIDE: ret 61 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) { 62 entry: 63 br label %for.body 64 65 for.body: ; preds = %for.body, %entry 66 %i.037 = phi i64 [ 0, %entry ], [ %inc, %for.body ] 67 %arrayidx = getelementptr inbounds i32, i32* %A, i64 %i.037 68 %0 = load i32, i32* %arrayidx, align 4 69 %arrayidx1 = getelementptr inbounds i32, i32* %B, i64 %i.037 70 %1 = load i32, i32* %arrayidx1, align 4 71 %add = add nsw i32 %1, %0 72 %arrayidx2 = getelementptr inbounds i32, i32* %C, i64 %i.037 73 %2 = load i32, i32* %arrayidx2, align 4 74 %add3 = add nsw i32 %add, %2 75 %arrayidx4 = getelementptr inbounds i32, i32* %E, i64 %i.037 76 %3 = load i32, i32* %arrayidx4, align 4 77 %add5 = add nsw i32 %add3, %3 78 %arrayidx6 = getelementptr inbounds i32, i32* %F, i64 %i.037 79 %4 = load i32, i32* %arrayidx6, align 4 80 %add7 = add nsw i32 %add5, %4 81 %arrayidx8 = getelementptr inbounds i32, i32* %out, i64 %i.037 82 store i32 %add7, i32* %arrayidx8, align 4 83 %5 = load i32, i32* %arrayidx, align 4 84 %6 = load i32, i32* %arrayidx1, align 4 85 %add11 = add nsw i32 %6, %5 86 %7 = load i32, i32* %arrayidx2, align 4 87 %add13 = add nsw i32 %add11, %7 88 %8 = load i32, i32* %arrayidx4, align 4 89 %add15 = add nsw i32 %add13, %8 90 %9 = load i32, i32* %arrayidx6, align 4 91 %add17 = add nsw i32 %add15, %9 92 %arrayidx18 = getelementptr inbounds i32, i32* %out2, i64 %i.037 93 store i32 %add17, i32* %arrayidx18, align 4 94 %inc = add i64 %i.037, 1 95 %exitcond = icmp eq i64 %inc, 256 96 br i1 %exitcond, label %for.end, label %for.body 97 98 for.end: ; preds = %for.body 99 ret i32 undef 100 } 101 102
