1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt < %s -loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -enable-if-conversion -dce -instcombine -S | FileCheck %s 3 4target 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 6define i32 @foo(i32* nocapture %A, i32* nocapture %B, i32 %n) { 7; CHECK-LABEL: @foo( 8; CHECK-NEXT: entry: 9; CHECK-NEXT: [[CMP26:%.*]] = icmp sgt i32 [[N:%.*]], 0 10; CHECK-NEXT: br i1 [[CMP26]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]] 11; CHECK: for.body.preheader: 12; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N]], -1 13; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64 14; CHECK-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1 15; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4 16; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[MIN_ITERS_CHECKED:%.*]] 17; CHECK: min.iters.checked: 18; CHECK-NEXT: [[TMP3:%.*]] = and i32 [[N]], 3 19; CHECK-NEXT: [[N_MOD_VF:%.*]] = zext i32 [[TMP3]] to i64 20; CHECK-NEXT: [[N_VEC:%.*]] = sub nsw i64 [[TMP2]], [[N_MOD_VF]] 21; CHECK-NEXT: [[CMP_ZERO:%.*]] = icmp eq i64 [[N_VEC]], 0 22; CHECK-NEXT: br i1 [[CMP_ZERO]], label [[SCALAR_PH]], label [[VECTOR_MEMCHECK:%.*]] 23; CHECK: vector.memcheck: 24; CHECK-NEXT: [[TMP4:%.*]] = add i32 [[N]], -1 25; CHECK-NEXT: [[TMP5:%.*]] = zext i32 [[TMP4]] to i64 26; CHECK-NEXT: [[TMP6:%.*]] = add nuw nsw i64 [[TMP5]], 1 27; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i32, i32* [[A:%.*]], i64 [[TMP6]] 28; CHECK-NEXT: [[SCEVGEP4:%.*]] = getelementptr i32, i32* [[B:%.*]], i64 [[TMP6]] 29; CHECK-NEXT: [[BOUND0:%.*]] = icmp ugt i32* [[SCEVGEP4]], [[A]] 30; CHECK-NEXT: [[BOUND1:%.*]] = icmp ugt i32* [[SCEVGEP]], [[B]] 31; CHECK-NEXT: [[MEMCHECK_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] 32; CHECK-NEXT: br i1 [[MEMCHECK_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]] 33; CHECK: vector.ph: 34; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 35; CHECK: vector.body: 36; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 37; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDEX]] 38; CHECK-NEXT: [[TMP8:%.*]] = bitcast i32* [[TMP7]] to <4 x i32>* 39; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, <4 x i32>* [[TMP8]], align 4, !alias.scope !0, !noalias !3 40; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDEX]] 41; CHECK-NEXT: [[TMP10:%.*]] = bitcast i32* [[TMP9]] to <4 x i32>* 42; CHECK-NEXT: [[WIDE_LOAD6:%.*]] = load <4 x i32>, <4 x i32>* [[TMP10]], align 4, !alias.scope !3 43; CHECK-NEXT: [[TMP11:%.*]] = icmp sgt <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD6]] 44; CHECK-NEXT: [[TMP12:%.*]] = icmp sgt <4 x i32> [[WIDE_LOAD]], <i32 19, i32 19, i32 19, i32 19> 45; CHECK-NEXT: [[TMP13:%.*]] = icmp slt <4 x i32> [[WIDE_LOAD6]], <i32 4, i32 4, i32 4, i32 4> 46; CHECK-NEXT: [[TMP14:%.*]] = select <4 x i1> [[TMP13]], <4 x i32> <i32 4, i32 4, i32 4, i32 4>, <4 x i32> <i32 5, i32 5, i32 5, i32 5> 47; CHECK-NEXT: [[TMP15:%.*]] = and <4 x i1> [[TMP12]], [[TMP11]] 48; CHECK-NEXT: [[PREDPHI:%.*]] = select <4 x i1> [[TMP15]], <4 x i32> <i32 3, i32 3, i32 3, i32 3>, <4 x i32> <i32 9, i32 9, i32 9, i32 9> 49; CHECK-NEXT: [[TMP16:%.*]] = xor <4 x i1> [[TMP12]], <i1 true, i1 true, i1 true, i1 true> 50; CHECK-NEXT: [[TMP17:%.*]] = and <4 x i1> [[TMP11]], [[TMP16]] 51; CHECK-NEXT: [[PREDPHI7:%.*]] = select <4 x i1> [[TMP17]], <4 x i32> [[TMP14]], <4 x i32> [[PREDPHI]] 52; CHECK-NEXT: [[TMP18:%.*]] = bitcast i32* [[TMP7]] to <4 x i32>* 53; CHECK-NEXT: store <4 x i32> [[PREDPHI7]], <4 x i32>* [[TMP18]], align 4, !alias.scope !0, !noalias !3 54; CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 4 55; CHECK-NEXT: [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 56; CHECK-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !5 57; CHECK: middle.block: 58; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP3]], 0 59; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]] 60; CHECK: scalar.ph: 61; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ], [ 0, [[MIN_ITERS_CHECKED]] ], [ 0, [[VECTOR_MEMCHECK]] ] 62; CHECK-NEXT: br label [[FOR_BODY:%.*]] 63; CHECK: for.body: 64; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[IF_END14:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ] 65; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]] 66; CHECK-NEXT: [[TMP20:%.*]] = load i32, i32* [[ARRAYIDX]], align 4 67; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]] 68; CHECK-NEXT: [[TMP21:%.*]] = load i32, i32* [[ARRAYIDX2]], align 4 69; CHECK-NEXT: [[CMP3:%.*]] = icmp sgt i32 [[TMP20]], [[TMP21]] 70; CHECK-NEXT: br i1 [[CMP3]], label [[IF_THEN:%.*]], label [[IF_END14]] 71; CHECK: if.then: 72; CHECK-NEXT: [[CMP6:%.*]] = icmp sgt i32 [[TMP20]], 19 73; CHECK-NEXT: br i1 [[CMP6]], label [[IF_END14]], label [[IF_ELSE:%.*]] 74; CHECK: if.else: 75; CHECK-NEXT: [[CMP10:%.*]] = icmp slt i32 [[TMP21]], 4 76; CHECK-NEXT: [[DOT:%.*]] = select i1 [[CMP10]], i32 4, i32 5 77; CHECK-NEXT: br label [[IF_END14]] 78; CHECK: if.end14: 79; CHECK-NEXT: [[X_0:%.*]] = phi i32 [ 9, [[FOR_BODY]] ], [ 3, [[IF_THEN]] ], [ [[DOT]], [[IF_ELSE]] ] 80; CHECK-NEXT: store i32 [[X_0]], i32* [[ARRAYIDX]], align 4 81; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1 82; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32 83; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]] 84; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop !8 85; CHECK: for.end.loopexit: 86; CHECK-NEXT: br label [[FOR_END]] 87; CHECK: for.end: 88; CHECK-NEXT: ret i32 undef 89; 90entry: 91 %cmp26 = icmp sgt i32 %n, 0 92 br i1 %cmp26, label %for.body, label %for.end 93 94for.body: 95 %indvars.iv = phi i64 [ %indvars.iv.next, %if.end14 ], [ 0, %entry ] 96 %arrayidx = getelementptr inbounds i32, i32* %A, i64 %indvars.iv 97 %0 = load i32, i32* %arrayidx, align 4 98 %arrayidx2 = getelementptr inbounds i32, i32* %B, i64 %indvars.iv 99 %1 = load i32, i32* %arrayidx2, align 4 100 %cmp3 = icmp sgt i32 %0, %1 101 br i1 %cmp3, label %if.then, label %if.end14 102 103if.then: 104 %cmp6 = icmp sgt i32 %0, 19 105 br i1 %cmp6, label %if.end14, label %if.else 106 107if.else: 108 %cmp10 = icmp slt i32 %1, 4 109 %. = select i1 %cmp10, i32 4, i32 5 110 br label %if.end14 111 112if.end14: 113 %x.0 = phi i32 [ 9, %for.body ], [ 3, %if.then ], [ %., %if.else ] ; <------------- A PHI with 3 entries that we can still vectorize. 114 store i32 %x.0, i32* %arrayidx, align 4 115 %indvars.iv.next = add i64 %indvars.iv, 1 116 %lftr.wideiv = trunc i64 %indvars.iv.next to i32 117 %exitcond = icmp eq i32 %lftr.wideiv, %n 118 br i1 %exitcond, label %for.end, label %for.body 119 120for.end: 121 ret i32 undef 122} 123 124