6257 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py2; RUN: opt < %s -passes=loop-vectorize -force-vector-interleave=1 -force-vector-width=2 -force-widen-divrem-via-safe-divisor=0 -S | FileCheck %s3; RUN: opt < %s -passes=loop-vectorize,instcombine -force-vector-interleave=1 -force-vector-width=2 -force-widen-divrem-via-safe-divisor=0 -S | FileCheck %s --check-prefix=IND4; RUN: opt < %s -passes=loop-vectorize,instcombine -force-vector-interleave=2 -force-vector-width=2 -force-widen-divrem-via-safe-divisor=0 -S | FileCheck %s --check-prefix=UNROLL5; RUN: opt < %s -passes=loop-vectorize -force-vector-interleave=2 -force-vector-width=2 -force-widen-divrem-via-safe-divisor=0 -S | FileCheck %s --check-prefix=UNROLL-NO-IC6; RUN: opt < %s -passes=loop-vectorize,instcombine -force-vector-interleave=2 -force-vector-width=4 -force-widen-divrem-via-safe-divisor=0 -enable-interleaved-mem-accesses -S | FileCheck %s --check-prefix=INTERLEAVE7 8target 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"9 10; Make sure that we can handle multiple integer induction variables.11;12define void @multi_int_induction(ptr %A, i32 %N) {13; CHECK-LABEL: @multi_int_induction(14; CHECK-NEXT: for.body.lr.ph:15; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -116; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i6417; CHECK-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 118; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 219; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]20; CHECK: vector.ph:21; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 222; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]23; CHECK-NEXT: [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i3224; CHECK-NEXT: [[IND_END:%.*]] = add i32 190, [[DOTCAST]]25; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]26; CHECK: vector.body:27; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]28; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 190, i32 191>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]29; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]30; CHECK-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP4]], align 431; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 232; CHECK-NEXT: [[VEC_IND_NEXT]] = add nsw <2 x i32> [[VEC_IND]], splat (i32 2)33; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]34; CHECK-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]35; CHECK: middle.block:36; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]37; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]38; CHECK: scalar.ph:39; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH:%.*]] ]40; CHECK-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 190, [[FOR_BODY_LR_PH]] ]41; CHECK-NEXT: br label [[FOR_BODY:%.*]]42; CHECK: for.body:43; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]44; CHECK-NEXT: [[COUNT_09:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]45; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]]46; CHECK-NEXT: store i32 [[COUNT_09]], ptr [[ARRAYIDX2]], align 447; CHECK-NEXT: [[INC]] = add nsw i32 [[COUNT_09]], 148; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 149; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i3250; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i32 [[LFTR_WIDEIV]], [[N]]51; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP3:![0-9]+]]52; CHECK: for.end:53; CHECK-NEXT: ret void54;55; IND-LABEL: @multi_int_induction(56; IND-NEXT: for.body.lr.ph:57; IND-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -158; IND-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i6459; IND-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 160; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP0]], 061; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]62; IND: vector.ph:63; IND-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 858993459064; IND-NEXT: [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i3265; IND-NEXT: [[IND_END:%.*]] = add i32 [[DOTCAST]], 19066; IND-NEXT: br label [[VECTOR_BODY:%.*]]67; IND: vector.body:68; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]69; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 190, i32 191>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]70; IND-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]71; IND-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP3]], align 472; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 273; IND-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <2 x i32> [[VEC_IND]], splat (i32 2)74; IND-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]75; IND-NEXT: br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]76; IND: middle.block:77; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]78; IND-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]79; IND: scalar.ph:80; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH:%.*]] ]81; IND-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 190, [[FOR_BODY_LR_PH]] ]82; IND-NEXT: br label [[FOR_BODY:%.*]]83; IND: for.body:84; IND-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]85; IND-NEXT: [[COUNT_09:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]86; IND-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]]87; IND-NEXT: store i32 [[COUNT_09]], ptr [[ARRAYIDX2]], align 488; IND-NEXT: [[INC]] = add nsw i32 [[COUNT_09]], 189; IND-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 190; IND-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i3291; IND-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i32 [[N]], [[LFTR_WIDEIV]]92; IND-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]93; IND: for.end:94; IND-NEXT: ret void95;96; UNROLL-LABEL: @multi_int_induction(97; UNROLL-NEXT: for.body.lr.ph:98; UNROLL-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -199; UNROLL-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64100; UNROLL-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1101; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3102; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]103; UNROLL: vector.ph:104; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934588105; UNROLL-NEXT: [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i32106; UNROLL-NEXT: [[IND_END:%.*]] = add i32 [[DOTCAST]], 190107; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]108; UNROLL: vector.body:109; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]110; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 190, i32 191>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]111; UNROLL-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)112; UNROLL-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]113; UNROLL-NEXT: [[TMP4:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP3]], i64 8114; UNROLL-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP3]], align 4115; UNROLL-NEXT: store <2 x i32> [[STEP_ADD]], ptr [[TMP4]], align 4116; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4117; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 4)118; UNROLL-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]119; UNROLL-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]120; UNROLL: middle.block:121; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]122; UNROLL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]123; UNROLL: scalar.ph:124; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH:%.*]] ]125; UNROLL-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 190, [[FOR_BODY_LR_PH]] ]126; UNROLL-NEXT: br label [[FOR_BODY:%.*]]127; UNROLL: for.body:128; UNROLL-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]129; UNROLL-NEXT: [[COUNT_09:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]130; UNROLL-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]]131; UNROLL-NEXT: store i32 [[COUNT_09]], ptr [[ARRAYIDX2]], align 4132; UNROLL-NEXT: [[INC]] = add nsw i32 [[COUNT_09]], 1133; UNROLL-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1134; UNROLL-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32135; UNROLL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i32 [[N]], [[LFTR_WIDEIV]]136; UNROLL-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]137; UNROLL: for.end:138; UNROLL-NEXT: ret void139;140; UNROLL-NO-IC-LABEL: @multi_int_induction(141; UNROLL-NO-IC-NEXT: for.body.lr.ph:142; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -1143; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64144; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1145; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4146; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]147; UNROLL-NO-IC: vector.ph:148; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4149; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]150; UNROLL-NO-IC-NEXT: [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i32151; UNROLL-NO-IC-NEXT: [[IND_END:%.*]] = add i32 190, [[DOTCAST]]152; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]153; UNROLL-NO-IC: vector.body:154; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]155; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 190, i32 191>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]156; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)157; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]158; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i64 2159; UNROLL-NO-IC-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP4]], align 4160; UNROLL-NO-IC-NEXT: store <2 x i32> [[STEP_ADD]], ptr [[TMP6]], align 4161; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4162; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add nsw <2 x i32> [[STEP_ADD]], splat (i32 2)163; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]164; UNROLL-NO-IC-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]165; UNROLL-NO-IC: middle.block:166; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]167; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]168; UNROLL-NO-IC: scalar.ph:169; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH:%.*]] ]170; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 190, [[FOR_BODY_LR_PH]] ]171; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]172; UNROLL-NO-IC: for.body:173; UNROLL-NO-IC-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]174; UNROLL-NO-IC-NEXT: [[COUNT_09:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]175; UNROLL-NO-IC-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]]176; UNROLL-NO-IC-NEXT: store i32 [[COUNT_09]], ptr [[ARRAYIDX2]], align 4177; UNROLL-NO-IC-NEXT: [[INC]] = add nsw i32 [[COUNT_09]], 1178; UNROLL-NO-IC-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1179; UNROLL-NO-IC-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32180; UNROLL-NO-IC-NEXT: [[EXITCOND:%.*]] = icmp ne i32 [[LFTR_WIDEIV]], [[N]]181; UNROLL-NO-IC-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP3:![0-9]+]]182; UNROLL-NO-IC: for.end:183; UNROLL-NO-IC-NEXT: ret void184;185; INTERLEAVE-LABEL: @multi_int_induction(186; INTERLEAVE-NEXT: for.body.lr.ph:187; INTERLEAVE-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -1188; INTERLEAVE-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64189; INTERLEAVE-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1190; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 7191; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]192; INTERLEAVE: vector.ph:193; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934584194; INTERLEAVE-NEXT: [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i32195; INTERLEAVE-NEXT: [[IND_END:%.*]] = add i32 [[DOTCAST]], 190196; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]197; INTERLEAVE: vector.body:198; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]199; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 190, i32 191, i32 192, i32 193>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]200; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], splat (i32 4)201; INTERLEAVE-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]202; INTERLEAVE-NEXT: [[TMP4:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP3]], i64 16203; INTERLEAVE-NEXT: store <4 x i32> [[VEC_IND]], ptr [[TMP3]], align 4204; INTERLEAVE-NEXT: store <4 x i32> [[STEP_ADD]], ptr [[TMP4]], align 4205; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8206; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 8)207; INTERLEAVE-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]208; INTERLEAVE-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]209; INTERLEAVE: middle.block:210; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]211; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]212; INTERLEAVE: scalar.ph:213; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH:%.*]] ]214; INTERLEAVE-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 190, [[FOR_BODY_LR_PH]] ]215; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]216; INTERLEAVE: for.body:217; INTERLEAVE-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]218; INTERLEAVE-NEXT: [[COUNT_09:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]219; INTERLEAVE-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]]220; INTERLEAVE-NEXT: store i32 [[COUNT_09]], ptr [[ARRAYIDX2]], align 4221; INTERLEAVE-NEXT: [[INC]] = add nsw i32 [[COUNT_09]], 1222; INTERLEAVE-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1223; INTERLEAVE-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32224; INTERLEAVE-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i32 [[N]], [[LFTR_WIDEIV]]225; INTERLEAVE-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]226; INTERLEAVE: for.end:227; INTERLEAVE-NEXT: ret void228;229for.body.lr.ph:230 br label %for.body231 232for.body:233 %indvars.iv = phi i64 [ 0, %for.body.lr.ph ], [ %indvars.iv.next, %for.body ]234 %count.09 = phi i32 [ 190, %for.body.lr.ph ], [ %inc, %for.body ]235 %arrayidx2 = getelementptr inbounds i32, ptr %A, i64 %indvars.iv236 store i32 %count.09, ptr %arrayidx2, align 4237 %inc = add nsw i32 %count.09, 1238 %indvars.iv.next = add i64 %indvars.iv, 1239 %lftr.wideiv = trunc i64 %indvars.iv.next to i32240 %exitcond = icmp ne i32 %lftr.wideiv, %N241 br i1 %exitcond, label %for.body, label %for.end242 243for.end:244 ret void245}246 247; Make sure we remove unneeded vectorization of induction variables.248; In order for instcombine to cleanup the vectorized induction variables that we249; create in the loop vectorizer we need to perform some form of redundancy250; elimination to get rid of multiple uses.251 252 253; Vectorized induction variable.254 255define void @scalar_use(ptr %a, float %b, i64 %offset, i64 %offset2, i64 %n) {256; CHECK-LABEL: @scalar_use(257; CHECK-NEXT: entry:258; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2259; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]260; CHECK: vector.memcheck:261; CHECK-NEXT: [[TMP0:%.*]] = shl i64 [[OFFSET:%.*]], 2262; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP0]]263; CHECK-NEXT: [[TMP1:%.*]] = shl i64 [[N]], 2264; CHECK-NEXT: [[TMP2:%.*]] = add i64 [[TMP1]], [[TMP0]]265; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP2]]266; CHECK-NEXT: [[TMP3:%.*]] = shl i64 [[OFFSET2:%.*]], 2267; CHECK-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]]268; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[TMP1]], [[TMP3]]269; CHECK-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP4]]270; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP3]]271; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[SCEVGEP2]], [[SCEVGEP1]]272; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]273; CHECK-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]274; CHECK: vector.ph:275; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 2276; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]277; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[B:%.*]], i64 0278; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer279; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]280; CHECK: vector.body:281; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]282; CHECK-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], [[OFFSET]]283; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP6]]284; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP7]], align 4, !alias.scope [[META4:![0-9]+]], !noalias [[META7:![0-9]+]]285; CHECK-NEXT: [[TMP9:%.*]] = add i64 [[INDEX]], [[OFFSET2]]286; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP9]]287; CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <2 x float>, ptr [[TMP10]], align 4, !alias.scope [[META7]]288; CHECK-NEXT: [[TMP12:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD4]]289; CHECK-NEXT: [[TMP13:%.*]] = fadd fast <2 x float> [[WIDE_LOAD]], [[TMP12]]290; CHECK-NEXT: store <2 x float> [[TMP13]], ptr [[TMP7]], align 4, !alias.scope [[META4]], !noalias [[META7]]291; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2292; CHECK-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]293; CHECK-NEXT: br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]294; CHECK: middle.block:295; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]296; CHECK-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]297; CHECK: scalar.ph:298; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]299; CHECK-NEXT: br label [[FOR_BODY:%.*]]300; CHECK: for.body:301; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]302; CHECK-NEXT: [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET]]303; CHECK-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IND_SUM]]304; CHECK-NEXT: [[L1:%.*]] = load float, ptr [[ARR_IDX]], align 4305; CHECK-NEXT: [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2]]306; CHECK-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IND_SUM2]]307; CHECK-NEXT: [[L2:%.*]] = load float, ptr [[ARR_IDX2]], align 4308; CHECK-NEXT: [[M:%.*]] = fmul fast float [[B]], [[L2]]309; CHECK-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]]310; CHECK-NEXT: store float [[AD]], ptr [[ARR_IDX]], align 4311; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1312; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]313; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]314; CHECK: loopexit:315; CHECK-NEXT: ret void316;317; IND-LABEL: @scalar_use(318; IND-NEXT: entry:319; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2320; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]321; IND: vector.memcheck:322; IND-NEXT: [[TMP0:%.*]] = shl i64 [[OFFSET:%.*]], 2323; IND-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP0]]324; IND-NEXT: [[TMP1:%.*]] = shl i64 [[N]], 2325; IND-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP1]]326; IND-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP0]]327; IND-NEXT: [[TMP3:%.*]] = shl i64 [[OFFSET2:%.*]], 2328; IND-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]]329; IND-NEXT: [[TMP4:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP1]]330; IND-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[TMP4]], i64 [[TMP3]]331; IND-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP3]]332; IND-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[SCEVGEP2]], [[SCEVGEP1]]333; IND-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]334; IND-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]335; IND: vector.ph:336; IND-NEXT: [[N_VEC:%.*]] = and i64 [[N]], -2337; IND-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[B:%.*]], i64 0338; IND-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer339; IND-NEXT: br label [[VECTOR_BODY:%.*]]340; IND: vector.body:341; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]342; IND-NEXT: [[TMP5:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]]343; IND-NEXT: [[TMP6:%.*]] = getelementptr float, ptr [[TMP5]], i64 [[OFFSET]]344; IND-NEXT: [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP6]], align 4, !alias.scope [[META4:![0-9]+]], !noalias [[META7:![0-9]+]]345; IND-NEXT: [[TMP7:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]]346; IND-NEXT: [[TMP8:%.*]] = getelementptr float, ptr [[TMP7]], i64 [[OFFSET2]]347; IND-NEXT: [[WIDE_LOAD4:%.*]] = load <2 x float>, ptr [[TMP8]], align 4, !alias.scope [[META7]]348; IND-NEXT: [[TMP9:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD4]]349; IND-NEXT: [[TMP10:%.*]] = fadd fast <2 x float> [[WIDE_LOAD]], [[TMP9]]350; IND-NEXT: store <2 x float> [[TMP10]], ptr [[TMP6]], align 4, !alias.scope [[META4]], !noalias [[META7]]351; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2352; IND-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]353; IND-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]354; IND: middle.block:355; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]356; IND-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]357; IND: scalar.ph:358; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]359; IND-NEXT: br label [[FOR_BODY:%.*]]360; IND: for.body:361; IND-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]362; IND-NEXT: [[TMP12:%.*]] = getelementptr float, ptr [[A]], i64 [[IV]]363; IND-NEXT: [[ARR_IDX:%.*]] = getelementptr float, ptr [[TMP12]], i64 [[OFFSET]]364; IND-NEXT: [[L1:%.*]] = load float, ptr [[ARR_IDX]], align 4365; IND-NEXT: [[TMP13:%.*]] = getelementptr float, ptr [[A]], i64 [[IV]]366; IND-NEXT: [[ARR_IDX2:%.*]] = getelementptr float, ptr [[TMP13]], i64 [[OFFSET2]]367; IND-NEXT: [[L2:%.*]] = load float, ptr [[ARR_IDX2]], align 4368; IND-NEXT: [[M:%.*]] = fmul fast float [[B]], [[L2]]369; IND-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]]370; IND-NEXT: store float [[AD]], ptr [[ARR_IDX]], align 4371; IND-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1372; IND-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]373; IND-NEXT: br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]374; IND: loopexit:375; IND-NEXT: ret void376;377; UNROLL-LABEL: @scalar_use(378; UNROLL-NEXT: entry:379; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4380; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]381; UNROLL: vector.memcheck:382; UNROLL-NEXT: [[TMP0:%.*]] = shl i64 [[OFFSET:%.*]], 2383; UNROLL-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP0]]384; UNROLL-NEXT: [[TMP1:%.*]] = shl i64 [[N]], 2385; UNROLL-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP1]]386; UNROLL-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP0]]387; UNROLL-NEXT: [[TMP3:%.*]] = shl i64 [[OFFSET2:%.*]], 2388; UNROLL-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]]389; UNROLL-NEXT: [[TMP4:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP1]]390; UNROLL-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[TMP4]], i64 [[TMP3]]391; UNROLL-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP3]]392; UNROLL-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[SCEVGEP2]], [[SCEVGEP1]]393; UNROLL-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]394; UNROLL-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]395; UNROLL: vector.ph:396; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[N]], -4397; UNROLL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[B:%.*]], i64 0398; UNROLL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer399; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]400; UNROLL: vector.body:401; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]402; UNROLL-NEXT: [[TMP5:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]]403; UNROLL-NEXT: [[TMP6:%.*]] = getelementptr float, ptr [[TMP5]], i64 [[OFFSET]]404; UNROLL-NEXT: [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP6]], i64 8405; UNROLL-NEXT: [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP6]], align 4, !alias.scope [[META4:![0-9]+]], !noalias [[META7:![0-9]+]]406; UNROLL-NEXT: [[WIDE_LOAD4:%.*]] = load <2 x float>, ptr [[TMP7]], align 4, !alias.scope [[META4]], !noalias [[META7]]407; UNROLL-NEXT: [[TMP8:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]]408; UNROLL-NEXT: [[TMP9:%.*]] = getelementptr float, ptr [[TMP8]], i64 [[OFFSET2]]409; UNROLL-NEXT: [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i64 8410; UNROLL-NEXT: [[WIDE_LOAD5:%.*]] = load <2 x float>, ptr [[TMP9]], align 4, !alias.scope [[META7]]411; UNROLL-NEXT: [[WIDE_LOAD6:%.*]] = load <2 x float>, ptr [[TMP10]], align 4, !alias.scope [[META7]]412; UNROLL-NEXT: [[TMP11:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD5]]413; UNROLL-NEXT: [[TMP12:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD6]]414; UNROLL-NEXT: [[TMP13:%.*]] = fadd fast <2 x float> [[WIDE_LOAD]], [[TMP11]]415; UNROLL-NEXT: [[TMP14:%.*]] = fadd fast <2 x float> [[WIDE_LOAD4]], [[TMP12]]416; UNROLL-NEXT: store <2 x float> [[TMP13]], ptr [[TMP6]], align 4, !alias.scope [[META4]], !noalias [[META7]]417; UNROLL-NEXT: store <2 x float> [[TMP14]], ptr [[TMP7]], align 4, !alias.scope [[META4]], !noalias [[META7]]418; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4419; UNROLL-NEXT: [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]420; UNROLL-NEXT: br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]421; UNROLL: middle.block:422; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]423; UNROLL-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]424; UNROLL: scalar.ph:425; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]426; UNROLL-NEXT: br label [[FOR_BODY:%.*]]427; UNROLL: for.body:428; UNROLL-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]429; UNROLL-NEXT: [[TMP16:%.*]] = getelementptr float, ptr [[A]], i64 [[IV]]430; UNROLL-NEXT: [[ARR_IDX:%.*]] = getelementptr float, ptr [[TMP16]], i64 [[OFFSET]]431; UNROLL-NEXT: [[L1:%.*]] = load float, ptr [[ARR_IDX]], align 4432; UNROLL-NEXT: [[TMP17:%.*]] = getelementptr float, ptr [[A]], i64 [[IV]]433; UNROLL-NEXT: [[ARR_IDX2:%.*]] = getelementptr float, ptr [[TMP17]], i64 [[OFFSET2]]434; UNROLL-NEXT: [[L2:%.*]] = load float, ptr [[ARR_IDX2]], align 4435; UNROLL-NEXT: [[M:%.*]] = fmul fast float [[B]], [[L2]]436; UNROLL-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]]437; UNROLL-NEXT: store float [[AD]], ptr [[ARR_IDX]], align 4438; UNROLL-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1439; UNROLL-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]440; UNROLL-NEXT: br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]441; UNROLL: loopexit:442; UNROLL-NEXT: ret void443;444; UNROLL-NO-IC-LABEL: @scalar_use(445; UNROLL-NO-IC-NEXT: entry:446; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4447; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]448; UNROLL-NO-IC: vector.memcheck:449; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = shl i64 [[OFFSET:%.*]], 2450; UNROLL-NO-IC-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP0]]451; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = shl i64 [[N]], 2452; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add i64 [[TMP1]], [[TMP0]]453; UNROLL-NO-IC-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP2]]454; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = shl i64 [[OFFSET2:%.*]], 2455; UNROLL-NO-IC-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]]456; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = add i64 [[TMP1]], [[TMP3]]457; UNROLL-NO-IC-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP4]]458; UNROLL-NO-IC-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP3]]459; UNROLL-NO-IC-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[SCEVGEP2]], [[SCEVGEP1]]460; UNROLL-NO-IC-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]461; UNROLL-NO-IC-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]462; UNROLL-NO-IC: vector.ph:463; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 4464; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]465; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[B:%.*]], i64 0466; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer467; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]468; UNROLL-NO-IC: vector.body:469; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]470; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], [[OFFSET]]471; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP6]]472; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i64 2473; UNROLL-NO-IC-NEXT: [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP7]], align 4, !alias.scope [[META4:![0-9]+]], !noalias [[META7:![0-9]+]]474; UNROLL-NO-IC-NEXT: [[WIDE_LOAD4:%.*]] = load <2 x float>, ptr [[TMP8]], align 4, !alias.scope [[META4]], !noalias [[META7]]475; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = add i64 [[INDEX]], [[OFFSET2]]476; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP10]]477; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP11]], i64 2478; UNROLL-NO-IC-NEXT: [[WIDE_LOAD5:%.*]] = load <2 x float>, ptr [[TMP11]], align 4, !alias.scope [[META7]]479; UNROLL-NO-IC-NEXT: [[WIDE_LOAD6:%.*]] = load <2 x float>, ptr [[TMP12]], align 4, !alias.scope [[META7]]480; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD5]]481; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD6]]482; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = fadd fast <2 x float> [[WIDE_LOAD]], [[TMP14]]483; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = fadd fast <2 x float> [[WIDE_LOAD4]], [[TMP15]]484; UNROLL-NO-IC-NEXT: store <2 x float> [[TMP16]], ptr [[TMP7]], align 4, !alias.scope [[META4]], !noalias [[META7]]485; UNROLL-NO-IC-NEXT: store <2 x float> [[TMP17]], ptr [[TMP8]], align 4, !alias.scope [[META4]], !noalias [[META7]]486; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4487; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]488; UNROLL-NO-IC-NEXT: br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]489; UNROLL-NO-IC: middle.block:490; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]491; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]492; UNROLL-NO-IC: scalar.ph:493; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]494; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]495; UNROLL-NO-IC: for.body:496; UNROLL-NO-IC-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]497; UNROLL-NO-IC-NEXT: [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET]]498; UNROLL-NO-IC-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IND_SUM]]499; UNROLL-NO-IC-NEXT: [[L1:%.*]] = load float, ptr [[ARR_IDX]], align 4500; UNROLL-NO-IC-NEXT: [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2]]501; UNROLL-NO-IC-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IND_SUM2]]502; UNROLL-NO-IC-NEXT: [[L2:%.*]] = load float, ptr [[ARR_IDX2]], align 4503; UNROLL-NO-IC-NEXT: [[M:%.*]] = fmul fast float [[B]], [[L2]]504; UNROLL-NO-IC-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]]505; UNROLL-NO-IC-NEXT: store float [[AD]], ptr [[ARR_IDX]], align 4506; UNROLL-NO-IC-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1507; UNROLL-NO-IC-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]508; UNROLL-NO-IC-NEXT: br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]509; UNROLL-NO-IC: loopexit:510; UNROLL-NO-IC-NEXT: ret void511;512; INTERLEAVE-LABEL: @scalar_use(513; INTERLEAVE-NEXT: entry:514; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8515; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]516; INTERLEAVE: vector.memcheck:517; INTERLEAVE-NEXT: [[TMP0:%.*]] = shl i64 [[OFFSET:%.*]], 2518; INTERLEAVE-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP0]]519; INTERLEAVE-NEXT: [[TMP1:%.*]] = shl i64 [[N]], 2520; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP1]]521; INTERLEAVE-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP0]]522; INTERLEAVE-NEXT: [[TMP3:%.*]] = shl i64 [[OFFSET2:%.*]], 2523; INTERLEAVE-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]]524; INTERLEAVE-NEXT: [[TMP4:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP1]]525; INTERLEAVE-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[TMP4]], i64 [[TMP3]]526; INTERLEAVE-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP3]]527; INTERLEAVE-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[SCEVGEP2]], [[SCEVGEP1]]528; INTERLEAVE-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]529; INTERLEAVE-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]530; INTERLEAVE: vector.ph:531; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i64 [[N]], -8532; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[B:%.*]], i64 0533; INTERLEAVE-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x float> [[BROADCAST_SPLATINSERT]], <4 x float> poison, <4 x i32> zeroinitializer534; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]535; INTERLEAVE: vector.body:536; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]537; INTERLEAVE-NEXT: [[TMP5:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]]538; INTERLEAVE-NEXT: [[TMP6:%.*]] = getelementptr float, ptr [[TMP5]], i64 [[OFFSET]]539; INTERLEAVE-NEXT: [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP6]], i64 16540; INTERLEAVE-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP6]], align 4, !alias.scope [[META4:![0-9]+]], !noalias [[META7:![0-9]+]]541; INTERLEAVE-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP7]], align 4, !alias.scope [[META4]], !noalias [[META7]]542; INTERLEAVE-NEXT: [[TMP8:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]]543; INTERLEAVE-NEXT: [[TMP9:%.*]] = getelementptr float, ptr [[TMP8]], i64 [[OFFSET2]]544; INTERLEAVE-NEXT: [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i64 16545; INTERLEAVE-NEXT: [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP9]], align 4, !alias.scope [[META7]]546; INTERLEAVE-NEXT: [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP10]], align 4, !alias.scope [[META7]]547; INTERLEAVE-NEXT: [[TMP11:%.*]] = fmul fast <4 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD5]]548; INTERLEAVE-NEXT: [[TMP12:%.*]] = fmul fast <4 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD6]]549; INTERLEAVE-NEXT: [[TMP13:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], [[TMP11]]550; INTERLEAVE-NEXT: [[TMP14:%.*]] = fadd fast <4 x float> [[WIDE_LOAD4]], [[TMP12]]551; INTERLEAVE-NEXT: store <4 x float> [[TMP13]], ptr [[TMP6]], align 4, !alias.scope [[META4]], !noalias [[META7]]552; INTERLEAVE-NEXT: store <4 x float> [[TMP14]], ptr [[TMP7]], align 4, !alias.scope [[META4]], !noalias [[META7]]553; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8554; INTERLEAVE-NEXT: [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]555; INTERLEAVE-NEXT: br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]556; INTERLEAVE: middle.block:557; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]558; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]559; INTERLEAVE: scalar.ph:560; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]561; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]562; INTERLEAVE: for.body:563; INTERLEAVE-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]564; INTERLEAVE-NEXT: [[TMP16:%.*]] = getelementptr float, ptr [[A]], i64 [[IV]]565; INTERLEAVE-NEXT: [[ARR_IDX:%.*]] = getelementptr float, ptr [[TMP16]], i64 [[OFFSET]]566; INTERLEAVE-NEXT: [[L1:%.*]] = load float, ptr [[ARR_IDX]], align 4567; INTERLEAVE-NEXT: [[TMP17:%.*]] = getelementptr float, ptr [[A]], i64 [[IV]]568; INTERLEAVE-NEXT: [[ARR_IDX2:%.*]] = getelementptr float, ptr [[TMP17]], i64 [[OFFSET2]]569; INTERLEAVE-NEXT: [[L2:%.*]] = load float, ptr [[ARR_IDX2]], align 4570; INTERLEAVE-NEXT: [[M:%.*]] = fmul fast float [[B]], [[L2]]571; INTERLEAVE-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]]572; INTERLEAVE-NEXT: store float [[AD]], ptr [[ARR_IDX]], align 4573; INTERLEAVE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1574; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]575; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]576; INTERLEAVE: loopexit:577; INTERLEAVE-NEXT: ret void578;579entry:580 br label %for.body581 582for.body:583 %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]584 %ind.sum = add i64 %iv, %offset585 %arr.idx = getelementptr inbounds float, ptr %a, i64 %ind.sum586 %l1 = load float, ptr %arr.idx, align 4587 %ind.sum2 = add i64 %iv, %offset2588 %arr.idx2 = getelementptr inbounds float, ptr %a, i64 %ind.sum2589 %l2 = load float, ptr %arr.idx2, align 4590 %m = fmul fast float %b, %l2591 %ad = fadd fast float %l1, %m592 store float %ad, ptr %arr.idx, align 4593 %iv.next = add nuw nsw i64 %iv, 1594 %exitcond = icmp eq i64 %iv.next, %n595 br i1 %exitcond, label %loopexit, label %for.body596 597loopexit:598 ret void599}600 601; Make sure we don't create a vector induction phi node that is unused.602; Scalarize the step vectors instead.603;604; for (int i = 0; i < n; ++i)605; sum += a[i];606;607;608;609;610 611define i64 @scalarize_induction_variable_01(ptr %a, i64 %n) {612; CHECK-LABEL: @scalarize_induction_variable_01(613; CHECK-NEXT: entry:614; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)615; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2616; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]617; CHECK: vector.ph:618; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 2619; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]620; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]621; CHECK: vector.body:622; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]623; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]624; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]625; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 8626; CHECK-NEXT: [[TMP3]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]627; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2628; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]629; CHECK-NEXT: br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]630; CHECK: middle.block:631; CHECK-NEXT: [[TMP5:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[TMP3]])632; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]633; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]634; CHECK: scalar.ph:635; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]636; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP5]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]637; CHECK-NEXT: br label [[FOR_BODY:%.*]]638; CHECK: for.body:639; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]640; CHECK-NEXT: [[SUM:%.*]] = phi i64 [ [[TMP8:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]641; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[I]]642; CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr [[TMP6]], align 8643; CHECK-NEXT: [[TMP8]] = add i64 [[TMP7]], [[SUM]]644; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1645; CHECK-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]646; CHECK-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP12:![0-9]+]]647; CHECK: for.end:648; CHECK-NEXT: [[TMP9:%.*]] = phi i64 [ [[TMP8]], [[FOR_BODY]] ], [ [[TMP5]], [[MIDDLE_BLOCK]] ]649; CHECK-NEXT: ret i64 [[TMP9]]650;651; IND-LABEL: @scalarize_induction_variable_01(652; IND-NEXT: entry:653; IND-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)654; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i64 [[N]], 2655; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]656; IND: vector.ph:657; IND-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775806658; IND-NEXT: br label [[VECTOR_BODY:%.*]]659; IND: vector.body:660; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]661; IND-NEXT: [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]662; IND-NEXT: [[TMP0:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]663; IND-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP0]], align 8664; IND-NEXT: [[TMP1]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]665; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2666; IND-NEXT: [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]667; IND-NEXT: br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]668; IND: middle.block:669; IND-NEXT: [[TMP3:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[TMP1]])670; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]671; IND-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]672; IND: scalar.ph:673; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]674; IND-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP3]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]675; IND-NEXT: br label [[FOR_BODY:%.*]]676; IND: for.body:677; IND-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]678; IND-NEXT: [[SUM:%.*]] = phi i64 [ [[TMP6:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]679; IND-NEXT: [[TMP4:%.*]] = getelementptr inbounds nuw i64, ptr [[A]], i64 [[I]]680; IND-NEXT: [[TMP5:%.*]] = load i64, ptr [[TMP4]], align 8681; IND-NEXT: [[TMP6]] = add i64 [[TMP5]], [[SUM]]682; IND-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1683; IND-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]684; IND-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP12:![0-9]+]]685; IND: for.end:686; IND-NEXT: [[TMP7:%.*]] = phi i64 [ [[TMP6]], [[FOR_BODY]] ], [ [[TMP3]], [[MIDDLE_BLOCK]] ]687; IND-NEXT: ret i64 [[TMP7]]688;689; UNROLL-LABEL: @scalarize_induction_variable_01(690; UNROLL-NEXT: entry:691; UNROLL-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)692; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i64 [[N]], 4693; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]694; UNROLL: vector.ph:695; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775804696; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]697; UNROLL: vector.body:698; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]699; UNROLL-NEXT: [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP2:%.*]], [[VECTOR_BODY]] ]700; UNROLL-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]701; UNROLL-NEXT: [[TMP0:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]702; UNROLL-NEXT: [[TMP1:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 16703; UNROLL-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP0]], align 8704; UNROLL-NEXT: [[WIDE_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP1]], align 8705; UNROLL-NEXT: [[TMP2]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]706; UNROLL-NEXT: [[TMP3]] = add <2 x i64> [[WIDE_LOAD2]], [[VEC_PHI1]]707; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4708; UNROLL-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]709; UNROLL-NEXT: br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]710; UNROLL: middle.block:711; UNROLL-NEXT: [[BIN_RDX:%.*]] = add <2 x i64> [[TMP3]], [[TMP2]]712; UNROLL-NEXT: [[TMP5:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])713; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]714; UNROLL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]715; UNROLL: scalar.ph:716; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]717; UNROLL-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP5]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]718; UNROLL-NEXT: br label [[FOR_BODY:%.*]]719; UNROLL: for.body:720; UNROLL-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]721; UNROLL-NEXT: [[SUM:%.*]] = phi i64 [ [[TMP8:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]722; UNROLL-NEXT: [[TMP6:%.*]] = getelementptr inbounds nuw i64, ptr [[A]], i64 [[I]]723; UNROLL-NEXT: [[TMP7:%.*]] = load i64, ptr [[TMP6]], align 8724; UNROLL-NEXT: [[TMP8]] = add i64 [[TMP7]], [[SUM]]725; UNROLL-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1726; UNROLL-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]727; UNROLL-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP12:![0-9]+]]728; UNROLL: for.end:729; UNROLL-NEXT: [[TMP9:%.*]] = phi i64 [ [[TMP8]], [[FOR_BODY]] ], [ [[TMP5]], [[MIDDLE_BLOCK]] ]730; UNROLL-NEXT: ret i64 [[TMP9]]731;732; UNROLL-NO-IC-LABEL: @scalarize_induction_variable_01(733; UNROLL-NO-IC-NEXT: entry:734; UNROLL-NO-IC-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)735; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 4736; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]737; UNROLL-NO-IC: vector.ph:738; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 4739; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]740; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]741; UNROLL-NO-IC: vector.body:742; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]743; UNROLL-NO-IC-NEXT: [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]744; UNROLL-NO-IC-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]745; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]746; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i64 2747; UNROLL-NO-IC-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 8748; UNROLL-NO-IC-NEXT: [[WIDE_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP3]], align 8749; UNROLL-NO-IC-NEXT: [[TMP4]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]750; UNROLL-NO-IC-NEXT: [[TMP5]] = add <2 x i64> [[WIDE_LOAD2]], [[VEC_PHI1]]751; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4752; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]753; UNROLL-NO-IC-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]754; UNROLL-NO-IC: middle.block:755; UNROLL-NO-IC-NEXT: [[BIN_RDX:%.*]] = add <2 x i64> [[TMP5]], [[TMP4]]756; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])757; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]758; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]759; UNROLL-NO-IC: scalar.ph:760; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]761; UNROLL-NO-IC-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]762; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]763; UNROLL-NO-IC: for.body:764; UNROLL-NO-IC-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]765; UNROLL-NO-IC-NEXT: [[SUM:%.*]] = phi i64 [ [[TMP10:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]766; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[I]]767; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = load i64, ptr [[TMP8]], align 8768; UNROLL-NO-IC-NEXT: [[TMP10]] = add i64 [[TMP9]], [[SUM]]769; UNROLL-NO-IC-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1770; UNROLL-NO-IC-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]771; UNROLL-NO-IC-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP12:![0-9]+]]772; UNROLL-NO-IC: for.end:773; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = phi i64 [ [[TMP10]], [[FOR_BODY]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]774; UNROLL-NO-IC-NEXT: ret i64 [[TMP11]]775;776; INTERLEAVE-LABEL: @scalarize_induction_variable_01(777; INTERLEAVE-NEXT: entry:778; INTERLEAVE-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)779; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i64 [[N]], 8780; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]781; INTERLEAVE: vector.ph:782; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775800783; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]784; INTERLEAVE: vector.body:785; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]786; INTERLEAVE-NEXT: [[VEC_PHI:%.*]] = phi <4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP2:%.*]], [[VECTOR_BODY]] ]787; INTERLEAVE-NEXT: [[VEC_PHI1:%.*]] = phi <4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]788; INTERLEAVE-NEXT: [[TMP0:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]789; INTERLEAVE-NEXT: [[TMP1:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 32790; INTERLEAVE-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP0]], align 8791; INTERLEAVE-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8792; INTERLEAVE-NEXT: [[TMP2]] = add <4 x i64> [[WIDE_LOAD]], [[VEC_PHI]]793; INTERLEAVE-NEXT: [[TMP3]] = add <4 x i64> [[WIDE_LOAD2]], [[VEC_PHI1]]794; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8795; INTERLEAVE-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]796; INTERLEAVE-NEXT: br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]797; INTERLEAVE: middle.block:798; INTERLEAVE-NEXT: [[BIN_RDX:%.*]] = add <4 x i64> [[TMP3]], [[TMP2]]799; INTERLEAVE-NEXT: [[TMP5:%.*]] = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> [[BIN_RDX]])800; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]801; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]802; INTERLEAVE: scalar.ph:803; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]804; INTERLEAVE-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP5]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]805; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]806; INTERLEAVE: for.body:807; INTERLEAVE-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]808; INTERLEAVE-NEXT: [[SUM:%.*]] = phi i64 [ [[TMP8:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]809; INTERLEAVE-NEXT: [[TMP6:%.*]] = getelementptr inbounds nuw i64, ptr [[A]], i64 [[I]]810; INTERLEAVE-NEXT: [[TMP7:%.*]] = load i64, ptr [[TMP6]], align 8811; INTERLEAVE-NEXT: [[TMP8]] = add i64 [[TMP7]], [[SUM]]812; INTERLEAVE-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1813; INTERLEAVE-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]814; INTERLEAVE-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP12:![0-9]+]]815; INTERLEAVE: for.end:816; INTERLEAVE-NEXT: [[TMP9:%.*]] = phi i64 [ [[TMP8]], [[FOR_BODY]] ], [ [[TMP5]], [[MIDDLE_BLOCK]] ]817; INTERLEAVE-NEXT: ret i64 [[TMP9]]818;819entry:820 br label %for.body821 822for.body:823 %i = phi i64 [ %i.next, %for.body ], [ 0, %entry ]824 %sum = phi i64 [ %2, %for.body ], [ 0, %entry ]825 %0 = getelementptr inbounds i64, ptr %a, i64 %i826 %1 = load i64, ptr %0, align 8827 %2 = add i64 %1, %sum828 %i.next = add nuw nsw i64 %i, 1829 %cond = icmp slt i64 %i.next, %n830 br i1 %cond, label %for.body, label %for.end831 832for.end:833 %3 = phi i64 [ %2, %for.body ]834 ret i64 %3835}836 837; Make sure we scalarize the step vectors used for the pointer arithmetic. We838; can't easily simplify vectorized step vectors.839;840; float s = 0;841; for (int i ; 0; i < n; i += 8)842; s += (a[i] + b[i] + 1.0f);843;844;845;846;847 848define float @scalarize_induction_variable_02(ptr %a, ptr %b, i64 %n) {849; CHECK-LABEL: @scalarize_induction_variable_02(850; CHECK-NEXT: entry:851; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 8)852; CHECK-NEXT: [[TMP0:%.*]] = add nsw i64 [[SMAX]], -1853; CHECK-NEXT: [[TMP1:%.*]] = lshr i64 [[TMP0]], 3854; CHECK-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1855; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 2856; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]857; CHECK: vector.ph:858; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 2859; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]860; CHECK-NEXT: [[IND_END:%.*]] = mul i64 [[N_VEC]], 8861; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]862; CHECK: vector.body:863; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]864; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]865; CHECK-NEXT: [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8866; CHECK-NEXT: [[TMP3:%.*]] = add i64 [[OFFSET_IDX]], 0867; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 8868; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[TMP3]]869; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP4]]870; CHECK-NEXT: [[TMP7:%.*]] = load float, ptr [[TMP5]], align 4871; CHECK-NEXT: [[TMP8:%.*]] = load float, ptr [[TMP6]], align 4872; CHECK-NEXT: [[TMP9:%.*]] = insertelement <2 x float> poison, float [[TMP7]], i32 0873; CHECK-NEXT: [[TMP10:%.*]] = insertelement <2 x float> [[TMP9]], float [[TMP8]], i32 1874; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[TMP3]]875; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP4]]876; CHECK-NEXT: [[TMP13:%.*]] = load float, ptr [[TMP11]], align 4877; CHECK-NEXT: [[TMP14:%.*]] = load float, ptr [[TMP12]], align 4878; CHECK-NEXT: [[TMP15:%.*]] = insertelement <2 x float> poison, float [[TMP13]], i32 0879; CHECK-NEXT: [[TMP16:%.*]] = insertelement <2 x float> [[TMP15]], float [[TMP14]], i32 1880; CHECK-NEXT: [[TMP17:%.*]] = fadd fast <2 x float> [[VEC_PHI]], splat (float 1.000000e+00)881; CHECK-NEXT: [[TMP18:%.*]] = fadd fast <2 x float> [[TMP17]], [[TMP10]]882; CHECK-NEXT: [[TMP19]] = fadd fast <2 x float> [[TMP18]], [[TMP16]]883; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2884; CHECK-NEXT: [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]885; CHECK-NEXT: br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]886; CHECK: middle.block:887; CHECK-NEXT: [[TMP21:%.*]] = call fast float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> [[TMP19]])888; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]889; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]890; CHECK: scalar.ph:891; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]892; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP21]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]893; CHECK-NEXT: br label [[FOR_BODY:%.*]]894; CHECK: for.body:895; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_BODY]] ]896; CHECK-NEXT: [[S:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP28:%.*]], [[FOR_BODY]] ]897; CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[I]]898; CHECK-NEXT: [[TMP23:%.*]] = load float, ptr [[TMP22]], align 4899; CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[I]]900; CHECK-NEXT: [[TMP25:%.*]] = load float, ptr [[TMP24]], align 4901; CHECK-NEXT: [[TMP26:%.*]] = fadd fast float [[S]], 1.000000e+00902; CHECK-NEXT: [[TMP27:%.*]] = fadd fast float [[TMP26]], [[TMP23]]903; CHECK-NEXT: [[TMP28]] = fadd fast float [[TMP27]], [[TMP25]]904; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 8905; CHECK-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]906; CHECK-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP14:![0-9]+]]907; CHECK: for.end:908; CHECK-NEXT: [[S_LCSSA:%.*]] = phi float [ [[TMP28]], [[FOR_BODY]] ], [ [[TMP21]], [[MIDDLE_BLOCK]] ]909; CHECK-NEXT: ret float [[S_LCSSA]]910;911; IND-LABEL: @scalarize_induction_variable_02(912; IND-NEXT: entry:913; IND-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 8)914; IND-NEXT: [[TMP0:%.*]] = add nsw i64 [[SMAX]], -1915; IND-NEXT: [[TMP1:%.*]] = lshr i64 [[TMP0]], 3916; IND-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1917; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp eq i64 [[TMP1]], 0918; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]919; IND: vector.ph:920; IND-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 4611686018427387902921; IND-NEXT: [[IND_END:%.*]] = shl i64 [[N_VEC]], 3922; IND-NEXT: br label [[VECTOR_BODY:%.*]]923; IND: vector.body:924; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]925; IND-NEXT: [[VEC_PHI:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]926; IND-NEXT: [[OFFSET_IDX:%.*]] = shl i64 [[INDEX]], 3927; IND-NEXT: [[TMP3:%.*]] = or disjoint i64 [[OFFSET_IDX]], 8928; IND-NEXT: [[TMP4:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[OFFSET_IDX]]929; IND-NEXT: [[TMP5:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP3]]930; IND-NEXT: [[TMP6:%.*]] = load float, ptr [[TMP4]], align 4931; IND-NEXT: [[TMP7:%.*]] = load float, ptr [[TMP5]], align 4932; IND-NEXT: [[TMP8:%.*]] = insertelement <2 x float> poison, float [[TMP6]], i64 0933; IND-NEXT: [[TMP9:%.*]] = insertelement <2 x float> [[TMP8]], float [[TMP7]], i64 1934; IND-NEXT: [[TMP10:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[OFFSET_IDX]]935; IND-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP3]]936; IND-NEXT: [[TMP12:%.*]] = load float, ptr [[TMP10]], align 4937; IND-NEXT: [[TMP13:%.*]] = load float, ptr [[TMP11]], align 4938; IND-NEXT: [[TMP14:%.*]] = insertelement <2 x float> poison, float [[TMP12]], i64 0939; IND-NEXT: [[TMP15:%.*]] = insertelement <2 x float> [[TMP14]], float [[TMP13]], i64 1940; IND-NEXT: [[TMP16:%.*]] = fadd fast <2 x float> [[VEC_PHI]], splat (float 1.000000e+00)941; IND-NEXT: [[TMP17:%.*]] = fadd fast <2 x float> [[TMP16]], [[TMP9]]942; IND-NEXT: [[TMP18]] = fadd fast <2 x float> [[TMP17]], [[TMP15]]943; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2944; IND-NEXT: [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]945; IND-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]946; IND: middle.block:947; IND-NEXT: [[TMP20:%.*]] = call fast float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> [[TMP18]])948; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]949; IND-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]950; IND: scalar.ph:951; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]952; IND-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP20]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]953; IND-NEXT: br label [[FOR_BODY:%.*]]954; IND: for.body:955; IND-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_BODY]] ]956; IND-NEXT: [[S:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP27:%.*]], [[FOR_BODY]] ]957; IND-NEXT: [[TMP21:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[I]]958; IND-NEXT: [[TMP22:%.*]] = load float, ptr [[TMP21]], align 4959; IND-NEXT: [[TMP23:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[I]]960; IND-NEXT: [[TMP24:%.*]] = load float, ptr [[TMP23]], align 4961; IND-NEXT: [[TMP25:%.*]] = fadd fast float [[S]], 1.000000e+00962; IND-NEXT: [[TMP26:%.*]] = fadd fast float [[TMP25]], [[TMP22]]963; IND-NEXT: [[TMP27]] = fadd fast float [[TMP26]], [[TMP24]]964; IND-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 8965; IND-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]966; IND-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP14:![0-9]+]]967; IND: for.end:968; IND-NEXT: [[S_LCSSA:%.*]] = phi float [ [[TMP27]], [[FOR_BODY]] ], [ [[TMP20]], [[MIDDLE_BLOCK]] ]969; IND-NEXT: ret float [[S_LCSSA]]970;971; UNROLL-LABEL: @scalarize_induction_variable_02(972; UNROLL-NEXT: entry:973; UNROLL-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 8)974; UNROLL-NEXT: [[TMP0:%.*]] = add nsw i64 [[SMAX]], -1975; UNROLL-NEXT: [[TMP1:%.*]] = lshr i64 [[TMP0]], 3976; UNROLL-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1977; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i64 [[N]], 25978; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]979; UNROLL: vector.ph:980; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 4611686018427387900981; UNROLL-NEXT: [[IND_END:%.*]] = shl i64 [[N_VEC]], 3982; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]983; UNROLL: vector.body:984; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]985; UNROLL-NEXT: [[VEC_PHI:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP34:%.*]], [[VECTOR_BODY]] ]986; UNROLL-NEXT: [[VEC_PHI1:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP35:%.*]], [[VECTOR_BODY]] ]987; UNROLL-NEXT: [[OFFSET_IDX:%.*]] = shl i64 [[INDEX]], 3988; UNROLL-NEXT: [[TMP3:%.*]] = or disjoint i64 [[OFFSET_IDX]], 8989; UNROLL-NEXT: [[TMP4:%.*]] = or disjoint i64 [[OFFSET_IDX]], 16990; UNROLL-NEXT: [[TMP5:%.*]] = or disjoint i64 [[OFFSET_IDX]], 24991; UNROLL-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[OFFSET_IDX]]992; UNROLL-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP3]]993; UNROLL-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP4]]994; UNROLL-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP5]]995; UNROLL-NEXT: [[TMP10:%.*]] = load float, ptr [[TMP6]], align 4996; UNROLL-NEXT: [[TMP11:%.*]] = load float, ptr [[TMP7]], align 4997; UNROLL-NEXT: [[TMP12:%.*]] = insertelement <2 x float> poison, float [[TMP10]], i64 0998; UNROLL-NEXT: [[TMP13:%.*]] = insertelement <2 x float> [[TMP12]], float [[TMP11]], i64 1999; UNROLL-NEXT: [[TMP14:%.*]] = load float, ptr [[TMP8]], align 41000; UNROLL-NEXT: [[TMP15:%.*]] = load float, ptr [[TMP9]], align 41001; UNROLL-NEXT: [[TMP16:%.*]] = insertelement <2 x float> poison, float [[TMP14]], i64 01002; UNROLL-NEXT: [[TMP17:%.*]] = insertelement <2 x float> [[TMP16]], float [[TMP15]], i64 11003; UNROLL-NEXT: [[TMP18:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[OFFSET_IDX]]1004; UNROLL-NEXT: [[TMP19:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP3]]1005; UNROLL-NEXT: [[TMP20:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP4]]1006; UNROLL-NEXT: [[TMP21:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP5]]1007; UNROLL-NEXT: [[TMP22:%.*]] = load float, ptr [[TMP18]], align 41008; UNROLL-NEXT: [[TMP23:%.*]] = load float, ptr [[TMP19]], align 41009; UNROLL-NEXT: [[TMP24:%.*]] = insertelement <2 x float> poison, float [[TMP22]], i64 01010; UNROLL-NEXT: [[TMP25:%.*]] = insertelement <2 x float> [[TMP24]], float [[TMP23]], i64 11011; UNROLL-NEXT: [[TMP26:%.*]] = load float, ptr [[TMP20]], align 41012; UNROLL-NEXT: [[TMP27:%.*]] = load float, ptr [[TMP21]], align 41013; UNROLL-NEXT: [[TMP28:%.*]] = insertelement <2 x float> poison, float [[TMP26]], i64 01014; UNROLL-NEXT: [[TMP29:%.*]] = insertelement <2 x float> [[TMP28]], float [[TMP27]], i64 11015; UNROLL-NEXT: [[TMP30:%.*]] = fadd fast <2 x float> [[VEC_PHI]], splat (float 1.000000e+00)1016; UNROLL-NEXT: [[TMP31:%.*]] = fadd fast <2 x float> [[VEC_PHI1]], splat (float 1.000000e+00)1017; UNROLL-NEXT: [[TMP32:%.*]] = fadd fast <2 x float> [[TMP30]], [[TMP13]]1018; UNROLL-NEXT: [[TMP33:%.*]] = fadd fast <2 x float> [[TMP31]], [[TMP17]]1019; UNROLL-NEXT: [[TMP34]] = fadd fast <2 x float> [[TMP32]], [[TMP25]]1020; UNROLL-NEXT: [[TMP35]] = fadd fast <2 x float> [[TMP33]], [[TMP29]]1021; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 41022; UNROLL-NEXT: [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]1023; UNROLL-NEXT: br i1 [[TMP36]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]1024; UNROLL: middle.block:1025; UNROLL-NEXT: [[BIN_RDX:%.*]] = fadd fast <2 x float> [[TMP35]], [[TMP34]]1026; UNROLL-NEXT: [[TMP37:%.*]] = call fast float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> [[BIN_RDX]])1027; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]1028; UNROLL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]1029; UNROLL: scalar.ph:1030; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]1031; UNROLL-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP37]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]1032; UNROLL-NEXT: br label [[FOR_BODY:%.*]]1033; UNROLL: for.body:1034; UNROLL-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_BODY]] ]1035; UNROLL-NEXT: [[S:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP44:%.*]], [[FOR_BODY]] ]1036; UNROLL-NEXT: [[TMP38:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[I]]1037; UNROLL-NEXT: [[TMP39:%.*]] = load float, ptr [[TMP38]], align 41038; UNROLL-NEXT: [[TMP40:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[I]]1039; UNROLL-NEXT: [[TMP41:%.*]] = load float, ptr [[TMP40]], align 41040; UNROLL-NEXT: [[TMP42:%.*]] = fadd fast float [[S]], 1.000000e+001041; UNROLL-NEXT: [[TMP43:%.*]] = fadd fast float [[TMP42]], [[TMP39]]1042; UNROLL-NEXT: [[TMP44]] = fadd fast float [[TMP43]], [[TMP41]]1043; UNROLL-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 81044; UNROLL-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]1045; UNROLL-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP14:![0-9]+]]1046; UNROLL: for.end:1047; UNROLL-NEXT: [[S_LCSSA:%.*]] = phi float [ [[TMP44]], [[FOR_BODY]] ], [ [[TMP37]], [[MIDDLE_BLOCK]] ]1048; UNROLL-NEXT: ret float [[S_LCSSA]]1049;1050; UNROLL-NO-IC-LABEL: @scalarize_induction_variable_02(1051; UNROLL-NO-IC-NEXT: entry:1052; UNROLL-NO-IC-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 8)1053; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add nsw i64 [[SMAX]], -11054; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = lshr i64 [[TMP0]], 31055; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 11056; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 41057; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]1058; UNROLL-NO-IC: vector.ph:1059; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 41060; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]1061; UNROLL-NO-IC-NEXT: [[IND_END:%.*]] = mul i64 [[N_VEC]], 81062; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]1063; UNROLL-NO-IC: vector.body:1064; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]1065; UNROLL-NO-IC-NEXT: [[VEC_PHI:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP35:%.*]], [[VECTOR_BODY]] ]1066; UNROLL-NO-IC-NEXT: [[VEC_PHI1:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP36:%.*]], [[VECTOR_BODY]] ]1067; UNROLL-NO-IC-NEXT: [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 81068; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = add i64 [[OFFSET_IDX]], 01069; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 81070; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = add i64 [[OFFSET_IDX]], 161071; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = add i64 [[OFFSET_IDX]], 241072; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[TMP3]]1073; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP4]]1074; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP5]]1075; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP6]]1076; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = load float, ptr [[TMP7]], align 41077; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = load float, ptr [[TMP8]], align 41078; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = insertelement <2 x float> poison, float [[TMP11]], i32 01079; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = insertelement <2 x float> [[TMP13]], float [[TMP12]], i32 11080; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = load float, ptr [[TMP9]], align 41081; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = load float, ptr [[TMP10]], align 41082; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = insertelement <2 x float> poison, float [[TMP15]], i32 01083; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = insertelement <2 x float> [[TMP17]], float [[TMP16]], i32 11084; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[TMP3]]1085; UNROLL-NO-IC-NEXT: [[TMP20:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP4]]1086; UNROLL-NO-IC-NEXT: [[TMP21:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP5]]1087; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP6]]1088; UNROLL-NO-IC-NEXT: [[TMP23:%.*]] = load float, ptr [[TMP19]], align 41089; UNROLL-NO-IC-NEXT: [[TMP24:%.*]] = load float, ptr [[TMP20]], align 41090; UNROLL-NO-IC-NEXT: [[TMP25:%.*]] = insertelement <2 x float> poison, float [[TMP23]], i32 01091; UNROLL-NO-IC-NEXT: [[TMP26:%.*]] = insertelement <2 x float> [[TMP25]], float [[TMP24]], i32 11092; UNROLL-NO-IC-NEXT: [[TMP27:%.*]] = load float, ptr [[TMP21]], align 41093; UNROLL-NO-IC-NEXT: [[TMP28:%.*]] = load float, ptr [[TMP22]], align 41094; UNROLL-NO-IC-NEXT: [[TMP29:%.*]] = insertelement <2 x float> poison, float [[TMP27]], i32 01095; UNROLL-NO-IC-NEXT: [[TMP30:%.*]] = insertelement <2 x float> [[TMP29]], float [[TMP28]], i32 11096; UNROLL-NO-IC-NEXT: [[TMP31:%.*]] = fadd fast <2 x float> [[VEC_PHI]], splat (float 1.000000e+00)1097; UNROLL-NO-IC-NEXT: [[TMP32:%.*]] = fadd fast <2 x float> [[VEC_PHI1]], splat (float 1.000000e+00)1098; UNROLL-NO-IC-NEXT: [[TMP33:%.*]] = fadd fast <2 x float> [[TMP31]], [[TMP14]]1099; UNROLL-NO-IC-NEXT: [[TMP34:%.*]] = fadd fast <2 x float> [[TMP32]], [[TMP18]]1100; UNROLL-NO-IC-NEXT: [[TMP35]] = fadd fast <2 x float> [[TMP33]], [[TMP26]]1101; UNROLL-NO-IC-NEXT: [[TMP36]] = fadd fast <2 x float> [[TMP34]], [[TMP30]]1102; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 41103; UNROLL-NO-IC-NEXT: [[TMP37:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]1104; UNROLL-NO-IC-NEXT: br i1 [[TMP37]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]1105; UNROLL-NO-IC: middle.block:1106; UNROLL-NO-IC-NEXT: [[BIN_RDX:%.*]] = fadd fast <2 x float> [[TMP36]], [[TMP35]]1107; UNROLL-NO-IC-NEXT: [[TMP38:%.*]] = call fast float @llvm.vector.reduce.fadd.v2f32(float 0.000000e+00, <2 x float> [[BIN_RDX]])1108; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]1109; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]1110; UNROLL-NO-IC: scalar.ph:1111; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]1112; UNROLL-NO-IC-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP38]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]1113; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]1114; UNROLL-NO-IC: for.body:1115; UNROLL-NO-IC-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_BODY]] ]1116; UNROLL-NO-IC-NEXT: [[S:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP45:%.*]], [[FOR_BODY]] ]1117; UNROLL-NO-IC-NEXT: [[TMP39:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[I]]1118; UNROLL-NO-IC-NEXT: [[TMP40:%.*]] = load float, ptr [[TMP39]], align 41119; UNROLL-NO-IC-NEXT: [[TMP41:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[I]]1120; UNROLL-NO-IC-NEXT: [[TMP42:%.*]] = load float, ptr [[TMP41]], align 41121; UNROLL-NO-IC-NEXT: [[TMP43:%.*]] = fadd fast float [[S]], 1.000000e+001122; UNROLL-NO-IC-NEXT: [[TMP44:%.*]] = fadd fast float [[TMP43]], [[TMP40]]1123; UNROLL-NO-IC-NEXT: [[TMP45]] = fadd fast float [[TMP44]], [[TMP42]]1124; UNROLL-NO-IC-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 81125; UNROLL-NO-IC-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]1126; UNROLL-NO-IC-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP14:![0-9]+]]1127; UNROLL-NO-IC: for.end:1128; UNROLL-NO-IC-NEXT: [[S_LCSSA:%.*]] = phi float [ [[TMP45]], [[FOR_BODY]] ], [ [[TMP38]], [[MIDDLE_BLOCK]] ]1129; UNROLL-NO-IC-NEXT: ret float [[S_LCSSA]]1130;1131; INTERLEAVE-LABEL: @scalarize_induction_variable_02(1132; INTERLEAVE-NEXT: entry:1133; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i64 [[N:%.*]], 651134; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]1135; INTERLEAVE: vector.ph:1136; INTERLEAVE-NEXT: [[TMP0:%.*]] = add nsw i64 [[N]], -11137; INTERLEAVE-NEXT: [[TMP1:%.*]] = lshr i64 [[TMP0]], 31138; INTERLEAVE-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 11139; INTERLEAVE-NEXT: [[N_MOD_VF:%.*]] = and i64 [[TMP2]], 71140; INTERLEAVE-NEXT: [[TMP3:%.*]] = icmp eq i64 [[N_MOD_VF]], 01141; INTERLEAVE-NEXT: [[TMP4:%.*]] = select i1 [[TMP3]], i64 8, i64 [[N_MOD_VF]]1142; INTERLEAVE-NEXT: [[N_VEC:%.*]] = sub nsw i64 [[TMP2]], [[TMP4]]1143; INTERLEAVE-NEXT: [[IND_END:%.*]] = shl i64 [[N_VEC]], 31144; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]1145; INTERLEAVE: vector.body:1146; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]1147; INTERLEAVE-NEXT: [[VEC_PHI:%.*]] = phi <4 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]1148; INTERLEAVE-NEXT: [[VEC_PHI1:%.*]] = phi <4 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]1149; INTERLEAVE-NEXT: [[OFFSET_IDX:%.*]] = shl i64 [[INDEX]], 31150; INTERLEAVE-NEXT: [[TMP5:%.*]] = or disjoint i64 [[OFFSET_IDX]], 321151; INTERLEAVE-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[OFFSET_IDX]]1152; INTERLEAVE-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP5]]1153; INTERLEAVE-NEXT: [[WIDE_VEC:%.*]] = load <32 x float>, ptr [[TMP6]], align 41154; INTERLEAVE-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <32 x float> [[WIDE_VEC]], <32 x float> poison, <4 x i32> <i32 0, i32 8, i32 16, i32 24>1155; INTERLEAVE-NEXT: [[WIDE_VEC2:%.*]] = load <32 x float>, ptr [[TMP7]], align 41156; INTERLEAVE-NEXT: [[STRIDED_VEC3:%.*]] = shufflevector <32 x float> [[WIDE_VEC2]], <32 x float> poison, <4 x i32> <i32 0, i32 8, i32 16, i32 24>1157; INTERLEAVE-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[OFFSET_IDX]]1158; INTERLEAVE-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP5]]1159; INTERLEAVE-NEXT: [[WIDE_VEC4:%.*]] = load <32 x float>, ptr [[TMP8]], align 41160; INTERLEAVE-NEXT: [[STRIDED_VEC5:%.*]] = shufflevector <32 x float> [[WIDE_VEC4]], <32 x float> poison, <4 x i32> <i32 0, i32 8, i32 16, i32 24>1161; INTERLEAVE-NEXT: [[WIDE_VEC6:%.*]] = load <32 x float>, ptr [[TMP9]], align 41162; INTERLEAVE-NEXT: [[STRIDED_VEC7:%.*]] = shufflevector <32 x float> [[WIDE_VEC6]], <32 x float> poison, <4 x i32> <i32 0, i32 8, i32 16, i32 24>1163; INTERLEAVE-NEXT: [[TMP10:%.*]] = fadd fast <4 x float> [[VEC_PHI]], splat (float 1.000000e+00)1164; INTERLEAVE-NEXT: [[TMP11:%.*]] = fadd fast <4 x float> [[VEC_PHI1]], splat (float 1.000000e+00)1165; INTERLEAVE-NEXT: [[TMP12:%.*]] = fadd fast <4 x float> [[TMP10]], [[STRIDED_VEC]]1166; INTERLEAVE-NEXT: [[TMP13:%.*]] = fadd fast <4 x float> [[TMP11]], [[STRIDED_VEC3]]1167; INTERLEAVE-NEXT: [[TMP14]] = fadd fast <4 x float> [[TMP12]], [[STRIDED_VEC5]]1168; INTERLEAVE-NEXT: [[TMP15]] = fadd fast <4 x float> [[TMP13]], [[STRIDED_VEC7]]1169; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 81170; INTERLEAVE-NEXT: [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]1171; INTERLEAVE-NEXT: br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]1172; INTERLEAVE: middle.block:1173; INTERLEAVE-NEXT: [[BIN_RDX:%.*]] = fadd fast <4 x float> [[TMP15]], [[TMP14]]1174; INTERLEAVE-NEXT: [[TMP17:%.*]] = call fast float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> [[BIN_RDX]])1175; INTERLEAVE-NEXT: br label [[SCALAR_PH]]1176; INTERLEAVE: scalar.ph:1177; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]1178; INTERLEAVE-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP17]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]1179; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]1180; INTERLEAVE: for.body:1181; INTERLEAVE-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_BODY]] ]1182; INTERLEAVE-NEXT: [[S:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP24:%.*]], [[FOR_BODY]] ]1183; INTERLEAVE-NEXT: [[TMP18:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[I]]1184; INTERLEAVE-NEXT: [[TMP19:%.*]] = load float, ptr [[TMP18]], align 41185; INTERLEAVE-NEXT: [[TMP20:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[I]]1186; INTERLEAVE-NEXT: [[TMP21:%.*]] = load float, ptr [[TMP20]], align 41187; INTERLEAVE-NEXT: [[TMP22:%.*]] = fadd fast float [[S]], 1.000000e+001188; INTERLEAVE-NEXT: [[TMP23:%.*]] = fadd fast float [[TMP22]], [[TMP19]]1189; INTERLEAVE-NEXT: [[TMP24]] = fadd fast float [[TMP23]], [[TMP21]]1190; INTERLEAVE-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 81191; INTERLEAVE-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]1192; INTERLEAVE-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END:%.*]], !llvm.loop [[LOOP14:![0-9]+]]1193; INTERLEAVE: for.end:1194; INTERLEAVE-NEXT: ret float [[TMP24]]1195;1196entry:1197 br label %for.body1198 1199for.body:1200 %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]1201 %s = phi float [ 0.0, %entry ], [ %6, %for.body ]1202 %0 = getelementptr inbounds float, ptr %a, i64 %i1203 %1 = load float, ptr %0, align 41204 %2 = getelementptr inbounds float, ptr %b, i64 %i1205 %3 = load float, ptr %2, align 41206 %4 = fadd fast float %s, 1.01207 %5 = fadd fast float %4, %11208 %6 = fadd fast float %5, %31209 %i.next = add nuw nsw i64 %i, 81210 %cond = icmp slt i64 %i.next, %n1211 br i1 %cond, label %for.body, label %for.end1212 1213for.end:1214 %s.lcssa = phi float [ %6, %for.body ]1215 ret float %s.lcssa1216}1217 1218; Make sure we scalarize the step vectors used for the pointer arithmetic. We1219; can't easily simplify vectorized step vectors. (Interleaved accesses.)1220;1221; for (int i = 0; i < n; ++i)1222; a[i].f ^= y;1223;1224 1225%pair.i32 = type { i32, i32 }1226define void @scalarize_induction_variable_03(ptr %p, i32 %y, i64 %n) {1227; CHECK-LABEL: @scalarize_induction_variable_03(1228; CHECK-NEXT: entry:1229; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)1230; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 21231; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]1232; CHECK: vector.ph:1233; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 21234; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]1235; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[Y:%.*]], i64 01236; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer1237; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]1238; CHECK: vector.body:1239; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]1240; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 01241; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 11242; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], ptr [[P:%.*]], i64 [[TMP0]], i32 11243; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[PAIR_I32]], ptr [[P]], i64 [[TMP1]], i32 11244; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[TMP2]], align 81245; CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[TMP3]], align 81246; CHECK-NEXT: [[TMP6:%.*]] = insertelement <2 x i32> poison, i32 [[TMP4]], i32 01247; CHECK-NEXT: [[TMP7:%.*]] = insertelement <2 x i32> [[TMP6]], i32 [[TMP5]], i32 11248; CHECK-NEXT: [[TMP8:%.*]] = xor <2 x i32> [[TMP7]], [[BROADCAST_SPLAT]]1249; CHECK-NEXT: [[TMP9:%.*]] = extractelement <2 x i32> [[TMP8]], i32 01250; CHECK-NEXT: [[TMP10:%.*]] = extractelement <2 x i32> [[TMP8]], i32 11251; CHECK-NEXT: store i32 [[TMP9]], ptr [[TMP2]], align 81252; CHECK-NEXT: store i32 [[TMP10]], ptr [[TMP3]], align 81253; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 21254; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]1255; CHECK-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]1256; CHECK: middle.block:1257; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]1258; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]1259; CHECK: scalar.ph:1260; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]1261; CHECK-NEXT: br label [[FOR_BODY:%.*]]1262; CHECK: for.body:1263; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]1264; CHECK-NEXT: [[F:%.*]] = getelementptr inbounds [[PAIR_I32]], ptr [[P]], i64 [[I]], i32 11265; CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[F]], align 81266; CHECK-NEXT: [[TMP13:%.*]] = xor i32 [[TMP12]], [[Y]]1267; CHECK-NEXT: store i32 [[TMP13]], ptr [[F]], align 81268; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 11269; CHECK-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]1270; CHECK-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP16:![0-9]+]]1271; CHECK: for.end:1272; CHECK-NEXT: ret void1273;1274; IND-LABEL: @scalarize_induction_variable_03(1275; IND-NEXT: entry:1276; IND-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)1277; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i64 [[N]], 21278; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]1279; IND: vector.ph:1280; IND-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 92233720368547758061281; IND-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[Y:%.*]], i64 01282; IND-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer1283; IND-NEXT: br label [[VECTOR_BODY:%.*]]1284; IND: vector.body:1285; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]1286; IND-NEXT: [[DOTSPLIT:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], ptr [[P:%.*]], i64 [[INDEX]]1287; IND-NEXT: [[TMP1:%.*]] = getelementptr inbounds nuw i8, ptr [[DOTSPLIT]], i64 41288; IND-NEXT: [[TMP13:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1289; IND-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[TMP13]], i64 121290; IND-NEXT: [[TMP3:%.*]] = load i32, ptr [[TMP1]], align 81291; IND-NEXT: [[TMP4:%.*]] = load i32, ptr [[TMP2]], align 81292; IND-NEXT: [[TMP5:%.*]] = insertelement <2 x i32> poison, i32 [[TMP3]], i64 01293; IND-NEXT: [[TMP6:%.*]] = insertelement <2 x i32> [[TMP5]], i32 [[TMP4]], i64 11294; IND-NEXT: [[TMP7:%.*]] = xor <2 x i32> [[TMP6]], [[BROADCAST_SPLAT]]1295; IND-NEXT: [[TMP8:%.*]] = extractelement <2 x i32> [[TMP7]], i64 01296; IND-NEXT: [[TMP9:%.*]] = extractelement <2 x i32> [[TMP7]], i64 11297; IND-NEXT: store i32 [[TMP8]], ptr [[TMP1]], align 81298; IND-NEXT: store i32 [[TMP9]], ptr [[TMP2]], align 81299; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 21300; IND-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]1301; IND-NEXT: br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]1302; IND: middle.block:1303; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]1304; IND-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]1305; IND: scalar.ph:1306; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]1307; IND-NEXT: br label [[FOR_BODY:%.*]]1308; IND: for.body:1309; IND-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]1310; IND-NEXT: [[F_SPLIT:%.*]] = getelementptr inbounds nuw [[PAIR_I32]], ptr [[P]], i64 [[I]]1311; IND-NEXT: [[F:%.*]] = getelementptr inbounds nuw i8, ptr [[F_SPLIT]], i64 41312; IND-NEXT: [[TMP11:%.*]] = load i32, ptr [[F]], align 81313; IND-NEXT: [[TMP12:%.*]] = xor i32 [[TMP11]], [[Y]]1314; IND-NEXT: store i32 [[TMP12]], ptr [[F]], align 81315; IND-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 11316; IND-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]1317; IND-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP16:![0-9]+]]1318; IND: for.end:1319; IND-NEXT: ret void1320;1321; UNROLL-LABEL: @scalarize_induction_variable_03(1322; UNROLL-NEXT: entry:1323; UNROLL-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)1324; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i64 [[N]], 41325; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]1326; UNROLL: vector.ph:1327; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 92233720368547758041328; UNROLL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[Y:%.*]], i64 01329; UNROLL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer1330; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]1331; UNROLL: vector.body:1332; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]1333; UNROLL-NEXT: [[DOTSPLIT:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], ptr [[P:%.*]], i64 [[INDEX]]1334; UNROLL-NEXT: [[TMP3:%.*]] = getelementptr inbounds nuw i8, ptr [[DOTSPLIT]], i64 41335; UNROLL-NEXT: [[TMP1:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1336; UNROLL-NEXT: [[TMP4:%.*]] = getelementptr i8, ptr [[TMP1]], i64 121337; UNROLL-NEXT: [[TMP24:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1338; UNROLL-NEXT: [[TMP5:%.*]] = getelementptr i8, ptr [[TMP24]], i64 201339; UNROLL-NEXT: [[TMP25:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1340; UNROLL-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr [[TMP25]], i64 281341; UNROLL-NEXT: [[TMP7:%.*]] = load i32, ptr [[TMP3]], align 81342; UNROLL-NEXT: [[TMP8:%.*]] = load i32, ptr [[TMP4]], align 81343; UNROLL-NEXT: [[TMP9:%.*]] = insertelement <2 x i32> poison, i32 [[TMP7]], i64 01344; UNROLL-NEXT: [[TMP10:%.*]] = insertelement <2 x i32> [[TMP9]], i32 [[TMP8]], i64 11345; UNROLL-NEXT: [[TMP11:%.*]] = load i32, ptr [[TMP5]], align 81346; UNROLL-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP6]], align 81347; UNROLL-NEXT: [[TMP13:%.*]] = insertelement <2 x i32> poison, i32 [[TMP11]], i64 01348; UNROLL-NEXT: [[TMP14:%.*]] = insertelement <2 x i32> [[TMP13]], i32 [[TMP12]], i64 11349; UNROLL-NEXT: [[TMP15:%.*]] = xor <2 x i32> [[TMP10]], [[BROADCAST_SPLAT]]1350; UNROLL-NEXT: [[TMP17:%.*]] = extractelement <2 x i32> [[TMP15]], i64 01351; UNROLL-NEXT: [[TMP18:%.*]] = extractelement <2 x i32> [[TMP15]], i64 11352; UNROLL-NEXT: [[TMP16:%.*]] = xor <2 x i32> [[TMP14]], [[BROADCAST_SPLAT]]1353; UNROLL-NEXT: [[TMP19:%.*]] = extractelement <2 x i32> [[TMP16]], i64 01354; UNROLL-NEXT: [[TMP20:%.*]] = extractelement <2 x i32> [[TMP16]], i64 11355; UNROLL-NEXT: store i32 [[TMP17]], ptr [[TMP3]], align 81356; UNROLL-NEXT: store i32 [[TMP18]], ptr [[TMP4]], align 81357; UNROLL-NEXT: store i32 [[TMP19]], ptr [[TMP5]], align 81358; UNROLL-NEXT: store i32 [[TMP20]], ptr [[TMP6]], align 81359; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 41360; UNROLL-NEXT: [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]1361; UNROLL-NEXT: br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]1362; UNROLL: middle.block:1363; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]1364; UNROLL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]1365; UNROLL: scalar.ph:1366; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]1367; UNROLL-NEXT: br label [[FOR_BODY:%.*]]1368; UNROLL: for.body:1369; UNROLL-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]1370; UNROLL-NEXT: [[F_SPLIT:%.*]] = getelementptr inbounds nuw [[PAIR_I32]], ptr [[P]], i64 [[I]]1371; UNROLL-NEXT: [[F:%.*]] = getelementptr inbounds nuw i8, ptr [[F_SPLIT]], i64 41372; UNROLL-NEXT: [[TMP22:%.*]] = load i32, ptr [[F]], align 81373; UNROLL-NEXT: [[TMP23:%.*]] = xor i32 [[TMP22]], [[Y]]1374; UNROLL-NEXT: store i32 [[TMP23]], ptr [[F]], align 81375; UNROLL-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 11376; UNROLL-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]1377; UNROLL-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP16:![0-9]+]]1378; UNROLL: for.end:1379; UNROLL-NEXT: ret void1380;1381; UNROLL-NO-IC-LABEL: @scalarize_induction_variable_03(1382; UNROLL-NO-IC-NEXT: entry:1383; UNROLL-NO-IC-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)1384; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 41385; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]1386; UNROLL-NO-IC: vector.ph:1387; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 41388; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]1389; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[Y:%.*]], i64 01390; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer1391; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]1392; UNROLL-NO-IC: vector.body:1393; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]1394; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 01395; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 11396; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], 21397; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 31398; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], ptr [[P:%.*]], i64 [[TMP0]], i32 11399; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[PAIR_I32]], ptr [[P]], i64 [[TMP1]], i32 11400; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[PAIR_I32]], ptr [[P]], i64 [[TMP2]], i32 11401; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[PAIR_I32]], ptr [[P]], i64 [[TMP3]], i32 11402; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = load i32, ptr [[TMP4]], align 81403; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = load i32, ptr [[TMP5]], align 81404; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = insertelement <2 x i32> poison, i32 [[TMP8]], i32 01405; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = insertelement <2 x i32> [[TMP10]], i32 [[TMP9]], i32 11406; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP6]], align 81407; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = load i32, ptr [[TMP7]], align 81408; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = insertelement <2 x i32> poison, i32 [[TMP12]], i32 01409; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = insertelement <2 x i32> [[TMP14]], i32 [[TMP13]], i32 11410; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = xor <2 x i32> [[TMP11]], [[BROADCAST_SPLAT]]1411; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = extractelement <2 x i32> [[TMP16]], i32 01412; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = extractelement <2 x i32> [[TMP16]], i32 11413; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = xor <2 x i32> [[TMP15]], [[BROADCAST_SPLAT]]1414; UNROLL-NO-IC-NEXT: [[TMP20:%.*]] = extractelement <2 x i32> [[TMP17]], i32 01415; UNROLL-NO-IC-NEXT: [[TMP21:%.*]] = extractelement <2 x i32> [[TMP17]], i32 11416; UNROLL-NO-IC-NEXT: store i32 [[TMP18]], ptr [[TMP4]], align 81417; UNROLL-NO-IC-NEXT: store i32 [[TMP19]], ptr [[TMP5]], align 81418; UNROLL-NO-IC-NEXT: store i32 [[TMP20]], ptr [[TMP6]], align 81419; UNROLL-NO-IC-NEXT: store i32 [[TMP21]], ptr [[TMP7]], align 81420; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 41421; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]1422; UNROLL-NO-IC-NEXT: br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]1423; UNROLL-NO-IC: middle.block:1424; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]1425; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]1426; UNROLL-NO-IC: scalar.ph:1427; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]1428; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]1429; UNROLL-NO-IC: for.body:1430; UNROLL-NO-IC-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]1431; UNROLL-NO-IC-NEXT: [[F:%.*]] = getelementptr inbounds [[PAIR_I32]], ptr [[P]], i64 [[I]], i32 11432; UNROLL-NO-IC-NEXT: [[TMP23:%.*]] = load i32, ptr [[F]], align 81433; UNROLL-NO-IC-NEXT: [[TMP24:%.*]] = xor i32 [[TMP23]], [[Y]]1434; UNROLL-NO-IC-NEXT: store i32 [[TMP24]], ptr [[F]], align 81435; UNROLL-NO-IC-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 11436; UNROLL-NO-IC-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]1437; UNROLL-NO-IC-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP16:![0-9]+]]1438; UNROLL-NO-IC: for.end:1439; UNROLL-NO-IC-NEXT: ret void1440;1441; INTERLEAVE-LABEL: @scalarize_induction_variable_03(1442; INTERLEAVE-NEXT: entry:1443; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i64 [[N:%.*]], 91444; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]1445; INTERLEAVE: vector.ph:1446; INTERLEAVE-NEXT: [[N_MOD_VF:%.*]] = and i64 [[N]], 71447; INTERLEAVE-NEXT: [[TMP0:%.*]] = icmp eq i64 [[N_MOD_VF]], 01448; INTERLEAVE-NEXT: [[TMP1:%.*]] = select i1 [[TMP0]], i64 8, i64 [[N_MOD_VF]]1449; INTERLEAVE-NEXT: [[N_VEC:%.*]] = sub nsw i64 [[N]], [[TMP1]]1450; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[Y:%.*]], i64 01451; INTERLEAVE-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer1452; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]1453; INTERLEAVE: vector.body:1454; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]1455; INTERLEAVE-NEXT: [[DOTSPLIT:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], ptr [[P:%.*]], i64 [[INDEX]]1456; INTERLEAVE-NEXT: [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[DOTSPLIT]], i64 41457; INTERLEAVE-NEXT: [[TMP3:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1458; INTERLEAVE-NEXT: [[TMP10:%.*]] = getelementptr i8, ptr [[TMP3]], i64 121459; INTERLEAVE-NEXT: [[TMP5:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1460; INTERLEAVE-NEXT: [[TMP11:%.*]] = getelementptr i8, ptr [[TMP5]], i64 201461; INTERLEAVE-NEXT: [[TMP7:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1462; INTERLEAVE-NEXT: [[TMP12:%.*]] = getelementptr i8, ptr [[TMP7]], i64 281463; INTERLEAVE-NEXT: [[TMP30:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1464; INTERLEAVE-NEXT: [[TMP13:%.*]] = getelementptr i8, ptr [[TMP30]], i64 361465; INTERLEAVE-NEXT: [[TMP31:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1466; INTERLEAVE-NEXT: [[TMP14:%.*]] = getelementptr i8, ptr [[TMP31]], i64 441467; INTERLEAVE-NEXT: [[TMP32:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1468; INTERLEAVE-NEXT: [[TMP15:%.*]] = getelementptr i8, ptr [[TMP32]], i64 521469; INTERLEAVE-NEXT: [[TMP33:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1470; INTERLEAVE-NEXT: [[TMP16:%.*]] = getelementptr i8, ptr [[TMP33]], i64 601471; INTERLEAVE-NEXT: [[WIDE_VEC:%.*]] = load <8 x i32>, ptr [[TMP9]], align 81472; INTERLEAVE-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <8 x i32> [[WIDE_VEC]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>1473; INTERLEAVE-NEXT: [[WIDE_VEC1:%.*]] = load <8 x i32>, ptr [[TMP13]], align 81474; INTERLEAVE-NEXT: [[STRIDED_VEC2:%.*]] = shufflevector <8 x i32> [[WIDE_VEC1]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>1475; INTERLEAVE-NEXT: [[TMP17:%.*]] = xor <4 x i32> [[STRIDED_VEC]], [[BROADCAST_SPLAT]]1476; INTERLEAVE-NEXT: [[TMP19:%.*]] = extractelement <4 x i32> [[TMP17]], i64 01477; INTERLEAVE-NEXT: [[TMP20:%.*]] = extractelement <4 x i32> [[TMP17]], i64 11478; INTERLEAVE-NEXT: [[TMP21:%.*]] = extractelement <4 x i32> [[TMP17]], i64 21479; INTERLEAVE-NEXT: [[TMP22:%.*]] = extractelement <4 x i32> [[TMP17]], i64 31480; INTERLEAVE-NEXT: [[TMP18:%.*]] = xor <4 x i32> [[STRIDED_VEC2]], [[BROADCAST_SPLAT]]1481; INTERLEAVE-NEXT: [[TMP23:%.*]] = extractelement <4 x i32> [[TMP18]], i64 01482; INTERLEAVE-NEXT: [[TMP24:%.*]] = extractelement <4 x i32> [[TMP18]], i64 11483; INTERLEAVE-NEXT: [[TMP25:%.*]] = extractelement <4 x i32> [[TMP18]], i64 21484; INTERLEAVE-NEXT: [[TMP26:%.*]] = extractelement <4 x i32> [[TMP18]], i64 31485; INTERLEAVE-NEXT: store i32 [[TMP19]], ptr [[TMP9]], align 81486; INTERLEAVE-NEXT: store i32 [[TMP20]], ptr [[TMP10]], align 81487; INTERLEAVE-NEXT: store i32 [[TMP21]], ptr [[TMP11]], align 81488; INTERLEAVE-NEXT: store i32 [[TMP22]], ptr [[TMP12]], align 81489; INTERLEAVE-NEXT: store i32 [[TMP23]], ptr [[TMP13]], align 81490; INTERLEAVE-NEXT: store i32 [[TMP24]], ptr [[TMP14]], align 81491; INTERLEAVE-NEXT: store i32 [[TMP25]], ptr [[TMP15]], align 81492; INTERLEAVE-NEXT: store i32 [[TMP26]], ptr [[TMP16]], align 81493; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 81494; INTERLEAVE-NEXT: [[TMP27:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]1495; INTERLEAVE-NEXT: br i1 [[TMP27]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]1496; INTERLEAVE: middle.block:1497; INTERLEAVE-NEXT: br label [[SCALAR_PH]]1498; INTERLEAVE: scalar.ph:1499; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]1500; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]1501; INTERLEAVE: for.body:1502; INTERLEAVE-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]1503; INTERLEAVE-NEXT: [[F_SPLIT:%.*]] = getelementptr inbounds [[PAIR_I32]], ptr [[P]], i64 [[I]]1504; INTERLEAVE-NEXT: [[F:%.*]] = getelementptr inbounds nuw i8, ptr [[F_SPLIT]], i64 41505; INTERLEAVE-NEXT: [[TMP28:%.*]] = load i32, ptr [[F]], align 81506; INTERLEAVE-NEXT: [[TMP29:%.*]] = xor i32 [[TMP28]], [[Y]]1507; INTERLEAVE-NEXT: store i32 [[TMP29]], ptr [[F]], align 81508; INTERLEAVE-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 11509; INTERLEAVE-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]1510; INTERLEAVE-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END:%.*]], !llvm.loop [[LOOP16:![0-9]+]]1511; INTERLEAVE: for.end:1512; INTERLEAVE-NEXT: ret void1513;1514entry:1515 br label %for.body1516 1517for.body:1518 %i = phi i64 [ %i.next, %for.body ], [ 0, %entry ]1519 %f = getelementptr inbounds %pair.i32, ptr %p, i64 %i, i32 11520 %0 = load i32, ptr %f, align 81521 %1 = xor i32 %0, %y1522 store i32 %1, ptr %f, align 81523 %i.next = add nuw nsw i64 %i, 11524 %cond = icmp slt i64 %i.next, %n1525 br i1 %cond, label %for.body, label %for.end1526 1527for.end:1528 ret void1529}1530 1531; Make sure we scalarize the step vectors used for the pointer arithmetic. We1532; can't easily simplify vectorized step vectors. (Interleaved accesses.)1533;1534; for (int i = 0; i < n; ++i)1535; p[i].f = a[i * 4]1536;1537 1538define void @scalarize_induction_variable_04(ptr %a, ptr %p, i32 %n) {1539; CHECK-LABEL: @scalarize_induction_variable_04(1540; CHECK-NEXT: entry:1541; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -11542; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i641543; CHECK-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 11544; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 21545; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]1546; CHECK: vector.memcheck:1547; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr [[P:%.*]], i64 41548; CHECK-NEXT: [[TMP3:%.*]] = add i32 [[N]], -11549; CHECK-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i641550; CHECK-NEXT: [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 31551; CHECK-NEXT: [[TMP6:%.*]] = add nuw nsw i64 [[TMP5]], 81552; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP6]]1553; CHECK-NEXT: [[TMP7:%.*]] = shl nuw nsw i64 [[TMP4]], 41554; CHECK-NEXT: [[TMP8:%.*]] = add nuw nsw i64 [[TMP7]], 41555; CHECK-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP8]]1556; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP2]]1557; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]]1558; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]1559; CHECK-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]1560; CHECK: vector.ph:1561; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 21562; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]1563; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]1564; CHECK: vector.body:1565; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]1566; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]1567; CHECK-NEXT: [[TMP9:%.*]] = add i64 [[INDEX]], 01568; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[INDEX]], 11569; CHECK-NEXT: [[TMP11:%.*]] = shl nsw <2 x i64> [[VEC_IND]], splat (i64 2)1570; CHECK-NEXT: [[TMP12:%.*]] = extractelement <2 x i64> [[TMP11]], i32 01571; CHECK-NEXT: [[TMP14:%.*]] = extractelement <2 x i64> [[TMP11]], i32 11572; CHECK-NEXT: [[TMP26:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP12]]1573; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP14]]1574; CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr [[TMP26]], align 1, !alias.scope [[META17:![0-9]+]]1575; CHECK-NEXT: [[TMP17:%.*]] = load i32, ptr [[TMP15]], align 1, !alias.scope [[META17]]1576; CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], ptr [[P]], i64 [[TMP9]], i32 11577; CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[PAIR_I32]], ptr [[P]], i64 [[TMP10]], i32 11578; CHECK-NEXT: store i32 [[TMP16]], ptr [[TMP18]], align 1, !alias.scope [[META20:![0-9]+]], !noalias [[META17]]1579; CHECK-NEXT: store i32 [[TMP17]], ptr [[TMP19]], align 1, !alias.scope [[META20]], !noalias [[META17]]1580; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 21581; CHECK-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <2 x i64> [[VEC_IND]], splat (i64 2)1582; CHECK-NEXT: [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]1583; CHECK-NEXT: br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]1584; CHECK: middle.block:1585; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]1586; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]1587; CHECK: scalar.ph:1588; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]1589; CHECK-NEXT: br label [[FOR_BODY:%.*]]1590; CHECK: for.body:1591; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]1592; CHECK-NEXT: [[TMP21:%.*]] = shl nsw i64 [[I]], 21593; CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP21]]1594; CHECK-NEXT: [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 11595; CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[PAIR_I32]], ptr [[P]], i64 [[I]], i32 11596; CHECK-NEXT: store i32 [[TMP23]], ptr [[TMP24]], align 11597; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 11598; CHECK-NEXT: [[TMP25:%.*]] = trunc i64 [[I_NEXT]] to i321599; CHECK-NEXT: [[COND:%.*]] = icmp eq i32 [[TMP25]], [[N]]1600; CHECK-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]1601; CHECK: for.end:1602; CHECK-NEXT: ret void1603;1604; IND-LABEL: @scalarize_induction_variable_04(1605; IND-NEXT: entry:1606; IND-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -11607; IND-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i641608; IND-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 11609; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP0]], 01610; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]1611; IND: vector.memcheck:1612; IND-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr [[P:%.*]], i64 41613; IND-NEXT: [[TMP3:%.*]] = add i32 [[N]], -11614; IND-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i641615; IND-NEXT: [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 31616; IND-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP5]]1617; IND-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[TMP6]], i64 81618; IND-NEXT: [[TMP7:%.*]] = shl nuw nsw i64 [[TMP4]], 41619; IND-NEXT: [[TMP8:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP7]]1620; IND-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[TMP8]], i64 41621; IND-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP2]]1622; IND-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]]1623; IND-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]1624; IND-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]1625; IND: vector.ph:1626; IND-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 85899345901627; IND-NEXT: br label [[VECTOR_BODY:%.*]]1628; IND: vector.body:1629; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]1630; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]1631; IND-NEXT: [[TMP10:%.*]] = shl nsw <2 x i64> [[VEC_IND]], splat (i64 2)1632; IND-NEXT: [[TMP11:%.*]] = extractelement <2 x i64> [[TMP10]], i64 01633; IND-NEXT: [[TMP13:%.*]] = extractelement <2 x i64> [[TMP10]], i64 11634; IND-NEXT: [[TMP12:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[TMP11]]1635; IND-NEXT: [[TMP24:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[TMP13]]1636; IND-NEXT: [[TMP14:%.*]] = load i32, ptr [[TMP12]], align 1, !alias.scope [[META17:![0-9]+]]1637; IND-NEXT: [[TMP15:%.*]] = load i32, ptr [[TMP24]], align 1, !alias.scope [[META17]]1638; IND-NEXT: [[DOTSPLIT:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], ptr [[P]], i64 [[INDEX]]1639; IND-NEXT: [[TMP16:%.*]] = getelementptr inbounds nuw i8, ptr [[DOTSPLIT]], i64 41640; IND-NEXT: [[TMP17:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1641; IND-NEXT: [[TMP18:%.*]] = getelementptr i8, ptr [[TMP17]], i64 121642; IND-NEXT: store i32 [[TMP14]], ptr [[TMP16]], align 1, !alias.scope [[META20:![0-9]+]], !noalias [[META17]]1643; IND-NEXT: store i32 [[TMP15]], ptr [[TMP18]], align 1, !alias.scope [[META20]], !noalias [[META17]]1644; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 21645; IND-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <2 x i64> [[VEC_IND]], splat (i64 2)1646; IND-NEXT: [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]1647; IND-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]1648; IND: middle.block:1649; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]1650; IND-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]1651; IND: scalar.ph:1652; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]1653; IND-NEXT: br label [[FOR_BODY:%.*]]1654; IND: for.body:1655; IND-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]1656; IND-NEXT: [[DOTIDX:%.*]] = shl nsw i64 [[I]], 41657; IND-NEXT: [[TMP20:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[DOTIDX]]1658; IND-NEXT: [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 11659; IND-NEXT: [[DOTSPLIT4:%.*]] = getelementptr inbounds nuw [[PAIR_I32]], ptr [[P]], i64 [[I]]1660; IND-NEXT: [[TMP22:%.*]] = getelementptr inbounds nuw i8, ptr [[DOTSPLIT4]], i64 41661; IND-NEXT: store i32 [[TMP21]], ptr [[TMP22]], align 11662; IND-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 11663; IND-NEXT: [[TMP23:%.*]] = trunc i64 [[I_NEXT]] to i321664; IND-NEXT: [[COND:%.*]] = icmp eq i32 [[N]], [[TMP23]]1665; IND-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]1666; IND: for.end:1667; IND-NEXT: ret void1668;1669; UNROLL-LABEL: @scalarize_induction_variable_04(1670; UNROLL-NEXT: entry:1671; UNROLL-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -11672; UNROLL-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i641673; UNROLL-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 11674; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 31675; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]1676; UNROLL: vector.memcheck:1677; UNROLL-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr [[P:%.*]], i64 41678; UNROLL-NEXT: [[TMP3:%.*]] = add i32 [[N]], -11679; UNROLL-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i641680; UNROLL-NEXT: [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 31681; UNROLL-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP5]]1682; UNROLL-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[TMP6]], i64 81683; UNROLL-NEXT: [[TMP7:%.*]] = shl nuw nsw i64 [[TMP4]], 41684; UNROLL-NEXT: [[TMP8:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP7]]1685; UNROLL-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[TMP8]], i64 41686; UNROLL-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP2]]1687; UNROLL-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]]1688; UNROLL-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]1689; UNROLL-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]1690; UNROLL: vector.ph:1691; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 85899345881692; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]1693; UNROLL: vector.body:1694; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]1695; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]1696; UNROLL-NEXT: [[TMP12:%.*]] = shl nsw <2 x i64> [[VEC_IND]], splat (i64 2)1697; UNROLL-NEXT: [[TMP16:%.*]] = extractelement <2 x i64> [[TMP12]], i64 01698; UNROLL-NEXT: [[TMP18:%.*]] = extractelement <2 x i64> [[TMP12]], i64 11699; UNROLL-NEXT: [[STEP_ADD:%.*]] = shl <2 x i64> [[VEC_IND]], splat (i64 2)1700; UNROLL-NEXT: [[TMP13:%.*]] = add <2 x i64> [[STEP_ADD]], splat (i64 8)1701; UNROLL-NEXT: [[TMP20:%.*]] = extractelement <2 x i64> [[TMP13]], i64 01702; UNROLL-NEXT: [[TMP35:%.*]] = extractelement <2 x i64> [[TMP13]], i64 11703; UNROLL-NEXT: [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP16]]1704; UNROLL-NEXT: [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP18]]1705; UNROLL-NEXT: [[TMP21:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP20]]1706; UNROLL-NEXT: [[TMP36:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP35]]1707; UNROLL-NEXT: [[TMP37:%.*]] = load i32, ptr [[TMP17]], align 1, !alias.scope [[META17:![0-9]+]]1708; UNROLL-NEXT: [[TMP38:%.*]] = load i32, ptr [[TMP19]], align 1, !alias.scope [[META17]]1709; UNROLL-NEXT: [[TMP39:%.*]] = load i32, ptr [[TMP21]], align 1, !alias.scope [[META17]]1710; UNROLL-NEXT: [[TMP22:%.*]] = load i32, ptr [[TMP36]], align 1, !alias.scope [[META17]]1711; UNROLL-NEXT: [[DOTSPLIT:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], ptr [[P]], i64 [[INDEX]]1712; UNROLL-NEXT: [[TMP23:%.*]] = getelementptr inbounds nuw i8, ptr [[DOTSPLIT]], i64 41713; UNROLL-NEXT: [[TMP24:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1714; UNROLL-NEXT: [[TMP25:%.*]] = getelementptr i8, ptr [[TMP24]], i64 121715; UNROLL-NEXT: [[TMP26:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1716; UNROLL-NEXT: [[TMP27:%.*]] = getelementptr i8, ptr [[TMP26]], i64 201717; UNROLL-NEXT: [[TMP28:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1718; UNROLL-NEXT: [[TMP29:%.*]] = getelementptr i8, ptr [[TMP28]], i64 281719; UNROLL-NEXT: store i32 [[TMP37]], ptr [[TMP23]], align 1, !alias.scope [[META20:![0-9]+]], !noalias [[META17]]1720; UNROLL-NEXT: store i32 [[TMP38]], ptr [[TMP25]], align 1, !alias.scope [[META20]], !noalias [[META17]]1721; UNROLL-NEXT: store i32 [[TMP39]], ptr [[TMP27]], align 1, !alias.scope [[META20]], !noalias [[META17]]1722; UNROLL-NEXT: store i32 [[TMP22]], ptr [[TMP29]], align 1, !alias.scope [[META20]], !noalias [[META17]]1723; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 41724; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 4)1725; UNROLL-NEXT: [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]1726; UNROLL-NEXT: br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]1727; UNROLL: middle.block:1728; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]1729; UNROLL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]1730; UNROLL: scalar.ph:1731; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]1732; UNROLL-NEXT: br label [[FOR_BODY:%.*]]1733; UNROLL: for.body:1734; UNROLL-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]1735; UNROLL-NEXT: [[DOTIDX:%.*]] = shl nsw i64 [[I]], 41736; UNROLL-NEXT: [[TMP31:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[DOTIDX]]1737; UNROLL-NEXT: [[TMP32:%.*]] = load i32, ptr [[TMP31]], align 11738; UNROLL-NEXT: [[DOTSPLIT6:%.*]] = getelementptr inbounds nuw [[PAIR_I32]], ptr [[P]], i64 [[I]]1739; UNROLL-NEXT: [[TMP33:%.*]] = getelementptr inbounds nuw i8, ptr [[DOTSPLIT6]], i64 41740; UNROLL-NEXT: store i32 [[TMP32]], ptr [[TMP33]], align 11741; UNROLL-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 11742; UNROLL-NEXT: [[TMP34:%.*]] = trunc i64 [[I_NEXT]] to i321743; UNROLL-NEXT: [[COND:%.*]] = icmp eq i32 [[N]], [[TMP34]]1744; UNROLL-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]1745; UNROLL: for.end:1746; UNROLL-NEXT: ret void1747;1748; UNROLL-NO-IC-LABEL: @scalarize_induction_variable_04(1749; UNROLL-NO-IC-NEXT: entry:1750; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -11751; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i641752; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 11753; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 41754; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]1755; UNROLL-NO-IC: vector.memcheck:1756; UNROLL-NO-IC-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr [[P:%.*]], i64 41757; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = add i32 [[N]], -11758; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i641759; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 31760; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = add nuw nsw i64 [[TMP5]], 81761; UNROLL-NO-IC-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP6]]1762; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = shl nuw nsw i64 [[TMP4]], 41763; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = add nuw nsw i64 [[TMP7]], 41764; UNROLL-NO-IC-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP8]]1765; UNROLL-NO-IC-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP2]]1766; UNROLL-NO-IC-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]]1767; UNROLL-NO-IC-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]1768; UNROLL-NO-IC-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]1769; UNROLL-NO-IC: vector.ph:1770; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 41771; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]1772; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]1773; UNROLL-NO-IC: vector.body:1774; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]1775; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]1776; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i64> [[VEC_IND]], splat (i64 2)1777; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = add i64 [[INDEX]], 01778; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = add i64 [[INDEX]], 11779; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = add i64 [[INDEX]], 21780; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = add i64 [[INDEX]], 31781; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = shl nsw <2 x i64> [[VEC_IND]], splat (i64 2)1782; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = extractelement <2 x i64> [[TMP13]], i32 01783; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = extractelement <2 x i64> [[TMP13]], i32 11784; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = shl nsw <2 x i64> [[STEP_ADD]], splat (i64 2)1785; UNROLL-NO-IC-NEXT: [[TMP21:%.*]] = extractelement <2 x i64> [[TMP14]], i32 01786; UNROLL-NO-IC-NEXT: [[TMP37:%.*]] = extractelement <2 x i64> [[TMP14]], i32 11787; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP17]]1788; UNROLL-NO-IC-NEXT: [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP19]]1789; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP21]]1790; UNROLL-NO-IC-NEXT: [[TMP38:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP37]]1791; UNROLL-NO-IC-NEXT: [[TMP23:%.*]] = load i32, ptr [[TMP18]], align 1, !alias.scope [[META17:![0-9]+]]1792; UNROLL-NO-IC-NEXT: [[TMP24:%.*]] = load i32, ptr [[TMP20]], align 1, !alias.scope [[META17]]1793; UNROLL-NO-IC-NEXT: [[TMP25:%.*]] = load i32, ptr [[TMP22]], align 1, !alias.scope [[META17]]1794; UNROLL-NO-IC-NEXT: [[TMP26:%.*]] = load i32, ptr [[TMP38]], align 1, !alias.scope [[META17]]1795; UNROLL-NO-IC-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], ptr [[P]], i64 [[TMP9]], i32 11796; UNROLL-NO-IC-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[PAIR_I32]], ptr [[P]], i64 [[TMP10]], i32 11797; UNROLL-NO-IC-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[PAIR_I32]], ptr [[P]], i64 [[TMP11]], i32 11798; UNROLL-NO-IC-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[PAIR_I32]], ptr [[P]], i64 [[TMP12]], i32 11799; UNROLL-NO-IC-NEXT: store i32 [[TMP23]], ptr [[TMP27]], align 1, !alias.scope [[META20:![0-9]+]], !noalias [[META17]]1800; UNROLL-NO-IC-NEXT: store i32 [[TMP24]], ptr [[TMP28]], align 1, !alias.scope [[META20]], !noalias [[META17]]1801; UNROLL-NO-IC-NEXT: store i32 [[TMP25]], ptr [[TMP29]], align 1, !alias.scope [[META20]], !noalias [[META17]]1802; UNROLL-NO-IC-NEXT: store i32 [[TMP26]], ptr [[TMP30]], align 1, !alias.scope [[META20]], !noalias [[META17]]1803; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 41804; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <2 x i64> [[STEP_ADD]], splat (i64 2)1805; UNROLL-NO-IC-NEXT: [[TMP31:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]1806; UNROLL-NO-IC-NEXT: br i1 [[TMP31]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]1807; UNROLL-NO-IC: middle.block:1808; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]1809; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]1810; UNROLL-NO-IC: scalar.ph:1811; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]1812; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]1813; UNROLL-NO-IC: for.body:1814; UNROLL-NO-IC-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]1815; UNROLL-NO-IC-NEXT: [[TMP32:%.*]] = shl nsw i64 [[I]], 21816; UNROLL-NO-IC-NEXT: [[TMP33:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP32]]1817; UNROLL-NO-IC-NEXT: [[TMP34:%.*]] = load i32, ptr [[TMP33]], align 11818; UNROLL-NO-IC-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[PAIR_I32]], ptr [[P]], i64 [[I]], i32 11819; UNROLL-NO-IC-NEXT: store i32 [[TMP34]], ptr [[TMP35]], align 11820; UNROLL-NO-IC-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 11821; UNROLL-NO-IC-NEXT: [[TMP36:%.*]] = trunc i64 [[I_NEXT]] to i321822; UNROLL-NO-IC-NEXT: [[COND:%.*]] = icmp eq i32 [[TMP36]], [[N]]1823; UNROLL-NO-IC-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]1824; UNROLL-NO-IC: for.end:1825; UNROLL-NO-IC-NEXT: ret void1826;1827; INTERLEAVE-LABEL: @scalarize_induction_variable_04(1828; INTERLEAVE-NEXT: entry:1829; INTERLEAVE-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -11830; INTERLEAVE-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i641831; INTERLEAVE-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 11832; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 81833; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]1834; INTERLEAVE: vector.memcheck:1835; INTERLEAVE-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr [[P:%.*]], i64 41836; INTERLEAVE-NEXT: [[TMP3:%.*]] = add i32 [[N]], -11837; INTERLEAVE-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i641838; INTERLEAVE-NEXT: [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 31839; INTERLEAVE-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP5]]1840; INTERLEAVE-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[TMP6]], i64 81841; INTERLEAVE-NEXT: [[TMP7:%.*]] = shl nuw nsw i64 [[TMP4]], 41842; INTERLEAVE-NEXT: [[TMP8:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP7]]1843; INTERLEAVE-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[TMP8]], i64 41844; INTERLEAVE-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP2]]1845; INTERLEAVE-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]]1846; INTERLEAVE-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]1847; INTERLEAVE-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]1848; INTERLEAVE: vector.ph:1849; INTERLEAVE-NEXT: [[N_MOD_VF:%.*]] = and i64 [[TMP2]], 71850; INTERLEAVE-NEXT: [[TMP9:%.*]] = icmp eq i64 [[N_MOD_VF]], 01851; INTERLEAVE-NEXT: [[TMP10:%.*]] = select i1 [[TMP9]], i64 8, i64 [[N_MOD_VF]]1852; INTERLEAVE-NEXT: [[N_VEC:%.*]] = sub nsw i64 [[TMP2]], [[TMP10]]1853; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]1854; INTERLEAVE: vector.body:1855; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]1856; INTERLEAVE-NEXT: [[TMP14:%.*]] = or disjoint i64 [[INDEX]], 41857; INTERLEAVE-NEXT: [[DOTIDX:%.*]] = shl nsw i64 [[INDEX]], 41858; INTERLEAVE-NEXT: [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[DOTIDX]]1859; INTERLEAVE-NEXT: [[DOTIDX5:%.*]] = shl nsw i64 [[TMP14]], 41860; INTERLEAVE-NEXT: [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[DOTIDX5]]1861; INTERLEAVE-NEXT: [[WIDE_VEC:%.*]] = load <16 x i32>, ptr [[TMP18]], align 1, !alias.scope [[META17:![0-9]+]]1862; INTERLEAVE-NEXT: [[TMP28:%.*]] = extractelement <16 x i32> [[WIDE_VEC]], i64 01863; INTERLEAVE-NEXT: [[TMP29:%.*]] = extractelement <16 x i32> [[WIDE_VEC]], i64 41864; INTERLEAVE-NEXT: [[TMP30:%.*]] = extractelement <16 x i32> [[WIDE_VEC]], i64 81865; INTERLEAVE-NEXT: [[TMP31:%.*]] = extractelement <16 x i32> [[WIDE_VEC]], i64 121866; INTERLEAVE-NEXT: [[WIDE_VEC3:%.*]] = load <16 x i32>, ptr [[TMP19]], align 1, !alias.scope [[META17]]1867; INTERLEAVE-NEXT: [[TMP32:%.*]] = extractelement <16 x i32> [[WIDE_VEC3]], i64 01868; INTERLEAVE-NEXT: [[TMP33:%.*]] = extractelement <16 x i32> [[WIDE_VEC3]], i64 41869; INTERLEAVE-NEXT: [[TMP34:%.*]] = extractelement <16 x i32> [[WIDE_VEC3]], i64 81870; INTERLEAVE-NEXT: [[TMP35:%.*]] = extractelement <16 x i32> [[WIDE_VEC3]], i64 121871; INTERLEAVE-NEXT: [[DOTSPLIT:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], ptr [[P]], i64 [[INDEX]]1872; INTERLEAVE-NEXT: [[TMP41:%.*]] = getelementptr inbounds nuw i8, ptr [[DOTSPLIT]], i64 41873; INTERLEAVE-NEXT: [[TMP15:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1874; INTERLEAVE-NEXT: [[TMP16:%.*]] = getelementptr i8, ptr [[TMP15]], i64 121875; INTERLEAVE-NEXT: [[TMP17:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1876; INTERLEAVE-NEXT: [[TMP42:%.*]] = getelementptr i8, ptr [[TMP17]], i64 201877; INTERLEAVE-NEXT: [[TMP43:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1878; INTERLEAVE-NEXT: [[TMP20:%.*]] = getelementptr i8, ptr [[TMP43]], i64 281879; INTERLEAVE-NEXT: [[DOTSPLIT9:%.*]] = getelementptr inbounds [[PAIR_I32]], ptr [[P]], i64 [[TMP14]]1880; INTERLEAVE-NEXT: [[TMP21:%.*]] = getelementptr inbounds nuw i8, ptr [[DOTSPLIT9]], i64 41881; INTERLEAVE-NEXT: [[TMP22:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1882; INTERLEAVE-NEXT: [[TMP23:%.*]] = getelementptr i8, ptr [[TMP22]], i64 441883; INTERLEAVE-NEXT: [[TMP24:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1884; INTERLEAVE-NEXT: [[TMP25:%.*]] = getelementptr i8, ptr [[TMP24]], i64 521885; INTERLEAVE-NEXT: [[TMP26:%.*]] = getelementptr [[PAIR_I32]], ptr [[P]], i64 [[INDEX]]1886; INTERLEAVE-NEXT: [[TMP27:%.*]] = getelementptr i8, ptr [[TMP26]], i64 601887; INTERLEAVE-NEXT: store i32 [[TMP28]], ptr [[TMP41]], align 1, !alias.scope [[META20:![0-9]+]], !noalias [[META17]]1888; INTERLEAVE-NEXT: store i32 [[TMP29]], ptr [[TMP16]], align 1, !alias.scope [[META20]], !noalias [[META17]]1889; INTERLEAVE-NEXT: store i32 [[TMP30]], ptr [[TMP42]], align 1, !alias.scope [[META20]], !noalias [[META17]]1890; INTERLEAVE-NEXT: store i32 [[TMP31]], ptr [[TMP20]], align 1, !alias.scope [[META20]], !noalias [[META17]]1891; INTERLEAVE-NEXT: store i32 [[TMP32]], ptr [[TMP21]], align 1, !alias.scope [[META20]], !noalias [[META17]]1892; INTERLEAVE-NEXT: store i32 [[TMP33]], ptr [[TMP23]], align 1, !alias.scope [[META20]], !noalias [[META17]]1893; INTERLEAVE-NEXT: store i32 [[TMP34]], ptr [[TMP25]], align 1, !alias.scope [[META20]], !noalias [[META17]]1894; INTERLEAVE-NEXT: store i32 [[TMP35]], ptr [[TMP27]], align 1, !alias.scope [[META20]], !noalias [[META17]]1895; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 81896; INTERLEAVE-NEXT: [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]1897; INTERLEAVE-NEXT: br i1 [[TMP36]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]1898; INTERLEAVE: middle.block:1899; INTERLEAVE-NEXT: br label [[SCALAR_PH]]1900; INTERLEAVE: scalar.ph:1901; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]1902; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]1903; INTERLEAVE: for.body:1904; INTERLEAVE-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]1905; INTERLEAVE-NEXT: [[DOTIDX6:%.*]] = shl nsw i64 [[I]], 41906; INTERLEAVE-NEXT: [[TMP37:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[DOTIDX6]]1907; INTERLEAVE-NEXT: [[TMP38:%.*]] = load i32, ptr [[TMP37]], align 11908; INTERLEAVE-NEXT: [[DOTSPLIT14:%.*]] = getelementptr inbounds [[PAIR_I32]], ptr [[P]], i64 [[I]]1909; INTERLEAVE-NEXT: [[TMP39:%.*]] = getelementptr inbounds nuw i8, ptr [[DOTSPLIT14]], i64 41910; INTERLEAVE-NEXT: store i32 [[TMP38]], ptr [[TMP39]], align 11911; INTERLEAVE-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 11912; INTERLEAVE-NEXT: [[TMP40:%.*]] = trunc i64 [[I_NEXT]] to i321913; INTERLEAVE-NEXT: [[COND:%.*]] = icmp eq i32 [[N]], [[TMP40]]1914; INTERLEAVE-NEXT: br i1 [[COND]], label [[FOR_END:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]1915; INTERLEAVE: for.end:1916; INTERLEAVE-NEXT: ret void1917;1918entry:1919 br label %for.body1920 1921for.body:1922 %i = phi i64 [ %i.next, %for.body ], [ 0, %entry]1923 %0 = shl nsw i64 %i, 21924 %1 = getelementptr inbounds i32, ptr %a, i64 %01925 %2 = load i32, ptr %1, align 11926 %3 = getelementptr inbounds %pair.i32, ptr %p, i64 %i, i32 11927 store i32 %2, ptr %3, align 11928 %i.next = add nuw nsw i64 %i, 11929 %4 = trunc i64 %i.next to i321930 %cond = icmp eq i32 %4, %n1931 br i1 %cond, label %for.end, label %for.body1932 1933for.end:1934 ret void1935}1936 1937; PR30542. Ensure we generate all the scalar steps for the induction variable.1938; The scalar induction variable is used by a getelementptr instruction1939; (uniform), and a udiv (non-uniform).1940;1941; int sum = 0;1942; for (int i = 0; i < n; ++i) {1943; int x = a[i];1944; if (c)1945; x /= i;1946; sum += x;1947; }1948;1949;1950;1951;1952 1953define i32 @scalarize_induction_variable_05(ptr %a, i32 %x, i1 %c, i32 %n) {1954; CHECK-LABEL: @scalarize_induction_variable_05(1955; CHECK-NEXT: entry:1956; CHECK-NEXT: [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[N:%.*]], i32 1)1957; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[SMAX]], 21958; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]1959; CHECK: vector.ph:1960; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[SMAX]], 21961; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[SMAX]], [[N_MOD_VF]]1962; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]1963; CHECK: vector.body:1964; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE2:%.*]] ]1965; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[PRED_UDIV_CONTINUE2]] ]1966; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]1967; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 41968; CHECK-NEXT: br i1 [[C:%.*]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]1969; CHECK: pred.udiv.if:1970; CHECK-NEXT: [[TMP4:%.*]] = add i32 [[INDEX]], 01971; CHECK-NEXT: [[TMP5:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 01972; CHECK-NEXT: [[TMP6:%.*]] = udiv i32 [[TMP5]], [[TMP4]]1973; CHECK-NEXT: [[TMP7:%.*]] = insertelement <2 x i32> poison, i32 [[TMP6]], i32 01974; CHECK-NEXT: br label [[PRED_UDIV_CONTINUE]]1975; CHECK: pred.udiv.continue:1976; CHECK-NEXT: [[TMP12:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP7]], [[PRED_UDIV_IF]] ]1977; CHECK-NEXT: br i1 [[C]], label [[PRED_UDIV_IF1:%.*]], label [[PRED_UDIV_CONTINUE2]]1978; CHECK: pred.udiv.if1:1979; CHECK-NEXT: [[TMP14:%.*]] = add i32 [[INDEX]], 11980; CHECK-NEXT: [[TMP8:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 11981; CHECK-NEXT: [[TMP9:%.*]] = udiv i32 [[TMP8]], [[TMP14]]1982; CHECK-NEXT: [[TMP10:%.*]] = insertelement <2 x i32> [[TMP12]], i32 [[TMP9]], i32 11983; CHECK-NEXT: br label [[PRED_UDIV_CONTINUE2]]1984; CHECK: pred.udiv.continue2:1985; CHECK-NEXT: [[TMP11:%.*]] = phi <2 x i32> [ [[TMP12]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP10]], [[PRED_UDIV_IF1]] ]1986; CHECK-NEXT: [[PREDPHI:%.*]] = select i1 [[C]], <2 x i32> [[TMP11]], <2 x i32> [[WIDE_LOAD]]1987; CHECK-NEXT: [[TMP16]] = add <2 x i32> [[PREDPHI]], [[VEC_PHI]]1988; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 21989; CHECK-NEXT: [[TMP13:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]1990; CHECK-NEXT: br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]1991; CHECK: middle.block:1992; CHECK-NEXT: [[TMP18:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[TMP16]])1993; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[SMAX]], [[N_VEC]]1994; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]1995; CHECK: scalar.ph:1996; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]1997; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP18]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]1998; CHECK-NEXT: br label [[FOR_BODY:%.*]]1999; CHECK: for.body:2000; CHECK-NEXT: [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[IF_END:%.*]] ]2001; CHECK-NEXT: [[SUM:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR4:%.*]], [[IF_END]] ]2002; CHECK-NEXT: [[VAR0:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[I]]2003; CHECK-NEXT: [[VAR1:%.*]] = load i32, ptr [[VAR0]], align 42004; CHECK-NEXT: br i1 [[C]], label [[IF_THEN:%.*]], label [[IF_END]]2005; CHECK: if.then:2006; CHECK-NEXT: [[VAR2:%.*]] = udiv i32 [[VAR1]], [[I]]2007; CHECK-NEXT: br label [[IF_END]]2008; CHECK: if.end:2009; CHECK-NEXT: [[VAR3:%.*]] = phi i32 [ [[VAR2]], [[IF_THEN]] ], [ [[VAR1]], [[FOR_BODY]] ]2010; CHECK-NEXT: [[VAR4]] = add i32 [[VAR3]], [[SUM]]2011; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i32 [[I]], 12012; CHECK-NEXT: [[COND:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]2013; CHECK-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP25:![0-9]+]]2014; CHECK: for.end:2015; CHECK-NEXT: [[VAR5:%.*]] = phi i32 [ [[VAR4]], [[IF_END]] ], [ [[TMP18]], [[MIDDLE_BLOCK]] ]2016; CHECK-NEXT: ret i32 [[VAR5]]2017;2018; IND-LABEL: @scalarize_induction_variable_05(2019; IND-NEXT: entry:2020; IND-NEXT: [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[N:%.*]], i32 1)2021; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i32 [[N]], 22022; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]2023; IND: vector.ph:2024; IND-NEXT: [[N_VEC:%.*]] = and i32 [[SMAX]], 21474836462025; IND-NEXT: br label [[VECTOR_BODY:%.*]]2026; IND: vector.body:2027; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE2:%.*]] ]2028; IND-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[PRED_UDIV_CONTINUE2]] ]2029; IND-NEXT: [[TMP0:%.*]] = sext i32 [[INDEX]] to i642030; IND-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP0]]2031; IND-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 42032; IND-NEXT: br i1 [[C:%.*]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]2033; IND: pred.udiv.if:2034; IND-NEXT: [[TMP2:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i64 02035; IND-NEXT: [[TMP3:%.*]] = udiv i32 [[TMP2]], [[INDEX]]2036; IND-NEXT: [[TMP4:%.*]] = insertelement <2 x i32> poison, i32 [[TMP3]], i64 02037; IND-NEXT: br label [[PRED_UDIV_CONTINUE]]2038; IND: pred.udiv.continue:2039; IND-NEXT: [[TMP5:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP4]], [[PRED_UDIV_IF]] ]2040; IND-NEXT: br i1 [[C]], label [[PRED_UDIV_IF1:%.*]], label [[PRED_UDIV_CONTINUE2]]2041; IND: pred.udiv.if1:2042; IND-NEXT: [[TMP6:%.*]] = or disjoint i32 [[INDEX]], 12043; IND-NEXT: [[TMP7:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i64 12044; IND-NEXT: [[TMP8:%.*]] = udiv i32 [[TMP7]], [[TMP6]]2045; IND-NEXT: [[TMP9:%.*]] = insertelement <2 x i32> [[TMP5]], i32 [[TMP8]], i64 12046; IND-NEXT: br label [[PRED_UDIV_CONTINUE2]]2047; IND: pred.udiv.continue2:2048; IND-NEXT: [[PREDPHI:%.*]] = phi <2 x i32> [ [[WIDE_LOAD]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP9]], [[PRED_UDIV_IF1]] ]2049; IND-NEXT: [[TMP13]] = add <2 x i32> [[PREDPHI]], [[VEC_PHI]]2050; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 22051; IND-NEXT: [[TMP14:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]2052; IND-NEXT: br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]2053; IND: middle.block:2054; IND-NEXT: [[TMP15:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[TMP13]])2055; IND-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[SMAX]], [[N_VEC]]2056; IND-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]2057; IND: scalar.ph:2058; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]2059; IND-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP15]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]2060; IND-NEXT: br label [[FOR_BODY:%.*]]2061; IND: for.body:2062; IND-NEXT: [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[IF_END:%.*]] ]2063; IND-NEXT: [[SUM:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR4:%.*]], [[IF_END]] ]2064; IND-NEXT: [[TMP16:%.*]] = zext nneg i32 [[I]] to i642065; IND-NEXT: [[VAR0:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[TMP16]]2066; IND-NEXT: [[VAR1:%.*]] = load i32, ptr [[VAR0]], align 42067; IND-NEXT: br i1 [[C]], label [[IF_THEN:%.*]], label [[IF_END]]2068; IND: if.then:2069; IND-NEXT: [[VAR2:%.*]] = udiv i32 [[VAR1]], [[I]]2070; IND-NEXT: br label [[IF_END]]2071; IND: if.end:2072; IND-NEXT: [[VAR3:%.*]] = phi i32 [ [[VAR2]], [[IF_THEN]] ], [ [[VAR1]], [[FOR_BODY]] ]2073; IND-NEXT: [[VAR4]] = add i32 [[VAR3]], [[SUM]]2074; IND-NEXT: [[I_NEXT]] = add nuw nsw i32 [[I]], 12075; IND-NEXT: [[COND:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]2076; IND-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP25:![0-9]+]]2077; IND: for.end:2078; IND-NEXT: [[VAR5:%.*]] = phi i32 [ [[VAR4]], [[IF_END]] ], [ [[TMP15]], [[MIDDLE_BLOCK]] ]2079; IND-NEXT: ret i32 [[VAR5]]2080;2081; UNROLL-LABEL: @scalarize_induction_variable_05(2082; UNROLL-NEXT: entry:2083; UNROLL-NEXT: [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[N:%.*]], i32 1)2084; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i32 [[N]], 42085; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]2086; UNROLL: vector.ph:2087; UNROLL-NEXT: [[N_VEC:%.*]] = and i32 [[SMAX]], 21474836442088; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]2089; UNROLL: vector.body:2090; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE8:%.*]] ]2091; UNROLL-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP22:%.*]], [[PRED_UDIV_CONTINUE8]] ]2092; UNROLL-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP23:%.*]], [[PRED_UDIV_CONTINUE8]] ]2093; UNROLL-NEXT: [[TMP0:%.*]] = sext i32 [[INDEX]] to i642094; UNROLL-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP0]]2095; UNROLL-NEXT: [[TMP2:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i64 82096; UNROLL-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 42097; UNROLL-NEXT: [[WIDE_LOAD2:%.*]] = load <2 x i32>, ptr [[TMP2]], align 42098; UNROLL-NEXT: br i1 [[C:%.*]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]2099; UNROLL: pred.udiv.if:2100; UNROLL-NEXT: [[TMP3:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i64 02101; UNROLL-NEXT: [[TMP4:%.*]] = udiv i32 [[TMP3]], [[INDEX]]2102; UNROLL-NEXT: [[TMP5:%.*]] = insertelement <2 x i32> poison, i32 [[TMP4]], i64 02103; UNROLL-NEXT: br label [[PRED_UDIV_CONTINUE]]2104; UNROLL: pred.udiv.continue:2105; UNROLL-NEXT: [[TMP6:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP5]], [[PRED_UDIV_IF]] ]2106; UNROLL-NEXT: br i1 [[C]], label [[PRED_UDIV_IF3:%.*]], label [[PRED_UDIV_CONTINUE4:%.*]]2107; UNROLL: pred.udiv.if3:2108; UNROLL-NEXT: [[TMP7:%.*]] = or disjoint i32 [[INDEX]], 12109; UNROLL-NEXT: [[TMP8:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i64 12110; UNROLL-NEXT: [[TMP9:%.*]] = udiv i32 [[TMP8]], [[TMP7]]2111; UNROLL-NEXT: [[TMP10:%.*]] = insertelement <2 x i32> [[TMP6]], i32 [[TMP9]], i64 12112; UNROLL-NEXT: br label [[PRED_UDIV_CONTINUE4]]2113; UNROLL: pred.udiv.continue4:2114; UNROLL-NEXT: [[TMP11:%.*]] = phi <2 x i32> [ [[TMP6]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP10]], [[PRED_UDIV_IF3]] ]2115; UNROLL-NEXT: br i1 [[C]], label [[PRED_UDIV_IF5:%.*]], label [[PRED_UDIV_CONTINUE6:%.*]]2116; UNROLL: pred.udiv.if5:2117; UNROLL-NEXT: [[TMP12:%.*]] = or disjoint i32 [[INDEX]], 22118; UNROLL-NEXT: [[TMP13:%.*]] = extractelement <2 x i32> [[WIDE_LOAD2]], i64 02119; UNROLL-NEXT: [[TMP14:%.*]] = udiv i32 [[TMP13]], [[TMP12]]2120; UNROLL-NEXT: [[TMP15:%.*]] = insertelement <2 x i32> poison, i32 [[TMP14]], i64 02121; UNROLL-NEXT: br label [[PRED_UDIV_CONTINUE6]]2122; UNROLL: pred.udiv.continue6:2123; UNROLL-NEXT: [[TMP16:%.*]] = phi <2 x i32> [ poison, [[PRED_UDIV_CONTINUE4]] ], [ [[TMP15]], [[PRED_UDIV_IF5]] ]2124; UNROLL-NEXT: br i1 [[C]], label [[PRED_UDIV_IF7:%.*]], label [[PRED_UDIV_CONTINUE8]]2125; UNROLL: pred.udiv.if7:2126; UNROLL-NEXT: [[TMP17:%.*]] = or disjoint i32 [[INDEX]], 32127; UNROLL-NEXT: [[TMP18:%.*]] = extractelement <2 x i32> [[WIDE_LOAD2]], i64 12128; UNROLL-NEXT: [[TMP19:%.*]] = udiv i32 [[TMP18]], [[TMP17]]2129; UNROLL-NEXT: [[TMP20:%.*]] = insertelement <2 x i32> [[TMP16]], i32 [[TMP19]], i64 12130; UNROLL-NEXT: br label [[PRED_UDIV_CONTINUE8]]2131; UNROLL: pred.udiv.continue8:2132; UNROLL-NEXT: [[PREDPHI9:%.*]] = phi <2 x i32> [ [[WIDE_LOAD2]], [[PRED_UDIV_CONTINUE6]] ], [ [[TMP20]], [[PRED_UDIV_IF7]] ]2133; UNROLL-NEXT: [[PREDPHI:%.*]] = phi <2 x i32> [ [[WIDE_LOAD]], [[PRED_UDIV_CONTINUE6]] ], [ [[TMP11]], [[PRED_UDIV_IF7]] ]2134; UNROLL-NEXT: [[TMP22]] = add <2 x i32> [[PREDPHI]], [[VEC_PHI]]2135; UNROLL-NEXT: [[TMP23]] = add <2 x i32> [[PREDPHI9]], [[VEC_PHI1]]2136; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 42137; UNROLL-NEXT: [[TMP24:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]2138; UNROLL-NEXT: br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]2139; UNROLL: middle.block:2140; UNROLL-NEXT: [[BIN_RDX:%.*]] = add <2 x i32> [[TMP23]], [[TMP22]]2141; UNROLL-NEXT: [[TMP25:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[BIN_RDX]])2142; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[SMAX]], [[N_VEC]]2143; UNROLL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]2144; UNROLL: scalar.ph:2145; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]2146; UNROLL-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP25]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]2147; UNROLL-NEXT: br label [[FOR_BODY:%.*]]2148; UNROLL: for.body:2149; UNROLL-NEXT: [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[IF_END:%.*]] ]2150; UNROLL-NEXT: [[SUM:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR4:%.*]], [[IF_END]] ]2151; UNROLL-NEXT: [[TMP26:%.*]] = zext nneg i32 [[I]] to i642152; UNROLL-NEXT: [[VAR0:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[TMP26]]2153; UNROLL-NEXT: [[VAR1:%.*]] = load i32, ptr [[VAR0]], align 42154; UNROLL-NEXT: br i1 [[C]], label [[IF_THEN:%.*]], label [[IF_END]]2155; UNROLL: if.then:2156; UNROLL-NEXT: [[VAR2:%.*]] = udiv i32 [[VAR1]], [[I]]2157; UNROLL-NEXT: br label [[IF_END]]2158; UNROLL: if.end:2159; UNROLL-NEXT: [[VAR3:%.*]] = phi i32 [ [[VAR2]], [[IF_THEN]] ], [ [[VAR1]], [[FOR_BODY]] ]2160; UNROLL-NEXT: [[VAR4]] = add i32 [[VAR3]], [[SUM]]2161; UNROLL-NEXT: [[I_NEXT]] = add nuw nsw i32 [[I]], 12162; UNROLL-NEXT: [[COND:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]2163; UNROLL-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP25:![0-9]+]]2164; UNROLL: for.end:2165; UNROLL-NEXT: [[VAR5:%.*]] = phi i32 [ [[VAR4]], [[IF_END]] ], [ [[TMP25]], [[MIDDLE_BLOCK]] ]2166; UNROLL-NEXT: ret i32 [[VAR5]]2167;2168; UNROLL-NO-IC-LABEL: @scalarize_induction_variable_05(2169; UNROLL-NO-IC-NEXT: entry:2170; UNROLL-NO-IC-NEXT: [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[N:%.*]], i32 1)2171; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[SMAX]], 42172; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]2173; UNROLL-NO-IC: vector.ph:2174; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[SMAX]], 42175; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i32 [[SMAX]], [[N_MOD_VF]]2176; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]2177; UNROLL-NO-IC: vector.body:2178; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE8:%.*]] ]2179; UNROLL-NO-IC-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP32:%.*]], [[PRED_UDIV_CONTINUE8]] ]2180; UNROLL-NO-IC-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP33:%.*]], [[PRED_UDIV_CONTINUE8]] ]2181; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]2182; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i64 22183; UNROLL-NO-IC-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP0]], align 42184; UNROLL-NO-IC-NEXT: [[WIDE_LOAD2:%.*]] = load <2 x i32>, ptr [[TMP2]], align 42185; UNROLL-NO-IC-NEXT: br i1 [[C:%.*]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]2186; UNROLL-NO-IC: pred.udiv.if:2187; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = add i32 [[INDEX]], 02188; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 02189; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = udiv i32 [[TMP8]], [[TMP7]]2190; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = insertelement <2 x i32> poison, i32 [[TMP9]], i32 02191; UNROLL-NO-IC-NEXT: br label [[PRED_UDIV_CONTINUE]]2192; UNROLL-NO-IC: pred.udiv.continue:2193; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP10]], [[PRED_UDIV_IF]] ]2194; UNROLL-NO-IC-NEXT: br i1 [[C]], label [[PRED_UDIV_IF3:%.*]], label [[PRED_UDIV_CONTINUE4:%.*]]2195; UNROLL-NO-IC: pred.udiv.if3:2196; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = add i32 [[INDEX]], 12197; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 12198; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = udiv i32 [[TMP14]], [[TMP13]]2199; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = insertelement <2 x i32> [[TMP11]], i32 [[TMP15]], i32 12200; UNROLL-NO-IC-NEXT: br label [[PRED_UDIV_CONTINUE4]]2201; UNROLL-NO-IC: pred.udiv.continue4:2202; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = phi <2 x i32> [ [[TMP11]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP16]], [[PRED_UDIV_IF3]] ]2203; UNROLL-NO-IC-NEXT: br i1 [[C]], label [[PRED_UDIV_IF5:%.*]], label [[PRED_UDIV_CONTINUE6:%.*]]2204; UNROLL-NO-IC: pred.udiv.if5:2205; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = add i32 [[INDEX]], 22206; UNROLL-NO-IC-NEXT: [[TMP20:%.*]] = extractelement <2 x i32> [[WIDE_LOAD2]], i32 02207; UNROLL-NO-IC-NEXT: [[TMP21:%.*]] = udiv i32 [[TMP20]], [[TMP19]]2208; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = insertelement <2 x i32> poison, i32 [[TMP21]], i32 02209; UNROLL-NO-IC-NEXT: br label [[PRED_UDIV_CONTINUE6]]2210; UNROLL-NO-IC: pred.udiv.continue6:2211; UNROLL-NO-IC-NEXT: [[TMP23:%.*]] = phi <2 x i32> [ poison, [[PRED_UDIV_CONTINUE4]] ], [ [[TMP22]], [[PRED_UDIV_IF5]] ]2212; UNROLL-NO-IC-NEXT: br i1 [[C]], label [[PRED_UDIV_IF7:%.*]], label [[PRED_UDIV_CONTINUE8]]2213; UNROLL-NO-IC: pred.udiv.if7:2214; UNROLL-NO-IC-NEXT: [[TMP25:%.*]] = add i32 [[INDEX]], 32215; UNROLL-NO-IC-NEXT: [[TMP26:%.*]] = extractelement <2 x i32> [[WIDE_LOAD2]], i32 12216; UNROLL-NO-IC-NEXT: [[TMP27:%.*]] = udiv i32 [[TMP26]], [[TMP25]]2217; UNROLL-NO-IC-NEXT: [[TMP28:%.*]] = insertelement <2 x i32> [[TMP23]], i32 [[TMP27]], i32 12218; UNROLL-NO-IC-NEXT: br label [[PRED_UDIV_CONTINUE8]]2219; UNROLL-NO-IC: pred.udiv.continue8:2220; UNROLL-NO-IC-NEXT: [[TMP29:%.*]] = phi <2 x i32> [ [[TMP23]], [[PRED_UDIV_CONTINUE6]] ], [ [[TMP28]], [[PRED_UDIV_IF7]] ]2221; UNROLL-NO-IC-NEXT: [[PREDPHI:%.*]] = select i1 [[C]], <2 x i32> [[TMP17]], <2 x i32> [[WIDE_LOAD]]2222; UNROLL-NO-IC-NEXT: [[PREDPHI9:%.*]] = select i1 [[C]], <2 x i32> [[TMP29]], <2 x i32> [[WIDE_LOAD2]]2223; UNROLL-NO-IC-NEXT: [[TMP32]] = add <2 x i32> [[PREDPHI]], [[VEC_PHI]]2224; UNROLL-NO-IC-NEXT: [[TMP33]] = add <2 x i32> [[PREDPHI9]], [[VEC_PHI1]]2225; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 42226; UNROLL-NO-IC-NEXT: [[TMP34:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]2227; UNROLL-NO-IC-NEXT: br i1 [[TMP34]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]2228; UNROLL-NO-IC: middle.block:2229; UNROLL-NO-IC-NEXT: [[BIN_RDX:%.*]] = add <2 x i32> [[TMP33]], [[TMP32]]2230; UNROLL-NO-IC-NEXT: [[TMP30:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[BIN_RDX]])2231; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[SMAX]], [[N_VEC]]2232; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]2233; UNROLL-NO-IC: scalar.ph:2234; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]2235; UNROLL-NO-IC-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP30]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]2236; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]2237; UNROLL-NO-IC: for.body:2238; UNROLL-NO-IC-NEXT: [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[IF_END:%.*]] ]2239; UNROLL-NO-IC-NEXT: [[SUM:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR4:%.*]], [[IF_END]] ]2240; UNROLL-NO-IC-NEXT: [[VAR0:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[I]]2241; UNROLL-NO-IC-NEXT: [[VAR1:%.*]] = load i32, ptr [[VAR0]], align 42242; UNROLL-NO-IC-NEXT: br i1 [[C]], label [[IF_THEN:%.*]], label [[IF_END]]2243; UNROLL-NO-IC: if.then:2244; UNROLL-NO-IC-NEXT: [[VAR2:%.*]] = udiv i32 [[VAR1]], [[I]]2245; UNROLL-NO-IC-NEXT: br label [[IF_END]]2246; UNROLL-NO-IC: if.end:2247; UNROLL-NO-IC-NEXT: [[VAR3:%.*]] = phi i32 [ [[VAR2]], [[IF_THEN]] ], [ [[VAR1]], [[FOR_BODY]] ]2248; UNROLL-NO-IC-NEXT: [[VAR4]] = add i32 [[VAR3]], [[SUM]]2249; UNROLL-NO-IC-NEXT: [[I_NEXT]] = add nuw nsw i32 [[I]], 12250; UNROLL-NO-IC-NEXT: [[COND:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]2251; UNROLL-NO-IC-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP25:![0-9]+]]2252; UNROLL-NO-IC: for.end:2253; UNROLL-NO-IC-NEXT: [[VAR5:%.*]] = phi i32 [ [[VAR4]], [[IF_END]] ], [ [[TMP30]], [[MIDDLE_BLOCK]] ]2254; UNROLL-NO-IC-NEXT: ret i32 [[VAR5]]2255;2256; INTERLEAVE-LABEL: @scalarize_induction_variable_05(2257; INTERLEAVE-NEXT: entry:2258; INTERLEAVE-NEXT: [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[N:%.*]], i32 1)2259; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i32 [[N]], 82260; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]2261; INTERLEAVE: vector.ph:2262; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i32 [[SMAX]], 21474836402263; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]2264; INTERLEAVE: vector.body:2265; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE16:%.*]] ]2266; INTERLEAVE-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP42:%.*]], [[PRED_UDIV_CONTINUE16]] ]2267; INTERLEAVE-NEXT: [[VEC_PHI1:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP43:%.*]], [[PRED_UDIV_CONTINUE16]] ]2268; INTERLEAVE-NEXT: [[TMP0:%.*]] = sext i32 [[INDEX]] to i642269; INTERLEAVE-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP0]]2270; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i64 162271; INTERLEAVE-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 42272; INTERLEAVE-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP2]], align 42273; INTERLEAVE-NEXT: br i1 [[C:%.*]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]2274; INTERLEAVE: pred.udiv.if:2275; INTERLEAVE-NEXT: [[TMP3:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i64 02276; INTERLEAVE-NEXT: [[TMP4:%.*]] = udiv i32 [[TMP3]], [[INDEX]]2277; INTERLEAVE-NEXT: [[TMP5:%.*]] = insertelement <4 x i32> poison, i32 [[TMP4]], i64 02278; INTERLEAVE-NEXT: br label [[PRED_UDIV_CONTINUE]]2279; INTERLEAVE: pred.udiv.continue:2280; INTERLEAVE-NEXT: [[TMP6:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP5]], [[PRED_UDIV_IF]] ]2281; INTERLEAVE-NEXT: br i1 [[C]], label [[PRED_UDIV_IF3:%.*]], label [[PRED_UDIV_CONTINUE4:%.*]]2282; INTERLEAVE: pred.udiv.if3:2283; INTERLEAVE-NEXT: [[TMP7:%.*]] = or disjoint i32 [[INDEX]], 12284; INTERLEAVE-NEXT: [[TMP8:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i64 12285; INTERLEAVE-NEXT: [[TMP9:%.*]] = udiv i32 [[TMP8]], [[TMP7]]2286; INTERLEAVE-NEXT: [[TMP10:%.*]] = insertelement <4 x i32> [[TMP6]], i32 [[TMP9]], i64 12287; INTERLEAVE-NEXT: br label [[PRED_UDIV_CONTINUE4]]2288; INTERLEAVE: pred.udiv.continue4:2289; INTERLEAVE-NEXT: [[TMP11:%.*]] = phi <4 x i32> [ [[TMP6]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP10]], [[PRED_UDIV_IF3]] ]2290; INTERLEAVE-NEXT: br i1 [[C]], label [[PRED_UDIV_IF5:%.*]], label [[PRED_UDIV_CONTINUE6:%.*]]2291; INTERLEAVE: pred.udiv.if5:2292; INTERLEAVE-NEXT: [[TMP12:%.*]] = or disjoint i32 [[INDEX]], 22293; INTERLEAVE-NEXT: [[TMP13:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i64 22294; INTERLEAVE-NEXT: [[TMP14:%.*]] = udiv i32 [[TMP13]], [[TMP12]]2295; INTERLEAVE-NEXT: [[TMP15:%.*]] = insertelement <4 x i32> [[TMP11]], i32 [[TMP14]], i64 22296; INTERLEAVE-NEXT: br label [[PRED_UDIV_CONTINUE6]]2297; INTERLEAVE: pred.udiv.continue6:2298; INTERLEAVE-NEXT: [[TMP16:%.*]] = phi <4 x i32> [ [[TMP11]], [[PRED_UDIV_CONTINUE4]] ], [ [[TMP15]], [[PRED_UDIV_IF5]] ]2299; INTERLEAVE-NEXT: br i1 [[C]], label [[PRED_UDIV_IF7:%.*]], label [[PRED_UDIV_CONTINUE8:%.*]]2300; INTERLEAVE: pred.udiv.if7:2301; INTERLEAVE-NEXT: [[TMP17:%.*]] = or disjoint i32 [[INDEX]], 32302; INTERLEAVE-NEXT: [[TMP18:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i64 32303; INTERLEAVE-NEXT: [[TMP19:%.*]] = udiv i32 [[TMP18]], [[TMP17]]2304; INTERLEAVE-NEXT: [[TMP20:%.*]] = insertelement <4 x i32> [[TMP16]], i32 [[TMP19]], i64 32305; INTERLEAVE-NEXT: br label [[PRED_UDIV_CONTINUE8]]2306; INTERLEAVE: pred.udiv.continue8:2307; INTERLEAVE-NEXT: [[TMP21:%.*]] = phi <4 x i32> [ [[TMP16]], [[PRED_UDIV_CONTINUE6]] ], [ [[TMP20]], [[PRED_UDIV_IF7]] ]2308; INTERLEAVE-NEXT: br i1 [[C]], label [[PRED_UDIV_IF9:%.*]], label [[PRED_UDIV_CONTINUE10:%.*]]2309; INTERLEAVE: pred.udiv.if9:2310; INTERLEAVE-NEXT: [[TMP22:%.*]] = or disjoint i32 [[INDEX]], 42311; INTERLEAVE-NEXT: [[TMP23:%.*]] = extractelement <4 x i32> [[WIDE_LOAD2]], i64 02312; INTERLEAVE-NEXT: [[TMP24:%.*]] = udiv i32 [[TMP23]], [[TMP22]]2313; INTERLEAVE-NEXT: [[TMP25:%.*]] = insertelement <4 x i32> poison, i32 [[TMP24]], i64 02314; INTERLEAVE-NEXT: br label [[PRED_UDIV_CONTINUE10]]2315; INTERLEAVE: pred.udiv.continue10:2316; INTERLEAVE-NEXT: [[TMP26:%.*]] = phi <4 x i32> [ poison, [[PRED_UDIV_CONTINUE8]] ], [ [[TMP25]], [[PRED_UDIV_IF9]] ]2317; INTERLEAVE-NEXT: br i1 [[C]], label [[PRED_UDIV_IF11:%.*]], label [[PRED_UDIV_CONTINUE12:%.*]]2318; INTERLEAVE: pred.udiv.if11:2319; INTERLEAVE-NEXT: [[TMP27:%.*]] = or disjoint i32 [[INDEX]], 52320; INTERLEAVE-NEXT: [[TMP28:%.*]] = extractelement <4 x i32> [[WIDE_LOAD2]], i64 12321; INTERLEAVE-NEXT: [[TMP29:%.*]] = udiv i32 [[TMP28]], [[TMP27]]2322; INTERLEAVE-NEXT: [[TMP30:%.*]] = insertelement <4 x i32> [[TMP26]], i32 [[TMP29]], i64 12323; INTERLEAVE-NEXT: br label [[PRED_UDIV_CONTINUE12]]2324; INTERLEAVE: pred.udiv.continue12:2325; INTERLEAVE-NEXT: [[TMP31:%.*]] = phi <4 x i32> [ [[TMP26]], [[PRED_UDIV_CONTINUE10]] ], [ [[TMP30]], [[PRED_UDIV_IF11]] ]2326; INTERLEAVE-NEXT: br i1 [[C]], label [[PRED_UDIV_IF13:%.*]], label [[PRED_UDIV_CONTINUE14:%.*]]2327; INTERLEAVE: pred.udiv.if13:2328; INTERLEAVE-NEXT: [[TMP32:%.*]] = or disjoint i32 [[INDEX]], 62329; INTERLEAVE-NEXT: [[TMP33:%.*]] = extractelement <4 x i32> [[WIDE_LOAD2]], i64 22330; INTERLEAVE-NEXT: [[TMP34:%.*]] = udiv i32 [[TMP33]], [[TMP32]]2331; INTERLEAVE-NEXT: [[TMP35:%.*]] = insertelement <4 x i32> [[TMP31]], i32 [[TMP34]], i64 22332; INTERLEAVE-NEXT: br label [[PRED_UDIV_CONTINUE14]]2333; INTERLEAVE: pred.udiv.continue14:2334; INTERLEAVE-NEXT: [[TMP36:%.*]] = phi <4 x i32> [ [[TMP31]], [[PRED_UDIV_CONTINUE12]] ], [ [[TMP35]], [[PRED_UDIV_IF13]] ]2335; INTERLEAVE-NEXT: br i1 [[C]], label [[PRED_UDIV_IF15:%.*]], label [[PRED_UDIV_CONTINUE16]]2336; INTERLEAVE: pred.udiv.if15:2337; INTERLEAVE-NEXT: [[TMP37:%.*]] = or disjoint i32 [[INDEX]], 72338; INTERLEAVE-NEXT: [[TMP38:%.*]] = extractelement <4 x i32> [[WIDE_LOAD2]], i64 32339; INTERLEAVE-NEXT: [[TMP39:%.*]] = udiv i32 [[TMP38]], [[TMP37]]2340; INTERLEAVE-NEXT: [[TMP40:%.*]] = insertelement <4 x i32> [[TMP36]], i32 [[TMP39]], i64 32341; INTERLEAVE-NEXT: br label [[PRED_UDIV_CONTINUE16]]2342; INTERLEAVE: pred.udiv.continue16:2343; INTERLEAVE-NEXT: [[PREDPHI17:%.*]] = phi <4 x i32> [ [[WIDE_LOAD2]], [[PRED_UDIV_CONTINUE14]] ], [ [[TMP40]], [[PRED_UDIV_IF15]] ]2344; INTERLEAVE-NEXT: [[PREDPHI:%.*]] = phi <4 x i32> [ [[WIDE_LOAD]], [[PRED_UDIV_CONTINUE14]] ], [ [[TMP21]], [[PRED_UDIV_IF15]] ]2345; INTERLEAVE-NEXT: [[TMP42]] = add <4 x i32> [[PREDPHI]], [[VEC_PHI]]2346; INTERLEAVE-NEXT: [[TMP43]] = add <4 x i32> [[PREDPHI17]], [[VEC_PHI1]]2347; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 82348; INTERLEAVE-NEXT: [[TMP44:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]2349; INTERLEAVE-NEXT: br i1 [[TMP44]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]2350; INTERLEAVE: middle.block:2351; INTERLEAVE-NEXT: [[BIN_RDX:%.*]] = add <4 x i32> [[TMP43]], [[TMP42]]2352; INTERLEAVE-NEXT: [[TMP45:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])2353; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[SMAX]], [[N_VEC]]2354; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]2355; INTERLEAVE: scalar.ph:2356; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]2357; INTERLEAVE-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP45]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]2358; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]2359; INTERLEAVE: for.body:2360; INTERLEAVE-NEXT: [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[IF_END:%.*]] ]2361; INTERLEAVE-NEXT: [[SUM:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR4:%.*]], [[IF_END]] ]2362; INTERLEAVE-NEXT: [[TMP46:%.*]] = zext nneg i32 [[I]] to i642363; INTERLEAVE-NEXT: [[VAR0:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[TMP46]]2364; INTERLEAVE-NEXT: [[VAR1:%.*]] = load i32, ptr [[VAR0]], align 42365; INTERLEAVE-NEXT: br i1 [[C]], label [[IF_THEN:%.*]], label [[IF_END]]2366; INTERLEAVE: if.then:2367; INTERLEAVE-NEXT: [[VAR2:%.*]] = udiv i32 [[VAR1]], [[I]]2368; INTERLEAVE-NEXT: br label [[IF_END]]2369; INTERLEAVE: if.end:2370; INTERLEAVE-NEXT: [[VAR3:%.*]] = phi i32 [ [[VAR2]], [[IF_THEN]] ], [ [[VAR1]], [[FOR_BODY]] ]2371; INTERLEAVE-NEXT: [[VAR4]] = add i32 [[VAR3]], [[SUM]]2372; INTERLEAVE-NEXT: [[I_NEXT]] = add nuw nsw i32 [[I]], 12373; INTERLEAVE-NEXT: [[COND:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]2374; INTERLEAVE-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP25:![0-9]+]]2375; INTERLEAVE: for.end:2376; INTERLEAVE-NEXT: [[VAR5:%.*]] = phi i32 [ [[VAR4]], [[IF_END]] ], [ [[TMP45]], [[MIDDLE_BLOCK]] ]2377; INTERLEAVE-NEXT: ret i32 [[VAR5]]2378;2379entry:2380 br label %for.body2381 2382for.body:2383 %i = phi i32 [ 0, %entry ], [ %i.next, %if.end ]2384 %sum = phi i32 [ 0, %entry ], [ %var4, %if.end ]2385 %var0 = getelementptr inbounds i32, ptr %a, i32 %i2386 %var1 = load i32, ptr %var0, align 42387 br i1 %c, label %if.then, label %if.end2388 2389if.then:2390 %var2 = udiv i32 %var1, %i2391 br label %if.end2392 2393if.end:2394 %var3 = phi i32 [ %var2, %if.then ], [ %var1, %for.body ]2395 %var4 = add i32 %var3, %sum2396 %i.next = add nuw nsw i32 %i, 12397 %cond = icmp slt i32 %i.next, %n2398 br i1 %cond, label %for.body, label %for.end2399 2400for.end:2401 %var5 = phi i32 [ %var4, %if.end ]2402 ret i32 %var52403}2404 2405; Ensure we generate both a vector and a scalar induction variable. In this2406; test, the induction variable is used by an instruction that will be2407; vectorized (trunc) as well as an instruction that will remain in scalar form2408; (gepelementptr).2409;2410;2411;2412 2413%pair.i16 = type { i16, i16 }2414define void @iv_vector_and_scalar_users(ptr %p, i32 %a, i32 %n) {2415; CHECK-LABEL: @iv_vector_and_scalar_users(2416; CHECK-NEXT: entry:2417; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -12418; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i642419; CHECK-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 12420; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 22421; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]2422; CHECK: vector.ph:2423; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 22424; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]2425; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[A:%.*]], i64 02426; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer2427; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]2428; CHECK: vector.body:2429; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]2430; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]2431; CHECK-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 02432; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 12433; CHECK-NEXT: [[TMP5:%.*]] = add <2 x i32> [[BROADCAST_SPLAT]], [[VEC_IND]]2434; CHECK-NEXT: [[TMP6:%.*]] = trunc <2 x i32> [[TMP5]] to <2 x i16>2435; CHECK-NEXT: [[TMP9:%.*]] = extractelement <2 x i16> [[TMP6]], i32 02436; CHECK-NEXT: [[TMP10:%.*]] = extractelement <2 x i16> [[TMP6]], i32 12437; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[PAIR_I16:%.*]], ptr [[P:%.*]], i64 [[TMP3]], i32 12438; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[PAIR_I16]], ptr [[P]], i64 [[TMP4]], i32 12439; CHECK-NEXT: store i16 [[TMP9]], ptr [[TMP7]], align 22440; CHECK-NEXT: store i16 [[TMP10]], ptr [[TMP8]], align 22441; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 22442; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)2443; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]2444; CHECK-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]2445; CHECK: middle.block:2446; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]2447; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]2448; CHECK: scalar.ph:2449; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]2450; CHECK-NEXT: br label [[FOR_BODY:%.*]]2451; CHECK: for.body:2452; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]2453; CHECK-NEXT: [[TMP12:%.*]] = trunc i64 [[I]] to i322454; CHECK-NEXT: [[TMP13:%.*]] = add i32 [[A]], [[TMP12]]2455; CHECK-NEXT: [[TMP14:%.*]] = trunc i32 [[TMP13]] to i162456; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[PAIR_I16]], ptr [[P]], i64 [[I]], i32 12457; CHECK-NEXT: store i16 [[TMP14]], ptr [[TMP15]], align 22458; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 12459; CHECK-NEXT: [[TMP16:%.*]] = trunc i64 [[I_NEXT]] to i322460; CHECK-NEXT: [[COND:%.*]] = icmp eq i32 [[TMP16]], [[N]]2461; CHECK-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]2462; CHECK: for.end:2463; CHECK-NEXT: ret void2464;2465; IND-LABEL: @iv_vector_and_scalar_users(2466; IND-NEXT: entry:2467; IND-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -12468; IND-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i642469; IND-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 12470; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP0]], 02471; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]2472; IND: vector.ph:2473; IND-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 85899345902474; IND-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[A:%.*]], i64 02475; IND-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer2476; IND-NEXT: br label [[VECTOR_BODY:%.*]]2477; IND: vector.body:2478; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]2479; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]2480; IND-NEXT: [[TMP4:%.*]] = add <2 x i32> [[BROADCAST_SPLAT]], [[VEC_IND]]2481; IND-NEXT: [[TMP5:%.*]] = trunc <2 x i32> [[TMP4]] to <2 x i16>2482; IND-NEXT: [[TMP8:%.*]] = extractelement <2 x i16> [[TMP5]], i64 02483; IND-NEXT: [[TMP9:%.*]] = extractelement <2 x i16> [[TMP5]], i64 12484; IND-NEXT: [[DOTSPLIT:%.*]] = getelementptr inbounds [[PAIR_I16:%.*]], ptr [[P:%.*]], i64 [[INDEX]]2485; IND-NEXT: [[TMP6:%.*]] = getelementptr inbounds nuw i8, ptr [[DOTSPLIT]], i64 22486; IND-NEXT: [[TMP16:%.*]] = getelementptr [[PAIR_I16]], ptr [[P]], i64 [[INDEX]]2487; IND-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[TMP16]], i64 62488; IND-NEXT: store i16 [[TMP8]], ptr [[TMP6]], align 22489; IND-NEXT: store i16 [[TMP9]], ptr [[TMP7]], align 22490; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 22491; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)2492; IND-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]2493; IND-NEXT: br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]2494; IND: middle.block:2495; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]2496; IND-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]2497; IND: scalar.ph:2498; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]2499; IND-NEXT: br label [[FOR_BODY:%.*]]2500; IND: for.body:2501; IND-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]2502; IND-NEXT: [[TMP11:%.*]] = trunc i64 [[I]] to i322503; IND-NEXT: [[TMP12:%.*]] = add i32 [[A]], [[TMP11]]2504; IND-NEXT: [[TMP13:%.*]] = trunc i32 [[TMP12]] to i162505; IND-NEXT: [[DOTSPLIT2:%.*]] = getelementptr inbounds nuw [[PAIR_I16]], ptr [[P]], i64 [[I]]2506; IND-NEXT: [[TMP14:%.*]] = getelementptr inbounds nuw i8, ptr [[DOTSPLIT2]], i64 22507; IND-NEXT: store i16 [[TMP13]], ptr [[TMP14]], align 22508; IND-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 12509; IND-NEXT: [[TMP15:%.*]] = trunc i64 [[I_NEXT]] to i322510; IND-NEXT: [[COND:%.*]] = icmp eq i32 [[N]], [[TMP15]]2511; IND-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]2512; IND: for.end:2513; IND-NEXT: ret void2514;2515; UNROLL-LABEL: @iv_vector_and_scalar_users(2516; UNROLL-NEXT: entry:2517; UNROLL-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -12518; UNROLL-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i642519; UNROLL-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 12520; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 32521; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]2522; UNROLL: vector.ph:2523; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 85899345882524; UNROLL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[A:%.*]], i64 02525; UNROLL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer2526; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]2527; UNROLL: vector.body:2528; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]2529; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]2530; UNROLL-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)2531; UNROLL-NEXT: [[TMP6:%.*]] = add <2 x i32> [[BROADCAST_SPLAT]], [[VEC_IND]]2532; UNROLL-NEXT: [[TMP7:%.*]] = add <2 x i32> [[BROADCAST_SPLAT]], [[STEP_ADD]]2533; UNROLL-NEXT: [[TMP8:%.*]] = trunc <2 x i32> [[TMP6]] to <2 x i16>2534; UNROLL-NEXT: [[TMP14:%.*]] = extractelement <2 x i16> [[TMP8]], i64 02535; UNROLL-NEXT: [[TMP15:%.*]] = extractelement <2 x i16> [[TMP8]], i64 12536; UNROLL-NEXT: [[TMP9:%.*]] = trunc <2 x i32> [[TMP7]] to <2 x i16>2537; UNROLL-NEXT: [[TMP16:%.*]] = extractelement <2 x i16> [[TMP9]], i64 02538; UNROLL-NEXT: [[TMP17:%.*]] = extractelement <2 x i16> [[TMP9]], i64 12539; UNROLL-NEXT: [[DOTSPLIT:%.*]] = getelementptr inbounds [[PAIR_I16:%.*]], ptr [[P:%.*]], i64 [[INDEX]]2540; UNROLL-NEXT: [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[DOTSPLIT]], i64 22541; UNROLL-NEXT: [[TMP24:%.*]] = getelementptr [[PAIR_I16]], ptr [[P]], i64 [[INDEX]]2542; UNROLL-NEXT: [[TMP11:%.*]] = getelementptr i8, ptr [[TMP24]], i64 62543; UNROLL-NEXT: [[TMP25:%.*]] = getelementptr [[PAIR_I16]], ptr [[P]], i64 [[INDEX]]2544; UNROLL-NEXT: [[TMP12:%.*]] = getelementptr i8, ptr [[TMP25]], i64 102545; UNROLL-NEXT: [[TMP26:%.*]] = getelementptr [[PAIR_I16]], ptr [[P]], i64 [[INDEX]]2546; UNROLL-NEXT: [[TMP13:%.*]] = getelementptr i8, ptr [[TMP26]], i64 142547; UNROLL-NEXT: store i16 [[TMP14]], ptr [[TMP10]], align 22548; UNROLL-NEXT: store i16 [[TMP15]], ptr [[TMP11]], align 22549; UNROLL-NEXT: store i16 [[TMP16]], ptr [[TMP12]], align 22550; UNROLL-NEXT: store i16 [[TMP17]], ptr [[TMP13]], align 22551; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 42552; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 4)2553; UNROLL-NEXT: [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]2554; UNROLL-NEXT: br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]2555; UNROLL: middle.block:2556; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]2557; UNROLL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]2558; UNROLL: scalar.ph:2559; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]2560; UNROLL-NEXT: br label [[FOR_BODY:%.*]]2561; UNROLL: for.body:2562; UNROLL-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]2563; UNROLL-NEXT: [[TMP19:%.*]] = trunc i64 [[I]] to i322564; UNROLL-NEXT: [[TMP20:%.*]] = add i32 [[A]], [[TMP19]]2565; UNROLL-NEXT: [[TMP21:%.*]] = trunc i32 [[TMP20]] to i162566; UNROLL-NEXT: [[DOTSPLIT4:%.*]] = getelementptr inbounds nuw [[PAIR_I16]], ptr [[P]], i64 [[I]]2567; UNROLL-NEXT: [[TMP22:%.*]] = getelementptr inbounds nuw i8, ptr [[DOTSPLIT4]], i64 22568; UNROLL-NEXT: store i16 [[TMP21]], ptr [[TMP22]], align 22569; UNROLL-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 12570; UNROLL-NEXT: [[TMP23:%.*]] = trunc i64 [[I_NEXT]] to i322571; UNROLL-NEXT: [[COND:%.*]] = icmp eq i32 [[N]], [[TMP23]]2572; UNROLL-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]2573; UNROLL: for.end:2574; UNROLL-NEXT: ret void2575;2576; UNROLL-NO-IC-LABEL: @iv_vector_and_scalar_users(2577; UNROLL-NO-IC-NEXT: entry:2578; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -12579; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i642580; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 12581; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 42582; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]2583; UNROLL-NO-IC: vector.ph:2584; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 42585; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]2586; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[A:%.*]], i64 02587; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer2588; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]2589; UNROLL-NO-IC: vector.body:2590; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]2591; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]2592; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)2593; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 02594; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 12595; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = add i64 [[INDEX]], 22596; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], 32597; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = add <2 x i32> [[BROADCAST_SPLAT]], [[VEC_IND]]2598; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = add <2 x i32> [[BROADCAST_SPLAT]], [[STEP_ADD]]2599; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = trunc <2 x i32> [[TMP7]] to <2 x i16>2600; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = extractelement <2 x i16> [[TMP9]], i32 02601; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = extractelement <2 x i16> [[TMP9]], i32 12602; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = trunc <2 x i32> [[TMP8]] to <2 x i16>2603; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = extractelement <2 x i16> [[TMP10]], i32 02604; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = extractelement <2 x i16> [[TMP10]], i32 12605; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[PAIR_I16:%.*]], ptr [[P:%.*]], i64 [[TMP3]], i32 12606; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[PAIR_I16]], ptr [[P]], i64 [[TMP4]], i32 12607; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[PAIR_I16]], ptr [[P]], i64 [[TMP5]], i32 12608; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[PAIR_I16]], ptr [[P]], i64 [[TMP6]], i32 12609; UNROLL-NO-IC-NEXT: store i16 [[TMP15]], ptr [[TMP11]], align 22610; UNROLL-NO-IC-NEXT: store i16 [[TMP16]], ptr [[TMP12]], align 22611; UNROLL-NO-IC-NEXT: store i16 [[TMP17]], ptr [[TMP13]], align 22612; UNROLL-NO-IC-NEXT: store i16 [[TMP18]], ptr [[TMP14]], align 22613; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 42614; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], splat (i32 2)2615; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]2616; UNROLL-NO-IC-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]2617; UNROLL-NO-IC: middle.block:2618; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]2619; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]2620; UNROLL-NO-IC: scalar.ph:2621; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]2622; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]2623; UNROLL-NO-IC: for.body:2624; UNROLL-NO-IC-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]2625; UNROLL-NO-IC-NEXT: [[TMP20:%.*]] = trunc i64 [[I]] to i322626; UNROLL-NO-IC-NEXT: [[TMP21:%.*]] = add i32 [[A]], [[TMP20]]2627; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = trunc i32 [[TMP21]] to i162628; UNROLL-NO-IC-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[PAIR_I16]], ptr [[P]], i64 [[I]], i32 12629; UNROLL-NO-IC-NEXT: store i16 [[TMP22]], ptr [[TMP23]], align 22630; UNROLL-NO-IC-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 12631; UNROLL-NO-IC-NEXT: [[TMP24:%.*]] = trunc i64 [[I_NEXT]] to i322632; UNROLL-NO-IC-NEXT: [[COND:%.*]] = icmp eq i32 [[TMP24]], [[N]]2633; UNROLL-NO-IC-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]2634; UNROLL-NO-IC: for.end:2635; UNROLL-NO-IC-NEXT: ret void2636;2637; INTERLEAVE-LABEL: @iv_vector_and_scalar_users(2638; INTERLEAVE-NEXT: entry:2639; INTERLEAVE-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -12640; INTERLEAVE-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i642641; INTERLEAVE-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 12642; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 72643; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]2644; INTERLEAVE: vector.ph:2645; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 85899345842646; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[A:%.*]], i64 02647; INTERLEAVE-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer2648; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]2649; INTERLEAVE: vector.body:2650; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]2651; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]2652; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], splat (i32 4)2653; INTERLEAVE-NEXT: [[TMP10:%.*]] = add <4 x i32> [[BROADCAST_SPLAT]], [[VEC_IND]]2654; INTERLEAVE-NEXT: [[TMP11:%.*]] = add <4 x i32> [[BROADCAST_SPLAT]], [[STEP_ADD]]2655; INTERLEAVE-NEXT: [[TMP12:%.*]] = trunc <4 x i32> [[TMP10]] to <4 x i16>2656; INTERLEAVE-NEXT: [[TMP22:%.*]] = extractelement <4 x i16> [[TMP12]], i64 02657; INTERLEAVE-NEXT: [[TMP23:%.*]] = extractelement <4 x i16> [[TMP12]], i64 12658; INTERLEAVE-NEXT: [[TMP24:%.*]] = extractelement <4 x i16> [[TMP12]], i64 22659; INTERLEAVE-NEXT: [[TMP25:%.*]] = extractelement <4 x i16> [[TMP12]], i64 32660; INTERLEAVE-NEXT: [[TMP13:%.*]] = trunc <4 x i32> [[TMP11]] to <4 x i16>2661; INTERLEAVE-NEXT: [[TMP26:%.*]] = extractelement <4 x i16> [[TMP13]], i64 02662; INTERLEAVE-NEXT: [[TMP27:%.*]] = extractelement <4 x i16> [[TMP13]], i64 12663; INTERLEAVE-NEXT: [[TMP28:%.*]] = extractelement <4 x i16> [[TMP13]], i64 22664; INTERLEAVE-NEXT: [[TMP29:%.*]] = extractelement <4 x i16> [[TMP13]], i64 32665; INTERLEAVE-NEXT: [[DOTSPLIT:%.*]] = getelementptr inbounds [[PAIR_I16:%.*]], ptr [[P:%.*]], i64 [[INDEX]]2666; INTERLEAVE-NEXT: [[TMP14:%.*]] = getelementptr inbounds nuw i8, ptr [[DOTSPLIT]], i64 22667; INTERLEAVE-NEXT: [[TMP8:%.*]] = getelementptr [[PAIR_I16]], ptr [[P]], i64 [[INDEX]]2668; INTERLEAVE-NEXT: [[TMP15:%.*]] = getelementptr i8, ptr [[TMP8]], i64 62669; INTERLEAVE-NEXT: [[TMP36:%.*]] = getelementptr [[PAIR_I16]], ptr [[P]], i64 [[INDEX]]2670; INTERLEAVE-NEXT: [[TMP16:%.*]] = getelementptr i8, ptr [[TMP36]], i64 102671; INTERLEAVE-NEXT: [[TMP37:%.*]] = getelementptr [[PAIR_I16]], ptr [[P]], i64 [[INDEX]]2672; INTERLEAVE-NEXT: [[TMP17:%.*]] = getelementptr i8, ptr [[TMP37]], i64 142673; INTERLEAVE-NEXT: [[TMP38:%.*]] = getelementptr [[PAIR_I16]], ptr [[P]], i64 [[INDEX]]2674; INTERLEAVE-NEXT: [[TMP18:%.*]] = getelementptr i8, ptr [[TMP38]], i64 182675; INTERLEAVE-NEXT: [[TMP39:%.*]] = getelementptr [[PAIR_I16]], ptr [[P]], i64 [[INDEX]]2676; INTERLEAVE-NEXT: [[TMP19:%.*]] = getelementptr i8, ptr [[TMP39]], i64 222677; INTERLEAVE-NEXT: [[TMP40:%.*]] = getelementptr [[PAIR_I16]], ptr [[P]], i64 [[INDEX]]2678; INTERLEAVE-NEXT: [[TMP20:%.*]] = getelementptr i8, ptr [[TMP40]], i64 262679; INTERLEAVE-NEXT: [[TMP41:%.*]] = getelementptr [[PAIR_I16]], ptr [[P]], i64 [[INDEX]]2680; INTERLEAVE-NEXT: [[TMP21:%.*]] = getelementptr i8, ptr [[TMP41]], i64 302681; INTERLEAVE-NEXT: store i16 [[TMP22]], ptr [[TMP14]], align 22682; INTERLEAVE-NEXT: store i16 [[TMP23]], ptr [[TMP15]], align 22683; INTERLEAVE-NEXT: store i16 [[TMP24]], ptr [[TMP16]], align 22684; INTERLEAVE-NEXT: store i16 [[TMP25]], ptr [[TMP17]], align 22685; INTERLEAVE-NEXT: store i16 [[TMP26]], ptr [[TMP18]], align 22686; INTERLEAVE-NEXT: store i16 [[TMP27]], ptr [[TMP19]], align 22687; INTERLEAVE-NEXT: store i16 [[TMP28]], ptr [[TMP20]], align 22688; INTERLEAVE-NEXT: store i16 [[TMP29]], ptr [[TMP21]], align 22689; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 82690; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 8)2691; INTERLEAVE-NEXT: [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]2692; INTERLEAVE-NEXT: br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]2693; INTERLEAVE: middle.block:2694; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]2695; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]2696; INTERLEAVE: scalar.ph:2697; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]2698; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]2699; INTERLEAVE: for.body:2700; INTERLEAVE-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]2701; INTERLEAVE-NEXT: [[TMP31:%.*]] = trunc i64 [[I]] to i322702; INTERLEAVE-NEXT: [[TMP32:%.*]] = add i32 [[A]], [[TMP31]]2703; INTERLEAVE-NEXT: [[TMP33:%.*]] = trunc i32 [[TMP32]] to i162704; INTERLEAVE-NEXT: [[DOTSPLIT8:%.*]] = getelementptr inbounds nuw [[PAIR_I16]], ptr [[P]], i64 [[I]]2705; INTERLEAVE-NEXT: [[TMP34:%.*]] = getelementptr inbounds nuw i8, ptr [[DOTSPLIT8]], i64 22706; INTERLEAVE-NEXT: store i16 [[TMP33]], ptr [[TMP34]], align 22707; INTERLEAVE-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 12708; INTERLEAVE-NEXT: [[TMP35:%.*]] = trunc i64 [[I_NEXT]] to i322709; INTERLEAVE-NEXT: [[COND:%.*]] = icmp eq i32 [[N]], [[TMP35]]2710; INTERLEAVE-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]2711; INTERLEAVE: for.end:2712; INTERLEAVE-NEXT: ret void2713;2714entry:2715 br label %for.body2716 2717for.body:2718 %i = phi i64 [ %i.next, %for.body ], [ 0, %entry ]2719 %0 = trunc i64 %i to i322720 %1 = add i32 %a, %02721 %2 = trunc i32 %1 to i162722 %3 = getelementptr inbounds %pair.i16, ptr %p, i64 %i, i32 12723 store i16 %2, ptr %3, align 22724 %i.next = add nuw nsw i64 %i, 12725 %4 = trunc i64 %i.next to i322726 %cond = icmp eq i32 %4, %n2727 br i1 %cond, label %for.end, label %for.body2728 2729for.end:2730 ret void2731}2732 2733; Make sure that the loop exit count computation does not overflow for i8 and2734; i16. The exit count of these loops is i8/i16 max + 1. If we don't cast the2735; induction variable to a bigger type the exit count computation will overflow2736; to 0.2737; PR175322738 2739define i32 @i8_loop() nounwind readnone ssp uwtable {2740; CHECK-LABEL: @i8_loop(2741; CHECK-NEXT: entry:2742; CHECK-NEXT: br label [[VECTOR_PH:%.*]]2743; CHECK: vector.ph:2744; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]2745; CHECK: vector.body:2746; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]2747; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ <i32 1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP0:%.*]], [[VECTOR_BODY]] ]2748; CHECK-NEXT: [[TMP0]] = and <2 x i32> [[VEC_PHI]], splat (i32 4)2749; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 22750; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 2562751; CHECK-NEXT: br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]2752; CHECK: middle.block:2753; CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[TMP0]])2754; CHECK-NEXT: br label [[LOOP:%.*]]2755; CHECK: exit:2756; CHECK-NEXT: ret i32 [[TMP2]]2757;2758; IND-LABEL: @i8_loop(2759; IND-NEXT: entry:2760; IND-NEXT: br label [[VECTOR_PH:%.*]]2761; IND: vector.ph:2762; IND-NEXT: br label [[VECTOR_BODY:%.*]]2763; IND: vector.body:2764; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]2765; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 22766; IND-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 2562767; IND-NEXT: br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]2768; IND: middle.block:2769; IND-NEXT: br label [[LOOP:%.*]]2770; IND: exit:2771; IND-NEXT: ret i32 02772;2773; UNROLL-LABEL: @i8_loop(2774; UNROLL-NEXT: entry:2775; UNROLL-NEXT: br label [[VECTOR_PH:%.*]]2776; UNROLL: vector.ph:2777; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]2778; UNROLL: vector.body:2779; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]2780; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 42781; UNROLL-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 2562782; UNROLL-NEXT: br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]2783; UNROLL: middle.block:2784; UNROLL-NEXT: br label [[LOOP:%.*]]2785; UNROLL: exit:2786; UNROLL-NEXT: ret i32 02787;2788; UNROLL-NO-IC-LABEL: @i8_loop(2789; UNROLL-NO-IC-NEXT: entry:2790; UNROLL-NO-IC-NEXT: br label [[VECTOR_PH:%.*]]2791; UNROLL-NO-IC: vector.ph:2792; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]2793; UNROLL-NO-IC: vector.body:2794; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]2795; UNROLL-NO-IC-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ <i32 1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP0:%.*]], [[VECTOR_BODY]] ]2796; UNROLL-NO-IC-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i32> [ splat (i32 -1), [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]2797; UNROLL-NO-IC-NEXT: [[TMP0]] = and <2 x i32> [[VEC_PHI]], splat (i32 4)2798; UNROLL-NO-IC-NEXT: [[TMP1]] = and <2 x i32> [[VEC_PHI1]], splat (i32 4)2799; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 42800; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 2562801; UNROLL-NO-IC-NEXT: br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]2802; UNROLL-NO-IC: middle.block:2803; UNROLL-NO-IC-NEXT: [[BIN_RDX:%.*]] = and <2 x i32> [[TMP1]], [[TMP0]]2804; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[BIN_RDX]])2805; UNROLL-NO-IC-NEXT: br label [[LOOP:%.*]]2806; UNROLL-NO-IC: exit:2807; UNROLL-NO-IC-NEXT: ret i32 [[TMP3]]2808;2809; INTERLEAVE-LABEL: @i8_loop(2810; INTERLEAVE-NEXT: entry:2811; INTERLEAVE-NEXT: br label [[VECTOR_PH:%.*]]2812; INTERLEAVE: vector.ph:2813; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]2814; INTERLEAVE: vector.body:2815; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]2816; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 82817; INTERLEAVE-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 2562818; INTERLEAVE-NEXT: br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]2819; INTERLEAVE: middle.block:2820; INTERLEAVE-NEXT: br label [[LOOP:%.*]]2821; INTERLEAVE: exit:2822; INTERLEAVE-NEXT: ret i32 02823;2824entry:2825 br label %loop2826 2827loop:2828 %a.0 = phi i32 [ 1, %entry ], [ %a.0.and, %loop ]2829 %b.0 = phi i8 [ 0, %entry ], [ %b.next, %loop ]2830 %a.0.and = and i32 %a.0, 42831 %b.next = add i8 %b.0, -12832 %ec = icmp eq i8 %b.next, 02833 br i1 %ec, label %exit, label %loop2834 2835exit:2836 ret i32 %a.0.and2837}2838 2839 2840define i32 @i16_loop() nounwind readnone ssp uwtable {2841; CHECK-LABEL: @i16_loop(2842; CHECK-NEXT: entry:2843; CHECK-NEXT: br label [[VECTOR_PH:%.*]]2844; CHECK: vector.ph:2845; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]2846; CHECK: vector.body:2847; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]2848; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ <i32 1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP0:%.*]], [[VECTOR_BODY]] ]2849; CHECK-NEXT: [[TMP0]] = and <2 x i32> [[VEC_PHI]], splat (i32 4)2850; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 22851; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 655362852; CHECK-NEXT: br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP29:![0-9]+]]2853; CHECK: middle.block:2854; CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[TMP0]])2855; CHECK-NEXT: br label [[LOOP:%.*]]2856; CHECK: exit:2857; CHECK-NEXT: ret i32 [[TMP2]]2858;2859; IND-LABEL: @i16_loop(2860; IND-NEXT: entry:2861; IND-NEXT: br label [[VECTOR_PH:%.*]]2862; IND: vector.ph:2863; IND-NEXT: br label [[VECTOR_BODY:%.*]]2864; IND: vector.body:2865; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]2866; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 22867; IND-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 655362868; IND-NEXT: br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP29:![0-9]+]]2869; IND: middle.block:2870; IND-NEXT: br label [[LOOP:%.*]]2871; IND: exit:2872; IND-NEXT: ret i32 02873;2874; UNROLL-LABEL: @i16_loop(2875; UNROLL-NEXT: entry:2876; UNROLL-NEXT: br label [[VECTOR_PH:%.*]]2877; UNROLL: vector.ph:2878; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]2879; UNROLL: vector.body:2880; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]2881; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 42882; UNROLL-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 655362883; UNROLL-NEXT: br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP29:![0-9]+]]2884; UNROLL: middle.block:2885; UNROLL-NEXT: br label [[LOOP:%.*]]2886; UNROLL: exit:2887; UNROLL-NEXT: ret i32 02888;2889; UNROLL-NO-IC-LABEL: @i16_loop(2890; UNROLL-NO-IC-NEXT: entry:2891; UNROLL-NO-IC-NEXT: br label [[VECTOR_PH:%.*]]2892; UNROLL-NO-IC: vector.ph:2893; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]2894; UNROLL-NO-IC: vector.body:2895; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]2896; UNROLL-NO-IC-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ <i32 1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP0:%.*]], [[VECTOR_BODY]] ]2897; UNROLL-NO-IC-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i32> [ splat (i32 -1), [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]2898; UNROLL-NO-IC-NEXT: [[TMP0]] = and <2 x i32> [[VEC_PHI]], splat (i32 4)2899; UNROLL-NO-IC-NEXT: [[TMP1]] = and <2 x i32> [[VEC_PHI1]], splat (i32 4)2900; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 42901; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 655362902; UNROLL-NO-IC-NEXT: br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP29:![0-9]+]]2903; UNROLL-NO-IC: middle.block:2904; UNROLL-NO-IC-NEXT: [[BIN_RDX:%.*]] = and <2 x i32> [[TMP1]], [[TMP0]]2905; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[BIN_RDX]])2906; UNROLL-NO-IC-NEXT: br label [[LOOP:%.*]]2907; UNROLL-NO-IC: exit:2908; UNROLL-NO-IC-NEXT: ret i32 [[TMP3]]2909;2910; INTERLEAVE-LABEL: @i16_loop(2911; INTERLEAVE-NEXT: entry:2912; INTERLEAVE-NEXT: br label [[VECTOR_PH:%.*]]2913; INTERLEAVE: vector.ph:2914; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]2915; INTERLEAVE: vector.body:2916; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]2917; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 82918; INTERLEAVE-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 655362919; INTERLEAVE-NEXT: br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP29:![0-9]+]]2920; INTERLEAVE: middle.block:2921; INTERLEAVE-NEXT: br label [[LOOP:%.*]]2922; INTERLEAVE: exit:2923; INTERLEAVE-NEXT: ret i32 02924;2925entry:2926 br label %loop2927 2928loop:2929 %a.0 = phi i32 [ 1, %entry ], [ %a.0.and, %loop ]2930 %b.0 = phi i16 [ 0, %entry ], [ %b.0.next, %loop ]2931 %a.0.and = and i32 %a.0, 42932 %b.0.next = add i16 %b.0, -12933 %ec = icmp eq i16 %b.0.next, 02934 br i1 %ec, label %exit, label %loop2935 2936exit:2937 ret i32 %a.0.and2938}2939 2940; This loop has a backedge taken count of i32_max. We need to check for this2941; condition and can skip vectorizing.2942 2943define i32 @max_i32_backedgetaken() {2944; CHECK-LABEL: @max_i32_backedgetaken(2945; CHECK-NEXT: entry:2946; CHECK-NEXT: br label [[LOOP:%.*]]2947; CHECK: loop:2948; CHECK-NEXT: [[A_0:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ], [ [[A_0_AND:%.*]], [[LOOP]] ]2949; CHECK-NEXT: [[B_0:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[B_NEXT:%.*]], [[LOOP]] ]2950; CHECK-NEXT: [[A_0_AND]] = and i32 [[A_0]], 42951; CHECK-NEXT: [[B_NEXT]] = add i32 [[B_0]], -12952; CHECK-NEXT: [[EC:%.*]] = icmp eq i32 [[B_NEXT]], 02953; CHECK-NEXT: br i1 [[EC]], label [[EXIT:%.*]], label [[LOOP]]2954; CHECK: exit:2955; CHECK-NEXT: [[A_0_AND_LCSSA:%.*]] = phi i32 [ [[A_0_AND]], [[LOOP]] ]2956; CHECK-NEXT: ret i32 [[A_0_AND_LCSSA]]2957;2958; IND-LABEL: @max_i32_backedgetaken(2959; IND-NEXT: entry:2960; IND-NEXT: br label [[LOOP:%.*]]2961; IND: loop:2962; IND-NEXT: [[B_0:%.*]] = phi i32 [ 0, [[SCALAR_PH:%.*]] ], [ [[B_NEXT:%.*]], [[LOOP]] ]2963; IND-NEXT: [[B_NEXT]] = add i32 [[B_0]], -12964; IND-NEXT: [[EC:%.*]] = icmp eq i32 [[B_NEXT]], 02965; IND-NEXT: br i1 [[EC]], label [[EXIT:%.*]], label [[LOOP]]2966; IND: exit:2967; IND-NEXT: ret i32 02968;2969; UNROLL-LABEL: @max_i32_backedgetaken(2970; UNROLL-NEXT: entry:2971; UNROLL-NEXT: br label [[LOOP:%.*]]2972; UNROLL: loop:2973; UNROLL-NEXT: [[B_0:%.*]] = phi i32 [ 0, [[SCALAR_PH:%.*]] ], [ [[B_NEXT:%.*]], [[LOOP]] ]2974; UNROLL-NEXT: [[B_NEXT]] = add i32 [[B_0]], -12975; UNROLL-NEXT: [[EC:%.*]] = icmp eq i32 [[B_NEXT]], 02976; UNROLL-NEXT: br i1 [[EC]], label [[EXIT:%.*]], label [[LOOP]]2977; UNROLL: exit:2978; UNROLL-NEXT: ret i32 02979;2980; UNROLL-NO-IC-LABEL: @max_i32_backedgetaken(2981; UNROLL-NO-IC-NEXT: entry:2982; UNROLL-NO-IC-NEXT: br label [[LOOP:%.*]]2983; UNROLL-NO-IC: loop:2984; UNROLL-NO-IC-NEXT: [[A_0:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ], [ [[A_0_AND:%.*]], [[LOOP]] ]2985; UNROLL-NO-IC-NEXT: [[B_0:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[B_NEXT:%.*]], [[LOOP]] ]2986; UNROLL-NO-IC-NEXT: [[A_0_AND]] = and i32 [[A_0]], 42987; UNROLL-NO-IC-NEXT: [[B_NEXT]] = add i32 [[B_0]], -12988; UNROLL-NO-IC-NEXT: [[EC:%.*]] = icmp eq i32 [[B_NEXT]], 02989; UNROLL-NO-IC-NEXT: br i1 [[EC]], label [[EXIT:%.*]], label [[LOOP]]2990; UNROLL-NO-IC: exit:2991; UNROLL-NO-IC-NEXT: [[A_0_AND_LCSSA:%.*]] = phi i32 [ [[A_0_AND]], [[LOOP]] ]2992; UNROLL-NO-IC-NEXT: ret i32 [[A_0_AND_LCSSA]]2993;2994; INTERLEAVE-LABEL: @max_i32_backedgetaken(2995; INTERLEAVE-NEXT: entry:2996; INTERLEAVE-NEXT: br label [[LOOP:%.*]]2997; INTERLEAVE: loop:2998; INTERLEAVE-NEXT: [[B_0:%.*]] = phi i32 [ 0, [[SCALAR_PH:%.*]] ], [ [[B_NEXT:%.*]], [[LOOP]] ]2999; INTERLEAVE-NEXT: [[B_NEXT]] = add i32 [[B_0]], -13000; INTERLEAVE-NEXT: [[EC:%.*]] = icmp eq i32 [[B_NEXT]], 03001; INTERLEAVE-NEXT: br i1 [[EC]], label [[EXIT:%.*]], label [[LOOP]]3002; INTERLEAVE: exit:3003; INTERLEAVE-NEXT: ret i32 03004;3005entry:3006 br label %loop3007 3008loop:3009 %a.0 = phi i32 [ 1, %entry ], [ %a.0.and, %loop ]3010 %b.0 = phi i32 [ 0, %entry ], [ %b.next, %loop ]3011 %a.0.and = and i32 %a.0, 43012 %b.next = add i32 %b.0, -13013 %ec = icmp eq i32 %b.next, 03014 br i1 %ec, label %exit, label %loop3015 3016exit:3017 ret i32 %a.0.and3018}3019 3020; When generating the overflow check we must sure that the induction start value3021; is defined before the branch to the scalar preheader.3022 3023 3024 3025@e = global i8 1, align 13026@d = common global i32 0, align 43027@c = common global i32 0, align 43028define i32 @testoverflowcheck() {3029; CHECK-LABEL: @testoverflowcheck(3030; CHECK-NEXT: entry:3031; CHECK-NEXT: [[DOTPR_I:%.*]] = load i8, ptr @e, align 13032; CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr @d, align 43033; CHECK-NEXT: [[C_PROMOTED_I:%.*]] = load i32, ptr @c, align 43034; CHECK-NEXT: [[TMP1:%.*]] = sub i8 -1, [[DOTPR_I]]3035; CHECK-NEXT: [[TMP2:%.*]] = zext i8 [[TMP1]] to i323036; CHECK-NEXT: [[TMP3:%.*]] = add nuw nsw i32 [[TMP2]], 13037; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP3]], 23038; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]3039; CHECK: vector.ph:3040; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP3]], 23041; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP3]], [[N_MOD_VF]]3042; CHECK-NEXT: [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i83043; CHECK-NEXT: [[IND_END:%.*]] = add i8 [[DOTPR_I]], [[DOTCAST]]3044; CHECK-NEXT: [[TMP4:%.*]] = insertelement <2 x i32> splat (i32 -1), i32 [[C_PROMOTED_I]], i32 03045; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP0]], i64 03046; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer3047; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]3048; CHECK: vector.body:3049; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]3050; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP4]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]3051; CHECK-NEXT: [[TMP5]] = and <2 x i32> [[BROADCAST_SPLAT]], [[VEC_PHI]]3052; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 23053; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]3054; CHECK-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]3055; CHECK: middle.block:3056; CHECK-NEXT: [[TMP7:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[TMP5]])3057; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP3]], [[N_VEC]]3058; CHECK-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]3059; CHECK: scalar.ph:3060; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[DOTPR_I]], [[ENTRY:%.*]] ]3061; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ [[C_PROMOTED_I]], [[ENTRY]] ]3062; CHECK-NEXT: br label [[COND_END_I:%.*]]3063; CHECK: cond.end.i:3064; CHECK-NEXT: [[INC4_I:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC_I:%.*]], [[COND_END_I]] ]3065; CHECK-NEXT: [[AND3_I:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[AND_I:%.*]], [[COND_END_I]] ]3066; CHECK-NEXT: [[AND_I]] = and i32 [[TMP0]], [[AND3_I]]3067; CHECK-NEXT: [[INC_I]] = add i8 [[INC4_I]], 13068; CHECK-NEXT: [[TOBOOL_I:%.*]] = icmp eq i8 [[INC_I]], 03069; CHECK-NEXT: br i1 [[TOBOOL_I]], label [[LOOPEXIT]], label [[COND_END_I]], !llvm.loop [[LOOP31:![0-9]+]]3070; CHECK: loopexit:3071; CHECK-NEXT: [[AND_I_LCSSA:%.*]] = phi i32 [ [[AND_I]], [[COND_END_I]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]3072; CHECK-NEXT: ret i32 [[AND_I_LCSSA]]3073;3074; IND-LABEL: @testoverflowcheck(3075; IND-NEXT: entry:3076; IND-NEXT: [[DOTPR_I:%.*]] = load i8, ptr @e, align 13077; IND-NEXT: [[TMP0:%.*]] = load i32, ptr @d, align 43078; IND-NEXT: [[C_PROMOTED_I:%.*]] = load i32, ptr @c, align 43079; IND-NEXT: [[TMP1:%.*]] = xor i8 [[DOTPR_I]], -13080; IND-NEXT: [[TMP2:%.*]] = zext i8 [[TMP1]] to i323081; IND-NEXT: [[TMP3:%.*]] = add nuw nsw i32 [[TMP2]], 13082; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp eq i8 [[DOTPR_I]], -13083; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]3084; IND: vector.ph:3085; IND-NEXT: [[N_VEC:%.*]] = and i32 [[TMP3]], 5103086; IND-NEXT: [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i83087; IND-NEXT: [[IND_END:%.*]] = add i8 [[DOTPR_I]], [[DOTCAST]]3088; IND-NEXT: [[TMP4:%.*]] = insertelement <2 x i32> <i32 poison, i32 -1>, i32 [[C_PROMOTED_I]], i64 03089; IND-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP0]], i64 03090; IND-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer3091; IND-NEXT: br label [[VECTOR_BODY:%.*]]3092; IND: vector.body:3093; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]3094; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 23095; IND-NEXT: [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]3096; IND-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]3097; IND: middle.block:3098; IND-NEXT: [[TMP6:%.*]] = and <2 x i32> [[TMP4]], [[BROADCAST_SPLAT]]3099; IND-NEXT: [[TMP7:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[TMP6]])3100; IND-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP3]], [[N_VEC]]3101; IND-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]3102; IND: scalar.ph:3103; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[DOTPR_I]], [[ENTRY:%.*]] ]3104; IND-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ [[C_PROMOTED_I]], [[ENTRY]] ]3105; IND-NEXT: br label [[COND_END_I:%.*]]3106; IND: cond.end.i:3107; IND-NEXT: [[INC4_I:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC_I:%.*]], [[COND_END_I]] ]3108; IND-NEXT: [[AND_I:%.*]] = and i32 [[BC_MERGE_RDX]], [[TMP0]]3109; IND-NEXT: [[INC_I]] = add i8 [[INC4_I]], 13110; IND-NEXT: [[TOBOOL_I:%.*]] = icmp eq i8 [[INC_I]], 03111; IND-NEXT: br i1 [[TOBOOL_I]], label [[LOOPEXIT]], label [[COND_END_I]], !llvm.loop [[LOOP31:![0-9]+]]3112; IND: loopexit:3113; IND-NEXT: [[AND_I_LCSSA:%.*]] = phi i32 [ [[AND_I]], [[COND_END_I]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]3114; IND-NEXT: ret i32 [[AND_I_LCSSA]]3115;3116; UNROLL-LABEL: @testoverflowcheck(3117; UNROLL-NEXT: entry:3118; UNROLL-NEXT: [[DOTPR_I:%.*]] = load i8, ptr @e, align 13119; UNROLL-NEXT: [[TMP0:%.*]] = load i32, ptr @d, align 43120; UNROLL-NEXT: [[C_PROMOTED_I:%.*]] = load i32, ptr @c, align 43121; UNROLL-NEXT: [[TMP1:%.*]] = xor i8 [[DOTPR_I]], -13122; UNROLL-NEXT: [[TMP2:%.*]] = zext i8 [[TMP1]] to i323123; UNROLL-NEXT: [[TMP3:%.*]] = add nuw nsw i32 [[TMP2]], 13124; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ugt i8 [[DOTPR_I]], -43125; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]3126; UNROLL: vector.ph:3127; UNROLL-NEXT: [[N_VEC:%.*]] = and i32 [[TMP3]], 5083128; UNROLL-NEXT: [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i83129; UNROLL-NEXT: [[IND_END:%.*]] = add i8 [[DOTPR_I]], [[DOTCAST]]3130; UNROLL-NEXT: [[TMP4:%.*]] = insertelement <2 x i32> <i32 poison, i32 -1>, i32 [[C_PROMOTED_I]], i64 03131; UNROLL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP0]], i64 03132; UNROLL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer3133; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]3134; UNROLL: vector.body:3135; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]3136; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 43137; UNROLL-NEXT: [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]3138; UNROLL-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]3139; UNROLL: middle.block:3140; UNROLL-NEXT: [[TMP6:%.*]] = and <2 x i32> [[TMP4]], [[BROADCAST_SPLAT]]3141; UNROLL-NEXT: [[TMP7:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[TMP6]])3142; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP3]], [[N_VEC]]3143; UNROLL-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]3144; UNROLL: scalar.ph:3145; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[DOTPR_I]], [[ENTRY:%.*]] ]3146; UNROLL-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ [[C_PROMOTED_I]], [[ENTRY]] ]3147; UNROLL-NEXT: br label [[COND_END_I:%.*]]3148; UNROLL: cond.end.i:3149; UNROLL-NEXT: [[INC4_I:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC_I:%.*]], [[COND_END_I]] ]3150; UNROLL-NEXT: [[AND_I:%.*]] = and i32 [[BC_MERGE_RDX]], [[TMP0]]3151; UNROLL-NEXT: [[INC_I]] = add i8 [[INC4_I]], 13152; UNROLL-NEXT: [[TOBOOL_I:%.*]] = icmp eq i8 [[INC_I]], 03153; UNROLL-NEXT: br i1 [[TOBOOL_I]], label [[LOOPEXIT]], label [[COND_END_I]], !llvm.loop [[LOOP31:![0-9]+]]3154; UNROLL: loopexit:3155; UNROLL-NEXT: [[AND_I_LCSSA:%.*]] = phi i32 [ [[AND_I]], [[COND_END_I]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]3156; UNROLL-NEXT: ret i32 [[AND_I_LCSSA]]3157;3158; UNROLL-NO-IC-LABEL: @testoverflowcheck(3159; UNROLL-NO-IC-NEXT: entry:3160; UNROLL-NO-IC-NEXT: [[DOTPR_I:%.*]] = load i8, ptr @e, align 13161; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = load i32, ptr @d, align 43162; UNROLL-NO-IC-NEXT: [[C_PROMOTED_I:%.*]] = load i32, ptr @c, align 43163; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = sub i8 -1, [[DOTPR_I]]3164; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = zext i8 [[TMP1]] to i323165; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = add nuw nsw i32 [[TMP2]], 13166; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP3]], 43167; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]3168; UNROLL-NO-IC: vector.ph:3169; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP3]], 43170; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP3]], [[N_MOD_VF]]3171; UNROLL-NO-IC-NEXT: [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i83172; UNROLL-NO-IC-NEXT: [[IND_END:%.*]] = add i8 [[DOTPR_I]], [[DOTCAST]]3173; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = insertelement <2 x i32> splat (i32 -1), i32 [[C_PROMOTED_I]], i32 03174; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP0]], i64 03175; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer3176; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]3177; UNROLL-NO-IC: vector.body:3178; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]3179; UNROLL-NO-IC-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP4]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]3180; UNROLL-NO-IC-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i32> [ splat (i32 -1), [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]3181; UNROLL-NO-IC-NEXT: [[TMP5]] = and <2 x i32> [[BROADCAST_SPLAT]], [[VEC_PHI]]3182; UNROLL-NO-IC-NEXT: [[TMP6]] = and <2 x i32> [[BROADCAST_SPLAT]], [[VEC_PHI1]]3183; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 43184; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]3185; UNROLL-NO-IC-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]3186; UNROLL-NO-IC: middle.block:3187; UNROLL-NO-IC-NEXT: [[BIN_RDX:%.*]] = and <2 x i32> [[TMP6]], [[TMP5]]3188; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[BIN_RDX]])3189; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP3]], [[N_VEC]]3190; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]3191; UNROLL-NO-IC: scalar.ph:3192; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[DOTPR_I]], [[ENTRY:%.*]] ]3193; UNROLL-NO-IC-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP8]], [[MIDDLE_BLOCK]] ], [ [[C_PROMOTED_I]], [[ENTRY]] ]3194; UNROLL-NO-IC-NEXT: br label [[COND_END_I:%.*]]3195; UNROLL-NO-IC: cond.end.i:3196; UNROLL-NO-IC-NEXT: [[INC4_I:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC_I:%.*]], [[COND_END_I]] ]3197; UNROLL-NO-IC-NEXT: [[AND3_I:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[AND_I:%.*]], [[COND_END_I]] ]3198; UNROLL-NO-IC-NEXT: [[AND_I]] = and i32 [[TMP0]], [[AND3_I]]3199; UNROLL-NO-IC-NEXT: [[INC_I]] = add i8 [[INC4_I]], 13200; UNROLL-NO-IC-NEXT: [[TOBOOL_I:%.*]] = icmp eq i8 [[INC_I]], 03201; UNROLL-NO-IC-NEXT: br i1 [[TOBOOL_I]], label [[LOOPEXIT]], label [[COND_END_I]], !llvm.loop [[LOOP31:![0-9]+]]3202; UNROLL-NO-IC: loopexit:3203; UNROLL-NO-IC-NEXT: [[AND_I_LCSSA:%.*]] = phi i32 [ [[AND_I]], [[COND_END_I]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]3204; UNROLL-NO-IC-NEXT: ret i32 [[AND_I_LCSSA]]3205;3206; INTERLEAVE-LABEL: @testoverflowcheck(3207; INTERLEAVE-NEXT: entry:3208; INTERLEAVE-NEXT: [[DOTPR_I:%.*]] = load i8, ptr @e, align 13209; INTERLEAVE-NEXT: [[TMP0:%.*]] = load i32, ptr @d, align 43210; INTERLEAVE-NEXT: [[C_PROMOTED_I:%.*]] = load i32, ptr @c, align 43211; INTERLEAVE-NEXT: [[TMP1:%.*]] = xor i8 [[DOTPR_I]], -13212; INTERLEAVE-NEXT: [[TMP2:%.*]] = zext i8 [[TMP1]] to i323213; INTERLEAVE-NEXT: [[TMP3:%.*]] = add nuw nsw i32 [[TMP2]], 13214; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ugt i8 [[DOTPR_I]], -83215; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]3216; INTERLEAVE: vector.ph:3217; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i32 [[TMP3]], 5043218; INTERLEAVE-NEXT: [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i83219; INTERLEAVE-NEXT: [[IND_END:%.*]] = add i8 [[DOTPR_I]], [[DOTCAST]]3220; INTERLEAVE-NEXT: [[TMP4:%.*]] = insertelement <4 x i32> <i32 poison, i32 -1, i32 -1, i32 -1>, i32 [[C_PROMOTED_I]], i64 03221; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TMP0]], i64 03222; INTERLEAVE-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer3223; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]3224; INTERLEAVE: vector.body:3225; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]3226; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 83227; INTERLEAVE-NEXT: [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]3228; INTERLEAVE-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]3229; INTERLEAVE: middle.block:3230; INTERLEAVE-NEXT: [[TMP6:%.*]] = and <4 x i32> [[TMP4]], [[BROADCAST_SPLAT]]3231; INTERLEAVE-NEXT: [[TMP7:%.*]] = call i32 @llvm.vector.reduce.and.v4i32(<4 x i32> [[TMP6]])3232; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP3]], [[N_VEC]]3233; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]3234; INTERLEAVE: scalar.ph:3235; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[DOTPR_I]], [[ENTRY:%.*]] ]3236; INTERLEAVE-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ [[C_PROMOTED_I]], [[ENTRY]] ]3237; INTERLEAVE-NEXT: br label [[COND_END_I:%.*]]3238; INTERLEAVE: cond.end.i:3239; INTERLEAVE-NEXT: [[INC4_I:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC_I:%.*]], [[COND_END_I]] ]3240; INTERLEAVE-NEXT: [[AND_I:%.*]] = and i32 [[BC_MERGE_RDX]], [[TMP0]]3241; INTERLEAVE-NEXT: [[INC_I]] = add i8 [[INC4_I]], 13242; INTERLEAVE-NEXT: [[TOBOOL_I:%.*]] = icmp eq i8 [[INC_I]], 03243; INTERLEAVE-NEXT: br i1 [[TOBOOL_I]], label [[LOOPEXIT]], label [[COND_END_I]], !llvm.loop [[LOOP31:![0-9]+]]3244; INTERLEAVE: loopexit:3245; INTERLEAVE-NEXT: [[AND_I_LCSSA:%.*]] = phi i32 [ [[AND_I]], [[COND_END_I]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]3246; INTERLEAVE-NEXT: ret i32 [[AND_I_LCSSA]]3247;3248entry:3249 %.pr.i = load i8, ptr @e, align 13250 %0 = load i32, ptr @d, align 43251 %c.promoted.i = load i32, ptr @c, align 43252 br label %cond.end.i3253 3254cond.end.i:3255 %inc4.i = phi i8 [ %.pr.i, %entry ], [ %inc.i, %cond.end.i ]3256 %and3.i = phi i32 [ %c.promoted.i, %entry ], [ %and.i, %cond.end.i ]3257 %and.i = and i32 %0, %and3.i3258 %inc.i = add i8 %inc4.i, 13259 %tobool.i = icmp eq i8 %inc.i, 03260 br i1 %tobool.i, label %loopexit, label %cond.end.i3261 3262loopexit:3263 ret i32 %and.i3264}3265 3266; The SCEV expression of %sphi is (zext i8 {%t,+,1}<%loop> to i32)3267; In order to recognize %sphi as an induction PHI and vectorize this loop,3268; we need to convert the SCEV expression into an AddRecExpr.3269; The expression gets converted to {zext i8 %t to i32,+,1}.3270 3271define void @wrappingindvars1(i8 %t, i32 %len, ptr %A) {3272; CHECK-LABEL: @wrappingindvars1(3273; CHECK-NEXT: entry:3274; CHECK-NEXT: [[ST:%.*]] = zext i8 [[T:%.*]] to i163275; CHECK-NEXT: [[EXT:%.*]] = zext i8 [[T]] to i323276; CHECK-NEXT: [[ECMP:%.*]] = icmp ult i16 [[ST]], 423277; CHECK-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]3278; CHECK: loop.preheader:3279; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 13280; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 23281; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]3282; CHECK: vector.scevcheck:3283; CHECK-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i83284; CHECK-NEXT: [[TMP2:%.*]] = add i8 [[T]], [[TMP1]]3285; CHECK-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[T]]3286; CHECK-NEXT: [[TMP4:%.*]] = icmp ugt i32 [[LEN]], 2553287; CHECK-NEXT: [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]3288; CHECK-NEXT: [[TMP6:%.*]] = trunc i32 [[LEN]] to i83289; CHECK-NEXT: [[TMP7:%.*]] = add i8 [[T]], [[TMP6]]3290; CHECK-NEXT: [[TMP8:%.*]] = icmp slt i8 [[TMP7]], [[T]]3291; CHECK-NEXT: [[TMP9:%.*]] = icmp ugt i32 [[LEN]], 2553292; CHECK-NEXT: [[TMP10:%.*]] = or i1 [[TMP8]], [[TMP9]]3293; CHECK-NEXT: [[TMP11:%.*]] = or i1 [[TMP5]], [[TMP10]]3294; CHECK-NEXT: br i1 [[TMP11]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]3295; CHECK: vector.ph:3296; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 23297; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]3298; CHECK-NEXT: [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i83299; CHECK-NEXT: [[IND_END:%.*]] = add i8 [[T]], [[DOTCAST]]3300; CHECK-NEXT: [[IND_END2:%.*]] = add i32 [[EXT]], [[N_VEC]]3301; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[EXT]], i64 03302; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer3303; CHECK-NEXT: [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>3304; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]3305; CHECK: vector.body:3306; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]3307; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]3308; CHECK-NEXT: [[DOTCAST4:%.*]] = trunc i32 [[INDEX]] to i83309; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[T]], [[DOTCAST4]]3310; CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i8 [[OFFSET_IDX]]3311; CHECK-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP13]], align 43312; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 23313; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)3314; CHECK-NEXT: [[TMP15:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]3315; CHECK-NEXT: br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]3316; CHECK: middle.block:3317; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]3318; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]3319; CHECK: scalar.ph:3320; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]3321; CHECK-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]3322; CHECK-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[EXT]], [[LOOP_PREHEADER]] ], [ [[EXT]], [[VECTOR_SCEVCHECK]] ]3323; CHECK-NEXT: br label [[LOOP:%.*]]3324; CHECK: loop:3325; CHECK-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]3326; CHECK-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]3327; CHECK-NEXT: [[SPHI:%.*]] = phi i32 [ [[IDX_INC_EXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]3328; CHECK-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, ptr [[A]], i8 [[IDX]]3329; CHECK-NEXT: store i32 [[SPHI]], ptr [[PTR]], align 43330; CHECK-NEXT: [[IDX_INC]] = add i8 [[IDX]], 13331; CHECK-NEXT: [[IDX_INC_EXT]] = zext i8 [[IDX_INC]] to i323332; CHECK-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 13333; CHECK-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]3334; CHECK-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP33:![0-9]+]]3335; CHECK: exit.loopexit:3336; CHECK-NEXT: br label [[EXIT]]3337; CHECK: exit:3338; CHECK-NEXT: ret void3339;3340; IND-LABEL: @wrappingindvars1(3341; IND-NEXT: entry:3342; IND-NEXT: [[EXT:%.*]] = zext i8 [[T:%.*]] to i323343; IND-NEXT: [[ECMP:%.*]] = icmp ult i8 [[T]], 423344; IND-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]3345; IND: loop.preheader:3346; IND-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 13347; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 23348; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]3349; IND: vector.scevcheck:3350; IND-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i83351; IND-NEXT: [[TMP2:%.*]] = xor i8 [[T]], -13352; IND-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]3353; IND-NEXT: [[TMP4:%.*]] = trunc i32 [[LEN]] to i83354; IND-NEXT: [[TMP5:%.*]] = add i8 [[T]], [[TMP4]]3355; IND-NEXT: [[TMP6:%.*]] = icmp slt i8 [[TMP5]], [[T]]3356; IND-NEXT: [[TMP7:%.*]] = icmp ugt i32 [[LEN]], 2553357; IND-NEXT: [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]3358; IND-NEXT: [[TMP9:%.*]] = or i1 [[TMP3]], [[TMP8]]3359; IND-NEXT: br i1 [[TMP9]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]3360; IND: vector.ph:3361; IND-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], 5103362; IND-NEXT: [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i83363; IND-NEXT: [[IND_END:%.*]] = add i8 [[T]], [[DOTCAST]]3364; IND-NEXT: [[IND_END2:%.*]] = add nuw nsw i32 [[N_VEC]], [[EXT]]3365; IND-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[EXT]], i64 03366; IND-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer3367; IND-NEXT: [[INDUCTION:%.*]] = add nuw nsw <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>3368; IND-NEXT: br label [[VECTOR_BODY:%.*]]3369; IND: vector.body:3370; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]3371; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]3372; IND-NEXT: [[DOTCAST4:%.*]] = trunc i32 [[INDEX]] to i83373; IND-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[T]], [[DOTCAST4]]3374; IND-NEXT: [[TMP10:%.*]] = sext i8 [[OFFSET_IDX]] to i643375; IND-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP10]]3376; IND-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP11]], align 43377; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 23378; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)3379; IND-NEXT: [[TMP12:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]3380; IND-NEXT: br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]3381; IND: middle.block:3382; IND-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]3383; IND-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]3384; IND: scalar.ph:3385; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]3386; IND-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]3387; IND-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[EXT]], [[LOOP_PREHEADER]] ], [ [[EXT]], [[VECTOR_SCEVCHECK]] ]3388; IND-NEXT: br label [[LOOP:%.*]]3389; IND: loop:3390; IND-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]3391; IND-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]3392; IND-NEXT: [[SPHI:%.*]] = phi i32 [ [[IDX_INC_EXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]3393; IND-NEXT: [[TMP13:%.*]] = sext i8 [[IDX]] to i643394; IND-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP13]]3395; IND-NEXT: store i32 [[SPHI]], ptr [[PTR]], align 43396; IND-NEXT: [[IDX_INC]] = add i8 [[IDX]], 13397; IND-NEXT: [[IDX_INC_EXT]] = zext i8 [[IDX_INC]] to i323398; IND-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 13399; IND-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]3400; IND-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP33:![0-9]+]]3401; IND: exit.loopexit:3402; IND-NEXT: br label [[EXIT]]3403; IND: exit:3404; IND-NEXT: ret void3405;3406; UNROLL-LABEL: @wrappingindvars1(3407; UNROLL-NEXT: entry:3408; UNROLL-NEXT: [[EXT:%.*]] = zext i8 [[T:%.*]] to i323409; UNROLL-NEXT: [[ECMP:%.*]] = icmp ult i8 [[T]], 423410; UNROLL-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]3411; UNROLL: loop.preheader:3412; UNROLL-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 13413; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 43414; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]3415; UNROLL: vector.scevcheck:3416; UNROLL-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i83417; UNROLL-NEXT: [[TMP2:%.*]] = xor i8 [[T]], -13418; UNROLL-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]3419; UNROLL-NEXT: [[TMP4:%.*]] = trunc i32 [[LEN]] to i83420; UNROLL-NEXT: [[TMP5:%.*]] = add i8 [[T]], [[TMP4]]3421; UNROLL-NEXT: [[TMP6:%.*]] = icmp slt i8 [[TMP5]], [[T]]3422; UNROLL-NEXT: [[TMP7:%.*]] = icmp ugt i32 [[LEN]], 2553423; UNROLL-NEXT: [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]3424; UNROLL-NEXT: [[TMP9:%.*]] = or i1 [[TMP3]], [[TMP8]]3425; UNROLL-NEXT: br i1 [[TMP9]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]3426; UNROLL: vector.ph:3427; UNROLL-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], 5083428; UNROLL-NEXT: [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i83429; UNROLL-NEXT: [[IND_END:%.*]] = add i8 [[T]], [[DOTCAST]]3430; UNROLL-NEXT: [[IND_END2:%.*]] = add nuw nsw i32 [[N_VEC]], [[EXT]]3431; UNROLL-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[EXT]], i64 03432; UNROLL-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer3433; UNROLL-NEXT: [[INDUCTION:%.*]] = add nuw nsw <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>3434; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]3435; UNROLL: vector.body:3436; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]3437; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]3438; UNROLL-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)3439; UNROLL-NEXT: [[DOTCAST4:%.*]] = trunc i32 [[INDEX]] to i83440; UNROLL-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[T]], [[DOTCAST4]]3441; UNROLL-NEXT: [[TMP10:%.*]] = sext i8 [[OFFSET_IDX]] to i643442; UNROLL-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP10]]3443; UNROLL-NEXT: [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP11]], i64 83444; UNROLL-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP11]], align 43445; UNROLL-NEXT: store <2 x i32> [[STEP_ADD]], ptr [[TMP12]], align 43446; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 43447; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 4)3448; UNROLL-NEXT: [[TMP13:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]3449; UNROLL-NEXT: br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]3450; UNROLL: middle.block:3451; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]3452; UNROLL-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]3453; UNROLL: scalar.ph:3454; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]3455; UNROLL-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]3456; UNROLL-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[EXT]], [[LOOP_PREHEADER]] ], [ [[EXT]], [[VECTOR_SCEVCHECK]] ]3457; UNROLL-NEXT: br label [[LOOP:%.*]]3458; UNROLL: loop:3459; UNROLL-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]3460; UNROLL-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]3461; UNROLL-NEXT: [[SPHI:%.*]] = phi i32 [ [[IDX_INC_EXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]3462; UNROLL-NEXT: [[TMP14:%.*]] = sext i8 [[IDX]] to i643463; UNROLL-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP14]]3464; UNROLL-NEXT: store i32 [[SPHI]], ptr [[PTR]], align 43465; UNROLL-NEXT: [[IDX_INC]] = add i8 [[IDX]], 13466; UNROLL-NEXT: [[IDX_INC_EXT]] = zext i8 [[IDX_INC]] to i323467; UNROLL-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 13468; UNROLL-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]3469; UNROLL-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP33:![0-9]+]]3470; UNROLL: exit.loopexit:3471; UNROLL-NEXT: br label [[EXIT]]3472; UNROLL: exit:3473; UNROLL-NEXT: ret void3474;3475; UNROLL-NO-IC-LABEL: @wrappingindvars1(3476; UNROLL-NO-IC-NEXT: entry:3477; UNROLL-NO-IC-NEXT: [[ST:%.*]] = zext i8 [[T:%.*]] to i163478; UNROLL-NO-IC-NEXT: [[EXT:%.*]] = zext i8 [[T]] to i323479; UNROLL-NO-IC-NEXT: [[ECMP:%.*]] = icmp ult i16 [[ST]], 423480; UNROLL-NO-IC-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]3481; UNROLL-NO-IC: loop.preheader:3482; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 13483; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 43484; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]3485; UNROLL-NO-IC: vector.scevcheck:3486; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i83487; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add i8 [[T]], [[TMP1]]3488; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[T]]3489; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = icmp ugt i32 [[LEN]], 2553490; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]3491; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = trunc i32 [[LEN]] to i83492; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = add i8 [[T]], [[TMP6]]3493; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = icmp slt i8 [[TMP7]], [[T]]3494; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = icmp ugt i32 [[LEN]], 2553495; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = or i1 [[TMP8]], [[TMP9]]3496; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = or i1 [[TMP5]], [[TMP10]]3497; UNROLL-NO-IC-NEXT: br i1 [[TMP11]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]3498; UNROLL-NO-IC: vector.ph:3499; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 43500; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]3501; UNROLL-NO-IC-NEXT: [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i83502; UNROLL-NO-IC-NEXT: [[IND_END:%.*]] = add i8 [[T]], [[DOTCAST]]3503; UNROLL-NO-IC-NEXT: [[IND_END2:%.*]] = add i32 [[EXT]], [[N_VEC]]3504; UNROLL-NO-IC-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[EXT]], i64 03505; UNROLL-NO-IC-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer3506; UNROLL-NO-IC-NEXT: [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>3507; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]3508; UNROLL-NO-IC: vector.body:3509; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]3510; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]3511; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)3512; UNROLL-NO-IC-NEXT: [[DOTCAST4:%.*]] = trunc i32 [[INDEX]] to i83513; UNROLL-NO-IC-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[T]], [[DOTCAST4]]3514; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i8 [[OFFSET_IDX]]3515; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i64 23516; UNROLL-NO-IC-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP13]], align 43517; UNROLL-NO-IC-NEXT: store <2 x i32> [[STEP_ADD]], ptr [[TMP15]], align 43518; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 43519; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], splat (i32 2)3520; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]3521; UNROLL-NO-IC-NEXT: br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]3522; UNROLL-NO-IC: middle.block:3523; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]3524; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]3525; UNROLL-NO-IC: scalar.ph:3526; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]3527; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]3528; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[EXT]], [[LOOP_PREHEADER]] ], [ [[EXT]], [[VECTOR_SCEVCHECK]] ]3529; UNROLL-NO-IC-NEXT: br label [[LOOP:%.*]]3530; UNROLL-NO-IC: loop:3531; UNROLL-NO-IC-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]3532; UNROLL-NO-IC-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]3533; UNROLL-NO-IC-NEXT: [[SPHI:%.*]] = phi i32 [ [[IDX_INC_EXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]3534; UNROLL-NO-IC-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, ptr [[A]], i8 [[IDX]]3535; UNROLL-NO-IC-NEXT: store i32 [[SPHI]], ptr [[PTR]], align 43536; UNROLL-NO-IC-NEXT: [[IDX_INC]] = add i8 [[IDX]], 13537; UNROLL-NO-IC-NEXT: [[IDX_INC_EXT]] = zext i8 [[IDX_INC]] to i323538; UNROLL-NO-IC-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 13539; UNROLL-NO-IC-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]3540; UNROLL-NO-IC-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP33:![0-9]+]]3541; UNROLL-NO-IC: exit.loopexit:3542; UNROLL-NO-IC-NEXT: br label [[EXIT]]3543; UNROLL-NO-IC: exit:3544; UNROLL-NO-IC-NEXT: ret void3545;3546; INTERLEAVE-LABEL: @wrappingindvars1(3547; INTERLEAVE-NEXT: entry:3548; INTERLEAVE-NEXT: [[EXT:%.*]] = zext i8 [[T:%.*]] to i323549; INTERLEAVE-NEXT: [[ECMP:%.*]] = icmp ult i8 [[T]], 423550; INTERLEAVE-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]3551; INTERLEAVE: loop.preheader:3552; INTERLEAVE-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 13553; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 83554; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]3555; INTERLEAVE: vector.scevcheck:3556; INTERLEAVE-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i83557; INTERLEAVE-NEXT: [[TMP2:%.*]] = xor i8 [[T]], -13558; INTERLEAVE-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]3559; INTERLEAVE-NEXT: [[TMP4:%.*]] = trunc i32 [[LEN]] to i83560; INTERLEAVE-NEXT: [[TMP5:%.*]] = add i8 [[T]], [[TMP4]]3561; INTERLEAVE-NEXT: [[TMP6:%.*]] = icmp slt i8 [[TMP5]], [[T]]3562; INTERLEAVE-NEXT: [[TMP7:%.*]] = icmp ugt i32 [[LEN]], 2553563; INTERLEAVE-NEXT: [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]3564; INTERLEAVE-NEXT: [[TMP9:%.*]] = or i1 [[TMP3]], [[TMP8]]3565; INTERLEAVE-NEXT: br i1 [[TMP9]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]3566; INTERLEAVE: vector.ph:3567; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], 5043568; INTERLEAVE-NEXT: [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i83569; INTERLEAVE-NEXT: [[IND_END:%.*]] = add i8 [[T]], [[DOTCAST]]3570; INTERLEAVE-NEXT: [[IND_END2:%.*]] = add nuw nsw i32 [[N_VEC]], [[EXT]]3571; INTERLEAVE-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[EXT]], i64 03572; INTERLEAVE-NEXT: [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer3573; INTERLEAVE-NEXT: [[INDUCTION:%.*]] = add nuw nsw <4 x i32> [[DOTSPLAT]], <i32 0, i32 1, i32 2, i32 3>3574; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]3575; INTERLEAVE: vector.body:3576; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]3577; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]3578; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], splat (i32 4)3579; INTERLEAVE-NEXT: [[DOTCAST4:%.*]] = trunc i32 [[INDEX]] to i83580; INTERLEAVE-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[T]], [[DOTCAST4]]3581; INTERLEAVE-NEXT: [[TMP10:%.*]] = sext i8 [[OFFSET_IDX]] to i643582; INTERLEAVE-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP10]]3583; INTERLEAVE-NEXT: [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP11]], i64 163584; INTERLEAVE-NEXT: store <4 x i32> [[VEC_IND]], ptr [[TMP11]], align 43585; INTERLEAVE-NEXT: store <4 x i32> [[STEP_ADD]], ptr [[TMP12]], align 43586; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 83587; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 8)3588; INTERLEAVE-NEXT: [[TMP13:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]3589; INTERLEAVE-NEXT: br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]3590; INTERLEAVE: middle.block:3591; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]3592; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]3593; INTERLEAVE: scalar.ph:3594; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]3595; INTERLEAVE-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]3596; INTERLEAVE-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[EXT]], [[LOOP_PREHEADER]] ], [ [[EXT]], [[VECTOR_SCEVCHECK]] ]3597; INTERLEAVE-NEXT: br label [[LOOP:%.*]]3598; INTERLEAVE: loop:3599; INTERLEAVE-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]3600; INTERLEAVE-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]3601; INTERLEAVE-NEXT: [[SPHI:%.*]] = phi i32 [ [[IDX_INC_EXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]3602; INTERLEAVE-NEXT: [[TMP14:%.*]] = sext i8 [[IDX]] to i643603; INTERLEAVE-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP14]]3604; INTERLEAVE-NEXT: store i32 [[SPHI]], ptr [[PTR]], align 43605; INTERLEAVE-NEXT: [[IDX_INC]] = add i8 [[IDX]], 13606; INTERLEAVE-NEXT: [[IDX_INC_EXT]] = zext i8 [[IDX_INC]] to i323607; INTERLEAVE-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 13608; INTERLEAVE-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]3609; INTERLEAVE-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP33:![0-9]+]]3610; INTERLEAVE: exit.loopexit:3611; INTERLEAVE-NEXT: br label [[EXIT]]3612; INTERLEAVE: exit:3613; INTERLEAVE-NEXT: ret void3614;3615 entry:3616 %st = zext i8 %t to i163617 %ext = zext i8 %t to i323618 %ecmp = icmp ult i16 %st, 423619 br i1 %ecmp, label %loop, label %exit3620 3621 loop:3622 3623 %idx = phi i8 [ %t, %entry ], [ %idx.inc, %loop ]3624 %idx.b = phi i32 [ 0, %entry ], [ %idx.b.inc, %loop ]3625 %sphi = phi i32 [ %ext, %entry ], [%idx.inc.ext, %loop]3626 3627 %ptr = getelementptr inbounds i32, ptr %A, i8 %idx3628 store i32 %sphi, ptr %ptr3629 3630 %idx.inc = add i8 %idx, 13631 %idx.inc.ext = zext i8 %idx.inc to i323632 %idx.b.inc = add nuw nsw i32 %idx.b, 13633 3634 %c = icmp ult i32 %idx.b, %len3635 br i1 %c, label %loop, label %exit3636 3637 exit:3638 ret void3639}3640 3641; The SCEV expression of %sphi is (4 * (zext i8 {%t,+,1}<%loop> to i32))3642; In order to recognize %sphi as an induction PHI and vectorize this loop,3643; we need to convert the SCEV expression into an AddRecExpr.3644; The expression gets converted to ({4 * (zext %t to i32),+,4}).3645define void @wrappingindvars2(i8 %t, i32 %len, ptr %A) {3646; CHECK-LABEL: @wrappingindvars2(3647; CHECK-NEXT: entry:3648; CHECK-NEXT: [[ST:%.*]] = zext i8 [[T:%.*]] to i163649; CHECK-NEXT: [[EXT:%.*]] = zext i8 [[T]] to i323650; CHECK-NEXT: [[EXT_MUL:%.*]] = mul i32 [[EXT]], 43651; CHECK-NEXT: [[ECMP:%.*]] = icmp ult i16 [[ST]], 423652; CHECK-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]3653; CHECK: loop.preheader:3654; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 13655; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 23656; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]3657; CHECK: vector.scevcheck:3658; CHECK-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i83659; CHECK-NEXT: [[TMP2:%.*]] = add i8 [[T]], [[TMP1]]3660; CHECK-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[T]]3661; CHECK-NEXT: [[TMP4:%.*]] = icmp ugt i32 [[LEN]], 2553662; CHECK-NEXT: [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]3663; CHECK-NEXT: [[TMP6:%.*]] = trunc i32 [[LEN]] to i83664; CHECK-NEXT: [[TMP7:%.*]] = add i8 [[T]], [[TMP6]]3665; CHECK-NEXT: [[TMP8:%.*]] = icmp slt i8 [[TMP7]], [[T]]3666; CHECK-NEXT: [[TMP9:%.*]] = icmp ugt i32 [[LEN]], 2553667; CHECK-NEXT: [[TMP10:%.*]] = or i1 [[TMP8]], [[TMP9]]3668; CHECK-NEXT: [[TMP11:%.*]] = or i1 [[TMP5]], [[TMP10]]3669; CHECK-NEXT: br i1 [[TMP11]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]3670; CHECK: vector.ph:3671; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 23672; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]3673; CHECK-NEXT: [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i83674; CHECK-NEXT: [[IND_END:%.*]] = add i8 [[T]], [[DOTCAST]]3675; CHECK-NEXT: [[TMP12:%.*]] = mul i32 [[N_VEC]], 43676; CHECK-NEXT: [[IND_END1:%.*]] = add i32 [[EXT_MUL]], [[TMP12]]3677; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[EXT_MUL]], i64 03678; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer3679; CHECK-NEXT: [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 4>3680; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]3681; CHECK: vector.body:3682; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]3683; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]3684; CHECK-NEXT: [[DOTCAST4:%.*]] = trunc i32 [[INDEX]] to i83685; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[T]], [[DOTCAST4]]3686; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i8 [[OFFSET_IDX]]3687; CHECK-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP14]], align 43688; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 23689; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 8)3690; CHECK-NEXT: [[TMP16:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]3691; CHECK-NEXT: br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]3692; CHECK: middle.block:3693; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]3694; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]3695; CHECK: scalar.ph:3696; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]3697; CHECK-NEXT: [[BC_RESUME_VAL2:%.*]] = phi i32 [ [[IND_END1]], [[MIDDLE_BLOCK]] ], [ [[EXT_MUL]], [[LOOP_PREHEADER]] ], [ [[EXT_MUL]], [[VECTOR_SCEVCHECK]] ]3698; CHECK-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]3699; CHECK-NEXT: br label [[LOOP:%.*]]3700; CHECK: loop:3701; CHECK-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]3702; CHECK-NEXT: [[SPHI:%.*]] = phi i32 [ [[MUL:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ]3703; CHECK-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]3704; CHECK-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, ptr [[A]], i8 [[IDX]]3705; CHECK-NEXT: store i32 [[SPHI]], ptr [[PTR]], align 43706; CHECK-NEXT: [[IDX_INC]] = add i8 [[IDX]], 13707; CHECK-NEXT: [[IDX_INC_EXT:%.*]] = zext i8 [[IDX_INC]] to i323708; CHECK-NEXT: [[MUL]] = mul i32 [[IDX_INC_EXT]], 43709; CHECK-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 13710; CHECK-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]3711; CHECK-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP35:![0-9]+]]3712; CHECK: exit.loopexit:3713; CHECK-NEXT: br label [[EXIT]]3714; CHECK: exit:3715; CHECK-NEXT: ret void3716;3717; IND-LABEL: @wrappingindvars2(3718; IND-NEXT: entry:3719; IND-NEXT: [[EXT:%.*]] = zext i8 [[T:%.*]] to i323720; IND-NEXT: [[EXT_MUL:%.*]] = shl nuw nsw i32 [[EXT]], 23721; IND-NEXT: [[ECMP:%.*]] = icmp ult i8 [[T]], 423722; IND-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]3723; IND: loop.preheader:3724; IND-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 13725; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 23726; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]3727; IND: vector.scevcheck:3728; IND-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i83729; IND-NEXT: [[TMP2:%.*]] = xor i8 [[T]], -13730; IND-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]3731; IND-NEXT: [[TMP4:%.*]] = trunc i32 [[LEN]] to i83732; IND-NEXT: [[TMP5:%.*]] = add i8 [[T]], [[TMP4]]3733; IND-NEXT: [[TMP6:%.*]] = icmp slt i8 [[TMP5]], [[T]]3734; IND-NEXT: [[TMP7:%.*]] = icmp ugt i32 [[LEN]], 2553735; IND-NEXT: [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]3736; IND-NEXT: [[TMP9:%.*]] = or i1 [[TMP3]], [[TMP8]]3737; IND-NEXT: br i1 [[TMP9]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]3738; IND: vector.ph:3739; IND-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], 5103740; IND-NEXT: [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i83741; IND-NEXT: [[IND_END:%.*]] = add i8 [[T]], [[DOTCAST]]3742; IND-NEXT: [[EXT_MUL5:%.*]] = add nuw nsw i32 [[N_VEC]], [[EXT]]3743; IND-NEXT: [[IND_END1:%.*]] = shl nuw nsw i32 [[EXT_MUL5]], 23744; IND-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[EXT_MUL]], i64 03745; IND-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer3746; IND-NEXT: [[INDUCTION:%.*]] = add nuw nsw <2 x i32> [[DOTSPLAT]], <i32 0, i32 4>3747; IND-NEXT: br label [[VECTOR_BODY:%.*]]3748; IND: vector.body:3749; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]3750; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]3751; IND-NEXT: [[DOTCAST4:%.*]] = trunc i32 [[INDEX]] to i83752; IND-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[T]], [[DOTCAST4]]3753; IND-NEXT: [[TMP10:%.*]] = sext i8 [[OFFSET_IDX]] to i643754; IND-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP10]]3755; IND-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP11]], align 43756; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 23757; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 8)3758; IND-NEXT: [[TMP12:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]3759; IND-NEXT: br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]3760; IND: middle.block:3761; IND-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]3762; IND-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]3763; IND: scalar.ph:3764; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]3765; IND-NEXT: [[BC_RESUME_VAL2:%.*]] = phi i32 [ [[IND_END1]], [[MIDDLE_BLOCK]] ], [ [[EXT_MUL]], [[LOOP_PREHEADER]] ], [ [[EXT_MUL]], [[VECTOR_SCEVCHECK]] ]3766; IND-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]3767; IND-NEXT: br label [[LOOP:%.*]]3768; IND: loop:3769; IND-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]3770; IND-NEXT: [[SPHI:%.*]] = phi i32 [ [[MUL:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ]3771; IND-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]3772; IND-NEXT: [[TMP13:%.*]] = sext i8 [[IDX]] to i643773; IND-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP13]]3774; IND-NEXT: store i32 [[SPHI]], ptr [[PTR]], align 43775; IND-NEXT: [[IDX_INC]] = add i8 [[IDX]], 13776; IND-NEXT: [[IDX_INC_EXT:%.*]] = zext i8 [[IDX_INC]] to i323777; IND-NEXT: [[MUL]] = shl nuw nsw i32 [[IDX_INC_EXT]], 23778; IND-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 13779; IND-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]3780; IND-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP35:![0-9]+]]3781; IND: exit.loopexit:3782; IND-NEXT: br label [[EXIT]]3783; IND: exit:3784; IND-NEXT: ret void3785;3786; UNROLL-LABEL: @wrappingindvars2(3787; UNROLL-NEXT: entry:3788; UNROLL-NEXT: [[EXT:%.*]] = zext i8 [[T:%.*]] to i323789; UNROLL-NEXT: [[EXT_MUL:%.*]] = shl nuw nsw i32 [[EXT]], 23790; UNROLL-NEXT: [[ECMP:%.*]] = icmp ult i8 [[T]], 423791; UNROLL-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]3792; UNROLL: loop.preheader:3793; UNROLL-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 13794; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 43795; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]3796; UNROLL: vector.scevcheck:3797; UNROLL-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i83798; UNROLL-NEXT: [[TMP2:%.*]] = xor i8 [[T]], -13799; UNROLL-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]3800; UNROLL-NEXT: [[TMP4:%.*]] = trunc i32 [[LEN]] to i83801; UNROLL-NEXT: [[TMP5:%.*]] = add i8 [[T]], [[TMP4]]3802; UNROLL-NEXT: [[TMP6:%.*]] = icmp slt i8 [[TMP5]], [[T]]3803; UNROLL-NEXT: [[TMP7:%.*]] = icmp ugt i32 [[LEN]], 2553804; UNROLL-NEXT: [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]3805; UNROLL-NEXT: [[TMP9:%.*]] = or i1 [[TMP3]], [[TMP8]]3806; UNROLL-NEXT: br i1 [[TMP9]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]3807; UNROLL: vector.ph:3808; UNROLL-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], 5083809; UNROLL-NEXT: [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i83810; UNROLL-NEXT: [[IND_END:%.*]] = add i8 [[T]], [[DOTCAST]]3811; UNROLL-NEXT: [[EXT_MUL5:%.*]] = add nuw nsw i32 [[N_VEC]], [[EXT]]3812; UNROLL-NEXT: [[IND_END1:%.*]] = shl nuw nsw i32 [[EXT_MUL5]], 23813; UNROLL-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[EXT_MUL]], i64 03814; UNROLL-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer3815; UNROLL-NEXT: [[INDUCTION:%.*]] = add nuw nsw <2 x i32> [[DOTSPLAT]], <i32 0, i32 4>3816; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]3817; UNROLL: vector.body:3818; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]3819; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]3820; UNROLL-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 8)3821; UNROLL-NEXT: [[DOTCAST4:%.*]] = trunc i32 [[INDEX]] to i83822; UNROLL-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[T]], [[DOTCAST4]]3823; UNROLL-NEXT: [[TMP10:%.*]] = sext i8 [[OFFSET_IDX]] to i643824; UNROLL-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP10]]3825; UNROLL-NEXT: [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP11]], i64 83826; UNROLL-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP11]], align 43827; UNROLL-NEXT: store <2 x i32> [[STEP_ADD]], ptr [[TMP12]], align 43828; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 43829; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 16)3830; UNROLL-NEXT: [[TMP13:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]3831; UNROLL-NEXT: br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]3832; UNROLL: middle.block:3833; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]3834; UNROLL-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]3835; UNROLL: scalar.ph:3836; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]3837; UNROLL-NEXT: [[BC_RESUME_VAL2:%.*]] = phi i32 [ [[IND_END1]], [[MIDDLE_BLOCK]] ], [ [[EXT_MUL]], [[LOOP_PREHEADER]] ], [ [[EXT_MUL]], [[VECTOR_SCEVCHECK]] ]3838; UNROLL-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]3839; UNROLL-NEXT: br label [[LOOP:%.*]]3840; UNROLL: loop:3841; UNROLL-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]3842; UNROLL-NEXT: [[SPHI:%.*]] = phi i32 [ [[MUL:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ]3843; UNROLL-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]3844; UNROLL-NEXT: [[TMP14:%.*]] = sext i8 [[IDX]] to i643845; UNROLL-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP14]]3846; UNROLL-NEXT: store i32 [[SPHI]], ptr [[PTR]], align 43847; UNROLL-NEXT: [[IDX_INC]] = add i8 [[IDX]], 13848; UNROLL-NEXT: [[IDX_INC_EXT:%.*]] = zext i8 [[IDX_INC]] to i323849; UNROLL-NEXT: [[MUL]] = shl nuw nsw i32 [[IDX_INC_EXT]], 23850; UNROLL-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 13851; UNROLL-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]3852; UNROLL-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP35:![0-9]+]]3853; UNROLL: exit.loopexit:3854; UNROLL-NEXT: br label [[EXIT]]3855; UNROLL: exit:3856; UNROLL-NEXT: ret void3857;3858; UNROLL-NO-IC-LABEL: @wrappingindvars2(3859; UNROLL-NO-IC-NEXT: entry:3860; UNROLL-NO-IC-NEXT: [[ST:%.*]] = zext i8 [[T:%.*]] to i163861; UNROLL-NO-IC-NEXT: [[EXT:%.*]] = zext i8 [[T]] to i323862; UNROLL-NO-IC-NEXT: [[EXT_MUL:%.*]] = mul i32 [[EXT]], 43863; UNROLL-NO-IC-NEXT: [[ECMP:%.*]] = icmp ult i16 [[ST]], 423864; UNROLL-NO-IC-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]3865; UNROLL-NO-IC: loop.preheader:3866; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 13867; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 43868; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]3869; UNROLL-NO-IC: vector.scevcheck:3870; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i83871; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add i8 [[T]], [[TMP1]]3872; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[T]]3873; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = icmp ugt i32 [[LEN]], 2553874; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]3875; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = trunc i32 [[LEN]] to i83876; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = add i8 [[T]], [[TMP6]]3877; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = icmp slt i8 [[TMP7]], [[T]]3878; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = icmp ugt i32 [[LEN]], 2553879; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = or i1 [[TMP8]], [[TMP9]]3880; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = or i1 [[TMP5]], [[TMP10]]3881; UNROLL-NO-IC-NEXT: br i1 [[TMP11]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]3882; UNROLL-NO-IC: vector.ph:3883; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 43884; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]3885; UNROLL-NO-IC-NEXT: [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i83886; UNROLL-NO-IC-NEXT: [[IND_END:%.*]] = add i8 [[T]], [[DOTCAST]]3887; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = mul i32 [[N_VEC]], 43888; UNROLL-NO-IC-NEXT: [[IND_END1:%.*]] = add i32 [[EXT_MUL]], [[TMP12]]3889; UNROLL-NO-IC-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[EXT_MUL]], i64 03890; UNROLL-NO-IC-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer3891; UNROLL-NO-IC-NEXT: [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 4>3892; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]3893; UNROLL-NO-IC: vector.body:3894; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]3895; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]3896; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 8)3897; UNROLL-NO-IC-NEXT: [[DOTCAST4:%.*]] = trunc i32 [[INDEX]] to i83898; UNROLL-NO-IC-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[T]], [[DOTCAST4]]3899; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i8 [[OFFSET_IDX]]3900; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i64 23901; UNROLL-NO-IC-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP14]], align 43902; UNROLL-NO-IC-NEXT: store <2 x i32> [[STEP_ADD]], ptr [[TMP16]], align 43903; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 43904; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], splat (i32 8)3905; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]3906; UNROLL-NO-IC-NEXT: br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]3907; UNROLL-NO-IC: middle.block:3908; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]3909; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]3910; UNROLL-NO-IC: scalar.ph:3911; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]3912; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL2:%.*]] = phi i32 [ [[IND_END1]], [[MIDDLE_BLOCK]] ], [ [[EXT_MUL]], [[LOOP_PREHEADER]] ], [ [[EXT_MUL]], [[VECTOR_SCEVCHECK]] ]3913; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]3914; UNROLL-NO-IC-NEXT: br label [[LOOP:%.*]]3915; UNROLL-NO-IC: loop:3916; UNROLL-NO-IC-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]3917; UNROLL-NO-IC-NEXT: [[SPHI:%.*]] = phi i32 [ [[MUL:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ]3918; UNROLL-NO-IC-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]3919; UNROLL-NO-IC-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, ptr [[A]], i8 [[IDX]]3920; UNROLL-NO-IC-NEXT: store i32 [[SPHI]], ptr [[PTR]], align 43921; UNROLL-NO-IC-NEXT: [[IDX_INC]] = add i8 [[IDX]], 13922; UNROLL-NO-IC-NEXT: [[IDX_INC_EXT:%.*]] = zext i8 [[IDX_INC]] to i323923; UNROLL-NO-IC-NEXT: [[MUL]] = mul i32 [[IDX_INC_EXT]], 43924; UNROLL-NO-IC-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 13925; UNROLL-NO-IC-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]3926; UNROLL-NO-IC-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP35:![0-9]+]]3927; UNROLL-NO-IC: exit.loopexit:3928; UNROLL-NO-IC-NEXT: br label [[EXIT]]3929; UNROLL-NO-IC: exit:3930; UNROLL-NO-IC-NEXT: ret void3931;3932; INTERLEAVE-LABEL: @wrappingindvars2(3933; INTERLEAVE-NEXT: entry:3934; INTERLEAVE-NEXT: [[EXT:%.*]] = zext i8 [[T:%.*]] to i323935; INTERLEAVE-NEXT: [[EXT_MUL:%.*]] = shl nuw nsw i32 [[EXT]], 23936; INTERLEAVE-NEXT: [[ECMP:%.*]] = icmp ult i8 [[T]], 423937; INTERLEAVE-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]3938; INTERLEAVE: loop.preheader:3939; INTERLEAVE-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 13940; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 83941; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]3942; INTERLEAVE: vector.scevcheck:3943; INTERLEAVE-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i83944; INTERLEAVE-NEXT: [[TMP2:%.*]] = xor i8 [[T]], -13945; INTERLEAVE-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]3946; INTERLEAVE-NEXT: [[TMP4:%.*]] = trunc i32 [[LEN]] to i83947; INTERLEAVE-NEXT: [[TMP5:%.*]] = add i8 [[T]], [[TMP4]]3948; INTERLEAVE-NEXT: [[TMP6:%.*]] = icmp slt i8 [[TMP5]], [[T]]3949; INTERLEAVE-NEXT: [[TMP7:%.*]] = icmp ugt i32 [[LEN]], 2553950; INTERLEAVE-NEXT: [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]3951; INTERLEAVE-NEXT: [[TMP9:%.*]] = or i1 [[TMP3]], [[TMP8]]3952; INTERLEAVE-NEXT: br i1 [[TMP9]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]3953; INTERLEAVE: vector.ph:3954; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], 5043955; INTERLEAVE-NEXT: [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i83956; INTERLEAVE-NEXT: [[IND_END:%.*]] = add i8 [[T]], [[DOTCAST]]3957; INTERLEAVE-NEXT: [[EXT_MUL5:%.*]] = add nuw nsw i32 [[N_VEC]], [[EXT]]3958; INTERLEAVE-NEXT: [[IND_END1:%.*]] = shl nuw nsw i32 [[EXT_MUL5]], 23959; INTERLEAVE-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[EXT_MUL]], i64 03960; INTERLEAVE-NEXT: [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer3961; INTERLEAVE-NEXT: [[INDUCTION:%.*]] = add nuw nsw <4 x i32> [[DOTSPLAT]], <i32 0, i32 4, i32 8, i32 12>3962; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]3963; INTERLEAVE: vector.body:3964; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]3965; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]3966; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], splat (i32 16)3967; INTERLEAVE-NEXT: [[DOTCAST4:%.*]] = trunc i32 [[INDEX]] to i83968; INTERLEAVE-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[T]], [[DOTCAST4]]3969; INTERLEAVE-NEXT: [[TMP10:%.*]] = sext i8 [[OFFSET_IDX]] to i643970; INTERLEAVE-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP10]]3971; INTERLEAVE-NEXT: [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP11]], i64 163972; INTERLEAVE-NEXT: store <4 x i32> [[VEC_IND]], ptr [[TMP11]], align 43973; INTERLEAVE-NEXT: store <4 x i32> [[STEP_ADD]], ptr [[TMP12]], align 43974; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 83975; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 32)3976; INTERLEAVE-NEXT: [[TMP13:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]3977; INTERLEAVE-NEXT: br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]3978; INTERLEAVE: middle.block:3979; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]3980; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]3981; INTERLEAVE: scalar.ph:3982; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]3983; INTERLEAVE-NEXT: [[BC_RESUME_VAL2:%.*]] = phi i32 [ [[IND_END1]], [[MIDDLE_BLOCK]] ], [ [[EXT_MUL]], [[LOOP_PREHEADER]] ], [ [[EXT_MUL]], [[VECTOR_SCEVCHECK]] ]3984; INTERLEAVE-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]3985; INTERLEAVE-NEXT: br label [[LOOP:%.*]]3986; INTERLEAVE: loop:3987; INTERLEAVE-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]3988; INTERLEAVE-NEXT: [[SPHI:%.*]] = phi i32 [ [[MUL:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ]3989; INTERLEAVE-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]3990; INTERLEAVE-NEXT: [[TMP14:%.*]] = sext i8 [[IDX]] to i643991; INTERLEAVE-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP14]]3992; INTERLEAVE-NEXT: store i32 [[SPHI]], ptr [[PTR]], align 43993; INTERLEAVE-NEXT: [[IDX_INC]] = add i8 [[IDX]], 13994; INTERLEAVE-NEXT: [[IDX_INC_EXT:%.*]] = zext i8 [[IDX_INC]] to i323995; INTERLEAVE-NEXT: [[MUL]] = shl nuw nsw i32 [[IDX_INC_EXT]], 23996; INTERLEAVE-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 13997; INTERLEAVE-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]3998; INTERLEAVE-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP35:![0-9]+]]3999; INTERLEAVE: exit.loopexit:4000; INTERLEAVE-NEXT: br label [[EXIT]]4001; INTERLEAVE: exit:4002; INTERLEAVE-NEXT: ret void4003;4004entry:4005 %st = zext i8 %t to i164006 %ext = zext i8 %t to i324007 %ext.mul = mul i32 %ext, 44008 4009 %ecmp = icmp ult i16 %st, 424010 br i1 %ecmp, label %loop, label %exit4011 4012 loop:4013 4014 %idx = phi i8 [ %t, %entry ], [ %idx.inc, %loop ]4015 %sphi = phi i32 [ %ext.mul, %entry ], [%mul, %loop]4016 %idx.b = phi i32 [ 0, %entry ], [ %idx.b.inc, %loop ]4017 4018 %ptr = getelementptr inbounds i32, ptr %A, i8 %idx4019 store i32 %sphi, ptr %ptr4020 4021 %idx.inc = add i8 %idx, 14022 %idx.inc.ext = zext i8 %idx.inc to i324023 %mul = mul i32 %idx.inc.ext, 44024 %idx.b.inc = add nuw nsw i32 %idx.b, 14025 4026 %c = icmp ult i32 %idx.b, %len4027 br i1 %c, label %loop, label %exit4028 4029 exit:4030 ret void4031}4032 4033; Check that we generate vectorized IVs in the pre-header4034; instead of widening the scalar IV inside the loop, when4035; we know how to do that.4036define void @veciv(ptr nocapture %a, i32 %start, i32 %k) {4037; CHECK-LABEL: @veciv(4038; CHECK-NEXT: for.body.preheader:4039; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K:%.*]], 24040; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]4041; CHECK: vector.ph:4042; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[K]], 24043; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[K]], [[N_MOD_VF]]4044; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]4045; CHECK: vector.body:4046; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4047; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4048; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]4049; CHECK-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP1]], align 44050; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 24051; CHECK-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <2 x i32> [[VEC_IND]], splat (i32 2)4052; CHECK-NEXT: [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]4053; CHECK-NEXT: br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]4054; CHECK: middle.block:4055; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[K]], [[N_VEC]]4056; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]4057; CHECK: scalar.ph:4058; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]4059; CHECK-NEXT: br label [[FOR_BODY:%.*]]4060; CHECK: for.body:4061; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4062; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[INDVARS_IV]]4063; CHECK-NEXT: store i32 [[INDVARS_IV]], ptr [[ARRAYIDX]], align 44064; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 14065; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]4066; CHECK-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP37:![0-9]+]]4067; CHECK: exit:4068; CHECK-NEXT: ret void4069;4070; IND-LABEL: @veciv(4071; IND-NEXT: for.body.preheader:4072; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K:%.*]], 24073; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]4074; IND: vector.ph:4075; IND-NEXT: [[N_VEC:%.*]] = and i32 [[K]], -24076; IND-NEXT: br label [[VECTOR_BODY:%.*]]4077; IND: vector.body:4078; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4079; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4080; IND-NEXT: [[TMP0:%.*]] = sext i32 [[INDEX]] to i644081; IND-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP0]]4082; IND-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP1]], align 44083; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 24084; IND-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <2 x i32> [[VEC_IND]], splat (i32 2)4085; IND-NEXT: [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]4086; IND-NEXT: br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]4087; IND: middle.block:4088; IND-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[K]], [[N_VEC]]4089; IND-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]4090; IND: scalar.ph:4091; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]4092; IND-NEXT: br label [[FOR_BODY:%.*]]4093; IND: for.body:4094; IND-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4095; IND-NEXT: [[TMP3:%.*]] = sext i32 [[INDVARS_IV]] to i644096; IND-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP3]]4097; IND-NEXT: store i32 [[INDVARS_IV]], ptr [[ARRAYIDX]], align 44098; IND-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 14099; IND-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]4100; IND-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP37:![0-9]+]]4101; IND: exit:4102; IND-NEXT: ret void4103;4104; UNROLL-LABEL: @veciv(4105; UNROLL-NEXT: for.body.preheader:4106; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K:%.*]], 44107; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]4108; UNROLL: vector.ph:4109; UNROLL-NEXT: [[N_VEC:%.*]] = and i32 [[K]], -44110; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]4111; UNROLL: vector.body:4112; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4113; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4114; UNROLL-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)4115; UNROLL-NEXT: [[TMP0:%.*]] = sext i32 [[INDEX]] to i644116; UNROLL-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP0]]4117; UNROLL-NEXT: [[TMP2:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i64 84118; UNROLL-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP1]], align 44119; UNROLL-NEXT: store <2 x i32> [[STEP_ADD]], ptr [[TMP2]], align 44120; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 44121; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 4)4122; UNROLL-NEXT: [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]4123; UNROLL-NEXT: br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]4124; UNROLL: middle.block:4125; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[K]], [[N_VEC]]4126; UNROLL-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]4127; UNROLL: scalar.ph:4128; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]4129; UNROLL-NEXT: br label [[FOR_BODY:%.*]]4130; UNROLL: for.body:4131; UNROLL-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4132; UNROLL-NEXT: [[TMP4:%.*]] = sext i32 [[INDVARS_IV]] to i644133; UNROLL-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP4]]4134; UNROLL-NEXT: store i32 [[INDVARS_IV]], ptr [[ARRAYIDX]], align 44135; UNROLL-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 14136; UNROLL-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]4137; UNROLL-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP37:![0-9]+]]4138; UNROLL: exit:4139; UNROLL-NEXT: ret void4140;4141; UNROLL-NO-IC-LABEL: @veciv(4142; UNROLL-NO-IC-NEXT: for.body.preheader:4143; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K:%.*]], 44144; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]4145; UNROLL-NO-IC: vector.ph:4146; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[K]], 44147; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i32 [[K]], [[N_MOD_VF]]4148; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]4149; UNROLL-NO-IC: vector.body:4150; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4151; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4152; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)4153; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]4154; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 24155; UNROLL-NO-IC-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP1]], align 44156; UNROLL-NO-IC-NEXT: store <2 x i32> [[STEP_ADD]], ptr [[TMP3]], align 44157; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 44158; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <2 x i32> [[STEP_ADD]], splat (i32 2)4159; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]4160; UNROLL-NO-IC-NEXT: br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]4161; UNROLL-NO-IC: middle.block:4162; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[K]], [[N_VEC]]4163; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]4164; UNROLL-NO-IC: scalar.ph:4165; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]4166; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]4167; UNROLL-NO-IC: for.body:4168; UNROLL-NO-IC-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4169; UNROLL-NO-IC-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[INDVARS_IV]]4170; UNROLL-NO-IC-NEXT: store i32 [[INDVARS_IV]], ptr [[ARRAYIDX]], align 44171; UNROLL-NO-IC-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 14172; UNROLL-NO-IC-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]4173; UNROLL-NO-IC-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP37:![0-9]+]]4174; UNROLL-NO-IC: exit:4175; UNROLL-NO-IC-NEXT: ret void4176;4177; INTERLEAVE-LABEL: @veciv(4178; INTERLEAVE-NEXT: for.body.preheader:4179; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K:%.*]], 84180; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]4181; INTERLEAVE: vector.ph:4182; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i32 [[K]], -84183; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]4184; INTERLEAVE: vector.body:4185; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4186; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4187; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], splat (i32 4)4188; INTERLEAVE-NEXT: [[TMP0:%.*]] = sext i32 [[INDEX]] to i644189; INTERLEAVE-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP0]]4190; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i64 164191; INTERLEAVE-NEXT: store <4 x i32> [[VEC_IND]], ptr [[TMP1]], align 44192; INTERLEAVE-NEXT: store <4 x i32> [[STEP_ADD]], ptr [[TMP2]], align 44193; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 84194; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 8)4195; INTERLEAVE-NEXT: [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]4196; INTERLEAVE-NEXT: br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]4197; INTERLEAVE: middle.block:4198; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[K]], [[N_VEC]]4199; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]4200; INTERLEAVE: scalar.ph:4201; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]4202; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]4203; INTERLEAVE: for.body:4204; INTERLEAVE-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4205; INTERLEAVE-NEXT: [[TMP4:%.*]] = sext i32 [[INDVARS_IV]] to i644206; INTERLEAVE-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP4]]4207; INTERLEAVE-NEXT: store i32 [[INDVARS_IV]], ptr [[ARRAYIDX]], align 44208; INTERLEAVE-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 14209; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]4210; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP37:![0-9]+]]4211; INTERLEAVE: exit:4212; INTERLEAVE-NEXT: ret void4213;4214for.body.preheader:4215 br label %for.body4216 4217for.body:4218 %indvars.iv = phi i32 [ %indvars.iv.next, %for.body ], [ 0, %for.body.preheader ]4219 %arrayidx = getelementptr inbounds i32, ptr %a, i32 %indvars.iv4220 store i32 %indvars.iv, ptr %arrayidx, align 44221 %indvars.iv.next = add nuw nsw i32 %indvars.iv, 14222 %exitcond = icmp eq i32 %indvars.iv.next, %k4223 br i1 %exitcond, label %exit, label %for.body4224 4225exit:4226 ret void4227}4228 4229define void @trunciv(ptr nocapture %a, i32 %start, i64 %k) {4230; CHECK-LABEL: @trunciv(4231; CHECK-NEXT: for.body.preheader:4232; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[K:%.*]], 24233; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]4234; CHECK: vector.scevcheck:4235; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[K]], -14236; CHECK-NEXT: [[TMP1:%.*]] = trunc i64 [[TMP0]] to i324237; CHECK-NEXT: [[TMP2:%.*]] = icmp slt i32 [[TMP1]], 04238; CHECK-NEXT: [[TMP3:%.*]] = icmp ugt i64 [[TMP0]], 42949672954239; CHECK-NEXT: [[TMP4:%.*]] = or i1 [[TMP2]], [[TMP3]]4240; CHECK-NEXT: br i1 [[TMP4]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]4241; CHECK: vector.ph:4242; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[K]], 24243; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[K]], [[N_MOD_VF]]4244; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]4245; CHECK: vector.body:4246; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4247; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4248; CHECK-NEXT: [[TMP5:%.*]] = trunc i64 [[INDEX]] to i324249; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[TMP5]]4250; CHECK-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP7]], align 44251; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 24252; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)4253; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]4254; CHECK-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]4255; CHECK: middle.block:4256; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[K]], [[N_VEC]]4257; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]4258; CHECK: scalar.ph:4259; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]4260; CHECK-NEXT: br label [[FOR_BODY:%.*]]4261; CHECK: for.body:4262; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4263; CHECK-NEXT: [[TRUNC_IV:%.*]] = trunc i64 [[INDVARS_IV]] to i324264; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[TRUNC_IV]]4265; CHECK-NEXT: store i32 [[TRUNC_IV]], ptr [[ARRAYIDX]], align 44266; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 14267; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[K]]4268; CHECK-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP39:![0-9]+]]4269; CHECK: exit:4270; CHECK-NEXT: ret void4271;4272; IND-LABEL: @trunciv(4273; IND-NEXT: for.body.preheader:4274; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[K:%.*]], 24275; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]4276; IND: vector.scevcheck:4277; IND-NEXT: [[DOTNOT:%.*]] = icmp ult i64 [[K]], 21474836494278; IND-NEXT: br i1 [[DOTNOT]], label [[VECTOR_PH:%.*]], label [[SCALAR_PH]]4279; IND: vector.ph:4280; IND-NEXT: [[N_VEC:%.*]] = and i64 [[K]], 42949672944281; IND-NEXT: br label [[VECTOR_BODY:%.*]]4282; IND: vector.body:4283; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4284; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4285; IND-NEXT: [[SEXT:%.*]] = shl i64 [[INDEX]], 324286; IND-NEXT: [[TMP0:%.*]] = ashr exact i64 [[SEXT]], 304287; IND-NEXT: [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[TMP0]]4288; IND-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP1]], align 44289; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 24290; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)4291; IND-NEXT: [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]4292; IND-NEXT: br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]4293; IND: middle.block:4294; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[K]], [[N_VEC]]4295; IND-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]4296; IND: scalar.ph:4297; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]4298; IND-NEXT: br label [[FOR_BODY:%.*]]4299; IND: for.body:4300; IND-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4301; IND-NEXT: [[TRUNC_IV:%.*]] = trunc i64 [[INDVARS_IV]] to i324302; IND-NEXT: [[SEXT1:%.*]] = shl i64 [[INDVARS_IV]], 324303; IND-NEXT: [[TMP3:%.*]] = ashr exact i64 [[SEXT1]], 304304; IND-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[TMP3]]4305; IND-NEXT: store i32 [[TRUNC_IV]], ptr [[ARRAYIDX]], align 44306; IND-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 14307; IND-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[K]]4308; IND-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP39:![0-9]+]]4309; IND: exit:4310; IND-NEXT: ret void4311;4312; UNROLL-LABEL: @trunciv(4313; UNROLL-NEXT: for.body.preheader:4314; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[K:%.*]], 44315; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]4316; UNROLL: vector.scevcheck:4317; UNROLL-NEXT: [[DOTNOT:%.*]] = icmp ult i64 [[K]], 21474836494318; UNROLL-NEXT: br i1 [[DOTNOT]], label [[VECTOR_PH:%.*]], label [[SCALAR_PH]]4319; UNROLL: vector.ph:4320; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[K]], 42949672924321; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]4322; UNROLL: vector.body:4323; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4324; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4325; UNROLL-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)4326; UNROLL-NEXT: [[SEXT:%.*]] = shl i64 [[INDEX]], 324327; UNROLL-NEXT: [[TMP0:%.*]] = ashr exact i64 [[SEXT]], 304328; UNROLL-NEXT: [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[TMP0]]4329; UNROLL-NEXT: [[TMP2:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i64 84330; UNROLL-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP1]], align 44331; UNROLL-NEXT: store <2 x i32> [[STEP_ADD]], ptr [[TMP2]], align 44332; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 44333; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 4)4334; UNROLL-NEXT: [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]4335; UNROLL-NEXT: br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]4336; UNROLL: middle.block:4337; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[K]], [[N_VEC]]4338; UNROLL-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]4339; UNROLL: scalar.ph:4340; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]4341; UNROLL-NEXT: br label [[FOR_BODY:%.*]]4342; UNROLL: for.body:4343; UNROLL-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4344; UNROLL-NEXT: [[TRUNC_IV:%.*]] = trunc i64 [[INDVARS_IV]] to i324345; UNROLL-NEXT: [[SEXT1:%.*]] = shl i64 [[INDVARS_IV]], 324346; UNROLL-NEXT: [[TMP4:%.*]] = ashr exact i64 [[SEXT1]], 304347; UNROLL-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[TMP4]]4348; UNROLL-NEXT: store i32 [[TRUNC_IV]], ptr [[ARRAYIDX]], align 44349; UNROLL-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 14350; UNROLL-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[K]]4351; UNROLL-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP39:![0-9]+]]4352; UNROLL: exit:4353; UNROLL-NEXT: ret void4354;4355; UNROLL-NO-IC-LABEL: @trunciv(4356; UNROLL-NO-IC-NEXT: for.body.preheader:4357; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[K:%.*]], 44358; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]4359; UNROLL-NO-IC: vector.scevcheck:4360; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i64 [[K]], -14361; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = trunc i64 [[TMP0]] to i324362; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = icmp slt i32 [[TMP1]], 04363; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = icmp ugt i64 [[TMP0]], 42949672954364; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = or i1 [[TMP2]], [[TMP3]]4365; UNROLL-NO-IC-NEXT: br i1 [[TMP4]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]4366; UNROLL-NO-IC: vector.ph:4367; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[K]], 44368; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[K]], [[N_MOD_VF]]4369; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]4370; UNROLL-NO-IC: vector.body:4371; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4372; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4373; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)4374; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = trunc i64 [[INDEX]] to i324375; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[TMP5]]4376; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i64 24377; UNROLL-NO-IC-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP7]], align 44378; UNROLL-NO-IC-NEXT: store <2 x i32> [[STEP_ADD]], ptr [[TMP9]], align 44379; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 44380; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], splat (i32 2)4381; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]4382; UNROLL-NO-IC-NEXT: br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]4383; UNROLL-NO-IC: middle.block:4384; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[K]], [[N_VEC]]4385; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]4386; UNROLL-NO-IC: scalar.ph:4387; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]4388; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]4389; UNROLL-NO-IC: for.body:4390; UNROLL-NO-IC-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4391; UNROLL-NO-IC-NEXT: [[TRUNC_IV:%.*]] = trunc i64 [[INDVARS_IV]] to i324392; UNROLL-NO-IC-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[TRUNC_IV]]4393; UNROLL-NO-IC-NEXT: store i32 [[TRUNC_IV]], ptr [[ARRAYIDX]], align 44394; UNROLL-NO-IC-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 14395; UNROLL-NO-IC-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[K]]4396; UNROLL-NO-IC-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP39:![0-9]+]]4397; UNROLL-NO-IC: exit:4398; UNROLL-NO-IC-NEXT: ret void4399;4400; INTERLEAVE-LABEL: @trunciv(4401; INTERLEAVE-NEXT: for.body.preheader:4402; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[K:%.*]], 84403; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]4404; INTERLEAVE: vector.scevcheck:4405; INTERLEAVE-NEXT: [[DOTNOT:%.*]] = icmp ult i64 [[K]], 21474836494406; INTERLEAVE-NEXT: br i1 [[DOTNOT]], label [[VECTOR_PH:%.*]], label [[SCALAR_PH]]4407; INTERLEAVE: vector.ph:4408; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i64 [[K]], 42949672884409; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]4410; INTERLEAVE: vector.body:4411; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4412; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4413; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], splat (i32 4)4414; INTERLEAVE-NEXT: [[SEXT:%.*]] = shl i64 [[INDEX]], 324415; INTERLEAVE-NEXT: [[TMP0:%.*]] = ashr exact i64 [[SEXT]], 304416; INTERLEAVE-NEXT: [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[TMP0]]4417; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i64 164418; INTERLEAVE-NEXT: store <4 x i32> [[VEC_IND]], ptr [[TMP1]], align 44419; INTERLEAVE-NEXT: store <4 x i32> [[STEP_ADD]], ptr [[TMP2]], align 44420; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 84421; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 8)4422; INTERLEAVE-NEXT: [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]4423; INTERLEAVE-NEXT: br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]4424; INTERLEAVE: middle.block:4425; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[K]], [[N_VEC]]4426; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]4427; INTERLEAVE: scalar.ph:4428; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]4429; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]4430; INTERLEAVE: for.body:4431; INTERLEAVE-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4432; INTERLEAVE-NEXT: [[TRUNC_IV:%.*]] = trunc i64 [[INDVARS_IV]] to i324433; INTERLEAVE-NEXT: [[SEXT1:%.*]] = shl i64 [[INDVARS_IV]], 324434; INTERLEAVE-NEXT: [[TMP4:%.*]] = ashr exact i64 [[SEXT1]], 304435; INTERLEAVE-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[TMP4]]4436; INTERLEAVE-NEXT: store i32 [[TRUNC_IV]], ptr [[ARRAYIDX]], align 44437; INTERLEAVE-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 14438; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[K]]4439; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP39:![0-9]+]]4440; INTERLEAVE: exit:4441; INTERLEAVE-NEXT: ret void4442;4443for.body.preheader:4444 br label %for.body4445 4446for.body:4447 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %for.body.preheader ]4448 %trunc.iv = trunc i64 %indvars.iv to i324449 %arrayidx = getelementptr inbounds i32, ptr %a, i32 %trunc.iv4450 store i32 %trunc.iv, ptr %arrayidx, align 44451 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 14452 %exitcond = icmp eq i64 %indvars.iv.next, %k4453 br i1 %exitcond, label %exit, label %for.body4454 4455exit:4456 ret void4457}4458 4459;4460;4461define void @nonprimary(ptr nocapture %a, i32 %start, i32 %i, i32 %k) {4462; CHECK-LABEL: @nonprimary(4463; CHECK-NEXT: for.body.preheader:4464; CHECK-NEXT: [[TMP0:%.*]] = sub i32 [[K:%.*]], [[I:%.*]]4465; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 24466; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]4467; CHECK: vector.ph:4468; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 24469; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]4470; CHECK-NEXT: [[IND_END:%.*]] = add i32 [[I]], [[N_VEC]]4471; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[I]], i64 04472; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer4473; CHECK-NEXT: [[INDUCTION:%.*]] = add nuw nsw <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>4474; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]4475; CHECK: vector.body:4476; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4477; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4478; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i32 [[I]], [[INDEX]]4479; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[OFFSET_IDX]]4480; CHECK-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP2]], align 44481; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 24482; CHECK-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <2 x i32> [[VEC_IND]], splat (i32 2)4483; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]4484; CHECK-NEXT: br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]4485; CHECK: middle.block:4486; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]4487; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]4488; CHECK: scalar.ph:4489; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[I]], [[FOR_BODY_PREHEADER:%.*]] ]4490; CHECK-NEXT: br label [[FOR_BODY:%.*]]4491; CHECK: for.body:4492; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4493; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[INDVARS_IV]]4494; CHECK-NEXT: store i32 [[INDVARS_IV]], ptr [[ARRAYIDX]], align 44495; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 14496; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]4497; CHECK-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]4498; CHECK: exit:4499; CHECK-NEXT: ret void4500;4501; IND-LABEL: @nonprimary(4502; IND-NEXT: for.body.preheader:4503; IND-NEXT: [[TMP0:%.*]] = sub i32 [[K:%.*]], [[I:%.*]]4504; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 24505; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]4506; IND: vector.ph:4507; IND-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], -24508; IND-NEXT: [[IND_END:%.*]] = add i32 [[I]], [[N_VEC]]4509; IND-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[I]], i64 04510; IND-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer4511; IND-NEXT: [[INDUCTION:%.*]] = add nuw nsw <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>4512; IND-NEXT: br label [[VECTOR_BODY:%.*]]4513; IND: vector.body:4514; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4515; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4516; IND-NEXT: [[OFFSET_IDX:%.*]] = add i32 [[I]], [[INDEX]]4517; IND-NEXT: [[TMP1:%.*]] = sext i32 [[OFFSET_IDX]] to i644518; IND-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP1]]4519; IND-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP2]], align 44520; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 24521; IND-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <2 x i32> [[VEC_IND]], splat (i32 2)4522; IND-NEXT: [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]4523; IND-NEXT: br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]4524; IND: middle.block:4525; IND-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]4526; IND-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]4527; IND: scalar.ph:4528; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[I]], [[FOR_BODY_PREHEADER:%.*]] ]4529; IND-NEXT: br label [[FOR_BODY:%.*]]4530; IND: for.body:4531; IND-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4532; IND-NEXT: [[TMP4:%.*]] = sext i32 [[INDVARS_IV]] to i644533; IND-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP4]]4534; IND-NEXT: store i32 [[INDVARS_IV]], ptr [[ARRAYIDX]], align 44535; IND-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 14536; IND-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]4537; IND-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]4538; IND: exit:4539; IND-NEXT: ret void4540;4541; UNROLL-LABEL: @nonprimary(4542; UNROLL-NEXT: for.body.preheader:4543; UNROLL-NEXT: [[TMP0:%.*]] = sub i32 [[K:%.*]], [[I:%.*]]4544; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 44545; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]4546; UNROLL: vector.ph:4547; UNROLL-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], -44548; UNROLL-NEXT: [[IND_END:%.*]] = add i32 [[I]], [[N_VEC]]4549; UNROLL-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[I]], i64 04550; UNROLL-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer4551; UNROLL-NEXT: [[INDUCTION:%.*]] = add nuw nsw <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>4552; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]4553; UNROLL: vector.body:4554; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4555; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4556; UNROLL-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)4557; UNROLL-NEXT: [[OFFSET_IDX:%.*]] = add i32 [[I]], [[INDEX]]4558; UNROLL-NEXT: [[TMP1:%.*]] = sext i32 [[OFFSET_IDX]] to i644559; UNROLL-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP1]]4560; UNROLL-NEXT: [[TMP3:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP2]], i64 84561; UNROLL-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP2]], align 44562; UNROLL-NEXT: store <2 x i32> [[STEP_ADD]], ptr [[TMP3]], align 44563; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 44564; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 4)4565; UNROLL-NEXT: [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]4566; UNROLL-NEXT: br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]4567; UNROLL: middle.block:4568; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]4569; UNROLL-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]4570; UNROLL: scalar.ph:4571; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[I]], [[FOR_BODY_PREHEADER:%.*]] ]4572; UNROLL-NEXT: br label [[FOR_BODY:%.*]]4573; UNROLL: for.body:4574; UNROLL-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4575; UNROLL-NEXT: [[TMP5:%.*]] = sext i32 [[INDVARS_IV]] to i644576; UNROLL-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP5]]4577; UNROLL-NEXT: store i32 [[INDVARS_IV]], ptr [[ARRAYIDX]], align 44578; UNROLL-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 14579; UNROLL-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]4580; UNROLL-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]4581; UNROLL: exit:4582; UNROLL-NEXT: ret void4583;4584; UNROLL-NO-IC-LABEL: @nonprimary(4585; UNROLL-NO-IC-NEXT: for.body.preheader:4586; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = sub i32 [[K:%.*]], [[I:%.*]]4587; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 44588; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]4589; UNROLL-NO-IC: vector.ph:4590; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 44591; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]4592; UNROLL-NO-IC-NEXT: [[IND_END:%.*]] = add i32 [[I]], [[N_VEC]]4593; UNROLL-NO-IC-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[I]], i64 04594; UNROLL-NO-IC-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer4595; UNROLL-NO-IC-NEXT: [[INDUCTION:%.*]] = add nuw nsw <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>4596; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]4597; UNROLL-NO-IC: vector.body:4598; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4599; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4600; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)4601; UNROLL-NO-IC-NEXT: [[OFFSET_IDX:%.*]] = add i32 [[I]], [[INDEX]]4602; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[OFFSET_IDX]]4603; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i64 24604; UNROLL-NO-IC-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP2]], align 44605; UNROLL-NO-IC-NEXT: store <2 x i32> [[STEP_ADD]], ptr [[TMP4]], align 44606; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 44607; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <2 x i32> [[STEP_ADD]], splat (i32 2)4608; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]4609; UNROLL-NO-IC-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]4610; UNROLL-NO-IC: middle.block:4611; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]4612; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]4613; UNROLL-NO-IC: scalar.ph:4614; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[I]], [[FOR_BODY_PREHEADER:%.*]] ]4615; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]4616; UNROLL-NO-IC: for.body:4617; UNROLL-NO-IC-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4618; UNROLL-NO-IC-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[INDVARS_IV]]4619; UNROLL-NO-IC-NEXT: store i32 [[INDVARS_IV]], ptr [[ARRAYIDX]], align 44620; UNROLL-NO-IC-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 14621; UNROLL-NO-IC-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]4622; UNROLL-NO-IC-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]4623; UNROLL-NO-IC: exit:4624; UNROLL-NO-IC-NEXT: ret void4625;4626; INTERLEAVE-LABEL: @nonprimary(4627; INTERLEAVE-NEXT: for.body.preheader:4628; INTERLEAVE-NEXT: [[TMP0:%.*]] = sub i32 [[K:%.*]], [[I:%.*]]4629; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 84630; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]4631; INTERLEAVE: vector.ph:4632; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], -84633; INTERLEAVE-NEXT: [[IND_END:%.*]] = add i32 [[I]], [[N_VEC]]4634; INTERLEAVE-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[I]], i64 04635; INTERLEAVE-NEXT: [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer4636; INTERLEAVE-NEXT: [[INDUCTION:%.*]] = add nuw nsw <4 x i32> [[DOTSPLAT]], <i32 0, i32 1, i32 2, i32 3>4637; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]4638; INTERLEAVE: vector.body:4639; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4640; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4641; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], splat (i32 4)4642; INTERLEAVE-NEXT: [[OFFSET_IDX:%.*]] = add i32 [[I]], [[INDEX]]4643; INTERLEAVE-NEXT: [[TMP1:%.*]] = sext i32 [[OFFSET_IDX]] to i644644; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP1]]4645; INTERLEAVE-NEXT: [[TMP3:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP2]], i64 164646; INTERLEAVE-NEXT: store <4 x i32> [[VEC_IND]], ptr [[TMP2]], align 44647; INTERLEAVE-NEXT: store <4 x i32> [[STEP_ADD]], ptr [[TMP3]], align 44648; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 84649; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 8)4650; INTERLEAVE-NEXT: [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]4651; INTERLEAVE-NEXT: br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]4652; INTERLEAVE: middle.block:4653; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]4654; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]4655; INTERLEAVE: scalar.ph:4656; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[I]], [[FOR_BODY_PREHEADER:%.*]] ]4657; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]4658; INTERLEAVE: for.body:4659; INTERLEAVE-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4660; INTERLEAVE-NEXT: [[TMP5:%.*]] = sext i32 [[INDVARS_IV]] to i644661; INTERLEAVE-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP5]]4662; INTERLEAVE-NEXT: store i32 [[INDVARS_IV]], ptr [[ARRAYIDX]], align 44663; INTERLEAVE-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 14664; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]4665; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]4666; INTERLEAVE: exit:4667; INTERLEAVE-NEXT: ret void4668;4669for.body.preheader:4670 br label %for.body4671 4672for.body:4673 %indvars.iv = phi i32 [ %indvars.iv.next, %for.body ], [ %i, %for.body.preheader ]4674 %arrayidx = getelementptr inbounds i32, ptr %a, i32 %indvars.iv4675 store i32 %indvars.iv, ptr %arrayidx, align 44676 %indvars.iv.next = add nuw nsw i32 %indvars.iv, 14677 %exitcond = icmp eq i32 %indvars.iv.next, %k4678 br i1 %exitcond, label %exit, label %for.body4679 4680exit:4681 ret void4682}4683 4684define void @non_primary_iv_trunc(ptr %a, i64 %n) {4685; CHECK-LABEL: @non_primary_iv_trunc(4686; CHECK-NEXT: entry:4687; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)4688; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 24689; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]4690; CHECK: vector.ph:4691; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 24692; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]4693; CHECK-NEXT: [[IND_END:%.*]] = mul i64 [[N_VEC]], 24694; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]4695; CHECK: vector.body:4696; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4697; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4698; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]4699; CHECK-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP1]], align 44700; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 24701; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 4)4702; CHECK-NEXT: [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]4703; CHECK-NEXT: br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]4704; CHECK: middle.block:4705; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]4706; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]4707; CHECK: scalar.ph:4708; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]4709; CHECK-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]4710; CHECK-NEXT: br label [[FOR_BODY:%.*]]4711; CHECK: for.body:4712; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4713; CHECK-NEXT: [[J:%.*]] = phi i64 [ [[J_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]4714; CHECK-NEXT: [[VAR0:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[I]]4715; CHECK-NEXT: [[VAR1:%.*]] = trunc i64 [[J]] to i324716; CHECK-NEXT: store i32 [[VAR1]], ptr [[VAR0]], align 44717; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 14718; CHECK-NEXT: [[J_NEXT]] = add nuw nsw i64 [[J]], 24719; CHECK-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]4720; CHECK-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP43:![0-9]+]]4721; CHECK: for.end:4722; CHECK-NEXT: ret void4723;4724; IND-LABEL: @non_primary_iv_trunc(4725; IND-NEXT: entry:4726; IND-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)4727; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i64 [[N]], 24728; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]4729; IND: vector.ph:4730; IND-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 92233720368547758064731; IND-NEXT: [[IND_END:%.*]] = shl nuw i64 [[N_VEC]], 14732; IND-NEXT: br label [[VECTOR_BODY:%.*]]4733; IND: vector.body:4734; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4735; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4736; IND-NEXT: [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]4737; IND-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP0]], align 44738; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 24739; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 4)4740; IND-NEXT: [[TMP1:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]4741; IND-NEXT: br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]4742; IND: middle.block:4743; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]4744; IND-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]4745; IND: scalar.ph:4746; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]4747; IND-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]4748; IND-NEXT: br label [[FOR_BODY:%.*]]4749; IND: for.body:4750; IND-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4751; IND-NEXT: [[J:%.*]] = phi i64 [ [[J_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]4752; IND-NEXT: [[VAR0:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[I]]4753; IND-NEXT: [[VAR1:%.*]] = trunc i64 [[J]] to i324754; IND-NEXT: store i32 [[VAR1]], ptr [[VAR0]], align 44755; IND-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 14756; IND-NEXT: [[J_NEXT]] = add nuw nsw i64 [[J]], 24757; IND-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]4758; IND-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP43:![0-9]+]]4759; IND: for.end:4760; IND-NEXT: ret void4761;4762; UNROLL-LABEL: @non_primary_iv_trunc(4763; UNROLL-NEXT: entry:4764; UNROLL-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)4765; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i64 [[N]], 44766; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]4767; UNROLL: vector.ph:4768; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 92233720368547758044769; UNROLL-NEXT: [[IND_END:%.*]] = shl nuw i64 [[N_VEC]], 14770; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]4771; UNROLL: vector.body:4772; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4773; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4774; UNROLL-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 4)4775; UNROLL-NEXT: [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]4776; UNROLL-NEXT: [[TMP1:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 84777; UNROLL-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP0]], align 44778; UNROLL-NEXT: store <2 x i32> [[STEP_ADD]], ptr [[TMP1]], align 44779; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 44780; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 8)4781; UNROLL-NEXT: [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]4782; UNROLL-NEXT: br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]4783; UNROLL: middle.block:4784; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]4785; UNROLL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]4786; UNROLL: scalar.ph:4787; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]4788; UNROLL-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]4789; UNROLL-NEXT: br label [[FOR_BODY:%.*]]4790; UNROLL: for.body:4791; UNROLL-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4792; UNROLL-NEXT: [[J:%.*]] = phi i64 [ [[J_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]4793; UNROLL-NEXT: [[VAR0:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[I]]4794; UNROLL-NEXT: [[VAR1:%.*]] = trunc i64 [[J]] to i324795; UNROLL-NEXT: store i32 [[VAR1]], ptr [[VAR0]], align 44796; UNROLL-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 14797; UNROLL-NEXT: [[J_NEXT]] = add nuw nsw i64 [[J]], 24798; UNROLL-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]4799; UNROLL-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP43:![0-9]+]]4800; UNROLL: for.end:4801; UNROLL-NEXT: ret void4802;4803; UNROLL-NO-IC-LABEL: @non_primary_iv_trunc(4804; UNROLL-NO-IC-NEXT: entry:4805; UNROLL-NO-IC-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)4806; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 44807; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]4808; UNROLL-NO-IC: vector.ph:4809; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 44810; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]4811; UNROLL-NO-IC-NEXT: [[IND_END:%.*]] = mul i64 [[N_VEC]], 24812; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]4813; UNROLL-NO-IC: vector.body:4814; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4815; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4816; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 4)4817; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]4818; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i64 24819; UNROLL-NO-IC-NEXT: store <2 x i32> [[VEC_IND]], ptr [[TMP1]], align 44820; UNROLL-NO-IC-NEXT: store <2 x i32> [[STEP_ADD]], ptr [[TMP3]], align 44821; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 44822; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], splat (i32 4)4823; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]4824; UNROLL-NO-IC-NEXT: br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]4825; UNROLL-NO-IC: middle.block:4826; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]4827; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]4828; UNROLL-NO-IC: scalar.ph:4829; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]4830; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]4831; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]4832; UNROLL-NO-IC: for.body:4833; UNROLL-NO-IC-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4834; UNROLL-NO-IC-NEXT: [[J:%.*]] = phi i64 [ [[J_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]4835; UNROLL-NO-IC-NEXT: [[VAR0:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[I]]4836; UNROLL-NO-IC-NEXT: [[VAR1:%.*]] = trunc i64 [[J]] to i324837; UNROLL-NO-IC-NEXT: store i32 [[VAR1]], ptr [[VAR0]], align 44838; UNROLL-NO-IC-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 14839; UNROLL-NO-IC-NEXT: [[J_NEXT]] = add nuw nsw i64 [[J]], 24840; UNROLL-NO-IC-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]4841; UNROLL-NO-IC-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP43:![0-9]+]]4842; UNROLL-NO-IC: for.end:4843; UNROLL-NO-IC-NEXT: ret void4844;4845; INTERLEAVE-LABEL: @non_primary_iv_trunc(4846; INTERLEAVE-NEXT: entry:4847; INTERLEAVE-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)4848; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp slt i64 [[N]], 84849; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]4850; INTERLEAVE: vector.ph:4851; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 92233720368547758004852; INTERLEAVE-NEXT: [[IND_END:%.*]] = shl nuw i64 [[N_VEC]], 14853; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]4854; INTERLEAVE: vector.body:4855; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]4856; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 2, i32 4, i32 6>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]4857; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], splat (i32 8)4858; INTERLEAVE-NEXT: [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]4859; INTERLEAVE-NEXT: [[TMP1:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 164860; INTERLEAVE-NEXT: store <4 x i32> [[VEC_IND]], ptr [[TMP0]], align 44861; INTERLEAVE-NEXT: store <4 x i32> [[STEP_ADD]], ptr [[TMP1]], align 44862; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 84863; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 16)4864; INTERLEAVE-NEXT: [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]4865; INTERLEAVE-NEXT: br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]4866; INTERLEAVE: middle.block:4867; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]4868; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]4869; INTERLEAVE: scalar.ph:4870; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]4871; INTERLEAVE-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]4872; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]4873; INTERLEAVE: for.body:4874; INTERLEAVE-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]4875; INTERLEAVE-NEXT: [[J:%.*]] = phi i64 [ [[J_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]4876; INTERLEAVE-NEXT: [[VAR0:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[I]]4877; INTERLEAVE-NEXT: [[VAR1:%.*]] = trunc i64 [[J]] to i324878; INTERLEAVE-NEXT: store i32 [[VAR1]], ptr [[VAR0]], align 44879; INTERLEAVE-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 14880; INTERLEAVE-NEXT: [[J_NEXT]] = add nuw nsw i64 [[J]], 24881; INTERLEAVE-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]4882; INTERLEAVE-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP43:![0-9]+]]4883; INTERLEAVE: for.end:4884; INTERLEAVE-NEXT: ret void4885;4886entry:4887 br label %for.body4888 4889for.body:4890 %i = phi i64 [ %i.next, %for.body ], [ 0, %entry ]4891 %j = phi i64 [ %j.next, %for.body ], [ 0, %entry ]4892 %var0 = getelementptr inbounds i32, ptr %a, i64 %i4893 %var1 = trunc i64 %j to i324894 store i32 %var1, ptr %var0, align 44895 %i.next = add nuw nsw i64 %i, 14896 %j.next = add nuw nsw i64 %j, 24897 %cond = icmp slt i64 %i.next, %n4898 br i1 %cond, label %for.body, label %for.end4899 4900for.end:4901 ret void4902}4903 4904; PR32419. Ensure we transform truncated non-primary induction variables. In4905; the test case below we replace %var1 with a new induction variable. Because4906; the truncated value is non-primary, we must compute an offset from the4907; primary induction variable.4908;4909;4910define i32 @PR32419(i32 %a, i16 %b) {4911; CHECK-LABEL: @PR32419(4912; CHECK-NEXT: entry:4913; CHECK-NEXT: br label [[VECTOR_PH:%.*]]4914; CHECK: vector.ph:4915; CHECK-NEXT: [[TMP0:%.*]] = insertelement <2 x i32> zeroinitializer, i32 [[A:%.*]], i32 04916; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]4917; CHECK: vector.body:4918; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UREM_CONTINUE2:%.*]] ]4919; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[PRED_UREM_CONTINUE2]] ]4920; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i16> [ <i16 -20, i16 -19>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_UREM_CONTINUE2]] ]4921; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i32 -20, [[INDEX]]4922; CHECK-NEXT: [[TMP1:%.*]] = trunc i32 [[OFFSET_IDX]] to i164923; CHECK-NEXT: [[TMP3:%.*]] = icmp ne <2 x i16> [[VEC_IND]], zeroinitializer4924; CHECK-NEXT: [[TMP4:%.*]] = extractelement <2 x i1> [[TMP3]], i32 04925; CHECK-NEXT: br i1 [[TMP4]], label [[PRED_UREM_IF:%.*]], label [[PRED_UREM_CONTINUE:%.*]]4926; CHECK: pred.urem.if:4927; CHECK-NEXT: [[TMP5:%.*]] = add i16 [[TMP1]], 04928; CHECK-NEXT: [[TMP6:%.*]] = urem i16 [[B:%.*]], [[TMP5]]4929; CHECK-NEXT: [[TMP7:%.*]] = insertelement <2 x i16> poison, i16 [[TMP6]], i32 04930; CHECK-NEXT: br label [[PRED_UREM_CONTINUE]]4931; CHECK: pred.urem.continue:4932; CHECK-NEXT: [[TMP8:%.*]] = phi <2 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP7]], [[PRED_UREM_IF]] ]4933; CHECK-NEXT: [[TMP9:%.*]] = extractelement <2 x i1> [[TMP3]], i32 14934; CHECK-NEXT: br i1 [[TMP9]], label [[PRED_UREM_IF1:%.*]], label [[PRED_UREM_CONTINUE2]]4935; CHECK: pred.urem.if1:4936; CHECK-NEXT: [[TMP10:%.*]] = add i16 [[TMP1]], 14937; CHECK-NEXT: [[TMP11:%.*]] = urem i16 [[B]], [[TMP10]]4938; CHECK-NEXT: [[TMP12:%.*]] = insertelement <2 x i16> [[TMP8]], i16 [[TMP11]], i32 14939; CHECK-NEXT: br label [[PRED_UREM_CONTINUE2]]4940; CHECK: pred.urem.continue2:4941; CHECK-NEXT: [[TMP13:%.*]] = phi <2 x i16> [ [[TMP8]], [[PRED_UREM_CONTINUE]] ], [ [[TMP12]], [[PRED_UREM_IF1]] ]4942; CHECK-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[TMP3]], <2 x i16> [[TMP13]], <2 x i16> zeroinitializer4943; CHECK-NEXT: [[TMP14:%.*]] = sext <2 x i16> [[PREDPHI]] to <2 x i32>4944; CHECK-NEXT: [[TMP15]] = or <2 x i32> [[VEC_PHI]], [[TMP14]]4945; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 24946; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i16> [[VEC_IND]], splat (i16 2)4947; CHECK-NEXT: [[TMP16:%.*]] = icmp eq i32 [[INDEX_NEXT]], 204948; CHECK-NEXT: br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]4949; CHECK: middle.block:4950; CHECK-NEXT: [[TMP17:%.*]] = call i32 @llvm.vector.reduce.or.v2i32(<2 x i32> [[TMP15]])4951; CHECK-NEXT: br label [[FOR_INC:%.*]]4952; CHECK: for.end:4953; CHECK-NEXT: ret i32 [[TMP17]]4954;4955; IND-LABEL: @PR32419(4956; IND-NEXT: entry:4957; IND-NEXT: br label [[VECTOR_PH:%.*]]4958; IND: vector.ph:4959; IND-NEXT: [[TMP0:%.*]] = insertelement <2 x i32> <i32 poison, i32 0>, i32 [[A:%.*]], i64 04960; IND-NEXT: br label [[VECTOR_BODY:%.*]]4961; IND: vector.body:4962; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UREM_CONTINUE2:%.*]] ]4963; IND-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[PRED_UREM_CONTINUE2]] ]4964; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i16> [ <i16 -20, i16 -19>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_UREM_CONTINUE2]] ]4965; IND-NEXT: [[TMP1:%.*]] = trunc i32 [[INDEX]] to i164966; IND-NEXT: [[TMP2:%.*]] = icmp ne <2 x i16> [[VEC_IND]], zeroinitializer4967; IND-NEXT: [[TMP3:%.*]] = extractelement <2 x i1> [[TMP2]], i64 04968; IND-NEXT: br i1 [[TMP3]], label [[PRED_UREM_IF:%.*]], label [[PRED_UREM_CONTINUE:%.*]]4969; IND: pred.urem.if:4970; IND-NEXT: [[TMP4:%.*]] = add i16 [[TMP1]], -204971; IND-NEXT: [[TMP5:%.*]] = urem i16 [[B:%.*]], [[TMP4]]4972; IND-NEXT: [[TMP6:%.*]] = insertelement <2 x i16> poison, i16 [[TMP5]], i64 04973; IND-NEXT: br label [[PRED_UREM_CONTINUE]]4974; IND: pred.urem.continue:4975; IND-NEXT: [[TMP7:%.*]] = phi <2 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP6]], [[PRED_UREM_IF]] ]4976; IND-NEXT: [[TMP8:%.*]] = extractelement <2 x i1> [[TMP2]], i64 14977; IND-NEXT: br i1 [[TMP8]], label [[PRED_UREM_IF1:%.*]], label [[PRED_UREM_CONTINUE2]]4978; IND: pred.urem.if1:4979; IND-NEXT: [[TMP9:%.*]] = add i16 [[TMP1]], -194980; IND-NEXT: [[TMP10:%.*]] = urem i16 [[B]], [[TMP9]]4981; IND-NEXT: [[TMP11:%.*]] = insertelement <2 x i16> [[TMP7]], i16 [[TMP10]], i64 14982; IND-NEXT: br label [[PRED_UREM_CONTINUE2]]4983; IND: pred.urem.continue2:4984; IND-NEXT: [[TMP12:%.*]] = phi <2 x i16> [ [[TMP7]], [[PRED_UREM_CONTINUE]] ], [ [[TMP11]], [[PRED_UREM_IF1]] ]4985; IND-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[TMP2]], <2 x i16> [[TMP12]], <2 x i16> zeroinitializer4986; IND-NEXT: [[TMP13:%.*]] = sext <2 x i16> [[PREDPHI]] to <2 x i32>4987; IND-NEXT: [[TMP14]] = or <2 x i32> [[VEC_PHI]], [[TMP13]]4988; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 24989; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i16> [[VEC_IND]], splat (i16 2)4990; IND-NEXT: [[TMP15:%.*]] = icmp eq i32 [[INDEX_NEXT]], 204991; IND-NEXT: br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]4992; IND: middle.block:4993; IND-NEXT: br label [[FOR_INC:%.*]]4994; IND: for.end:4995; IND-NEXT: [[VAR7:%.*]] = call i32 @llvm.vector.reduce.or.v2i32(<2 x i32> [[TMP14]])4996; IND-NEXT: ret i32 [[VAR7]]4997;4998; UNROLL-LABEL: @PR32419(4999; UNROLL-NEXT: entry:5000; UNROLL-NEXT: br label [[VECTOR_PH:%.*]]5001; UNROLL: vector.ph:5002; UNROLL-NEXT: [[TMP0:%.*]] = insertelement <2 x i32> <i32 poison, i32 0>, i32 [[A:%.*]], i64 05003; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]5004; UNROLL: vector.body:5005; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UREM_CONTINUE7:%.*]] ]5006; UNROLL-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP26:%.*]], [[PRED_UREM_CONTINUE7]] ]5007; UNROLL-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP27:%.*]], [[PRED_UREM_CONTINUE7]] ]5008; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i16> [ <i16 -20, i16 -19>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_UREM_CONTINUE7]] ]5009; UNROLL-NEXT: [[TMP1:%.*]] = trunc i32 [[INDEX]] to i165010; UNROLL-NEXT: [[TMP2:%.*]] = icmp ne <2 x i16> [[VEC_IND]], zeroinitializer5011; UNROLL-NEXT: [[TMP3:%.*]] = icmp ne <2 x i16> [[VEC_IND]], splat (i16 -2)5012; UNROLL-NEXT: [[TMP4:%.*]] = extractelement <2 x i1> [[TMP2]], i64 05013; UNROLL-NEXT: br i1 [[TMP4]], label [[PRED_UREM_IF:%.*]], label [[PRED_UREM_CONTINUE:%.*]]5014; UNROLL: pred.urem.if:5015; UNROLL-NEXT: [[TMP5:%.*]] = add i16 [[TMP1]], -205016; UNROLL-NEXT: [[TMP6:%.*]] = urem i16 [[B:%.*]], [[TMP5]]5017; UNROLL-NEXT: [[TMP7:%.*]] = insertelement <2 x i16> poison, i16 [[TMP6]], i64 05018; UNROLL-NEXT: br label [[PRED_UREM_CONTINUE]]5019; UNROLL: pred.urem.continue:5020; UNROLL-NEXT: [[TMP8:%.*]] = phi <2 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP7]], [[PRED_UREM_IF]] ]5021; UNROLL-NEXT: [[TMP9:%.*]] = extractelement <2 x i1> [[TMP2]], i64 15022; UNROLL-NEXT: br i1 [[TMP9]], label [[PRED_UREM_IF2:%.*]], label [[PRED_UREM_CONTINUE3:%.*]]5023; UNROLL: pred.urem.if2:5024; UNROLL-NEXT: [[TMP10:%.*]] = add i16 [[TMP1]], -195025; UNROLL-NEXT: [[TMP11:%.*]] = urem i16 [[B]], [[TMP10]]5026; UNROLL-NEXT: [[TMP12:%.*]] = insertelement <2 x i16> [[TMP8]], i16 [[TMP11]], i64 15027; UNROLL-NEXT: br label [[PRED_UREM_CONTINUE3]]5028; UNROLL: pred.urem.continue3:5029; UNROLL-NEXT: [[TMP13:%.*]] = phi <2 x i16> [ [[TMP8]], [[PRED_UREM_CONTINUE]] ], [ [[TMP12]], [[PRED_UREM_IF2]] ]5030; UNROLL-NEXT: [[TMP14:%.*]] = extractelement <2 x i1> [[TMP3]], i64 05031; UNROLL-NEXT: br i1 [[TMP14]], label [[PRED_UREM_IF4:%.*]], label [[PRED_UREM_CONTINUE5:%.*]]5032; UNROLL: pred.urem.if4:5033; UNROLL-NEXT: [[TMP15:%.*]] = add i16 [[TMP1]], -185034; UNROLL-NEXT: [[TMP16:%.*]] = urem i16 [[B]], [[TMP15]]5035; UNROLL-NEXT: [[TMP17:%.*]] = insertelement <2 x i16> poison, i16 [[TMP16]], i64 05036; UNROLL-NEXT: br label [[PRED_UREM_CONTINUE5]]5037; UNROLL: pred.urem.continue5:5038; UNROLL-NEXT: [[TMP18:%.*]] = phi <2 x i16> [ poison, [[PRED_UREM_CONTINUE3]] ], [ [[TMP17]], [[PRED_UREM_IF4]] ]5039; UNROLL-NEXT: [[TMP19:%.*]] = extractelement <2 x i1> [[TMP3]], i64 15040; UNROLL-NEXT: br i1 [[TMP19]], label [[PRED_UREM_IF6:%.*]], label [[PRED_UREM_CONTINUE7]]5041; UNROLL: pred.urem.if6:5042; UNROLL-NEXT: [[TMP20:%.*]] = add i16 [[TMP1]], -175043; UNROLL-NEXT: [[TMP21:%.*]] = urem i16 [[B]], [[TMP20]]5044; UNROLL-NEXT: [[TMP22:%.*]] = insertelement <2 x i16> [[TMP18]], i16 [[TMP21]], i64 15045; UNROLL-NEXT: br label [[PRED_UREM_CONTINUE7]]5046; UNROLL: pred.urem.continue7:5047; UNROLL-NEXT: [[TMP23:%.*]] = phi <2 x i16> [ [[TMP18]], [[PRED_UREM_CONTINUE5]] ], [ [[TMP22]], [[PRED_UREM_IF6]] ]5048; UNROLL-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[TMP2]], <2 x i16> [[TMP13]], <2 x i16> zeroinitializer5049; UNROLL-NEXT: [[PREDPHI8:%.*]] = select <2 x i1> [[TMP3]], <2 x i16> [[TMP23]], <2 x i16> zeroinitializer5050; UNROLL-NEXT: [[TMP24:%.*]] = sext <2 x i16> [[PREDPHI]] to <2 x i32>5051; UNROLL-NEXT: [[TMP25:%.*]] = sext <2 x i16> [[PREDPHI8]] to <2 x i32>5052; UNROLL-NEXT: [[TMP26]] = or <2 x i32> [[VEC_PHI]], [[TMP24]]5053; UNROLL-NEXT: [[TMP27]] = or <2 x i32> [[VEC_PHI1]], [[TMP25]]5054; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 45055; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i16> [[VEC_IND]], splat (i16 4)5056; UNROLL-NEXT: [[TMP28:%.*]] = icmp eq i32 [[INDEX_NEXT]], 205057; UNROLL-NEXT: br i1 [[TMP28]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]5058; UNROLL: middle.block:5059; UNROLL-NEXT: br label [[FOR_INC:%.*]]5060; UNROLL: for.end:5061; UNROLL-NEXT: [[BIN_RDX:%.*]] = or <2 x i32> [[TMP27]], [[TMP26]]5062; UNROLL-NEXT: [[VAR7:%.*]] = call i32 @llvm.vector.reduce.or.v2i32(<2 x i32> [[BIN_RDX]])5063; UNROLL-NEXT: ret i32 [[VAR7]]5064;5065; UNROLL-NO-IC-LABEL: @PR32419(5066; UNROLL-NO-IC-NEXT: entry:5067; UNROLL-NO-IC-NEXT: br label [[VECTOR_PH:%.*]]5068; UNROLL-NO-IC: vector.ph:5069; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = insertelement <2 x i32> zeroinitializer, i32 [[A:%.*]], i32 05070; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]5071; UNROLL-NO-IC: vector.body:5072; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UREM_CONTINUE7:%.*]] ]5073; UNROLL-NO-IC-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP28:%.*]], [[PRED_UREM_CONTINUE7]] ]5074; UNROLL-NO-IC-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP29:%.*]], [[PRED_UREM_CONTINUE7]] ]5075; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i16> [ <i16 -20, i16 -19>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_UREM_CONTINUE7]] ]5076; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i16> [[VEC_IND]], splat (i16 2)5077; UNROLL-NO-IC-NEXT: [[OFFSET_IDX:%.*]] = add i32 -20, [[INDEX]]5078; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = trunc i32 [[OFFSET_IDX]] to i165079; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = icmp ne <2 x i16> [[VEC_IND]], zeroinitializer5080; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = icmp ne <2 x i16> [[STEP_ADD]], zeroinitializer5081; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = extractelement <2 x i1> [[TMP4]], i32 05082; UNROLL-NO-IC-NEXT: br i1 [[TMP6]], label [[PRED_UREM_IF:%.*]], label [[PRED_UREM_CONTINUE:%.*]]5083; UNROLL-NO-IC: pred.urem.if:5084; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = add i16 [[TMP1]], 05085; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = urem i16 [[B:%.*]], [[TMP7]]5086; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = insertelement <2 x i16> poison, i16 [[TMP8]], i32 05087; UNROLL-NO-IC-NEXT: br label [[PRED_UREM_CONTINUE]]5088; UNROLL-NO-IC: pred.urem.continue:5089; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = phi <2 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP9]], [[PRED_UREM_IF]] ]5090; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = extractelement <2 x i1> [[TMP4]], i32 15091; UNROLL-NO-IC-NEXT: br i1 [[TMP11]], label [[PRED_UREM_IF2:%.*]], label [[PRED_UREM_CONTINUE3:%.*]]5092; UNROLL-NO-IC: pred.urem.if2:5093; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = add i16 [[TMP1]], 15094; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = urem i16 [[B]], [[TMP12]]5095; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = insertelement <2 x i16> [[TMP10]], i16 [[TMP13]], i32 15096; UNROLL-NO-IC-NEXT: br label [[PRED_UREM_CONTINUE3]]5097; UNROLL-NO-IC: pred.urem.continue3:5098; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = phi <2 x i16> [ [[TMP10]], [[PRED_UREM_CONTINUE]] ], [ [[TMP14]], [[PRED_UREM_IF2]] ]5099; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = extractelement <2 x i1> [[TMP5]], i32 05100; UNROLL-NO-IC-NEXT: br i1 [[TMP16]], label [[PRED_UREM_IF4:%.*]], label [[PRED_UREM_CONTINUE5:%.*]]5101; UNROLL-NO-IC: pred.urem.if4:5102; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = add i16 [[TMP1]], 25103; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = urem i16 [[B]], [[TMP17]]5104; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = insertelement <2 x i16> poison, i16 [[TMP18]], i32 05105; UNROLL-NO-IC-NEXT: br label [[PRED_UREM_CONTINUE5]]5106; UNROLL-NO-IC: pred.urem.continue5:5107; UNROLL-NO-IC-NEXT: [[TMP20:%.*]] = phi <2 x i16> [ poison, [[PRED_UREM_CONTINUE3]] ], [ [[TMP19]], [[PRED_UREM_IF4]] ]5108; UNROLL-NO-IC-NEXT: [[TMP21:%.*]] = extractelement <2 x i1> [[TMP5]], i32 15109; UNROLL-NO-IC-NEXT: br i1 [[TMP21]], label [[PRED_UREM_IF6:%.*]], label [[PRED_UREM_CONTINUE7]]5110; UNROLL-NO-IC: pred.urem.if6:5111; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = add i16 [[TMP1]], 35112; UNROLL-NO-IC-NEXT: [[TMP23:%.*]] = urem i16 [[B]], [[TMP22]]5113; UNROLL-NO-IC-NEXT: [[TMP24:%.*]] = insertelement <2 x i16> [[TMP20]], i16 [[TMP23]], i32 15114; UNROLL-NO-IC-NEXT: br label [[PRED_UREM_CONTINUE7]]5115; UNROLL-NO-IC: pred.urem.continue7:5116; UNROLL-NO-IC-NEXT: [[TMP25:%.*]] = phi <2 x i16> [ [[TMP20]], [[PRED_UREM_CONTINUE5]] ], [ [[TMP24]], [[PRED_UREM_IF6]] ]5117; UNROLL-NO-IC-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[TMP4]], <2 x i16> [[TMP15]], <2 x i16> zeroinitializer5118; UNROLL-NO-IC-NEXT: [[PREDPHI8:%.*]] = select <2 x i1> [[TMP5]], <2 x i16> [[TMP25]], <2 x i16> zeroinitializer5119; UNROLL-NO-IC-NEXT: [[TMP26:%.*]] = sext <2 x i16> [[PREDPHI]] to <2 x i32>5120; UNROLL-NO-IC-NEXT: [[TMP27:%.*]] = sext <2 x i16> [[PREDPHI8]] to <2 x i32>5121; UNROLL-NO-IC-NEXT: [[TMP28]] = or <2 x i32> [[VEC_PHI]], [[TMP26]]5122; UNROLL-NO-IC-NEXT: [[TMP29]] = or <2 x i32> [[VEC_PHI1]], [[TMP27]]5123; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 45124; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i16> [[STEP_ADD]], splat (i16 2)5125; UNROLL-NO-IC-NEXT: [[TMP30:%.*]] = icmp eq i32 [[INDEX_NEXT]], 205126; UNROLL-NO-IC-NEXT: br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]5127; UNROLL-NO-IC: middle.block:5128; UNROLL-NO-IC-NEXT: [[BIN_RDX:%.*]] = or <2 x i32> [[TMP29]], [[TMP28]]5129; UNROLL-NO-IC-NEXT: [[TMP31:%.*]] = call i32 @llvm.vector.reduce.or.v2i32(<2 x i32> [[BIN_RDX]])5130; UNROLL-NO-IC-NEXT: br label [[FOR_INC:%.*]]5131; UNROLL-NO-IC: for.end:5132; UNROLL-NO-IC-NEXT: ret i32 [[TMP31]]5133;5134; INTERLEAVE-LABEL: @PR32419(5135; INTERLEAVE-NEXT: entry:5136; INTERLEAVE-NEXT: br label [[VECTOR_PH:%.*]]5137; INTERLEAVE: vector.ph:5138; INTERLEAVE-NEXT: [[TMP0:%.*]] = insertelement <4 x i32> <i32 poison, i32 0, i32 0, i32 0>, i32 [[A:%.*]], i64 05139; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]5140; INTERLEAVE: vector.body:5141; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UREM_CONTINUE15:%.*]] ]5142; INTERLEAVE-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP46:%.*]], [[PRED_UREM_CONTINUE15]] ]5143; INTERLEAVE-NEXT: [[VEC_PHI1:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP47:%.*]], [[PRED_UREM_CONTINUE15]] ]5144; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i16> [ <i16 -20, i16 -19, i16 -18, i16 -17>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_UREM_CONTINUE15]] ]5145; INTERLEAVE-NEXT: [[TMP1:%.*]] = trunc i32 [[INDEX]] to i165146; INTERLEAVE-NEXT: [[TMP2:%.*]] = icmp ne <4 x i16> [[VEC_IND]], zeroinitializer5147; INTERLEAVE-NEXT: [[TMP3:%.*]] = icmp ne <4 x i16> [[VEC_IND]], splat (i16 -4)5148; INTERLEAVE-NEXT: [[TMP4:%.*]] = extractelement <4 x i1> [[TMP2]], i64 05149; INTERLEAVE-NEXT: br i1 [[TMP4]], label [[PRED_UREM_IF:%.*]], label [[PRED_UREM_CONTINUE:%.*]]5150; INTERLEAVE: pred.urem.if:5151; INTERLEAVE-NEXT: [[TMP5:%.*]] = add i16 [[TMP1]], -205152; INTERLEAVE-NEXT: [[TMP6:%.*]] = urem i16 [[B:%.*]], [[TMP5]]5153; INTERLEAVE-NEXT: [[TMP7:%.*]] = insertelement <4 x i16> poison, i16 [[TMP6]], i64 05154; INTERLEAVE-NEXT: br label [[PRED_UREM_CONTINUE]]5155; INTERLEAVE: pred.urem.continue:5156; INTERLEAVE-NEXT: [[TMP8:%.*]] = phi <4 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP7]], [[PRED_UREM_IF]] ]5157; INTERLEAVE-NEXT: [[TMP9:%.*]] = extractelement <4 x i1> [[TMP2]], i64 15158; INTERLEAVE-NEXT: br i1 [[TMP9]], label [[PRED_UREM_IF2:%.*]], label [[PRED_UREM_CONTINUE3:%.*]]5159; INTERLEAVE: pred.urem.if2:5160; INTERLEAVE-NEXT: [[TMP10:%.*]] = add i16 [[TMP1]], -195161; INTERLEAVE-NEXT: [[TMP11:%.*]] = urem i16 [[B]], [[TMP10]]5162; INTERLEAVE-NEXT: [[TMP12:%.*]] = insertelement <4 x i16> [[TMP8]], i16 [[TMP11]], i64 15163; INTERLEAVE-NEXT: br label [[PRED_UREM_CONTINUE3]]5164; INTERLEAVE: pred.urem.continue3:5165; INTERLEAVE-NEXT: [[TMP13:%.*]] = phi <4 x i16> [ [[TMP8]], [[PRED_UREM_CONTINUE]] ], [ [[TMP12]], [[PRED_UREM_IF2]] ]5166; INTERLEAVE-NEXT: [[TMP14:%.*]] = extractelement <4 x i1> [[TMP2]], i64 25167; INTERLEAVE-NEXT: br i1 [[TMP14]], label [[PRED_UREM_IF4:%.*]], label [[PRED_UREM_CONTINUE5:%.*]]5168; INTERLEAVE: pred.urem.if4:5169; INTERLEAVE-NEXT: [[TMP15:%.*]] = add i16 [[TMP1]], -185170; INTERLEAVE-NEXT: [[TMP16:%.*]] = urem i16 [[B]], [[TMP15]]5171; INTERLEAVE-NEXT: [[TMP17:%.*]] = insertelement <4 x i16> [[TMP13]], i16 [[TMP16]], i64 25172; INTERLEAVE-NEXT: br label [[PRED_UREM_CONTINUE5]]5173; INTERLEAVE: pred.urem.continue5:5174; INTERLEAVE-NEXT: [[TMP18:%.*]] = phi <4 x i16> [ [[TMP13]], [[PRED_UREM_CONTINUE3]] ], [ [[TMP17]], [[PRED_UREM_IF4]] ]5175; INTERLEAVE-NEXT: [[TMP19:%.*]] = extractelement <4 x i1> [[TMP2]], i64 35176; INTERLEAVE-NEXT: br i1 [[TMP19]], label [[PRED_UREM_IF6:%.*]], label [[PRED_UREM_CONTINUE7:%.*]]5177; INTERLEAVE: pred.urem.if6:5178; INTERLEAVE-NEXT: [[TMP20:%.*]] = add i16 [[TMP1]], -175179; INTERLEAVE-NEXT: [[TMP21:%.*]] = urem i16 [[B]], [[TMP20]]5180; INTERLEAVE-NEXT: [[TMP22:%.*]] = insertelement <4 x i16> [[TMP18]], i16 [[TMP21]], i64 35181; INTERLEAVE-NEXT: br label [[PRED_UREM_CONTINUE7]]5182; INTERLEAVE: pred.urem.continue7:5183; INTERLEAVE-NEXT: [[TMP23:%.*]] = phi <4 x i16> [ [[TMP18]], [[PRED_UREM_CONTINUE5]] ], [ [[TMP22]], [[PRED_UREM_IF6]] ]5184; INTERLEAVE-NEXT: [[TMP24:%.*]] = extractelement <4 x i1> [[TMP3]], i64 05185; INTERLEAVE-NEXT: br i1 [[TMP24]], label [[PRED_UREM_IF8:%.*]], label [[PRED_UREM_CONTINUE9:%.*]]5186; INTERLEAVE: pred.urem.if8:5187; INTERLEAVE-NEXT: [[TMP25:%.*]] = add i16 [[TMP1]], -165188; INTERLEAVE-NEXT: [[TMP26:%.*]] = urem i16 [[B]], [[TMP25]]5189; INTERLEAVE-NEXT: [[TMP27:%.*]] = insertelement <4 x i16> poison, i16 [[TMP26]], i64 05190; INTERLEAVE-NEXT: br label [[PRED_UREM_CONTINUE9]]5191; INTERLEAVE: pred.urem.continue9:5192; INTERLEAVE-NEXT: [[TMP28:%.*]] = phi <4 x i16> [ poison, [[PRED_UREM_CONTINUE7]] ], [ [[TMP27]], [[PRED_UREM_IF8]] ]5193; INTERLEAVE-NEXT: [[TMP29:%.*]] = extractelement <4 x i1> [[TMP3]], i64 15194; INTERLEAVE-NEXT: br i1 [[TMP29]], label [[PRED_UREM_IF10:%.*]], label [[PRED_UREM_CONTINUE11:%.*]]5195; INTERLEAVE: pred.urem.if10:5196; INTERLEAVE-NEXT: [[TMP30:%.*]] = add i16 [[TMP1]], -155197; INTERLEAVE-NEXT: [[TMP31:%.*]] = urem i16 [[B]], [[TMP30]]5198; INTERLEAVE-NEXT: [[TMP32:%.*]] = insertelement <4 x i16> [[TMP28]], i16 [[TMP31]], i64 15199; INTERLEAVE-NEXT: br label [[PRED_UREM_CONTINUE11]]5200; INTERLEAVE: pred.urem.continue11:5201; INTERLEAVE-NEXT: [[TMP33:%.*]] = phi <4 x i16> [ [[TMP28]], [[PRED_UREM_CONTINUE9]] ], [ [[TMP32]], [[PRED_UREM_IF10]] ]5202; INTERLEAVE-NEXT: [[TMP34:%.*]] = extractelement <4 x i1> [[TMP3]], i64 25203; INTERLEAVE-NEXT: br i1 [[TMP34]], label [[PRED_UREM_IF12:%.*]], label [[PRED_UREM_CONTINUE13:%.*]]5204; INTERLEAVE: pred.urem.if12:5205; INTERLEAVE-NEXT: [[TMP35:%.*]] = add i16 [[TMP1]], -145206; INTERLEAVE-NEXT: [[TMP36:%.*]] = urem i16 [[B]], [[TMP35]]5207; INTERLEAVE-NEXT: [[TMP37:%.*]] = insertelement <4 x i16> [[TMP33]], i16 [[TMP36]], i64 25208; INTERLEAVE-NEXT: br label [[PRED_UREM_CONTINUE13]]5209; INTERLEAVE: pred.urem.continue13:5210; INTERLEAVE-NEXT: [[TMP38:%.*]] = phi <4 x i16> [ [[TMP33]], [[PRED_UREM_CONTINUE11]] ], [ [[TMP37]], [[PRED_UREM_IF12]] ]5211; INTERLEAVE-NEXT: [[TMP39:%.*]] = extractelement <4 x i1> [[TMP3]], i64 35212; INTERLEAVE-NEXT: br i1 [[TMP39]], label [[PRED_UREM_IF14:%.*]], label [[PRED_UREM_CONTINUE15]]5213; INTERLEAVE: pred.urem.if14:5214; INTERLEAVE-NEXT: [[TMP40:%.*]] = add i16 [[TMP1]], -135215; INTERLEAVE-NEXT: [[TMP41:%.*]] = urem i16 [[B]], [[TMP40]]5216; INTERLEAVE-NEXT: [[TMP42:%.*]] = insertelement <4 x i16> [[TMP38]], i16 [[TMP41]], i64 35217; INTERLEAVE-NEXT: br label [[PRED_UREM_CONTINUE15]]5218; INTERLEAVE: pred.urem.continue15:5219; INTERLEAVE-NEXT: [[TMP43:%.*]] = phi <4 x i16> [ [[TMP38]], [[PRED_UREM_CONTINUE13]] ], [ [[TMP42]], [[PRED_UREM_IF14]] ]5220; INTERLEAVE-NEXT: [[PREDPHI:%.*]] = select <4 x i1> [[TMP2]], <4 x i16> [[TMP23]], <4 x i16> zeroinitializer5221; INTERLEAVE-NEXT: [[PREDPHI16:%.*]] = select <4 x i1> [[TMP3]], <4 x i16> [[TMP43]], <4 x i16> zeroinitializer5222; INTERLEAVE-NEXT: [[TMP44:%.*]] = sext <4 x i16> [[PREDPHI]] to <4 x i32>5223; INTERLEAVE-NEXT: [[TMP45:%.*]] = sext <4 x i16> [[PREDPHI16]] to <4 x i32>5224; INTERLEAVE-NEXT: [[TMP46]] = or <4 x i32> [[VEC_PHI]], [[TMP44]]5225; INTERLEAVE-NEXT: [[TMP47]] = or <4 x i32> [[VEC_PHI1]], [[TMP45]]5226; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 85227; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i16> [[VEC_IND]], splat (i16 8)5228; INTERLEAVE-NEXT: [[TMP48:%.*]] = icmp eq i32 [[INDEX_NEXT]], 165229; INTERLEAVE-NEXT: br i1 [[TMP48]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]5230; INTERLEAVE: middle.block:5231; INTERLEAVE-NEXT: br label [[SCALAR_PH:%.*]]5232; INTERLEAVE: scalar.ph:5233; INTERLEAVE-NEXT: [[BIN_RDX:%.*]] = or <4 x i32> [[TMP47]], [[TMP46]]5234; INTERLEAVE-NEXT: [[BC_MERGE_RDX:%.*]] = call i32 @llvm.vector.reduce.or.v4i32(<4 x i32> [[BIN_RDX]])5235; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]5236; INTERLEAVE: for.body:5237; INTERLEAVE-NEXT: [[I:%.*]] = phi i32 [ -4, [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_INC:%.*]] ]5238; INTERLEAVE-NEXT: [[VAR0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR6:%.*]], [[FOR_INC]] ]5239; INTERLEAVE-NEXT: [[VAR1:%.*]] = trunc i32 [[I]] to i165240; INTERLEAVE-NEXT: [[VAR2:%.*]] = icmp eq i16 [[VAR1]], 05241; INTERLEAVE-NEXT: br i1 [[VAR2]], label [[FOR_INC]], label [[FOR_COND:%.*]]5242; INTERLEAVE: for.cond:5243; INTERLEAVE-NEXT: [[VAR3:%.*]] = urem i16 [[B]], [[VAR1]]5244; INTERLEAVE-NEXT: [[TMP50:%.*]] = sext i16 [[VAR3]] to i325245; INTERLEAVE-NEXT: br label [[FOR_INC]]5246; INTERLEAVE: for.inc:5247; INTERLEAVE-NEXT: [[VAR4:%.*]] = phi i32 [ [[TMP50]], [[FOR_COND]] ], [ 0, [[FOR_BODY]] ]5248; INTERLEAVE-NEXT: [[VAR6]] = or i32 [[VAR0]], [[VAR4]]5249; INTERLEAVE-NEXT: [[I_NEXT]] = add nsw i32 [[I]], 15250; INTERLEAVE-NEXT: [[COND:%.*]] = icmp eq i32 [[I_NEXT]], 05251; INTERLEAVE-NEXT: br i1 [[COND]], label [[FOR_END:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]5252; INTERLEAVE: for.end:5253; INTERLEAVE-NEXT: ret i32 [[VAR6]]5254;5255entry:5256 br label %for.body5257 5258for.body:5259 %i = phi i32 [ -20, %entry ], [ %i.next, %for.inc ]5260 %var0 = phi i32 [ %a, %entry ], [ %var6, %for.inc ]5261 %var1 = trunc i32 %i to i165262 %var2 = icmp eq i16 %var1, 05263 br i1 %var2, label %for.inc, label %for.cond5264 5265for.cond:5266 %var3 = urem i16 %b, %var15267 br label %for.inc5268 5269for.inc:5270 %var4 = phi i16 [ %var3, %for.cond ], [ 0, %for.body ]5271 %var5 = sext i16 %var4 to i325272 %var6 = or i32 %var0, %var55273 %i.next = add nsw i32 %i, 15274 %cond = icmp eq i32 %i.next, 05275 br i1 %cond, label %for.end, label %for.body5276 5277for.end:5278 %var7 = phi i32 [ %var6, %for.inc ]5279 ret i32 %var75280}5281 5282; Ensure that the shuffle vector for first order recurrence is inserted5283; correctly after all the phis. These new phis correspond to new IVs5284; that are generated by optimizing non-free truncs of IVs to IVs themselves.5285; This also ensures the first-order recurrence splice recipe is placed5286; correctly if it is fed by an induction.5287define i64 @trunc_with_first_order_recurrence() {5288; CHECK-LABEL: @trunc_with_first_order_recurrence(5289; CHECK-NEXT: entry:5290; CHECK-NEXT: br label [[VECTOR_PH:%.*]]5291; CHECK: vector.ph:5292; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]5293; CHECK: vector.body:5294; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]5295; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]5296; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]5297; CHECK-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 42>, [[VECTOR_PH]] ], [ [[VEC_IND2:%.*]], [[VECTOR_BODY]] ]5298; CHECK-NEXT: [[VEC_IND2]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT3:%.*]], [[VECTOR_BODY]] ]5299; CHECK-NEXT: [[VEC_IND4:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT5:%.*]], [[VECTOR_BODY]] ]5300; CHECK-NEXT: [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND2]], <2 x i32> <i32 1, i32 2>5301; CHECK-NEXT: [[TMP1:%.*]] = mul <2 x i32> [[VEC_IND]], [[VEC_IND2]]5302; CHECK-NEXT: [[TMP2:%.*]] = add <2 x i32> [[TMP1]], splat (i32 42)5303; CHECK-NEXT: [[TMP3:%.*]] = add <2 x i32> [[TMP0]], [[VEC_IND2]]5304; CHECK-NEXT: [[TMP4:%.*]] = add <2 x i32> [[TMP3]], [[TMP2]]5305; CHECK-NEXT: [[TMP5:%.*]] = sext <2 x i32> [[TMP4]] to <2 x i64>5306; CHECK-NEXT: [[TMP6:%.*]] = add <2 x i64> [[VEC_PHI]], [[TMP5]]5307; CHECK-NEXT: [[TMP7:%.*]] = shl <2 x i32> [[VEC_IND4]], splat (i32 1)5308; CHECK-NEXT: [[TMP8:%.*]] = add <2 x i32> [[TMP2]], [[TMP7]]5309; CHECK-NEXT: [[TMP9:%.*]] = sext <2 x i32> [[TMP8]] to <2 x i64>5310; CHECK-NEXT: [[TMP10]] = add <2 x i64> [[TMP6]], [[TMP9]]5311; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 25312; CHECK-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <2 x i32> [[VEC_IND]], splat (i32 2)5313; CHECK-NEXT: [[VEC_IND_NEXT3]] = add <2 x i32> [[VEC_IND2]], splat (i32 2)5314; CHECK-NEXT: [[VEC_IND_NEXT5]] = add <2 x i32> [[VEC_IND4]], splat (i32 2)5315; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1125316; CHECK-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]5317; CHECK: middle.block:5318; CHECK-NEXT: [[TMP12:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[TMP10]])5319; CHECK-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[VEC_IND2]], i32 15320; CHECK-NEXT: br label [[SCALAR_PH:%.*]]5321; CHECK: scalar.ph:5322; CHECK-NEXT: br label [[LOOP:%.*]]5323; CHECK: exit:5324; CHECK-NEXT: [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ]5325; CHECK-NEXT: ret i64 [[DOTLCSSA]]5326; CHECK: loop:5327; CHECK-NEXT: [[C5:%.*]] = phi i64 [ [[C23]], [[LOOP]] ], [ [[TMP12]], [[SCALAR_PH]] ]5328; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ 113, [[SCALAR_PH]] ]5329; CHECK-NEXT: [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ 113, [[SCALAR_PH]] ]5330; CHECK-NEXT: [[Y:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT]], [[SCALAR_PH]] ]5331; CHECK-NEXT: [[C6]] = trunc i64 [[INDVARS_IV]] to i325332; CHECK-NEXT: [[C8:%.*]] = mul i32 [[X]], [[C6]]5333; CHECK-NEXT: [[C9:%.*]] = add i32 [[C8]], 425334; CHECK-NEXT: [[C10:%.*]] = add i32 [[Y]], [[C6]]5335; CHECK-NEXT: [[C11:%.*]] = add i32 [[C10]], [[C9]]5336; CHECK-NEXT: [[C12:%.*]] = sext i32 [[C11]] to i645337; CHECK-NEXT: [[C13:%.*]] = add i64 [[C5]], [[C12]]5338; CHECK-NEXT: [[INDVARS_IV_TR:%.*]] = trunc i64 [[INDVARS_IV]] to i325339; CHECK-NEXT: [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 15340; CHECK-NEXT: [[C15:%.*]] = add i32 [[C9]], [[C14]]5341; CHECK-NEXT: [[C16:%.*]] = sext i32 [[C15]] to i645342; CHECK-NEXT: [[C23]] = add i64 [[C13]], [[C16]]5343; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 15344; CHECK-NEXT: [[C24]] = add nuw nsw i32 [[X]], 15345; CHECK-NEXT: [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 1145346; CHECK-NEXT: br i1 [[EXITCOND_I]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP46:![0-9]+]]5347;5348; IND-LABEL: @trunc_with_first_order_recurrence(5349; IND-NEXT: entry:5350; IND-NEXT: br label [[VECTOR_PH:%.*]]5351; IND: vector.ph:5352; IND-NEXT: br label [[VECTOR_BODY:%.*]]5353; IND: vector.body:5354; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]5355; IND-NEXT: [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]5356; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]5357; IND-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 42>, [[VECTOR_PH]] ], [ [[VEC_IND2:%.*]], [[VECTOR_BODY]] ]5358; IND-NEXT: [[VEC_IND2]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT3:%.*]], [[VECTOR_BODY]] ]5359; IND-NEXT: [[VEC_IND4:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT5:%.*]], [[VECTOR_BODY]] ]5360; IND-NEXT: [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND2]], <2 x i32> <i32 1, i32 2>5361; IND-NEXT: [[TMP1:%.*]] = mul <2 x i32> [[VEC_IND]], [[VEC_IND2]]5362; IND-NEXT: [[TMP2:%.*]] = add <2 x i32> [[TMP1]], splat (i32 42)5363; IND-NEXT: [[TMP3:%.*]] = add <2 x i32> [[TMP0]], [[VEC_IND2]]5364; IND-NEXT: [[TMP4:%.*]] = add <2 x i32> [[TMP3]], [[TMP2]]5365; IND-NEXT: [[TMP5:%.*]] = sext <2 x i32> [[TMP4]] to <2 x i64>5366; IND-NEXT: [[TMP6:%.*]] = add <2 x i64> [[VEC_PHI]], [[TMP5]]5367; IND-NEXT: [[TMP7:%.*]] = shl <2 x i32> [[VEC_IND4]], splat (i32 1)5368; IND-NEXT: [[TMP8:%.*]] = add <2 x i32> [[TMP2]], [[TMP7]]5369; IND-NEXT: [[TMP9:%.*]] = sext <2 x i32> [[TMP8]] to <2 x i64>5370; IND-NEXT: [[TMP10]] = add <2 x i64> [[TMP6]], [[TMP9]]5371; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 25372; IND-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <2 x i32> [[VEC_IND]], splat (i32 2)5373; IND-NEXT: [[VEC_IND_NEXT3]] = add <2 x i32> [[VEC_IND2]], splat (i32 2)5374; IND-NEXT: [[VEC_IND_NEXT5]] = add <2 x i32> [[VEC_IND4]], splat (i32 2)5375; IND-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1125376; IND-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]5377; IND: middle.block:5378; IND-NEXT: br label [[SCALAR_PH:%.*]]5379; IND: scalar.ph:5380; IND-NEXT: [[SCALAR_RECUR_INIT:%.*]] = extractelement <2 x i32> [[VEC_IND2]], i64 15381; IND-NEXT: [[BC_MERGE_RDX:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[TMP10]])5382; IND-NEXT: br label [[LOOP:%.*]]5383; IND: exit:5384; IND-NEXT: ret i64 [[DOTLCSSA:%.*]]5385; IND: loop:5386; IND-NEXT: [[C5:%.*]] = phi i64 [ [[DOTLCSSA]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]5387; IND-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ 113, [[SCALAR_PH]] ]5388; IND-NEXT: [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ 113, [[SCALAR_PH]] ]5389; IND-NEXT: [[Y:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]5390; IND-NEXT: [[C6]] = trunc i64 [[INDVARS_IV]] to i325391; IND-NEXT: [[C8:%.*]] = mul i32 [[X]], [[C6]]5392; IND-NEXT: [[C9:%.*]] = add i32 [[C8]], 425393; IND-NEXT: [[C10:%.*]] = add i32 [[Y]], [[C6]]5394; IND-NEXT: [[C11:%.*]] = add i32 [[C10]], [[C9]]5395; IND-NEXT: [[C12:%.*]] = sext i32 [[C11]] to i645396; IND-NEXT: [[C13:%.*]] = add i64 [[C5]], [[C12]]5397; IND-NEXT: [[INDVARS_IV_TR:%.*]] = trunc i64 [[INDVARS_IV]] to i325398; IND-NEXT: [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 15399; IND-NEXT: [[C15:%.*]] = add i32 [[C9]], [[C14]]5400; IND-NEXT: [[C16:%.*]] = sext i32 [[C15]] to i645401; IND-NEXT: [[DOTLCSSA]] = add i64 [[C13]], [[C16]]5402; IND-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 15403; IND-NEXT: [[C24]] = add nuw nsw i32 [[X]], 15404; IND-NEXT: [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 1145405; IND-NEXT: br i1 [[EXITCOND_I]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP46:![0-9]+]]5406;5407; UNROLL-LABEL: @trunc_with_first_order_recurrence(5408; UNROLL-NEXT: entry:5409; UNROLL-NEXT: br label [[VECTOR_PH:%.*]]5410; UNROLL: vector.ph:5411; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]5412; UNROLL: vector.body:5413; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]5414; UNROLL-NEXT: [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]5415; UNROLL-NEXT: [[VEC_PHI2:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[VECTOR_BODY]] ]5416; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]5417; UNROLL-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 42>, [[VECTOR_PH]] ], [ [[STEP_ADD7:%.*]], [[VECTOR_BODY]] ]5418; UNROLL-NEXT: [[VEC_IND3:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT4:%.*]], [[VECTOR_BODY]] ]5419; UNROLL-NEXT: [[VEC_IND5:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT6:%.*]], [[VECTOR_BODY]] ]5420; UNROLL-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)5421; UNROLL-NEXT: [[STEP_ADD7]] = add <2 x i32> [[VEC_IND3]], splat (i32 2)5422; UNROLL-NEXT: [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND3]], <2 x i32> <i32 1, i32 2>5423; UNROLL-NEXT: [[TMP1:%.*]] = shufflevector <2 x i32> [[VEC_IND3]], <2 x i32> [[STEP_ADD7]], <2 x i32> <i32 1, i32 2>5424; UNROLL-NEXT: [[TMP2:%.*]] = mul <2 x i32> [[VEC_IND]], [[VEC_IND3]]5425; UNROLL-NEXT: [[TMP3:%.*]] = mul <2 x i32> [[STEP_ADD]], [[STEP_ADD7]]5426; UNROLL-NEXT: [[TMP4:%.*]] = add <2 x i32> [[TMP2]], splat (i32 42)5427; UNROLL-NEXT: [[TMP5:%.*]] = add <2 x i32> [[TMP3]], splat (i32 42)5428; UNROLL-NEXT: [[TMP6:%.*]] = add <2 x i32> [[TMP0]], [[VEC_IND3]]5429; UNROLL-NEXT: [[TMP7:%.*]] = add <2 x i32> [[TMP1]], [[STEP_ADD7]]5430; UNROLL-NEXT: [[TMP8:%.*]] = add <2 x i32> [[TMP6]], [[TMP4]]5431; UNROLL-NEXT: [[TMP9:%.*]] = add <2 x i32> [[TMP7]], [[TMP5]]5432; UNROLL-NEXT: [[TMP10:%.*]] = sext <2 x i32> [[TMP8]] to <2 x i64>5433; UNROLL-NEXT: [[TMP11:%.*]] = sext <2 x i32> [[TMP9]] to <2 x i64>5434; UNROLL-NEXT: [[TMP12:%.*]] = add <2 x i64> [[VEC_PHI]], [[TMP10]]5435; UNROLL-NEXT: [[TMP13:%.*]] = add <2 x i64> [[VEC_PHI2]], [[TMP11]]5436; UNROLL-NEXT: [[TMP14:%.*]] = shl <2 x i32> [[VEC_IND5]], splat (i32 1)5437; UNROLL-NEXT: [[STEP_ADD8:%.*]] = shl <2 x i32> [[VEC_IND5]], splat (i32 1)5438; UNROLL-NEXT: [[TMP15:%.*]] = add <2 x i32> [[STEP_ADD8]], splat (i32 4)5439; UNROLL-NEXT: [[TMP16:%.*]] = add <2 x i32> [[TMP4]], [[TMP14]]5440; UNROLL-NEXT: [[TMP17:%.*]] = add <2 x i32> [[TMP5]], [[TMP15]]5441; UNROLL-NEXT: [[TMP18:%.*]] = sext <2 x i32> [[TMP16]] to <2 x i64>5442; UNROLL-NEXT: [[TMP19:%.*]] = sext <2 x i32> [[TMP17]] to <2 x i64>5443; UNROLL-NEXT: [[TMP20]] = add <2 x i64> [[TMP12]], [[TMP18]]5444; UNROLL-NEXT: [[TMP21]] = add <2 x i64> [[TMP13]], [[TMP19]]5445; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 45446; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 4)5447; UNROLL-NEXT: [[VEC_IND_NEXT4]] = add <2 x i32> [[VEC_IND3]], splat (i32 4)5448; UNROLL-NEXT: [[VEC_IND_NEXT6]] = add <2 x i32> [[VEC_IND5]], splat (i32 4)5449; UNROLL-NEXT: [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1125450; UNROLL-NEXT: br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]5451; UNROLL: middle.block:5452; UNROLL-NEXT: br label [[SCALAR_PH:%.*]]5453; UNROLL: scalar.ph:5454; UNROLL-NEXT: [[SCALAR_RECUR_INIT:%.*]] = extractelement <2 x i32> [[STEP_ADD7]], i64 15455; UNROLL-NEXT: [[BIN_RDX:%.*]] = add <2 x i64> [[TMP21]], [[TMP20]]5456; UNROLL-NEXT: [[BC_MERGE_RDX:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])5457; UNROLL-NEXT: br label [[LOOP:%.*]]5458; UNROLL: exit:5459; UNROLL-NEXT: ret i64 [[DOTLCSSA:%.*]]5460; UNROLL: loop:5461; UNROLL-NEXT: [[C5:%.*]] = phi i64 [ [[DOTLCSSA]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]5462; UNROLL-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ 113, [[SCALAR_PH]] ]5463; UNROLL-NEXT: [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ 113, [[SCALAR_PH]] ]5464; UNROLL-NEXT: [[Y:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]5465; UNROLL-NEXT: [[C6]] = trunc i64 [[INDVARS_IV]] to i325466; UNROLL-NEXT: [[C8:%.*]] = mul i32 [[X]], [[C6]]5467; UNROLL-NEXT: [[C9:%.*]] = add i32 [[C8]], 425468; UNROLL-NEXT: [[C10:%.*]] = add i32 [[Y]], [[C6]]5469; UNROLL-NEXT: [[C11:%.*]] = add i32 [[C10]], [[C9]]5470; UNROLL-NEXT: [[C12:%.*]] = sext i32 [[C11]] to i645471; UNROLL-NEXT: [[C13:%.*]] = add i64 [[C5]], [[C12]]5472; UNROLL-NEXT: [[INDVARS_IV_TR:%.*]] = trunc i64 [[INDVARS_IV]] to i325473; UNROLL-NEXT: [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 15474; UNROLL-NEXT: [[C15:%.*]] = add i32 [[C9]], [[C14]]5475; UNROLL-NEXT: [[C16:%.*]] = sext i32 [[C15]] to i645476; UNROLL-NEXT: [[DOTLCSSA]] = add i64 [[C13]], [[C16]]5477; UNROLL-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 15478; UNROLL-NEXT: [[C24]] = add nuw nsw i32 [[X]], 15479; UNROLL-NEXT: [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 1145480; UNROLL-NEXT: br i1 [[EXITCOND_I]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP46:![0-9]+]]5481;5482; UNROLL-NO-IC-LABEL: @trunc_with_first_order_recurrence(5483; UNROLL-NO-IC-NEXT: entry:5484; UNROLL-NO-IC-NEXT: br label [[VECTOR_PH:%.*]]5485; UNROLL-NO-IC: vector.ph:5486; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]5487; UNROLL-NO-IC: vector.body:5488; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]5489; UNROLL-NO-IC-NEXT: [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]5490; UNROLL-NO-IC-NEXT: [[VEC_PHI2:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[VECTOR_BODY]] ]5491; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]5492; UNROLL-NO-IC-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 42>, [[VECTOR_PH]] ], [ [[STEP_ADD7:%.*]], [[VECTOR_BODY]] ]5493; UNROLL-NO-IC-NEXT: [[VEC_IND3:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT4:%.*]], [[VECTOR_BODY]] ]5494; UNROLL-NO-IC-NEXT: [[VEC_IND5:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT6:%.*]], [[VECTOR_BODY]] ]5495; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)5496; UNROLL-NO-IC-NEXT: [[STEP_ADD7]] = add <2 x i32> [[VEC_IND3]], splat (i32 2)5497; UNROLL-NO-IC-NEXT: [[STEP_ADD8:%.*]] = add <2 x i32> [[VEC_IND5]], splat (i32 2)5498; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND3]], <2 x i32> <i32 1, i32 2>5499; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = shufflevector <2 x i32> [[VEC_IND3]], <2 x i32> [[STEP_ADD7]], <2 x i32> <i32 1, i32 2>5500; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = mul <2 x i32> [[VEC_IND]], [[VEC_IND3]]5501; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = mul <2 x i32> [[STEP_ADD]], [[STEP_ADD7]]5502; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = add <2 x i32> [[TMP2]], splat (i32 42)5503; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = add <2 x i32> [[TMP3]], splat (i32 42)5504; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = add <2 x i32> [[TMP0]], [[VEC_IND3]]5505; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = add <2 x i32> [[TMP1]], [[STEP_ADD7]]5506; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = add <2 x i32> [[TMP6]], [[TMP4]]5507; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = add <2 x i32> [[TMP7]], [[TMP5]]5508; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = sext <2 x i32> [[TMP8]] to <2 x i64>5509; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = sext <2 x i32> [[TMP9]] to <2 x i64>5510; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = add <2 x i64> [[VEC_PHI]], [[TMP10]]5511; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = add <2 x i64> [[VEC_PHI2]], [[TMP11]]5512; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = shl <2 x i32> [[VEC_IND5]], splat (i32 1)5513; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = shl <2 x i32> [[STEP_ADD8]], splat (i32 1)5514; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = add <2 x i32> [[TMP4]], [[TMP14]]5515; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = add <2 x i32> [[TMP5]], [[TMP15]]5516; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = sext <2 x i32> [[TMP16]] to <2 x i64>5517; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = sext <2 x i32> [[TMP17]] to <2 x i64>5518; UNROLL-NO-IC-NEXT: [[TMP20]] = add <2 x i64> [[TMP12]], [[TMP18]]5519; UNROLL-NO-IC-NEXT: [[TMP21]] = add <2 x i64> [[TMP13]], [[TMP19]]5520; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 45521; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add nuw nsw <2 x i32> [[STEP_ADD]], splat (i32 2)5522; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT4]] = add <2 x i32> [[STEP_ADD7]], splat (i32 2)5523; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT6]] = add <2 x i32> [[STEP_ADD8]], splat (i32 2)5524; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1125525; UNROLL-NO-IC-NEXT: br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]5526; UNROLL-NO-IC: middle.block:5527; UNROLL-NO-IC-NEXT: [[BIN_RDX:%.*]] = add <2 x i64> [[TMP21]], [[TMP20]]5528; UNROLL-NO-IC-NEXT: [[TMP23:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])5529; UNROLL-NO-IC-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[STEP_ADD7]], i32 15530; UNROLL-NO-IC-NEXT: br label [[SCALAR_PH:%.*]]5531; UNROLL-NO-IC: scalar.ph:5532; UNROLL-NO-IC-NEXT: br label [[LOOP:%.*]]5533; UNROLL-NO-IC: exit:5534; UNROLL-NO-IC-NEXT: [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ]5535; UNROLL-NO-IC-NEXT: ret i64 [[DOTLCSSA]]5536; UNROLL-NO-IC: loop:5537; UNROLL-NO-IC-NEXT: [[C5:%.*]] = phi i64 [ [[C23]], [[LOOP]] ], [ [[TMP23]], [[SCALAR_PH]] ]5538; UNROLL-NO-IC-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ 113, [[SCALAR_PH]] ]5539; UNROLL-NO-IC-NEXT: [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ 113, [[SCALAR_PH]] ]5540; UNROLL-NO-IC-NEXT: [[Y:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT]], [[SCALAR_PH]] ]5541; UNROLL-NO-IC-NEXT: [[C6]] = trunc i64 [[INDVARS_IV]] to i325542; UNROLL-NO-IC-NEXT: [[C8:%.*]] = mul i32 [[X]], [[C6]]5543; UNROLL-NO-IC-NEXT: [[C9:%.*]] = add i32 [[C8]], 425544; UNROLL-NO-IC-NEXT: [[C10:%.*]] = add i32 [[Y]], [[C6]]5545; UNROLL-NO-IC-NEXT: [[C11:%.*]] = add i32 [[C10]], [[C9]]5546; UNROLL-NO-IC-NEXT: [[C12:%.*]] = sext i32 [[C11]] to i645547; UNROLL-NO-IC-NEXT: [[C13:%.*]] = add i64 [[C5]], [[C12]]5548; UNROLL-NO-IC-NEXT: [[INDVARS_IV_TR:%.*]] = trunc i64 [[INDVARS_IV]] to i325549; UNROLL-NO-IC-NEXT: [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 15550; UNROLL-NO-IC-NEXT: [[C15:%.*]] = add i32 [[C9]], [[C14]]5551; UNROLL-NO-IC-NEXT: [[C16:%.*]] = sext i32 [[C15]] to i645552; UNROLL-NO-IC-NEXT: [[C23]] = add i64 [[C13]], [[C16]]5553; UNROLL-NO-IC-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 15554; UNROLL-NO-IC-NEXT: [[C24]] = add nuw nsw i32 [[X]], 15555; UNROLL-NO-IC-NEXT: [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 1145556; UNROLL-NO-IC-NEXT: br i1 [[EXITCOND_I]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP46:![0-9]+]]5557;5558; INTERLEAVE-LABEL: @trunc_with_first_order_recurrence(5559; INTERLEAVE-NEXT: entry:5560; INTERLEAVE-NEXT: br label [[VECTOR_PH:%.*]]5561; INTERLEAVE: vector.ph:5562; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]5563; INTERLEAVE: vector.body:5564; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]5565; INTERLEAVE-NEXT: [[VEC_PHI:%.*]] = phi <4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]5566; INTERLEAVE-NEXT: [[VEC_PHI2:%.*]] = phi <4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[VECTOR_BODY]] ]5567; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 1, i32 2, i32 3, i32 4>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]5568; INTERLEAVE-NEXT: [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 42>, [[VECTOR_PH]] ], [ [[STEP_ADD7:%.*]], [[VECTOR_BODY]] ]5569; INTERLEAVE-NEXT: [[VEC_IND3:%.*]] = phi <4 x i32> [ <i32 1, i32 2, i32 3, i32 4>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT4:%.*]], [[VECTOR_BODY]] ]5570; INTERLEAVE-NEXT: [[VEC_IND5:%.*]] = phi <4 x i32> [ <i32 1, i32 2, i32 3, i32 4>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT6:%.*]], [[VECTOR_BODY]] ]5571; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], splat (i32 4)5572; INTERLEAVE-NEXT: [[STEP_ADD7]] = add <4 x i32> [[VEC_IND3]], splat (i32 4)5573; INTERLEAVE-NEXT: [[TMP0:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[VEC_IND3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>5574; INTERLEAVE-NEXT: [[TMP1:%.*]] = shufflevector <4 x i32> [[VEC_IND3]], <4 x i32> [[STEP_ADD7]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>5575; INTERLEAVE-NEXT: [[TMP2:%.*]] = mul <4 x i32> [[VEC_IND]], [[VEC_IND3]]5576; INTERLEAVE-NEXT: [[TMP3:%.*]] = mul <4 x i32> [[STEP_ADD]], [[STEP_ADD7]]5577; INTERLEAVE-NEXT: [[TMP4:%.*]] = add <4 x i32> [[TMP2]], splat (i32 42)5578; INTERLEAVE-NEXT: [[TMP5:%.*]] = add <4 x i32> [[TMP3]], splat (i32 42)5579; INTERLEAVE-NEXT: [[TMP6:%.*]] = add <4 x i32> [[TMP0]], [[VEC_IND3]]5580; INTERLEAVE-NEXT: [[TMP7:%.*]] = add <4 x i32> [[TMP1]], [[STEP_ADD7]]5581; INTERLEAVE-NEXT: [[TMP8:%.*]] = add <4 x i32> [[TMP6]], [[TMP4]]5582; INTERLEAVE-NEXT: [[TMP9:%.*]] = add <4 x i32> [[TMP7]], [[TMP5]]5583; INTERLEAVE-NEXT: [[TMP10:%.*]] = sext <4 x i32> [[TMP8]] to <4 x i64>5584; INTERLEAVE-NEXT: [[TMP11:%.*]] = sext <4 x i32> [[TMP9]] to <4 x i64>5585; INTERLEAVE-NEXT: [[TMP12:%.*]] = add <4 x i64> [[VEC_PHI]], [[TMP10]]5586; INTERLEAVE-NEXT: [[TMP13:%.*]] = add <4 x i64> [[VEC_PHI2]], [[TMP11]]5587; INTERLEAVE-NEXT: [[TMP14:%.*]] = shl <4 x i32> [[VEC_IND5]], splat (i32 1)5588; INTERLEAVE-NEXT: [[STEP_ADD8:%.*]] = shl <4 x i32> [[VEC_IND5]], splat (i32 1)5589; INTERLEAVE-NEXT: [[TMP15:%.*]] = add <4 x i32> [[STEP_ADD8]], splat (i32 8)5590; INTERLEAVE-NEXT: [[TMP16:%.*]] = add <4 x i32> [[TMP4]], [[TMP14]]5591; INTERLEAVE-NEXT: [[TMP17:%.*]] = add <4 x i32> [[TMP5]], [[TMP15]]5592; INTERLEAVE-NEXT: [[TMP18:%.*]] = sext <4 x i32> [[TMP16]] to <4 x i64>5593; INTERLEAVE-NEXT: [[TMP19:%.*]] = sext <4 x i32> [[TMP17]] to <4 x i64>5594; INTERLEAVE-NEXT: [[TMP20]] = add <4 x i64> [[TMP12]], [[TMP18]]5595; INTERLEAVE-NEXT: [[TMP21]] = add <4 x i64> [[TMP13]], [[TMP19]]5596; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 85597; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 8)5598; INTERLEAVE-NEXT: [[VEC_IND_NEXT4]] = add <4 x i32> [[VEC_IND3]], splat (i32 8)5599; INTERLEAVE-NEXT: [[VEC_IND_NEXT6]] = add <4 x i32> [[VEC_IND5]], splat (i32 8)5600; INTERLEAVE-NEXT: [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1125601; INTERLEAVE-NEXT: br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP46:![0-9]+]]5602; INTERLEAVE: middle.block:5603; INTERLEAVE-NEXT: br label [[SCALAR_PH:%.*]]5604; INTERLEAVE: scalar.ph:5605; INTERLEAVE-NEXT: [[SCALAR_RECUR_INIT:%.*]] = extractelement <4 x i32> [[STEP_ADD7]], i64 35606; INTERLEAVE-NEXT: [[BIN_RDX:%.*]] = add <4 x i64> [[TMP21]], [[TMP20]]5607; INTERLEAVE-NEXT: [[BC_MERGE_RDX:%.*]] = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> [[BIN_RDX]])5608; INTERLEAVE-NEXT: br label [[LOOP:%.*]]5609; INTERLEAVE: exit:5610; INTERLEAVE-NEXT: ret i64 [[DOTLCSSA:%.*]]5611; INTERLEAVE: loop:5612; INTERLEAVE-NEXT: [[C5:%.*]] = phi i64 [ [[DOTLCSSA]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]5613; INTERLEAVE-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ 113, [[SCALAR_PH]] ]5614; INTERLEAVE-NEXT: [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ 113, [[SCALAR_PH]] ]5615; INTERLEAVE-NEXT: [[Y:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]5616; INTERLEAVE-NEXT: [[C6]] = trunc i64 [[INDVARS_IV]] to i325617; INTERLEAVE-NEXT: [[C8:%.*]] = mul i32 [[X]], [[C6]]5618; INTERLEAVE-NEXT: [[C9:%.*]] = add i32 [[C8]], 425619; INTERLEAVE-NEXT: [[C10:%.*]] = add i32 [[Y]], [[C6]]5620; INTERLEAVE-NEXT: [[C11:%.*]] = add i32 [[C10]], [[C9]]5621; INTERLEAVE-NEXT: [[C12:%.*]] = sext i32 [[C11]] to i645622; INTERLEAVE-NEXT: [[C13:%.*]] = add i64 [[C5]], [[C12]]5623; INTERLEAVE-NEXT: [[INDVARS_IV_TR:%.*]] = trunc i64 [[INDVARS_IV]] to i325624; INTERLEAVE-NEXT: [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 15625; INTERLEAVE-NEXT: [[C15:%.*]] = add i32 [[C9]], [[C14]]5626; INTERLEAVE-NEXT: [[C16:%.*]] = sext i32 [[C15]] to i645627; INTERLEAVE-NEXT: [[DOTLCSSA]] = add i64 [[C13]], [[C16]]5628; INTERLEAVE-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 15629; INTERLEAVE-NEXT: [[C24]] = add nuw nsw i32 [[X]], 15630; INTERLEAVE-NEXT: [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 1145631; INTERLEAVE-NEXT: br i1 [[EXITCOND_I]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP47:![0-9]+]]5632;5633entry:5634 br label %loop5635 5636exit: ; preds = %loop5637 %.lcssa = phi i64 [ %c23, %loop ]5638 ret i64 %.lcssa5639 5640loop: ; preds = %loop, %entry5641 %c5 = phi i64 [ %c23, %loop ], [ 0, %entry ]5642 %indvars.iv = phi i64 [ %indvars.iv.next, %loop ], [ 1, %entry ]5643 %x = phi i32 [ %c24, %loop ], [ 1, %entry ]5644 %y = phi i32 [ %c6, %loop ], [ 42, %entry ]5645 %c6 = trunc i64 %indvars.iv to i325646 %c8 = mul i32 %x, %c65647 %c9 = add i32 %c8, 425648 %c10 = add i32 %y, %c65649 %c11 = add i32 %c10, %c95650 %c12 = sext i32 %c11 to i645651 %c13 = add i64 %c5, %c125652 %indvars.iv.tr = trunc i64 %indvars.iv to i325653 %c14 = shl i32 %indvars.iv.tr, 15654 %c15 = add i32 %c9, %c145655 %c16 = sext i32 %c15 to i645656 %c23 = add i64 %c13, %c165657 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 15658 %c24 = add nuw nsw i32 %x, 15659 %exitcond.i = icmp eq i64 %indvars.iv.next, 1145660 br i1 %exitcond.i, label %exit, label %loop5661 5662}5663 5664; Test case for PR52460.5665define void @pr52460_first_order_recurrence_truncated_iv(ptr noalias %src, ptr %dst) {5666;5667; CHECK-LABEL: @pr52460_first_order_recurrence_truncated_iv(5668; CHECK-NEXT: entry:5669; CHECK-NEXT: br label [[VECTOR_PH:%.*]]5670; CHECK: vector.ph:5671; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]5672; CHECK: vector.body:5673; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]5674; CHECK-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[VEC_IND:%.*]], [[VECTOR_BODY]] ]5675; CHECK-NEXT: [[VEC_IND]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]5676; CHECK-NEXT: [[TMP0:%.*]] = trunc i64 [[INDEX]] to i325677; CHECK-NEXT: [[TMP2:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>5678; CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[SRC:%.*]], align 45679; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP3]], i64 05680; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer5681; CHECK-NEXT: [[TMP4:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT]], [[TMP2]]5682; CHECK-NEXT: [[TMP5:%.*]] = getelementptr i32, ptr [[DST:%.*]], i32 [[TMP0]]5683; CHECK-NEXT: [[TMP6:%.*]] = add <2 x i32> [[VEC_IND]], [[TMP4]]5684; CHECK-NEXT: store <2 x i32> [[TMP6]], ptr [[TMP5]], align 45685; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 25686; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)5687; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1005688; CHECK-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP47:![0-9]+]]5689; CHECK: middle.block:5690; CHECK-NEXT: br label [[LOOP:%.*]]5691; CHECK: exit:5692; CHECK-NEXT: ret void5693;5694; IND-LABEL: @pr52460_first_order_recurrence_truncated_iv(5695; IND-NEXT: entry:5696; IND-NEXT: br label [[VECTOR_PH:%.*]]5697; IND: vector.ph:5698; IND-NEXT: br label [[VECTOR_BODY:%.*]]5699; IND: vector.body:5700; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]5701; IND-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[VEC_IND:%.*]], [[VECTOR_BODY]] ]5702; IND-NEXT: [[VEC_IND]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]5703; IND-NEXT: [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>5704; IND-NEXT: [[TMP1:%.*]] = load i32, ptr [[SRC:%.*]], align 45705; IND-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP1]], i64 05706; IND-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer5707; IND-NEXT: [[TMP2:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT]], [[TMP0]]5708; IND-NEXT: [[SEXT:%.*]] = shl i64 [[INDEX]], 325709; IND-NEXT: [[TMP3:%.*]] = ashr exact i64 [[SEXT]], 305710; IND-NEXT: [[TMP4:%.*]] = getelementptr i8, ptr [[DST:%.*]], i64 [[TMP3]]5711; IND-NEXT: [[TMP5:%.*]] = add <2 x i32> [[VEC_IND]], [[TMP2]]5712; IND-NEXT: store <2 x i32> [[TMP5]], ptr [[TMP4]], align 45713; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 25714; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)5715; IND-NEXT: [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1005716; IND-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP47:![0-9]+]]5717; IND: middle.block:5718; IND-NEXT: br label [[LOOP:%.*]]5719; IND: exit:5720; IND-NEXT: ret void5721;5722; UNROLL-LABEL: @pr52460_first_order_recurrence_truncated_iv(5723; UNROLL-NEXT: entry:5724; UNROLL-NEXT: br label [[VECTOR_PH:%.*]]5725; UNROLL: vector.ph:5726; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]5727; UNROLL: vector.body:5728; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]5729; UNROLL-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD:%.*]], [[VECTOR_BODY]] ]5730; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]5731; UNROLL-NEXT: [[STEP_ADD]] = add <2 x i32> [[VEC_IND]], splat (i32 2)5732; UNROLL-NEXT: [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>5733; UNROLL-NEXT: [[TMP1:%.*]] = shufflevector <2 x i32> [[VEC_IND]], <2 x i32> [[STEP_ADD]], <2 x i32> <i32 1, i32 2>5734; UNROLL-NEXT: [[TMP2:%.*]] = load i32, ptr [[SRC:%.*]], align 45735; UNROLL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP2]], i64 05736; UNROLL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer5737; UNROLL-NEXT: [[TMP3:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT]], [[TMP0]]5738; UNROLL-NEXT: [[TMP4:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT]], [[TMP1]]5739; UNROLL-NEXT: [[SEXT:%.*]] = shl i64 [[INDEX]], 325740; UNROLL-NEXT: [[TMP5:%.*]] = ashr exact i64 [[SEXT]], 305741; UNROLL-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr [[DST:%.*]], i64 [[TMP5]]5742; UNROLL-NEXT: [[TMP7:%.*]] = add <2 x i32> [[VEC_IND]], [[TMP3]]5743; UNROLL-NEXT: [[TMP8:%.*]] = add <2 x i32> [[STEP_ADD]], [[TMP4]]5744; UNROLL-NEXT: [[TMP9:%.*]] = getelementptr i8, ptr [[TMP6]], i64 85745; UNROLL-NEXT: store <2 x i32> [[TMP7]], ptr [[TMP6]], align 45746; UNROLL-NEXT: store <2 x i32> [[TMP8]], ptr [[TMP9]], align 45747; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 45748; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 4)5749; UNROLL-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1005750; UNROLL-NEXT: br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP47:![0-9]+]]5751; UNROLL: middle.block:5752; UNROLL-NEXT: br label [[LOOP:%.*]]5753; UNROLL: exit:5754; UNROLL-NEXT: ret void5755;5756; UNROLL-NO-IC-LABEL: @pr52460_first_order_recurrence_truncated_iv(5757; UNROLL-NO-IC-NEXT: entry:5758; UNROLL-NO-IC-NEXT: br label [[VECTOR_PH:%.*]]5759; UNROLL-NO-IC: vector.ph:5760; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]5761; UNROLL-NO-IC: vector.body:5762; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]5763; UNROLL-NO-IC-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD:%.*]], [[VECTOR_BODY]] ]5764; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]5765; UNROLL-NO-IC-NEXT: [[STEP_ADD]] = add <2 x i32> [[VEC_IND]], splat (i32 2)5766; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = trunc i64 [[INDEX]] to i325767; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>5768; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = shufflevector <2 x i32> [[VEC_IND]], <2 x i32> [[STEP_ADD]], <2 x i32> <i32 1, i32 2>5769; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = load i32, ptr [[SRC:%.*]], align 45770; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP4]], i64 05771; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer5772; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT]], [[TMP2]]5773; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT]], [[TMP3]]5774; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = getelementptr i32, ptr [[DST:%.*]], i32 [[TMP0]]5775; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = add <2 x i32> [[VEC_IND]], [[TMP5]]5776; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = add <2 x i32> [[STEP_ADD]], [[TMP6]]5777; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = getelementptr i32, ptr [[TMP7]], i64 25778; UNROLL-NO-IC-NEXT: store <2 x i32> [[TMP8]], ptr [[TMP7]], align 45779; UNROLL-NO-IC-NEXT: store <2 x i32> [[TMP9]], ptr [[TMP11]], align 45780; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 45781; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], splat (i32 2)5782; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1005783; UNROLL-NO-IC-NEXT: br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP47:![0-9]+]]5784; UNROLL-NO-IC: middle.block:5785; UNROLL-NO-IC-NEXT: br label [[LOOP:%.*]]5786; UNROLL-NO-IC: exit:5787; UNROLL-NO-IC-NEXT: ret void5788;5789; INTERLEAVE-LABEL: @pr52460_first_order_recurrence_truncated_iv(5790; INTERLEAVE-NEXT: entry:5791; INTERLEAVE-NEXT: br label [[VECTOR_PH:%.*]]5792; INTERLEAVE: vector.ph:5793; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]5794; INTERLEAVE: vector.body:5795; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]5796; INTERLEAVE-NEXT: [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD:%.*]], [[VECTOR_BODY]] ]5797; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]5798; INTERLEAVE-NEXT: [[STEP_ADD]] = add <4 x i32> [[VEC_IND]], splat (i32 4)5799; INTERLEAVE-NEXT: [[TMP0:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[VEC_IND]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>5800; INTERLEAVE-NEXT: [[TMP1:%.*]] = shufflevector <4 x i32> [[VEC_IND]], <4 x i32> [[STEP_ADD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>5801; INTERLEAVE-NEXT: [[TMP2:%.*]] = load i32, ptr [[SRC:%.*]], align 45802; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TMP2]], i64 05803; INTERLEAVE-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer5804; INTERLEAVE-NEXT: [[TMP3:%.*]] = mul nsw <4 x i32> [[BROADCAST_SPLAT]], [[TMP0]]5805; INTERLEAVE-NEXT: [[TMP4:%.*]] = mul nsw <4 x i32> [[BROADCAST_SPLAT]], [[TMP1]]5806; INTERLEAVE-NEXT: [[SEXT:%.*]] = shl i64 [[INDEX]], 325807; INTERLEAVE-NEXT: [[TMP5:%.*]] = ashr exact i64 [[SEXT]], 305808; INTERLEAVE-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr [[DST:%.*]], i64 [[TMP5]]5809; INTERLEAVE-NEXT: [[TMP7:%.*]] = add <4 x i32> [[VEC_IND]], [[TMP3]]5810; INTERLEAVE-NEXT: [[TMP8:%.*]] = add <4 x i32> [[STEP_ADD]], [[TMP4]]5811; INTERLEAVE-NEXT: [[TMP9:%.*]] = getelementptr i8, ptr [[TMP6]], i64 165812; INTERLEAVE-NEXT: store <4 x i32> [[TMP7]], ptr [[TMP6]], align 45813; INTERLEAVE-NEXT: store <4 x i32> [[TMP8]], ptr [[TMP9]], align 45814; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 85815; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 8)5816; INTERLEAVE-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 965817; INTERLEAVE-NEXT: br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]5818; INTERLEAVE: middle.block:5819; INTERLEAVE-NEXT: br label [[SCALAR_PH:%.*]]5820; INTERLEAVE: scalar.ph:5821; INTERLEAVE-NEXT: [[SCALAR_RECUR_INIT:%.*]] = extractelement <4 x i32> [[STEP_ADD]], i64 35822; INTERLEAVE-NEXT: br label [[LOOP:%.*]]5823; INTERLEAVE: loop:5824; INTERLEAVE-NEXT: [[IV:%.*]] = phi i64 [ 96, [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]5825; INTERLEAVE-NEXT: [[TRUNC_IV:%.*]] = phi i32 [ 96, [[SCALAR_PH]] ], [ [[TRUNC_IV_NEXT:%.*]], [[LOOP]] ]5826; INTERLEAVE-NEXT: [[RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_TRUNC:%.*]], [[LOOP]] ]5827; INTERLEAVE-NEXT: [[LV:%.*]] = load i32, ptr [[SRC]], align 45828; INTERLEAVE-NEXT: [[MUL:%.*]] = mul nsw i32 [[LV]], [[RECUR]]5829; INTERLEAVE-NEXT: [[TRUNC_IV_NEXT]] = add i32 [[TRUNC_IV]], 15830; INTERLEAVE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 15831; INTERLEAVE-NEXT: [[IV_TRUNC]] = trunc i64 [[IV]] to i325832; INTERLEAVE-NEXT: [[SEXT2:%.*]] = shl i64 [[IV]], 325833; INTERLEAVE-NEXT: [[TMP11:%.*]] = ashr exact i64 [[SEXT2]], 305834; INTERLEAVE-NEXT: [[DST_GEP:%.*]] = getelementptr i8, ptr [[DST]], i64 [[TMP11]]5835; INTERLEAVE-NEXT: [[ADD:%.*]] = add i32 [[MUL]], [[IV_TRUNC]]5836; INTERLEAVE-NEXT: store i32 [[ADD]], ptr [[DST_GEP]], align 45837; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[TRUNC_IV_NEXT]], 1005838; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP49:![0-9]+]]5839; INTERLEAVE: exit:5840; INTERLEAVE-NEXT: ret void5841;5842entry:5843 br label %loop5844 5845loop:5846 %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]5847 %trunc.iv = phi i32 [ 0, %entry ], [ %trunc.iv.next, %loop ]5848 %recur = phi i32 [ 0, %entry ], [ %iv.trunc, %loop ]5849 %lv = load i32, ptr %src, align 45850 %mul = mul nsw i32 %lv, %recur5851 %trunc.iv.next = add i32 %trunc.iv, 15852 %iv.next = add nuw nsw i64 %iv, 15853 %iv.trunc = trunc i64 %iv to i325854 %dst.gep = getelementptr i32, ptr %dst, i32 %iv.trunc5855 %add = add i32 %iv.trunc, %mul5856 store i32 %add, ptr %dst.gep5857 %exitcond = icmp eq i32 %trunc.iv.next, 1005858 br i1 %exitcond, label %exit, label %loop5859 5860exit:5861 ret void5862}5863 5864; Test case where %iv.2.ext and %iv.2.conv become redundant due to the SCEV5865; predicates generated for the vector loop. They should be removed in the5866; vector loop.5867define void @test_optimized_cast_induction_feeding_first_order_recurrence(i64 %n, i32 %step, ptr %ptr) {5868;5869; CHECK-LABEL: @test_optimized_cast_induction_feeding_first_order_recurrence(5870; CHECK-NEXT: entry:5871; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 25872; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]5873; CHECK: vector.scevcheck:5874; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[N]], -15875; CHECK-NEXT: [[TMP1:%.*]] = trunc i32 [[STEP:%.*]] to i85876; CHECK-NEXT: [[TMP2:%.*]] = sub i8 0, [[TMP1]]5877; CHECK-NEXT: [[TMP3:%.*]] = icmp slt i8 [[TMP1]], 05878; CHECK-NEXT: [[TMP4:%.*]] = select i1 [[TMP3]], i8 [[TMP2]], i8 [[TMP1]]5879; CHECK-NEXT: [[TMP5:%.*]] = trunc i64 [[TMP0]] to i85880; CHECK-NEXT: [[MUL:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 [[TMP4]], i8 [[TMP5]])5881; CHECK-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i8, i1 } [[MUL]], 05882; CHECK-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i8, i1 } [[MUL]], 15883; CHECK-NEXT: [[TMP6:%.*]] = sub i8 0, [[MUL_RESULT]]5884; CHECK-NEXT: [[TMP7:%.*]] = icmp slt i8 [[MUL_RESULT]], 05885; CHECK-NEXT: [[TMP8:%.*]] = icmp sgt i8 [[TMP6]], 05886; CHECK-NEXT: [[TMP9:%.*]] = select i1 [[TMP3]], i1 [[TMP8]], i1 [[TMP7]]5887; CHECK-NEXT: [[TMP10:%.*]] = or i1 [[TMP9]], [[MUL_OVERFLOW]]5888; CHECK-NEXT: [[TMP11:%.*]] = icmp ugt i64 [[TMP0]], 2555889; CHECK-NEXT: [[TMP12:%.*]] = icmp ne i8 [[TMP1]], 05890; CHECK-NEXT: [[TMP13:%.*]] = and i1 [[TMP11]], [[TMP12]]5891; CHECK-NEXT: [[TMP14:%.*]] = or i1 [[TMP10]], [[TMP13]]5892; CHECK-NEXT: [[TMP15:%.*]] = sext i8 [[TMP1]] to i325893; CHECK-NEXT: [[IDENT_CHECK:%.*]] = icmp ne i32 [[STEP]], [[TMP15]]5894; CHECK-NEXT: [[TMP16:%.*]] = or i1 [[TMP14]], [[IDENT_CHECK]]5895; CHECK-NEXT: br i1 [[TMP16]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]5896; CHECK: vector.ph:5897; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 25898; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]5899; CHECK-NEXT: [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i325900; CHECK-NEXT: [[IND_END:%.*]] = mul i32 [[DOTCAST]], [[STEP]]5901; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[STEP]], i64 05902; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer5903; CHECK-NEXT: [[TMP19:%.*]] = mul nsw <2 x i32> <i32 0, i32 1>, [[DOTSPLAT]]5904; CHECK-NEXT: [[INDUCTION:%.*]] = add nsw <2 x i32> zeroinitializer, [[TMP19]]5905; CHECK-NEXT: [[TMP18:%.*]] = mul nsw i32 [[STEP]], 25906; CHECK-NEXT: [[DOTSPLATINSERT2:%.*]] = insertelement <2 x i32> poison, i32 [[TMP18]], i64 05907; CHECK-NEXT: [[DOTSPLAT3:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT2]], <2 x i32> poison, <2 x i32> zeroinitializer5908; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]5909; CHECK: vector.body:5910; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]5911; CHECK-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[VEC_IND:%.*]], [[VECTOR_BODY]] ]5912; CHECK-NEXT: [[VEC_IND]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]5913; CHECK-NEXT: [[TMP20:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>5914; CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i64 [[INDEX]]5915; CHECK-NEXT: store <2 x i32> [[TMP20]], ptr [[TMP21]], align 45916; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 25917; CHECK-NEXT: [[VEC_IND_NEXT]] = add nsw <2 x i32> [[VEC_IND]], [[DOTSPLAT3]]5918; CHECK-NEXT: [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]5919; CHECK-NEXT: br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]5920; CHECK: middle.block:5921; CHECK-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[VEC_IND]], i32 15922; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]5923; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]5924; CHECK: scalar.ph:5925; CHECK-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]5926; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]5927; CHECK-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]5928; CHECK-NEXT: br label [[LOOP:%.*]]5929; CHECK: loop:5930; CHECK-NEXT: [[FOR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_2_CONV:%.*]], [[LOOP]] ]5931; CHECK-NEXT: [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP]] ]5932; CHECK-NEXT: [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], [[LOOP]] ]5933; CHECK-NEXT: [[IV_2_EXT:%.*]] = shl i32 [[IV_2]], 245934; CHECK-NEXT: [[IV_2_CONV]] = ashr exact i32 [[IV_2_EXT]], 245935; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[PTR]], i64 [[IV_1]]5936; CHECK-NEXT: store i32 [[FOR]], ptr [[GEP]], align 45937; CHECK-NEXT: [[IV_2_NEXT]] = add nsw i32 [[IV_2_CONV]], [[STEP]]5938; CHECK-NEXT: [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 15939; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]5940; CHECK-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP49:![0-9]+]]5941; CHECK: exit:5942; CHECK-NEXT: ret void5943;5944; IND-LABEL: @test_optimized_cast_induction_feeding_first_order_recurrence(5945; IND-NEXT: entry:5946; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 25947; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]5948; IND: vector.scevcheck:5949; IND-NEXT: [[TMP0:%.*]] = add i64 [[N]], -15950; IND-NEXT: [[TMP1:%.*]] = trunc i32 [[STEP:%.*]] to i85951; IND-NEXT: [[TMP2:%.*]] = icmp slt i8 [[TMP1]], 05952; IND-NEXT: [[TMP3:%.*]] = call i8 @llvm.abs.i8(i8 [[TMP1]], i1 false)5953; IND-NEXT: [[TMP4:%.*]] = trunc i64 [[TMP0]] to i85954; IND-NEXT: [[MUL:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 [[TMP3]], i8 [[TMP4]])5955; IND-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i8, i1 } [[MUL]], 05956; IND-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i8, i1 } [[MUL]], 15957; IND-NEXT: [[TMP5:%.*]] = icmp slt i8 [[MUL_RESULT]], 05958; IND-NEXT: [[TMP6:%.*]] = icmp ugt i8 [[MUL_RESULT]], -1285959; IND-NEXT: [[TMP7:%.*]] = select i1 [[TMP2]], i1 [[TMP6]], i1 [[TMP5]]5960; IND-NEXT: [[TMP8:%.*]] = or i1 [[TMP7]], [[MUL_OVERFLOW]]5961; IND-NEXT: [[TMP9:%.*]] = icmp ugt i64 [[TMP0]], 2555962; IND-NEXT: [[TMP10:%.*]] = icmp ne i8 [[TMP1]], 05963; IND-NEXT: [[TMP11:%.*]] = and i1 [[TMP9]], [[TMP10]]5964; IND-NEXT: [[TMP12:%.*]] = or i1 [[TMP8]], [[TMP11]]5965; IND-NEXT: [[TMP13:%.*]] = add i32 [[STEP]], -1285966; IND-NEXT: [[IDENT_CHECK:%.*]] = icmp ult i32 [[TMP13]], -2565967; IND-NEXT: [[TMP14:%.*]] = or i1 [[TMP12]], [[IDENT_CHECK]]5968; IND-NEXT: br i1 [[TMP14]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]5969; IND: vector.ph:5970; IND-NEXT: [[N_VEC:%.*]] = and i64 [[N]], -25971; IND-NEXT: [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i325972; IND-NEXT: [[IND_END:%.*]] = mul i32 [[STEP]], [[DOTCAST]]5973; IND-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[STEP]], i64 05974; IND-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer5975; IND-NEXT: [[TMP15:%.*]] = mul nuw nsw <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>5976; IND-NEXT: [[TMP16:%.*]] = shl nsw i32 [[STEP]], 15977; IND-NEXT: [[DOTSPLATINSERT2:%.*]] = insertelement <2 x i32> poison, i32 [[TMP16]], i64 05978; IND-NEXT: [[DOTSPLAT3:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT2]], <2 x i32> poison, <2 x i32> zeroinitializer5979; IND-NEXT: br label [[VECTOR_BODY:%.*]]5980; IND: vector.body:5981; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]5982; IND-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[VEC_IND:%.*]], [[VECTOR_BODY]] ]5983; IND-NEXT: [[VEC_IND]] = phi <2 x i32> [ [[TMP15]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]5984; IND-NEXT: [[TMP17:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>5985; IND-NEXT: [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i64 [[INDEX]]5986; IND-NEXT: store <2 x i32> [[TMP17]], ptr [[TMP18]], align 45987; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 25988; IND-NEXT: [[VEC_IND_NEXT]] = add nsw <2 x i32> [[VEC_IND]], [[DOTSPLAT3]]5989; IND-NEXT: [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]5990; IND-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]5991; IND: middle.block:5992; IND-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[VEC_IND]], i64 15993; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]5994; IND-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]5995; IND: scalar.ph:5996; IND-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]5997; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]5998; IND-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]5999; IND-NEXT: br label [[LOOP:%.*]]6000; IND: loop:6001; IND-NEXT: [[FOR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_2_CONV:%.*]], [[LOOP]] ]6002; IND-NEXT: [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP]] ]6003; IND-NEXT: [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], [[LOOP]] ]6004; IND-NEXT: [[IV_2_EXT:%.*]] = shl i32 [[IV_2]], 246005; IND-NEXT: [[IV_2_CONV]] = ashr exact i32 [[IV_2_EXT]], 246006; IND-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[PTR]], i64 [[IV_1]]6007; IND-NEXT: store i32 [[FOR]], ptr [[GEP]], align 46008; IND-NEXT: [[IV_2_NEXT]] = add nsw i32 [[IV_2_CONV]], [[STEP]]6009; IND-NEXT: [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 16010; IND-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]6011; IND-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP49:![0-9]+]]6012; IND: exit:6013; IND-NEXT: ret void6014;6015; UNROLL-LABEL: @test_optimized_cast_induction_feeding_first_order_recurrence(6016; UNROLL-NEXT: entry:6017; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 46018; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]6019; UNROLL: vector.scevcheck:6020; UNROLL-NEXT: [[TMP0:%.*]] = add i64 [[N]], -16021; UNROLL-NEXT: [[TMP1:%.*]] = trunc i32 [[STEP:%.*]] to i86022; UNROLL-NEXT: [[TMP2:%.*]] = icmp slt i8 [[TMP1]], 06023; UNROLL-NEXT: [[TMP3:%.*]] = call i8 @llvm.abs.i8(i8 [[TMP1]], i1 false)6024; UNROLL-NEXT: [[TMP4:%.*]] = trunc i64 [[TMP0]] to i86025; UNROLL-NEXT: [[MUL:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 [[TMP3]], i8 [[TMP4]])6026; UNROLL-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i8, i1 } [[MUL]], 06027; UNROLL-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i8, i1 } [[MUL]], 16028; UNROLL-NEXT: [[TMP5:%.*]] = icmp slt i8 [[MUL_RESULT]], 06029; UNROLL-NEXT: [[TMP6:%.*]] = icmp ugt i8 [[MUL_RESULT]], -1286030; UNROLL-NEXT: [[TMP7:%.*]] = select i1 [[TMP2]], i1 [[TMP6]], i1 [[TMP5]]6031; UNROLL-NEXT: [[TMP8:%.*]] = or i1 [[TMP7]], [[MUL_OVERFLOW]]6032; UNROLL-NEXT: [[TMP9:%.*]] = icmp ugt i64 [[TMP0]], 2556033; UNROLL-NEXT: [[TMP10:%.*]] = icmp ne i8 [[TMP1]], 06034; UNROLL-NEXT: [[TMP11:%.*]] = and i1 [[TMP9]], [[TMP10]]6035; UNROLL-NEXT: [[TMP12:%.*]] = or i1 [[TMP8]], [[TMP11]]6036; UNROLL-NEXT: [[TMP13:%.*]] = add i32 [[STEP]], -1286037; UNROLL-NEXT: [[IDENT_CHECK:%.*]] = icmp ult i32 [[TMP13]], -2566038; UNROLL-NEXT: [[TMP14:%.*]] = or i1 [[TMP12]], [[IDENT_CHECK]]6039; UNROLL-NEXT: br i1 [[TMP14]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]6040; UNROLL: vector.ph:6041; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[N]], -46042; UNROLL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[STEP]], i64 06043; UNROLL-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer6044; UNROLL-NEXT: [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i326045; UNROLL-NEXT: [[IND_END:%.*]] = mul i32 [[STEP]], [[DOTCAST]]6046; UNROLL-NEXT: [[TMP16:%.*]] = shl <2 x i32> [[DOTSPLAT]], splat (i32 1)6047; UNROLL-NEXT: [[TMP17:%.*]] = mul nuw nsw <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>6048; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]6049; UNROLL: vector.body:6050; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]6051; UNROLL-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD:%.*]], [[VECTOR_BODY]] ]6052; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[TMP17]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]6053; UNROLL-NEXT: [[STEP_ADD]] = add <2 x i32> [[VEC_IND]], [[TMP16]]6054; UNROLL-NEXT: [[TMP18:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>6055; UNROLL-NEXT: [[TMP19:%.*]] = shufflevector <2 x i32> [[VEC_IND]], <2 x i32> [[STEP_ADD]], <2 x i32> <i32 1, i32 2>6056; UNROLL-NEXT: [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i64 [[INDEX]]6057; UNROLL-NEXT: [[TMP21:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP20]], i64 86058; UNROLL-NEXT: store <2 x i32> [[TMP18]], ptr [[TMP20]], align 46059; UNROLL-NEXT: store <2 x i32> [[TMP19]], ptr [[TMP21]], align 46060; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 46061; UNROLL-NEXT: [[VEC_IND_NEXT]] = add nsw <2 x i32> [[STEP_ADD]], [[TMP16]]6062; UNROLL-NEXT: [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]6063; UNROLL-NEXT: br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]6064; UNROLL: middle.block:6065; UNROLL-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[STEP_ADD]], i64 16066; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]6067; UNROLL-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]6068; UNROLL: scalar.ph:6069; UNROLL-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]6070; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]6071; UNROLL-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]6072; UNROLL-NEXT: br label [[LOOP:%.*]]6073; UNROLL: loop:6074; UNROLL-NEXT: [[FOR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_2_CONV:%.*]], [[LOOP]] ]6075; UNROLL-NEXT: [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP]] ]6076; UNROLL-NEXT: [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], [[LOOP]] ]6077; UNROLL-NEXT: [[IV_2_EXT:%.*]] = shl i32 [[IV_2]], 246078; UNROLL-NEXT: [[IV_2_CONV]] = ashr exact i32 [[IV_2_EXT]], 246079; UNROLL-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[PTR]], i64 [[IV_1]]6080; UNROLL-NEXT: store i32 [[FOR]], ptr [[GEP]], align 46081; UNROLL-NEXT: [[IV_2_NEXT]] = add nsw i32 [[IV_2_CONV]], [[STEP]]6082; UNROLL-NEXT: [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 16083; UNROLL-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]6084; UNROLL-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP49:![0-9]+]]6085; UNROLL: exit:6086; UNROLL-NEXT: ret void6087;6088; UNROLL-NO-IC-LABEL: @test_optimized_cast_induction_feeding_first_order_recurrence(6089; UNROLL-NO-IC-NEXT: entry:6090; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 46091; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]6092; UNROLL-NO-IC: vector.scevcheck:6093; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i64 [[N]], -16094; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = trunc i32 [[STEP:%.*]] to i86095; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = sub i8 0, [[TMP1]]6096; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = icmp slt i8 [[TMP1]], 06097; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = select i1 [[TMP3]], i8 [[TMP2]], i8 [[TMP1]]6098; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = trunc i64 [[TMP0]] to i86099; UNROLL-NO-IC-NEXT: [[MUL:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 [[TMP4]], i8 [[TMP5]])6100; UNROLL-NO-IC-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i8, i1 } [[MUL]], 06101; UNROLL-NO-IC-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i8, i1 } [[MUL]], 16102; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = sub i8 0, [[MUL_RESULT]]6103; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = icmp slt i8 [[MUL_RESULT]], 06104; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = icmp sgt i8 [[TMP6]], 06105; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = select i1 [[TMP3]], i1 [[TMP8]], i1 [[TMP7]]6106; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = or i1 [[TMP9]], [[MUL_OVERFLOW]]6107; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = icmp ugt i64 [[TMP0]], 2556108; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = icmp ne i8 [[TMP1]], 06109; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = and i1 [[TMP11]], [[TMP12]]6110; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = or i1 [[TMP10]], [[TMP13]]6111; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = sext i8 [[TMP1]] to i326112; UNROLL-NO-IC-NEXT: [[IDENT_CHECK:%.*]] = icmp ne i32 [[STEP]], [[TMP15]]6113; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = or i1 [[TMP14]], [[IDENT_CHECK]]6114; UNROLL-NO-IC-NEXT: br i1 [[TMP16]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]6115; UNROLL-NO-IC: vector.ph:6116; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 46117; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]6118; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[STEP]], i64 06119; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer6120; UNROLL-NO-IC-NEXT: [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i326121; UNROLL-NO-IC-NEXT: [[IND_END:%.*]] = mul i32 [[DOTCAST]], [[STEP]]6122; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = mul <2 x i32> splat (i32 2), [[BROADCAST_SPLAT]]6123; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = mul nsw <2 x i32> <i32 0, i32 1>, [[BROADCAST_SPLAT]]6124; UNROLL-NO-IC-NEXT: [[INDUCTION:%.*]] = add nsw <2 x i32> zeroinitializer, [[TMP19]]6125; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]6126; UNROLL-NO-IC: vector.body:6127; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]6128; UNROLL-NO-IC-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD:%.*]], [[VECTOR_BODY]] ]6129; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]6130; UNROLL-NO-IC-NEXT: [[STEP_ADD]] = add <2 x i32> [[VEC_IND]], [[TMP17]]6131; UNROLL-NO-IC-NEXT: [[TMP20:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>6132; UNROLL-NO-IC-NEXT: [[TMP21:%.*]] = shufflevector <2 x i32> [[VEC_IND]], <2 x i32> [[STEP_ADD]], <2 x i32> <i32 1, i32 2>6133; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i64 [[INDEX]]6134; UNROLL-NO-IC-NEXT: [[TMP24:%.*]] = getelementptr inbounds i32, ptr [[TMP22]], i64 26135; UNROLL-NO-IC-NEXT: store <2 x i32> [[TMP20]], ptr [[TMP22]], align 46136; UNROLL-NO-IC-NEXT: store <2 x i32> [[TMP21]], ptr [[TMP24]], align 46137; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 46138; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add nsw <2 x i32> [[STEP_ADD]], [[TMP17]]6139; UNROLL-NO-IC-NEXT: [[TMP25:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]6140; UNROLL-NO-IC-NEXT: br i1 [[TMP25]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]6141; UNROLL-NO-IC: middle.block:6142; UNROLL-NO-IC-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[STEP_ADD]], i32 16143; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]6144; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]6145; UNROLL-NO-IC: scalar.ph:6146; UNROLL-NO-IC-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]6147; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]6148; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]6149; UNROLL-NO-IC-NEXT: br label [[LOOP:%.*]]6150; UNROLL-NO-IC: loop:6151; UNROLL-NO-IC-NEXT: [[FOR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_2_CONV:%.*]], [[LOOP]] ]6152; UNROLL-NO-IC-NEXT: [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP]] ]6153; UNROLL-NO-IC-NEXT: [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], [[LOOP]] ]6154; UNROLL-NO-IC-NEXT: [[IV_2_EXT:%.*]] = shl i32 [[IV_2]], 246155; UNROLL-NO-IC-NEXT: [[IV_2_CONV]] = ashr exact i32 [[IV_2_EXT]], 246156; UNROLL-NO-IC-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[PTR]], i64 [[IV_1]]6157; UNROLL-NO-IC-NEXT: store i32 [[FOR]], ptr [[GEP]], align 46158; UNROLL-NO-IC-NEXT: [[IV_2_NEXT]] = add nsw i32 [[IV_2_CONV]], [[STEP]]6159; UNROLL-NO-IC-NEXT: [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 16160; UNROLL-NO-IC-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]6161; UNROLL-NO-IC-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP49:![0-9]+]]6162; UNROLL-NO-IC: exit:6163; UNROLL-NO-IC-NEXT: ret void6164;6165; INTERLEAVE-LABEL: @test_optimized_cast_induction_feeding_first_order_recurrence(6166; INTERLEAVE-NEXT: entry:6167; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 86168; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]6169; INTERLEAVE: vector.scevcheck:6170; INTERLEAVE-NEXT: [[TMP0:%.*]] = add i64 [[N]], -16171; INTERLEAVE-NEXT: [[TMP1:%.*]] = trunc i32 [[STEP:%.*]] to i86172; INTERLEAVE-NEXT: [[TMP2:%.*]] = icmp slt i8 [[TMP1]], 06173; INTERLEAVE-NEXT: [[TMP3:%.*]] = call i8 @llvm.abs.i8(i8 [[TMP1]], i1 false)6174; INTERLEAVE-NEXT: [[TMP4:%.*]] = trunc i64 [[TMP0]] to i86175; INTERLEAVE-NEXT: [[MUL:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 [[TMP3]], i8 [[TMP4]])6176; INTERLEAVE-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i8, i1 } [[MUL]], 06177; INTERLEAVE-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i8, i1 } [[MUL]], 16178; INTERLEAVE-NEXT: [[TMP5:%.*]] = icmp slt i8 [[MUL_RESULT]], 06179; INTERLEAVE-NEXT: [[TMP6:%.*]] = icmp ugt i8 [[MUL_RESULT]], -1286180; INTERLEAVE-NEXT: [[TMP7:%.*]] = select i1 [[TMP2]], i1 [[TMP6]], i1 [[TMP5]]6181; INTERLEAVE-NEXT: [[TMP8:%.*]] = or i1 [[TMP7]], [[MUL_OVERFLOW]]6182; INTERLEAVE-NEXT: [[TMP9:%.*]] = icmp ugt i64 [[TMP0]], 2556183; INTERLEAVE-NEXT: [[TMP10:%.*]] = icmp ne i8 [[TMP1]], 06184; INTERLEAVE-NEXT: [[TMP11:%.*]] = and i1 [[TMP9]], [[TMP10]]6185; INTERLEAVE-NEXT: [[TMP12:%.*]] = or i1 [[TMP8]], [[TMP11]]6186; INTERLEAVE-NEXT: [[TMP13:%.*]] = add i32 [[STEP]], -1286187; INTERLEAVE-NEXT: [[IDENT_CHECK:%.*]] = icmp ult i32 [[TMP13]], -2566188; INTERLEAVE-NEXT: [[TMP14:%.*]] = or i1 [[TMP12]], [[IDENT_CHECK]]6189; INTERLEAVE-NEXT: br i1 [[TMP14]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]6190; INTERLEAVE: vector.ph:6191; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i64 [[N]], -86192; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[STEP]], i64 06193; INTERLEAVE-NEXT: [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer6194; INTERLEAVE-NEXT: [[DOTCAST:%.*]] = trunc i64 [[N_VEC]] to i326195; INTERLEAVE-NEXT: [[IND_END:%.*]] = mul i32 [[STEP]], [[DOTCAST]]6196; INTERLEAVE-NEXT: [[TMP16:%.*]] = shl <4 x i32> [[DOTSPLAT]], splat (i32 2)6197; INTERLEAVE-NEXT: [[TMP17:%.*]] = mul nsw <4 x i32> [[DOTSPLAT]], <i32 0, i32 1, i32 2, i32 3>6198; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]6199; INTERLEAVE: vector.body:6200; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]6201; INTERLEAVE-NEXT: [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD:%.*]], [[VECTOR_BODY]] ]6202; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ [[TMP17]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]6203; INTERLEAVE-NEXT: [[STEP_ADD]] = add <4 x i32> [[VEC_IND]], [[TMP16]]6204; INTERLEAVE-NEXT: [[TMP18:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[VEC_IND]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>6205; INTERLEAVE-NEXT: [[TMP19:%.*]] = shufflevector <4 x i32> [[VEC_IND]], <4 x i32> [[STEP_ADD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>6206; INTERLEAVE-NEXT: [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i64 [[INDEX]]6207; INTERLEAVE-NEXT: [[TMP21:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP20]], i64 166208; INTERLEAVE-NEXT: store <4 x i32> [[TMP18]], ptr [[TMP20]], align 46209; INTERLEAVE-NEXT: store <4 x i32> [[TMP19]], ptr [[TMP21]], align 46210; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 86211; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add nsw <4 x i32> [[STEP_ADD]], [[TMP16]]6212; INTERLEAVE-NEXT: [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]6213; INTERLEAVE-NEXT: br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]6214; INTERLEAVE: middle.block:6215; INTERLEAVE-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[STEP_ADD]], i64 36216; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]6217; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]6218; INTERLEAVE: scalar.ph:6219; INTERLEAVE-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]6220; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]6221; INTERLEAVE-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]6222; INTERLEAVE-NEXT: br label [[LOOP:%.*]]6223; INTERLEAVE: loop:6224; INTERLEAVE-NEXT: [[FOR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_2_CONV:%.*]], [[LOOP]] ]6225; INTERLEAVE-NEXT: [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP]] ]6226; INTERLEAVE-NEXT: [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], [[LOOP]] ]6227; INTERLEAVE-NEXT: [[IV_2_EXT:%.*]] = shl i32 [[IV_2]], 246228; INTERLEAVE-NEXT: [[IV_2_CONV]] = ashr exact i32 [[IV_2_EXT]], 246229; INTERLEAVE-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[PTR]], i64 [[IV_1]]6230; INTERLEAVE-NEXT: store i32 [[FOR]], ptr [[GEP]], align 46231; INTERLEAVE-NEXT: [[IV_2_NEXT]] = add nsw i32 [[IV_2_CONV]], [[STEP]]6232; INTERLEAVE-NEXT: [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 16233; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]6234; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP51:![0-9]+]]6235; INTERLEAVE: exit:6236; INTERLEAVE-NEXT: ret void6237;6238entry:6239 br label %loop6240 6241loop:6242 %for = phi i32 [ 0, %entry ], [ %iv.2.conv, %loop ]6243 %iv.1 = phi i64 [ 0, %entry ], [ %iv.1.next, %loop ]6244 %iv.2 = phi i32 [ 0, %entry ], [ %iv.2.next, %loop ]6245 %iv.2.ext = shl i32 %iv.2, 246246 %iv.2.conv = ashr exact i32 %iv.2.ext, 246247 %gep = getelementptr inbounds i32, ptr %ptr, i64 %iv.16248 store i32 %for, ptr %gep, align 46249 %iv.2.next = add nsw i32 %iv.2.conv, %step6250 %iv.1.next = add nuw nsw i64 %iv.1, 16251 %exitcond = icmp eq i64 %iv.1.next, %n6252 br i1 %exitcond, label %exit, label %loop6253 6254exit:6255 ret void6256}6257