; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -loop-vectorize -force-vector-interleave=1 -force-vector-width=2 -S | FileCheck %s
; RUN: opt < %s -loop-vectorize -force-vector-interleave=1 -force-vector-width=2 -instcombine -S | FileCheck %s --check-prefix=IND
; RUN: opt < %s -loop-vectorize -force-vector-interleave=2 -force-vector-width=2 -instcombine -S | FileCheck %s --check-prefix=UNROLL
; RUN: opt < %s -loop-vectorize -force-vector-interleave=2 -force-vector-width=2 -S | FileCheck %s --check-prefix=UNROLL-NO-IC
; RUN: opt < %s -loop-vectorize -force-vector-interleave=2 -force-vector-width=4 -enable-interleaved-mem-accesses -instcombine -S | FileCheck %s --check-prefix=INTERLEAVE
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
; Make sure that we can handle multiple integer induction variables.
;
define void @multi_int_induction(i32* %A, i32 %N) {
; CHECK-LABEL: @multi_int_induction(
; CHECK-NEXT: for.body.lr.ph:
; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; CHECK-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; CHECK-NEXT: [[CAST_VTC:%.*]] = trunc i64 [[N_VEC]] to i32
; CHECK-NEXT: [[IND_END:%.*]] = add i32 190, [[CAST_VTC]]
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 190, i32 191>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP3]]
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP4]], i32 0
; CHECK-NEXT: [[TMP6:%.*]] = bitcast i32* [[TMP5]] to <2 x i32>*
; CHECK-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP6]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; CHECK-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH:%.*]] ]
; CHECK-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 190, [[FOR_BODY_LR_PH]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[COUNT_09:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
; CHECK-NEXT: store i32 [[COUNT_09]], i32* [[ARRAYIDX2]], align 4
; CHECK-NEXT: [[INC]] = add nsw i32 [[COUNT_09]], 1
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i32 [[LFTR_WIDEIV]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP2:![0-9]+]]
; CHECK: for.end:
; CHECK-NEXT: ret void
;
; IND-LABEL: @multi_int_induction(
; IND-NEXT: for.body.lr.ph:
; IND-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; IND-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; IND-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP0]], 0
; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934590
; IND-NEXT: [[CAST_VTC:%.*]] = trunc i64 [[N_VEC]] to i32
; IND-NEXT: [[IND_END:%.*]] = add i32 [[CAST_VTC]], 190
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 190, i32 191>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDEX]]
; IND-NEXT: [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <2 x i32>*
; IND-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP4]], align 4
; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; IND-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; IND-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; IND-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH:%.*]] ]
; IND-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 190, [[FOR_BODY_LR_PH]] ]
; IND-NEXT: br label [[FOR_BODY:%.*]]
; IND: for.body:
; IND-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; IND-NEXT: [[COUNT_09:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; IND-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
; IND-NEXT: store i32 [[COUNT_09]], i32* [[ARRAYIDX2]], align 4
; IND-NEXT: [[INC]] = add nsw i32 [[COUNT_09]], 1
; IND-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; IND-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; IND-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; IND-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
; IND: for.end:
; IND-NEXT: ret void
;
; UNROLL-LABEL: @multi_int_induction(
; UNROLL-NEXT: for.body.lr.ph:
; UNROLL-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; UNROLL-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; UNROLL-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3
; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934588
; UNROLL-NEXT: [[CAST_VTC:%.*]] = trunc i64 [[N_VEC]] to i32
; UNROLL-NEXT: [[IND_END:%.*]] = add i32 [[CAST_VTC]], 190
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 190, i32 191>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; UNROLL-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDEX]]
; UNROLL-NEXT: [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP4]], align 4
; UNROLL-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP3]], i64 2
; UNROLL-NEXT: [[TMP6:%.*]] = bitcast i32* [[TMP5]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP6]], align 4
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 4, i32 4>
; UNROLL-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH:%.*]] ]
; UNROLL-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 190, [[FOR_BODY_LR_PH]] ]
; UNROLL-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL: for.body:
; UNROLL-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; UNROLL-NEXT: [[COUNT_09:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; UNROLL-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
; UNROLL-NEXT: store i32 [[COUNT_09]], i32* [[ARRAYIDX2]], align 4
; UNROLL-NEXT: [[INC]] = add nsw i32 [[COUNT_09]], 1
; UNROLL-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; UNROLL-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; UNROLL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; UNROLL-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
; UNROLL: for.end:
; UNROLL-NEXT: ret void
;
; UNROLL-NO-IC-LABEL: @multi_int_induction(
; UNROLL-NO-IC-NEXT: for.body.lr.ph:
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT: [[CAST_VTC:%.*]] = trunc i64 [[N_VEC]] to i32
; UNROLL-NO-IC-NEXT: [[IND_END:%.*]] = add i32 190, [[CAST_VTC]]
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 190, i32 191>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 2
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP3]]
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP4]]
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP5]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = bitcast i32* [[TMP7]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP8]], align 4
; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP5]], i32 2
; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = bitcast i32* [[TMP9]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP10]], align 4
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH:%.*]] ]
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 190, [[FOR_BODY_LR_PH]] ]
; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC: for.body:
; UNROLL-NO-IC-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[COUNT_09:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
; UNROLL-NO-IC-NEXT: store i32 [[COUNT_09]], i32* [[ARRAYIDX2]], align 4
; UNROLL-NO-IC-NEXT: [[INC]] = add nsw i32 [[COUNT_09]], 1
; UNROLL-NO-IC-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; UNROLL-NO-IC-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; UNROLL-NO-IC-NEXT: [[EXITCOND:%.*]] = icmp ne i32 [[LFTR_WIDEIV]], [[N]]
; UNROLL-NO-IC-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP2:![0-9]+]]
; UNROLL-NO-IC: for.end:
; UNROLL-NO-IC-NEXT: ret void
;
; INTERLEAVE-LABEL: @multi_int_induction(
; INTERLEAVE-NEXT: for.body.lr.ph:
; INTERLEAVE-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; INTERLEAVE-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; INTERLEAVE-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 7
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934584
; INTERLEAVE-NEXT: [[CAST_VTC:%.*]] = trunc i64 [[N_VEC]] to i32
; INTERLEAVE-NEXT: [[IND_END:%.*]] = add i32 [[CAST_VTC]], 190
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 190, i32 191, i32 192, i32 193>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; INTERLEAVE-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDEX]]
; INTERLEAVE-NEXT: [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[VEC_IND]], <4 x i32>* [[TMP4]], align 4
; INTERLEAVE-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP3]], i64 4
; INTERLEAVE-NEXT: [[TMP6:%.*]] = bitcast i32* [[TMP5]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[STEP_ADD]], <4 x i32>* [[TMP6]], align 4
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 8, i32 8, i32 8, i32 8>
; INTERLEAVE-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_LR_PH:%.*]] ]
; INTERLEAVE-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 190, [[FOR_BODY_LR_PH]] ]
; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]
; INTERLEAVE: for.body:
; INTERLEAVE-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; INTERLEAVE-NEXT: [[COUNT_09:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; INTERLEAVE-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
; INTERLEAVE-NEXT: store i32 [[COUNT_09]], i32* [[ARRAYIDX2]], align 4
; INTERLEAVE-NEXT: [[INC]] = add nsw i32 [[COUNT_09]], 1
; INTERLEAVE-NEXT: [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
; INTERLEAVE-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; INTERLEAVE-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; INTERLEAVE-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
; INTERLEAVE: for.end:
; INTERLEAVE-NEXT: ret void
;
for.body.lr.ph:
br label %for.body
for.body:
%indvars.iv = phi i64 [ 0, %for.body.lr.ph ], [ %indvars.iv.next, %for.body ]
%count.09 = phi i32 [ 190, %for.body.lr.ph ], [ %inc, %for.body ]
%arrayidx2 = getelementptr inbounds i32, i32* %A, i64 %indvars.iv
store i32 %count.09, i32* %arrayidx2, align 4
%inc = add nsw i32 %count.09, 1
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp ne i32 %lftr.wideiv, %N
br i1 %exitcond, label %for.body, label %for.end
for.end:
ret void
}
; Make sure we remove unneeded vectorization of induction variables.
; In order for instcombine to cleanup the vectorized induction variables that we
; create in the loop vectorizer we need to perform some form of redundancy
; elimination to get rid of multiple uses.
; Vectorized induction variable.
define void @scalar_use(float* %a, float %b, i64 %offset, i64 %offset2, i64 %n) {
; CHECK-LABEL: @scalar_use(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; CHECK: vector.memcheck:
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr float, float* [[A:%.*]], i64 [[OFFSET:%.*]]
; CHECK-NEXT: [[SCEVGEP1:%.*]] = bitcast float* [[SCEVGEP]] to i8*
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[N]], [[OFFSET]]
; CHECK-NEXT: [[SCEVGEP2:%.*]] = getelementptr float, float* [[A]], i64 [[TMP0]]
; CHECK-NEXT: [[SCEVGEP23:%.*]] = bitcast float* [[SCEVGEP2]] to i8*
; CHECK-NEXT: [[SCEVGEP4:%.*]] = getelementptr float, float* [[A]], i64 [[OFFSET2:%.*]]
; CHECK-NEXT: [[SCEVGEP45:%.*]] = bitcast float* [[SCEVGEP4]] to i8*
; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[N]], [[OFFSET2]]
; CHECK-NEXT: [[SCEVGEP6:%.*]] = getelementptr float, float* [[A]], i64 [[TMP1]]
; CHECK-NEXT: [[SCEVGEP67:%.*]] = bitcast float* [[SCEVGEP6]] to i8*
; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult i8* [[SCEVGEP1]], [[SCEVGEP67]]
; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP45]], [[SCEVGEP23]]
; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; CHECK-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[B:%.*]], i32 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT: [[TMP3:%.*]] = add i64 [[TMP2]], [[OFFSET]]
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP3]]
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds float, float* [[TMP4]], i32 0
; CHECK-NEXT: [[TMP6:%.*]] = bitcast float* [[TMP5]] to <2 x float>*
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x float>, <2 x float>* [[TMP6]], align 4, !alias.scope !4, !noalias !7
; CHECK-NEXT: [[TMP7:%.*]] = add i64 [[TMP2]], [[OFFSET2]]
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP7]]
; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, float* [[TMP8]], i32 0
; CHECK-NEXT: [[TMP10:%.*]] = bitcast float* [[TMP9]] to <2 x float>*
; CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <2 x float>, <2 x float>* [[TMP10]], align 4, !alias.scope !7
; CHECK-NEXT: [[TMP11:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD8]]
; CHECK-NEXT: [[TMP12:%.*]] = fadd fast <2 x float> [[WIDE_LOAD]], [[TMP11]]
; CHECK-NEXT: [[TMP13:%.*]] = bitcast float* [[TMP5]] to <2 x float>*
; CHECK-NEXT: store <2 x float> [[TMP12]], <2 x float>* [[TMP13]], align 4, !alias.scope !4, !noalias !7
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET]]
; CHECK-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM]]
; CHECK-NEXT: [[L1:%.*]] = load float, float* [[ARR_IDX]], align 4
; CHECK-NEXT: [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2]]
; CHECK-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM2]]
; CHECK-NEXT: [[L2:%.*]] = load float, float* [[ARR_IDX2]], align 4
; CHECK-NEXT: [[M:%.*]] = fmul fast float [[B]], [[L2]]
; CHECK-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]]
; CHECK-NEXT: store float [[AD]], float* [[ARR_IDX]], align 4
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
; CHECK: loopexit:
; CHECK-NEXT: ret void
;
; IND-LABEL: @scalar_use(
; IND-NEXT: entry:
; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2
; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; IND: vector.memcheck:
; IND-NEXT: [[SCEVGEP:%.*]] = getelementptr float, float* [[A:%.*]], i64 [[OFFSET:%.*]]
; IND-NEXT: [[TMP0:%.*]] = add i64 [[N]], [[OFFSET]]
; IND-NEXT: [[SCEVGEP2:%.*]] = getelementptr float, float* [[A]], i64 [[TMP0]]
; IND-NEXT: [[SCEVGEP4:%.*]] = getelementptr float, float* [[A]], i64 [[OFFSET2:%.*]]
; IND-NEXT: [[TMP1:%.*]] = add i64 [[N]], [[OFFSET2]]
; IND-NEXT: [[SCEVGEP6:%.*]] = getelementptr float, float* [[A]], i64 [[TMP1]]
; IND-NEXT: [[BOUND0:%.*]] = icmp ult float* [[SCEVGEP]], [[SCEVGEP6]]
; IND-NEXT: [[BOUND1:%.*]] = icmp ult float* [[SCEVGEP4]], [[SCEVGEP2]]
; IND-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; IND-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: [[N_VEC:%.*]] = and i64 [[N]], -2
; IND-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[B:%.*]], i64 0
; IND-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], [[OFFSET]]
; IND-NEXT: [[TMP3:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP2]]
; IND-NEXT: [[TMP4:%.*]] = bitcast float* [[TMP3]] to <2 x float>*
; IND-NEXT: [[WIDE_LOAD:%.*]] = load <2 x float>, <2 x float>* [[TMP4]], align 4, !alias.scope !4, !noalias !7
; IND-NEXT: [[TMP5:%.*]] = add i64 [[INDEX]], [[OFFSET2]]
; IND-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP5]]
; IND-NEXT: [[TMP7:%.*]] = bitcast float* [[TMP6]] to <2 x float>*
; IND-NEXT: [[WIDE_LOAD8:%.*]] = load <2 x float>, <2 x float>* [[TMP7]], align 4, !alias.scope !7
; IND-NEXT: [[TMP8:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD8]]
; IND-NEXT: [[TMP9:%.*]] = fadd fast <2 x float> [[WIDE_LOAD]], [[TMP8]]
; IND-NEXT: [[TMP10:%.*]] = bitcast float* [[TMP3]] to <2 x float>*
; IND-NEXT: store <2 x float> [[TMP9]], <2 x float>* [[TMP10]], align 4, !alias.scope !4, !noalias !7
; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; IND-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; IND-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
; IND-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
; IND-NEXT: br label [[FOR_BODY:%.*]]
; IND: for.body:
; IND-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; IND-NEXT: [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET]]
; IND-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM]]
; IND-NEXT: [[L1:%.*]] = load float, float* [[ARR_IDX]], align 4
; IND-NEXT: [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2]]
; IND-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM2]]
; IND-NEXT: [[L2:%.*]] = load float, float* [[ARR_IDX2]], align 4
; IND-NEXT: [[M:%.*]] = fmul fast float [[L2]], [[B]]
; IND-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]]
; IND-NEXT: store float [[AD]], float* [[ARR_IDX]], align 4
; IND-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; IND-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; IND-NEXT: br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
; IND: loopexit:
; IND-NEXT: ret void
;
; UNROLL-LABEL: @scalar_use(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; UNROLL: vector.memcheck:
; UNROLL-NEXT: [[SCEVGEP:%.*]] = getelementptr float, float* [[A:%.*]], i64 [[OFFSET:%.*]]
; UNROLL-NEXT: [[TMP0:%.*]] = add i64 [[N]], [[OFFSET]]
; UNROLL-NEXT: [[SCEVGEP2:%.*]] = getelementptr float, float* [[A]], i64 [[TMP0]]
; UNROLL-NEXT: [[SCEVGEP4:%.*]] = getelementptr float, float* [[A]], i64 [[OFFSET2:%.*]]
; UNROLL-NEXT: [[TMP1:%.*]] = add i64 [[N]], [[OFFSET2]]
; UNROLL-NEXT: [[SCEVGEP6:%.*]] = getelementptr float, float* [[A]], i64 [[TMP1]]
; UNROLL-NEXT: [[BOUND0:%.*]] = icmp ult float* [[SCEVGEP]], [[SCEVGEP6]]
; UNROLL-NEXT: [[BOUND1:%.*]] = icmp ult float* [[SCEVGEP4]], [[SCEVGEP2]]
; UNROLL-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; UNROLL-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[N]], -4
; UNROLL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[B:%.*]], i64 0
; UNROLL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer
; UNROLL-NEXT: [[BROADCAST_SPLATINSERT11:%.*]] = insertelement <2 x float> poison, float [[B]], i64 0
; UNROLL-NEXT: [[BROADCAST_SPLAT12:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT11]], <2 x float> poison, <2 x i32> zeroinitializer
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], [[OFFSET]]
; UNROLL-NEXT: [[TMP3:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP2]]
; UNROLL-NEXT: [[TMP4:%.*]] = bitcast float* [[TMP3]] to <2 x float>*
; UNROLL-NEXT: [[WIDE_LOAD:%.*]] = load <2 x float>, <2 x float>* [[TMP4]], align 4, !alias.scope !4, !noalias !7
; UNROLL-NEXT: [[TMP5:%.*]] = getelementptr inbounds float, float* [[TMP3]], i64 2
; UNROLL-NEXT: [[TMP6:%.*]] = bitcast float* [[TMP5]] to <2 x float>*
; UNROLL-NEXT: [[WIDE_LOAD8:%.*]] = load <2 x float>, <2 x float>* [[TMP6]], align 4, !alias.scope !4, !noalias !7
; UNROLL-NEXT: [[TMP7:%.*]] = add i64 [[INDEX]], [[OFFSET2]]
; UNROLL-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP7]]
; UNROLL-NEXT: [[TMP9:%.*]] = bitcast float* [[TMP8]] to <2 x float>*
; UNROLL-NEXT: [[WIDE_LOAD9:%.*]] = load <2 x float>, <2 x float>* [[TMP9]], align 4, !alias.scope !7
; UNROLL-NEXT: [[TMP10:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 2
; UNROLL-NEXT: [[TMP11:%.*]] = bitcast float* [[TMP10]] to <2 x float>*
; UNROLL-NEXT: [[WIDE_LOAD10:%.*]] = load <2 x float>, <2 x float>* [[TMP11]], align 4, !alias.scope !7
; UNROLL-NEXT: [[TMP12:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD9]]
; UNROLL-NEXT: [[TMP13:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT12]], [[WIDE_LOAD10]]
; UNROLL-NEXT: [[TMP14:%.*]] = fadd fast <2 x float> [[WIDE_LOAD]], [[TMP12]]
; UNROLL-NEXT: [[TMP15:%.*]] = fadd fast <2 x float> [[WIDE_LOAD8]], [[TMP13]]
; UNROLL-NEXT: [[TMP16:%.*]] = bitcast float* [[TMP3]] to <2 x float>*
; UNROLL-NEXT: store <2 x float> [[TMP14]], <2 x float>* [[TMP16]], align 4, !alias.scope !4, !noalias !7
; UNROLL-NEXT: [[TMP17:%.*]] = bitcast float* [[TMP5]] to <2 x float>*
; UNROLL-NEXT: store <2 x float> [[TMP15]], <2 x float>* [[TMP17]], align 4, !alias.scope !4, !noalias !7
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NEXT: [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
; UNROLL-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
; UNROLL-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL: for.body:
; UNROLL-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; UNROLL-NEXT: [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET]]
; UNROLL-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM]]
; UNROLL-NEXT: [[L1:%.*]] = load float, float* [[ARR_IDX]], align 4
; UNROLL-NEXT: [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2]]
; UNROLL-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM2]]
; UNROLL-NEXT: [[L2:%.*]] = load float, float* [[ARR_IDX2]], align 4
; UNROLL-NEXT: [[M:%.*]] = fmul fast float [[L2]], [[B]]
; UNROLL-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]]
; UNROLL-NEXT: store float [[AD]], float* [[ARR_IDX]], align 4
; UNROLL-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; UNROLL-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; UNROLL-NEXT: br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
; UNROLL: loopexit:
; UNROLL-NEXT: ret void
;
; UNROLL-NO-IC-LABEL: @scalar_use(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; UNROLL-NO-IC: vector.memcheck:
; UNROLL-NO-IC-NEXT: [[SCEVGEP:%.*]] = getelementptr float, float* [[A:%.*]], i64 [[OFFSET:%.*]]
; UNROLL-NO-IC-NEXT: [[SCEVGEP1:%.*]] = bitcast float* [[SCEVGEP]] to i8*
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i64 [[N]], [[OFFSET]]
; UNROLL-NO-IC-NEXT: [[SCEVGEP2:%.*]] = getelementptr float, float* [[A]], i64 [[TMP0]]
; UNROLL-NO-IC-NEXT: [[SCEVGEP23:%.*]] = bitcast float* [[SCEVGEP2]] to i8*
; UNROLL-NO-IC-NEXT: [[SCEVGEP4:%.*]] = getelementptr float, float* [[A]], i64 [[OFFSET2:%.*]]
; UNROLL-NO-IC-NEXT: [[SCEVGEP45:%.*]] = bitcast float* [[SCEVGEP4]] to i8*
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = add i64 [[N]], [[OFFSET2]]
; UNROLL-NO-IC-NEXT: [[SCEVGEP6:%.*]] = getelementptr float, float* [[A]], i64 [[TMP1]]
; UNROLL-NO-IC-NEXT: [[SCEVGEP67:%.*]] = bitcast float* [[SCEVGEP6]] to i8*
; UNROLL-NO-IC-NEXT: [[BOUND0:%.*]] = icmp ult i8* [[SCEVGEP1]], [[SCEVGEP67]]
; UNROLL-NO-IC-NEXT: [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP45]], [[SCEVGEP23]]
; UNROLL-NO-IC-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; UNROLL-NO-IC-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x float> poison, float [[B:%.*]], i32 0
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT]], <2 x float> poison, <2 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT11:%.*]] = insertelement <2 x float> poison, float [[B]], i32 0
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT12:%.*]] = shufflevector <2 x float> [[BROADCAST_SPLATINSERT11]], <2 x float> poison, <2 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 2
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = add i64 [[TMP2]], [[OFFSET]]
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = add i64 [[TMP3]], [[OFFSET]]
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP4]]
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP5]]
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, float* [[TMP6]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = bitcast float* [[TMP8]] to <2 x float>*
; UNROLL-NO-IC-NEXT: [[WIDE_LOAD:%.*]] = load <2 x float>, <2 x float>* [[TMP9]], align 4, !alias.scope !4, !noalias !7
; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = getelementptr inbounds float, float* [[TMP6]], i32 2
; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = bitcast float* [[TMP10]] to <2 x float>*
; UNROLL-NO-IC-NEXT: [[WIDE_LOAD8:%.*]] = load <2 x float>, <2 x float>* [[TMP11]], align 4, !alias.scope !4, !noalias !7
; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = add i64 [[TMP2]], [[OFFSET2]]
; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = add i64 [[TMP3]], [[OFFSET2]]
; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP12]]
; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP13]]
; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = getelementptr inbounds float, float* [[TMP14]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = bitcast float* [[TMP16]] to <2 x float>*
; UNROLL-NO-IC-NEXT: [[WIDE_LOAD9:%.*]] = load <2 x float>, <2 x float>* [[TMP17]], align 4, !alias.scope !7
; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = getelementptr inbounds float, float* [[TMP14]], i32 2
; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = bitcast float* [[TMP18]] to <2 x float>*
; UNROLL-NO-IC-NEXT: [[WIDE_LOAD10:%.*]] = load <2 x float>, <2 x float>* [[TMP19]], align 4, !alias.scope !7
; UNROLL-NO-IC-NEXT: [[TMP20:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD9]]
; UNROLL-NO-IC-NEXT: [[TMP21:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT12]], [[WIDE_LOAD10]]
; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = fadd fast <2 x float> [[WIDE_LOAD]], [[TMP20]]
; UNROLL-NO-IC-NEXT: [[TMP23:%.*]] = fadd fast <2 x float> [[WIDE_LOAD8]], [[TMP21]]
; UNROLL-NO-IC-NEXT: [[TMP24:%.*]] = bitcast float* [[TMP8]] to <2 x float>*
; UNROLL-NO-IC-NEXT: store <2 x float> [[TMP22]], <2 x float>* [[TMP24]], align 4, !alias.scope !4, !noalias !7
; UNROLL-NO-IC-NEXT: [[TMP25:%.*]] = bitcast float* [[TMP10]] to <2 x float>*
; UNROLL-NO-IC-NEXT: store <2 x float> [[TMP23]], <2 x float>* [[TMP25]], align 4, !alias.scope !4, !noalias !7
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC: for.body:
; UNROLL-NO-IC-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET]]
; UNROLL-NO-IC-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM]]
; UNROLL-NO-IC-NEXT: [[L1:%.*]] = load float, float* [[ARR_IDX]], align 4
; UNROLL-NO-IC-NEXT: [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2]]
; UNROLL-NO-IC-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM2]]
; UNROLL-NO-IC-NEXT: [[L2:%.*]] = load float, float* [[ARR_IDX2]], align 4
; UNROLL-NO-IC-NEXT: [[M:%.*]] = fmul fast float [[B]], [[L2]]
; UNROLL-NO-IC-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]]
; UNROLL-NO-IC-NEXT: store float [[AD]], float* [[ARR_IDX]], align 4
; UNROLL-NO-IC-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; UNROLL-NO-IC-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; UNROLL-NO-IC-NEXT: br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
; UNROLL-NO-IC: loopexit:
; UNROLL-NO-IC-NEXT: ret void
;
; INTERLEAVE-LABEL: @scalar_use(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; INTERLEAVE: vector.memcheck:
; INTERLEAVE-NEXT: [[SCEVGEP:%.*]] = getelementptr float, float* [[A:%.*]], i64 [[OFFSET:%.*]]
; INTERLEAVE-NEXT: [[TMP0:%.*]] = add i64 [[N]], [[OFFSET]]
; INTERLEAVE-NEXT: [[SCEVGEP2:%.*]] = getelementptr float, float* [[A]], i64 [[TMP0]]
; INTERLEAVE-NEXT: [[SCEVGEP4:%.*]] = getelementptr float, float* [[A]], i64 [[OFFSET2:%.*]]
; INTERLEAVE-NEXT: [[TMP1:%.*]] = add i64 [[N]], [[OFFSET2]]
; INTERLEAVE-NEXT: [[SCEVGEP6:%.*]] = getelementptr float, float* [[A]], i64 [[TMP1]]
; INTERLEAVE-NEXT: [[BOUND0:%.*]] = icmp ult float* [[SCEVGEP]], [[SCEVGEP6]]
; INTERLEAVE-NEXT: [[BOUND1:%.*]] = icmp ult float* [[SCEVGEP4]], [[SCEVGEP2]]
; INTERLEAVE-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; INTERLEAVE-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i64 [[N]], -8
; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x float> poison, float [[B:%.*]], i64 0
; INTERLEAVE-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x float> [[BROADCAST_SPLATINSERT]], <4 x float> poison, <4 x i32> zeroinitializer
; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT11:%.*]] = insertelement <4 x float> poison, float [[B]], i64 0
; INTERLEAVE-NEXT: [[BROADCAST_SPLAT12:%.*]] = shufflevector <4 x float> [[BROADCAST_SPLATINSERT11]], <4 x float> poison, <4 x i32> zeroinitializer
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], [[OFFSET]]
; INTERLEAVE-NEXT: [[TMP3:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP2]]
; INTERLEAVE-NEXT: [[TMP4:%.*]] = bitcast float* [[TMP3]] to <4 x float>*
; INTERLEAVE-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP4]], align 4, !alias.scope !4, !noalias !7
; INTERLEAVE-NEXT: [[TMP5:%.*]] = getelementptr inbounds float, float* [[TMP3]], i64 4
; INTERLEAVE-NEXT: [[TMP6:%.*]] = bitcast float* [[TMP5]] to <4 x float>*
; INTERLEAVE-NEXT: [[WIDE_LOAD8:%.*]] = load <4 x float>, <4 x float>* [[TMP6]], align 4, !alias.scope !4, !noalias !7
; INTERLEAVE-NEXT: [[TMP7:%.*]] = add i64 [[INDEX]], [[OFFSET2]]
; INTERLEAVE-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP7]]
; INTERLEAVE-NEXT: [[TMP9:%.*]] = bitcast float* [[TMP8]] to <4 x float>*
; INTERLEAVE-NEXT: [[WIDE_LOAD9:%.*]] = load <4 x float>, <4 x float>* [[TMP9]], align 4, !alias.scope !7
; INTERLEAVE-NEXT: [[TMP10:%.*]] = getelementptr inbounds float, float* [[TMP8]], i64 4
; INTERLEAVE-NEXT: [[TMP11:%.*]] = bitcast float* [[TMP10]] to <4 x float>*
; INTERLEAVE-NEXT: [[WIDE_LOAD10:%.*]] = load <4 x float>, <4 x float>* [[TMP11]], align 4, !alias.scope !7
; INTERLEAVE-NEXT: [[TMP12:%.*]] = fmul fast <4 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD9]]
; INTERLEAVE-NEXT: [[TMP13:%.*]] = fmul fast <4 x float> [[BROADCAST_SPLAT12]], [[WIDE_LOAD10]]
; INTERLEAVE-NEXT: [[TMP14:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], [[TMP12]]
; INTERLEAVE-NEXT: [[TMP15:%.*]] = fadd fast <4 x float> [[WIDE_LOAD8]], [[TMP13]]
; INTERLEAVE-NEXT: [[TMP16:%.*]] = bitcast float* [[TMP3]] to <4 x float>*
; INTERLEAVE-NEXT: store <4 x float> [[TMP14]], <4 x float>* [[TMP16]], align 4, !alias.scope !4, !noalias !7
; INTERLEAVE-NEXT: [[TMP17:%.*]] = bitcast float* [[TMP5]] to <4 x float>*
; INTERLEAVE-NEXT: store <4 x float> [[TMP15]], <4 x float>* [[TMP17]], align 4, !alias.scope !4, !noalias !7
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; INTERLEAVE-NEXT: [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]
; INTERLEAVE: for.body:
; INTERLEAVE-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; INTERLEAVE-NEXT: [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET]]
; INTERLEAVE-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM]]
; INTERLEAVE-NEXT: [[L1:%.*]] = load float, float* [[ARR_IDX]], align 4
; INTERLEAVE-NEXT: [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2]]
; INTERLEAVE-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[IND_SUM2]]
; INTERLEAVE-NEXT: [[L2:%.*]] = load float, float* [[ARR_IDX2]], align 4
; INTERLEAVE-NEXT: [[M:%.*]] = fmul fast float [[L2]], [[B]]
; INTERLEAVE-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]]
; INTERLEAVE-NEXT: store float [[AD]], float* [[ARR_IDX]], align 4
; INTERLEAVE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
; INTERLEAVE: loopexit:
; INTERLEAVE-NEXT: ret void
;
entry:
br label %for.body
for.body:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
%ind.sum = add i64 %iv, %offset
%arr.idx = getelementptr inbounds float, float* %a, i64 %ind.sum
%l1 = load float, float* %arr.idx, align 4
%ind.sum2 = add i64 %iv, %offset2
%arr.idx2 = getelementptr inbounds float, float* %a, i64 %ind.sum2
%l2 = load float, float* %arr.idx2, align 4
%m = fmul fast float %b, %l2
%ad = fadd fast float %l1, %m
store float %ad, float* %arr.idx, align 4
%iv.next = add nuw nsw i64 %iv, 1
%exitcond = icmp eq i64 %iv.next, %n
br i1 %exitcond, label %loopexit, label %for.body
loopexit:
ret void
}
; Make sure we don't create a vector induction phi node that is unused.
; Scalarize the step vectors instead.
;
; for (int i = 0; i < n; ++i)
; sum += a[i];
;
;
;
;
define i64 @scalarize_induction_variable_01(i64 *%a, i64 %n) {
; CHECK-LABEL: @scalarize_induction_variable_01(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i64, i64* [[A:%.*]], i64 [[TMP0]]
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i64, i64* [[TMP1]], i32 0
; CHECK-NEXT: [[TMP3:%.*]] = bitcast i64* [[TMP2]] to <2 x i64>*
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i64>, <2 x i64>* [[TMP3]], align 8
; CHECK-NEXT: [[TMP4]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP6:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[TMP4]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[SUM:%.*]] = phi i64 [ [[TMP9:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 [[I]]
; CHECK-NEXT: [[TMP8:%.*]] = load i64, i64* [[TMP7]], align 8
; CHECK-NEXT: [[TMP9]] = add i64 [[TMP8]], [[SUM]]
; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; CHECK-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP12:![0-9]+]]
; CHECK: for.end:
; CHECK-NEXT: [[TMP10:%.*]] = phi i64 [ [[TMP9]], [[FOR_BODY]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i64 [[TMP10]]
;
; IND-LABEL: @scalarize_induction_variable_01(
; IND-NEXT: entry:
; IND-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2
; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775806
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP2:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[TMP0:%.*]] = getelementptr inbounds i64, i64* [[A:%.*]], i64 [[INDEX]]
; IND-NEXT: [[TMP1:%.*]] = bitcast i64* [[TMP0]] to <2 x i64>*
; IND-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i64>, <2 x i64>* [[TMP1]], align 8
; IND-NEXT: [[TMP2]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; IND-NEXT: [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; IND-NEXT: br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[TMP4:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[TMP2]])
; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
; IND-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; IND-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP4]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; IND-NEXT: br label [[FOR_BODY:%.*]]
; IND: for.body:
; IND-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; IND-NEXT: [[SUM:%.*]] = phi i64 [ [[TMP7:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; IND-NEXT: [[TMP5:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 [[I]]
; IND-NEXT: [[TMP6:%.*]] = load i64, i64* [[TMP5]], align 8
; IND-NEXT: [[TMP7]] = add i64 [[TMP6]], [[SUM]]
; IND-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; IND-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; IND-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP12:![0-9]+]]
; IND: for.end:
; IND-NEXT: [[TMP8:%.*]] = phi i64 [ [[TMP7]], [[FOR_BODY]] ], [ [[TMP4]], [[MIDDLE_BLOCK]] ]
; IND-NEXT: ret i64 [[TMP8]]
;
; UNROLL-LABEL: @scalarize_induction_variable_01(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 4
; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775804
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[TMP0:%.*]] = getelementptr inbounds i64, i64* [[A:%.*]], i64 [[INDEX]]
; UNROLL-NEXT: [[TMP1:%.*]] = bitcast i64* [[TMP0]] to <2 x i64>*
; UNROLL-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i64>, <2 x i64>* [[TMP1]], align 8
; UNROLL-NEXT: [[TMP2:%.*]] = getelementptr inbounds i64, i64* [[TMP0]], i64 2
; UNROLL-NEXT: [[TMP3:%.*]] = bitcast i64* [[TMP2]] to <2 x i64>*
; UNROLL-NEXT: [[WIDE_LOAD2:%.*]] = load <2 x i64>, <2 x i64>* [[TMP3]], align 8
; UNROLL-NEXT: [[TMP4]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
; UNROLL-NEXT: [[TMP5]] = add <2 x i64> [[WIDE_LOAD2]], [[VEC_PHI1]]
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NEXT: [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[BIN_RDX:%.*]] = add <2 x i64> [[TMP5]], [[TMP4]]
; UNROLL-NEXT: [[TMP7:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])
; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; UNROLL-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL: for.body:
; UNROLL-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[SUM:%.*]] = phi i64 [ [[TMP10:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[TMP8:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 [[I]]
; UNROLL-NEXT: [[TMP9:%.*]] = load i64, i64* [[TMP8]], align 8
; UNROLL-NEXT: [[TMP10]] = add i64 [[TMP9]], [[SUM]]
; UNROLL-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; UNROLL-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; UNROLL-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP12:![0-9]+]]
; UNROLL: for.end:
; UNROLL-NEXT: [[TMP11:%.*]] = phi i64 [ [[TMP10]], [[FOR_BODY]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
; UNROLL-NEXT: ret i64 [[TMP11]]
;
; UNROLL-NO-IC-LABEL: @scalarize_induction_variable_01(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 4
; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 4
; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 2
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = getelementptr inbounds i64, i64* [[A:%.*]], i64 [[TMP0]]
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 [[TMP1]]
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = getelementptr inbounds i64, i64* [[TMP2]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = bitcast i64* [[TMP4]] to <2 x i64>*
; UNROLL-NO-IC-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i64>, <2 x i64>* [[TMP5]], align 8
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = getelementptr inbounds i64, i64* [[TMP2]], i32 2
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = bitcast i64* [[TMP6]] to <2 x i64>*
; UNROLL-NO-IC-NEXT: [[WIDE_LOAD2:%.*]] = load <2 x i64>, <2 x i64>* [[TMP7]], align 8
; UNROLL-NO-IC-NEXT: [[TMP8]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
; UNROLL-NO-IC-NEXT: [[TMP9]] = add <2 x i64> [[WIDE_LOAD2]], [[VEC_PHI1]]
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[BIN_RDX:%.*]] = add <2 x i64> [[TMP9]], [[TMP8]]
; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; UNROLL-NO-IC-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[TMP11]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC: for.body:
; UNROLL-NO-IC-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[SUM:%.*]] = phi i64 [ [[TMP14:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 [[I]]
; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = load i64, i64* [[TMP12]], align 8
; UNROLL-NO-IC-NEXT: [[TMP14]] = add i64 [[TMP13]], [[SUM]]
; UNROLL-NO-IC-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; UNROLL-NO-IC-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; UNROLL-NO-IC-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP12:![0-9]+]]
; UNROLL-NO-IC: for.end:
; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = phi i64 [ [[TMP14]], [[FOR_BODY]] ], [ [[TMP11]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: ret i64 [[TMP15]]
;
; INTERLEAVE-LABEL: @scalarize_induction_variable_01(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 8
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775800
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_PHI:%.*]] = phi <4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_PHI1:%.*]] = phi <4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[TMP0:%.*]] = getelementptr inbounds i64, i64* [[A:%.*]], i64 [[INDEX]]
; INTERLEAVE-NEXT: [[TMP1:%.*]] = bitcast i64* [[TMP0]] to <4 x i64>*
; INTERLEAVE-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i64>, <4 x i64>* [[TMP1]], align 8
; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr inbounds i64, i64* [[TMP0]], i64 4
; INTERLEAVE-NEXT: [[TMP3:%.*]] = bitcast i64* [[TMP2]] to <4 x i64>*
; INTERLEAVE-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x i64>, <4 x i64>* [[TMP3]], align 8
; INTERLEAVE-NEXT: [[TMP4]] = add <4 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
; INTERLEAVE-NEXT: [[TMP5]] = add <4 x i64> [[WIDE_LOAD2]], [[VEC_PHI1]]
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; INTERLEAVE-NEXT: [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: [[BIN_RDX:%.*]] = add <4 x i64> [[TMP5]], [[TMP4]]
; INTERLEAVE-NEXT: [[TMP7:%.*]] = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> [[BIN_RDX]])
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; INTERLEAVE-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]
; INTERLEAVE: for.body:
; INTERLEAVE-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[SUM:%.*]] = phi i64 [ [[TMP10:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[TMP8:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 [[I]]
; INTERLEAVE-NEXT: [[TMP9:%.*]] = load i64, i64* [[TMP8]], align 8
; INTERLEAVE-NEXT: [[TMP10]] = add i64 [[TMP9]], [[SUM]]
; INTERLEAVE-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; INTERLEAVE-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; INTERLEAVE-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP12:![0-9]+]]
; INTERLEAVE: for.end:
; INTERLEAVE-NEXT: [[TMP11:%.*]] = phi i64 [ [[TMP10]], [[FOR_BODY]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: ret i64 [[TMP11]]
;
entry:
br label %for.body
for.body:
%i = phi i64 [ %i.next, %for.body ], [ 0, %entry ]
%sum = phi i64 [ %2, %for.body ], [ 0, %entry ]
%0 = getelementptr inbounds i64, i64* %a, i64 %i
%1 = load i64, i64* %0, align 8
%2 = add i64 %1, %sum
%i.next = add nuw nsw i64 %i, 1
%cond = icmp slt i64 %i.next, %n
br i1 %cond, label %for.body, label %for.end
for.end:
%3 = phi i64 [ %2, %for.body ]
ret i64 %3
}
; Make sure we scalarize the step vectors used for the pointer arithmetic. We
; can't easily simplify vectorized step vectors.
;
; float s = 0;
; for (int i ; 0; i < n; i += 8)
; s += (a[i] + b[i] + 1.0f);
;
;
;
;
define float @scalarize_induction_variable_02(float* %a, float* %b, i64 %n) {
; CHECK-LABEL: @scalarize_induction_variable_02(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 8)
; CHECK-NEXT: [[TMP0:%.*]] = add nsw i64 [[SMAX]], -1
; CHECK-NEXT: [[TMP1:%.*]] = lshr i64 [[TMP0]], 3
; CHECK-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; CHECK-NEXT: [[IND_END:%.*]] = mul i64 [[N_VEC]], 8
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
; CHECK-NEXT: [[TMP3:%.*]] = add i64 [[OFFSET_IDX]], 0
; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 8
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[TMP3]]
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP4]]
; CHECK-NEXT: [[TMP7:%.*]] = load float, float* [[TMP5]], align 4
; CHECK-NEXT: [[TMP8:%.*]] = load float, float* [[TMP6]], align 4
; CHECK-NEXT: [[TMP9:%.*]] = insertelement <2 x float> poison, float [[TMP7]], i32 0
; CHECK-NEXT: [[TMP10:%.*]] = insertelement <2 x float> [[TMP9]], float [[TMP8]], i32 1
; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[TMP3]]
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP4]]
; CHECK-NEXT: [[TMP13:%.*]] = load float, float* [[TMP11]], align 4
; CHECK-NEXT: [[TMP14:%.*]] = load float, float* [[TMP12]], align 4
; CHECK-NEXT: [[TMP15:%.*]] = insertelement <2 x float> poison, float [[TMP13]], i32 0
; CHECK-NEXT: [[TMP16:%.*]] = insertelement <2 x float> [[TMP15]], float [[TMP14]], i32 1
; CHECK-NEXT: [[TMP17:%.*]] = fadd fast <2 x float> [[VEC_PHI]], <float 1.000000e+00, float 1.000000e+00>
; CHECK-NEXT: [[TMP18:%.*]] = fadd fast <2 x float> [[TMP17]], [[TMP10]]
; CHECK-NEXT: [[TMP19]] = fadd fast <2 x float> [[TMP18]], [[TMP16]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP21:%.*]] = call fast float @llvm.vector.reduce.fadd.v2f32(float -0.000000e+00, <2 x float> [[TMP19]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP21]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[S:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP28:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[I]]
; CHECK-NEXT: [[TMP23:%.*]] = load float, float* [[TMP22]], align 4
; CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[I]]
; CHECK-NEXT: [[TMP25:%.*]] = load float, float* [[TMP24]], align 4
; CHECK-NEXT: [[TMP26:%.*]] = fadd fast float [[S]], 1.000000e+00
; CHECK-NEXT: [[TMP27:%.*]] = fadd fast float [[TMP26]], [[TMP23]]
; CHECK-NEXT: [[TMP28]] = fadd fast float [[TMP27]], [[TMP25]]
; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 8
; CHECK-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP14:![0-9]+]]
; CHECK: for.end:
; CHECK-NEXT: [[S_LCSSA:%.*]] = phi float [ [[TMP28]], [[FOR_BODY]] ], [ [[TMP21]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret float [[S_LCSSA]]
;
; IND-LABEL: @scalarize_induction_variable_02(
; IND-NEXT: entry:
; IND-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 8)
; IND-NEXT: [[TMP0:%.*]] = add nsw i64 [[SMAX]], -1
; IND-NEXT: [[TMP1:%.*]] = lshr i64 [[TMP0]], 3
; IND-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 8
; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 4611686018427387902
; IND-NEXT: [[IND_END:%.*]] = shl i64 [[N_VEC]], 3
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VEC_PHI:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[OFFSET_IDX:%.*]] = shl i64 [[INDEX]], 3
; IND-NEXT: [[TMP3:%.*]] = or i64 [[OFFSET_IDX]], 8
; IND-NEXT: [[TMP4:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[OFFSET_IDX]]
; IND-NEXT: [[TMP5:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP3]]
; IND-NEXT: [[TMP6:%.*]] = load float, float* [[TMP4]], align 4
; IND-NEXT: [[TMP7:%.*]] = load float, float* [[TMP5]], align 4
; IND-NEXT: [[TMP8:%.*]] = insertelement <2 x float> poison, float [[TMP6]], i64 0
; IND-NEXT: [[TMP9:%.*]] = insertelement <2 x float> [[TMP8]], float [[TMP7]], i64 1
; IND-NEXT: [[TMP10:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[OFFSET_IDX]]
; IND-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP3]]
; IND-NEXT: [[TMP12:%.*]] = load float, float* [[TMP10]], align 4
; IND-NEXT: [[TMP13:%.*]] = load float, float* [[TMP11]], align 4
; IND-NEXT: [[TMP14:%.*]] = insertelement <2 x float> poison, float [[TMP12]], i64 0
; IND-NEXT: [[TMP15:%.*]] = insertelement <2 x float> [[TMP14]], float [[TMP13]], i64 1
; IND-NEXT: [[TMP16:%.*]] = fadd fast <2 x float> [[VEC_PHI]], <float 1.000000e+00, float 1.000000e+00>
; IND-NEXT: [[TMP17:%.*]] = fadd fast <2 x float> [[TMP16]], [[TMP9]]
; IND-NEXT: [[TMP18]] = fadd fast <2 x float> [[TMP17]], [[TMP15]]
; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; IND-NEXT: [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; IND-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[TMP20:%.*]] = call fast float @llvm.vector.reduce.fadd.v2f32(float -0.000000e+00, <2 x float> [[TMP18]])
; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; IND-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; IND-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP20]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
; IND-NEXT: br label [[FOR_BODY:%.*]]
; IND: for.body:
; IND-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_BODY]] ]
; IND-NEXT: [[S:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP27:%.*]], [[FOR_BODY]] ]
; IND-NEXT: [[TMP21:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[I]]
; IND-NEXT: [[TMP22:%.*]] = load float, float* [[TMP21]], align 4
; IND-NEXT: [[TMP23:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[I]]
; IND-NEXT: [[TMP24:%.*]] = load float, float* [[TMP23]], align 4
; IND-NEXT: [[TMP25:%.*]] = fadd fast float [[S]], 1.000000e+00
; IND-NEXT: [[TMP26:%.*]] = fadd fast float [[TMP25]], [[TMP22]]
; IND-NEXT: [[TMP27]] = fadd fast float [[TMP26]], [[TMP24]]
; IND-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 8
; IND-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; IND-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP14:![0-9]+]]
; IND: for.end:
; IND-NEXT: [[S_LCSSA:%.*]] = phi float [ [[TMP27]], [[FOR_BODY]] ], [ [[TMP20]], [[MIDDLE_BLOCK]] ]
; IND-NEXT: ret float [[S_LCSSA]]
;
; UNROLL-LABEL: @scalarize_induction_variable_02(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 8)
; UNROLL-NEXT: [[TMP0:%.*]] = add nsw i64 [[SMAX]], -1
; UNROLL-NEXT: [[TMP1:%.*]] = lshr i64 [[TMP0]], 3
; UNROLL-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 24
; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 4611686018427387900
; UNROLL-NEXT: [[IND_END:%.*]] = shl i64 [[N_VEC]], 3
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_PHI:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP34:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_PHI1:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP35:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[OFFSET_IDX:%.*]] = shl i64 [[INDEX]], 3
; UNROLL-NEXT: [[TMP3:%.*]] = or i64 [[OFFSET_IDX]], 8
; UNROLL-NEXT: [[TMP4:%.*]] = or i64 [[OFFSET_IDX]], 16
; UNROLL-NEXT: [[TMP5:%.*]] = or i64 [[OFFSET_IDX]], 24
; UNROLL-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[OFFSET_IDX]]
; UNROLL-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP3]]
; UNROLL-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP4]]
; UNROLL-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP5]]
; UNROLL-NEXT: [[TMP10:%.*]] = load float, float* [[TMP6]], align 4
; UNROLL-NEXT: [[TMP11:%.*]] = load float, float* [[TMP7]], align 4
; UNROLL-NEXT: [[TMP12:%.*]] = insertelement <2 x float> poison, float [[TMP10]], i64 0
; UNROLL-NEXT: [[TMP13:%.*]] = insertelement <2 x float> [[TMP12]], float [[TMP11]], i64 1
; UNROLL-NEXT: [[TMP14:%.*]] = load float, float* [[TMP8]], align 4
; UNROLL-NEXT: [[TMP15:%.*]] = load float, float* [[TMP9]], align 4
; UNROLL-NEXT: [[TMP16:%.*]] = insertelement <2 x float> poison, float [[TMP14]], i64 0
; UNROLL-NEXT: [[TMP17:%.*]] = insertelement <2 x float> [[TMP16]], float [[TMP15]], i64 1
; UNROLL-NEXT: [[TMP18:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[OFFSET_IDX]]
; UNROLL-NEXT: [[TMP19:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP3]]
; UNROLL-NEXT: [[TMP20:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP4]]
; UNROLL-NEXT: [[TMP21:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP5]]
; UNROLL-NEXT: [[TMP22:%.*]] = load float, float* [[TMP18]], align 4
; UNROLL-NEXT: [[TMP23:%.*]] = load float, float* [[TMP19]], align 4
; UNROLL-NEXT: [[TMP24:%.*]] = insertelement <2 x float> poison, float [[TMP22]], i64 0
; UNROLL-NEXT: [[TMP25:%.*]] = insertelement <2 x float> [[TMP24]], float [[TMP23]], i64 1
; UNROLL-NEXT: [[TMP26:%.*]] = load float, float* [[TMP20]], align 4
; UNROLL-NEXT: [[TMP27:%.*]] = load float, float* [[TMP21]], align 4
; UNROLL-NEXT: [[TMP28:%.*]] = insertelement <2 x float> poison, float [[TMP26]], i64 0
; UNROLL-NEXT: [[TMP29:%.*]] = insertelement <2 x float> [[TMP28]], float [[TMP27]], i64 1
; UNROLL-NEXT: [[TMP30:%.*]] = fadd fast <2 x float> [[VEC_PHI]], <float 1.000000e+00, float 1.000000e+00>
; UNROLL-NEXT: [[TMP31:%.*]] = fadd fast <2 x float> [[VEC_PHI1]], <float 1.000000e+00, float 1.000000e+00>
; UNROLL-NEXT: [[TMP32:%.*]] = fadd fast <2 x float> [[TMP30]], [[TMP13]]
; UNROLL-NEXT: [[TMP33:%.*]] = fadd fast <2 x float> [[TMP31]], [[TMP17]]
; UNROLL-NEXT: [[TMP34]] = fadd fast <2 x float> [[TMP32]], [[TMP25]]
; UNROLL-NEXT: [[TMP35]] = fadd fast <2 x float> [[TMP33]], [[TMP29]]
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NEXT: [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[TMP36]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[BIN_RDX:%.*]] = fadd fast <2 x float> [[TMP35]], [[TMP34]]
; UNROLL-NEXT: [[TMP37:%.*]] = call fast float @llvm.vector.reduce.fadd.v2f32(float -0.000000e+00, <2 x float> [[BIN_RDX]])
; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; UNROLL-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP37]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
; UNROLL-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL: for.body:
; UNROLL-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_BODY]] ]
; UNROLL-NEXT: [[S:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP44:%.*]], [[FOR_BODY]] ]
; UNROLL-NEXT: [[TMP38:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[I]]
; UNROLL-NEXT: [[TMP39:%.*]] = load float, float* [[TMP38]], align 4
; UNROLL-NEXT: [[TMP40:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[I]]
; UNROLL-NEXT: [[TMP41:%.*]] = load float, float* [[TMP40]], align 4
; UNROLL-NEXT: [[TMP42:%.*]] = fadd fast float [[S]], 1.000000e+00
; UNROLL-NEXT: [[TMP43:%.*]] = fadd fast float [[TMP42]], [[TMP39]]
; UNROLL-NEXT: [[TMP44]] = fadd fast float [[TMP43]], [[TMP41]]
; UNROLL-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 8
; UNROLL-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; UNROLL-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP14:![0-9]+]]
; UNROLL: for.end:
; UNROLL-NEXT: [[S_LCSSA:%.*]] = phi float [ [[TMP44]], [[FOR_BODY]] ], [ [[TMP37]], [[MIDDLE_BLOCK]] ]
; UNROLL-NEXT: ret float [[S_LCSSA]]
;
; UNROLL-NO-IC-LABEL: @scalarize_induction_variable_02(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 8)
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add nsw i64 [[SMAX]], -1
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = lshr i64 [[TMP0]], 3
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT: [[IND_END:%.*]] = mul i64 [[N_VEC]], 8
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP35:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI1:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP36:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = add i64 [[OFFSET_IDX]], 0
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 8
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = add i64 [[OFFSET_IDX]], 16
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = add i64 [[OFFSET_IDX]], 24
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[TMP3]]
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP4]]
; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP5]]
; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP6]]
; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = load float, float* [[TMP7]], align 4
; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = load float, float* [[TMP8]], align 4
; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = insertelement <2 x float> poison, float [[TMP11]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = insertelement <2 x float> [[TMP13]], float [[TMP12]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = load float, float* [[TMP9]], align 4
; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = load float, float* [[TMP10]], align 4
; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = insertelement <2 x float> poison, float [[TMP15]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = insertelement <2 x float> [[TMP17]], float [[TMP16]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[TMP3]]
; UNROLL-NO-IC-NEXT: [[TMP20:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP4]]
; UNROLL-NO-IC-NEXT: [[TMP21:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP5]]
; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP6]]
; UNROLL-NO-IC-NEXT: [[TMP23:%.*]] = load float, float* [[TMP19]], align 4
; UNROLL-NO-IC-NEXT: [[TMP24:%.*]] = load float, float* [[TMP20]], align 4
; UNROLL-NO-IC-NEXT: [[TMP25:%.*]] = insertelement <2 x float> poison, float [[TMP23]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP26:%.*]] = insertelement <2 x float> [[TMP25]], float [[TMP24]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP27:%.*]] = load float, float* [[TMP21]], align 4
; UNROLL-NO-IC-NEXT: [[TMP28:%.*]] = load float, float* [[TMP22]], align 4
; UNROLL-NO-IC-NEXT: [[TMP29:%.*]] = insertelement <2 x float> poison, float [[TMP27]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP30:%.*]] = insertelement <2 x float> [[TMP29]], float [[TMP28]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP31:%.*]] = fadd fast <2 x float> [[VEC_PHI]], <float 1.000000e+00, float 1.000000e+00>
; UNROLL-NO-IC-NEXT: [[TMP32:%.*]] = fadd fast <2 x float> [[VEC_PHI1]], <float 1.000000e+00, float 1.000000e+00>
; UNROLL-NO-IC-NEXT: [[TMP33:%.*]] = fadd fast <2 x float> [[TMP31]], [[TMP14]]
; UNROLL-NO-IC-NEXT: [[TMP34:%.*]] = fadd fast <2 x float> [[TMP32]], [[TMP18]]
; UNROLL-NO-IC-NEXT: [[TMP35]] = fadd fast <2 x float> [[TMP33]], [[TMP26]]
; UNROLL-NO-IC-NEXT: [[TMP36]] = fadd fast <2 x float> [[TMP34]], [[TMP30]]
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[TMP37:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP37]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[BIN_RDX:%.*]] = fadd fast <2 x float> [[TMP36]], [[TMP35]]
; UNROLL-NO-IC-NEXT: [[TMP38:%.*]] = call fast float @llvm.vector.reduce.fadd.v2f32(float -0.000000e+00, <2 x float> [[BIN_RDX]])
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; UNROLL-NO-IC-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[TMP38]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC: for.body:
; UNROLL-NO-IC-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[S:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP45:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[TMP39:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[I]]
; UNROLL-NO-IC-NEXT: [[TMP40:%.*]] = load float, float* [[TMP39]], align 4
; UNROLL-NO-IC-NEXT: [[TMP41:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[I]]
; UNROLL-NO-IC-NEXT: [[TMP42:%.*]] = load float, float* [[TMP41]], align 4
; UNROLL-NO-IC-NEXT: [[TMP43:%.*]] = fadd fast float [[S]], 1.000000e+00
; UNROLL-NO-IC-NEXT: [[TMP44:%.*]] = fadd fast float [[TMP43]], [[TMP40]]
; UNROLL-NO-IC-NEXT: [[TMP45]] = fadd fast float [[TMP44]], [[TMP42]]
; UNROLL-NO-IC-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 8
; UNROLL-NO-IC-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; UNROLL-NO-IC-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP14:![0-9]+]]
; UNROLL-NO-IC: for.end:
; UNROLL-NO-IC-NEXT: [[S_LCSSA:%.*]] = phi float [ [[TMP45]], [[FOR_BODY]] ], [ [[TMP38]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: ret float [[S_LCSSA]]
;
; INTERLEAVE-LABEL: @scalarize_induction_variable_02(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 8)
; INTERLEAVE-NEXT: [[TMP0:%.*]] = add nsw i64 [[SMAX]], -1
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 64
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: [[TMP1:%.*]] = lshr i64 [[TMP0]], 3
; INTERLEAVE-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; INTERLEAVE-NEXT: [[N_MOD_VF:%.*]] = and i64 [[TMP2]], 7
; INTERLEAVE-NEXT: [[TMP3:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
; INTERLEAVE-NEXT: [[TMP4:%.*]] = select i1 [[TMP3]], i64 8, i64 [[N_MOD_VF]]
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = sub nsw i64 [[TMP2]], [[TMP4]]
; INTERLEAVE-NEXT: [[IND_END:%.*]] = shl i64 [[N_VEC]], 3
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_PHI:%.*]] = phi <4 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_PHI1:%.*]] = phi <4 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[OFFSET_IDX:%.*]] = shl i64 [[INDEX]], 3
; INTERLEAVE-NEXT: [[TMP5:%.*]] = or i64 [[OFFSET_IDX]], 32
; INTERLEAVE-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[OFFSET_IDX]]
; INTERLEAVE-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP5]]
; INTERLEAVE-NEXT: [[TMP8:%.*]] = bitcast float* [[TMP6]] to <32 x float>*
; INTERLEAVE-NEXT: [[TMP9:%.*]] = bitcast float* [[TMP7]] to <32 x float>*
; INTERLEAVE-NEXT: [[WIDE_VEC:%.*]] = load <32 x float>, <32 x float>* [[TMP8]], align 4
; INTERLEAVE-NEXT: [[WIDE_VEC2:%.*]] = load <32 x float>, <32 x float>* [[TMP9]], align 4
; 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>
; 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>
; INTERLEAVE-NEXT: [[TMP10:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[OFFSET_IDX]]
; INTERLEAVE-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP5]]
; INTERLEAVE-NEXT: [[TMP12:%.*]] = bitcast float* [[TMP10]] to <32 x float>*
; INTERLEAVE-NEXT: [[TMP13:%.*]] = bitcast float* [[TMP11]] to <32 x float>*
; INTERLEAVE-NEXT: [[WIDE_VEC4:%.*]] = load <32 x float>, <32 x float>* [[TMP12]], align 4
; INTERLEAVE-NEXT: [[WIDE_VEC5:%.*]] = load <32 x float>, <32 x float>* [[TMP13]], align 4
; INTERLEAVE-NEXT: [[STRIDED_VEC6:%.*]] = shufflevector <32 x float> [[WIDE_VEC4]], <32 x float> poison, <4 x i32> <i32 0, i32 8, i32 16, i32 24>
; INTERLEAVE-NEXT: [[STRIDED_VEC7:%.*]] = shufflevector <32 x float> [[WIDE_VEC5]], <32 x float> poison, <4 x i32> <i32 0, i32 8, i32 16, i32 24>
; INTERLEAVE-NEXT: [[TMP14:%.*]] = fadd fast <4 x float> [[VEC_PHI]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>
; INTERLEAVE-NEXT: [[TMP15:%.*]] = fadd fast <4 x float> [[VEC_PHI1]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>
; INTERLEAVE-NEXT: [[TMP16:%.*]] = fadd fast <4 x float> [[TMP14]], [[STRIDED_VEC]]
; INTERLEAVE-NEXT: [[TMP17:%.*]] = fadd fast <4 x float> [[TMP15]], [[STRIDED_VEC3]]
; INTERLEAVE-NEXT: [[TMP18]] = fadd fast <4 x float> [[TMP16]], [[STRIDED_VEC6]]
; INTERLEAVE-NEXT: [[TMP19]] = fadd fast <4 x float> [[TMP17]], [[STRIDED_VEC7]]
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; INTERLEAVE-NEXT: [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: [[BIN_RDX:%.*]] = fadd fast <4 x float> [[TMP19]], [[TMP18]]
; INTERLEAVE-NEXT: [[TMP21:%.*]] = call fast float @llvm.vector.reduce.fadd.v4f32(float -0.000000e+00, <4 x float> [[BIN_RDX]])
; INTERLEAVE-NEXT: br label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; INTERLEAVE-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP21]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]
; INTERLEAVE: for.body:
; INTERLEAVE-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_BODY]] ]
; INTERLEAVE-NEXT: [[S:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP28:%.*]], [[FOR_BODY]] ]
; INTERLEAVE-NEXT: [[TMP22:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[I]]
; INTERLEAVE-NEXT: [[TMP23:%.*]] = load float, float* [[TMP22]], align 4
; INTERLEAVE-NEXT: [[TMP24:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[I]]
; INTERLEAVE-NEXT: [[TMP25:%.*]] = load float, float* [[TMP24]], align 4
; INTERLEAVE-NEXT: [[TMP26:%.*]] = fadd fast float [[S]], 1.000000e+00
; INTERLEAVE-NEXT: [[TMP27:%.*]] = fadd fast float [[TMP26]], [[TMP23]]
; INTERLEAVE-NEXT: [[TMP28]] = fadd fast float [[TMP27]], [[TMP25]]
; INTERLEAVE-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 8
; INTERLEAVE-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; INTERLEAVE-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END:%.*]], !llvm.loop [[LOOP14:![0-9]+]]
; INTERLEAVE: for.end:
; INTERLEAVE-NEXT: ret float [[TMP28]]
;
entry:
br label %for.body
for.body:
%i = phi i64 [ 0, %entry ], [ %i.next, %for.body ]
%s = phi float [ 0.0, %entry ], [ %6, %for.body ]
%0 = getelementptr inbounds float, float* %a, i64 %i
%1 = load float, float* %0, align 4
%2 = getelementptr inbounds float, float* %b, i64 %i
%3 = load float, float* %2, align 4
%4 = fadd fast float %s, 1.0
%5 = fadd fast float %4, %1
%6 = fadd fast float %5, %3
%i.next = add nuw nsw i64 %i, 8
%cond = icmp slt i64 %i.next, %n
br i1 %cond, label %for.body, label %for.end
for.end:
%s.lcssa = phi float [ %6, %for.body ]
ret float %s.lcssa
}
; Make sure we scalarize the step vectors used for the pointer arithmetic. We
; can't easily simplify vectorized step vectors. (Interleaved accesses.)
;
; for (int i = 0; i < n; ++i)
; a[i].f ^= y;
;
%pair.i32 = type { i32, i32 }
define void @scalarize_induction_variable_03(%pair.i32 *%p, i32 %y, i64 %n) {
; CHECK-LABEL: @scalarize_induction_variable_03(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[Y:%.*]], i32 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 1
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 [[TMP0]], i32 1
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP1]], i32 1
; CHECK-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 8
; CHECK-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP3]], align 8
; CHECK-NEXT: [[TMP6:%.*]] = insertelement <2 x i32> poison, i32 [[TMP4]], i32 0
; CHECK-NEXT: [[TMP7:%.*]] = insertelement <2 x i32> [[TMP6]], i32 [[TMP5]], i32 1
; CHECK-NEXT: [[TMP8:%.*]] = xor <2 x i32> [[TMP7]], [[BROADCAST_SPLAT]]
; CHECK-NEXT: [[TMP9:%.*]] = extractelement <2 x i32> [[TMP8]], i32 0
; CHECK-NEXT: store i32 [[TMP9]], i32* [[TMP2]], align 8
; CHECK-NEXT: [[TMP10:%.*]] = extractelement <2 x i32> [[TMP8]], i32 1
; CHECK-NEXT: store i32 [[TMP10]], i32* [[TMP3]], align 8
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[F:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
; CHECK-NEXT: [[TMP12:%.*]] = load i32, i32* [[F]], align 8
; CHECK-NEXT: [[TMP13:%.*]] = xor i32 [[TMP12]], [[Y]]
; CHECK-NEXT: store i32 [[TMP13]], i32* [[F]], align 8
; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; CHECK-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP16:![0-9]+]]
; CHECK: for.end:
; CHECK-NEXT: ret void
;
; IND-LABEL: @scalarize_induction_variable_03(
; IND-NEXT: entry:
; IND-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2
; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775806
; IND-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[Y:%.*]], i64 0
; IND-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[TMP0:%.*]] = or i64 [[INDEX]], 1
; IND-NEXT: [[TMP1:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 [[INDEX]], i32 1
; IND-NEXT: [[TMP2:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP0]], i32 1
; IND-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 8
; IND-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP2]], align 8
; IND-NEXT: [[TMP5:%.*]] = insertelement <2 x i32> poison, i32 [[TMP3]], i64 0
; IND-NEXT: [[TMP6:%.*]] = insertelement <2 x i32> [[TMP5]], i32 [[TMP4]], i64 1
; IND-NEXT: [[TMP7:%.*]] = xor <2 x i32> [[TMP6]], [[BROADCAST_SPLAT]]
; IND-NEXT: [[TMP8:%.*]] = extractelement <2 x i32> [[TMP7]], i64 0
; IND-NEXT: store i32 [[TMP8]], i32* [[TMP1]], align 8
; IND-NEXT: [[TMP9:%.*]] = extractelement <2 x i32> [[TMP7]], i64 1
; IND-NEXT: store i32 [[TMP9]], i32* [[TMP2]], align 8
; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; IND-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; IND-NEXT: br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
; IND-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; IND-NEXT: br label [[FOR_BODY:%.*]]
; IND: for.body:
; IND-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; IND-NEXT: [[F:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
; IND-NEXT: [[TMP11:%.*]] = load i32, i32* [[F]], align 8
; IND-NEXT: [[TMP12:%.*]] = xor i32 [[TMP11]], [[Y]]
; IND-NEXT: store i32 [[TMP12]], i32* [[F]], align 8
; IND-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; IND-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; IND-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP16:![0-9]+]]
; IND: for.end:
; IND-NEXT: ret void
;
; UNROLL-LABEL: @scalarize_induction_variable_03(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 4
; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775804
; UNROLL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[Y:%.*]], i64 0
; UNROLL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <2 x i32> poison, i32 [[Y]], i64 0
; UNROLL-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT1]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[TMP0:%.*]] = or i64 [[INDEX]], 1
; UNROLL-NEXT: [[TMP1:%.*]] = or i64 [[INDEX]], 2
; UNROLL-NEXT: [[TMP2:%.*]] = or i64 [[INDEX]], 3
; UNROLL-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 [[INDEX]], i32 1
; UNROLL-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP0]], i32 1
; UNROLL-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP1]], i32 1
; UNROLL-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP2]], i32 1
; UNROLL-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP3]], align 8
; UNROLL-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP4]], align 8
; UNROLL-NEXT: [[TMP9:%.*]] = insertelement <2 x i32> poison, i32 [[TMP7]], i64 0
; UNROLL-NEXT: [[TMP10:%.*]] = insertelement <2 x i32> [[TMP9]], i32 [[TMP8]], i64 1
; UNROLL-NEXT: [[TMP11:%.*]] = load i32, i32* [[TMP5]], align 8
; UNROLL-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP6]], align 8
; UNROLL-NEXT: [[TMP13:%.*]] = insertelement <2 x i32> poison, i32 [[TMP11]], i64 0
; UNROLL-NEXT: [[TMP14:%.*]] = insertelement <2 x i32> [[TMP13]], i32 [[TMP12]], i64 1
; UNROLL-NEXT: [[TMP15:%.*]] = xor <2 x i32> [[TMP10]], [[BROADCAST_SPLAT]]
; UNROLL-NEXT: [[TMP16:%.*]] = xor <2 x i32> [[TMP14]], [[BROADCAST_SPLAT2]]
; UNROLL-NEXT: [[TMP17:%.*]] = extractelement <2 x i32> [[TMP15]], i64 0
; UNROLL-NEXT: store i32 [[TMP17]], i32* [[TMP3]], align 8
; UNROLL-NEXT: [[TMP18:%.*]] = extractelement <2 x i32> [[TMP15]], i64 1
; UNROLL-NEXT: store i32 [[TMP18]], i32* [[TMP4]], align 8
; UNROLL-NEXT: [[TMP19:%.*]] = extractelement <2 x i32> [[TMP16]], i64 0
; UNROLL-NEXT: store i32 [[TMP19]], i32* [[TMP5]], align 8
; UNROLL-NEXT: [[TMP20:%.*]] = extractelement <2 x i32> [[TMP16]], i64 1
; UNROLL-NEXT: store i32 [[TMP20]], i32* [[TMP6]], align 8
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NEXT: [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; UNROLL-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL: for.body:
; UNROLL-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[F:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
; UNROLL-NEXT: [[TMP22:%.*]] = load i32, i32* [[F]], align 8
; UNROLL-NEXT: [[TMP23:%.*]] = xor i32 [[TMP22]], [[Y]]
; UNROLL-NEXT: store i32 [[TMP23]], i32* [[F]], align 8
; UNROLL-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; UNROLL-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; UNROLL-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP16:![0-9]+]]
; UNROLL: for.end:
; UNROLL-NEXT: ret void
;
; UNROLL-NO-IC-LABEL: @scalarize_induction_variable_03(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 4
; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 4
; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[Y:%.*]], i32 0
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <2 x i32> poison, i32 [[Y]], i32 0
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT1]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 1
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], 2
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 3
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 [[TMP0]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP1]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP2]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP3]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP4]], align 8
; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP5]], align 8
; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = insertelement <2 x i32> poison, i32 [[TMP8]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = insertelement <2 x i32> [[TMP10]], i32 [[TMP9]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP6]], align 8
; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP7]], align 8
; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = insertelement <2 x i32> poison, i32 [[TMP12]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = insertelement <2 x i32> [[TMP14]], i32 [[TMP13]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = xor <2 x i32> [[TMP11]], [[BROADCAST_SPLAT]]
; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = xor <2 x i32> [[TMP15]], [[BROADCAST_SPLAT2]]
; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = extractelement <2 x i32> [[TMP16]], i32 0
; UNROLL-NO-IC-NEXT: store i32 [[TMP18]], i32* [[TMP4]], align 8
; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = extractelement <2 x i32> [[TMP16]], i32 1
; UNROLL-NO-IC-NEXT: store i32 [[TMP19]], i32* [[TMP5]], align 8
; UNROLL-NO-IC-NEXT: [[TMP20:%.*]] = extractelement <2 x i32> [[TMP17]], i32 0
; UNROLL-NO-IC-NEXT: store i32 [[TMP20]], i32* [[TMP6]], align 8
; UNROLL-NO-IC-NEXT: [[TMP21:%.*]] = extractelement <2 x i32> [[TMP17]], i32 1
; UNROLL-NO-IC-NEXT: store i32 [[TMP21]], i32* [[TMP7]], align 8
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC: for.body:
; UNROLL-NO-IC-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[F:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP23:%.*]] = load i32, i32* [[F]], align 8
; UNROLL-NO-IC-NEXT: [[TMP24:%.*]] = xor i32 [[TMP23]], [[Y]]
; UNROLL-NO-IC-NEXT: store i32 [[TMP24]], i32* [[F]], align 8
; UNROLL-NO-IC-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; UNROLL-NO-IC-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; UNROLL-NO-IC-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP16:![0-9]+]]
; UNROLL-NO-IC: for.end:
; UNROLL-NO-IC-NEXT: ret void
;
; INTERLEAVE-LABEL: @scalarize_induction_variable_03(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 9
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: [[N_MOD_VF:%.*]] = and i64 [[SMAX]], 7
; INTERLEAVE-NEXT: [[TMP0:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
; INTERLEAVE-NEXT: [[TMP1:%.*]] = select i1 [[TMP0]], i64 8, i64 [[N_MOD_VF]]
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = sub nsw i64 [[SMAX]], [[TMP1]]
; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[Y:%.*]], i64 0
; INTERLEAVE-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <4 x i32> poison, i32 [[Y]], i64 0
; INTERLEAVE-NEXT: [[BROADCAST_SPLAT4:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT3]], <4 x i32> poison, <4 x i32> zeroinitializer
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[TMP2:%.*]] = or i64 [[INDEX]], 1
; INTERLEAVE-NEXT: [[TMP3:%.*]] = or i64 [[INDEX]], 2
; INTERLEAVE-NEXT: [[TMP4:%.*]] = or i64 [[INDEX]], 3
; INTERLEAVE-NEXT: [[TMP5:%.*]] = or i64 [[INDEX]], 4
; INTERLEAVE-NEXT: [[TMP6:%.*]] = or i64 [[INDEX]], 5
; INTERLEAVE-NEXT: [[TMP7:%.*]] = or i64 [[INDEX]], 6
; INTERLEAVE-NEXT: [[TMP8:%.*]] = or i64 [[INDEX]], 7
; INTERLEAVE-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 [[INDEX]], i32 1
; INTERLEAVE-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP2]], i32 1
; INTERLEAVE-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP3]], i32 1
; INTERLEAVE-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP4]], i32 1
; INTERLEAVE-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP5]], i32 1
; INTERLEAVE-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP6]], i32 1
; INTERLEAVE-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP7]], i32 1
; INTERLEAVE-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP8]], i32 1
; INTERLEAVE-NEXT: [[TMP17:%.*]] = bitcast i32* [[TMP9]] to <8 x i32>*
; INTERLEAVE-NEXT: [[TMP18:%.*]] = bitcast i32* [[TMP13]] to <8 x i32>*
; INTERLEAVE-NEXT: [[WIDE_VEC:%.*]] = load <8 x i32>, <8 x i32>* [[TMP17]], align 8
; INTERLEAVE-NEXT: [[WIDE_VEC1:%.*]] = load <8 x i32>, <8 x i32>* [[TMP18]], align 8
; 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>
; 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>
; INTERLEAVE-NEXT: [[TMP19:%.*]] = xor <4 x i32> [[STRIDED_VEC]], [[BROADCAST_SPLAT]]
; INTERLEAVE-NEXT: [[TMP20:%.*]] = xor <4 x i32> [[STRIDED_VEC2]], [[BROADCAST_SPLAT4]]
; INTERLEAVE-NEXT: [[TMP21:%.*]] = extractelement <4 x i32> [[TMP19]], i64 0
; INTERLEAVE-NEXT: store i32 [[TMP21]], i32* [[TMP9]], align 8
; INTERLEAVE-NEXT: [[TMP22:%.*]] = extractelement <4 x i32> [[TMP19]], i64 1
; INTERLEAVE-NEXT: store i32 [[TMP22]], i32* [[TMP10]], align 8
; INTERLEAVE-NEXT: [[TMP23:%.*]] = extractelement <4 x i32> [[TMP19]], i64 2
; INTERLEAVE-NEXT: store i32 [[TMP23]], i32* [[TMP11]], align 8
; INTERLEAVE-NEXT: [[TMP24:%.*]] = extractelement <4 x i32> [[TMP19]], i64 3
; INTERLEAVE-NEXT: store i32 [[TMP24]], i32* [[TMP12]], align 8
; INTERLEAVE-NEXT: [[TMP25:%.*]] = extractelement <4 x i32> [[TMP20]], i64 0
; INTERLEAVE-NEXT: store i32 [[TMP25]], i32* [[TMP13]], align 8
; INTERLEAVE-NEXT: [[TMP26:%.*]] = extractelement <4 x i32> [[TMP20]], i64 1
; INTERLEAVE-NEXT: store i32 [[TMP26]], i32* [[TMP14]], align 8
; INTERLEAVE-NEXT: [[TMP27:%.*]] = extractelement <4 x i32> [[TMP20]], i64 2
; INTERLEAVE-NEXT: store i32 [[TMP27]], i32* [[TMP15]], align 8
; INTERLEAVE-NEXT: [[TMP28:%.*]] = extractelement <4 x i32> [[TMP20]], i64 3
; INTERLEAVE-NEXT: store i32 [[TMP28]], i32* [[TMP16]], align 8
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; INTERLEAVE-NEXT: [[TMP29:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP29]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: br label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]
; INTERLEAVE: for.body:
; INTERLEAVE-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[F:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
; INTERLEAVE-NEXT: [[TMP30:%.*]] = load i32, i32* [[F]], align 8
; INTERLEAVE-NEXT: [[TMP31:%.*]] = xor i32 [[TMP30]], [[Y]]
; INTERLEAVE-NEXT: store i32 [[TMP31]], i32* [[F]], align 8
; INTERLEAVE-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; INTERLEAVE-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; INTERLEAVE-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END:%.*]], !llvm.loop [[LOOP16:![0-9]+]]
; INTERLEAVE: for.end:
; INTERLEAVE-NEXT: ret void
;
entry:
br label %for.body
for.body:
%i = phi i64 [ %i.next, %for.body ], [ 0, %entry ]
%f = getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %i, i32 1
%0 = load i32, i32* %f, align 8
%1 = xor i32 %0, %y
store i32 %1, i32* %f, align 8
%i.next = add nuw nsw i64 %i, 1
%cond = icmp slt i64 %i.next, %n
br i1 %cond, label %for.body, label %for.end
for.end:
ret void
}
; Make sure we scalarize the step vectors used for the pointer arithmetic. We
; can't easily simplify vectorized step vectors. (Interleaved accesses.)
;
; for (int i = 0; i < n; ++i)
; p[i].f = a[i * 4]
;
define void @scalarize_induction_variable_04(i32* %a, %pair.i32* %p, i32 %n) {
; CHECK-LABEL: @scalarize_induction_variable_04(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[A4:%.*]] = bitcast i32* [[A:%.*]] to i8*
; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; CHECK-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; CHECK: vector.memcheck:
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 0, i32 1
; CHECK-NEXT: [[SCEVGEP1:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
; CHECK-NEXT: [[TMP3:%.*]] = add i32 [[N]], -1
; CHECK-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
; CHECK-NEXT: [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1
; CHECK-NEXT: [[SCEVGEP2:%.*]] = getelementptr [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP5]]
; CHECK-NEXT: [[SCEVGEP23:%.*]] = bitcast %pair.i32* [[SCEVGEP2]] to i8*
; CHECK-NEXT: [[TMP6:%.*]] = shl nuw nsw i64 [[TMP4]], 2
; CHECK-NEXT: [[TMP7:%.*]] = add nuw nsw i64 [[TMP6]], 1
; CHECK-NEXT: [[SCEVGEP5:%.*]] = getelementptr i32, i32* [[A]], i64 [[TMP7]]
; CHECK-NEXT: [[SCEVGEP56:%.*]] = bitcast i32* [[SCEVGEP5]] to i8*
; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult i8* [[SCEVGEP1]], [[SCEVGEP56]]
; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult i8* [[A4]], [[SCEVGEP23]]
; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; CHECK-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP8:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT: [[TMP9:%.*]] = add i64 [[INDEX]], 1
; CHECK-NEXT: [[TMP10:%.*]] = shl nsw <2 x i64> [[VEC_IND]], <i64 2, i64 2>
; CHECK-NEXT: [[TMP11:%.*]] = extractelement <2 x i64> [[TMP10]], i32 0
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP11]]
; CHECK-NEXT: [[TMP13:%.*]] = extractelement <2 x i64> [[TMP10]], i32 1
; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP13]]
; CHECK-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP12]], align 1, !alias.scope !17
; CHECK-NEXT: [[TMP16:%.*]] = load i32, i32* [[TMP14]], align 1, !alias.scope !17
; CHECK-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP8]], i32 1
; CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP9]], i32 1
; CHECK-NEXT: store i32 [[TMP15]], i32* [[TMP17]], align 1, !alias.scope !20, !noalias !17
; CHECK-NEXT: store i32 [[TMP16]], i32* [[TMP18]], align 1, !alias.scope !20, !noalias !17
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], <i64 2, i64 2>
; CHECK-NEXT: [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[TMP20:%.*]] = shl nsw i64 [[I]], 2
; CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP20]]
; CHECK-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 1
; CHECK-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
; CHECK-NEXT: store i32 [[TMP22]], i32* [[TMP23]], align 1
; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; CHECK-NEXT: [[TMP24:%.*]] = trunc i64 [[I_NEXT]] to i32
; CHECK-NEXT: [[COND:%.*]] = icmp eq i32 [[TMP24]], [[N]]
; CHECK-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
; CHECK: for.end:
; CHECK-NEXT: ret void
;
; IND-LABEL: @scalarize_induction_variable_04(
; IND-NEXT: entry:
; IND-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; IND-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; IND-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP0]], 0
; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; IND: vector.memcheck:
; IND-NEXT: [[SCEVGEP:%.*]] = getelementptr [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 0, i32 1
; IND-NEXT: [[TMP3:%.*]] = add i32 [[N]], -1
; IND-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
; IND-NEXT: [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1
; IND-NEXT: [[TMP6:%.*]] = shl nuw nsw i64 [[TMP4]], 2
; IND-NEXT: [[TMP7:%.*]] = or i64 [[TMP6]], 1
; IND-NEXT: [[SCEVGEP5:%.*]] = getelementptr i32, i32* [[A:%.*]], i64 [[TMP7]]
; IND-NEXT: [[BOUND0:%.*]] = icmp ult i32* [[SCEVGEP]], [[SCEVGEP5]]
; IND-NEXT: [[TMP8:%.*]] = getelementptr [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP5]], i32 0
; IND-NEXT: [[BOUND1:%.*]] = icmp ugt i32* [[TMP8]], [[A]]
; IND-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; IND-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934590
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[TMP9:%.*]] = or i64 [[INDEX]], 1
; IND-NEXT: [[TMP10:%.*]] = shl nsw <2 x i64> [[VEC_IND]], <i64 2, i64 2>
; IND-NEXT: [[TMP11:%.*]] = extractelement <2 x i64> [[TMP10]], i64 0
; IND-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP11]]
; IND-NEXT: [[TMP13:%.*]] = extractelement <2 x i64> [[TMP10]], i64 1
; IND-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP13]]
; IND-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP12]], align 1, !alias.scope !17
; IND-NEXT: [[TMP16:%.*]] = load i32, i32* [[TMP14]], align 1, !alias.scope !17
; IND-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[INDEX]], i32 1
; IND-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP9]], i32 1
; IND-NEXT: store i32 [[TMP15]], i32* [[TMP17]], align 1, !alias.scope !20, !noalias !17
; IND-NEXT: store i32 [[TMP16]], i32* [[TMP18]], align 1, !alias.scope !20, !noalias !17
; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], <i64 2, i64 2>
; IND-NEXT: [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; IND-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; IND-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
; IND-NEXT: br label [[FOR_BODY:%.*]]
; IND: for.body:
; IND-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; IND-NEXT: [[TMP20:%.*]] = shl nsw i64 [[I]], 2
; IND-NEXT: [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP20]]
; IND-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 1
; IND-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
; IND-NEXT: store i32 [[TMP22]], i32* [[TMP23]], align 1
; IND-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; IND-NEXT: [[TMP24:%.*]] = trunc i64 [[I_NEXT]] to i32
; IND-NEXT: [[COND:%.*]] = icmp eq i32 [[TMP24]], [[N]]
; IND-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
; IND: for.end:
; IND-NEXT: ret void
;
; UNROLL-LABEL: @scalarize_induction_variable_04(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; UNROLL-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; UNROLL-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3
; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; UNROLL: vector.memcheck:
; UNROLL-NEXT: [[SCEVGEP:%.*]] = getelementptr [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 0, i32 1
; UNROLL-NEXT: [[TMP3:%.*]] = add i32 [[N]], -1
; UNROLL-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
; UNROLL-NEXT: [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1
; UNROLL-NEXT: [[TMP6:%.*]] = shl nuw nsw i64 [[TMP4]], 2
; UNROLL-NEXT: [[TMP7:%.*]] = or i64 [[TMP6]], 1
; UNROLL-NEXT: [[SCEVGEP5:%.*]] = getelementptr i32, i32* [[A:%.*]], i64 [[TMP7]]
; UNROLL-NEXT: [[BOUND0:%.*]] = icmp ult i32* [[SCEVGEP]], [[SCEVGEP5]]
; UNROLL-NEXT: [[TMP8:%.*]] = getelementptr [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP5]], i32 0
; UNROLL-NEXT: [[BOUND1:%.*]] = icmp ugt i32* [[TMP8]], [[A]]
; UNROLL-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; UNROLL-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934588
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[TMP9:%.*]] = or i64 [[INDEX]], 1
; UNROLL-NEXT: [[TMP10:%.*]] = or i64 [[INDEX]], 2
; UNROLL-NEXT: [[TMP11:%.*]] = or i64 [[INDEX]], 3
; UNROLL-NEXT: [[TMP12:%.*]] = shl nsw <2 x i64> [[VEC_IND]], <i64 2, i64 2>
; UNROLL-NEXT: [[STEP_ADD:%.*]] = shl <2 x i64> [[VEC_IND]], <i64 2, i64 2>
; UNROLL-NEXT: [[TMP13:%.*]] = add <2 x i64> [[STEP_ADD]], <i64 8, i64 8>
; UNROLL-NEXT: [[TMP14:%.*]] = extractelement <2 x i64> [[TMP12]], i64 0
; UNROLL-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP14]]
; UNROLL-NEXT: [[TMP16:%.*]] = extractelement <2 x i64> [[TMP12]], i64 1
; UNROLL-NEXT: [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP16]]
; UNROLL-NEXT: [[TMP18:%.*]] = extractelement <2 x i64> [[TMP13]], i64 0
; UNROLL-NEXT: [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP18]]
; UNROLL-NEXT: [[TMP20:%.*]] = extractelement <2 x i64> [[TMP13]], i64 1
; UNROLL-NEXT: [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP20]]
; UNROLL-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP15]], align 1, !alias.scope !17
; UNROLL-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP17]], align 1, !alias.scope !17
; UNROLL-NEXT: [[TMP24:%.*]] = load i32, i32* [[TMP19]], align 1, !alias.scope !17
; UNROLL-NEXT: [[TMP25:%.*]] = load i32, i32* [[TMP21]], align 1, !alias.scope !17
; UNROLL-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[INDEX]], i32 1
; UNROLL-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP9]], i32 1
; UNROLL-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP10]], i32 1
; UNROLL-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP11]], i32 1
; UNROLL-NEXT: store i32 [[TMP22]], i32* [[TMP26]], align 1, !alias.scope !20, !noalias !17
; UNROLL-NEXT: store i32 [[TMP23]], i32* [[TMP27]], align 1, !alias.scope !20, !noalias !17
; UNROLL-NEXT: store i32 [[TMP24]], i32* [[TMP28]], align 1, !alias.scope !20, !noalias !17
; UNROLL-NEXT: store i32 [[TMP25]], i32* [[TMP29]], align 1, !alias.scope !20, !noalias !17
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], <i64 4, i64 4>
; UNROLL-NEXT: [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
; UNROLL-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL: for.body:
; UNROLL-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[TMP31:%.*]] = shl nsw i64 [[I]], 2
; UNROLL-NEXT: [[TMP32:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP31]]
; UNROLL-NEXT: [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 1
; UNROLL-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
; UNROLL-NEXT: store i32 [[TMP33]], i32* [[TMP34]], align 1
; UNROLL-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; UNROLL-NEXT: [[TMP35:%.*]] = trunc i64 [[I_NEXT]] to i32
; UNROLL-NEXT: [[COND:%.*]] = icmp eq i32 [[TMP35]], [[N]]
; UNROLL-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
; UNROLL: for.end:
; UNROLL-NEXT: ret void
;
; UNROLL-NO-IC-LABEL: @scalarize_induction_variable_04(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: [[A4:%.*]] = bitcast i32* [[A:%.*]] to i8*
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; UNROLL-NO-IC: vector.memcheck:
; UNROLL-NO-IC-NEXT: [[SCEVGEP:%.*]] = getelementptr [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 0, i32 1
; UNROLL-NO-IC-NEXT: [[SCEVGEP1:%.*]] = bitcast i32* [[SCEVGEP]] to i8*
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = add i32 [[N]], -1
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1
; UNROLL-NO-IC-NEXT: [[SCEVGEP2:%.*]] = getelementptr [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP5]]
; UNROLL-NO-IC-NEXT: [[SCEVGEP23:%.*]] = bitcast %pair.i32* [[SCEVGEP2]] to i8*
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = shl nuw nsw i64 [[TMP4]], 2
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = add nuw nsw i64 [[TMP6]], 1
; UNROLL-NO-IC-NEXT: [[SCEVGEP5:%.*]] = getelementptr i32, i32* [[A]], i64 [[TMP7]]
; UNROLL-NO-IC-NEXT: [[SCEVGEP56:%.*]] = bitcast i32* [[SCEVGEP5]] to i8*
; UNROLL-NO-IC-NEXT: [[BOUND0:%.*]] = icmp ult i8* [[SCEVGEP1]], [[SCEVGEP56]]
; UNROLL-NO-IC-NEXT: [[BOUND1:%.*]] = icmp ult i8* [[A4]], [[SCEVGEP23]]
; UNROLL-NO-IC-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; UNROLL-NO-IC-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i64> [[VEC_IND]], <i64 2, i64 2>
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = add i64 [[INDEX]], 1
; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = add i64 [[INDEX]], 2
; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = add i64 [[INDEX]], 3
; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = shl nsw <2 x i64> [[VEC_IND]], <i64 2, i64 2>
; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = shl nsw <2 x i64> [[STEP_ADD]], <i64 2, i64 2>
; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = extractelement <2 x i64> [[TMP12]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP14]]
; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = extractelement <2 x i64> [[TMP12]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP16]]
; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = extractelement <2 x i64> [[TMP13]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP18]]
; UNROLL-NO-IC-NEXT: [[TMP20:%.*]] = extractelement <2 x i64> [[TMP13]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP20]]
; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP15]], align 1, !alias.scope !17
; UNROLL-NO-IC-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP17]], align 1, !alias.scope !17
; UNROLL-NO-IC-NEXT: [[TMP24:%.*]] = load i32, i32* [[TMP19]], align 1, !alias.scope !17
; UNROLL-NO-IC-NEXT: [[TMP25:%.*]] = load i32, i32* [[TMP21]], align 1, !alias.scope !17
; UNROLL-NO-IC-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP8]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP9]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP10]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP11]], i32 1
; UNROLL-NO-IC-NEXT: store i32 [[TMP22]], i32* [[TMP26]], align 1, !alias.scope !20, !noalias !17
; UNROLL-NO-IC-NEXT: store i32 [[TMP23]], i32* [[TMP27]], align 1, !alias.scope !20, !noalias !17
; UNROLL-NO-IC-NEXT: store i32 [[TMP24]], i32* [[TMP28]], align 1, !alias.scope !20, !noalias !17
; UNROLL-NO-IC-NEXT: store i32 [[TMP25]], i32* [[TMP29]], align 1, !alias.scope !20, !noalias !17
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i64> [[STEP_ADD]], <i64 2, i64 2>
; UNROLL-NO-IC-NEXT: [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC: for.body:
; UNROLL-NO-IC-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[TMP31:%.*]] = shl nsw i64 [[I]], 2
; UNROLL-NO-IC-NEXT: [[TMP32:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP31]]
; UNROLL-NO-IC-NEXT: [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 1
; UNROLL-NO-IC-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
; UNROLL-NO-IC-NEXT: store i32 [[TMP33]], i32* [[TMP34]], align 1
; UNROLL-NO-IC-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; UNROLL-NO-IC-NEXT: [[TMP35:%.*]] = trunc i64 [[I_NEXT]] to i32
; UNROLL-NO-IC-NEXT: [[COND:%.*]] = icmp eq i32 [[TMP35]], [[N]]
; UNROLL-NO-IC-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
; UNROLL-NO-IC: for.end:
; UNROLL-NO-IC-NEXT: ret void
;
; INTERLEAVE-LABEL: @scalarize_induction_variable_04(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; INTERLEAVE-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; INTERLEAVE-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 8
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; INTERLEAVE: vector.memcheck:
; INTERLEAVE-NEXT: [[SCEVGEP:%.*]] = getelementptr [[PAIR_I32:%.*]], %pair.i32* [[P:%.*]], i64 0, i32 1
; INTERLEAVE-NEXT: [[TMP3:%.*]] = add i32 [[N]], -1
; INTERLEAVE-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
; INTERLEAVE-NEXT: [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1
; INTERLEAVE-NEXT: [[TMP6:%.*]] = shl nuw nsw i64 [[TMP4]], 2
; INTERLEAVE-NEXT: [[TMP7:%.*]] = or i64 [[TMP6]], 1
; INTERLEAVE-NEXT: [[SCEVGEP5:%.*]] = getelementptr i32, i32* [[A:%.*]], i64 [[TMP7]]
; INTERLEAVE-NEXT: [[BOUND0:%.*]] = icmp ult i32* [[SCEVGEP]], [[SCEVGEP5]]
; INTERLEAVE-NEXT: [[TMP8:%.*]] = getelementptr [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP5]], i32 0
; INTERLEAVE-NEXT: [[BOUND1:%.*]] = icmp ugt i32* [[TMP8]], [[A]]
; INTERLEAVE-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; INTERLEAVE-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: [[N_MOD_VF:%.*]] = and i64 [[TMP2]], 7
; INTERLEAVE-NEXT: [[TMP9:%.*]] = icmp eq i64 [[N_MOD_VF]], 0
; INTERLEAVE-NEXT: [[TMP10:%.*]] = select i1 [[TMP9]], i64 8, i64 [[N_MOD_VF]]
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = sub nsw i64 [[TMP2]], [[TMP10]]
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[TMP11:%.*]] = or i64 [[INDEX]], 1
; INTERLEAVE-NEXT: [[TMP12:%.*]] = or i64 [[INDEX]], 2
; INTERLEAVE-NEXT: [[TMP13:%.*]] = or i64 [[INDEX]], 3
; INTERLEAVE-NEXT: [[TMP14:%.*]] = or i64 [[INDEX]], 4
; INTERLEAVE-NEXT: [[TMP15:%.*]] = or i64 [[INDEX]], 5
; INTERLEAVE-NEXT: [[TMP16:%.*]] = or i64 [[INDEX]], 6
; INTERLEAVE-NEXT: [[TMP17:%.*]] = or i64 [[INDEX]], 7
; INTERLEAVE-NEXT: [[TMP18:%.*]] = shl nsw i64 [[INDEX]], 2
; INTERLEAVE-NEXT: [[TMP19:%.*]] = shl nsw i64 [[TMP14]], 2
; INTERLEAVE-NEXT: [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP18]]
; INTERLEAVE-NEXT: [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP19]]
; INTERLEAVE-NEXT: [[TMP22:%.*]] = bitcast i32* [[TMP20]] to <16 x i32>*
; INTERLEAVE-NEXT: [[TMP23:%.*]] = bitcast i32* [[TMP21]] to <16 x i32>*
; INTERLEAVE-NEXT: [[WIDE_VEC:%.*]] = load <16 x i32>, <16 x i32>* [[TMP22]], align 1
; INTERLEAVE-NEXT: [[WIDE_VEC7:%.*]] = load <16 x i32>, <16 x i32>* [[TMP23]], align 1
; INTERLEAVE-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[INDEX]], i32 1
; INTERLEAVE-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP11]], i32 1
; INTERLEAVE-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP12]], i32 1
; INTERLEAVE-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP13]], i32 1
; INTERLEAVE-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP14]], i32 1
; INTERLEAVE-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP15]], i32 1
; INTERLEAVE-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP16]], i32 1
; INTERLEAVE-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[TMP17]], i32 1
; INTERLEAVE-NEXT: [[TMP32:%.*]] = extractelement <16 x i32> [[WIDE_VEC]], i64 0
; INTERLEAVE-NEXT: store i32 [[TMP32]], i32* [[TMP24]], align 1, !alias.scope !17, !noalias !20
; INTERLEAVE-NEXT: [[TMP33:%.*]] = extractelement <16 x i32> [[WIDE_VEC]], i64 4
; INTERLEAVE-NEXT: store i32 [[TMP33]], i32* [[TMP25]], align 1, !alias.scope !17, !noalias !20
; INTERLEAVE-NEXT: [[TMP34:%.*]] = extractelement <16 x i32> [[WIDE_VEC]], i64 8
; INTERLEAVE-NEXT: store i32 [[TMP34]], i32* [[TMP26]], align 1, !alias.scope !17, !noalias !20
; INTERLEAVE-NEXT: [[TMP35:%.*]] = extractelement <16 x i32> [[WIDE_VEC]], i64 12
; INTERLEAVE-NEXT: store i32 [[TMP35]], i32* [[TMP27]], align 1, !alias.scope !17, !noalias !20
; INTERLEAVE-NEXT: [[TMP36:%.*]] = extractelement <16 x i32> [[WIDE_VEC7]], i64 0
; INTERLEAVE-NEXT: store i32 [[TMP36]], i32* [[TMP28]], align 1, !alias.scope !17, !noalias !20
; INTERLEAVE-NEXT: [[TMP37:%.*]] = extractelement <16 x i32> [[WIDE_VEC7]], i64 4
; INTERLEAVE-NEXT: store i32 [[TMP37]], i32* [[TMP29]], align 1, !alias.scope !17, !noalias !20
; INTERLEAVE-NEXT: [[TMP38:%.*]] = extractelement <16 x i32> [[WIDE_VEC7]], i64 8
; INTERLEAVE-NEXT: store i32 [[TMP38]], i32* [[TMP30]], align 1, !alias.scope !17, !noalias !20
; INTERLEAVE-NEXT: [[TMP39:%.*]] = extractelement <16 x i32> [[WIDE_VEC7]], i64 12
; INTERLEAVE-NEXT: store i32 [[TMP39]], i32* [[TMP31]], align 1, !alias.scope !17, !noalias !20
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; INTERLEAVE-NEXT: [[TMP40:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP40]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: br label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]
; INTERLEAVE: for.body:
; INTERLEAVE-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[TMP41:%.*]] = shl nsw i64 [[I]], 2
; INTERLEAVE-NEXT: [[TMP42:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP41]]
; INTERLEAVE-NEXT: [[TMP43:%.*]] = load i32, i32* [[TMP42]], align 1
; INTERLEAVE-NEXT: [[TMP44:%.*]] = getelementptr inbounds [[PAIR_I32]], %pair.i32* [[P]], i64 [[I]], i32 1
; INTERLEAVE-NEXT: store i32 [[TMP43]], i32* [[TMP44]], align 1
; INTERLEAVE-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; INTERLEAVE-NEXT: [[TMP45:%.*]] = trunc i64 [[I_NEXT]] to i32
; INTERLEAVE-NEXT: [[COND:%.*]] = icmp eq i32 [[TMP45]], [[N]]
; INTERLEAVE-NEXT: br i1 [[COND]], label [[FOR_END:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
; INTERLEAVE: for.end:
; INTERLEAVE-NEXT: ret void
;
entry:
br label %for.body
for.body:
%i = phi i64 [ %i.next, %for.body ], [ 0, %entry]
%0 = shl nsw i64 %i, 2
%1 = getelementptr inbounds i32, i32* %a, i64 %0
%2 = load i32, i32* %1, align 1
%3 = getelementptr inbounds %pair.i32, %pair.i32* %p, i64 %i, i32 1
store i32 %2, i32* %3, align 1
%i.next = add nuw nsw i64 %i, 1
%4 = trunc i64 %i.next to i32
%cond = icmp eq i32 %4, %n
br i1 %cond, label %for.end, label %for.body
for.end:
ret void
}
; PR30542. Ensure we generate all the scalar steps for the induction variable.
; The scalar induction variable is used by a getelementptr instruction
; (uniform), and a udiv (non-uniform).
;
; int sum = 0;
; for (int i = 0; i < n; ++i) {
; int x = a[i];
; if (c)
; x /= i;
; sum += x;
; }
;
;
;
;
define i32 @scalarize_induction_variable_05(i32* %a, i32 %x, i1 %c, i32 %n) {
; CHECK-LABEL: @scalarize_induction_variable_05(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[N:%.*]], i32 1)
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[SMAX]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[SMAX]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[SMAX]], [[N_MOD_VF]]
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i1> poison, i1 [[C:%.*]], i32 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i1> [[BROADCAST_SPLATINSERT]], <2 x i1> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE2:%.*]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[PRED_UDIV_CONTINUE2]] ]
; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 0
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP0]]
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i32 0
; CHECK-NEXT: [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <2 x i32>*
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i32>, <2 x i32>* [[TMP3]], align 4
; CHECK-NEXT: [[TMP4:%.*]] = extractelement <2 x i1> [[BROADCAST_SPLAT]], i32 0
; CHECK-NEXT: br i1 [[TMP4]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
; CHECK: pred.udiv.if:
; CHECK-NEXT: [[TMP5:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 0
; CHECK-NEXT: [[TMP6:%.*]] = udiv i32 [[TMP5]], [[TMP0]]
; CHECK-NEXT: [[TMP7:%.*]] = insertelement <2 x i32> poison, i32 [[TMP6]], i32 0
; CHECK-NEXT: br label [[PRED_UDIV_CONTINUE]]
; CHECK: pred.udiv.continue:
; CHECK-NEXT: [[TMP8:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP7]], [[PRED_UDIV_IF]] ]
; CHECK-NEXT: [[TMP9:%.*]] = extractelement <2 x i1> [[BROADCAST_SPLAT]], i32 1
; CHECK-NEXT: br i1 [[TMP9]], label [[PRED_UDIV_IF1:%.*]], label [[PRED_UDIV_CONTINUE2]]
; CHECK: pred.udiv.if1:
; CHECK-NEXT: [[TMP10:%.*]] = add i32 [[INDEX]], 1
; CHECK-NEXT: [[TMP11:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 1
; CHECK-NEXT: [[TMP12:%.*]] = udiv i32 [[TMP11]], [[TMP10]]
; CHECK-NEXT: [[TMP13:%.*]] = insertelement <2 x i32> [[TMP8]], i32 [[TMP12]], i32 1
; CHECK-NEXT: br label [[PRED_UDIV_CONTINUE2]]
; CHECK: pred.udiv.continue2:
; CHECK-NEXT: [[TMP14:%.*]] = phi <2 x i32> [ [[TMP8]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP13]], [[PRED_UDIV_IF1]] ]
; CHECK-NEXT: [[TMP15:%.*]] = xor <2 x i1> [[BROADCAST_SPLAT]], <i1 true, i1 true>
; CHECK-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[TMP15]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP14]]
; CHECK-NEXT: [[TMP16]] = add <2 x i32> [[PREDPHI]], [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; CHECK-NEXT: [[TMP17:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP18:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[TMP16]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[SMAX]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[TMP18]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[IF_END:%.*]] ]
; CHECK-NEXT: [[SUM:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR4:%.*]], [[IF_END]] ]
; CHECK-NEXT: [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[I]]
; CHECK-NEXT: [[VAR1:%.*]] = load i32, i32* [[VAR0]], align 4
; CHECK-NEXT: br i1 [[C]], label [[IF_THEN:%.*]], label [[IF_END]]
; CHECK: if.then:
; CHECK-NEXT: [[VAR2:%.*]] = udiv i32 [[VAR1]], [[I]]
; CHECK-NEXT: br label [[IF_END]]
; CHECK: if.end:
; CHECK-NEXT: [[VAR3:%.*]] = phi i32 [ [[VAR2]], [[IF_THEN]] ], [ [[VAR1]], [[FOR_BODY]] ]
; CHECK-NEXT: [[VAR4]] = add i32 [[VAR3]], [[SUM]]
; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i32 [[I]], 1
; CHECK-NEXT: [[COND:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP25:![0-9]+]]
; CHECK: for.end:
; CHECK-NEXT: [[VAR5:%.*]] = phi i32 [ [[VAR4]], [[IF_END]] ], [ [[TMP18]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i32 [[VAR5]]
;
; IND-LABEL: @scalarize_induction_variable_05(
; IND-NEXT: entry:
; IND-NEXT: [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[N:%.*]], i32 1)
; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[SMAX]], 2
; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: [[N_VEC:%.*]] = and i32 [[SMAX]], 2147483646
; IND-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i1> poison, i1 [[C:%.*]], i64 0
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE2:%.*]] ]
; IND-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[PRED_UDIV_CONTINUE2]] ]
; IND-NEXT: [[TMP0:%.*]] = sext i32 [[INDEX]] to i64
; IND-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP0]]
; IND-NEXT: [[TMP2:%.*]] = bitcast i32* [[TMP1]] to <2 x i32>*
; IND-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i32>, <2 x i32>* [[TMP2]], align 4
; IND-NEXT: br i1 [[C]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
; IND: pred.udiv.if:
; IND-NEXT: [[TMP3:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i64 0
; IND-NEXT: [[TMP4:%.*]] = udiv i32 [[TMP3]], [[INDEX]]
; IND-NEXT: [[TMP5:%.*]] = insertelement <2 x i32> poison, i32 [[TMP4]], i64 0
; IND-NEXT: br label [[PRED_UDIV_CONTINUE]]
; IND: pred.udiv.continue:
; IND-NEXT: [[TMP6:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP5]], [[PRED_UDIV_IF]] ]
; IND-NEXT: br i1 [[C]], label [[PRED_UDIV_IF1:%.*]], label [[PRED_UDIV_CONTINUE2]]
; IND: pred.udiv.if1:
; IND-NEXT: [[TMP7:%.*]] = or i32 [[INDEX]], 1
; IND-NEXT: [[TMP8:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i64 1
; IND-NEXT: [[TMP9:%.*]] = udiv i32 [[TMP8]], [[TMP7]]
; IND-NEXT: [[TMP10:%.*]] = insertelement <2 x i32> [[TMP6]], i32 [[TMP9]], i64 1
; IND-NEXT: br label [[PRED_UDIV_CONTINUE2]]
; IND: pred.udiv.continue2:
; IND-NEXT: [[TMP11:%.*]] = phi <2 x i32> [ [[TMP6]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP10]], [[PRED_UDIV_IF1]] ]
; IND-NEXT: [[TMP12:%.*]] = xor <2 x i1> [[BROADCAST_SPLATINSERT]], <i1 true, i1 poison>
; IND-NEXT: [[TMP13:%.*]] = shufflevector <2 x i1> [[TMP12]], <2 x i1> poison, <2 x i32> zeroinitializer
; IND-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[TMP13]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP11]]
; IND-NEXT: [[TMP14]] = add <2 x i32> [[PREDPHI]], [[VEC_PHI]]
; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; IND-NEXT: [[TMP15:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; IND-NEXT: br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[TMP16:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[TMP14]])
; IND-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[SMAX]], [[N_VEC]]
; IND-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; IND-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP16]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; IND-NEXT: br label [[FOR_BODY:%.*]]
; IND: for.body:
; IND-NEXT: [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[IF_END:%.*]] ]
; IND-NEXT: [[SUM:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR4:%.*]], [[IF_END]] ]
; IND-NEXT: [[TMP17:%.*]] = zext i32 [[I]] to i64
; IND-NEXT: [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP17]]
; IND-NEXT: [[VAR1:%.*]] = load i32, i32* [[VAR0]], align 4
; IND-NEXT: br i1 [[C]], label [[IF_THEN:%.*]], label [[IF_END]]
; IND: if.then:
; IND-NEXT: [[VAR2:%.*]] = udiv i32 [[VAR1]], [[I]]
; IND-NEXT: br label [[IF_END]]
; IND: if.end:
; IND-NEXT: [[VAR3:%.*]] = phi i32 [ [[VAR2]], [[IF_THEN]] ], [ [[VAR1]], [[FOR_BODY]] ]
; IND-NEXT: [[VAR4]] = add i32 [[VAR3]], [[SUM]]
; IND-NEXT: [[I_NEXT]] = add nuw nsw i32 [[I]], 1
; IND-NEXT: [[COND:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
; IND-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP25:![0-9]+]]
; IND: for.end:
; IND-NEXT: [[VAR5:%.*]] = phi i32 [ [[VAR4]], [[IF_END]] ], [ [[TMP16]], [[MIDDLE_BLOCK]] ]
; IND-NEXT: ret i32 [[VAR5]]
;
; UNROLL-LABEL: @scalarize_induction_variable_05(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[N:%.*]], i32 1)
; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[SMAX]], 4
; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: [[N_VEC:%.*]] = and i32 [[SMAX]], 2147483644
; UNROLL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i1> poison, i1 [[C:%.*]], i64 0
; UNROLL-NEXT: [[BROADCAST_SPLATINSERT5:%.*]] = insertelement <2 x i1> poison, i1 [[C]], i64 0
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE10:%.*]] ]
; UNROLL-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP28:%.*]], [[PRED_UDIV_CONTINUE10]] ]
; UNROLL-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP29:%.*]], [[PRED_UDIV_CONTINUE10]] ]
; UNROLL-NEXT: [[TMP0:%.*]] = or i32 [[INDEX]], 2
; UNROLL-NEXT: [[TMP1:%.*]] = sext i32 [[INDEX]] to i64
; UNROLL-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP1]]
; UNROLL-NEXT: [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <2 x i32>*
; UNROLL-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i32>, <2 x i32>* [[TMP3]], align 4
; UNROLL-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 2
; UNROLL-NEXT: [[TMP5:%.*]] = bitcast i32* [[TMP4]] to <2 x i32>*
; UNROLL-NEXT: [[WIDE_LOAD2:%.*]] = load <2 x i32>, <2 x i32>* [[TMP5]], align 4
; UNROLL-NEXT: br i1 [[C]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
; UNROLL: pred.udiv.if:
; UNROLL-NEXT: [[TMP6:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i64 0
; UNROLL-NEXT: [[TMP7:%.*]] = udiv i32 [[TMP6]], [[INDEX]]
; UNROLL-NEXT: [[TMP8:%.*]] = insertelement <2 x i32> poison, i32 [[TMP7]], i64 0
; UNROLL-NEXT: br label [[PRED_UDIV_CONTINUE]]
; UNROLL: pred.udiv.continue:
; UNROLL-NEXT: [[TMP9:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP8]], [[PRED_UDIV_IF]] ]
; UNROLL-NEXT: br i1 [[C]], label [[PRED_UDIV_IF3:%.*]], label [[PRED_UDIV_CONTINUE4:%.*]]
; UNROLL: pred.udiv.if3:
; UNROLL-NEXT: [[TMP10:%.*]] = or i32 [[INDEX]], 1
; UNROLL-NEXT: [[TMP11:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i64 1
; UNROLL-NEXT: [[TMP12:%.*]] = udiv i32 [[TMP11]], [[TMP10]]
; UNROLL-NEXT: [[TMP13:%.*]] = insertelement <2 x i32> [[TMP9]], i32 [[TMP12]], i64 1
; UNROLL-NEXT: br label [[PRED_UDIV_CONTINUE4]]
; UNROLL: pred.udiv.continue4:
; UNROLL-NEXT: [[TMP14:%.*]] = phi <2 x i32> [ [[TMP9]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP13]], [[PRED_UDIV_IF3]] ]
; UNROLL-NEXT: br i1 [[C]], label [[PRED_UDIV_IF7:%.*]], label [[PRED_UDIV_CONTINUE8:%.*]]
; UNROLL: pred.udiv.if7:
; UNROLL-NEXT: [[TMP15:%.*]] = extractelement <2 x i32> [[WIDE_LOAD2]], i64 0
; UNROLL-NEXT: [[TMP16:%.*]] = udiv i32 [[TMP15]], [[TMP0]]
; UNROLL-NEXT: [[TMP17:%.*]] = insertelement <2 x i32> poison, i32 [[TMP16]], i64 0
; UNROLL-NEXT: br label [[PRED_UDIV_CONTINUE8]]
; UNROLL: pred.udiv.continue8:
; UNROLL-NEXT: [[TMP18:%.*]] = phi <2 x i32> [ poison, [[PRED_UDIV_CONTINUE4]] ], [ [[TMP17]], [[PRED_UDIV_IF7]] ]
; UNROLL-NEXT: br i1 [[C]], label [[PRED_UDIV_IF9:%.*]], label [[PRED_UDIV_CONTINUE10]]
; UNROLL: pred.udiv.if9:
; UNROLL-NEXT: [[TMP19:%.*]] = or i32 [[INDEX]], 3
; UNROLL-NEXT: [[TMP20:%.*]] = extractelement <2 x i32> [[WIDE_LOAD2]], i64 1
; UNROLL-NEXT: [[TMP21:%.*]] = udiv i32 [[TMP20]], [[TMP19]]
; UNROLL-NEXT: [[TMP22:%.*]] = insertelement <2 x i32> [[TMP18]], i32 [[TMP21]], i64 1
; UNROLL-NEXT: br label [[PRED_UDIV_CONTINUE10]]
; UNROLL: pred.udiv.continue10:
; UNROLL-NEXT: [[TMP23:%.*]] = phi <2 x i32> [ [[TMP18]], [[PRED_UDIV_CONTINUE8]] ], [ [[TMP22]], [[PRED_UDIV_IF9]] ]
; UNROLL-NEXT: [[TMP24:%.*]] = xor <2 x i1> [[BROADCAST_SPLATINSERT]], <i1 true, i1 poison>
; UNROLL-NEXT: [[TMP25:%.*]] = shufflevector <2 x i1> [[TMP24]], <2 x i1> poison, <2 x i32> zeroinitializer
; UNROLL-NEXT: [[TMP26:%.*]] = xor <2 x i1> [[BROADCAST_SPLATINSERT5]], <i1 true, i1 poison>
; UNROLL-NEXT: [[TMP27:%.*]] = shufflevector <2 x i1> [[TMP26]], <2 x i1> poison, <2 x i32> zeroinitializer
; UNROLL-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[TMP25]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP14]]
; UNROLL-NEXT: [[PREDPHI11:%.*]] = select <2 x i1> [[TMP27]], <2 x i32> [[WIDE_LOAD2]], <2 x i32> [[TMP23]]
; UNROLL-NEXT: [[TMP28]] = add <2 x i32> [[PREDPHI]], [[VEC_PHI]]
; UNROLL-NEXT: [[TMP29]] = add <2 x i32> [[PREDPHI11]], [[VEC_PHI1]]
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NEXT: [[TMP30:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[BIN_RDX:%.*]] = add <2 x i32> [[TMP29]], [[TMP28]]
; UNROLL-NEXT: [[TMP31:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[BIN_RDX]])
; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[SMAX]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; UNROLL-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP31]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL: for.body:
; UNROLL-NEXT: [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[IF_END:%.*]] ]
; UNROLL-NEXT: [[SUM:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR4:%.*]], [[IF_END]] ]
; UNROLL-NEXT: [[TMP32:%.*]] = zext i32 [[I]] to i64
; UNROLL-NEXT: [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP32]]
; UNROLL-NEXT: [[VAR1:%.*]] = load i32, i32* [[VAR0]], align 4
; UNROLL-NEXT: br i1 [[C]], label [[IF_THEN:%.*]], label [[IF_END]]
; UNROLL: if.then:
; UNROLL-NEXT: [[VAR2:%.*]] = udiv i32 [[VAR1]], [[I]]
; UNROLL-NEXT: br label [[IF_END]]
; UNROLL: if.end:
; UNROLL-NEXT: [[VAR3:%.*]] = phi i32 [ [[VAR2]], [[IF_THEN]] ], [ [[VAR1]], [[FOR_BODY]] ]
; UNROLL-NEXT: [[VAR4]] = add i32 [[VAR3]], [[SUM]]
; UNROLL-NEXT: [[I_NEXT]] = add nuw nsw i32 [[I]], 1
; UNROLL-NEXT: [[COND:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
; UNROLL-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP25:![0-9]+]]
; UNROLL: for.end:
; UNROLL-NEXT: [[VAR5:%.*]] = phi i32 [ [[VAR4]], [[IF_END]] ], [ [[TMP31]], [[MIDDLE_BLOCK]] ]
; UNROLL-NEXT: ret i32 [[VAR5]]
;
; UNROLL-NO-IC-LABEL: @scalarize_induction_variable_05(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[N:%.*]], i32 1)
; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[SMAX]], 4
; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[SMAX]], 4
; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i32 [[SMAX]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i1> poison, i1 [[C:%.*]], i32 0
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i1> [[BROADCAST_SPLATINSERT]], <2 x i1> poison, <2 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT5:%.*]] = insertelement <2 x i1> poison, i1 [[C]], i32 0
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT6:%.*]] = shufflevector <2 x i1> [[BROADCAST_SPLATINSERT5]], <2 x i1> poison, <2 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE10:%.*]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP32:%.*]], [[PRED_UDIV_CONTINUE10]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP33:%.*]], [[PRED_UDIV_CONTINUE10]] ]
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 0
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = add i32 [[INDEX]], 2
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP0]]
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[TMP1]]
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = bitcast i32* [[TMP4]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i32>, <2 x i32>* [[TMP5]], align 4
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 2
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = bitcast i32* [[TMP6]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: [[WIDE_LOAD2:%.*]] = load <2 x i32>, <2 x i32>* [[TMP7]], align 4
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = extractelement <2 x i1> [[BROADCAST_SPLAT]], i32 0
; UNROLL-NO-IC-NEXT: br i1 [[TMP8]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
; UNROLL-NO-IC: pred.udiv.if:
; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = udiv i32 [[TMP9]], [[TMP0]]
; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = insertelement <2 x i32> poison, i32 [[TMP10]], i32 0
; UNROLL-NO-IC-NEXT: br label [[PRED_UDIV_CONTINUE]]
; UNROLL-NO-IC: pred.udiv.continue:
; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = phi <2 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP11]], [[PRED_UDIV_IF]] ]
; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = extractelement <2 x i1> [[BROADCAST_SPLAT]], i32 1
; UNROLL-NO-IC-NEXT: br i1 [[TMP13]], label [[PRED_UDIV_IF3:%.*]], label [[PRED_UDIV_CONTINUE4:%.*]]
; UNROLL-NO-IC: pred.udiv.if3:
; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = add i32 [[INDEX]], 1
; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = udiv i32 [[TMP15]], [[TMP14]]
; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = insertelement <2 x i32> [[TMP12]], i32 [[TMP16]], i32 1
; UNROLL-NO-IC-NEXT: br label [[PRED_UDIV_CONTINUE4]]
; UNROLL-NO-IC: pred.udiv.continue4:
; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = phi <2 x i32> [ [[TMP12]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP17]], [[PRED_UDIV_IF3]] ]
; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = extractelement <2 x i1> [[BROADCAST_SPLAT6]], i32 0
; UNROLL-NO-IC-NEXT: br i1 [[TMP19]], label [[PRED_UDIV_IF7:%.*]], label [[PRED_UDIV_CONTINUE8:%.*]]
; UNROLL-NO-IC: pred.udiv.if7:
; UNROLL-NO-IC-NEXT: [[TMP20:%.*]] = extractelement <2 x i32> [[WIDE_LOAD2]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP21:%.*]] = udiv i32 [[TMP20]], [[TMP1]]
; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = insertelement <2 x i32> poison, i32 [[TMP21]], i32 0
; UNROLL-NO-IC-NEXT: br label [[PRED_UDIV_CONTINUE8]]
; UNROLL-NO-IC: pred.udiv.continue8:
; UNROLL-NO-IC-NEXT: [[TMP23:%.*]] = phi <2 x i32> [ poison, [[PRED_UDIV_CONTINUE4]] ], [ [[TMP22]], [[PRED_UDIV_IF7]] ]
; UNROLL-NO-IC-NEXT: [[TMP24:%.*]] = extractelement <2 x i1> [[BROADCAST_SPLAT6]], i32 1
; UNROLL-NO-IC-NEXT: br i1 [[TMP24]], label [[PRED_UDIV_IF9:%.*]], label [[PRED_UDIV_CONTINUE10]]
; UNROLL-NO-IC: pred.udiv.if9:
; UNROLL-NO-IC-NEXT: [[TMP25:%.*]] = add i32 [[INDEX]], 3
; UNROLL-NO-IC-NEXT: [[TMP26:%.*]] = extractelement <2 x i32> [[WIDE_LOAD2]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP27:%.*]] = udiv i32 [[TMP26]], [[TMP25]]
; UNROLL-NO-IC-NEXT: [[TMP28:%.*]] = insertelement <2 x i32> [[TMP23]], i32 [[TMP27]], i32 1
; UNROLL-NO-IC-NEXT: br label [[PRED_UDIV_CONTINUE10]]
; UNROLL-NO-IC: pred.udiv.continue10:
; UNROLL-NO-IC-NEXT: [[TMP29:%.*]] = phi <2 x i32> [ [[TMP23]], [[PRED_UDIV_CONTINUE8]] ], [ [[TMP28]], [[PRED_UDIV_IF9]] ]
; UNROLL-NO-IC-NEXT: [[TMP30:%.*]] = xor <2 x i1> [[BROADCAST_SPLAT]], <i1 true, i1 true>
; UNROLL-NO-IC-NEXT: [[TMP31:%.*]] = xor <2 x i1> [[BROADCAST_SPLAT6]], <i1 true, i1 true>
; UNROLL-NO-IC-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[TMP30]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP18]]
; UNROLL-NO-IC-NEXT: [[PREDPHI11:%.*]] = select <2 x i1> [[TMP31]], <2 x i32> [[WIDE_LOAD2]], <2 x i32> [[TMP29]]
; UNROLL-NO-IC-NEXT: [[TMP32]] = add <2 x i32> [[PREDPHI]], [[VEC_PHI]]
; UNROLL-NO-IC-NEXT: [[TMP33]] = add <2 x i32> [[PREDPHI11]], [[VEC_PHI1]]
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[TMP34:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP34]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[BIN_RDX:%.*]] = add <2 x i32> [[TMP33]], [[TMP32]]
; UNROLL-NO-IC-NEXT: [[TMP35:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[BIN_RDX]])
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[SMAX]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; UNROLL-NO-IC-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[TMP35]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC: for.body:
; UNROLL-NO-IC-NEXT: [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[IF_END:%.*]] ]
; UNROLL-NO-IC-NEXT: [[SUM:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR4:%.*]], [[IF_END]] ]
; UNROLL-NO-IC-NEXT: [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[I]]
; UNROLL-NO-IC-NEXT: [[VAR1:%.*]] = load i32, i32* [[VAR0]], align 4
; UNROLL-NO-IC-NEXT: br i1 [[C]], label [[IF_THEN:%.*]], label [[IF_END]]
; UNROLL-NO-IC: if.then:
; UNROLL-NO-IC-NEXT: [[VAR2:%.*]] = udiv i32 [[VAR1]], [[I]]
; UNROLL-NO-IC-NEXT: br label [[IF_END]]
; UNROLL-NO-IC: if.end:
; UNROLL-NO-IC-NEXT: [[VAR3:%.*]] = phi i32 [ [[VAR2]], [[IF_THEN]] ], [ [[VAR1]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VAR4]] = add i32 [[VAR3]], [[SUM]]
; UNROLL-NO-IC-NEXT: [[I_NEXT]] = add nuw nsw i32 [[I]], 1
; UNROLL-NO-IC-NEXT: [[COND:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
; UNROLL-NO-IC-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP25:![0-9]+]]
; UNROLL-NO-IC: for.end:
; UNROLL-NO-IC-NEXT: [[VAR5:%.*]] = phi i32 [ [[VAR4]], [[IF_END]] ], [ [[TMP35]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: ret i32 [[VAR5]]
;
; INTERLEAVE-LABEL: @scalarize_induction_variable_05(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[N:%.*]], i32 1)
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[SMAX]], 8
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i32 [[SMAX]], 2147483640
; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[C:%.*]], i64 0
; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT9:%.*]] = insertelement <4 x i1> poison, i1 [[C]], i64 0
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE18:%.*]] ]
; INTERLEAVE-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP48:%.*]], [[PRED_UDIV_CONTINUE18]] ]
; INTERLEAVE-NEXT: [[VEC_PHI1:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP49:%.*]], [[PRED_UDIV_CONTINUE18]] ]
; INTERLEAVE-NEXT: [[TMP0:%.*]] = or i32 [[INDEX]], 4
; INTERLEAVE-NEXT: [[TMP1:%.*]] = sext i32 [[INDEX]] to i64
; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP1]]
; INTERLEAVE-NEXT: [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <4 x i32>*
; INTERLEAVE-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, <4 x i32>* [[TMP3]], align 4
; INTERLEAVE-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 4
; INTERLEAVE-NEXT: [[TMP5:%.*]] = bitcast i32* [[TMP4]] to <4 x i32>*
; INTERLEAVE-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x i32>, <4 x i32>* [[TMP5]], align 4
; INTERLEAVE-NEXT: br i1 [[C]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
; INTERLEAVE: pred.udiv.if:
; INTERLEAVE-NEXT: [[TMP6:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i64 0
; INTERLEAVE-NEXT: [[TMP7:%.*]] = udiv i32 [[TMP6]], [[INDEX]]
; INTERLEAVE-NEXT: [[TMP8:%.*]] = insertelement <4 x i32> poison, i32 [[TMP7]], i64 0
; INTERLEAVE-NEXT: br label [[PRED_UDIV_CONTINUE]]
; INTERLEAVE: pred.udiv.continue:
; INTERLEAVE-NEXT: [[TMP9:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP8]], [[PRED_UDIV_IF]] ]
; INTERLEAVE-NEXT: br i1 [[C]], label [[PRED_UDIV_IF3:%.*]], label [[PRED_UDIV_CONTINUE4:%.*]]
; INTERLEAVE: pred.udiv.if3:
; INTERLEAVE-NEXT: [[TMP10:%.*]] = or i32 [[INDEX]], 1
; INTERLEAVE-NEXT: [[TMP11:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i64 1
; INTERLEAVE-NEXT: [[TMP12:%.*]] = udiv i32 [[TMP11]], [[TMP10]]
; INTERLEAVE-NEXT: [[TMP13:%.*]] = insertelement <4 x i32> [[TMP9]], i32 [[TMP12]], i64 1
; INTERLEAVE-NEXT: br label [[PRED_UDIV_CONTINUE4]]
; INTERLEAVE: pred.udiv.continue4:
; INTERLEAVE-NEXT: [[TMP14:%.*]] = phi <4 x i32> [ [[TMP9]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP13]], [[PRED_UDIV_IF3]] ]
; INTERLEAVE-NEXT: br i1 [[C]], label [[PRED_UDIV_IF5:%.*]], label [[PRED_UDIV_CONTINUE6:%.*]]
; INTERLEAVE: pred.udiv.if5:
; INTERLEAVE-NEXT: [[TMP15:%.*]] = or i32 [[INDEX]], 2
; INTERLEAVE-NEXT: [[TMP16:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i64 2
; INTERLEAVE-NEXT: [[TMP17:%.*]] = udiv i32 [[TMP16]], [[TMP15]]
; INTERLEAVE-NEXT: [[TMP18:%.*]] = insertelement <4 x i32> [[TMP14]], i32 [[TMP17]], i64 2
; INTERLEAVE-NEXT: br label [[PRED_UDIV_CONTINUE6]]
; INTERLEAVE: pred.udiv.continue6:
; INTERLEAVE-NEXT: [[TMP19:%.*]] = phi <4 x i32> [ [[TMP14]], [[PRED_UDIV_CONTINUE4]] ], [ [[TMP18]], [[PRED_UDIV_IF5]] ]
; INTERLEAVE-NEXT: br i1 [[C]], label [[PRED_UDIV_IF7:%.*]], label [[PRED_UDIV_CONTINUE8:%.*]]
; INTERLEAVE: pred.udiv.if7:
; INTERLEAVE-NEXT: [[TMP20:%.*]] = or i32 [[INDEX]], 3
; INTERLEAVE-NEXT: [[TMP21:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i64 3
; INTERLEAVE-NEXT: [[TMP22:%.*]] = udiv i32 [[TMP21]], [[TMP20]]
; INTERLEAVE-NEXT: [[TMP23:%.*]] = insertelement <4 x i32> [[TMP19]], i32 [[TMP22]], i64 3
; INTERLEAVE-NEXT: br label [[PRED_UDIV_CONTINUE8]]
; INTERLEAVE: pred.udiv.continue8:
; INTERLEAVE-NEXT: [[TMP24:%.*]] = phi <4 x i32> [ [[TMP19]], [[PRED_UDIV_CONTINUE6]] ], [ [[TMP23]], [[PRED_UDIV_IF7]] ]
; INTERLEAVE-NEXT: br i1 [[C]], label [[PRED_UDIV_IF11:%.*]], label [[PRED_UDIV_CONTINUE12:%.*]]
; INTERLEAVE: pred.udiv.if11:
; INTERLEAVE-NEXT: [[TMP25:%.*]] = extractelement <4 x i32> [[WIDE_LOAD2]], i64 0
; INTERLEAVE-NEXT: [[TMP26:%.*]] = udiv i32 [[TMP25]], [[TMP0]]
; INTERLEAVE-NEXT: [[TMP27:%.*]] = insertelement <4 x i32> poison, i32 [[TMP26]], i64 0
; INTERLEAVE-NEXT: br label [[PRED_UDIV_CONTINUE12]]
; INTERLEAVE: pred.udiv.continue12:
; INTERLEAVE-NEXT: [[TMP28:%.*]] = phi <4 x i32> [ poison, [[PRED_UDIV_CONTINUE8]] ], [ [[TMP27]], [[PRED_UDIV_IF11]] ]
; INTERLEAVE-NEXT: br i1 [[C]], label [[PRED_UDIV_IF13:%.*]], label [[PRED_UDIV_CONTINUE14:%.*]]
; INTERLEAVE: pred.udiv.if13:
; INTERLEAVE-NEXT: [[TMP29:%.*]] = or i32 [[INDEX]], 5
; INTERLEAVE-NEXT: [[TMP30:%.*]] = extractelement <4 x i32> [[WIDE_LOAD2]], i64 1
; INTERLEAVE-NEXT: [[TMP31:%.*]] = udiv i32 [[TMP30]], [[TMP29]]
; INTERLEAVE-NEXT: [[TMP32:%.*]] = insertelement <4 x i32> [[TMP28]], i32 [[TMP31]], i64 1
; INTERLEAVE-NEXT: br label [[PRED_UDIV_CONTINUE14]]
; INTERLEAVE: pred.udiv.continue14:
; INTERLEAVE-NEXT: [[TMP33:%.*]] = phi <4 x i32> [ [[TMP28]], [[PRED_UDIV_CONTINUE12]] ], [ [[TMP32]], [[PRED_UDIV_IF13]] ]
; INTERLEAVE-NEXT: br i1 [[C]], label [[PRED_UDIV_IF15:%.*]], label [[PRED_UDIV_CONTINUE16:%.*]]
; INTERLEAVE: pred.udiv.if15:
; INTERLEAVE-NEXT: [[TMP34:%.*]] = or i32 [[INDEX]], 6
; INTERLEAVE-NEXT: [[TMP35:%.*]] = extractelement <4 x i32> [[WIDE_LOAD2]], i64 2
; INTERLEAVE-NEXT: [[TMP36:%.*]] = udiv i32 [[TMP35]], [[TMP34]]
; INTERLEAVE-NEXT: [[TMP37:%.*]] = insertelement <4 x i32> [[TMP33]], i32 [[TMP36]], i64 2
; INTERLEAVE-NEXT: br label [[PRED_UDIV_CONTINUE16]]
; INTERLEAVE: pred.udiv.continue16:
; INTERLEAVE-NEXT: [[TMP38:%.*]] = phi <4 x i32> [ [[TMP33]], [[PRED_UDIV_CONTINUE14]] ], [ [[TMP37]], [[PRED_UDIV_IF15]] ]
; INTERLEAVE-NEXT: br i1 [[C]], label [[PRED_UDIV_IF17:%.*]], label [[PRED_UDIV_CONTINUE18]]
; INTERLEAVE: pred.udiv.if17:
; INTERLEAVE-NEXT: [[TMP39:%.*]] = or i32 [[INDEX]], 7
; INTERLEAVE-NEXT: [[TMP40:%.*]] = extractelement <4 x i32> [[WIDE_LOAD2]], i64 3
; INTERLEAVE-NEXT: [[TMP41:%.*]] = udiv i32 [[TMP40]], [[TMP39]]
; INTERLEAVE-NEXT: [[TMP42:%.*]] = insertelement <4 x i32> [[TMP38]], i32 [[TMP41]], i64 3
; INTERLEAVE-NEXT: br label [[PRED_UDIV_CONTINUE18]]
; INTERLEAVE: pred.udiv.continue18:
; INTERLEAVE-NEXT: [[TMP43:%.*]] = phi <4 x i32> [ [[TMP38]], [[PRED_UDIV_CONTINUE16]] ], [ [[TMP42]], [[PRED_UDIV_IF17]] ]
; INTERLEAVE-NEXT: [[TMP44:%.*]] = xor <4 x i1> [[BROADCAST_SPLATINSERT]], <i1 true, i1 poison, i1 poison, i1 poison>
; INTERLEAVE-NEXT: [[TMP45:%.*]] = shufflevector <4 x i1> [[TMP44]], <4 x i1> poison, <4 x i32> zeroinitializer
; INTERLEAVE-NEXT: [[TMP46:%.*]] = xor <4 x i1> [[BROADCAST_SPLATINSERT9]], <i1 true, i1 poison, i1 poison, i1 poison>
; INTERLEAVE-NEXT: [[TMP47:%.*]] = shufflevector <4 x i1> [[TMP46]], <4 x i1> poison, <4 x i32> zeroinitializer
; INTERLEAVE-NEXT: [[PREDPHI:%.*]] = select <4 x i1> [[TMP45]], <4 x i32> [[WIDE_LOAD]], <4 x i32> [[TMP24]]
; INTERLEAVE-NEXT: [[PREDPHI19:%.*]] = select <4 x i1> [[TMP47]], <4 x i32> [[WIDE_LOAD2]], <4 x i32> [[TMP43]]
; INTERLEAVE-NEXT: [[TMP48]] = add <4 x i32> [[PREDPHI]], [[VEC_PHI]]
; INTERLEAVE-NEXT: [[TMP49]] = add <4 x i32> [[PREDPHI19]], [[VEC_PHI1]]
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
; INTERLEAVE-NEXT: [[TMP50:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP50]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: [[BIN_RDX:%.*]] = add <4 x i32> [[TMP49]], [[TMP48]]
; INTERLEAVE-NEXT: [[TMP51:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[SMAX]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; INTERLEAVE-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP51]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]
; INTERLEAVE: for.body:
; INTERLEAVE-NEXT: [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[IF_END:%.*]] ]
; INTERLEAVE-NEXT: [[SUM:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR4:%.*]], [[IF_END]] ]
; INTERLEAVE-NEXT: [[TMP52:%.*]] = zext i32 [[I]] to i64
; INTERLEAVE-NEXT: [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP52]]
; INTERLEAVE-NEXT: [[VAR1:%.*]] = load i32, i32* [[VAR0]], align 4
; INTERLEAVE-NEXT: br i1 [[C]], label [[IF_THEN:%.*]], label [[IF_END]]
; INTERLEAVE: if.then:
; INTERLEAVE-NEXT: [[VAR2:%.*]] = udiv i32 [[VAR1]], [[I]]
; INTERLEAVE-NEXT: br label [[IF_END]]
; INTERLEAVE: if.end:
; INTERLEAVE-NEXT: [[VAR3:%.*]] = phi i32 [ [[VAR2]], [[IF_THEN]] ], [ [[VAR1]], [[FOR_BODY]] ]
; INTERLEAVE-NEXT: [[VAR4]] = add i32 [[VAR3]], [[SUM]]
; INTERLEAVE-NEXT: [[I_NEXT]] = add nuw nsw i32 [[I]], 1
; INTERLEAVE-NEXT: [[COND:%.*]] = icmp slt i32 [[I_NEXT]], [[N]]
; INTERLEAVE-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP25:![0-9]+]]
; INTERLEAVE: for.end:
; INTERLEAVE-NEXT: [[VAR5:%.*]] = phi i32 [ [[VAR4]], [[IF_END]] ], [ [[TMP51]], [[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: ret i32 [[VAR5]]
;
entry:
br label %for.body
for.body:
%i = phi i32 [ 0, %entry ], [ %i.next, %if.end ]
%sum = phi i32 [ 0, %entry ], [ %var4, %if.end ]
%var0 = getelementptr inbounds i32, i32* %a, i32 %i
%var1 = load i32, i32* %var0, align 4
br i1 %c, label %if.then, label %if.end
if.then:
%var2 = udiv i32 %var1, %i
br label %if.end
if.end:
%var3 = phi i32 [ %var2, %if.then ], [ %var1, %for.body ]
%var4 = add i32 %var3, %sum
%i.next = add nuw nsw i32 %i, 1
%cond = icmp slt i32 %i.next, %n
br i1 %cond, label %for.body, label %for.end
for.end:
%var5 = phi i32 [ %var4, %if.end ]
ret i32 %var5
}
; Ensure we generate both a vector and a scalar induction variable. In this
; test, the induction variable is used by an instruction that will be
; vectorized (trunc) as well as an instruction that will remain in scalar form
; (gepelementptr).
;
;
;
%pair.i16 = type { i16, i16 }
define void @iv_vector_and_scalar_users(%pair.i16* %p, i32 %a, i32 %n) {
; CHECK-LABEL: @iv_vector_and_scalar_users(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; CHECK-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[A:%.*]], i32 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 1
; CHECK-NEXT: [[TMP5:%.*]] = add <2 x i32> [[BROADCAST_SPLAT]], [[VEC_IND]]
; CHECK-NEXT: [[TMP6:%.*]] = trunc <2 x i32> [[TMP5]] to <2 x i16>
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[PAIR_I16:%.*]], %pair.i16* [[P:%.*]], i64 [[TMP3]], i32 1
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP4]], i32 1
; CHECK-NEXT: [[TMP9:%.*]] = extractelement <2 x i16> [[TMP6]], i32 0
; CHECK-NEXT: store i16 [[TMP9]], i16* [[TMP7]], align 2
; CHECK-NEXT: [[TMP10:%.*]] = extractelement <2 x i16> [[TMP6]], i32 1
; CHECK-NEXT: store i16 [[TMP10]], i16* [[TMP8]], align 2
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[TMP12:%.*]] = trunc i64 [[I]] to i32
; CHECK-NEXT: [[TMP13:%.*]] = add i32 [[A]], [[TMP12]]
; CHECK-NEXT: [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[I]], i32 1
; CHECK-NEXT: store i16 [[TMP14]], i16* [[TMP15]], align 2
; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; CHECK-NEXT: [[TMP16:%.*]] = trunc i64 [[I_NEXT]] to i32
; CHECK-NEXT: [[COND:%.*]] = icmp eq i32 [[TMP16]], [[N]]
; CHECK-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]
; CHECK: for.end:
; CHECK-NEXT: ret void
;
; IND-LABEL: @iv_vector_and_scalar_users(
; IND-NEXT: entry:
; IND-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; IND-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; IND-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp eq i32 [[TMP0]], 0
; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934590
; IND-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[A:%.*]], i64 0
; IND-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[TMP3:%.*]] = or i64 [[INDEX]], 1
; IND-NEXT: [[TMP4:%.*]] = add <2 x i32> [[BROADCAST_SPLAT]], [[VEC_IND]]
; IND-NEXT: [[TMP5:%.*]] = trunc <2 x i32> [[TMP4]] to <2 x i16>
; IND-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[PAIR_I16:%.*]], %pair.i16* [[P:%.*]], i64 [[INDEX]], i32 1
; IND-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP3]], i32 1
; IND-NEXT: [[TMP8:%.*]] = extractelement <2 x i16> [[TMP5]], i64 0
; IND-NEXT: store i16 [[TMP8]], i16* [[TMP6]], align 2
; IND-NEXT: [[TMP9:%.*]] = extractelement <2 x i16> [[TMP5]], i64 1
; IND-NEXT: store i16 [[TMP9]], i16* [[TMP7]], align 2
; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; IND-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; IND-NEXT: br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; IND-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; IND-NEXT: br label [[FOR_BODY:%.*]]
; IND: for.body:
; IND-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; IND-NEXT: [[TMP11:%.*]] = trunc i64 [[I]] to i32
; IND-NEXT: [[TMP12:%.*]] = add i32 [[TMP11]], [[A]]
; IND-NEXT: [[TMP13:%.*]] = trunc i32 [[TMP12]] to i16
; IND-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[I]], i32 1
; IND-NEXT: store i16 [[TMP13]], i16* [[TMP14]], align 2
; IND-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; IND-NEXT: [[TMP15:%.*]] = trunc i64 [[I_NEXT]] to i32
; IND-NEXT: [[COND:%.*]] = icmp eq i32 [[TMP15]], [[N]]
; IND-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]
; IND: for.end:
; IND-NEXT: ret void
;
; UNROLL-LABEL: @iv_vector_and_scalar_users(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; UNROLL-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; UNROLL-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 3
; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934588
; UNROLL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[A:%.*]], i64 0
; UNROLL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NEXT: [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <2 x i32> poison, i32 [[A]], i64 0
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[TMP3:%.*]] = or i64 [[INDEX]], 1
; UNROLL-NEXT: [[TMP4:%.*]] = or i64 [[INDEX]], 2
; UNROLL-NEXT: [[TMP5:%.*]] = or i64 [[INDEX]], 3
; UNROLL-NEXT: [[TMP6:%.*]] = add <2 x i32> [[BROADCAST_SPLAT]], [[VEC_IND]]
; UNROLL-NEXT: [[TMP7:%.*]] = add <2 x i32> [[BROADCAST_SPLATINSERT2]], <i32 2, i32 poison>
; UNROLL-NEXT: [[TMP8:%.*]] = shufflevector <2 x i32> [[TMP7]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NEXT: [[TMP9:%.*]] = add <2 x i32> [[TMP8]], [[VEC_IND]]
; UNROLL-NEXT: [[TMP10:%.*]] = trunc <2 x i32> [[TMP6]] to <2 x i16>
; UNROLL-NEXT: [[TMP11:%.*]] = trunc <2 x i32> [[TMP9]] to <2 x i16>
; UNROLL-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[PAIR_I16:%.*]], %pair.i16* [[P:%.*]], i64 [[INDEX]], i32 1
; UNROLL-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP3]], i32 1
; UNROLL-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP4]], i32 1
; UNROLL-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP5]], i32 1
; UNROLL-NEXT: [[TMP16:%.*]] = extractelement <2 x i16> [[TMP10]], i64 0
; UNROLL-NEXT: store i16 [[TMP16]], i16* [[TMP12]], align 2
; UNROLL-NEXT: [[TMP17:%.*]] = extractelement <2 x i16> [[TMP10]], i64 1
; UNROLL-NEXT: store i16 [[TMP17]], i16* [[TMP13]], align 2
; UNROLL-NEXT: [[TMP18:%.*]] = extractelement <2 x i16> [[TMP11]], i64 0
; UNROLL-NEXT: store i16 [[TMP18]], i16* [[TMP14]], align 2
; UNROLL-NEXT: [[TMP19:%.*]] = extractelement <2 x i16> [[TMP11]], i64 1
; UNROLL-NEXT: store i16 [[TMP19]], i16* [[TMP15]], align 2
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 4, i32 4>
; UNROLL-NEXT: [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; UNROLL-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL: for.body:
; UNROLL-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[TMP21:%.*]] = trunc i64 [[I]] to i32
; UNROLL-NEXT: [[TMP22:%.*]] = add i32 [[TMP21]], [[A]]
; UNROLL-NEXT: [[TMP23:%.*]] = trunc i32 [[TMP22]] to i16
; UNROLL-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[I]], i32 1
; UNROLL-NEXT: store i16 [[TMP23]], i16* [[TMP24]], align 2
; UNROLL-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; UNROLL-NEXT: [[TMP25:%.*]] = trunc i64 [[I_NEXT]] to i32
; UNROLL-NEXT: [[COND:%.*]] = icmp eq i32 [[TMP25]], [[N]]
; UNROLL-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]
; UNROLL: for.end:
; UNROLL-NEXT: ret void
;
; UNROLL-NO-IC-LABEL: @iv_vector_and_scalar_users(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[A:%.*]], i32 0
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <2 x i32> poison, i32 [[A]], i32 0
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT3:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT2]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 1
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = add i64 [[INDEX]], 2
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], 3
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = add <2 x i32> [[BROADCAST_SPLAT]], [[VEC_IND]]
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = add <2 x i32> [[BROADCAST_SPLAT3]], [[STEP_ADD]]
; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = trunc <2 x i32> [[TMP7]] to <2 x i16>
; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = trunc <2 x i32> [[TMP8]] to <2 x i16>
; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[PAIR_I16:%.*]], %pair.i16* [[P:%.*]], i64 [[TMP3]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP4]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP5]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP6]], i32 1
; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = extractelement <2 x i16> [[TMP9]], i32 0
; UNROLL-NO-IC-NEXT: store i16 [[TMP15]], i16* [[TMP11]], align 2
; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = extractelement <2 x i16> [[TMP9]], i32 1
; UNROLL-NO-IC-NEXT: store i16 [[TMP16]], i16* [[TMP12]], align 2
; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = extractelement <2 x i16> [[TMP10]], i32 0
; UNROLL-NO-IC-NEXT: store i16 [[TMP17]], i16* [[TMP13]], align 2
; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = extractelement <2 x i16> [[TMP10]], i32 1
; UNROLL-NO-IC-NEXT: store i16 [[TMP18]], i16* [[TMP14]], align 2
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC: for.body:
; UNROLL-NO-IC-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[TMP20:%.*]] = trunc i64 [[I]] to i32
; UNROLL-NO-IC-NEXT: [[TMP21:%.*]] = add i32 [[A]], [[TMP20]]
; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = trunc i32 [[TMP21]] to i16
; UNROLL-NO-IC-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[I]], i32 1
; UNROLL-NO-IC-NEXT: store i16 [[TMP22]], i16* [[TMP23]], align 2
; UNROLL-NO-IC-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; UNROLL-NO-IC-NEXT: [[TMP24:%.*]] = trunc i64 [[I_NEXT]] to i32
; UNROLL-NO-IC-NEXT: [[COND:%.*]] = icmp eq i32 [[TMP24]], [[N]]
; UNROLL-NO-IC-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]
; UNROLL-NO-IC: for.end:
; UNROLL-NO-IC-NEXT: ret void
;
; INTERLEAVE-LABEL: @iv_vector_and_scalar_users(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; INTERLEAVE-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; INTERLEAVE-NEXT: [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 7
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i64 [[TMP2]], 8589934584
; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[A:%.*]], i64 0
; INTERLEAVE-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <4 x i32> poison, i32 [[A]], i64 0
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[TMP3:%.*]] = or i64 [[INDEX]], 1
; INTERLEAVE-NEXT: [[TMP4:%.*]] = or i64 [[INDEX]], 2
; INTERLEAVE-NEXT: [[TMP5:%.*]] = or i64 [[INDEX]], 3
; INTERLEAVE-NEXT: [[TMP6:%.*]] = or i64 [[INDEX]], 4
; INTERLEAVE-NEXT: [[TMP7:%.*]] = or i64 [[INDEX]], 5
; INTERLEAVE-NEXT: [[TMP8:%.*]] = or i64 [[INDEX]], 6
; INTERLEAVE-NEXT: [[TMP9:%.*]] = or i64 [[INDEX]], 7
; INTERLEAVE-NEXT: [[TMP10:%.*]] = add <4 x i32> [[BROADCAST_SPLAT]], [[VEC_IND]]
; INTERLEAVE-NEXT: [[TMP11:%.*]] = add <4 x i32> [[BROADCAST_SPLATINSERT2]], <i32 4, i32 poison, i32 poison, i32 poison>
; INTERLEAVE-NEXT: [[TMP12:%.*]] = shufflevector <4 x i32> [[TMP11]], <4 x i32> poison, <4 x i32> zeroinitializer
; INTERLEAVE-NEXT: [[TMP13:%.*]] = add <4 x i32> [[TMP12]], [[VEC_IND]]
; INTERLEAVE-NEXT: [[TMP14:%.*]] = trunc <4 x i32> [[TMP10]] to <4 x i16>
; INTERLEAVE-NEXT: [[TMP15:%.*]] = trunc <4 x i32> [[TMP13]] to <4 x i16>
; INTERLEAVE-NEXT: [[TMP16:%.*]] = getelementptr inbounds [[PAIR_I16:%.*]], %pair.i16* [[P:%.*]], i64 [[INDEX]], i32 1
; INTERLEAVE-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP3]], i32 1
; INTERLEAVE-NEXT: [[TMP18:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP4]], i32 1
; INTERLEAVE-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP5]], i32 1
; INTERLEAVE-NEXT: [[TMP20:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP6]], i32 1
; INTERLEAVE-NEXT: [[TMP21:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP7]], i32 1
; INTERLEAVE-NEXT: [[TMP22:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP8]], i32 1
; INTERLEAVE-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[TMP9]], i32 1
; INTERLEAVE-NEXT: [[TMP24:%.*]] = extractelement <4 x i16> [[TMP14]], i64 0
; INTERLEAVE-NEXT: store i16 [[TMP24]], i16* [[TMP16]], align 2
; INTERLEAVE-NEXT: [[TMP25:%.*]] = extractelement <4 x i16> [[TMP14]], i64 1
; INTERLEAVE-NEXT: store i16 [[TMP25]], i16* [[TMP17]], align 2
; INTERLEAVE-NEXT: [[TMP26:%.*]] = extractelement <4 x i16> [[TMP14]], i64 2
; INTERLEAVE-NEXT: store i16 [[TMP26]], i16* [[TMP18]], align 2
; INTERLEAVE-NEXT: [[TMP27:%.*]] = extractelement <4 x i16> [[TMP14]], i64 3
; INTERLEAVE-NEXT: store i16 [[TMP27]], i16* [[TMP19]], align 2
; INTERLEAVE-NEXT: [[TMP28:%.*]] = extractelement <4 x i16> [[TMP15]], i64 0
; INTERLEAVE-NEXT: store i16 [[TMP28]], i16* [[TMP20]], align 2
; INTERLEAVE-NEXT: [[TMP29:%.*]] = extractelement <4 x i16> [[TMP15]], i64 1
; INTERLEAVE-NEXT: store i16 [[TMP29]], i16* [[TMP21]], align 2
; INTERLEAVE-NEXT: [[TMP30:%.*]] = extractelement <4 x i16> [[TMP15]], i64 2
; INTERLEAVE-NEXT: store i16 [[TMP30]], i16* [[TMP22]], align 2
; INTERLEAVE-NEXT: [[TMP31:%.*]] = extractelement <4 x i16> [[TMP15]], i64 3
; INTERLEAVE-NEXT: store i16 [[TMP31]], i16* [[TMP23]], align 2
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 8, i32 8, i32 8, i32 8>
; INTERLEAVE-NEXT: [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP32]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]
; INTERLEAVE: for.body:
; INTERLEAVE-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[TMP33:%.*]] = trunc i64 [[I]] to i32
; INTERLEAVE-NEXT: [[TMP34:%.*]] = add i32 [[TMP33]], [[A]]
; INTERLEAVE-NEXT: [[TMP35:%.*]] = trunc i32 [[TMP34]] to i16
; INTERLEAVE-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[PAIR_I16]], %pair.i16* [[P]], i64 [[I]], i32 1
; INTERLEAVE-NEXT: store i16 [[TMP35]], i16* [[TMP36]], align 2
; INTERLEAVE-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; INTERLEAVE-NEXT: [[TMP37:%.*]] = trunc i64 [[I_NEXT]] to i32
; INTERLEAVE-NEXT: [[COND:%.*]] = icmp eq i32 [[TMP37]], [[N]]
; INTERLEAVE-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]
; INTERLEAVE: for.end:
; INTERLEAVE-NEXT: ret void
;
entry:
br label %for.body
for.body:
%i = phi i64 [ %i.next, %for.body ], [ 0, %entry ]
%0 = trunc i64 %i to i32
%1 = add i32 %a, %0
%2 = trunc i32 %1 to i16
%3 = getelementptr inbounds %pair.i16, %pair.i16* %p, i64 %i, i32 1
store i16 %2, i16* %3, align 2
%i.next = add nuw nsw i64 %i, 1
%4 = trunc i64 %i.next to i32
%cond = icmp eq i32 %4, %n
br i1 %cond, label %for.end, label %for.body
for.end:
ret void
}
; Make sure that the loop exit count computation does not overflow for i8 and
; i16. The exit count of these loops is i8/i16 max + 1. If we don't cast the
; induction variable to a bigger type the exit count computation will overflow
; to 0.
; PR17532
define i32 @i8_loop() nounwind readnone ssp uwtable {
; CHECK-LABEL: @i8_loop(
; CHECK-NEXT: entry:
; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ <i32 1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP0:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP0]] = and <2 x i32> [[VEC_PHI]], <i32 4, i32 4>
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
; CHECK-NEXT: br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[TMP0]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 256, 256
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ 1, [[ENTRY]] ], [ [[TMP2]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[A_0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[A_0_AND:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[B_0:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[B_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[A_0_AND]] = and i32 [[A_0]], 4
; CHECK-NEXT: [[B_NEXT]] = add i8 [[B_0]], -1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i8 [[B_NEXT]], 0
; CHECK-NEXT: br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP29:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: [[A_0_AND_LCSSA:%.*]] = phi i32 [ [[A_0_AND]], [[LOOP]] ], [ [[TMP2]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i32 [[A_0_AND_LCSSA]]
;
; IND-LABEL: @i8_loop(
; IND-NEXT: entry:
; IND-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; IND-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
; IND-NEXT: br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
; IND: middle.block:
; IND-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: br label [[LOOP:%.*]]
; IND: loop:
; IND-NEXT: br i1 poison, label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP29:![0-9]+]]
; IND: exit:
; IND-NEXT: ret i32 0
;
; UNROLL-LABEL: @i8_loop(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
; UNROLL-NEXT: br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: br label [[LOOP:%.*]]
; UNROLL: loop:
; UNROLL-NEXT: br i1 poison, label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP29:![0-9]+]]
; UNROLL: exit:
; UNROLL-NEXT: ret i32 0
;
; UNROLL-NO-IC-LABEL: @i8_loop(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ <i32 1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP0:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i32> [ <i32 -1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[TMP0]] = and <2 x i32> [[VEC_PHI]], <i32 4, i32 4>
; UNROLL-NO-IC-NEXT: [[TMP1]] = and <2 x i32> [[VEC_PHI1]], <i32 4, i32 4>
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
; UNROLL-NO-IC-NEXT: br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[BIN_RDX:%.*]] = and <2 x i32> [[TMP1]], [[TMP0]]
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[BIN_RDX]])
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i32 256, 256
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; UNROLL-NO-IC-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ 1, [[ENTRY]] ], [ [[TMP3]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: br label [[LOOP:%.*]]
; UNROLL-NO-IC: loop:
; UNROLL-NO-IC-NEXT: [[A_0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[A_0_AND:%.*]], [[LOOP]] ]
; UNROLL-NO-IC-NEXT: [[B_0:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[B_NEXT:%.*]], [[LOOP]] ]
; UNROLL-NO-IC-NEXT: [[A_0_AND]] = and i32 [[A_0]], 4
; UNROLL-NO-IC-NEXT: [[B_NEXT]] = add i8 [[B_0]], -1
; UNROLL-NO-IC-NEXT: [[EC:%.*]] = icmp eq i8 [[B_NEXT]], 0
; UNROLL-NO-IC-NEXT: br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP29:![0-9]+]]
; UNROLL-NO-IC: exit:
; UNROLL-NO-IC-NEXT: [[A_0_AND_LCSSA:%.*]] = phi i32 [ [[A_0_AND]], [[LOOP]] ], [ [[TMP3]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: ret i32 [[A_0_AND_LCSSA]]
;
; INTERLEAVE-LABEL: @i8_loop(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
; INTERLEAVE-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
; INTERLEAVE-NEXT: br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: br label [[LOOP:%.*]]
; INTERLEAVE: loop:
; INTERLEAVE-NEXT: br i1 poison, label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP29:![0-9]+]]
; INTERLEAVE: exit:
; INTERLEAVE-NEXT: ret i32 0
;
entry:
br label %loop
loop:
%a.0 = phi i32 [ 1, %entry ], [ %a.0.and, %loop ]
%b.0 = phi i8 [ 0, %entry ], [ %b.next, %loop ]
%a.0.and = and i32 %a.0, 4
%b.next = add i8 %b.0, -1
%ec = icmp eq i8 %b.next, 0
br i1 %ec, label %exit, label %loop
exit:
ret i32 %a.0.and
}
define i32 @i16_loop() nounwind readnone ssp uwtable {
; CHECK-LABEL: @i16_loop(
; CHECK-NEXT: entry:
; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ <i32 1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP0:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP0]] = and <2 x i32> [[VEC_PHI]], <i32 4, i32 4>
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 65536
; CHECK-NEXT: br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[TMP0]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 65536, 65536
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ 1, [[ENTRY]] ], [ [[TMP2]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[A_0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[A_0_AND:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[B_0:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[B_0_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[A_0_AND]] = and i32 [[A_0]], 4
; CHECK-NEXT: [[B_0_NEXT]] = add i16 [[B_0]], -1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i16 [[B_0_NEXT]], 0
; CHECK-NEXT: br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP31:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: [[A_0_AND_LCSSA:%.*]] = phi i32 [ [[A_0_AND]], [[LOOP]] ], [ [[TMP2]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i32 [[A_0_AND_LCSSA]]
;
; IND-LABEL: @i16_loop(
; IND-NEXT: entry:
; IND-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; IND-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 65536
; IND-NEXT: br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
; IND: middle.block:
; IND-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: br label [[LOOP:%.*]]
; IND: loop:
; IND-NEXT: br i1 poison, label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP31:![0-9]+]]
; IND: exit:
; IND-NEXT: ret i32 0
;
; UNROLL-LABEL: @i16_loop(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 65536
; UNROLL-NEXT: br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: br label [[LOOP:%.*]]
; UNROLL: loop:
; UNROLL-NEXT: br i1 poison, label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP31:![0-9]+]]
; UNROLL: exit:
; UNROLL-NEXT: ret i32 0
;
; UNROLL-NO-IC-LABEL: @i16_loop(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ <i32 1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP0:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i32> [ <i32 -1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[TMP0]] = and <2 x i32> [[VEC_PHI]], <i32 4, i32 4>
; UNROLL-NO-IC-NEXT: [[TMP1]] = and <2 x i32> [[VEC_PHI1]], <i32 4, i32 4>
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 65536
; UNROLL-NO-IC-NEXT: br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[BIN_RDX:%.*]] = and <2 x i32> [[TMP1]], [[TMP0]]
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[BIN_RDX]])
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i32 65536, 65536
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; UNROLL-NO-IC-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ 1, [[ENTRY]] ], [ [[TMP3]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: br label [[LOOP:%.*]]
; UNROLL-NO-IC: loop:
; UNROLL-NO-IC-NEXT: [[A_0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[A_0_AND:%.*]], [[LOOP]] ]
; UNROLL-NO-IC-NEXT: [[B_0:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[B_0_NEXT:%.*]], [[LOOP]] ]
; UNROLL-NO-IC-NEXT: [[A_0_AND]] = and i32 [[A_0]], 4
; UNROLL-NO-IC-NEXT: [[B_0_NEXT]] = add i16 [[B_0]], -1
; UNROLL-NO-IC-NEXT: [[EC:%.*]] = icmp eq i16 [[B_0_NEXT]], 0
; UNROLL-NO-IC-NEXT: br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP31:![0-9]+]]
; UNROLL-NO-IC: exit:
; UNROLL-NO-IC-NEXT: [[A_0_AND_LCSSA:%.*]] = phi i32 [ [[A_0_AND]], [[LOOP]] ], [ [[TMP3]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: ret i32 [[A_0_AND_LCSSA]]
;
; INTERLEAVE-LABEL: @i16_loop(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
; INTERLEAVE-NEXT: [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 65536
; INTERLEAVE-NEXT: br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: br label [[LOOP:%.*]]
; INTERLEAVE: loop:
; INTERLEAVE-NEXT: br i1 poison, label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP31:![0-9]+]]
; INTERLEAVE: exit:
; INTERLEAVE-NEXT: ret i32 0
;
entry:
br label %loop
loop:
%a.0 = phi i32 [ 1, %entry ], [ %a.0.and, %loop ]
%b.0 = phi i16 [ 0, %entry ], [ %b.0.next, %loop ]
%a.0.and = and i32 %a.0, 4
%b.0.next = add i16 %b.0, -1
%ec = icmp eq i16 %b.0.next, 0
br i1 %ec, label %exit, label %loop
exit:
ret i32 %a.0.and
}
; This loop has a backedge taken count of i32_max. We need to check for this
; condition and branch directly to the scalar loop.
define i32 @max_i32_backedgetaken() nounwind readnone ssp uwtable {
; CHECK-LABEL: @max_i32_backedgetaken(
; CHECK-NEXT: entry:
; CHECK-NEXT: br i1 true, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ <i32 1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP0:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP0]] = and <2 x i32> [[VEC_PHI]], <i32 4, i32 4>
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 0
; CHECK-NEXT: br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[TMP0]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 0, 0
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ 1, [[ENTRY]] ], [ [[TMP2]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[A_0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[A_0_AND:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[B_0:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[B_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[A_0_AND]] = and i32 [[A_0]], 4
; CHECK-NEXT: [[B_NEXT]] = add i32 [[B_0]], -1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i32 [[B_NEXT]], 0
; CHECK-NEXT: br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP33:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: [[A_0_AND_LCSSA:%.*]] = phi i32 [ [[A_0_AND]], [[LOOP]] ], [ [[TMP2]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i32 [[A_0_AND_LCSSA]]
;
; IND-LABEL: @max_i32_backedgetaken(
; IND-NEXT: entry:
; IND-NEXT: br i1 true, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: br i1 poison, label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
; IND: middle.block:
; IND-NEXT: br i1 poison, label [[EXIT:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: br label [[LOOP:%.*]]
; IND: loop:
; IND-NEXT: [[B_0:%.*]] = phi i32 [ 0, [[SCALAR_PH]] ], [ [[B_NEXT:%.*]], [[LOOP]] ]
; IND-NEXT: [[B_NEXT]] = add i32 [[B_0]], -1
; IND-NEXT: [[EC:%.*]] = icmp eq i32 [[B_NEXT]], 0
; IND-NEXT: br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP33:![0-9]+]]
; IND: exit:
; IND-NEXT: ret i32 0
;
; UNROLL-LABEL: @max_i32_backedgetaken(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: br i1 true, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: br i1 poison, label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: br i1 poison, label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: br label [[LOOP:%.*]]
; UNROLL: loop:
; UNROLL-NEXT: [[B_0:%.*]] = phi i32 [ 0, [[SCALAR_PH]] ], [ [[B_NEXT:%.*]], [[LOOP]] ]
; UNROLL-NEXT: [[B_NEXT]] = add i32 [[B_0]], -1
; UNROLL-NEXT: [[EC:%.*]] = icmp eq i32 [[B_NEXT]], 0
; UNROLL-NEXT: br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP33:![0-9]+]]
; UNROLL: exit:
; UNROLL-NEXT: ret i32 0
;
; UNROLL-NO-IC-LABEL: @max_i32_backedgetaken(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: br i1 true, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ <i32 1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP0:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i32> [ <i32 -1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[TMP0]] = and <2 x i32> [[VEC_PHI]], <i32 4, i32 4>
; UNROLL-NO-IC-NEXT: [[TMP1]] = and <2 x i32> [[VEC_PHI1]], <i32 4, i32 4>
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 0
; UNROLL-NO-IC-NEXT: br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[BIN_RDX:%.*]] = and <2 x i32> [[TMP1]], [[TMP0]]
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[BIN_RDX]])
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i32 0, 0
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; UNROLL-NO-IC-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ 1, [[ENTRY]] ], [ [[TMP3]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: br label [[LOOP:%.*]]
; UNROLL-NO-IC: loop:
; UNROLL-NO-IC-NEXT: [[A_0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[A_0_AND:%.*]], [[LOOP]] ]
; UNROLL-NO-IC-NEXT: [[B_0:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[B_NEXT:%.*]], [[LOOP]] ]
; UNROLL-NO-IC-NEXT: [[A_0_AND]] = and i32 [[A_0]], 4
; UNROLL-NO-IC-NEXT: [[B_NEXT]] = add i32 [[B_0]], -1
; UNROLL-NO-IC-NEXT: [[EC:%.*]] = icmp eq i32 [[B_NEXT]], 0
; UNROLL-NO-IC-NEXT: br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP33:![0-9]+]]
; UNROLL-NO-IC: exit:
; UNROLL-NO-IC-NEXT: [[A_0_AND_LCSSA:%.*]] = phi i32 [ [[A_0_AND]], [[LOOP]] ], [ [[TMP3]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: ret i32 [[A_0_AND_LCSSA]]
;
; INTERLEAVE-LABEL: @max_i32_backedgetaken(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: br i1 true, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: br i1 poison, label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: br i1 poison, label [[EXIT:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: br label [[LOOP:%.*]]
; INTERLEAVE: loop:
; INTERLEAVE-NEXT: [[B_0:%.*]] = phi i32 [ 0, [[SCALAR_PH]] ], [ [[B_NEXT:%.*]], [[LOOP]] ]
; INTERLEAVE-NEXT: [[B_NEXT]] = add i32 [[B_0]], -1
; INTERLEAVE-NEXT: [[EC:%.*]] = icmp eq i32 [[B_NEXT]], 0
; INTERLEAVE-NEXT: br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP33:![0-9]+]]
; INTERLEAVE: exit:
; INTERLEAVE-NEXT: ret i32 0
;
entry:
br label %loop
loop:
%a.0 = phi i32 [ 1, %entry ], [ %a.0.and, %loop ]
%b.0 = phi i32 [ 0, %entry ], [ %b.next, %loop ]
%a.0.and = and i32 %a.0, 4
%b.next = add i32 %b.0, -1
%ec = icmp eq i32 %b.next, 0
br i1 %ec, label %exit, label %loop
exit:
ret i32 %a.0.and
}
; When generating the overflow check we must sure that the induction start value
; is defined before the branch to the scalar preheader.
@e = global i8 1, align 1
@d = common global i32 0, align 4
@c = common global i32 0, align 4
define i32 @testoverflowcheck() {
; CHECK-LABEL: @testoverflowcheck(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[DOTPR_I:%.*]] = load i8, i8* @e, align 1
; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* @d, align 4
; CHECK-NEXT: [[C_PROMOTED_I:%.*]] = load i32, i32* @c, align 4
; CHECK-NEXT: [[TMP1:%.*]] = sub i8 -1, [[DOTPR_I]]
; CHECK-NEXT: [[TMP2:%.*]] = zext i8 [[TMP1]] to i32
; CHECK-NEXT: [[TMP3:%.*]] = add nuw nsw i32 [[TMP2]], 1
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP3]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP3]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP3]], [[N_MOD_VF]]
; CHECK-NEXT: [[CAST_VTC:%.*]] = trunc i32 [[N_VEC]] to i8
; CHECK-NEXT: [[IND_END:%.*]] = add i8 [[DOTPR_I]], [[CAST_VTC]]
; CHECK-NEXT: [[TMP4:%.*]] = insertelement <2 x i32> <i32 -1, i32 -1>, i32 [[C_PROMOTED_I]], i32 0
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP0]], i32 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP4]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP5]] = and <2 x i32> [[BROADCAST_SPLAT]], [[VEC_PHI]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP7:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[TMP5]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP3]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[DOTPR_I]], [[ENTRY:%.*]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[C_PROMOTED_I]], [[ENTRY]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[COND_END_I:%.*]]
; CHECK: cond.end.i:
; CHECK-NEXT: [[INC4_I:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC_I:%.*]], [[COND_END_I]] ]
; CHECK-NEXT: [[AND3_I:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[AND_I:%.*]], [[COND_END_I]] ]
; CHECK-NEXT: [[AND_I]] = and i32 [[TMP0]], [[AND3_I]]
; CHECK-NEXT: [[INC_I]] = add i8 [[INC4_I]], 1
; CHECK-NEXT: [[TOBOOL_I:%.*]] = icmp eq i8 [[INC_I]], 0
; CHECK-NEXT: br i1 [[TOBOOL_I]], label [[LOOPEXIT]], label [[COND_END_I]], !llvm.loop [[LOOP35:![0-9]+]]
; CHECK: loopexit:
; CHECK-NEXT: [[AND_I_LCSSA:%.*]] = phi i32 [ [[AND_I]], [[COND_END_I]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i32 [[AND_I_LCSSA]]
;
; IND-LABEL: @testoverflowcheck(
; IND-NEXT: entry:
; IND-NEXT: [[DOTPR_I:%.*]] = load i8, i8* @e, align 1
; IND-NEXT: [[TMP0:%.*]] = load i32, i32* @d, align 4
; IND-NEXT: [[C_PROMOTED_I:%.*]] = load i32, i32* @c, align 4
; IND-NEXT: [[TMP1:%.*]] = xor i8 [[DOTPR_I]], -1
; IND-NEXT: [[TMP2:%.*]] = zext i8 [[TMP1]] to i32
; IND-NEXT: [[TMP3:%.*]] = add nuw nsw i32 [[TMP2]], 1
; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp eq i8 [[DOTPR_I]], -1
; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: [[N_VEC:%.*]] = and i32 [[TMP3]], 510
; IND-NEXT: [[CAST_VTC:%.*]] = trunc i32 [[N_VEC]] to i8
; IND-NEXT: [[IND_END:%.*]] = add i8 [[DOTPR_I]], [[CAST_VTC]]
; IND-NEXT: [[TMP4:%.*]] = insertelement <2 x i32> <i32 poison, i32 -1>, i32 [[C_PROMOTED_I]], i64 0
; IND-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP0]], i64 0
; IND-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; IND-NEXT: [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; IND-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[TMP6:%.*]] = and <2 x i32> [[TMP4]], [[BROADCAST_SPLAT]]
; IND-NEXT: [[TMP7:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[TMP6]])
; IND-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP3]], [[N_VEC]]
; IND-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[DOTPR_I]], [[ENTRY:%.*]] ]
; IND-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ [[C_PROMOTED_I]], [[ENTRY]] ]
; IND-NEXT: br label [[COND_END_I:%.*]]
; IND: cond.end.i:
; IND-NEXT: [[INC4_I:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC_I:%.*]], [[COND_END_I]] ]
; IND-NEXT: [[TMP8:%.*]] = and i32 [[BC_MERGE_RDX]], [[TMP0]]
; IND-NEXT: [[INC_I]] = add i8 [[INC4_I]], 1
; IND-NEXT: [[TOBOOL_I:%.*]] = icmp eq i8 [[INC_I]], 0
; IND-NEXT: br i1 [[TOBOOL_I]], label [[LOOPEXIT]], label [[COND_END_I]], !llvm.loop [[LOOP35:![0-9]+]]
; IND: loopexit:
; IND-NEXT: [[AND_I_LCSSA:%.*]] = phi i32 [ [[TMP8]], [[COND_END_I]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
; IND-NEXT: ret i32 [[AND_I_LCSSA]]
;
; UNROLL-LABEL: @testoverflowcheck(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: [[DOTPR_I:%.*]] = load i8, i8* @e, align 1
; UNROLL-NEXT: [[TMP0:%.*]] = load i32, i32* @d, align 4
; UNROLL-NEXT: [[C_PROMOTED_I:%.*]] = load i32, i32* @c, align 4
; UNROLL-NEXT: [[TMP1:%.*]] = xor i8 [[DOTPR_I]], -1
; UNROLL-NEXT: [[TMP2:%.*]] = zext i8 [[TMP1]] to i32
; UNROLL-NEXT: [[TMP3:%.*]] = add nuw nsw i32 [[TMP2]], 1
; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ugt i8 [[DOTPR_I]], -4
; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: [[N_VEC:%.*]] = and i32 [[TMP3]], 508
; UNROLL-NEXT: [[CAST_VTC:%.*]] = trunc i32 [[N_VEC]] to i8
; UNROLL-NEXT: [[IND_END:%.*]] = add i8 [[DOTPR_I]], [[CAST_VTC]]
; UNROLL-NEXT: [[TMP4:%.*]] = insertelement <2 x i32> <i32 poison, i32 -1>, i32 [[C_PROMOTED_I]], i64 0
; UNROLL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP0]], i64 0
; UNROLL-NEXT: [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <2 x i32> poison, i32 [[TMP0]], i64 0
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NEXT: [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[TMP6:%.*]] = and <2 x i32> [[BROADCAST_SPLATINSERT2]], [[BROADCAST_SPLATINSERT]]
; UNROLL-NEXT: [[TMP7:%.*]] = shufflevector <2 x i32> [[TMP6]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NEXT: [[BIN_RDX:%.*]] = and <2 x i32> [[TMP7]], [[TMP4]]
; UNROLL-NEXT: [[TMP8:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[BIN_RDX]])
; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP3]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[DOTPR_I]], [[ENTRY:%.*]] ]
; UNROLL-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP8]], [[MIDDLE_BLOCK]] ], [ [[C_PROMOTED_I]], [[ENTRY]] ]
; UNROLL-NEXT: br label [[COND_END_I:%.*]]
; UNROLL: cond.end.i:
; UNROLL-NEXT: [[INC4_I:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC_I:%.*]], [[COND_END_I]] ]
; UNROLL-NEXT: [[TMP9:%.*]] = and i32 [[BC_MERGE_RDX]], [[TMP0]]
; UNROLL-NEXT: [[INC_I]] = add i8 [[INC4_I]], 1
; UNROLL-NEXT: [[TOBOOL_I:%.*]] = icmp eq i8 [[INC_I]], 0
; UNROLL-NEXT: br i1 [[TOBOOL_I]], label [[LOOPEXIT]], label [[COND_END_I]], !llvm.loop [[LOOP35:![0-9]+]]
; UNROLL: loopexit:
; UNROLL-NEXT: [[AND_I_LCSSA:%.*]] = phi i32 [ [[TMP9]], [[COND_END_I]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
; UNROLL-NEXT: ret i32 [[AND_I_LCSSA]]
;
; UNROLL-NO-IC-LABEL: @testoverflowcheck(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: [[DOTPR_I:%.*]] = load i8, i8* @e, align 1
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = load i32, i32* @d, align 4
; UNROLL-NO-IC-NEXT: [[C_PROMOTED_I:%.*]] = load i32, i32* @c, align 4
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = sub i8 -1, [[DOTPR_I]]
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = zext i8 [[TMP1]] to i32
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = add nuw nsw i32 [[TMP2]], 1
; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP3]], 4
; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP3]], 4
; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP3]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT: [[CAST_VTC:%.*]] = trunc i32 [[N_VEC]] to i8
; UNROLL-NO-IC-NEXT: [[IND_END:%.*]] = add i8 [[DOTPR_I]], [[CAST_VTC]]
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = insertelement <2 x i32> <i32 -1, i32 -1>, i32 [[C_PROMOTED_I]], i32 0
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP0]], i32 0
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <2 x i32> poison, i32 [[TMP0]], i32 0
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT3:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT2]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP4]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i32> [ <i32 -1, i32 -1>, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[TMP5]] = and <2 x i32> [[BROADCAST_SPLAT]], [[VEC_PHI]]
; UNROLL-NO-IC-NEXT: [[TMP6]] = and <2 x i32> [[BROADCAST_SPLAT3]], [[VEC_PHI1]]
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[BIN_RDX:%.*]] = and <2 x i32> [[TMP6]], [[TMP5]]
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[BIN_RDX]])
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP3]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[DOTPR_I]], [[ENTRY:%.*]] ]
; UNROLL-NO-IC-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[C_PROMOTED_I]], [[ENTRY]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: br label [[COND_END_I:%.*]]
; UNROLL-NO-IC: cond.end.i:
; UNROLL-NO-IC-NEXT: [[INC4_I:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC_I:%.*]], [[COND_END_I]] ]
; UNROLL-NO-IC-NEXT: [[AND3_I:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[AND_I:%.*]], [[COND_END_I]] ]
; UNROLL-NO-IC-NEXT: [[AND_I]] = and i32 [[TMP0]], [[AND3_I]]
; UNROLL-NO-IC-NEXT: [[INC_I]] = add i8 [[INC4_I]], 1
; UNROLL-NO-IC-NEXT: [[TOBOOL_I:%.*]] = icmp eq i8 [[INC_I]], 0
; UNROLL-NO-IC-NEXT: br i1 [[TOBOOL_I]], label [[LOOPEXIT]], label [[COND_END_I]], !llvm.loop [[LOOP35:![0-9]+]]
; UNROLL-NO-IC: loopexit:
; UNROLL-NO-IC-NEXT: [[AND_I_LCSSA:%.*]] = phi i32 [ [[AND_I]], [[COND_END_I]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: ret i32 [[AND_I_LCSSA]]
;
; INTERLEAVE-LABEL: @testoverflowcheck(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: [[DOTPR_I:%.*]] = load i8, i8* @e, align 1
; INTERLEAVE-NEXT: [[TMP0:%.*]] = load i32, i32* @d, align 4
; INTERLEAVE-NEXT: [[C_PROMOTED_I:%.*]] = load i32, i32* @c, align 4
; INTERLEAVE-NEXT: [[TMP1:%.*]] = xor i8 [[DOTPR_I]], -1
; INTERLEAVE-NEXT: [[TMP2:%.*]] = zext i8 [[TMP1]] to i32
; INTERLEAVE-NEXT: [[TMP3:%.*]] = add nuw nsw i32 [[TMP2]], 1
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ugt i8 [[DOTPR_I]], -8
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i32 [[TMP3]], 504
; INTERLEAVE-NEXT: [[CAST_VTC:%.*]] = trunc i32 [[N_VEC]] to i8
; INTERLEAVE-NEXT: [[IND_END:%.*]] = add i8 [[DOTPR_I]], [[CAST_VTC]]
; INTERLEAVE-NEXT: [[TMP4:%.*]] = insertelement <4 x i32> <i32 poison, i32 -1, i32 -1, i32 -1>, i32 [[C_PROMOTED_I]], i64 0
; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TMP0]], i64 0
; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <4 x i32> poison, i32 [[TMP0]], i64 0
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
; INTERLEAVE-NEXT: [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: [[TMP6:%.*]] = and <4 x i32> [[BROADCAST_SPLATINSERT2]], [[BROADCAST_SPLATINSERT]]
; INTERLEAVE-NEXT: [[TMP7:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> poison, <4 x i32> zeroinitializer
; INTERLEAVE-NEXT: [[BIN_RDX:%.*]] = and <4 x i32> [[TMP7]], [[TMP4]]
; INTERLEAVE-NEXT: [[TMP8:%.*]] = call i32 @llvm.vector.reduce.and.v4i32(<4 x i32> [[BIN_RDX]])
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP3]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[DOTPR_I]], [[ENTRY:%.*]] ]
; INTERLEAVE-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP8]], [[MIDDLE_BLOCK]] ], [ [[C_PROMOTED_I]], [[ENTRY]] ]
; INTERLEAVE-NEXT: br label [[COND_END_I:%.*]]
; INTERLEAVE: cond.end.i:
; INTERLEAVE-NEXT: [[INC4_I:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC_I:%.*]], [[COND_END_I]] ]
; INTERLEAVE-NEXT: [[TMP9:%.*]] = and i32 [[BC_MERGE_RDX]], [[TMP0]]
; INTERLEAVE-NEXT: [[INC_I]] = add i8 [[INC4_I]], 1
; INTERLEAVE-NEXT: [[TOBOOL_I:%.*]] = icmp eq i8 [[INC_I]], 0
; INTERLEAVE-NEXT: br i1 [[TOBOOL_I]], label [[LOOPEXIT]], label [[COND_END_I]], !llvm.loop [[LOOP35:![0-9]+]]
; INTERLEAVE: loopexit:
; INTERLEAVE-NEXT: [[AND_I_LCSSA:%.*]] = phi i32 [ [[TMP9]], [[COND_END_I]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: ret i32 [[AND_I_LCSSA]]
;
entry:
%.pr.i = load i8, i8* @e, align 1
%0 = load i32, i32* @d, align 4
%c.promoted.i = load i32, i32* @c, align 4
br label %cond.end.i
cond.end.i:
%inc4.i = phi i8 [ %.pr.i, %entry ], [ %inc.i, %cond.end.i ]
%and3.i = phi i32 [ %c.promoted.i, %entry ], [ %and.i, %cond.end.i ]
%and.i = and i32 %0, %and3.i
%inc.i = add i8 %inc4.i, 1
%tobool.i = icmp eq i8 %inc.i, 0
br i1 %tobool.i, label %loopexit, label %cond.end.i
loopexit:
ret i32 %and.i
}
; The SCEV expression of %sphi is (zext i8 {%t,+,1}<%loop> to i32)
; In order to recognize %sphi as an induction PHI and vectorize this loop,
; we need to convert the SCEV expression into an AddRecExpr.
; The expression gets converted to {zext i8 %t to i32,+,1}.
define void @wrappingindvars1(i8 %t, i32 %len, i32 *%A) {
; CHECK-LABEL: @wrappingindvars1(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[ST:%.*]] = zext i8 [[T:%.*]] to i16
; CHECK-NEXT: [[EXT:%.*]] = zext i8 [[T]] to i32
; CHECK-NEXT: [[ECMP:%.*]] = icmp ult i16 [[ST]], 42
; CHECK-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
; CHECK: loop.preheader:
; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; CHECK: vector.scevcheck:
; CHECK-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
; CHECK-NEXT: [[TMP2:%.*]] = add i8 [[T]], [[TMP1]]
; CHECK-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[T]]
; CHECK-NEXT: [[TMP4:%.*]] = icmp ugt i32 [[LEN]], 255
; CHECK-NEXT: [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]
; CHECK-NEXT: [[TMP6:%.*]] = trunc i32 [[LEN]] to i8
; CHECK-NEXT: [[TMP7:%.*]] = add i8 [[T]], [[TMP6]]
; CHECK-NEXT: [[TMP8:%.*]] = icmp slt i8 [[TMP7]], [[T]]
; CHECK-NEXT: [[TMP9:%.*]] = icmp ugt i32 [[LEN]], 255
; CHECK-NEXT: [[TMP10:%.*]] = or i1 [[TMP8]], [[TMP9]]
; CHECK-NEXT: [[TMP11:%.*]] = or i1 [[TMP5]], [[TMP10]]
; CHECK-NEXT: br i1 [[TMP11]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]
; CHECK-NEXT: [[CAST_VTC:%.*]] = trunc i32 [[N_VEC]] to i8
; CHECK-NEXT: [[IND_END:%.*]] = add i8 [[T]], [[CAST_VTC]]
; CHECK-NEXT: [[IND_END2:%.*]] = add i32 [[EXT]], [[N_VEC]]
; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[EXT]], i32 0
; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP12:%.*]] = trunc i32 [[INDEX]] to i8
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[T]], [[TMP12]]
; CHECK-NEXT: [[TMP13:%.*]] = add i8 [[OFFSET_IDX]], 0
; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i8 [[TMP13]]
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[TMP14]], i32 0
; CHECK-NEXT: [[TMP16:%.*]] = bitcast i32* [[TMP15]] to <2 x i32>*
; CHECK-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP16]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; CHECK-NEXT: [[TMP17:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
; CHECK-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; CHECK-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[EXT]], [[LOOP_PREHEADER]] ], [ [[EXT]], [[VECTOR_SCEVCHECK]] ]
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[SPHI:%.*]] = phi i32 [ [[IDX_INC_EXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i8 [[IDX]]
; CHECK-NEXT: store i32 [[SPHI]], i32* [[PTR]], align 4
; CHECK-NEXT: [[IDX_INC]] = add i8 [[IDX]], 1
; CHECK-NEXT: [[IDX_INC_EXT]] = zext i8 [[IDX_INC]] to i32
; CHECK-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
; CHECK-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
; CHECK-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP37:![0-9]+]]
; CHECK: exit.loopexit:
; CHECK-NEXT: br label [[EXIT]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
; IND-LABEL: @wrappingindvars1(
; IND-NEXT: entry:
; IND-NEXT: [[EXT:%.*]] = zext i8 [[T:%.*]] to i32
; IND-NEXT: [[ECMP:%.*]] = icmp ult i8 [[T]], 42
; IND-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
; IND: loop.preheader:
; IND-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 2
; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; IND: vector.scevcheck:
; IND-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
; IND-NEXT: [[TMP2:%.*]] = xor i8 [[T]], -1
; IND-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]
; IND-NEXT: [[TMP4:%.*]] = trunc i32 [[LEN]] to i8
; IND-NEXT: [[TMP5:%.*]] = add i8 [[TMP4]], [[T]]
; IND-NEXT: [[TMP6:%.*]] = icmp slt i8 [[TMP5]], [[T]]
; IND-NEXT: [[TMP7:%.*]] = icmp ugt i32 [[LEN]], 255
; IND-NEXT: [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]
; IND-NEXT: [[TMP9:%.*]] = or i1 [[TMP3]], [[TMP8]]
; IND-NEXT: br i1 [[TMP9]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], -2
; IND-NEXT: [[CAST_VTC:%.*]] = trunc i32 [[N_VEC]] to i8
; IND-NEXT: [[IND_END:%.*]] = add i8 [[CAST_VTC]], [[T]]
; IND-NEXT: [[IND_END2:%.*]] = add i32 [[N_VEC]], [[EXT]]
; IND-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[EXT]], i64 0
; IND-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; IND-NEXT: [[INDUCTION:%.*]] = add nuw nsw <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[TMP10:%.*]] = trunc i32 [[INDEX]] to i8
; IND-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[TMP10]], [[T]]
; IND-NEXT: [[TMP11:%.*]] = sext i8 [[OFFSET_IDX]] to i64
; IND-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP11]]
; IND-NEXT: [[TMP13:%.*]] = bitcast i32* [[TMP12]] to <2 x i32>*
; IND-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP13]], align 4
; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; IND-NEXT: [[TMP14:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; IND-NEXT: br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
; IND-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
; IND-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; IND-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[EXT]], [[LOOP_PREHEADER]] ], [ [[EXT]], [[VECTOR_SCEVCHECK]] ]
; IND-NEXT: br label [[LOOP:%.*]]
; IND: loop:
; IND-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; IND-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; IND-NEXT: [[SPHI:%.*]] = phi i32 [ [[IDX_INC_EXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]
; IND-NEXT: [[TMP15:%.*]] = sext i8 [[IDX]] to i64
; IND-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP15]]
; IND-NEXT: store i32 [[SPHI]], i32* [[PTR]], align 4
; IND-NEXT: [[IDX_INC]] = add i8 [[IDX]], 1
; IND-NEXT: [[IDX_INC_EXT]] = zext i8 [[IDX_INC]] to i32
; IND-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
; IND-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
; IND-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP37:![0-9]+]]
; IND: exit.loopexit:
; IND-NEXT: br label [[EXIT]]
; IND: exit:
; IND-NEXT: ret void
;
; UNROLL-LABEL: @wrappingindvars1(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: [[EXT:%.*]] = zext i8 [[T:%.*]] to i32
; UNROLL-NEXT: [[ECMP:%.*]] = icmp ult i8 [[T]], 42
; UNROLL-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
; UNROLL: loop.preheader:
; UNROLL-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 4
; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; UNROLL: vector.scevcheck:
; UNROLL-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
; UNROLL-NEXT: [[TMP2:%.*]] = xor i8 [[T]], -1
; UNROLL-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]
; UNROLL-NEXT: [[TMP4:%.*]] = trunc i32 [[LEN]] to i8
; UNROLL-NEXT: [[TMP5:%.*]] = add i8 [[TMP4]], [[T]]
; UNROLL-NEXT: [[TMP6:%.*]] = icmp slt i8 [[TMP5]], [[T]]
; UNROLL-NEXT: [[TMP7:%.*]] = icmp ugt i32 [[LEN]], 255
; UNROLL-NEXT: [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]
; UNROLL-NEXT: [[TMP9:%.*]] = or i1 [[TMP3]], [[TMP8]]
; UNROLL-NEXT: br i1 [[TMP9]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], -4
; UNROLL-NEXT: [[CAST_VTC:%.*]] = trunc i32 [[N_VEC]] to i8
; UNROLL-NEXT: [[IND_END:%.*]] = add i8 [[CAST_VTC]], [[T]]
; UNROLL-NEXT: [[IND_END2:%.*]] = add i32 [[N_VEC]], [[EXT]]
; UNROLL-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[EXT]], i64 0
; UNROLL-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NEXT: [[INDUCTION:%.*]] = add nuw nsw <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; UNROLL-NEXT: [[TMP10:%.*]] = trunc i32 [[INDEX]] to i8
; UNROLL-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[TMP10]], [[T]]
; UNROLL-NEXT: [[TMP11:%.*]] = sext i8 [[OFFSET_IDX]] to i64
; UNROLL-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP11]]
; UNROLL-NEXT: [[TMP13:%.*]] = bitcast i32* [[TMP12]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP13]], align 4
; UNROLL-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[TMP12]], i64 2
; UNROLL-NEXT: [[TMP15:%.*]] = bitcast i32* [[TMP14]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP15]], align 4
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 4, i32 4>
; UNROLL-NEXT: [[TMP16:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
; UNROLL-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; UNROLL-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[EXT]], [[LOOP_PREHEADER]] ], [ [[EXT]], [[VECTOR_SCEVCHECK]] ]
; UNROLL-NEXT: br label [[LOOP:%.*]]
; UNROLL: loop:
; UNROLL-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[SPHI:%.*]] = phi i32 [ [[IDX_INC_EXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[TMP17:%.*]] = sext i8 [[IDX]] to i64
; UNROLL-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP17]]
; UNROLL-NEXT: store i32 [[SPHI]], i32* [[PTR]], align 4
; UNROLL-NEXT: [[IDX_INC]] = add i8 [[IDX]], 1
; UNROLL-NEXT: [[IDX_INC_EXT]] = zext i8 [[IDX_INC]] to i32
; UNROLL-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
; UNROLL-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
; UNROLL-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP37:![0-9]+]]
; UNROLL: exit.loopexit:
; UNROLL-NEXT: br label [[EXIT]]
; UNROLL: exit:
; UNROLL-NEXT: ret void
;
; UNROLL-NO-IC-LABEL: @wrappingindvars1(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: [[ST:%.*]] = zext i8 [[T:%.*]] to i16
; UNROLL-NO-IC-NEXT: [[EXT:%.*]] = zext i8 [[T]] to i32
; UNROLL-NO-IC-NEXT: [[ECMP:%.*]] = icmp ult i16 [[ST]], 42
; UNROLL-NO-IC-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
; UNROLL-NO-IC: loop.preheader:
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 4
; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; UNROLL-NO-IC: vector.scevcheck:
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add i8 [[T]], [[TMP1]]
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[T]]
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = icmp ugt i32 [[LEN]], 255
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = trunc i32 [[LEN]] to i8
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = add i8 [[T]], [[TMP6]]
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = icmp slt i8 [[TMP7]], [[T]]
; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = icmp ugt i32 [[LEN]], 255
; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = or i1 [[TMP8]], [[TMP9]]
; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = or i1 [[TMP5]], [[TMP10]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP11]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 4
; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT: [[CAST_VTC:%.*]] = trunc i32 [[N_VEC]] to i8
; UNROLL-NO-IC-NEXT: [[IND_END:%.*]] = add i8 [[T]], [[CAST_VTC]]
; UNROLL-NO-IC-NEXT: [[IND_END2:%.*]] = add i32 [[EXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[EXT]], i32 0
; UNROLL-NO-IC-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT: [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = trunc i32 [[INDEX]] to i8
; UNROLL-NO-IC-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[T]], [[TMP12]]
; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = add i8 [[OFFSET_IDX]], 0
; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = add i8 [[OFFSET_IDX]], 2
; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i8 [[TMP13]]
; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[A]], i8 [[TMP14]]
; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[TMP15]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = bitcast i32* [[TMP17]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP18]], align 4
; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[TMP15]], i32 2
; UNROLL-NO-IC-NEXT: [[TMP20:%.*]] = bitcast i32* [[TMP19]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP20]], align 4
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP21:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[EXT]], [[LOOP_PREHEADER]] ], [ [[EXT]], [[VECTOR_SCEVCHECK]] ]
; UNROLL-NO-IC-NEXT: br label [[LOOP:%.*]]
; UNROLL-NO-IC: loop:
; UNROLL-NO-IC-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[SPHI:%.*]] = phi i32 [ [[IDX_INC_EXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i8 [[IDX]]
; UNROLL-NO-IC-NEXT: store i32 [[SPHI]], i32* [[PTR]], align 4
; UNROLL-NO-IC-NEXT: [[IDX_INC]] = add i8 [[IDX]], 1
; UNROLL-NO-IC-NEXT: [[IDX_INC_EXT]] = zext i8 [[IDX_INC]] to i32
; UNROLL-NO-IC-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
; UNROLL-NO-IC-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
; UNROLL-NO-IC-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP37:![0-9]+]]
; UNROLL-NO-IC: exit.loopexit:
; UNROLL-NO-IC-NEXT: br label [[EXIT]]
; UNROLL-NO-IC: exit:
; UNROLL-NO-IC-NEXT: ret void
;
; INTERLEAVE-LABEL: @wrappingindvars1(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: [[EXT:%.*]] = zext i8 [[T:%.*]] to i32
; INTERLEAVE-NEXT: [[ECMP:%.*]] = icmp ult i8 [[T]], 42
; INTERLEAVE-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
; INTERLEAVE: loop.preheader:
; INTERLEAVE-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 8
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; INTERLEAVE: vector.scevcheck:
; INTERLEAVE-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
; INTERLEAVE-NEXT: [[TMP2:%.*]] = xor i8 [[T]], -1
; INTERLEAVE-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]
; INTERLEAVE-NEXT: [[TMP4:%.*]] = trunc i32 [[LEN]] to i8
; INTERLEAVE-NEXT: [[TMP5:%.*]] = add i8 [[TMP4]], [[T]]
; INTERLEAVE-NEXT: [[TMP6:%.*]] = icmp slt i8 [[TMP5]], [[T]]
; INTERLEAVE-NEXT: [[TMP7:%.*]] = icmp ugt i32 [[LEN]], 255
; INTERLEAVE-NEXT: [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]
; INTERLEAVE-NEXT: [[TMP9:%.*]] = or i1 [[TMP3]], [[TMP8]]
; INTERLEAVE-NEXT: br i1 [[TMP9]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], -8
; INTERLEAVE-NEXT: [[CAST_VTC:%.*]] = trunc i32 [[N_VEC]] to i8
; INTERLEAVE-NEXT: [[IND_END:%.*]] = add i8 [[CAST_VTC]], [[T]]
; INTERLEAVE-NEXT: [[IND_END2:%.*]] = add i32 [[N_VEC]], [[EXT]]
; INTERLEAVE-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[EXT]], i64 0
; INTERLEAVE-NEXT: [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; INTERLEAVE-NEXT: [[INDUCTION:%.*]] = add nuw nsw <4 x i32> [[DOTSPLAT]], <i32 0, i32 1, i32 2, i32 3>
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; INTERLEAVE-NEXT: [[TMP10:%.*]] = trunc i32 [[INDEX]] to i8
; INTERLEAVE-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[TMP10]], [[T]]
; INTERLEAVE-NEXT: [[TMP11:%.*]] = sext i8 [[OFFSET_IDX]] to i64
; INTERLEAVE-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP11]]
; INTERLEAVE-NEXT: [[TMP13:%.*]] = bitcast i32* [[TMP12]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[VEC_IND]], <4 x i32>* [[TMP13]], align 4
; INTERLEAVE-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[TMP12]], i64 4
; INTERLEAVE-NEXT: [[TMP15:%.*]] = bitcast i32* [[TMP14]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[STEP_ADD]], <4 x i32>* [[TMP15]], align 4
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 8, i32 8, i32 8, i32 8>
; INTERLEAVE-NEXT: [[TMP16:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
; INTERLEAVE-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; INTERLEAVE-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[IND_END2]], [[MIDDLE_BLOCK]] ], [ [[EXT]], [[LOOP_PREHEADER]] ], [ [[EXT]], [[VECTOR_SCEVCHECK]] ]
; INTERLEAVE-NEXT: br label [[LOOP:%.*]]
; INTERLEAVE: loop:
; INTERLEAVE-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[SPHI:%.*]] = phi i32 [ [[IDX_INC_EXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[TMP17:%.*]] = sext i8 [[IDX]] to i64
; INTERLEAVE-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP17]]
; INTERLEAVE-NEXT: store i32 [[SPHI]], i32* [[PTR]], align 4
; INTERLEAVE-NEXT: [[IDX_INC]] = add i8 [[IDX]], 1
; INTERLEAVE-NEXT: [[IDX_INC_EXT]] = zext i8 [[IDX_INC]] to i32
; INTERLEAVE-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
; INTERLEAVE-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
; INTERLEAVE-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP37:![0-9]+]]
; INTERLEAVE: exit.loopexit:
; INTERLEAVE-NEXT: br label [[EXIT]]
; INTERLEAVE: exit:
; INTERLEAVE-NEXT: ret void
;
entry:
%st = zext i8 %t to i16
%ext = zext i8 %t to i32
%ecmp = icmp ult i16 %st, 42
br i1 %ecmp, label %loop, label %exit
loop:
%idx = phi i8 [ %t, %entry ], [ %idx.inc, %loop ]
%idx.b = phi i32 [ 0, %entry ], [ %idx.b.inc, %loop ]
%sphi = phi i32 [ %ext, %entry ], [%idx.inc.ext, %loop]
%ptr = getelementptr inbounds i32, i32* %A, i8 %idx
store i32 %sphi, i32* %ptr
%idx.inc = add i8 %idx, 1
%idx.inc.ext = zext i8 %idx.inc to i32
%idx.b.inc = add nuw nsw i32 %idx.b, 1
%c = icmp ult i32 %idx.b, %len
br i1 %c, label %loop, label %exit
exit:
ret void
}
; The SCEV expression of %sphi is (4 * (zext i8 {%t,+,1}<%loop> to i32))
; In order to recognize %sphi as an induction PHI and vectorize this loop,
; we need to convert the SCEV expression into an AddRecExpr.
; The expression gets converted to ({4 * (zext %t to i32),+,4}).
define void @wrappingindvars2(i8 %t, i32 %len, i32 *%A) {
; CHECK-LABEL: @wrappingindvars2(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[ST:%.*]] = zext i8 [[T:%.*]] to i16
; CHECK-NEXT: [[EXT:%.*]] = zext i8 [[T]] to i32
; CHECK-NEXT: [[EXT_MUL:%.*]] = mul i32 [[EXT]], 4
; CHECK-NEXT: [[ECMP:%.*]] = icmp ult i16 [[ST]], 42
; CHECK-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
; CHECK: loop.preheader:
; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; CHECK: vector.scevcheck:
; CHECK-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
; CHECK-NEXT: [[TMP2:%.*]] = add i8 [[T]], [[TMP1]]
; CHECK-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[T]]
; CHECK-NEXT: [[TMP4:%.*]] = icmp ugt i32 [[LEN]], 255
; CHECK-NEXT: [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]
; CHECK-NEXT: [[TMP6:%.*]] = trunc i32 [[LEN]] to i8
; CHECK-NEXT: [[TMP7:%.*]] = add i8 [[T]], [[TMP6]]
; CHECK-NEXT: [[TMP8:%.*]] = icmp slt i8 [[TMP7]], [[T]]
; CHECK-NEXT: [[TMP9:%.*]] = icmp ugt i32 [[LEN]], 255
; CHECK-NEXT: [[TMP10:%.*]] = or i1 [[TMP8]], [[TMP9]]
; CHECK-NEXT: [[TMP11:%.*]] = or i1 [[TMP5]], [[TMP10]]
; CHECK-NEXT: br i1 [[TMP11]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]
; CHECK-NEXT: [[CAST_VTC:%.*]] = trunc i32 [[N_VEC]] to i8
; CHECK-NEXT: [[IND_END:%.*]] = add i8 [[T]], [[CAST_VTC]]
; CHECK-NEXT: [[TMP12:%.*]] = mul i32 [[N_VEC]], 4
; CHECK-NEXT: [[IND_END1:%.*]] = add i32 [[EXT_MUL]], [[TMP12]]
; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[EXT_MUL]], i32 0
; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 4>
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP13:%.*]] = trunc i32 [[INDEX]] to i8
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[T]], [[TMP13]]
; CHECK-NEXT: [[TMP14:%.*]] = add i8 [[OFFSET_IDX]], 0
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i8 [[TMP14]]
; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[TMP15]], i32 0
; CHECK-NEXT: [[TMP17:%.*]] = bitcast i32* [[TMP16]] to <2 x i32>*
; CHECK-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP17]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 8, i32 8>
; CHECK-NEXT: [[TMP18:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
; CHECK-NEXT: [[BC_RESUME_VAL2:%.*]] = phi i32 [ [[IND_END1]], [[MIDDLE_BLOCK]] ], [ [[EXT_MUL]], [[LOOP_PREHEADER]] ], [ [[EXT_MUL]], [[VECTOR_SCEVCHECK]] ]
; CHECK-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[SPHI:%.*]] = phi i32 [ [[MUL:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i8 [[IDX]]
; CHECK-NEXT: store i32 [[SPHI]], i32* [[PTR]], align 4
; CHECK-NEXT: [[IDX_INC]] = add i8 [[IDX]], 1
; CHECK-NEXT: [[IDX_INC_EXT:%.*]] = zext i8 [[IDX_INC]] to i32
; CHECK-NEXT: [[MUL]] = mul i32 [[IDX_INC_EXT]], 4
; CHECK-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
; CHECK-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
; CHECK-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP39:![0-9]+]]
; CHECK: exit.loopexit:
; CHECK-NEXT: br label [[EXIT]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
; IND-LABEL: @wrappingindvars2(
; IND-NEXT: entry:
; IND-NEXT: [[EXT:%.*]] = zext i8 [[T:%.*]] to i32
; IND-NEXT: [[EXT_MUL:%.*]] = shl nuw nsw i32 [[EXT]], 2
; IND-NEXT: [[ECMP:%.*]] = icmp ult i8 [[T]], 42
; IND-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
; IND: loop.preheader:
; IND-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 2
; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; IND: vector.scevcheck:
; IND-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
; IND-NEXT: [[TMP2:%.*]] = xor i8 [[T]], -1
; IND-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]
; IND-NEXT: [[TMP4:%.*]] = trunc i32 [[LEN]] to i8
; IND-NEXT: [[TMP5:%.*]] = add i8 [[TMP4]], [[T]]
; IND-NEXT: [[TMP6:%.*]] = icmp slt i8 [[TMP5]], [[T]]
; IND-NEXT: [[TMP7:%.*]] = icmp ugt i32 [[LEN]], 255
; IND-NEXT: [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]
; IND-NEXT: [[TMP9:%.*]] = or i1 [[TMP3]], [[TMP8]]
; IND-NEXT: br i1 [[TMP9]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], -2
; IND-NEXT: [[CAST_VTC:%.*]] = trunc i32 [[N_VEC]] to i8
; IND-NEXT: [[IND_END:%.*]] = add i8 [[CAST_VTC]], [[T]]
; IND-NEXT: [[TMP10:%.*]] = add i32 [[N_VEC]], [[EXT]]
; IND-NEXT: [[IND_END1:%.*]] = shl i32 [[TMP10]], 2
; IND-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[EXT_MUL]], i64 0
; IND-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; IND-NEXT: [[INDUCTION:%.*]] = add nuw nsw <2 x i32> [[DOTSPLAT]], <i32 0, i32 4>
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[TMP11:%.*]] = trunc i32 [[INDEX]] to i8
; IND-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[TMP11]], [[T]]
; IND-NEXT: [[TMP12:%.*]] = sext i8 [[OFFSET_IDX]] to i64
; IND-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP12]]
; IND-NEXT: [[TMP14:%.*]] = bitcast i32* [[TMP13]] to <2 x i32>*
; IND-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP14]], align 4
; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 8, i32 8>
; IND-NEXT: [[TMP15:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; IND-NEXT: br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
; IND-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
; IND-NEXT: [[BC_RESUME_VAL2:%.*]] = phi i32 [ [[IND_END1]], [[MIDDLE_BLOCK]] ], [ [[EXT_MUL]], [[LOOP_PREHEADER]] ], [ [[EXT_MUL]], [[VECTOR_SCEVCHECK]] ]
; IND-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; IND-NEXT: br label [[LOOP:%.*]]
; IND: loop:
; IND-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; IND-NEXT: [[SPHI:%.*]] = phi i32 [ [[MUL:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ]
; IND-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]
; IND-NEXT: [[TMP16:%.*]] = sext i8 [[IDX]] to i64
; IND-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP16]]
; IND-NEXT: store i32 [[SPHI]], i32* [[PTR]], align 4
; IND-NEXT: [[IDX_INC]] = add i8 [[IDX]], 1
; IND-NEXT: [[IDX_INC_EXT:%.*]] = zext i8 [[IDX_INC]] to i32
; IND-NEXT: [[MUL]] = shl nuw nsw i32 [[IDX_INC_EXT]], 2
; IND-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
; IND-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
; IND-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP39:![0-9]+]]
; IND: exit.loopexit:
; IND-NEXT: br label [[EXIT]]
; IND: exit:
; IND-NEXT: ret void
;
; UNROLL-LABEL: @wrappingindvars2(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: [[EXT:%.*]] = zext i8 [[T:%.*]] to i32
; UNROLL-NEXT: [[EXT_MUL:%.*]] = shl nuw nsw i32 [[EXT]], 2
; UNROLL-NEXT: [[ECMP:%.*]] = icmp ult i8 [[T]], 42
; UNROLL-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
; UNROLL: loop.preheader:
; UNROLL-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 4
; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; UNROLL: vector.scevcheck:
; UNROLL-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
; UNROLL-NEXT: [[TMP2:%.*]] = xor i8 [[T]], -1
; UNROLL-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]
; UNROLL-NEXT: [[TMP4:%.*]] = trunc i32 [[LEN]] to i8
; UNROLL-NEXT: [[TMP5:%.*]] = add i8 [[TMP4]], [[T]]
; UNROLL-NEXT: [[TMP6:%.*]] = icmp slt i8 [[TMP5]], [[T]]
; UNROLL-NEXT: [[TMP7:%.*]] = icmp ugt i32 [[LEN]], 255
; UNROLL-NEXT: [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]
; UNROLL-NEXT: [[TMP9:%.*]] = or i1 [[TMP3]], [[TMP8]]
; UNROLL-NEXT: br i1 [[TMP9]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], -4
; UNROLL-NEXT: [[CAST_VTC:%.*]] = trunc i32 [[N_VEC]] to i8
; UNROLL-NEXT: [[IND_END:%.*]] = add i8 [[CAST_VTC]], [[T]]
; UNROLL-NEXT: [[TMP10:%.*]] = add i32 [[N_VEC]], [[EXT]]
; UNROLL-NEXT: [[IND_END1:%.*]] = shl i32 [[TMP10]], 2
; UNROLL-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[EXT_MUL]], i64 0
; UNROLL-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NEXT: [[INDUCTION:%.*]] = add nuw nsw <2 x i32> [[DOTSPLAT]], <i32 0, i32 4>
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 8, i32 8>
; UNROLL-NEXT: [[TMP11:%.*]] = trunc i32 [[INDEX]] to i8
; UNROLL-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[TMP11]], [[T]]
; UNROLL-NEXT: [[TMP12:%.*]] = sext i8 [[OFFSET_IDX]] to i64
; UNROLL-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP12]]
; UNROLL-NEXT: [[TMP14:%.*]] = bitcast i32* [[TMP13]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP14]], align 4
; UNROLL-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[TMP13]], i64 2
; UNROLL-NEXT: [[TMP16:%.*]] = bitcast i32* [[TMP15]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP16]], align 4
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 16, i32 16>
; UNROLL-NEXT: [[TMP17:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
; UNROLL-NEXT: [[BC_RESUME_VAL2:%.*]] = phi i32 [ [[IND_END1]], [[MIDDLE_BLOCK]] ], [ [[EXT_MUL]], [[LOOP_PREHEADER]] ], [ [[EXT_MUL]], [[VECTOR_SCEVCHECK]] ]
; UNROLL-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; UNROLL-NEXT: br label [[LOOP:%.*]]
; UNROLL: loop:
; UNROLL-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[SPHI:%.*]] = phi i32 [ [[MUL:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[TMP18:%.*]] = sext i8 [[IDX]] to i64
; UNROLL-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP18]]
; UNROLL-NEXT: store i32 [[SPHI]], i32* [[PTR]], align 4
; UNROLL-NEXT: [[IDX_INC]] = add i8 [[IDX]], 1
; UNROLL-NEXT: [[IDX_INC_EXT:%.*]] = zext i8 [[IDX_INC]] to i32
; UNROLL-NEXT: [[MUL]] = shl nuw nsw i32 [[IDX_INC_EXT]], 2
; UNROLL-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
; UNROLL-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
; UNROLL-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP39:![0-9]+]]
; UNROLL: exit.loopexit:
; UNROLL-NEXT: br label [[EXIT]]
; UNROLL: exit:
; UNROLL-NEXT: ret void
;
; UNROLL-NO-IC-LABEL: @wrappingindvars2(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: [[ST:%.*]] = zext i8 [[T:%.*]] to i16
; UNROLL-NO-IC-NEXT: [[EXT:%.*]] = zext i8 [[T]] to i32
; UNROLL-NO-IC-NEXT: [[EXT_MUL:%.*]] = mul i32 [[EXT]], 4
; UNROLL-NO-IC-NEXT: [[ECMP:%.*]] = icmp ult i16 [[ST]], 42
; UNROLL-NO-IC-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
; UNROLL-NO-IC: loop.preheader:
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 4
; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; UNROLL-NO-IC: vector.scevcheck:
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add i8 [[T]], [[TMP1]]
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[T]]
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = icmp ugt i32 [[LEN]], 255
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = trunc i32 [[LEN]] to i8
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = add i8 [[T]], [[TMP6]]
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = icmp slt i8 [[TMP7]], [[T]]
; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = icmp ugt i32 [[LEN]], 255
; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = or i1 [[TMP8]], [[TMP9]]
; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = or i1 [[TMP5]], [[TMP10]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP11]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 4
; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT: [[CAST_VTC:%.*]] = trunc i32 [[N_VEC]] to i8
; UNROLL-NO-IC-NEXT: [[IND_END:%.*]] = add i8 [[T]], [[CAST_VTC]]
; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = mul i32 [[N_VEC]], 4
; UNROLL-NO-IC-NEXT: [[IND_END1:%.*]] = add i32 [[EXT_MUL]], [[TMP12]]
; UNROLL-NO-IC-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[EXT_MUL]], i32 0
; UNROLL-NO-IC-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT: [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 4>
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 8, i32 8>
; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = trunc i32 [[INDEX]] to i8
; UNROLL-NO-IC-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[T]], [[TMP13]]
; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = add i8 [[OFFSET_IDX]], 0
; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = add i8 [[OFFSET_IDX]], 2
; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i8 [[TMP14]]
; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = getelementptr inbounds i32, i32* [[A]], i8 [[TMP15]]
; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = getelementptr inbounds i32, i32* [[TMP16]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = bitcast i32* [[TMP18]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP19]], align 4
; UNROLL-NO-IC-NEXT: [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[TMP16]], i32 2
; UNROLL-NO-IC-NEXT: [[TMP21:%.*]] = bitcast i32* [[TMP20]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP21]], align 4
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], <i32 8, i32 8>
; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL2:%.*]] = phi i32 [ [[IND_END1]], [[MIDDLE_BLOCK]] ], [ [[EXT_MUL]], [[LOOP_PREHEADER]] ], [ [[EXT_MUL]], [[VECTOR_SCEVCHECK]] ]
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; UNROLL-NO-IC-NEXT: br label [[LOOP:%.*]]
; UNROLL-NO-IC: loop:
; UNROLL-NO-IC-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[SPHI:%.*]] = phi i32 [ [[MUL:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i8 [[IDX]]
; UNROLL-NO-IC-NEXT: store i32 [[SPHI]], i32* [[PTR]], align 4
; UNROLL-NO-IC-NEXT: [[IDX_INC]] = add i8 [[IDX]], 1
; UNROLL-NO-IC-NEXT: [[IDX_INC_EXT:%.*]] = zext i8 [[IDX_INC]] to i32
; UNROLL-NO-IC-NEXT: [[MUL]] = mul i32 [[IDX_INC_EXT]], 4
; UNROLL-NO-IC-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
; UNROLL-NO-IC-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
; UNROLL-NO-IC-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP39:![0-9]+]]
; UNROLL-NO-IC: exit.loopexit:
; UNROLL-NO-IC-NEXT: br label [[EXIT]]
; UNROLL-NO-IC: exit:
; UNROLL-NO-IC-NEXT: ret void
;
; INTERLEAVE-LABEL: @wrappingindvars2(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: [[EXT:%.*]] = zext i8 [[T:%.*]] to i32
; INTERLEAVE-NEXT: [[EXT_MUL:%.*]] = shl nuw nsw i32 [[EXT]], 2
; INTERLEAVE-NEXT: [[ECMP:%.*]] = icmp ult i8 [[T]], 42
; INTERLEAVE-NEXT: br i1 [[ECMP]], label [[LOOP_PREHEADER:%.*]], label [[EXIT:%.*]]
; INTERLEAVE: loop.preheader:
; INTERLEAVE-NEXT: [[TMP0:%.*]] = add i32 [[LEN:%.*]], 1
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 8
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; INTERLEAVE: vector.scevcheck:
; INTERLEAVE-NEXT: [[TMP1:%.*]] = trunc i32 [[LEN]] to i8
; INTERLEAVE-NEXT: [[TMP2:%.*]] = xor i8 [[T]], -1
; INTERLEAVE-NEXT: [[TMP3:%.*]] = icmp ult i8 [[TMP2]], [[TMP1]]
; INTERLEAVE-NEXT: [[TMP4:%.*]] = trunc i32 [[LEN]] to i8
; INTERLEAVE-NEXT: [[TMP5:%.*]] = add i8 [[TMP4]], [[T]]
; INTERLEAVE-NEXT: [[TMP6:%.*]] = icmp slt i8 [[TMP5]], [[T]]
; INTERLEAVE-NEXT: [[TMP7:%.*]] = icmp ugt i32 [[LEN]], 255
; INTERLEAVE-NEXT: [[TMP8:%.*]] = or i1 [[TMP6]], [[TMP7]]
; INTERLEAVE-NEXT: [[TMP9:%.*]] = or i1 [[TMP3]], [[TMP8]]
; INTERLEAVE-NEXT: br i1 [[TMP9]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], -8
; INTERLEAVE-NEXT: [[CAST_VTC:%.*]] = trunc i32 [[N_VEC]] to i8
; INTERLEAVE-NEXT: [[IND_END:%.*]] = add i8 [[CAST_VTC]], [[T]]
; INTERLEAVE-NEXT: [[TMP10:%.*]] = add i32 [[N_VEC]], [[EXT]]
; INTERLEAVE-NEXT: [[IND_END1:%.*]] = shl i32 [[TMP10]], 2
; INTERLEAVE-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[EXT_MUL]], i64 0
; INTERLEAVE-NEXT: [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; INTERLEAVE-NEXT: [[INDUCTION:%.*]] = add nuw nsw <4 x i32> [[DOTSPLAT]], <i32 0, i32 4, i32 8, i32 12>
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], <i32 16, i32 16, i32 16, i32 16>
; INTERLEAVE-NEXT: [[TMP11:%.*]] = trunc i32 [[INDEX]] to i8
; INTERLEAVE-NEXT: [[OFFSET_IDX:%.*]] = add i8 [[TMP11]], [[T]]
; INTERLEAVE-NEXT: [[TMP12:%.*]] = sext i8 [[OFFSET_IDX]] to i64
; INTERLEAVE-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP12]]
; INTERLEAVE-NEXT: [[TMP14:%.*]] = bitcast i32* [[TMP13]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[VEC_IND]], <4 x i32>* [[TMP14]], align 4
; INTERLEAVE-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[TMP13]], i64 4
; INTERLEAVE-NEXT: [[TMP16:%.*]] = bitcast i32* [[TMP15]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[STEP_ADD]], <4 x i32>* [[TMP16]], align 4
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 32, i32 32, i32 32, i32 32>
; INTERLEAVE-NEXT: [[TMP17:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[T]], [[LOOP_PREHEADER]] ], [ [[T]], [[VECTOR_SCEVCHECK]] ]
; INTERLEAVE-NEXT: [[BC_RESUME_VAL2:%.*]] = phi i32 [ [[IND_END1]], [[MIDDLE_BLOCK]] ], [ [[EXT_MUL]], [[LOOP_PREHEADER]] ], [ [[EXT_MUL]], [[VECTOR_SCEVCHECK]] ]
; INTERLEAVE-NEXT: [[BC_RESUME_VAL3:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; INTERLEAVE-NEXT: br label [[LOOP:%.*]]
; INTERLEAVE: loop:
; INTERLEAVE-NEXT: [[IDX:%.*]] = phi i8 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[SPHI:%.*]] = phi i32 [ [[MUL:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[IDX_B:%.*]] = phi i32 [ [[IDX_B_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL3]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[TMP18:%.*]] = sext i8 [[IDX]] to i64
; INTERLEAVE-NEXT: [[PTR:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP18]]
; INTERLEAVE-NEXT: store i32 [[SPHI]], i32* [[PTR]], align 4
; INTERLEAVE-NEXT: [[IDX_INC]] = add i8 [[IDX]], 1
; INTERLEAVE-NEXT: [[IDX_INC_EXT:%.*]] = zext i8 [[IDX_INC]] to i32
; INTERLEAVE-NEXT: [[MUL]] = shl nuw nsw i32 [[IDX_INC_EXT]], 2
; INTERLEAVE-NEXT: [[IDX_B_INC]] = add nuw nsw i32 [[IDX_B]], 1
; INTERLEAVE-NEXT: [[C:%.*]] = icmp ult i32 [[IDX_B]], [[LEN]]
; INTERLEAVE-NEXT: br i1 [[C]], label [[LOOP]], label [[EXIT_LOOPEXIT]], !llvm.loop [[LOOP39:![0-9]+]]
; INTERLEAVE: exit.loopexit:
; INTERLEAVE-NEXT: br label [[EXIT]]
; INTERLEAVE: exit:
; INTERLEAVE-NEXT: ret void
;
entry:
%st = zext i8 %t to i16
%ext = zext i8 %t to i32
%ext.mul = mul i32 %ext, 4
%ecmp = icmp ult i16 %st, 42
br i1 %ecmp, label %loop, label %exit
loop:
%idx = phi i8 [ %t, %entry ], [ %idx.inc, %loop ]
%sphi = phi i32 [ %ext.mul, %entry ], [%mul, %loop]
%idx.b = phi i32 [ 0, %entry ], [ %idx.b.inc, %loop ]
%ptr = getelementptr inbounds i32, i32* %A, i8 %idx
store i32 %sphi, i32* %ptr
%idx.inc = add i8 %idx, 1
%idx.inc.ext = zext i8 %idx.inc to i32
%mul = mul i32 %idx.inc.ext, 4
%idx.b.inc = add nuw nsw i32 %idx.b, 1
%c = icmp ult i32 %idx.b, %len
br i1 %c, label %loop, label %exit
exit:
ret void
}
; Check that we generate vectorized IVs in the pre-header
; instead of widening the scalar IV inside the loop, when
; we know how to do that.
define void @veciv(i32* nocapture %a, i32 %start, i32 %k) {
; CHECK-LABEL: @veciv(
; CHECK-NEXT: for.body.preheader:
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K:%.*]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[K]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[K]], [[N_MOD_VF]]
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 0
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP0]]
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i32 0
; CHECK-NEXT: [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <2 x i32>*
; CHECK-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP3]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[K]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[INDVARS_IV]]
; CHECK-NEXT: store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
; IND-LABEL: @veciv(
; IND-NEXT: for.body.preheader:
; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K:%.*]], 2
; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: [[N_VEC:%.*]] = and i32 [[K]], -2
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[TMP0:%.*]] = sext i32 [[INDEX]] to i64
; IND-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP0]]
; IND-NEXT: [[TMP2:%.*]] = bitcast i32* [[TMP1]] to <2 x i32>*
; IND-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP2]], align 4
; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; IND-NEXT: [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; IND-NEXT: br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[N_VEC]], [[K]]
; IND-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
; IND-NEXT: br label [[FOR_BODY:%.*]]
; IND: for.body:
; IND-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; IND-NEXT: [[TMP4:%.*]] = sext i32 [[INDVARS_IV]] to i64
; IND-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP4]]
; IND-NEXT: store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
; IND-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
; IND-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
; IND-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
; IND: exit:
; IND-NEXT: ret void
;
; UNROLL-LABEL: @veciv(
; UNROLL-NEXT: for.body.preheader:
; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K:%.*]], 4
; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: [[N_VEC:%.*]] = and i32 [[K]], -4
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; UNROLL-NEXT: [[TMP0:%.*]] = sext i32 [[INDEX]] to i64
; UNROLL-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP0]]
; UNROLL-NEXT: [[TMP2:%.*]] = bitcast i32* [[TMP1]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP2]], align 4
; UNROLL-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 2
; UNROLL-NEXT: [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP4]], align 4
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 4, i32 4>
; UNROLL-NEXT: [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[N_VEC]], [[K]]
; UNROLL-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
; UNROLL-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL: for.body:
; UNROLL-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[TMP6:%.*]] = sext i32 [[INDVARS_IV]] to i64
; UNROLL-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP6]]
; UNROLL-NEXT: store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
; UNROLL-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
; UNROLL-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
; UNROLL-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
; UNROLL: exit:
; UNROLL-NEXT: ret void
;
; UNROLL-NO-IC-LABEL: @veciv(
; UNROLL-NO-IC-NEXT: for.body.preheader:
; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K:%.*]], 4
; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[K]], 4
; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i32 [[K]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 0
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = add i32 [[INDEX]], 2
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP0]]
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[TMP1]]
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = bitcast i32* [[TMP4]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP5]], align 4
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 2
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = bitcast i32* [[TMP6]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP7]], align 4
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[K]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC: for.body:
; UNROLL-NO-IC-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[INDVARS_IV]]
; UNROLL-NO-IC-NEXT: store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
; UNROLL-NO-IC-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
; UNROLL-NO-IC-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
; UNROLL-NO-IC-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
; UNROLL-NO-IC: exit:
; UNROLL-NO-IC-NEXT: ret void
;
; INTERLEAVE-LABEL: @veciv(
; INTERLEAVE-NEXT: for.body.preheader:
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[K:%.*]], 8
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i32 [[K]], -8
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; INTERLEAVE-NEXT: [[TMP0:%.*]] = sext i32 [[INDEX]] to i64
; INTERLEAVE-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP0]]
; INTERLEAVE-NEXT: [[TMP2:%.*]] = bitcast i32* [[TMP1]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[VEC_IND]], <4 x i32>* [[TMP2]], align 4
; INTERLEAVE-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i64 4
; INTERLEAVE-NEXT: [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[STEP_ADD]], <4 x i32>* [[TMP4]], align 4
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 8, i32 8, i32 8, i32 8>
; INTERLEAVE-NEXT: [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[N_VEC]], [[K]]
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]
; INTERLEAVE: for.body:
; INTERLEAVE-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[TMP6:%.*]] = sext i32 [[INDVARS_IV]] to i64
; INTERLEAVE-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP6]]
; INTERLEAVE-NEXT: store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
; INTERLEAVE-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
; INTERLEAVE: exit:
; INTERLEAVE-NEXT: ret void
;
for.body.preheader:
br label %for.body
for.body:
%indvars.iv = phi i32 [ %indvars.iv.next, %for.body ], [ 0, %for.body.preheader ]
%arrayidx = getelementptr inbounds i32, i32* %a, i32 %indvars.iv
store i32 %indvars.iv, i32* %arrayidx, align 4
%indvars.iv.next = add nuw nsw i32 %indvars.iv, 1
%exitcond = icmp eq i32 %indvars.iv.next, %k
br i1 %exitcond, label %exit, label %for.body
exit:
ret void
}
define void @trunciv(i32* nocapture %a, i32 %start, i64 %k) {
; CHECK-LABEL: @trunciv(
; CHECK-NEXT: for.body.preheader:
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[K:%.*]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; CHECK: vector.scevcheck:
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[K]], -1
; CHECK-NEXT: [[TMP1:%.*]] = trunc i64 [[TMP0]] to i32
; CHECK-NEXT: [[TMP2:%.*]] = icmp slt i32 [[TMP1]], 0
; CHECK-NEXT: [[TMP3:%.*]] = icmp ugt i64 [[TMP0]], 4294967295
; CHECK-NEXT: [[TMP4:%.*]] = or i1 [[TMP2]], [[TMP3]]
; CHECK-NEXT: br i1 [[TMP4]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[K]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[K]], [[N_MOD_VF]]
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP5:%.*]] = trunc i64 [[INDEX]] to i32
; CHECK-NEXT: [[TMP6:%.*]] = add i32 [[TMP5]], 0
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP6]]
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[TMP7]], i32 0
; CHECK-NEXT: [[TMP9:%.*]] = bitcast i32* [[TMP8]] to <2 x i32>*
; CHECK-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP9]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; CHECK-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[K]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[TRUNC_IV:%.*]] = trunc i64 [[INDVARS_IV]] to i32
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[TRUNC_IV]]
; CHECK-NEXT: store i32 [[TRUNC_IV]], i32* [[ARRAYIDX]], align 4
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[K]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP43:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
; IND-LABEL: @trunciv(
; IND-NEXT: for.body.preheader:
; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[K:%.*]], 2
; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; IND: vector.scevcheck:
; IND-NEXT: [[TMP0:%.*]] = add i64 [[K]], -1
; IND-NEXT: [[DOTNOT:%.*]] = icmp ult i64 [[TMP0]], 2147483648
; IND-NEXT: br i1 [[DOTNOT]], label [[VECTOR_PH:%.*]], label [[SCALAR_PH]]
; IND: vector.ph:
; IND-NEXT: [[N_VEC:%.*]] = and i64 [[K]], -2
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[SEXT:%.*]] = shl i64 [[INDEX]], 32
; IND-NEXT: [[TMP1:%.*]] = ashr exact i64 [[SEXT]], 32
; IND-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP1]]
; IND-NEXT: [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <2 x i32>*
; IND-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP3]], align 4
; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; IND-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; IND-NEXT: br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[K]]
; IND-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; IND-NEXT: br label [[FOR_BODY:%.*]]
; IND: for.body:
; IND-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; IND-NEXT: [[TRUNC_IV:%.*]] = trunc i64 [[INDVARS_IV]] to i32
; IND-NEXT: [[SEXT1:%.*]] = shl i64 [[INDVARS_IV]], 32
; IND-NEXT: [[TMP5:%.*]] = ashr exact i64 [[SEXT1]], 32
; IND-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP5]]
; IND-NEXT: store i32 [[TRUNC_IV]], i32* [[ARRAYIDX]], align 4
; IND-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; IND-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[K]]
; IND-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP43:![0-9]+]]
; IND: exit:
; IND-NEXT: ret void
;
; UNROLL-LABEL: @trunciv(
; UNROLL-NEXT: for.body.preheader:
; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[K:%.*]], 4
; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; UNROLL: vector.scevcheck:
; UNROLL-NEXT: [[TMP0:%.*]] = add i64 [[K]], -1
; UNROLL-NEXT: [[DOTNOT:%.*]] = icmp ult i64 [[TMP0]], 2147483648
; UNROLL-NEXT: br i1 [[DOTNOT]], label [[VECTOR_PH:%.*]], label [[SCALAR_PH]]
; UNROLL: vector.ph:
; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[K]], -4
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; UNROLL-NEXT: [[SEXT:%.*]] = shl i64 [[INDEX]], 32
; UNROLL-NEXT: [[TMP1:%.*]] = ashr exact i64 [[SEXT]], 32
; UNROLL-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP1]]
; UNROLL-NEXT: [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP3]], align 4
; UNROLL-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 2
; UNROLL-NEXT: [[TMP5:%.*]] = bitcast i32* [[TMP4]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP5]], align 4
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 4, i32 4>
; UNROLL-NEXT: [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[K]]
; UNROLL-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; UNROLL-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL: for.body:
; UNROLL-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[TRUNC_IV:%.*]] = trunc i64 [[INDVARS_IV]] to i32
; UNROLL-NEXT: [[SEXT2:%.*]] = shl i64 [[INDVARS_IV]], 32
; UNROLL-NEXT: [[TMP7:%.*]] = ashr exact i64 [[SEXT2]], 32
; UNROLL-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP7]]
; UNROLL-NEXT: store i32 [[TRUNC_IV]], i32* [[ARRAYIDX]], align 4
; UNROLL-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; UNROLL-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[K]]
; UNROLL-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP43:![0-9]+]]
; UNROLL: exit:
; UNROLL-NEXT: ret void
;
; UNROLL-NO-IC-LABEL: @trunciv(
; UNROLL-NO-IC-NEXT: for.body.preheader:
; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[K:%.*]], 4
; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; UNROLL-NO-IC: vector.scevcheck:
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i64 [[K]], -1
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = trunc i64 [[TMP0]] to i32
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = icmp slt i32 [[TMP1]], 0
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = icmp ugt i64 [[TMP0]], 4294967295
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = or i1 [[TMP2]], [[TMP3]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP4]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[K]], 4
; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[K]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = trunc i64 [[INDEX]] to i32
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = add i32 [[TMP5]], 0
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = add i32 [[TMP5]], 2
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP6]]
; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[TMP7]]
; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[TMP8]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = bitcast i32* [[TMP10]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP11]], align 4
; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[TMP8]], i32 2
; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = bitcast i32* [[TMP12]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP13]], align 4
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[K]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC: for.body:
; UNROLL-NO-IC-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[TRUNC_IV:%.*]] = trunc i64 [[INDVARS_IV]] to i32
; UNROLL-NO-IC-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[TRUNC_IV]]
; UNROLL-NO-IC-NEXT: store i32 [[TRUNC_IV]], i32* [[ARRAYIDX]], align 4
; UNROLL-NO-IC-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; UNROLL-NO-IC-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[K]]
; UNROLL-NO-IC-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP43:![0-9]+]]
; UNROLL-NO-IC: exit:
; UNROLL-NO-IC-NEXT: ret void
;
; INTERLEAVE-LABEL: @trunciv(
; INTERLEAVE-NEXT: for.body.preheader:
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[K:%.*]], 8
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; INTERLEAVE: vector.scevcheck:
; INTERLEAVE-NEXT: [[TMP0:%.*]] = add i64 [[K]], -1
; INTERLEAVE-NEXT: [[DOTNOT:%.*]] = icmp ult i64 [[TMP0]], 2147483648
; INTERLEAVE-NEXT: br i1 [[DOTNOT]], label [[VECTOR_PH:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i64 [[K]], -8
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; INTERLEAVE-NEXT: [[SEXT:%.*]] = shl i64 [[INDEX]], 32
; INTERLEAVE-NEXT: [[TMP1:%.*]] = ashr exact i64 [[SEXT]], 32
; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP1]]
; INTERLEAVE-NEXT: [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[VEC_IND]], <4 x i32>* [[TMP3]], align 4
; INTERLEAVE-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 4
; INTERLEAVE-NEXT: [[TMP5:%.*]] = bitcast i32* [[TMP4]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[STEP_ADD]], <4 x i32>* [[TMP5]], align 4
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 8, i32 8, i32 8, i32 8>
; INTERLEAVE-NEXT: [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[K]]
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]
; INTERLEAVE: for.body:
; INTERLEAVE-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[TRUNC_IV:%.*]] = trunc i64 [[INDVARS_IV]] to i32
; INTERLEAVE-NEXT: [[SEXT2:%.*]] = shl i64 [[INDVARS_IV]], 32
; INTERLEAVE-NEXT: [[TMP7:%.*]] = ashr exact i64 [[SEXT2]], 32
; INTERLEAVE-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP7]]
; INTERLEAVE-NEXT: store i32 [[TRUNC_IV]], i32* [[ARRAYIDX]], align 4
; INTERLEAVE-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[K]]
; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP43:![0-9]+]]
; INTERLEAVE: exit:
; INTERLEAVE-NEXT: ret void
;
for.body.preheader:
br label %for.body
for.body:
%indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %for.body.preheader ]
%trunc.iv = trunc i64 %indvars.iv to i32
%arrayidx = getelementptr inbounds i32, i32* %a, i32 %trunc.iv
store i32 %trunc.iv, i32* %arrayidx, align 4
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
%exitcond = icmp eq i64 %indvars.iv.next, %k
br i1 %exitcond, label %exit, label %for.body
exit:
ret void
}
;
;
define void @nonprimary(i32* nocapture %a, i32 %start, i32 %i, i32 %k) {
; CHECK-LABEL: @nonprimary(
; CHECK-NEXT: for.body.preheader:
; CHECK-NEXT: [[TMP0:%.*]] = sub i32 [[K:%.*]], [[I:%.*]]
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]
; CHECK-NEXT: [[IND_END:%.*]] = add i32 [[I]], [[N_VEC]]
; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[I]], i32 0
; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i32 [[I]], [[INDEX]]
; CHECK-NEXT: [[TMP1:%.*]] = add i32 [[OFFSET_IDX]], 0
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP1]]
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 0
; CHECK-NEXT: [[TMP4:%.*]] = bitcast i32* [[TMP3]] to <2 x i32>*
; CHECK-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP4]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[I]], [[FOR_BODY_PREHEADER:%.*]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[INDVARS_IV]]
; CHECK-NEXT: store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
; IND-LABEL: @nonprimary(
; IND-NEXT: for.body.preheader:
; IND-NEXT: [[TMP0:%.*]] = sub i32 [[K:%.*]], [[I:%.*]]
; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 2
; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], -2
; IND-NEXT: [[IND_END:%.*]] = add i32 [[N_VEC]], [[I]]
; IND-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[I]], i64 0
; IND-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; IND-NEXT: [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[OFFSET_IDX:%.*]] = add i32 [[INDEX]], [[I]]
; IND-NEXT: [[TMP1:%.*]] = sext i32 [[OFFSET_IDX]] to i64
; IND-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP1]]
; IND-NEXT: [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <2 x i32>*
; IND-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP3]], align 4
; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; IND-NEXT: [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; IND-NEXT: br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
; IND-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[I]], [[FOR_BODY_PREHEADER:%.*]] ]
; IND-NEXT: br label [[FOR_BODY:%.*]]
; IND: for.body:
; IND-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; IND-NEXT: [[TMP5:%.*]] = sext i32 [[INDVARS_IV]] to i64
; IND-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP5]]
; IND-NEXT: store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
; IND-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
; IND-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
; IND-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]
; IND: exit:
; IND-NEXT: ret void
;
; UNROLL-LABEL: @nonprimary(
; UNROLL-NEXT: for.body.preheader:
; UNROLL-NEXT: [[TMP0:%.*]] = sub i32 [[K:%.*]], [[I:%.*]]
; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 4
; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], -4
; UNROLL-NEXT: [[IND_END:%.*]] = add i32 [[N_VEC]], [[I]]
; UNROLL-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[I]], i64 0
; UNROLL-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NEXT: [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; UNROLL-NEXT: [[OFFSET_IDX:%.*]] = add i32 [[INDEX]], [[I]]
; UNROLL-NEXT: [[TMP1:%.*]] = sext i32 [[OFFSET_IDX]] to i64
; UNROLL-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP1]]
; UNROLL-NEXT: [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP3]], align 4
; UNROLL-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 2
; UNROLL-NEXT: [[TMP5:%.*]] = bitcast i32* [[TMP4]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP5]], align 4
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 4, i32 4>
; UNROLL-NEXT: [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[I]], [[FOR_BODY_PREHEADER:%.*]] ]
; UNROLL-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL: for.body:
; UNROLL-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[TMP7:%.*]] = sext i32 [[INDVARS_IV]] to i64
; UNROLL-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP7]]
; UNROLL-NEXT: store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
; UNROLL-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
; UNROLL-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
; UNROLL-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]
; UNROLL: exit:
; UNROLL-NEXT: ret void
;
; UNROLL-NO-IC-LABEL: @nonprimary(
; UNROLL-NO-IC-NEXT: for.body.preheader:
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = sub i32 [[K:%.*]], [[I:%.*]]
; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 4
; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[TMP0]], 4
; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i32 [[TMP0]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT: [[IND_END:%.*]] = add i32 [[I]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[I]], i32 0
; UNROLL-NO-IC-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT: [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[OFFSET_IDX:%.*]] = add i32 [[I]], [[INDEX]]
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = add i32 [[OFFSET_IDX]], 0
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add i32 [[OFFSET_IDX]], 2
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP1]]
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[TMP2]]
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP3]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = bitcast i32* [[TMP5]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP6]], align 4
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP3]], i32 2
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = bitcast i32* [[TMP7]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP8]], align 4
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[I]], [[FOR_BODY_PREHEADER:%.*]] ]
; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC: for.body:
; UNROLL-NO-IC-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[INDVARS_IV]]
; UNROLL-NO-IC-NEXT: store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
; UNROLL-NO-IC-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
; UNROLL-NO-IC-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
; UNROLL-NO-IC-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]
; UNROLL-NO-IC: exit:
; UNROLL-NO-IC-NEXT: ret void
;
; INTERLEAVE-LABEL: @nonprimary(
; INTERLEAVE-NEXT: for.body.preheader:
; INTERLEAVE-NEXT: [[TMP0:%.*]] = sub i32 [[K:%.*]], [[I:%.*]]
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP0]], 8
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i32 [[TMP0]], -8
; INTERLEAVE-NEXT: [[IND_END:%.*]] = add i32 [[N_VEC]], [[I]]
; INTERLEAVE-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[I]], i64 0
; INTERLEAVE-NEXT: [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; INTERLEAVE-NEXT: [[INDUCTION:%.*]] = add <4 x i32> [[DOTSPLAT]], <i32 0, i32 1, i32 2, i32 3>
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; INTERLEAVE-NEXT: [[OFFSET_IDX:%.*]] = add i32 [[INDEX]], [[I]]
; INTERLEAVE-NEXT: [[TMP1:%.*]] = sext i32 [[OFFSET_IDX]] to i64
; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP1]]
; INTERLEAVE-NEXT: [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[VEC_IND]], <4 x i32>* [[TMP3]], align 4
; INTERLEAVE-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 4
; INTERLEAVE-NEXT: [[TMP5:%.*]] = bitcast i32* [[TMP4]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[STEP_ADD]], <4 x i32>* [[TMP5]], align 4
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 8, i32 8, i32 8, i32 8>
; INTERLEAVE-NEXT: [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[TMP0]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[I]], [[FOR_BODY_PREHEADER:%.*]] ]
; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]
; INTERLEAVE: for.body:
; INTERLEAVE-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[TMP7:%.*]] = sext i32 [[INDVARS_IV]] to i64
; INTERLEAVE-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP7]]
; INTERLEAVE-NEXT: store i32 [[INDVARS_IV]], i32* [[ARRAYIDX]], align 4
; INTERLEAVE-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[K]]
; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]
; INTERLEAVE: exit:
; INTERLEAVE-NEXT: ret void
;
for.body.preheader:
br label %for.body
for.body:
%indvars.iv = phi i32 [ %indvars.iv.next, %for.body ], [ %i, %for.body.preheader ]
%arrayidx = getelementptr inbounds i32, i32* %a, i32 %indvars.iv
store i32 %indvars.iv, i32* %arrayidx, align 4
%indvars.iv.next = add nuw nsw i32 %indvars.iv, 1
%exitcond = icmp eq i32 %indvars.iv.next, %k
br i1 %exitcond, label %exit, label %for.body
exit:
ret void
}
define void @non_primary_iv_trunc(i32* %a, i64 %n) {
; CHECK-LABEL: @non_primary_iv_trunc(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]
; CHECK-NEXT: [[IND_END:%.*]] = mul i64 [[N_VEC]], 2
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP0]]
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[TMP1]], i32 0
; CHECK-NEXT: [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <2 x i32>*
; CHECK-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP3]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 4, i32 4>
; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP46:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; CHECK-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[J:%.*]] = phi i64 [ [[J_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[I]]
; CHECK-NEXT: [[VAR1:%.*]] = trunc i64 [[J]] to i32
; CHECK-NEXT: store i32 [[VAR1]], i32* [[VAR0]], align 4
; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; CHECK-NEXT: [[J_NEXT]] = add nuw nsw i64 [[J]], 2
; CHECK-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP47:![0-9]+]]
; CHECK: for.end:
; CHECK-NEXT: ret void
;
; IND-LABEL: @non_primary_iv_trunc(
; IND-NEXT: entry:
; IND-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 2
; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775806
; IND-NEXT: [[IND_END:%.*]] = shl nuw i64 [[N_VEC]], 1
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[TMP0:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDEX]]
; IND-NEXT: [[TMP1:%.*]] = bitcast i32* [[TMP0]] to <2 x i32>*
; IND-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP1]], align 4
; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 4, i32 4>
; IND-NEXT: [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; IND-NEXT: br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP46:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
; IND-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; IND-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; IND-NEXT: br label [[FOR_BODY:%.*]]
; IND: for.body:
; IND-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; IND-NEXT: [[J:%.*]] = phi i64 [ [[J_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; IND-NEXT: [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[I]]
; IND-NEXT: [[VAR1:%.*]] = trunc i64 [[J]] to i32
; IND-NEXT: store i32 [[VAR1]], i32* [[VAR0]], align 4
; IND-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; IND-NEXT: [[J_NEXT]] = add nuw nsw i64 [[J]], 2
; IND-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; IND-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP47:![0-9]+]]
; IND: for.end:
; IND-NEXT: ret void
;
; UNROLL-LABEL: @non_primary_iv_trunc(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 4
; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775804
; UNROLL-NEXT: [[IND_END:%.*]] = shl nuw i64 [[N_VEC]], 1
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 4, i32 4>
; UNROLL-NEXT: [[TMP0:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDEX]]
; UNROLL-NEXT: [[TMP1:%.*]] = bitcast i32* [[TMP0]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP1]], align 4
; UNROLL-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 2
; UNROLL-NEXT: [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP3]], align 4
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 8, i32 8>
; UNROLL-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP46:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; UNROLL-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL: for.body:
; UNROLL-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[J:%.*]] = phi i64 [ [[J_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[I]]
; UNROLL-NEXT: [[VAR1:%.*]] = trunc i64 [[J]] to i32
; UNROLL-NEXT: store i32 [[VAR1]], i32* [[VAR0]], align 4
; UNROLL-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; UNROLL-NEXT: [[J_NEXT]] = add nuw nsw i64 [[J]], 2
; UNROLL-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; UNROLL-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP47:![0-9]+]]
; UNROLL: for.end:
; UNROLL-NEXT: ret void
;
; UNROLL-NO-IC-LABEL: @non_primary_iv_trunc(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 4
; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SMAX]], 4
; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[SMAX]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT: [[IND_END:%.*]] = mul i64 [[N_VEC]], 2
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 4, i32 4>
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 2
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP0]]
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP1]]
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = bitcast i32* [[TMP4]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[VEC_IND]], <2 x i32>* [[TMP5]], align 4
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 2
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = bitcast i32* [[TMP6]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[STEP_ADD]], <2 x i32>* [[TMP7]], align 4
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], <i32 4, i32 4>
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP46:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC: for.body:
; UNROLL-NO-IC-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[J:%.*]] = phi i64 [ [[J_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[I]]
; UNROLL-NO-IC-NEXT: [[VAR1:%.*]] = trunc i64 [[J]] to i32
; UNROLL-NO-IC-NEXT: store i32 [[VAR1]], i32* [[VAR0]], align 4
; UNROLL-NO-IC-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; UNROLL-NO-IC-NEXT: [[J_NEXT]] = add nuw nsw i64 [[J]], 2
; UNROLL-NO-IC-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; UNROLL-NO-IC-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP47:![0-9]+]]
; UNROLL-NO-IC: for.end:
; UNROLL-NO-IC-NEXT: ret void
;
; INTERLEAVE-LABEL: @non_primary_iv_trunc(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: [[SMAX:%.*]] = call i64 @llvm.smax.i64(i64 [[N:%.*]], i64 1)
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SMAX]], 8
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i64 [[SMAX]], 9223372036854775800
; INTERLEAVE-NEXT: [[IND_END:%.*]] = shl nuw i64 [[N_VEC]], 1
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 2, i32 4, i32 6>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], <i32 8, i32 8, i32 8, i32 8>
; INTERLEAVE-NEXT: [[TMP0:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDEX]]
; INTERLEAVE-NEXT: [[TMP1:%.*]] = bitcast i32* [[TMP0]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[VEC_IND]], <4 x i32>* [[TMP1]], align 4
; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, i32* [[TMP0]], i64 4
; INTERLEAVE-NEXT: [[TMP3:%.*]] = bitcast i32* [[TMP2]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[STEP_ADD]], <4 x i32>* [[TMP3]], align 4
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 16, i32 16, i32 16, i32 16>
; INTERLEAVE-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP46:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SMAX]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; INTERLEAVE-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]
; INTERLEAVE: for.body:
; INTERLEAVE-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[J:%.*]] = phi i64 [ [[J_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[VAR0:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[I]]
; INTERLEAVE-NEXT: [[VAR1:%.*]] = trunc i64 [[J]] to i32
; INTERLEAVE-NEXT: store i32 [[VAR1]], i32* [[VAR0]], align 4
; INTERLEAVE-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; INTERLEAVE-NEXT: [[J_NEXT]] = add nuw nsw i64 [[J]], 2
; INTERLEAVE-NEXT: [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
; INTERLEAVE-NEXT: br i1 [[COND]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP47:![0-9]+]]
; INTERLEAVE: for.end:
; INTERLEAVE-NEXT: ret void
;
entry:
br label %for.body
for.body:
%i = phi i64 [ %i.next, %for.body ], [ 0, %entry ]
%j = phi i64 [ %j.next, %for.body ], [ 0, %entry ]
%var0 = getelementptr inbounds i32, i32* %a, i64 %i
%var1 = trunc i64 %j to i32
store i32 %var1, i32* %var0, align 4
%i.next = add nuw nsw i64 %i, 1
%j.next = add nuw nsw i64 %j, 2
%cond = icmp slt i64 %i.next, %n
br i1 %cond, label %for.body, label %for.end
for.end:
ret void
}
; PR32419. Ensure we transform truncated non-primary induction variables. In
; the test case below we replace %var1 with a new induction variable. Because
; the truncated value is non-primary, we must compute an offset from the
; primary induction variable.
;
;
define i32 @PR32419(i32 %a, i16 %b) {
; CHECK-LABEL: @PR32419(
; CHECK-NEXT: entry:
; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[TMP0:%.*]] = insertelement <2 x i32> zeroinitializer, i32 [[A:%.*]], i32 0
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UREM_CONTINUE2:%.*]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[PRED_UREM_CONTINUE2]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i16> [ <i16 -20, i16 -19>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_UREM_CONTINUE2]] ]
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i32 -20, [[INDEX]]
; CHECK-NEXT: [[TMP1:%.*]] = trunc i32 [[OFFSET_IDX]] to i16
; CHECK-NEXT: [[TMP2:%.*]] = icmp eq <2 x i16> [[VEC_IND]], zeroinitializer
; CHECK-NEXT: [[TMP3:%.*]] = xor <2 x i1> [[TMP2]], <i1 true, i1 true>
; CHECK-NEXT: [[TMP4:%.*]] = extractelement <2 x i1> [[TMP3]], i32 0
; CHECK-NEXT: br i1 [[TMP4]], label [[PRED_UREM_IF:%.*]], label [[PRED_UREM_CONTINUE:%.*]]
; CHECK: pred.urem.if:
; CHECK-NEXT: [[TMP5:%.*]] = add i16 [[TMP1]], 0
; CHECK-NEXT: [[TMP6:%.*]] = urem i16 [[B:%.*]], [[TMP5]]
; CHECK-NEXT: [[TMP7:%.*]] = insertelement <2 x i16> poison, i16 [[TMP6]], i32 0
; CHECK-NEXT: br label [[PRED_UREM_CONTINUE]]
; CHECK: pred.urem.continue:
; CHECK-NEXT: [[TMP8:%.*]] = phi <2 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP7]], [[PRED_UREM_IF]] ]
; CHECK-NEXT: [[TMP9:%.*]] = extractelement <2 x i1> [[TMP3]], i32 1
; CHECK-NEXT: br i1 [[TMP9]], label [[PRED_UREM_IF1:%.*]], label [[PRED_UREM_CONTINUE2]]
; CHECK: pred.urem.if1:
; CHECK-NEXT: [[TMP10:%.*]] = add i16 [[TMP1]], 1
; CHECK-NEXT: [[TMP11:%.*]] = urem i16 [[B]], [[TMP10]]
; CHECK-NEXT: [[TMP12:%.*]] = insertelement <2 x i16> [[TMP8]], i16 [[TMP11]], i32 1
; CHECK-NEXT: br label [[PRED_UREM_CONTINUE2]]
; CHECK: pred.urem.continue2:
; CHECK-NEXT: [[TMP13:%.*]] = phi <2 x i16> [ [[TMP8]], [[PRED_UREM_CONTINUE]] ], [ [[TMP12]], [[PRED_UREM_IF1]] ]
; CHECK-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[TMP2]], <2 x i16> zeroinitializer, <2 x i16> [[TMP13]]
; CHECK-NEXT: [[TMP14:%.*]] = sext <2 x i16> [[PREDPHI]] to <2 x i32>
; CHECK-NEXT: [[TMP15]] = or <2 x i32> [[VEC_PHI]], [[TMP14]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i16> [[VEC_IND]], <i16 2, i16 2>
; CHECK-NEXT: [[TMP16:%.*]] = icmp eq i32 [[INDEX_NEXT]], 20
; CHECK-NEXT: br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP17:%.*]] = call i32 @llvm.vector.reduce.or.v2i32(<2 x i32> [[TMP15]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 20, 20
; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[MIDDLE_BLOCK]] ], [ -20, [[ENTRY:%.*]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[A]], [[ENTRY]] ], [ [[TMP17]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[FOR_BODY:%.*]]
; CHECK: for.body:
; CHECK-NEXT: [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_INC:%.*]] ]
; CHECK-NEXT: [[VAR0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR6:%.*]], [[FOR_INC]] ]
; CHECK-NEXT: [[VAR1:%.*]] = trunc i32 [[I]] to i16
; CHECK-NEXT: [[VAR2:%.*]] = icmp eq i16 [[VAR1]], 0
; CHECK-NEXT: br i1 [[VAR2]], label [[FOR_INC]], label [[FOR_COND:%.*]]
; CHECK: for.cond:
; CHECK-NEXT: [[VAR3:%.*]] = urem i16 [[B]], [[VAR1]]
; CHECK-NEXT: br label [[FOR_INC]]
; CHECK: for.inc:
; CHECK-NEXT: [[VAR4:%.*]] = phi i16 [ [[VAR3]], [[FOR_COND]] ], [ 0, [[FOR_BODY]] ]
; CHECK-NEXT: [[VAR5:%.*]] = sext i16 [[VAR4]] to i32
; CHECK-NEXT: [[VAR6]] = or i32 [[VAR0]], [[VAR5]]
; CHECK-NEXT: [[I_NEXT]] = add nsw i32 [[I]], 1
; CHECK-NEXT: [[COND:%.*]] = icmp eq i32 [[I_NEXT]], 0
; CHECK-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP49:![0-9]+]]
; CHECK: for.end:
; CHECK-NEXT: [[VAR7:%.*]] = phi i32 [ [[VAR6]], [[FOR_INC]] ], [ [[TMP17]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i32 [[VAR7]]
;
; IND-LABEL: @PR32419(
; IND-NEXT: entry:
; IND-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: [[TMP0:%.*]] = insertelement <2 x i32> <i32 poison, i32 0>, i32 [[A:%.*]], i64 0
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UREM_CONTINUE2:%.*]] ]
; IND-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[PRED_UREM_CONTINUE2]] ]
; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i16> [ <i16 -20, i16 -19>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_UREM_CONTINUE2]] ]
; IND-NEXT: [[TMP1:%.*]] = trunc i32 [[INDEX]] to i16
; IND-NEXT: [[TMP2:%.*]] = icmp eq <2 x i16> [[VEC_IND]], zeroinitializer
; IND-NEXT: [[TMP3:%.*]] = xor <2 x i1> [[TMP2]], <i1 true, i1 true>
; IND-NEXT: [[TMP4:%.*]] = extractelement <2 x i1> [[TMP3]], i64 0
; IND-NEXT: br i1 [[TMP4]], label [[PRED_UREM_IF:%.*]], label [[PRED_UREM_CONTINUE:%.*]]
; IND: pred.urem.if:
; IND-NEXT: [[TMP5:%.*]] = add i16 [[TMP1]], -20
; IND-NEXT: [[TMP6:%.*]] = urem i16 [[B:%.*]], [[TMP5]]
; IND-NEXT: [[TMP7:%.*]] = insertelement <2 x i16> poison, i16 [[TMP6]], i64 0
; IND-NEXT: br label [[PRED_UREM_CONTINUE]]
; IND: pred.urem.continue:
; IND-NEXT: [[TMP8:%.*]] = phi <2 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP7]], [[PRED_UREM_IF]] ]
; IND-NEXT: [[TMP9:%.*]] = extractelement <2 x i1> [[TMP3]], i64 1
; IND-NEXT: br i1 [[TMP9]], label [[PRED_UREM_IF1:%.*]], label [[PRED_UREM_CONTINUE2]]
; IND: pred.urem.if1:
; IND-NEXT: [[TMP10:%.*]] = add i16 [[TMP1]], -19
; IND-NEXT: [[TMP11:%.*]] = urem i16 [[B]], [[TMP10]]
; IND-NEXT: [[TMP12:%.*]] = insertelement <2 x i16> [[TMP8]], i16 [[TMP11]], i64 1
; IND-NEXT: br label [[PRED_UREM_CONTINUE2]]
; IND: pred.urem.continue2:
; IND-NEXT: [[TMP13:%.*]] = phi <2 x i16> [ [[TMP8]], [[PRED_UREM_CONTINUE]] ], [ [[TMP12]], [[PRED_UREM_IF1]] ]
; IND-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[TMP2]], <2 x i16> zeroinitializer, <2 x i16> [[TMP13]]
; IND-NEXT: [[TMP14:%.*]] = sext <2 x i16> [[PREDPHI]] to <2 x i32>
; IND-NEXT: [[TMP15]] = or <2 x i32> [[VEC_PHI]], [[TMP14]]
; IND-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i16> [[VEC_IND]], <i16 2, i16 2>
; IND-NEXT: [[TMP16:%.*]] = icmp eq i32 [[INDEX_NEXT]], 20
; IND-NEXT: br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[TMP17:%.*]] = call i32 @llvm.vector.reduce.or.v2i32(<2 x i32> [[TMP15]])
; IND-NEXT: br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: br label [[FOR_BODY:%.*]]
; IND: for.body:
; IND-NEXT: br i1 poison, label [[FOR_INC:%.*]], label [[FOR_COND:%.*]]
; IND: for.cond:
; IND-NEXT: br label [[FOR_INC]]
; IND: for.inc:
; IND-NEXT: br i1 poison, label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP49:![0-9]+]]
; IND: for.end:
; IND-NEXT: [[VAR7:%.*]] = phi i32 [ poison, [[FOR_INC]] ], [ [[TMP17]], [[MIDDLE_BLOCK]] ]
; IND-NEXT: ret i32 [[VAR7]]
;
; UNROLL-LABEL: @PR32419(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: [[TMP0:%.*]] = insertelement <2 x i32> <i32 poison, i32 0>, i32 [[A:%.*]], i64 0
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UREM_CONTINUE8:%.*]] ]
; UNROLL-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP28:%.*]], [[PRED_UREM_CONTINUE8]] ]
; UNROLL-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP29:%.*]], [[PRED_UREM_CONTINUE8]] ]
; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i16> [ <i16 -20, i16 -19>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_UREM_CONTINUE8]] ]
; UNROLL-NEXT: [[TMP1:%.*]] = trunc i32 [[INDEX]] to i16
; UNROLL-NEXT: [[TMP2:%.*]] = icmp eq <2 x i16> [[VEC_IND]], zeroinitializer
; UNROLL-NEXT: [[TMP3:%.*]] = icmp eq <2 x i16> [[VEC_IND]], <i16 -2, i16 -2>
; UNROLL-NEXT: [[TMP4:%.*]] = xor <2 x i1> [[TMP2]], <i1 true, i1 true>
; UNROLL-NEXT: [[TMP5:%.*]] = xor <2 x i1> [[TMP3]], <i1 true, i1 true>
; UNROLL-NEXT: [[TMP6:%.*]] = extractelement <2 x i1> [[TMP4]], i64 0
; UNROLL-NEXT: br i1 [[TMP6]], label [[PRED_UREM_IF:%.*]], label [[PRED_UREM_CONTINUE:%.*]]
; UNROLL: pred.urem.if:
; UNROLL-NEXT: [[TMP7:%.*]] = add i16 [[TMP1]], -20
; UNROLL-NEXT: [[TMP8:%.*]] = urem i16 [[B:%.*]], [[TMP7]]
; UNROLL-NEXT: [[TMP9:%.*]] = insertelement <2 x i16> poison, i16 [[TMP8]], i64 0
; UNROLL-NEXT: br label [[PRED_UREM_CONTINUE]]
; UNROLL: pred.urem.continue:
; UNROLL-NEXT: [[TMP10:%.*]] = phi <2 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP9]], [[PRED_UREM_IF]] ]
; UNROLL-NEXT: [[TMP11:%.*]] = extractelement <2 x i1> [[TMP4]], i64 1
; UNROLL-NEXT: br i1 [[TMP11]], label [[PRED_UREM_IF3:%.*]], label [[PRED_UREM_CONTINUE4:%.*]]
; UNROLL: pred.urem.if3:
; UNROLL-NEXT: [[TMP12:%.*]] = add i16 [[TMP1]], -19
; UNROLL-NEXT: [[TMP13:%.*]] = urem i16 [[B]], [[TMP12]]
; UNROLL-NEXT: [[TMP14:%.*]] = insertelement <2 x i16> [[TMP10]], i16 [[TMP13]], i64 1
; UNROLL-NEXT: br label [[PRED_UREM_CONTINUE4]]
; UNROLL: pred.urem.continue4:
; UNROLL-NEXT: [[TMP15:%.*]] = phi <2 x i16> [ [[TMP10]], [[PRED_UREM_CONTINUE]] ], [ [[TMP14]], [[PRED_UREM_IF3]] ]
; UNROLL-NEXT: [[TMP16:%.*]] = extractelement <2 x i1> [[TMP5]], i64 0
; UNROLL-NEXT: br i1 [[TMP16]], label [[PRED_UREM_IF5:%.*]], label [[PRED_UREM_CONTINUE6:%.*]]
; UNROLL: pred.urem.if5:
; UNROLL-NEXT: [[TMP17:%.*]] = add i16 [[TMP1]], -18
; UNROLL-NEXT: [[TMP18:%.*]] = urem i16 [[B]], [[TMP17]]
; UNROLL-NEXT: [[TMP19:%.*]] = insertelement <2 x i16> poison, i16 [[TMP18]], i64 0
; UNROLL-NEXT: br label [[PRED_UREM_CONTINUE6]]
; UNROLL: pred.urem.continue6:
; UNROLL-NEXT: [[TMP20:%.*]] = phi <2 x i16> [ poison, [[PRED_UREM_CONTINUE4]] ], [ [[TMP19]], [[PRED_UREM_IF5]] ]
; UNROLL-NEXT: [[TMP21:%.*]] = extractelement <2 x i1> [[TMP5]], i64 1
; UNROLL-NEXT: br i1 [[TMP21]], label [[PRED_UREM_IF7:%.*]], label [[PRED_UREM_CONTINUE8]]
; UNROLL: pred.urem.if7:
; UNROLL-NEXT: [[TMP22:%.*]] = add i16 [[TMP1]], -17
; UNROLL-NEXT: [[TMP23:%.*]] = urem i16 [[B]], [[TMP22]]
; UNROLL-NEXT: [[TMP24:%.*]] = insertelement <2 x i16> [[TMP20]], i16 [[TMP23]], i64 1
; UNROLL-NEXT: br label [[PRED_UREM_CONTINUE8]]
; UNROLL: pred.urem.continue8:
; UNROLL-NEXT: [[TMP25:%.*]] = phi <2 x i16> [ [[TMP20]], [[PRED_UREM_CONTINUE6]] ], [ [[TMP24]], [[PRED_UREM_IF7]] ]
; UNROLL-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[TMP2]], <2 x i16> zeroinitializer, <2 x i16> [[TMP15]]
; UNROLL-NEXT: [[PREDPHI9:%.*]] = select <2 x i1> [[TMP3]], <2 x i16> zeroinitializer, <2 x i16> [[TMP25]]
; UNROLL-NEXT: [[TMP26:%.*]] = sext <2 x i16> [[PREDPHI]] to <2 x i32>
; UNROLL-NEXT: [[TMP27:%.*]] = sext <2 x i16> [[PREDPHI9]] to <2 x i32>
; UNROLL-NEXT: [[TMP28]] = or <2 x i32> [[VEC_PHI]], [[TMP26]]
; UNROLL-NEXT: [[TMP29]] = or <2 x i32> [[VEC_PHI1]], [[TMP27]]
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i16> [[VEC_IND]], <i16 4, i16 4>
; UNROLL-NEXT: [[TMP30:%.*]] = icmp eq i32 [[INDEX_NEXT]], 20
; UNROLL-NEXT: br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[BIN_RDX:%.*]] = or <2 x i32> [[TMP29]], [[TMP28]]
; UNROLL-NEXT: [[TMP31:%.*]] = call i32 @llvm.vector.reduce.or.v2i32(<2 x i32> [[BIN_RDX]])
; UNROLL-NEXT: br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL: for.body:
; UNROLL-NEXT: br i1 poison, label [[FOR_INC:%.*]], label [[FOR_COND:%.*]]
; UNROLL: for.cond:
; UNROLL-NEXT: br label [[FOR_INC]]
; UNROLL: for.inc:
; UNROLL-NEXT: br i1 poison, label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP49:![0-9]+]]
; UNROLL: for.end:
; UNROLL-NEXT: [[VAR7:%.*]] = phi i32 [ poison, [[FOR_INC]] ], [ [[TMP31]], [[MIDDLE_BLOCK]] ]
; UNROLL-NEXT: ret i32 [[VAR7]]
;
; UNROLL-NO-IC-LABEL: @PR32419(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = insertelement <2 x i32> zeroinitializer, i32 [[A:%.*]], i32 0
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UREM_CONTINUE8:%.*]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP28:%.*]], [[PRED_UREM_CONTINUE8]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI1:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP29:%.*]], [[PRED_UREM_CONTINUE8]] ]
; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i16> [ <i16 -20, i16 -19>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_UREM_CONTINUE8]] ]
; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i16> [[VEC_IND]], <i16 2, i16 2>
; UNROLL-NO-IC-NEXT: [[OFFSET_IDX:%.*]] = add i32 -20, [[INDEX]]
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = trunc i32 [[OFFSET_IDX]] to i16
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = icmp eq <2 x i16> [[VEC_IND]], zeroinitializer
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = icmp eq <2 x i16> [[STEP_ADD]], zeroinitializer
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = xor <2 x i1> [[TMP2]], <i1 true, i1 true>
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = xor <2 x i1> [[TMP3]], <i1 true, i1 true>
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = extractelement <2 x i1> [[TMP4]], i32 0
; UNROLL-NO-IC-NEXT: br i1 [[TMP6]], label [[PRED_UREM_IF:%.*]], label [[PRED_UREM_CONTINUE:%.*]]
; UNROLL-NO-IC: pred.urem.if:
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = add i16 [[TMP1]], 0
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = urem i16 [[B:%.*]], [[TMP7]]
; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = insertelement <2 x i16> poison, i16 [[TMP8]], i32 0
; UNROLL-NO-IC-NEXT: br label [[PRED_UREM_CONTINUE]]
; UNROLL-NO-IC: pred.urem.continue:
; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = phi <2 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP9]], [[PRED_UREM_IF]] ]
; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = extractelement <2 x i1> [[TMP4]], i32 1
; UNROLL-NO-IC-NEXT: br i1 [[TMP11]], label [[PRED_UREM_IF3:%.*]], label [[PRED_UREM_CONTINUE4:%.*]]
; UNROLL-NO-IC: pred.urem.if3:
; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = add i16 [[TMP1]], 1
; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = urem i16 [[B]], [[TMP12]]
; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = insertelement <2 x i16> [[TMP10]], i16 [[TMP13]], i32 1
; UNROLL-NO-IC-NEXT: br label [[PRED_UREM_CONTINUE4]]
; UNROLL-NO-IC: pred.urem.continue4:
; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = phi <2 x i16> [ [[TMP10]], [[PRED_UREM_CONTINUE]] ], [ [[TMP14]], [[PRED_UREM_IF3]] ]
; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = extractelement <2 x i1> [[TMP5]], i32 0
; UNROLL-NO-IC-NEXT: br i1 [[TMP16]], label [[PRED_UREM_IF5:%.*]], label [[PRED_UREM_CONTINUE6:%.*]]
; UNROLL-NO-IC: pred.urem.if5:
; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = add i16 [[TMP1]], 2
; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = urem i16 [[B]], [[TMP17]]
; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = insertelement <2 x i16> poison, i16 [[TMP18]], i32 0
; UNROLL-NO-IC-NEXT: br label [[PRED_UREM_CONTINUE6]]
; UNROLL-NO-IC: pred.urem.continue6:
; UNROLL-NO-IC-NEXT: [[TMP20:%.*]] = phi <2 x i16> [ poison, [[PRED_UREM_CONTINUE4]] ], [ [[TMP19]], [[PRED_UREM_IF5]] ]
; UNROLL-NO-IC-NEXT: [[TMP21:%.*]] = extractelement <2 x i1> [[TMP5]], i32 1
; UNROLL-NO-IC-NEXT: br i1 [[TMP21]], label [[PRED_UREM_IF7:%.*]], label [[PRED_UREM_CONTINUE8]]
; UNROLL-NO-IC: pred.urem.if7:
; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = add i16 [[TMP1]], 3
; UNROLL-NO-IC-NEXT: [[TMP23:%.*]] = urem i16 [[B]], [[TMP22]]
; UNROLL-NO-IC-NEXT: [[TMP24:%.*]] = insertelement <2 x i16> [[TMP20]], i16 [[TMP23]], i32 1
; UNROLL-NO-IC-NEXT: br label [[PRED_UREM_CONTINUE8]]
; UNROLL-NO-IC: pred.urem.continue8:
; UNROLL-NO-IC-NEXT: [[TMP25:%.*]] = phi <2 x i16> [ [[TMP20]], [[PRED_UREM_CONTINUE6]] ], [ [[TMP24]], [[PRED_UREM_IF7]] ]
; UNROLL-NO-IC-NEXT: [[PREDPHI:%.*]] = select <2 x i1> [[TMP2]], <2 x i16> zeroinitializer, <2 x i16> [[TMP15]]
; UNROLL-NO-IC-NEXT: [[PREDPHI9:%.*]] = select <2 x i1> [[TMP3]], <2 x i16> zeroinitializer, <2 x i16> [[TMP25]]
; UNROLL-NO-IC-NEXT: [[TMP26:%.*]] = sext <2 x i16> [[PREDPHI]] to <2 x i32>
; UNROLL-NO-IC-NEXT: [[TMP27:%.*]] = sext <2 x i16> [[PREDPHI9]] to <2 x i32>
; UNROLL-NO-IC-NEXT: [[TMP28]] = or <2 x i32> [[VEC_PHI]], [[TMP26]]
; UNROLL-NO-IC-NEXT: [[TMP29]] = or <2 x i32> [[VEC_PHI1]], [[TMP27]]
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i16> [[STEP_ADD]], <i16 2, i16 2>
; UNROLL-NO-IC-NEXT: [[TMP30:%.*]] = icmp eq i32 [[INDEX_NEXT]], 20
; UNROLL-NO-IC-NEXT: br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[BIN_RDX:%.*]] = or <2 x i32> [[TMP29]], [[TMP28]]
; UNROLL-NO-IC-NEXT: [[TMP31:%.*]] = call i32 @llvm.vector.reduce.or.v2i32(<2 x i32> [[BIN_RDX]])
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i32 20, 20
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[MIDDLE_BLOCK]] ], [ -20, [[ENTRY:%.*]] ]
; UNROLL-NO-IC-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[A]], [[ENTRY]] ], [ [[TMP31]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC: for.body:
; UNROLL-NO-IC-NEXT: [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_INC:%.*]] ]
; UNROLL-NO-IC-NEXT: [[VAR0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR6:%.*]], [[FOR_INC]] ]
; UNROLL-NO-IC-NEXT: [[VAR1:%.*]] = trunc i32 [[I]] to i16
; UNROLL-NO-IC-NEXT: [[VAR2:%.*]] = icmp eq i16 [[VAR1]], 0
; UNROLL-NO-IC-NEXT: br i1 [[VAR2]], label [[FOR_INC]], label [[FOR_COND:%.*]]
; UNROLL-NO-IC: for.cond:
; UNROLL-NO-IC-NEXT: [[VAR3:%.*]] = urem i16 [[B]], [[VAR1]]
; UNROLL-NO-IC-NEXT: br label [[FOR_INC]]
; UNROLL-NO-IC: for.inc:
; UNROLL-NO-IC-NEXT: [[VAR4:%.*]] = phi i16 [ [[VAR3]], [[FOR_COND]] ], [ 0, [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VAR5:%.*]] = sext i16 [[VAR4]] to i32
; UNROLL-NO-IC-NEXT: [[VAR6]] = or i32 [[VAR0]], [[VAR5]]
; UNROLL-NO-IC-NEXT: [[I_NEXT]] = add nsw i32 [[I]], 1
; UNROLL-NO-IC-NEXT: [[COND:%.*]] = icmp eq i32 [[I_NEXT]], 0
; UNROLL-NO-IC-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP49:![0-9]+]]
; UNROLL-NO-IC: for.end:
; UNROLL-NO-IC-NEXT: [[VAR7:%.*]] = phi i32 [ [[VAR6]], [[FOR_INC]] ], [ [[TMP31]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: ret i32 [[VAR7]]
;
; INTERLEAVE-LABEL: @PR32419(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: [[TMP0:%.*]] = insertelement <4 x i32> <i32 poison, i32 0, i32 0, i32 0>, i32 [[A:%.*]], i64 0
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UREM_CONTINUE16:%.*]] ]
; INTERLEAVE-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP48:%.*]], [[PRED_UREM_CONTINUE16]] ]
; INTERLEAVE-NEXT: [[VEC_PHI1:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP49:%.*]], [[PRED_UREM_CONTINUE16]] ]
; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i16> [ <i16 -20, i16 -19, i16 -18, i16 -17>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_UREM_CONTINUE16]] ]
; INTERLEAVE-NEXT: [[TMP1:%.*]] = trunc i32 [[INDEX]] to i16
; INTERLEAVE-NEXT: [[TMP2:%.*]] = icmp eq <4 x i16> [[VEC_IND]], zeroinitializer
; INTERLEAVE-NEXT: [[TMP3:%.*]] = icmp eq <4 x i16> [[VEC_IND]], <i16 -4, i16 -4, i16 -4, i16 -4>
; INTERLEAVE-NEXT: [[TMP4:%.*]] = xor <4 x i1> [[TMP2]], <i1 true, i1 true, i1 true, i1 true>
; INTERLEAVE-NEXT: [[TMP5:%.*]] = xor <4 x i1> [[TMP3]], <i1 true, i1 true, i1 true, i1 true>
; INTERLEAVE-NEXT: [[TMP6:%.*]] = extractelement <4 x i1> [[TMP4]], i64 0
; INTERLEAVE-NEXT: br i1 [[TMP6]], label [[PRED_UREM_IF:%.*]], label [[PRED_UREM_CONTINUE:%.*]]
; INTERLEAVE: pred.urem.if:
; INTERLEAVE-NEXT: [[TMP7:%.*]] = add i16 [[TMP1]], -20
; INTERLEAVE-NEXT: [[TMP8:%.*]] = urem i16 [[B:%.*]], [[TMP7]]
; INTERLEAVE-NEXT: [[TMP9:%.*]] = insertelement <4 x i16> poison, i16 [[TMP8]], i64 0
; INTERLEAVE-NEXT: br label [[PRED_UREM_CONTINUE]]
; INTERLEAVE: pred.urem.continue:
; INTERLEAVE-NEXT: [[TMP10:%.*]] = phi <4 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP9]], [[PRED_UREM_IF]] ]
; INTERLEAVE-NEXT: [[TMP11:%.*]] = extractelement <4 x i1> [[TMP4]], i64 1
; INTERLEAVE-NEXT: br i1 [[TMP11]], label [[PRED_UREM_IF3:%.*]], label [[PRED_UREM_CONTINUE4:%.*]]
; INTERLEAVE: pred.urem.if3:
; INTERLEAVE-NEXT: [[TMP12:%.*]] = add i16 [[TMP1]], -19
; INTERLEAVE-NEXT: [[TMP13:%.*]] = urem i16 [[B]], [[TMP12]]
; INTERLEAVE-NEXT: [[TMP14:%.*]] = insertelement <4 x i16> [[TMP10]], i16 [[TMP13]], i64 1
; INTERLEAVE-NEXT: br label [[PRED_UREM_CONTINUE4]]
; INTERLEAVE: pred.urem.continue4:
; INTERLEAVE-NEXT: [[TMP15:%.*]] = phi <4 x i16> [ [[TMP10]], [[PRED_UREM_CONTINUE]] ], [ [[TMP14]], [[PRED_UREM_IF3]] ]
; INTERLEAVE-NEXT: [[TMP16:%.*]] = extractelement <4 x i1> [[TMP4]], i64 2
; INTERLEAVE-NEXT: br i1 [[TMP16]], label [[PRED_UREM_IF5:%.*]], label [[PRED_UREM_CONTINUE6:%.*]]
; INTERLEAVE: pred.urem.if5:
; INTERLEAVE-NEXT: [[TMP17:%.*]] = add i16 [[TMP1]], -18
; INTERLEAVE-NEXT: [[TMP18:%.*]] = urem i16 [[B]], [[TMP17]]
; INTERLEAVE-NEXT: [[TMP19:%.*]] = insertelement <4 x i16> [[TMP15]], i16 [[TMP18]], i64 2
; INTERLEAVE-NEXT: br label [[PRED_UREM_CONTINUE6]]
; INTERLEAVE: pred.urem.continue6:
; INTERLEAVE-NEXT: [[TMP20:%.*]] = phi <4 x i16> [ [[TMP15]], [[PRED_UREM_CONTINUE4]] ], [ [[TMP19]], [[PRED_UREM_IF5]] ]
; INTERLEAVE-NEXT: [[TMP21:%.*]] = extractelement <4 x i1> [[TMP4]], i64 3
; INTERLEAVE-NEXT: br i1 [[TMP21]], label [[PRED_UREM_IF7:%.*]], label [[PRED_UREM_CONTINUE8:%.*]]
; INTERLEAVE: pred.urem.if7:
; INTERLEAVE-NEXT: [[TMP22:%.*]] = add i16 [[TMP1]], -17
; INTERLEAVE-NEXT: [[TMP23:%.*]] = urem i16 [[B]], [[TMP22]]
; INTERLEAVE-NEXT: [[TMP24:%.*]] = insertelement <4 x i16> [[TMP20]], i16 [[TMP23]], i64 3
; INTERLEAVE-NEXT: br label [[PRED_UREM_CONTINUE8]]
; INTERLEAVE: pred.urem.continue8:
; INTERLEAVE-NEXT: [[TMP25:%.*]] = phi <4 x i16> [ [[TMP20]], [[PRED_UREM_CONTINUE6]] ], [ [[TMP24]], [[PRED_UREM_IF7]] ]
; INTERLEAVE-NEXT: [[TMP26:%.*]] = extractelement <4 x i1> [[TMP5]], i64 0
; INTERLEAVE-NEXT: br i1 [[TMP26]], label [[PRED_UREM_IF9:%.*]], label [[PRED_UREM_CONTINUE10:%.*]]
; INTERLEAVE: pred.urem.if9:
; INTERLEAVE-NEXT: [[TMP27:%.*]] = add i16 [[TMP1]], -16
; INTERLEAVE-NEXT: [[TMP28:%.*]] = urem i16 [[B]], [[TMP27]]
; INTERLEAVE-NEXT: [[TMP29:%.*]] = insertelement <4 x i16> poison, i16 [[TMP28]], i64 0
; INTERLEAVE-NEXT: br label [[PRED_UREM_CONTINUE10]]
; INTERLEAVE: pred.urem.continue10:
; INTERLEAVE-NEXT: [[TMP30:%.*]] = phi <4 x i16> [ poison, [[PRED_UREM_CONTINUE8]] ], [ [[TMP29]], [[PRED_UREM_IF9]] ]
; INTERLEAVE-NEXT: [[TMP31:%.*]] = extractelement <4 x i1> [[TMP5]], i64 1
; INTERLEAVE-NEXT: br i1 [[TMP31]], label [[PRED_UREM_IF11:%.*]], label [[PRED_UREM_CONTINUE12:%.*]]
; INTERLEAVE: pred.urem.if11:
; INTERLEAVE-NEXT: [[TMP32:%.*]] = add i16 [[TMP1]], -15
; INTERLEAVE-NEXT: [[TMP33:%.*]] = urem i16 [[B]], [[TMP32]]
; INTERLEAVE-NEXT: [[TMP34:%.*]] = insertelement <4 x i16> [[TMP30]], i16 [[TMP33]], i64 1
; INTERLEAVE-NEXT: br label [[PRED_UREM_CONTINUE12]]
; INTERLEAVE: pred.urem.continue12:
; INTERLEAVE-NEXT: [[TMP35:%.*]] = phi <4 x i16> [ [[TMP30]], [[PRED_UREM_CONTINUE10]] ], [ [[TMP34]], [[PRED_UREM_IF11]] ]
; INTERLEAVE-NEXT: [[TMP36:%.*]] = extractelement <4 x i1> [[TMP5]], i64 2
; INTERLEAVE-NEXT: br i1 [[TMP36]], label [[PRED_UREM_IF13:%.*]], label [[PRED_UREM_CONTINUE14:%.*]]
; INTERLEAVE: pred.urem.if13:
; INTERLEAVE-NEXT: [[TMP37:%.*]] = add i16 [[TMP1]], -14
; INTERLEAVE-NEXT: [[TMP38:%.*]] = urem i16 [[B]], [[TMP37]]
; INTERLEAVE-NEXT: [[TMP39:%.*]] = insertelement <4 x i16> [[TMP35]], i16 [[TMP38]], i64 2
; INTERLEAVE-NEXT: br label [[PRED_UREM_CONTINUE14]]
; INTERLEAVE: pred.urem.continue14:
; INTERLEAVE-NEXT: [[TMP40:%.*]] = phi <4 x i16> [ [[TMP35]], [[PRED_UREM_CONTINUE12]] ], [ [[TMP39]], [[PRED_UREM_IF13]] ]
; INTERLEAVE-NEXT: [[TMP41:%.*]] = extractelement <4 x i1> [[TMP5]], i64 3
; INTERLEAVE-NEXT: br i1 [[TMP41]], label [[PRED_UREM_IF15:%.*]], label [[PRED_UREM_CONTINUE16]]
; INTERLEAVE: pred.urem.if15:
; INTERLEAVE-NEXT: [[TMP42:%.*]] = add i16 [[TMP1]], -13
; INTERLEAVE-NEXT: [[TMP43:%.*]] = urem i16 [[B]], [[TMP42]]
; INTERLEAVE-NEXT: [[TMP44:%.*]] = insertelement <4 x i16> [[TMP40]], i16 [[TMP43]], i64 3
; INTERLEAVE-NEXT: br label [[PRED_UREM_CONTINUE16]]
; INTERLEAVE: pred.urem.continue16:
; INTERLEAVE-NEXT: [[TMP45:%.*]] = phi <4 x i16> [ [[TMP40]], [[PRED_UREM_CONTINUE14]] ], [ [[TMP44]], [[PRED_UREM_IF15]] ]
; INTERLEAVE-NEXT: [[PREDPHI:%.*]] = select <4 x i1> [[TMP2]], <4 x i16> zeroinitializer, <4 x i16> [[TMP25]]
; INTERLEAVE-NEXT: [[PREDPHI17:%.*]] = select <4 x i1> [[TMP3]], <4 x i16> zeroinitializer, <4 x i16> [[TMP45]]
; INTERLEAVE-NEXT: [[TMP46:%.*]] = sext <4 x i16> [[PREDPHI]] to <4 x i32>
; INTERLEAVE-NEXT: [[TMP47:%.*]] = sext <4 x i16> [[PREDPHI17]] to <4 x i32>
; INTERLEAVE-NEXT: [[TMP48]] = or <4 x i32> [[VEC_PHI]], [[TMP46]]
; INTERLEAVE-NEXT: [[TMP49]] = or <4 x i32> [[VEC_PHI1]], [[TMP47]]
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i16> [[VEC_IND]], <i16 8, i16 8, i16 8, i16 8>
; INTERLEAVE-NEXT: [[TMP50:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
; INTERLEAVE-NEXT: br i1 [[TMP50]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: [[BIN_RDX:%.*]] = or <4 x i32> [[TMP49]], [[TMP48]]
; INTERLEAVE-NEXT: [[TMP51:%.*]] = call i32 @llvm.vector.reduce.or.v4i32(<4 x i32> [[BIN_RDX]])
; INTERLEAVE-NEXT: br i1 false, label [[FOR_END:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ -4, [[MIDDLE_BLOCK]] ], [ -20, [[ENTRY:%.*]] ]
; INTERLEAVE-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP51]], [[MIDDLE_BLOCK]] ], [ [[A]], [[ENTRY]] ]
; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]]
; INTERLEAVE: for.body:
; INTERLEAVE-NEXT: [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_INC:%.*]] ]
; INTERLEAVE-NEXT: [[VAR0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[VAR6:%.*]], [[FOR_INC]] ]
; INTERLEAVE-NEXT: [[VAR1:%.*]] = trunc i32 [[I]] to i16
; INTERLEAVE-NEXT: [[VAR2:%.*]] = icmp eq i16 [[VAR1]], 0
; INTERLEAVE-NEXT: br i1 [[VAR2]], label [[FOR_INC]], label [[FOR_COND:%.*]]
; INTERLEAVE: for.cond:
; INTERLEAVE-NEXT: [[VAR3:%.*]] = urem i16 [[B]], [[VAR1]]
; INTERLEAVE-NEXT: br label [[FOR_INC]]
; INTERLEAVE: for.inc:
; INTERLEAVE-NEXT: [[VAR4:%.*]] = phi i16 [ [[VAR3]], [[FOR_COND]] ], [ 0, [[FOR_BODY]] ]
; INTERLEAVE-NEXT: [[VAR5:%.*]] = sext i16 [[VAR4]] to i32
; INTERLEAVE-NEXT: [[VAR6]] = or i32 [[VAR0]], [[VAR5]]
; INTERLEAVE-NEXT: [[I_NEXT]] = add nsw i32 [[I]], 1
; INTERLEAVE-NEXT: [[COND:%.*]] = icmp eq i32 [[I_NEXT]], 0
; INTERLEAVE-NEXT: br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP49:![0-9]+]]
; INTERLEAVE: for.end:
; INTERLEAVE-NEXT: [[VAR7:%.*]] = phi i32 [ [[VAR6]], [[FOR_INC]] ], [ [[TMP51]], [[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: ret i32 [[VAR7]]
;
entry:
br label %for.body
for.body:
%i = phi i32 [ -20, %entry ], [ %i.next, %for.inc ]
%var0 = phi i32 [ %a, %entry ], [ %var6, %for.inc ]
%var1 = trunc i32 %i to i16
%var2 = icmp eq i16 %var1, 0
br i1 %var2, label %for.inc, label %for.cond
for.cond:
%var3 = urem i16 %b, %var1
br label %for.inc
for.inc:
%var4 = phi i16 [ %var3, %for.cond ], [ 0, %for.body ]
%var5 = sext i16 %var4 to i32
%var6 = or i32 %var0, %var5
%i.next = add nsw i32 %i, 1
%cond = icmp eq i32 %i.next, 0
br i1 %cond, label %for.end, label %for.body
for.end:
%var7 = phi i32 [ %var6, %for.inc ]
ret i32 %var7
}
; Ensure that the shuffle vector for first order recurrence is inserted
; correctly after all the phis. These new phis correspond to new IVs
; that are generated by optimizing non-free truncs of IVs to IVs themselves.
; This also ensures the first-order recurrence splice recipe is placed
; correctly if it is fed by an induction.
define i64 @trunc_with_first_order_recurrence() {
; CHECK-LABEL: @trunc_with_first_order_recurrence(
; CHECK-NEXT: entry:
; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 42>, [[VECTOR_PH]] ], [ [[VEC_IND2:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND2]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT3:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND4:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT5:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND2]], <2 x i32> <i32 1, i32 2>
; CHECK-NEXT: [[TMP1:%.*]] = mul <2 x i32> [[VEC_IND]], [[VEC_IND2]]
; CHECK-NEXT: [[TMP2:%.*]] = add <2 x i32> [[TMP1]], <i32 42, i32 42>
; CHECK-NEXT: [[TMP3:%.*]] = add <2 x i32> [[TMP0]], [[VEC_IND2]]
; CHECK-NEXT: [[TMP4:%.*]] = add <2 x i32> [[TMP3]], [[TMP2]]
; CHECK-NEXT: [[TMP5:%.*]] = sext <2 x i32> [[TMP4]] to <2 x i64>
; CHECK-NEXT: [[TMP6:%.*]] = add <2 x i64> [[VEC_PHI]], [[TMP5]]
; CHECK-NEXT: [[TMP7:%.*]] = shl <2 x i32> [[VEC_IND4]], <i32 1, i32 1>
; CHECK-NEXT: [[TMP8:%.*]] = add <2 x i32> [[TMP2]], [[TMP7]]
; CHECK-NEXT: [[TMP9:%.*]] = sext <2 x i32> [[TMP8]] to <2 x i64>
; CHECK-NEXT: [[TMP10]] = add <2 x i64> [[TMP6]], [[TMP9]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; CHECK-NEXT: [[VEC_IND_NEXT3]] = add <2 x i32> [[VEC_IND2]], <i32 2, i32 2>
; CHECK-NEXT: [[VEC_IND_NEXT5]] = add <2 x i32> [[VEC_IND4]], <i32 2, i32 2>
; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 112
; CHECK-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[TMP12:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[TMP10]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 113, 112
; CHECK-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[VEC_IND2]], i32 1
; CHECK-NEXT: [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <2 x i32> [[VEC_IND2]], i32 0
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 42, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 113, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY]] ]
; CHECK-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ 113, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: exit:
; CHECK-NEXT: [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i64 [[DOTLCSSA]]
; CHECK: loop:
; CHECK-NEXT: [[C5:%.*]] = phi i64 [ [[C23]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[SCALAR_RECUR:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
; CHECK-NEXT: [[C6]] = trunc i64 [[INDVARS_IV]] to i32
; CHECK-NEXT: [[C8:%.*]] = mul i32 [[X]], [[C6]]
; CHECK-NEXT: [[C9:%.*]] = add i32 [[C8]], 42
; CHECK-NEXT: [[C10:%.*]] = add i32 [[SCALAR_RECUR]], [[C6]]
; CHECK-NEXT: [[C11:%.*]] = add i32 [[C10]], [[C9]]
; CHECK-NEXT: [[C12:%.*]] = sext i32 [[C11]] to i64
; CHECK-NEXT: [[C13:%.*]] = add i64 [[C5]], [[C12]]
; CHECK-NEXT: [[INDVARS_IV_TR:%.*]] = trunc i64 [[INDVARS_IV]] to i32
; CHECK-NEXT: [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 1
; CHECK-NEXT: [[C15:%.*]] = add i32 [[C9]], [[C14]]
; CHECK-NEXT: [[C16:%.*]] = sext i32 [[C15]] to i64
; CHECK-NEXT: [[C23]] = add i64 [[C13]], [[C16]]
; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; CHECK-NEXT: [[C24]] = add nuw nsw i32 [[X]], 1
; CHECK-NEXT: [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 114
; CHECK-NEXT: br i1 [[EXITCOND_I]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP51:![0-9]+]]
;
; IND-LABEL: @trunc_with_first_order_recurrence(
; IND-NEXT: entry:
; IND-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 42>, [[VECTOR_PH]] ], [ [[VEC_IND2:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VEC_IND2]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT3:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VEC_IND4:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT5:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND2]], <2 x i32> <i32 1, i32 2>
; IND-NEXT: [[TMP1:%.*]] = mul <2 x i32> [[VEC_IND]], [[VEC_IND2]]
; IND-NEXT: [[TMP2:%.*]] = add <2 x i32> [[TMP1]], <i32 42, i32 42>
; IND-NEXT: [[TMP3:%.*]] = add <2 x i32> [[TMP0]], [[VEC_IND2]]
; IND-NEXT: [[TMP4:%.*]] = add <2 x i32> [[TMP3]], [[TMP2]]
; IND-NEXT: [[TMP5:%.*]] = sext <2 x i32> [[TMP4]] to <2 x i64>
; IND-NEXT: [[TMP6:%.*]] = add <2 x i64> [[VEC_PHI]], [[TMP5]]
; IND-NEXT: [[TMP7:%.*]] = shl <2 x i32> [[VEC_IND4]], <i32 1, i32 1>
; IND-NEXT: [[TMP8:%.*]] = add <2 x i32> [[TMP2]], [[TMP7]]
; IND-NEXT: [[TMP9:%.*]] = sext <2 x i32> [[TMP8]] to <2 x i64>
; IND-NEXT: [[TMP10]] = add <2 x i64> [[TMP6]], [[TMP9]]
; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; IND-NEXT: [[VEC_IND_NEXT3]] = add <2 x i32> [[VEC_IND2]], <i32 2, i32 2>
; IND-NEXT: [[VEC_IND_NEXT5]] = add <2 x i32> [[VEC_IND4]], <i32 2, i32 2>
; IND-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 112
; IND-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[TMP12:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[TMP10]])
; IND-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[VEC_IND2]], i64 1
; IND-NEXT: br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 42, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1, [[ENTRY]] ], [ 113, [[MIDDLE_BLOCK]] ]
; IND-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ 1, [[ENTRY]] ], [ 113, [[MIDDLE_BLOCK]] ]
; IND-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
; IND-NEXT: br label [[LOOP:%.*]]
; IND: exit:
; IND-NEXT: [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
; IND-NEXT: ret i64 [[DOTLCSSA]]
; IND: loop:
; IND-NEXT: [[C5:%.*]] = phi i64 [ [[C23]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; IND-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; IND-NEXT: [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; IND-NEXT: [[SCALAR_RECUR:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
; IND-NEXT: [[C6]] = trunc i64 [[INDVARS_IV]] to i32
; IND-NEXT: [[C8:%.*]] = mul i32 [[X]], [[C6]]
; IND-NEXT: [[C9:%.*]] = add i32 [[C8]], 42
; IND-NEXT: [[C10:%.*]] = add i32 [[SCALAR_RECUR]], [[C6]]
; IND-NEXT: [[C11:%.*]] = add i32 [[C10]], [[C9]]
; IND-NEXT: [[C12:%.*]] = sext i32 [[C11]] to i64
; IND-NEXT: [[C13:%.*]] = add i64 [[C5]], [[C12]]
; IND-NEXT: [[INDVARS_IV_TR:%.*]] = trunc i64 [[INDVARS_IV]] to i32
; IND-NEXT: [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 1
; IND-NEXT: [[C15:%.*]] = add i32 [[C9]], [[C14]]
; IND-NEXT: [[C16:%.*]] = sext i32 [[C15]] to i64
; IND-NEXT: [[C23]] = add i64 [[C13]], [[C16]]
; IND-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; IND-NEXT: [[C24]] = add nuw nsw i32 [[X]], 1
; IND-NEXT: [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 114
; IND-NEXT: br i1 [[EXITCOND_I]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP51:![0-9]+]]
;
; UNROLL-LABEL: @trunc_with_first_order_recurrence(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_PHI2:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 42>, [[VECTOR_PH]] ], [ [[STEP_ADD5:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_IND4:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT7:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_IND8:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT11:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; UNROLL-NEXT: [[STEP_ADD5]] = add <2 x i32> [[VEC_IND4]], <i32 2, i32 2>
; UNROLL-NEXT: [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND4]], <2 x i32> <i32 1, i32 2>
; UNROLL-NEXT: [[TMP1:%.*]] = shufflevector <2 x i32> [[VEC_IND4]], <2 x i32> [[STEP_ADD5]], <2 x i32> <i32 1, i32 2>
; UNROLL-NEXT: [[TMP2:%.*]] = mul <2 x i32> [[VEC_IND]], [[VEC_IND4]]
; UNROLL-NEXT: [[TMP3:%.*]] = mul <2 x i32> [[STEP_ADD]], [[STEP_ADD5]]
; UNROLL-NEXT: [[TMP4:%.*]] = add <2 x i32> [[TMP2]], <i32 42, i32 42>
; UNROLL-NEXT: [[TMP5:%.*]] = add <2 x i32> [[TMP3]], <i32 42, i32 42>
; UNROLL-NEXT: [[TMP6:%.*]] = add <2 x i32> [[TMP0]], [[VEC_IND4]]
; UNROLL-NEXT: [[TMP7:%.*]] = add <2 x i32> [[TMP1]], [[STEP_ADD5]]
; UNROLL-NEXT: [[TMP8:%.*]] = add <2 x i32> [[TMP6]], [[TMP4]]
; UNROLL-NEXT: [[TMP9:%.*]] = add <2 x i32> [[TMP7]], [[TMP5]]
; UNROLL-NEXT: [[TMP10:%.*]] = sext <2 x i32> [[TMP8]] to <2 x i64>
; UNROLL-NEXT: [[TMP11:%.*]] = sext <2 x i32> [[TMP9]] to <2 x i64>
; UNROLL-NEXT: [[TMP12:%.*]] = add <2 x i64> [[VEC_PHI]], [[TMP10]]
; UNROLL-NEXT: [[TMP13:%.*]] = add <2 x i64> [[VEC_PHI2]], [[TMP11]]
; UNROLL-NEXT: [[TMP14:%.*]] = shl <2 x i32> [[VEC_IND8]], <i32 1, i32 1>
; UNROLL-NEXT: [[STEP_ADD9:%.*]] = shl <2 x i32> [[VEC_IND8]], <i32 1, i32 1>
; UNROLL-NEXT: [[TMP15:%.*]] = add <2 x i32> [[STEP_ADD9]], <i32 4, i32 4>
; UNROLL-NEXT: [[TMP16:%.*]] = add <2 x i32> [[TMP4]], [[TMP14]]
; UNROLL-NEXT: [[TMP17:%.*]] = add <2 x i32> [[TMP5]], [[TMP15]]
; UNROLL-NEXT: [[TMP18:%.*]] = sext <2 x i32> [[TMP16]] to <2 x i64>
; UNROLL-NEXT: [[TMP19:%.*]] = sext <2 x i32> [[TMP17]] to <2 x i64>
; UNROLL-NEXT: [[TMP20]] = add <2 x i64> [[TMP12]], [[TMP18]]
; UNROLL-NEXT: [[TMP21]] = add <2 x i64> [[TMP13]], [[TMP19]]
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 4, i32 4>
; UNROLL-NEXT: [[VEC_IND_NEXT7]] = add <2 x i32> [[VEC_IND4]], <i32 4, i32 4>
; UNROLL-NEXT: [[VEC_IND_NEXT11]] = add <2 x i32> [[VEC_IND8]], <i32 4, i32 4>
; UNROLL-NEXT: [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], 112
; UNROLL-NEXT: br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[BIN_RDX:%.*]] = add <2 x i64> [[TMP21]], [[TMP20]]
; UNROLL-NEXT: [[TMP23:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])
; UNROLL-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[STEP_ADD5]], i64 1
; UNROLL-NEXT: br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 42, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1, [[ENTRY]] ], [ 113, [[MIDDLE_BLOCK]] ]
; UNROLL-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ 1, [[ENTRY]] ], [ 113, [[MIDDLE_BLOCK]] ]
; UNROLL-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[TMP23]], [[MIDDLE_BLOCK]] ]
; UNROLL-NEXT: br label [[LOOP:%.*]]
; UNROLL: exit:
; UNROLL-NEXT: [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ], [ [[TMP23]], [[MIDDLE_BLOCK]] ]
; UNROLL-NEXT: ret i64 [[DOTLCSSA]]
; UNROLL: loop:
; UNROLL-NEXT: [[C5:%.*]] = phi i64 [ [[C23]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[SCALAR_RECUR:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
; UNROLL-NEXT: [[C6]] = trunc i64 [[INDVARS_IV]] to i32
; UNROLL-NEXT: [[C8:%.*]] = mul i32 [[X]], [[C6]]
; UNROLL-NEXT: [[C9:%.*]] = add i32 [[C8]], 42
; UNROLL-NEXT: [[C10:%.*]] = add i32 [[SCALAR_RECUR]], [[C6]]
; UNROLL-NEXT: [[C11:%.*]] = add i32 [[C10]], [[C9]]
; UNROLL-NEXT: [[C12:%.*]] = sext i32 [[C11]] to i64
; UNROLL-NEXT: [[C13:%.*]] = add i64 [[C5]], [[C12]]
; UNROLL-NEXT: [[INDVARS_IV_TR:%.*]] = trunc i64 [[INDVARS_IV]] to i32
; UNROLL-NEXT: [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 1
; UNROLL-NEXT: [[C15:%.*]] = add i32 [[C9]], [[C14]]
; UNROLL-NEXT: [[C16:%.*]] = sext i32 [[C15]] to i64
; UNROLL-NEXT: [[C23]] = add i64 [[C13]], [[C16]]
; UNROLL-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; UNROLL-NEXT: [[C24]] = add nuw nsw i32 [[X]], 1
; UNROLL-NEXT: [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 114
; UNROLL-NEXT: br i1 [[EXITCOND_I]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP51:![0-9]+]]
;
; UNROLL-NO-IC-LABEL: @trunc_with_first_order_recurrence(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_PHI2:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 42>, [[VECTOR_PH]] ], [ [[STEP_ADD5:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_IND4:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT7:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_IND8:%.*]] = phi <2 x i32> [ <i32 1, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT11:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[STEP_ADD5]] = add <2 x i32> [[VEC_IND4]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[STEP_ADD9:%.*]] = add <2 x i32> [[VEC_IND8]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND4]], <2 x i32> <i32 1, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = shufflevector <2 x i32> [[VEC_IND4]], <2 x i32> [[STEP_ADD5]], <2 x i32> <i32 1, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = mul <2 x i32> [[VEC_IND]], [[VEC_IND4]]
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = mul <2 x i32> [[STEP_ADD]], [[STEP_ADD5]]
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = add <2 x i32> [[TMP2]], <i32 42, i32 42>
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = add <2 x i32> [[TMP3]], <i32 42, i32 42>
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = add <2 x i32> [[TMP0]], [[VEC_IND4]]
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = add <2 x i32> [[TMP1]], [[STEP_ADD5]]
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = add <2 x i32> [[TMP6]], [[TMP4]]
; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = add <2 x i32> [[TMP7]], [[TMP5]]
; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = sext <2 x i32> [[TMP8]] to <2 x i64>
; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = sext <2 x i32> [[TMP9]] to <2 x i64>
; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = add <2 x i64> [[VEC_PHI]], [[TMP10]]
; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = add <2 x i64> [[VEC_PHI2]], [[TMP11]]
; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = shl <2 x i32> [[VEC_IND8]], <i32 1, i32 1>
; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = shl <2 x i32> [[STEP_ADD9]], <i32 1, i32 1>
; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = add <2 x i32> [[TMP4]], [[TMP14]]
; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = add <2 x i32> [[TMP5]], [[TMP15]]
; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = sext <2 x i32> [[TMP16]] to <2 x i64>
; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = sext <2 x i32> [[TMP17]] to <2 x i64>
; UNROLL-NO-IC-NEXT: [[TMP20]] = add <2 x i64> [[TMP12]], [[TMP18]]
; UNROLL-NO-IC-NEXT: [[TMP21]] = add <2 x i64> [[TMP13]], [[TMP19]]
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT7]] = add <2 x i32> [[STEP_ADD5]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT11]] = add <2 x i32> [[STEP_ADD9]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], 112
; UNROLL-NO-IC-NEXT: br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[BIN_RDX:%.*]] = add <2 x i64> [[TMP21]], [[TMP20]]
; UNROLL-NO-IC-NEXT: [[TMP23:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 113, 112
; UNROLL-NO-IC-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[STEP_ADD5]], i32 1
; UNROLL-NO-IC-NEXT: [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <2 x i32> [[STEP_ADD5]], i32 0
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 42, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 113, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY]] ]
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ 113, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY]] ]
; UNROLL-NO-IC-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[TMP23]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: br label [[LOOP:%.*]]
; UNROLL-NO-IC: exit:
; UNROLL-NO-IC-NEXT: [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ], [ [[TMP23]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: ret i64 [[DOTLCSSA]]
; UNROLL-NO-IC: loop:
; UNROLL-NO-IC-NEXT: [[C5:%.*]] = phi i64 [ [[C23]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[SCALAR_RECUR:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT: [[C6]] = trunc i64 [[INDVARS_IV]] to i32
; UNROLL-NO-IC-NEXT: [[C8:%.*]] = mul i32 [[X]], [[C6]]
; UNROLL-NO-IC-NEXT: [[C9:%.*]] = add i32 [[C8]], 42
; UNROLL-NO-IC-NEXT: [[C10:%.*]] = add i32 [[SCALAR_RECUR]], [[C6]]
; UNROLL-NO-IC-NEXT: [[C11:%.*]] = add i32 [[C10]], [[C9]]
; UNROLL-NO-IC-NEXT: [[C12:%.*]] = sext i32 [[C11]] to i64
; UNROLL-NO-IC-NEXT: [[C13:%.*]] = add i64 [[C5]], [[C12]]
; UNROLL-NO-IC-NEXT: [[INDVARS_IV_TR:%.*]] = trunc i64 [[INDVARS_IV]] to i32
; UNROLL-NO-IC-NEXT: [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 1
; UNROLL-NO-IC-NEXT: [[C15:%.*]] = add i32 [[C9]], [[C14]]
; UNROLL-NO-IC-NEXT: [[C16:%.*]] = sext i32 [[C15]] to i64
; UNROLL-NO-IC-NEXT: [[C23]] = add i64 [[C13]], [[C16]]
; UNROLL-NO-IC-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; UNROLL-NO-IC-NEXT: [[C24]] = add nuw nsw i32 [[X]], 1
; UNROLL-NO-IC-NEXT: [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 114
; UNROLL-NO-IC-NEXT: br i1 [[EXITCOND_I]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP51:![0-9]+]]
;
; INTERLEAVE-LABEL: @trunc_with_first_order_recurrence(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_PHI:%.*]] = phi <4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_PHI2:%.*]] = phi <4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 1, i32 2, i32 3, i32 4>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 42>, [[VECTOR_PH]] ], [ [[STEP_ADD5:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_IND4:%.*]] = phi <4 x i32> [ <i32 1, i32 2, i32 3, i32 4>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT7:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_IND8:%.*]] = phi <4 x i32> [ <i32 1, i32 2, i32 3, i32 4>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT11:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; INTERLEAVE-NEXT: [[STEP_ADD5]] = add <4 x i32> [[VEC_IND4]], <i32 4, i32 4, i32 4, i32 4>
; INTERLEAVE-NEXT: [[TMP0:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[VEC_IND4]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; INTERLEAVE-NEXT: [[TMP1:%.*]] = shufflevector <4 x i32> [[VEC_IND4]], <4 x i32> [[STEP_ADD5]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; INTERLEAVE-NEXT: [[TMP2:%.*]] = mul <4 x i32> [[VEC_IND]], [[VEC_IND4]]
; INTERLEAVE-NEXT: [[TMP3:%.*]] = mul <4 x i32> [[STEP_ADD]], [[STEP_ADD5]]
; INTERLEAVE-NEXT: [[TMP4:%.*]] = add <4 x i32> [[TMP2]], <i32 42, i32 42, i32 42, i32 42>
; INTERLEAVE-NEXT: [[TMP5:%.*]] = add <4 x i32> [[TMP3]], <i32 42, i32 42, i32 42, i32 42>
; INTERLEAVE-NEXT: [[TMP6:%.*]] = add <4 x i32> [[TMP0]], [[VEC_IND4]]
; INTERLEAVE-NEXT: [[TMP7:%.*]] = add <4 x i32> [[TMP1]], [[STEP_ADD5]]
; INTERLEAVE-NEXT: [[TMP8:%.*]] = add <4 x i32> [[TMP6]], [[TMP4]]
; INTERLEAVE-NEXT: [[TMP9:%.*]] = add <4 x i32> [[TMP7]], [[TMP5]]
; INTERLEAVE-NEXT: [[TMP10:%.*]] = sext <4 x i32> [[TMP8]] to <4 x i64>
; INTERLEAVE-NEXT: [[TMP11:%.*]] = sext <4 x i32> [[TMP9]] to <4 x i64>
; INTERLEAVE-NEXT: [[TMP12:%.*]] = add <4 x i64> [[VEC_PHI]], [[TMP10]]
; INTERLEAVE-NEXT: [[TMP13:%.*]] = add <4 x i64> [[VEC_PHI2]], [[TMP11]]
; INTERLEAVE-NEXT: [[TMP14:%.*]] = shl <4 x i32> [[VEC_IND8]], <i32 1, i32 1, i32 1, i32 1>
; INTERLEAVE-NEXT: [[STEP_ADD9:%.*]] = shl <4 x i32> [[VEC_IND8]], <i32 1, i32 1, i32 1, i32 1>
; INTERLEAVE-NEXT: [[TMP15:%.*]] = add <4 x i32> [[STEP_ADD9]], <i32 8, i32 8, i32 8, i32 8>
; INTERLEAVE-NEXT: [[TMP16:%.*]] = add <4 x i32> [[TMP4]], [[TMP14]]
; INTERLEAVE-NEXT: [[TMP17:%.*]] = add <4 x i32> [[TMP5]], [[TMP15]]
; INTERLEAVE-NEXT: [[TMP18:%.*]] = sext <4 x i32> [[TMP16]] to <4 x i64>
; INTERLEAVE-NEXT: [[TMP19:%.*]] = sext <4 x i32> [[TMP17]] to <4 x i64>
; INTERLEAVE-NEXT: [[TMP20]] = add <4 x i64> [[TMP12]], [[TMP18]]
; INTERLEAVE-NEXT: [[TMP21]] = add <4 x i64> [[TMP13]], [[TMP19]]
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 8, i32 8, i32 8, i32 8>
; INTERLEAVE-NEXT: [[VEC_IND_NEXT7]] = add <4 x i32> [[VEC_IND4]], <i32 8, i32 8, i32 8, i32 8>
; INTERLEAVE-NEXT: [[VEC_IND_NEXT11]] = add <4 x i32> [[VEC_IND8]], <i32 8, i32 8, i32 8, i32 8>
; INTERLEAVE-NEXT: [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], 112
; INTERLEAVE-NEXT: br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: [[BIN_RDX:%.*]] = add <4 x i64> [[TMP21]], [[TMP20]]
; INTERLEAVE-NEXT: [[TMP23:%.*]] = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> [[BIN_RDX]])
; INTERLEAVE-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[STEP_ADD5]], i64 3
; INTERLEAVE-NEXT: br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 42, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1, [[ENTRY]] ], [ 113, [[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ 1, [[ENTRY]] ], [ 113, [[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: [[BC_MERGE_RDX:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[TMP23]], [[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: br label [[LOOP:%.*]]
; INTERLEAVE: exit:
; INTERLEAVE-NEXT: [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ], [ [[TMP23]], [[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: ret i64 [[DOTLCSSA]]
; INTERLEAVE: loop:
; INTERLEAVE-NEXT: [[C5:%.*]] = phi i64 [ [[C23]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[SCALAR_RECUR:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
; INTERLEAVE-NEXT: [[C6]] = trunc i64 [[INDVARS_IV]] to i32
; INTERLEAVE-NEXT: [[C8:%.*]] = mul i32 [[X]], [[C6]]
; INTERLEAVE-NEXT: [[C9:%.*]] = add i32 [[C8]], 42
; INTERLEAVE-NEXT: [[C10:%.*]] = add i32 [[SCALAR_RECUR]], [[C6]]
; INTERLEAVE-NEXT: [[C11:%.*]] = add i32 [[C10]], [[C9]]
; INTERLEAVE-NEXT: [[C12:%.*]] = sext i32 [[C11]] to i64
; INTERLEAVE-NEXT: [[C13:%.*]] = add i64 [[C5]], [[C12]]
; INTERLEAVE-NEXT: [[INDVARS_IV_TR:%.*]] = trunc i64 [[INDVARS_IV]] to i32
; INTERLEAVE-NEXT: [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 1
; INTERLEAVE-NEXT: [[C15:%.*]] = add i32 [[C9]], [[C14]]
; INTERLEAVE-NEXT: [[C16:%.*]] = sext i32 [[C15]] to i64
; INTERLEAVE-NEXT: [[C23]] = add i64 [[C13]], [[C16]]
; INTERLEAVE-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; INTERLEAVE-NEXT: [[C24]] = add nuw nsw i32 [[X]], 1
; INTERLEAVE-NEXT: [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 114
; INTERLEAVE-NEXT: br i1 [[EXITCOND_I]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP51:![0-9]+]]
;
entry:
br label %loop
exit: ; preds = %loop
%.lcssa = phi i64 [ %c23, %loop ]
ret i64 %.lcssa
loop: ; preds = %loop, %entry
%c5 = phi i64 [ %c23, %loop ], [ 0, %entry ]
%indvars.iv = phi i64 [ %indvars.iv.next, %loop ], [ 1, %entry ]
%x = phi i32 [ %c24, %loop ], [ 1, %entry ]
%y = phi i32 [ %c6, %loop ], [ 42, %entry ]
%c6 = trunc i64 %indvars.iv to i32
%c8 = mul i32 %x, %c6
%c9 = add i32 %c8, 42
%c10 = add i32 %y, %c6
%c11 = add i32 %c10, %c9
%c12 = sext i32 %c11 to i64
%c13 = add i64 %c5, %c12
%indvars.iv.tr = trunc i64 %indvars.iv to i32
%c14 = shl i32 %indvars.iv.tr, 1
%c15 = add i32 %c9, %c14
%c16 = sext i32 %c15 to i64
%c23 = add i64 %c13, %c16
%indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
%c24 = add nuw nsw i32 %x, 1
%exitcond.i = icmp eq i64 %indvars.iv.next, 114
br i1 %exitcond.i, label %exit, label %loop
}
; Test case for PR52460.
define void @pr52460_first_order_recurrence_truncated_iv(i32* noalias %src, i32* %dst) {
;
; CHECK-LABEL: @pr52460_first_order_recurrence_truncated_iv(
; CHECK-NEXT: entry:
; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[VEC_IND:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP0:%.*]] = trunc i64 [[INDEX]] to i32
; CHECK-NEXT: [[TMP1:%.*]] = add i32 [[TMP0]], 0
; CHECK-NEXT: [[TMP2:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>
; CHECK-NEXT: [[TMP3:%.*]] = load i32, i32* [[SRC:%.*]], align 4
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP3]], i32 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP4:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT]], [[TMP2]]
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr i32, i32* [[DST:%.*]], i32 [[TMP1]]
; CHECK-NEXT: [[TMP6:%.*]] = add <2 x i32> [[VEC_IND]], [[TMP4]]
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr i32, i32* [[TMP5]], i32 0
; CHECK-NEXT: [[TMP8:%.*]] = bitcast i32* [[TMP7]] to <2 x i32>*
; CHECK-NEXT: store <2 x i32> [[TMP6]], <2 x i32>* [[TMP8]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
; CHECK-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP52:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, 100
; CHECK-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[VEC_IND]], i32 1
; CHECK-NEXT: [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <2 x i32> [[VEC_IND]], i32 0
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; CHECK-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[TRUNC_IV:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[TRUNC_IV_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_TRUNC:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[LV:%.*]] = load i32, i32* [[SRC]], align 4
; CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[LV]], [[SCALAR_RECUR]]
; CHECK-NEXT: [[TRUNC_IV_NEXT]] = add i32 [[TRUNC_IV]], 1
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[IV_TRUNC]] = trunc i64 [[IV]] to i32
; CHECK-NEXT: [[DST_GEP:%.*]] = getelementptr i32, i32* [[DST]], i32 [[IV_TRUNC]]
; CHECK-NEXT: [[ADD:%.*]] = add i32 [[IV_TRUNC]], [[MUL]]
; CHECK-NEXT: store i32 [[ADD]], i32* [[DST_GEP]], align 4
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[TRUNC_IV_NEXT]], 100
; CHECK-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP53:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
; IND-LABEL: @pr52460_first_order_recurrence_truncated_iv(
; IND-NEXT: entry:
; IND-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[VEC_IND:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VEC_IND]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>
; IND-NEXT: [[TMP1:%.*]] = load i32, i32* [[SRC:%.*]], align 4
; IND-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP1]], i64 0
; IND-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; IND-NEXT: [[TMP2:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT]], [[TMP0]]
; IND-NEXT: [[SEXT:%.*]] = shl i64 [[INDEX]], 32
; IND-NEXT: [[TMP3:%.*]] = ashr exact i64 [[SEXT]], 32
; IND-NEXT: [[TMP4:%.*]] = getelementptr i32, i32* [[DST:%.*]], i64 [[TMP3]]
; IND-NEXT: [[TMP5:%.*]] = add <2 x i32> [[VEC_IND]], [[TMP2]]
; IND-NEXT: [[TMP6:%.*]] = bitcast i32* [[TMP4]] to <2 x i32>*
; IND-NEXT: store <2 x i32> [[TMP5]], <2 x i32>* [[TMP6]], align 4
; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; IND-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
; IND-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP52:![0-9]+]]
; IND: middle.block:
; IND-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: br label [[LOOP:%.*]]
; IND: loop:
; IND-NEXT: br i1 poison, label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP53:![0-9]+]]
; IND: exit:
; IND-NEXT: ret void
;
; UNROLL-LABEL: @pr52460_first_order_recurrence_truncated_iv(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[STEP_ADD]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; UNROLL-NEXT: [[TMP0:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>
; UNROLL-NEXT: [[TMP1:%.*]] = shufflevector <2 x i32> [[VEC_IND]], <2 x i32> [[STEP_ADD]], <2 x i32> <i32 1, i32 2>
; UNROLL-NEXT: [[TMP2:%.*]] = load i32, i32* [[SRC:%.*]], align 4
; UNROLL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP2]], i64 0
; UNROLL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NEXT: [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <2 x i32> poison, i32 [[TMP2]], i64 0
; UNROLL-NEXT: [[BROADCAST_SPLAT4:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT3]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NEXT: [[TMP3:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT]], [[TMP0]]
; UNROLL-NEXT: [[TMP4:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT4]], [[TMP1]]
; UNROLL-NEXT: [[SEXT:%.*]] = shl i64 [[INDEX]], 32
; UNROLL-NEXT: [[TMP5:%.*]] = ashr exact i64 [[SEXT]], 32
; UNROLL-NEXT: [[TMP6:%.*]] = getelementptr i32, i32* [[DST:%.*]], i64 [[TMP5]]
; UNROLL-NEXT: [[TMP7:%.*]] = add <2 x i32> [[VEC_IND]], [[TMP3]]
; UNROLL-NEXT: [[TMP8:%.*]] = add <2 x i32> [[STEP_ADD]], [[TMP4]]
; UNROLL-NEXT: [[TMP9:%.*]] = bitcast i32* [[TMP6]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[TMP7]], <2 x i32>* [[TMP9]], align 4
; UNROLL-NEXT: [[TMP10:%.*]] = getelementptr i32, i32* [[TMP6]], i64 2
; UNROLL-NEXT: [[TMP11:%.*]] = bitcast i32* [[TMP10]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[TMP8]], <2 x i32>* [[TMP11]], align 4
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], <i32 4, i32 4>
; UNROLL-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
; UNROLL-NEXT: br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP52:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: br label [[LOOP:%.*]]
; UNROLL: loop:
; UNROLL-NEXT: br i1 poison, label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP53:![0-9]+]]
; UNROLL: exit:
; UNROLL-NEXT: ret void
;
; UNROLL-NO-IC-LABEL: @pr52460_first_order_recurrence_truncated_iv(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[STEP_ADD]] = add <2 x i32> [[VEC_IND]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = trunc i64 [[INDEX]] to i32
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = add i32 [[TMP0]], 0
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = add i32 [[TMP0]], 2
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = shufflevector <2 x i32> [[VEC_IND]], <2 x i32> [[STEP_ADD]], <2 x i32> <i32 1, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = load i32, i32* [[SRC:%.*]], align 4
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP5]], i32 0
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = load i32, i32* [[SRC]], align 4
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <2 x i32> poison, i32 [[TMP6]], i32 0
; UNROLL-NO-IC-NEXT: [[BROADCAST_SPLAT4:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT3]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT]], [[TMP3]]
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT4]], [[TMP4]]
; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = getelementptr i32, i32* [[DST:%.*]], i32 [[TMP1]]
; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = getelementptr i32, i32* [[DST]], i32 [[TMP2]]
; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = add <2 x i32> [[VEC_IND]], [[TMP7]]
; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = add <2 x i32> [[STEP_ADD]], [[TMP8]]
; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = getelementptr i32, i32* [[TMP9]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = bitcast i32* [[TMP13]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[TMP11]], <2 x i32>* [[TMP14]], align 4
; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = getelementptr i32, i32* [[TMP9]], i32 2
; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = bitcast i32* [[TMP15]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[TMP12]], <2 x i32>* [[TMP16]], align 4
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], <i32 2, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
; UNROLL-NO-IC-NEXT: br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP52:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, 100
; UNROLL-NO-IC-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[STEP_ADD]], i32 1
; UNROLL-NO-IC-NEXT: [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <2 x i32> [[STEP_ADD]], i32 0
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-IC-NEXT: br label [[LOOP:%.*]]
; UNROLL-NO-IC: loop:
; UNROLL-NO-IC-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; UNROLL-NO-IC-NEXT: [[TRUNC_IV:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[TRUNC_IV_NEXT:%.*]], [[LOOP]] ]
; UNROLL-NO-IC-NEXT: [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_TRUNC:%.*]], [[LOOP]] ]
; UNROLL-NO-IC-NEXT: [[LV:%.*]] = load i32, i32* [[SRC]], align 4
; UNROLL-NO-IC-NEXT: [[MUL:%.*]] = mul nsw i32 [[LV]], [[SCALAR_RECUR]]
; UNROLL-NO-IC-NEXT: [[TRUNC_IV_NEXT]] = add i32 [[TRUNC_IV]], 1
; UNROLL-NO-IC-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; UNROLL-NO-IC-NEXT: [[IV_TRUNC]] = trunc i64 [[IV]] to i32
; UNROLL-NO-IC-NEXT: [[DST_GEP:%.*]] = getelementptr i32, i32* [[DST]], i32 [[IV_TRUNC]]
; UNROLL-NO-IC-NEXT: [[ADD:%.*]] = add i32 [[IV_TRUNC]], [[MUL]]
; UNROLL-NO-IC-NEXT: store i32 [[ADD]], i32* [[DST_GEP]], align 4
; UNROLL-NO-IC-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[TRUNC_IV_NEXT]], 100
; UNROLL-NO-IC-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP53:![0-9]+]]
; UNROLL-NO-IC: exit:
; UNROLL-NO-IC-NEXT: ret void
;
; INTERLEAVE-LABEL: @pr52460_first_order_recurrence_truncated_iv(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[STEP_ADD]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; 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>
; 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>
; INTERLEAVE-NEXT: [[TMP2:%.*]] = load i32, i32* [[SRC:%.*]], align 4
; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TMP2]], i64 0
; INTERLEAVE-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; INTERLEAVE-NEXT: [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <4 x i32> poison, i32 [[TMP2]], i64 0
; INTERLEAVE-NEXT: [[BROADCAST_SPLAT4:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT3]], <4 x i32> poison, <4 x i32> zeroinitializer
; INTERLEAVE-NEXT: [[TMP3:%.*]] = mul nsw <4 x i32> [[BROADCAST_SPLAT]], [[TMP0]]
; INTERLEAVE-NEXT: [[TMP4:%.*]] = mul nsw <4 x i32> [[BROADCAST_SPLAT4]], [[TMP1]]
; INTERLEAVE-NEXT: [[SEXT:%.*]] = shl i64 [[INDEX]], 32
; INTERLEAVE-NEXT: [[TMP5:%.*]] = ashr exact i64 [[SEXT]], 32
; INTERLEAVE-NEXT: [[TMP6:%.*]] = getelementptr i32, i32* [[DST:%.*]], i64 [[TMP5]]
; INTERLEAVE-NEXT: [[TMP7:%.*]] = add <4 x i32> [[VEC_IND]], [[TMP3]]
; INTERLEAVE-NEXT: [[TMP8:%.*]] = add <4 x i32> [[STEP_ADD]], [[TMP4]]
; INTERLEAVE-NEXT: [[TMP9:%.*]] = bitcast i32* [[TMP6]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[TMP7]], <4 x i32>* [[TMP9]], align 4
; INTERLEAVE-NEXT: [[TMP10:%.*]] = getelementptr i32, i32* [[TMP6]], i64 4
; INTERLEAVE-NEXT: [[TMP11:%.*]] = bitcast i32* [[TMP10]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[TMP8]], <4 x i32>* [[TMP11]], align 4
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 8, i32 8, i32 8, i32 8>
; INTERLEAVE-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 96
; INTERLEAVE-NEXT: br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP52:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[STEP_ADD]], i64 3
; INTERLEAVE-NEXT: br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ 96, [[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ 96, [[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: br label [[LOOP:%.*]]
; INTERLEAVE: loop:
; INTERLEAVE-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; INTERLEAVE-NEXT: [[TRUNC_IV:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[TRUNC_IV_NEXT:%.*]], [[LOOP]] ]
; INTERLEAVE-NEXT: [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_TRUNC:%.*]], [[LOOP]] ]
; INTERLEAVE-NEXT: [[LV:%.*]] = load i32, i32* [[SRC]], align 4
; INTERLEAVE-NEXT: [[MUL:%.*]] = mul nsw i32 [[LV]], [[SCALAR_RECUR]]
; INTERLEAVE-NEXT: [[TRUNC_IV_NEXT]] = add i32 [[TRUNC_IV]], 1
; INTERLEAVE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; INTERLEAVE-NEXT: [[IV_TRUNC]] = trunc i64 [[IV]] to i32
; INTERLEAVE-NEXT: [[SEXT5:%.*]] = shl i64 [[IV]], 32
; INTERLEAVE-NEXT: [[TMP13:%.*]] = ashr exact i64 [[SEXT5]], 32
; INTERLEAVE-NEXT: [[DST_GEP:%.*]] = getelementptr i32, i32* [[DST]], i64 [[TMP13]]
; INTERLEAVE-NEXT: [[ADD:%.*]] = add i32 [[MUL]], [[IV_TRUNC]]
; INTERLEAVE-NEXT: store i32 [[ADD]], i32* [[DST_GEP]], align 4
; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[TRUNC_IV_NEXT]], 100
; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP53:![0-9]+]]
; INTERLEAVE: exit:
; INTERLEAVE-NEXT: ret void
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%trunc.iv = phi i32 [ 0, %entry ], [ %trunc.iv.next, %loop ]
%recur = phi i32 [ 0, %entry ], [ %iv.trunc, %loop ]
%lv = load i32, i32* %src, align 4
%mul = mul nsw i32 %lv, %recur
%trunc.iv.next = add i32 %trunc.iv, 1
%iv.next = add nuw nsw i64 %iv, 1
%iv.trunc = trunc i64 %iv to i32
%dst.gep = getelementptr i32, i32* %dst, i32 %iv.trunc
%add = add i32 %iv.trunc, %mul
store i32 %add, i32* %dst.gep
%exitcond = icmp eq i32 %trunc.iv.next, 100
br i1 %exitcond, label %exit, label %loop
exit:
ret void
}
; Test case where %iv.2.ext and %iv.2.conv become redundant due to the SCEV
; predicates generated for the vector loop. They should be removed in the
; vector loop.
define void @test_optimized_cast_induction_feeding_first_order_recurrence(i64 %n, i32 %step, i32* %ptr) {
;
; CHECK-LABEL: @test_optimized_cast_induction_feeding_first_order_recurrence(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; CHECK: vector.scevcheck:
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[N]], -1
; CHECK-NEXT: [[TMP1:%.*]] = trunc i32 [[STEP:%.*]] to i8
; CHECK-NEXT: [[TMP2:%.*]] = sub i8 0, [[TMP1]]
; CHECK-NEXT: [[TMP3:%.*]] = icmp slt i8 [[TMP1]], 0
; CHECK-NEXT: [[TMP4:%.*]] = select i1 [[TMP3]], i8 [[TMP2]], i8 [[TMP1]]
; CHECK-NEXT: [[TMP5:%.*]] = trunc i64 [[TMP0]] to i8
; CHECK-NEXT: [[MUL:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 [[TMP4]], i8 [[TMP5]])
; CHECK-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i8, i1 } [[MUL]], 0
; CHECK-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i8, i1 } [[MUL]], 1
; CHECK-NEXT: [[TMP6:%.*]] = sub i8 0, [[MUL_RESULT]]
; CHECK-NEXT: [[TMP7:%.*]] = icmp slt i8 [[MUL_RESULT]], 0
; CHECK-NEXT: [[TMP8:%.*]] = icmp sgt i8 [[TMP6]], 0
; CHECK-NEXT: [[TMP9:%.*]] = select i1 [[TMP3]], i1 [[TMP8]], i1 [[TMP7]]
; CHECK-NEXT: [[TMP10:%.*]] = or i1 [[TMP9]], [[MUL_OVERFLOW]]
; CHECK-NEXT: [[TMP11:%.*]] = icmp ugt i64 [[TMP0]], 255
; CHECK-NEXT: [[TMP12:%.*]] = icmp ne i8 [[TMP1]], 0
; CHECK-NEXT: [[TMP13:%.*]] = and i1 [[TMP11]], [[TMP12]]
; CHECK-NEXT: [[TMP14:%.*]] = or i1 [[TMP10]], [[TMP13]]
; CHECK-NEXT: [[TMP15:%.*]] = sext i8 [[TMP1]] to i32
; CHECK-NEXT: [[IDENT_CHECK:%.*]] = icmp ne i32 [[STEP]], [[TMP15]]
; CHECK-NEXT: [[TMP16:%.*]] = or i1 [[TMP14]], [[IDENT_CHECK]]
; CHECK-NEXT: br i1 [[TMP16]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; CHECK-NEXT: [[CAST_VTC:%.*]] = trunc i64 [[N_VEC]] to i32
; CHECK-NEXT: [[IND_END:%.*]] = mul i32 [[CAST_VTC]], [[STEP]]
; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[STEP]], i32 0
; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP17:%.*]] = mul <2 x i32> <i32 0, i32 1>, [[DOTSPLAT]]
; CHECK-NEXT: [[INDUCTION:%.*]] = add <2 x i32> zeroinitializer, [[TMP17]]
; CHECK-NEXT: [[TMP18:%.*]] = mul i32 [[STEP]], 2
; CHECK-NEXT: [[DOTSPLATINSERT2:%.*]] = insertelement <2 x i32> poison, i32 [[TMP18]], i32 0
; CHECK-NEXT: [[DOTSPLAT3:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT2]], <2 x i32> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[VEC_IND:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_IND]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP19:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT: [[TMP20:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>
; CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[PTR:%.*]], i64 [[TMP19]]
; CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds i32, i32* [[TMP21]], i32 0
; CHECK-NEXT: [[TMP23:%.*]] = bitcast i32* [[TMP22]] to <2 x i32>*
; CHECK-NEXT: store <2 x i32> [[TMP20]], <2 x i32>* [[TMP23]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], [[DOTSPLAT3]]
; CHECK-NEXT: [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP54:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; CHECK-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[VEC_IND]], i32 1
; CHECK-NEXT: [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <2 x i32> [[VEC_IND]], i32 0
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; CHECK-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_2_CONV:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[IV_2_EXT:%.*]] = shl i32 [[IV_2]], 24
; CHECK-NEXT: [[IV_2_CONV]] = ashr exact i32 [[IV_2_EXT]], 24
; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, i32* [[PTR]], i64 [[IV_1]]
; CHECK-NEXT: store i32 [[SCALAR_RECUR]], i32* [[GEP]], align 4
; CHECK-NEXT: [[IV_2_NEXT]] = add nsw i32 [[IV_2_CONV]], [[STEP]]
; CHECK-NEXT: [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP55:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
; IND-LABEL: @test_optimized_cast_induction_feeding_first_order_recurrence(
; IND-NEXT: entry:
; IND-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2
; IND-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; IND: vector.scevcheck:
; IND-NEXT: [[TMP0:%.*]] = add i64 [[N]], -1
; IND-NEXT: [[TMP1:%.*]] = trunc i32 [[STEP:%.*]] to i8
; IND-NEXT: [[TMP2:%.*]] = icmp slt i8 [[TMP1]], 0
; IND-NEXT: [[TMP3:%.*]] = call i8 @llvm.abs.i8(i8 [[TMP1]], i1 false)
; IND-NEXT: [[TMP4:%.*]] = trunc i64 [[TMP0]] to i8
; IND-NEXT: [[MUL:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 [[TMP3]], i8 [[TMP4]])
; IND-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i8, i1 } [[MUL]], 0
; IND-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i8, i1 } [[MUL]], 1
; IND-NEXT: [[TMP5:%.*]] = icmp slt i8 [[MUL_RESULT]], 0
; IND-NEXT: [[TMP6:%.*]] = icmp ugt i8 [[MUL_RESULT]], -128
; IND-NEXT: [[TMP7:%.*]] = select i1 [[TMP2]], i1 [[TMP6]], i1 [[TMP5]]
; IND-NEXT: [[TMP8:%.*]] = or i1 [[TMP7]], [[MUL_OVERFLOW]]
; IND-NEXT: [[TMP9:%.*]] = icmp ugt i64 [[TMP0]], 255
; IND-NEXT: [[TMP10:%.*]] = icmp ne i8 [[TMP1]], 0
; IND-NEXT: [[TMP11:%.*]] = and i1 [[TMP9]], [[TMP10]]
; IND-NEXT: [[TMP12:%.*]] = or i1 [[TMP8]], [[TMP11]]
; IND-NEXT: [[TMP13:%.*]] = add i32 [[STEP]], -128
; IND-NEXT: [[TMP14:%.*]] = icmp ult i32 [[TMP13]], -256
; IND-NEXT: [[TMP15:%.*]] = or i1 [[TMP12]], [[TMP14]]
; IND-NEXT: br i1 [[TMP15]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; IND: vector.ph:
; IND-NEXT: [[N_VEC:%.*]] = and i64 [[N]], -2
; IND-NEXT: [[CAST_VTC:%.*]] = trunc i64 [[N_VEC]] to i32
; IND-NEXT: [[IND_END:%.*]] = mul i32 [[CAST_VTC]], [[STEP]]
; IND-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[STEP]], i64 0
; IND-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; IND-NEXT: [[TMP16:%.*]] = mul nuw <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
; IND-NEXT: [[TMP17:%.*]] = shl i32 [[STEP]], 1
; IND-NEXT: [[DOTSPLATINSERT2:%.*]] = insertelement <2 x i32> poison, i32 [[TMP17]], i64 0
; IND-NEXT: [[DOTSPLAT3:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT2]], <2 x i32> poison, <2 x i32> zeroinitializer
; IND-NEXT: br label [[VECTOR_BODY:%.*]]
; IND: vector.body:
; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[VEC_IND:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[VEC_IND]] = phi <2 x i32> [ [[TMP16]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; IND-NEXT: [[TMP18:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>
; IND-NEXT: [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[PTR:%.*]], i64 [[INDEX]]
; IND-NEXT: [[TMP20:%.*]] = bitcast i32* [[TMP19]] to <2 x i32>*
; IND-NEXT: store <2 x i32> [[TMP18]], <2 x i32>* [[TMP20]], align 4
; IND-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; IND-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], [[DOTSPLAT3]]
; IND-NEXT: [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; IND-NEXT: br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP54:![0-9]+]]
; IND: middle.block:
; IND-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
; IND-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[VEC_IND]], i64 1
; IND-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; IND: scalar.ph:
; IND-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; IND-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
; IND-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY]] ], [ [[IND_END]], [[MIDDLE_BLOCK]] ]
; IND-NEXT: br label [[LOOP:%.*]]
; IND: loop:
; IND-NEXT: [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_2_CONV:%.*]], [[LOOP]] ]
; IND-NEXT: [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP]] ]
; IND-NEXT: [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], [[LOOP]] ]
; IND-NEXT: [[IV_2_EXT:%.*]] = shl i32 [[IV_2]], 24
; IND-NEXT: [[IV_2_CONV]] = ashr exact i32 [[IV_2_EXT]], 24
; IND-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, i32* [[PTR]], i64 [[IV_1]]
; IND-NEXT: store i32 [[SCALAR_RECUR]], i32* [[GEP]], align 4
; IND-NEXT: [[IV_2_NEXT]] = add nsw i32 [[IV_2_CONV]], [[STEP]]
; IND-NEXT: [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 1
; IND-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]
; IND-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP55:![0-9]+]]
; IND: exit:
; IND-NEXT: ret void
;
; UNROLL-LABEL: @test_optimized_cast_induction_feeding_first_order_recurrence(
; UNROLL-NEXT: entry:
; UNROLL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
; UNROLL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; UNROLL: vector.scevcheck:
; UNROLL-NEXT: [[TMP0:%.*]] = add i64 [[N]], -1
; UNROLL-NEXT: [[TMP1:%.*]] = trunc i32 [[STEP:%.*]] to i8
; UNROLL-NEXT: [[TMP2:%.*]] = icmp slt i8 [[TMP1]], 0
; UNROLL-NEXT: [[TMP3:%.*]] = call i8 @llvm.abs.i8(i8 [[TMP1]], i1 false)
; UNROLL-NEXT: [[TMP4:%.*]] = trunc i64 [[TMP0]] to i8
; UNROLL-NEXT: [[MUL:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 [[TMP3]], i8 [[TMP4]])
; UNROLL-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i8, i1 } [[MUL]], 0
; UNROLL-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i8, i1 } [[MUL]], 1
; UNROLL-NEXT: [[TMP5:%.*]] = icmp slt i8 [[MUL_RESULT]], 0
; UNROLL-NEXT: [[TMP6:%.*]] = icmp ugt i8 [[MUL_RESULT]], -128
; UNROLL-NEXT: [[TMP7:%.*]] = select i1 [[TMP2]], i1 [[TMP6]], i1 [[TMP5]]
; UNROLL-NEXT: [[TMP8:%.*]] = or i1 [[TMP7]], [[MUL_OVERFLOW]]
; UNROLL-NEXT: [[TMP9:%.*]] = icmp ugt i64 [[TMP0]], 255
; UNROLL-NEXT: [[TMP10:%.*]] = icmp ne i8 [[TMP1]], 0
; UNROLL-NEXT: [[TMP11:%.*]] = and i1 [[TMP9]], [[TMP10]]
; UNROLL-NEXT: [[TMP12:%.*]] = or i1 [[TMP8]], [[TMP11]]
; UNROLL-NEXT: [[TMP13:%.*]] = add i32 [[STEP]], -128
; UNROLL-NEXT: [[TMP14:%.*]] = icmp ult i32 [[TMP13]], -256
; UNROLL-NEXT: [[TMP15:%.*]] = or i1 [[TMP12]], [[TMP14]]
; UNROLL-NEXT: br i1 [[TMP15]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; UNROLL: vector.ph:
; UNROLL-NEXT: [[N_VEC:%.*]] = and i64 [[N]], -4
; UNROLL-NEXT: [[CAST_VTC:%.*]] = trunc i64 [[N_VEC]] to i32
; UNROLL-NEXT: [[IND_END:%.*]] = mul i32 [[CAST_VTC]], [[STEP]]
; UNROLL-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[STEP]], i64 0
; UNROLL-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NEXT: [[TMP16:%.*]] = mul nuw <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
; UNROLL-NEXT: [[TMP17:%.*]] = shl i32 [[STEP]], 1
; UNROLL-NEXT: [[DOTSPLATINSERT2:%.*]] = insertelement <2 x i32> poison, i32 [[TMP17]], i64 0
; UNROLL-NEXT: [[DOTSPLAT3:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT2]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL: vector.body:
; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[TMP16]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NEXT: [[STEP_ADD]] = add <2 x i32> [[VEC_IND]], [[DOTSPLAT3]]
; UNROLL-NEXT: [[TMP18:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>
; UNROLL-NEXT: [[TMP19:%.*]] = shufflevector <2 x i32> [[VEC_IND]], <2 x i32> [[STEP_ADD]], <2 x i32> <i32 1, i32 2>
; UNROLL-NEXT: [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[PTR:%.*]], i64 [[INDEX]]
; UNROLL-NEXT: [[TMP21:%.*]] = bitcast i32* [[TMP20]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[TMP18]], <2 x i32>* [[TMP21]], align 4
; UNROLL-NEXT: [[TMP22:%.*]] = getelementptr inbounds i32, i32* [[TMP20]], i64 2
; UNROLL-NEXT: [[TMP23:%.*]] = bitcast i32* [[TMP22]] to <2 x i32>*
; UNROLL-NEXT: store <2 x i32> [[TMP19]], <2 x i32>* [[TMP23]], align 4
; UNROLL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], [[DOTSPLAT3]]
; UNROLL-NEXT: [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NEXT: br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP54:![0-9]+]]
; UNROLL: middle.block:
; UNROLL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
; UNROLL-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[STEP_ADD]], i64 1
; UNROLL-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL: scalar.ph:
; UNROLL-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
; UNROLL-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY]] ], [ [[IND_END]], [[MIDDLE_BLOCK]] ]
; UNROLL-NEXT: br label [[LOOP:%.*]]
; UNROLL: loop:
; UNROLL-NEXT: [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_2_CONV:%.*]], [[LOOP]] ]
; UNROLL-NEXT: [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP]] ]
; UNROLL-NEXT: [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], [[LOOP]] ]
; UNROLL-NEXT: [[IV_2_EXT:%.*]] = shl i32 [[IV_2]], 24
; UNROLL-NEXT: [[IV_2_CONV]] = ashr exact i32 [[IV_2_EXT]], 24
; UNROLL-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, i32* [[PTR]], i64 [[IV_1]]
; UNROLL-NEXT: store i32 [[SCALAR_RECUR]], i32* [[GEP]], align 4
; UNROLL-NEXT: [[IV_2_NEXT]] = add nsw i32 [[IV_2_CONV]], [[STEP]]
; UNROLL-NEXT: [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 1
; UNROLL-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]
; UNROLL-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP55:![0-9]+]]
; UNROLL: exit:
; UNROLL-NEXT: ret void
;
; UNROLL-NO-IC-LABEL: @test_optimized_cast_induction_feeding_first_order_recurrence(
; UNROLL-NO-IC-NEXT: entry:
; UNROLL-NO-IC-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
; UNROLL-NO-IC-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; UNROLL-NO-IC: vector.scevcheck:
; UNROLL-NO-IC-NEXT: [[TMP0:%.*]] = add i64 [[N]], -1
; UNROLL-NO-IC-NEXT: [[TMP1:%.*]] = trunc i32 [[STEP:%.*]] to i8
; UNROLL-NO-IC-NEXT: [[TMP2:%.*]] = sub i8 0, [[TMP1]]
; UNROLL-NO-IC-NEXT: [[TMP3:%.*]] = icmp slt i8 [[TMP1]], 0
; UNROLL-NO-IC-NEXT: [[TMP4:%.*]] = select i1 [[TMP3]], i8 [[TMP2]], i8 [[TMP1]]
; UNROLL-NO-IC-NEXT: [[TMP5:%.*]] = trunc i64 [[TMP0]] to i8
; UNROLL-NO-IC-NEXT: [[MUL:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 [[TMP4]], i8 [[TMP5]])
; UNROLL-NO-IC-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i8, i1 } [[MUL]], 0
; UNROLL-NO-IC-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i8, i1 } [[MUL]], 1
; UNROLL-NO-IC-NEXT: [[TMP6:%.*]] = sub i8 0, [[MUL_RESULT]]
; UNROLL-NO-IC-NEXT: [[TMP7:%.*]] = icmp slt i8 [[MUL_RESULT]], 0
; UNROLL-NO-IC-NEXT: [[TMP8:%.*]] = icmp sgt i8 [[TMP6]], 0
; UNROLL-NO-IC-NEXT: [[TMP9:%.*]] = select i1 [[TMP3]], i1 [[TMP8]], i1 [[TMP7]]
; UNROLL-NO-IC-NEXT: [[TMP10:%.*]] = or i1 [[TMP9]], [[MUL_OVERFLOW]]
; UNROLL-NO-IC-NEXT: [[TMP11:%.*]] = icmp ugt i64 [[TMP0]], 255
; UNROLL-NO-IC-NEXT: [[TMP12:%.*]] = icmp ne i8 [[TMP1]], 0
; UNROLL-NO-IC-NEXT: [[TMP13:%.*]] = and i1 [[TMP11]], [[TMP12]]
; UNROLL-NO-IC-NEXT: [[TMP14:%.*]] = or i1 [[TMP10]], [[TMP13]]
; UNROLL-NO-IC-NEXT: [[TMP15:%.*]] = sext i8 [[TMP1]] to i32
; UNROLL-NO-IC-NEXT: [[IDENT_CHECK:%.*]] = icmp ne i32 [[STEP]], [[TMP15]]
; UNROLL-NO-IC-NEXT: [[TMP16:%.*]] = or i1 [[TMP14]], [[IDENT_CHECK]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP16]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC: vector.ph:
; UNROLL-NO-IC-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
; UNROLL-NO-IC-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT: [[CAST_VTC:%.*]] = trunc i64 [[N_VEC]] to i32
; UNROLL-NO-IC-NEXT: [[IND_END:%.*]] = mul i32 [[CAST_VTC]], [[STEP]]
; UNROLL-NO-IC-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[STEP]], i32 0
; UNROLL-NO-IC-NEXT: [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT: [[TMP17:%.*]] = mul <2 x i32> <i32 0, i32 1>, [[DOTSPLAT]]
; UNROLL-NO-IC-NEXT: [[INDUCTION:%.*]] = add <2 x i32> zeroinitializer, [[TMP17]]
; UNROLL-NO-IC-NEXT: [[TMP18:%.*]] = mul i32 [[STEP]], 2
; UNROLL-NO-IC-NEXT: [[DOTSPLATINSERT2:%.*]] = insertelement <2 x i32> poison, i32 [[TMP18]], i32 0
; UNROLL-NO-IC-NEXT: [[DOTSPLAT3:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT2]], <2 x i32> poison, <2 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT: br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC: vector.body:
; UNROLL-NO-IC-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i32> [ <i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT: [[STEP_ADD]] = add <2 x i32> [[VEC_IND]], [[DOTSPLAT3]]
; UNROLL-NO-IC-NEXT: [[TMP19:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-IC-NEXT: [[TMP20:%.*]] = add i64 [[INDEX]], 2
; UNROLL-NO-IC-NEXT: [[TMP21:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP22:%.*]] = shufflevector <2 x i32> [[VEC_IND]], <2 x i32> [[STEP_ADD]], <2 x i32> <i32 1, i32 2>
; UNROLL-NO-IC-NEXT: [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[PTR:%.*]], i64 [[TMP19]]
; UNROLL-NO-IC-NEXT: [[TMP24:%.*]] = getelementptr inbounds i32, i32* [[PTR]], i64 [[TMP20]]
; UNROLL-NO-IC-NEXT: [[TMP25:%.*]] = getelementptr inbounds i32, i32* [[TMP23]], i32 0
; UNROLL-NO-IC-NEXT: [[TMP26:%.*]] = bitcast i32* [[TMP25]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[TMP21]], <2 x i32>* [[TMP26]], align 4
; UNROLL-NO-IC-NEXT: [[TMP27:%.*]] = getelementptr inbounds i32, i32* [[TMP23]], i32 2
; UNROLL-NO-IC-NEXT: [[TMP28:%.*]] = bitcast i32* [[TMP27]] to <2 x i32>*
; UNROLL-NO-IC-NEXT: store <2 x i32> [[TMP22]], <2 x i32>* [[TMP28]], align 4
; UNROLL-NO-IC-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT: [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], [[DOTSPLAT3]]
; UNROLL-NO-IC-NEXT: [[TMP29:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: br i1 [[TMP29]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP54:![0-9]+]]
; UNROLL-NO-IC: middle.block:
; UNROLL-NO-IC-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; UNROLL-NO-IC-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[STEP_ADD]], i32 1
; UNROLL-NO-IC-NEXT: [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <2 x i32> [[STEP_ADD]], i32 0
; UNROLL-NO-IC-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC: scalar.ph:
; UNROLL-NO-IC-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; UNROLL-NO-IC-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
; UNROLL-NO-IC-NEXT: br label [[LOOP:%.*]]
; UNROLL-NO-IC: loop:
; UNROLL-NO-IC-NEXT: [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_2_CONV:%.*]], [[LOOP]] ]
; UNROLL-NO-IC-NEXT: [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP]] ]
; UNROLL-NO-IC-NEXT: [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], [[LOOP]] ]
; UNROLL-NO-IC-NEXT: [[IV_2_EXT:%.*]] = shl i32 [[IV_2]], 24
; UNROLL-NO-IC-NEXT: [[IV_2_CONV]] = ashr exact i32 [[IV_2_EXT]], 24
; UNROLL-NO-IC-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, i32* [[PTR]], i64 [[IV_1]]
; UNROLL-NO-IC-NEXT: store i32 [[SCALAR_RECUR]], i32* [[GEP]], align 4
; UNROLL-NO-IC-NEXT: [[IV_2_NEXT]] = add nsw i32 [[IV_2_CONV]], [[STEP]]
; UNROLL-NO-IC-NEXT: [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 1
; UNROLL-NO-IC-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]
; UNROLL-NO-IC-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP55:![0-9]+]]
; UNROLL-NO-IC: exit:
; UNROLL-NO-IC-NEXT: ret void
;
; INTERLEAVE-LABEL: @test_optimized_cast_induction_feeding_first_order_recurrence(
; INTERLEAVE-NEXT: entry:
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; INTERLEAVE: vector.scevcheck:
; INTERLEAVE-NEXT: [[TMP0:%.*]] = add i64 [[N]], -1
; INTERLEAVE-NEXT: [[TMP1:%.*]] = trunc i32 [[STEP:%.*]] to i8
; INTERLEAVE-NEXT: [[TMP2:%.*]] = icmp slt i8 [[TMP1]], 0
; INTERLEAVE-NEXT: [[TMP3:%.*]] = call i8 @llvm.abs.i8(i8 [[TMP1]], i1 false)
; INTERLEAVE-NEXT: [[TMP4:%.*]] = trunc i64 [[TMP0]] to i8
; INTERLEAVE-NEXT: [[MUL:%.*]] = call { i8, i1 } @llvm.umul.with.overflow.i8(i8 [[TMP3]], i8 [[TMP4]])
; INTERLEAVE-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i8, i1 } [[MUL]], 0
; INTERLEAVE-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i8, i1 } [[MUL]], 1
; INTERLEAVE-NEXT: [[TMP5:%.*]] = icmp slt i8 [[MUL_RESULT]], 0
; INTERLEAVE-NEXT: [[TMP6:%.*]] = icmp ugt i8 [[MUL_RESULT]], -128
; INTERLEAVE-NEXT: [[TMP7:%.*]] = select i1 [[TMP2]], i1 [[TMP6]], i1 [[TMP5]]
; INTERLEAVE-NEXT: [[TMP8:%.*]] = or i1 [[TMP7]], [[MUL_OVERFLOW]]
; INTERLEAVE-NEXT: [[TMP9:%.*]] = icmp ugt i64 [[TMP0]], 255
; INTERLEAVE-NEXT: [[TMP10:%.*]] = icmp ne i8 [[TMP1]], 0
; INTERLEAVE-NEXT: [[TMP11:%.*]] = and i1 [[TMP9]], [[TMP10]]
; INTERLEAVE-NEXT: [[TMP12:%.*]] = or i1 [[TMP8]], [[TMP11]]
; INTERLEAVE-NEXT: [[TMP13:%.*]] = add i32 [[STEP]], -128
; INTERLEAVE-NEXT: [[TMP14:%.*]] = icmp ult i32 [[TMP13]], -256
; INTERLEAVE-NEXT: [[TMP15:%.*]] = or i1 [[TMP12]], [[TMP14]]
; INTERLEAVE-NEXT: br i1 [[TMP15]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; INTERLEAVE: vector.ph:
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = and i64 [[N]], -8
; INTERLEAVE-NEXT: [[CAST_VTC:%.*]] = trunc i64 [[N_VEC]] to i32
; INTERLEAVE-NEXT: [[IND_END:%.*]] = mul i32 [[CAST_VTC]], [[STEP]]
; INTERLEAVE-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[STEP]], i64 0
; INTERLEAVE-NEXT: [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; INTERLEAVE-NEXT: [[TMP16:%.*]] = mul <4 x i32> [[DOTSPLAT]], <i32 0, i32 1, i32 2, i32 3>
; INTERLEAVE-NEXT: [[TMP17:%.*]] = shl i32 [[STEP]], 2
; INTERLEAVE-NEXT: [[DOTSPLATINSERT2:%.*]] = insertelement <4 x i32> poison, i32 [[TMP17]], i64 0
; INTERLEAVE-NEXT: [[DOTSPLAT3:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT2]], <4 x i32> poison, <4 x i32> zeroinitializer
; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]]
; INTERLEAVE: vector.body:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[STEP_ADD:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ [[TMP16]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[STEP_ADD]] = add <4 x i32> [[VEC_IND]], [[DOTSPLAT3]]
; 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>
; 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>
; INTERLEAVE-NEXT: [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[PTR:%.*]], i64 [[INDEX]]
; INTERLEAVE-NEXT: [[TMP21:%.*]] = bitcast i32* [[TMP20]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[TMP18]], <4 x i32>* [[TMP21]], align 4
; INTERLEAVE-NEXT: [[TMP22:%.*]] = getelementptr inbounds i32, i32* [[TMP20]], i64 4
; INTERLEAVE-NEXT: [[TMP23:%.*]] = bitcast i32* [[TMP22]] to <4 x i32>*
; INTERLEAVE-NEXT: store <4 x i32> [[TMP19]], <4 x i32>* [[TMP23]], align 4
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[STEP_ADD]], [[DOTSPLAT3]]
; INTERLEAVE-NEXT: [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP54:![0-9]+]]
; INTERLEAVE: middle.block:
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]]
; INTERLEAVE-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[STEP_ADD]], i64 3
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; INTERLEAVE: scalar.ph:
; INTERLEAVE-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY]] ], [ [[N_VEC]], [[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i32 [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ENTRY]] ], [ [[IND_END]], [[MIDDLE_BLOCK]] ]
; INTERLEAVE-NEXT: br label [[LOOP:%.*]]
; INTERLEAVE: loop:
; INTERLEAVE-NEXT: [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[IV_2_CONV:%.*]], [[LOOP]] ]
; INTERLEAVE-NEXT: [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP]] ]
; INTERLEAVE-NEXT: [[IV_2:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], [[LOOP]] ]
; INTERLEAVE-NEXT: [[IV_2_EXT:%.*]] = shl i32 [[IV_2]], 24
; INTERLEAVE-NEXT: [[IV_2_CONV]] = ashr exact i32 [[IV_2_EXT]], 24
; INTERLEAVE-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, i32* [[PTR]], i64 [[IV_1]]
; INTERLEAVE-NEXT: store i32 [[SCALAR_RECUR]], i32* [[GEP]], align 4
; INTERLEAVE-NEXT: [[IV_2_NEXT]] = add nsw i32 [[IV_2_CONV]], [[STEP]]
; INTERLEAVE-NEXT: [[IV_1_NEXT]] = add nuw nsw i64 [[IV_1]], 1
; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_1_NEXT]], [[N]]
; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP55:![0-9]+]]
; INTERLEAVE: exit:
; INTERLEAVE-NEXT: ret void
;
entry:
br label %loop
loop:
%for = phi i32 [ 0, %entry ], [ %iv.2.conv, %loop ]
%iv.1 = phi i64 [ 0, %entry ], [ %iv.1.next, %loop ]
%iv.2 = phi i32 [ 0, %entry ], [ %iv.2.next, %loop ]
%iv.2.ext = shl i32 %iv.2, 24
%iv.2.conv = ashr exact i32 %iv.2.ext, 24
%gep = getelementptr inbounds i32, i32* %ptr, i64 %iv.1
store i32 %for, i32* %gep, align 4
%iv.2.next = add nsw i32 %iv.2.conv, %step
%iv.1.next = add nuw nsw i64 %iv.1, 1
%exitcond = icmp eq i64 %iv.1.next, %n
br i1 %exitcond, label %exit, label %loop
exit:
ret void
}