x264 vectorization

get_ref.print-after.loop-vectorize.ll

; *** IR Dump After LoopVectorizePass on get_ref ***
; Function Attrs: nofree norecurse nosync nounwind memory(readwrite, inaccessiblemem: none) uwtable
define internal ptr @get_ref(ptr noundef %dst, ptr nocapture noundef %i_dst_stride, ptr nocapture noundef readonly %src, i32 noundef %i_src_stride, i32 noundef %mvx, i32 noundef %mvy, i32 noundef %i_width, i32 noundef %i_height, ptr nocapture noundef readonly %weight) #0 {
entry:
  %and = and i32 %mvy, 3
  %shl = shl nuw nsw i32 %and, 2
  %and1 = and i32 %mvx, 3
  %add = or disjoint i32 %shl, %and1
  %shr = ashr i32 %mvy, 2
  %mul = mul i32 %shr, %i_src_stride
  %shr2 = ashr i32 %mvx, 2
  %add3 = add i32 %mul, %shr2
  %idxprom = zext nneg i32 %add to i64
  %arrayidx = getelementptr inbounds [16 x i8], ptr @hpel_ref0, i64 0, i64 %idxprom
  %0 = load i8, ptr %arrayidx, align 1, !tbaa !15
  %idxprom4 = zext i8 %0 to i64
  %arrayidx5 = getelementptr inbounds ptr, ptr %src, i64 %idxprom4
  %1 = load ptr, ptr %arrayidx5, align 8, !tbaa !9
  %idx.ext = sext i32 %add3 to i64
  %add.ptr = getelementptr i8, ptr %1, i64 %idx.ext
  %cmp = icmp eq i32 %and, 3
  %mul7 = select i1 %cmp, i32 %i_src_stride, i32 0
  %idx.ext8 = sext i32 %mul7 to i64
  %add.ptr9 = getelementptr i8, ptr %add.ptr, i64 %idx.ext8
  %and10 = and i32 %add, 5
  %tobool.not = icmp eq i32 %and10, 0
  br i1 %tobool.not, label %if.else, label %if.then

if.then:                                          ; preds = %entry
  %arrayidx12 = getelementptr inbounds [16 x i8], ptr @hpel_ref1, i64 0, i64 %idxprom
  %2 = load i8, ptr %arrayidx12, align 1, !tbaa !15
  %idxprom13 = zext i8 %2 to i64
  %arrayidx14 = getelementptr inbounds ptr, ptr %src, i64 %idxprom13
  %3 = load ptr, ptr %arrayidx14, align 8, !tbaa !9
  %add.ptr16 = getelementptr i8, ptr %3, i64 %idx.ext
  %cmp18 = icmp eq i32 %and1, 3
  %idx.ext20 = zext i1 %cmp18 to i64
  %add.ptr21 = getelementptr i8, ptr %add.ptr16, i64 %idx.ext20
  %cmp29.i = icmp sgt i32 %i_height, 0
  br i1 %cmp29.i, label %for.cond1.preheader.lr.ph.i, label %pixel_avg.exit

for.cond1.preheader.lr.ph.i:                      ; preds = %if.then
  %4 = load i32, ptr %i_dst_stride, align 4, !tbaa !13
  %cmp227.i = icmp sgt i32 %i_width, 0
  %idx.ext.i = sext i32 %4 to i64
  %idx.ext12.i = sext i32 %i_src_stride to i64
  br i1 %cmp227.i, label %for.cond1.preheader.us.preheader.i, label %pixel_avg.exit

for.cond1.preheader.us.preheader.i:               ; preds = %for.cond1.preheader.lr.ph.i
  %wide.trip.count.i = zext nneg i32 %i_width to i64
  %5 = add i32 %i_height, -1
  %6 = zext i32 %5 to i64
  %7 = mul i64 %idx.ext.i, %6
  %8 = add i64 %7, %wide.trip.count.i
  %scevgep = getelementptr i8, ptr %dst, i64 %8
  %9 = mul i64 %idx.ext12.i, %6
  %10 = add i64 %9, %idx.ext8
  %11 = add i64 %10, %idx.ext
  %12 = add i64 %11, %wide.trip.count.i
  %scevgep154 = getelementptr i8, ptr %1, i64 %12
  %13 = add i64 %9, %idx.ext
  %14 = add i64 %13, %idx.ext20
  %15 = add i64 %14, %wide.trip.count.i
  %scevgep155 = getelementptr i8, ptr %3, i64 %15
  br label %iter.check

iter.check:                                       ; preds = %for.cond1.for.cond.cleanup3_crit_edge.us.i, %for.cond1.preheader.us.preheader.i
  %y.033.us.i = phi i32 [ %inc17.us.i, %for.cond1.for.cond.cleanup3_crit_edge.us.i ], [ 0, %for.cond1.preheader.us.preheader.i ]
  %dst.addr.032.us.i = phi ptr [ %add.ptr.us.i, %for.cond1.for.cond.cleanup3_crit_edge.us.i ], [ %dst, %for.cond1.preheader.us.preheader.i ]
  %src1.addr.031.us.i = phi ptr [ %add.ptr13.us.i, %for.cond1.for.cond.cleanup3_crit_edge.us.i ], [ %add.ptr9, %for.cond1.preheader.us.preheader.i ]
  %src2.addr.030.us.i = phi ptr [ %add.ptr15.us.i, %for.cond1.for.cond.cleanup3_crit_edge.us.i ], [ %add.ptr21, %for.cond1.preheader.us.preheader.i ]
  %min.iters.check = icmp ult i64 %wide.trip.count.i, 8
  br i1 %min.iters.check, label %vec.epilog.scalar.ph, label %vector.memcheck

vector.memcheck:                                  ; preds = %iter.check
  %bound0 = icmp ult ptr %dst, %scevgep154
  %bound1 = icmp ult ptr %add.ptr9, %scevgep
  %found.conflict = and i1 %bound0, %bound1
  %stride.check = icmp slt i64 %idx.ext.i, 0
  %16 = or i1 %found.conflict, %stride.check
  %stride.check156 = icmp slt i64 %idx.ext12.i, 0
  %17 = or i1 %16, %stride.check156
  %bound0157 = icmp ult ptr %dst, %scevgep155
  %bound1158 = icmp ult ptr %add.ptr21, %scevgep
  %found.conflict159 = and i1 %bound0157, %bound1158
  %stride.check160 = icmp slt i64 %idx.ext.i, 0
  %18 = or i1 %found.conflict159, %stride.check160
  %stride.check161 = icmp slt i64 %idx.ext12.i, 0
  %19 = or i1 %18, %stride.check161
  %conflict.rdx = or i1 %17, %19
  br i1 %conflict.rdx, label %vec.epilog.scalar.ph, label %vector.main.loop.iter.check

vector.main.loop.iter.check:                      ; preds = %vector.memcheck
  %min.iters.check162 = icmp ult i64 %wide.trip.count.i, 32
  br i1 %min.iters.check162, label %vec.epilog.ph, label %vector.ph

vector.ph:                                        ; preds = %vector.main.loop.iter.check
  %n.mod.vf = urem i64 %wide.trip.count.i, 32
  %n.vec = sub i64 %wide.trip.count.i, %n.mod.vf
  br label %vector.body

vector.body:                                      ; preds = %vector.body, %vector.ph
  %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
  %20 = add i64 %index, 0
  %21 = add i64 %index, 16
  %22 = getelementptr inbounds i8, ptr %src1.addr.031.us.i, i64 %20
  %23 = getelementptr inbounds i8, ptr %src1.addr.031.us.i, i64 %21
  %24 = getelementptr inbounds i8, ptr %22, i32 0
  %25 = getelementptr inbounds i8, ptr %22, i32 16
  %wide.load = load <16 x i8>, ptr %24, align 1, !tbaa !15, !alias.scope !144
  %wide.load163 = load <16 x i8>, ptr %25, align 1, !tbaa !15, !alias.scope !144
  %26 = zext <16 x i8> %wide.load to <16 x i16>
  %27 = zext <16 x i8> %wide.load163 to <16 x i16>
  %28 = getelementptr inbounds i8, ptr %src2.addr.030.us.i, i64 %20
  %29 = getelementptr inbounds i8, ptr %src2.addr.030.us.i, i64 %21
  %30 = getelementptr inbounds i8, ptr %28, i32 0
  %31 = getelementptr inbounds i8, ptr %28, i32 16
  %wide.load164 = load <16 x i8>, ptr %30, align 1, !tbaa !15, !alias.scope !147
  %wide.load165 = load <16 x i8>, ptr %31, align 1, !tbaa !15, !alias.scope !147
  %32 = zext <16 x i8> %wide.load164 to <16 x i16>
  %33 = zext <16 x i8> %wide.load165 to <16 x i16>
  %34 = add nuw nsw <16 x i16> %26, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
  %35 = add nuw nsw <16 x i16> %27, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
  %36 = add nuw nsw <16 x i16> %34, %32
  %37 = add nuw nsw <16 x i16> %35, %33
  %38 = lshr <16 x i16> %36, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
  %39 = lshr <16 x i16> %37, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
  %40 = trunc <16 x i16> %38 to <16 x i8>
  %41 = trunc <16 x i16> %39 to <16 x i8>
  %42 = getelementptr inbounds i8, ptr %dst.addr.032.us.i, i64 %20
  %43 = getelementptr inbounds i8, ptr %dst.addr.032.us.i, i64 %21
  %44 = getelementptr inbounds i8, ptr %42, i32 0
  %45 = getelementptr inbounds i8, ptr %42, i32 16
  store <16 x i8> %40, ptr %44, align 1, !tbaa !15, !alias.scope !149, !noalias !151
  store <16 x i8> %41, ptr %45, align 1, !tbaa !15, !alias.scope !149, !noalias !151
  %index.next = add nuw i64 %index, 32
  %46 = icmp eq i64 %index.next, %n.vec
  br i1 %46, label %middle.block, label %vector.body, !llvm.loop !152

middle.block:                                     ; preds = %vector.body
  %cmp.n = icmp eq i64 %wide.trip.count.i, %n.vec
  br i1 %cmp.n, label %for.cond1.for.cond.cleanup3_crit_edge.us.i, label %vec.epilog.iter.check

vec.epilog.iter.check:                            ; preds = %middle.block
  %n.vec.remaining = sub i64 %wide.trip.count.i, %n.vec
  %min.epilog.iters.check = icmp ult i64 %n.vec.remaining, 8
  br i1 %min.epilog.iters.check, label %vec.epilog.scalar.ph, label %vec.epilog.ph

vec.epilog.ph:                                    ; preds = %vector.main.loop.iter.check, %vec.epilog.iter.check
  %vec.epilog.resume.val = phi i64 [ %n.vec, %vec.epilog.iter.check ], [ 0, %vector.main.loop.iter.check ]
  %n.mod.vf166 = urem i64 %wide.trip.count.i, 8
  %n.vec167 = sub i64 %wide.trip.count.i, %n.mod.vf166
  br label %vec.epilog.vector.body

vec.epilog.vector.body:                           ; preds = %vec.epilog.vector.body, %vec.epilog.ph
  %index169 = phi i64 [ %vec.epilog.resume.val, %vec.epilog.ph ], [ %index.next172, %vec.epilog.vector.body ]
  %47 = add i64 %index169, 0
  %48 = getelementptr inbounds i8, ptr %src1.addr.031.us.i, i64 %47
  %49 = getelementptr inbounds i8, ptr %48, i32 0
  %wide.load170 = load <8 x i8>, ptr %49, align 1, !tbaa !15, !alias.scope !153
  %50 = zext <8 x i8> %wide.load170 to <8 x i16>
  %51 = getelementptr inbounds i8, ptr %src2.addr.030.us.i, i64 %47
  %52 = getelementptr inbounds i8, ptr %51, i32 0
  %wide.load171 = load <8 x i8>, ptr %52, align 1, !tbaa !15, !alias.scope !156
  %53 = zext <8 x i8> %wide.load171 to <8 x i16>
  %54 = add nuw nsw <8 x i16> %50, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
  %55 = add nuw nsw <8 x i16> %54, %53
  %56 = lshr <8 x i16> %55, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
  %57 = trunc <8 x i16> %56 to <8 x i8>
  %58 = getelementptr inbounds i8, ptr %dst.addr.032.us.i, i64 %47
  %59 = getelementptr inbounds i8, ptr %58, i32 0
  store <8 x i8> %57, ptr %59, align 1, !tbaa !15, !alias.scope !158, !noalias !160
  %index.next172 = add nuw i64 %index169, 8
  %60 = icmp eq i64 %index.next172, %n.vec167
  br i1 %60, label %vec.epilog.middle.block, label %vec.epilog.vector.body, !llvm.loop !161

vec.epilog.middle.block:                          ; preds = %vec.epilog.vector.body
  %cmp.n168 = icmp eq i64 %wide.trip.count.i, %n.vec167
  br i1 %cmp.n168, label %for.cond1.for.cond.cleanup3_crit_edge.us.i, label %vec.epilog.scalar.ph

vec.epilog.scalar.ph:                             ; preds = %vector.memcheck, %iter.check, %vec.epilog.iter.check, %vec.epilog.middle.block
  %bc.resume.val = phi i64 [ %n.vec167, %vec.epilog.middle.block ], [ %n.vec, %vec.epilog.iter.check ], [ 0, %vector.memcheck ], [ 0, %iter.check ]
  br label %for.body4.us.i

for.body4.us.i:                                   ; preds = %for.body4.us.i, %vec.epilog.scalar.ph
  %indvars.iv.i = phi i64 [ %bc.resume.val, %vec.epilog.scalar.ph ], [ %indvars.iv.next.i, %for.body4.us.i ]
  %arrayidx.us.i = getelementptr inbounds i8, ptr %src1.addr.031.us.i, i64 %indvars.iv.i
  %61 = load i8, ptr %arrayidx.us.i, align 1, !tbaa !15
  %conv.us.i = zext i8 %61 to i16
  %arrayidx6.us.i = getelementptr inbounds i8, ptr %src2.addr.030.us.i, i64 %indvars.iv.i
  %62 = load i8, ptr %arrayidx6.us.i, align 1, !tbaa !15
  %conv7.us.i = zext i8 %62 to i16
  %add.us.i = add nuw nsw i16 %conv.us.i, 1
  %add8.us.i = add nuw nsw i16 %add.us.i, %conv7.us.i
  %shr.us.i = lshr i16 %add8.us.i, 1
  %conv9.us.i = trunc i16 %shr.us.i to i8
  %arrayidx11.us.i = getelementptr inbounds i8, ptr %dst.addr.032.us.i, i64 %indvars.iv.i
  store i8 %conv9.us.i, ptr %arrayidx11.us.i, align 1, !tbaa !15
  %indvars.iv.next.i = add nuw nsw i64 %indvars.iv.i, 1
  %exitcond.not.i = icmp eq i64 %indvars.iv.next.i, %wide.trip.count.i
  br i1 %exitcond.not.i, label %for.cond1.for.cond.cleanup3_crit_edge.us.i, label %for.body4.us.i, !llvm.loop !162

for.cond1.for.cond.cleanup3_crit_edge.us.i:       ; preds = %vec.epilog.middle.block, %middle.block, %for.body4.us.i
  %add.ptr.us.i = getelementptr inbounds i8, ptr %dst.addr.032.us.i, i64 %idx.ext.i
  %add.ptr13.us.i = getelementptr inbounds i8, ptr %src1.addr.031.us.i, i64 %idx.ext12.i
  %add.ptr15.us.i = getelementptr inbounds i8, ptr %src2.addr.030.us.i, i64 %idx.ext12.i
  %inc17.us.i = add nuw nsw i32 %y.033.us.i, 1
  %exitcond36.not.i = icmp eq i32 %inc17.us.i, %i_height
  br i1 %exitcond36.not.i, label %pixel_avg.exit.loopexit, label %iter.check, !llvm.loop !100

pixel_avg.exit.loopexit:                          ; preds = %for.cond1.for.cond.cleanup3_crit_edge.us.i
  br label %pixel_avg.exit

pixel_avg.exit:                                   ; preds = %pixel_avg.exit.loopexit, %if.then, %for.cond1.preheader.lr.ph.i
  %weightfn = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 5
  %63 = load ptr, ptr %weightfn, align 16, !tbaa !101
  %tobool22.not = icmp eq ptr %63, null
  br i1 %tobool22.not, label %cleanup, label %if.then23

if.then23:                                        ; preds = %pixel_avg.exit
  %64 = load i32, ptr %i_dst_stride, align 4, !tbaa !13
  %i_denom.i = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 2
  %65 = load i32, ptr %i_denom.i, align 16, !tbaa !103
  %cmp.i = icmp sgt i32 %65, 0
  br i1 %cmp.i, label %for.cond.preheader.i, label %for.cond17.preheader.i

for.cond17.preheader.i:                           ; preds = %if.then23
  br i1 %cmp29.i, label %for.cond23.preheader.lr.ph.i, label %cleanup

for.cond23.preheader.lr.ph.i:                     ; preds = %for.cond17.preheader.i
  %cmp2477.i = icmp sgt i32 %i_width, 0
  %i_scale31.i = getelementptr %struct.x264_weight_t, ptr %weight, i64 0, i32 3
  %i_offset33.i = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 4
  %idx.ext43.i = sext i32 %64 to i64
  br i1 %cmp2477.i, label %for.cond23.preheader.us.preheader.i, label %cleanup

for.cond23.preheader.us.preheader.i:              ; preds = %for.cond23.preheader.lr.ph.i
  %wide.trip.count.i61 = zext nneg i32 %i_width to i64
  %66 = add i32 %i_height, -1
  %67 = zext i32 %66 to i64
  %68 = mul i64 %idx.ext43.i, %67
  %69 = add i64 %68, %wide.trip.count.i61
  %scevgep174 = getelementptr i8, ptr %dst, i64 %69
  %scevgep175 = getelementptr i8, ptr %weight, i64 44
  br label %for.cond23.preheader.us.i

for.cond23.preheader.us.i:                        ; preds = %for.cond23.for.cond.cleanup26_crit_edge.us.i, %for.cond23.preheader.us.preheader.i
  %y16.082.us.i = phi i32 [ %inc42.us.i, %for.cond23.for.cond.cleanup26_crit_edge.us.i ], [ 0, %for.cond23.preheader.us.preheader.i ]
  %dst.addr.181.us.i = phi ptr [ %add.ptr44.us.i, %for.cond23.for.cond.cleanup26_crit_edge.us.i ], [ %dst, %for.cond23.preheader.us.preheader.i ]
  %min.iters.check181 = icmp ult i64 %wide.trip.count.i61, 8
  br i1 %min.iters.check181, label %scalar.ph, label %vector.memcheck173

vector.memcheck173:                               ; preds = %for.cond23.preheader.us.i
  %bound0176 = icmp ult ptr %dst, %scevgep175
  %bound1177 = icmp ult ptr %i_scale31.i, %scevgep174
  %found.conflict178 = and i1 %bound0176, %bound1177
  %stride.check179 = icmp slt i64 %idx.ext43.i, 0
  %70 = or i1 %found.conflict178, %stride.check179
  br i1 %70, label %scalar.ph, label %vector.ph182

vector.ph182:                                     ; preds = %vector.memcheck173
  %n.mod.vf183 = urem i64 %wide.trip.count.i61, 4
  %n.vec184 = sub i64 %wide.trip.count.i61, %n.mod.vf183
  br label %vector.body187

vector.body187:                                   ; preds = %vector.body187, %vector.ph182
  %index188 = phi i64 [ 0, %vector.ph182 ], [ %index.next192, %vector.body187 ]
  %71 = add i64 %index188, 0
  %72 = getelementptr inbounds i8, ptr %dst.addr.181.us.i, i64 %71
  %73 = getelementptr inbounds i8, ptr %72, i32 0
  %wide.load189 = load <4 x i8>, ptr %73, align 1, !tbaa !15, !alias.scope !163, !noalias !166
  %74 = zext <4 x i8> %wide.load189 to <4 x i32>
  %75 = load i32, ptr %i_scale31.i, align 4, !tbaa !104, !alias.scope !166
  %broadcast.splatinsert = insertelement <4 x i32> poison, i32 %75, i64 0
  %broadcast.splat = shufflevector <4 x i32> %broadcast.splatinsert, <4 x i32> poison, <4 x i32> zeroinitializer
  %76 = mul nsw <4 x i32> %broadcast.splat, %74
  %77 = load i32, ptr %i_offset33.i, align 8, !tbaa !108, !alias.scope !166
  %broadcast.splatinsert190 = insertelement <4 x i32> poison, i32 %77, i64 0
  %broadcast.splat191 = shufflevector <4 x i32> %broadcast.splatinsert190, <4 x i32> poison, <4 x i32> zeroinitializer
  %78 = add nsw <4 x i32> %76, %broadcast.splat191
  %79 = icmp ult <4 x i32> %78, <i32 256, i32 256, i32 256, i32 256>
  %80 = icmp sgt <4 x i32> %78, zeroinitializer
  %81 = sext <4 x i1> %80 to <4 x i32>
  %82 = select <4 x i1> %79, <4 x i32> %78, <4 x i32> %81
  %83 = trunc <4 x i32> %82 to <4 x i8>
  store <4 x i8> %83, ptr %73, align 1, !tbaa !15, !alias.scope !163, !noalias !166
  %index.next192 = add nuw i64 %index188, 4
  %84 = icmp eq i64 %index.next192, %n.vec184
  br i1 %84, label %middle.block180, label %vector.body187, !llvm.loop !168

middle.block180:                                  ; preds = %vector.body187
  %cmp.n186 = icmp eq i64 %wide.trip.count.i61, %n.vec184
  br i1 %cmp.n186, label %for.cond23.for.cond.cleanup26_crit_edge.us.i, label %scalar.ph

scalar.ph:                                        ; preds = %vector.memcheck173, %for.cond23.preheader.us.i, %middle.block180
  %bc.resume.val185 = phi i64 [ %n.vec184, %middle.block180 ], [ 0, %for.cond23.preheader.us.i ], [ 0, %vector.memcheck173 ]
  br label %for.body27.us.i

for.body27.us.i:                                  ; preds = %for.body27.us.i, %scalar.ph
  %indvars.iv.i62 = phi i64 [ %bc.resume.val185, %scalar.ph ], [ %indvars.iv.next.i63, %for.body27.us.i ]
  %arrayidx29.us.i = getelementptr inbounds i8, ptr %dst.addr.181.us.i, i64 %indvars.iv.i62
  %85 = load i8, ptr %arrayidx29.us.i, align 1, !tbaa !15
  %conv30.us.i = zext i8 %85 to i32
  %86 = load i32, ptr %i_scale31.i, align 4, !tbaa !104
  %mul32.us.i = mul nsw i32 %86, %conv30.us.i
  %87 = load i32, ptr %i_offset33.i, align 8, !tbaa !108
  %add34.us.i = add nsw i32 %mul32.us.i, %87
  %tobool.not.i72.us.i = icmp ult i32 %add34.us.i, 256
  %88 = icmp sgt i32 %add34.us.i, 0
  %shr.i73.us.i = sext i1 %88 to i32
  %cond.i74.us.i = select i1 %tobool.not.i72.us.i, i32 %add34.us.i, i32 %shr.i73.us.i
  %conv.i75.us.i = trunc i32 %cond.i74.us.i to i8
  store i8 %conv.i75.us.i, ptr %arrayidx29.us.i, align 1, !tbaa !15
  %indvars.iv.next.i63 = add nuw nsw i64 %indvars.iv.i62, 1
  %exitcond.not.i64 = icmp eq i64 %indvars.iv.next.i63, %wide.trip.count.i61
  br i1 %exitcond.not.i64, label %for.cond23.for.cond.cleanup26_crit_edge.us.i, label %for.body27.us.i, !llvm.loop !169

for.cond23.for.cond.cleanup26_crit_edge.us.i:     ; preds = %middle.block180, %for.body27.us.i
  %inc42.us.i = add nuw nsw i32 %y16.082.us.i, 1
  %add.ptr44.us.i = getelementptr i8, ptr %dst.addr.181.us.i, i64 %idx.ext43.i
  %exitcond93.not.i = icmp eq i32 %inc42.us.i, %i_height
  br i1 %exitcond93.not.i, label %cleanup.loopexit153, label %for.cond23.preheader.us.i, !llvm.loop !113

for.cond.preheader.i:                             ; preds = %if.then23
  br i1 %cmp29.i, label %for.cond2.preheader.lr.ph.i, label %cleanup

for.cond2.preheader.lr.ph.i:                      ; preds = %for.cond.preheader.i
  %cmp383.i = icmp sgt i32 %i_width, 0
  %i_scale.i = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 3
  %i_offset.i = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 4
  %idx.ext.i65 = sext i32 %64 to i64
  br i1 %cmp383.i, label %for.cond2.preheader.us.preheader.i, label %cleanup

for.cond2.preheader.us.preheader.i:               ; preds = %for.cond2.preheader.lr.ph.i
  %wide.trip.count97.i = zext nneg i32 %i_width to i64
  %89 = add i32 %i_height, -1
  %90 = zext i32 %89 to i64
  %91 = mul i64 %idx.ext.i65, %90
  %92 = add i64 %91, %wide.trip.count97.i
  %scevgep194 = getelementptr i8, ptr %dst, i64 %92
  %scevgep195 = getelementptr i8, ptr %weight, i64 44
  br label %for.cond2.preheader.us.i

for.cond2.preheader.us.i:                         ; preds = %for.cond2.for.cond.cleanup4_crit_edge.us.i, %for.cond2.preheader.us.preheader.i
  %y.088.us.i = phi i32 [ %inc12.us.i, %for.cond2.for.cond.cleanup4_crit_edge.us.i ], [ 0, %for.cond2.preheader.us.preheader.i ]
  %dst.addr.087.us.i = phi ptr [ %add.ptr.us.i71, %for.cond2.for.cond.cleanup4_crit_edge.us.i ], [ %dst, %for.cond2.preheader.us.preheader.i ]
  %min.iters.check202 = icmp ult i64 %wide.trip.count97.i, 4
  br i1 %min.iters.check202, label %scalar.ph201, label %vector.memcheck193

vector.memcheck193:                               ; preds = %for.cond2.preheader.us.i
  %bound0196 = icmp ult ptr %dst, %scevgep195
  %bound1197 = icmp ult ptr %i_denom.i, %scevgep194
  %found.conflict198 = and i1 %bound0196, %bound1197
  %stride.check199 = icmp slt i64 %idx.ext.i65, 0
  %93 = or i1 %found.conflict198, %stride.check199
  br i1 %93, label %scalar.ph201, label %vector.ph203

vector.ph203:                                     ; preds = %vector.memcheck193
  %n.mod.vf204 = urem i64 %wide.trip.count97.i, 4
  %n.vec205 = sub i64 %wide.trip.count97.i, %n.mod.vf204
  br label %vector.body208

vector.body208:                                   ; preds = %vector.body208, %vector.ph203
  %index209 = phi i64 [ 0, %vector.ph203 ], [ %index.next217, %vector.body208 ]
  %94 = add i64 %index209, 0
  %95 = getelementptr inbounds i8, ptr %dst.addr.087.us.i, i64 %94
  %96 = getelementptr inbounds i8, ptr %95, i32 0
  %wide.load210 = load <4 x i8>, ptr %96, align 1, !tbaa !15, !alias.scope !170, !noalias !173
  %97 = zext <4 x i8> %wide.load210 to <4 x i32>
  %98 = load i32, ptr %i_scale.i, align 4, !tbaa !104, !alias.scope !173
  %broadcast.splatinsert211 = insertelement <4 x i32> poison, i32 %98, i64 0
  %broadcast.splat212 = shufflevector <4 x i32> %broadcast.splatinsert211, <4 x i32> poison, <4 x i32> zeroinitializer
  %99 = mul nsw <4 x i32> %broadcast.splat212, %97
  %100 = load i32, ptr %i_denom.i, align 16, !tbaa !103, !alias.scope !173
  %broadcast.splatinsert213 = insertelement <4 x i32> poison, i32 %100, i64 0
  %broadcast.splat214 = shufflevector <4 x i32> %broadcast.splatinsert213, <4 x i32> poison, <4 x i32> zeroinitializer
  %101 = add nsw <4 x i32> %broadcast.splat214, <i32 -1, i32 -1, i32 -1, i32 -1>
  %102 = shl nuw <4 x i32> <i32 1, i32 1, i32 1, i32 1>, %101
  %103 = add nsw <4 x i32> %102, %99
  %104 = ashr <4 x i32> %103, %broadcast.splat214
  %105 = load i32, ptr %i_offset.i, align 8, !tbaa !108, !alias.scope !173
  %broadcast.splatinsert215 = insertelement <4 x i32> poison, i32 %105, i64 0
  %broadcast.splat216 = shufflevector <4 x i32> %broadcast.splatinsert215, <4 x i32> poison, <4 x i32> zeroinitializer
  %106 = add nsw <4 x i32> %104, %broadcast.splat216
  %107 = icmp ult <4 x i32> %106, <i32 256, i32 256, i32 256, i32 256>
  %108 = icmp sgt <4 x i32> %106, zeroinitializer
  %109 = sext <4 x i1> %108 to <4 x i32>
  %110 = select <4 x i1> %107, <4 x i32> %106, <4 x i32> %109
  %111 = trunc <4 x i32> %110 to <4 x i8>
  store <4 x i8> %111, ptr %96, align 1, !tbaa !15, !alias.scope !170, !noalias !173
  %index.next217 = add nuw i64 %index209, 4
  %112 = icmp eq i64 %index.next217, %n.vec205
  br i1 %112, label %middle.block200, label %vector.body208, !llvm.loop !175

middle.block200:                                  ; preds = %vector.body208
  %cmp.n207 = icmp eq i64 %wide.trip.count97.i, %n.vec205
  br i1 %cmp.n207, label %for.cond2.for.cond.cleanup4_crit_edge.us.i, label %scalar.ph201

scalar.ph201:                                     ; preds = %vector.memcheck193, %for.cond2.preheader.us.i, %middle.block200
  %bc.resume.val206 = phi i64 [ %n.vec205, %middle.block200 ], [ 0, %for.cond2.preheader.us.i ], [ 0, %vector.memcheck193 ]
  br label %for.body5.us.i

for.body5.us.i:                                   ; preds = %for.body5.us.i, %scalar.ph201
  %indvars.iv94.i = phi i64 [ %bc.resume.val206, %scalar.ph201 ], [ %indvars.iv.next95.i, %for.body5.us.i ]
  %arrayidx.us.i66 = getelementptr inbounds i8, ptr %dst.addr.087.us.i, i64 %indvars.iv94.i
  %113 = load i8, ptr %arrayidx.us.i66, align 1, !tbaa !15
  %conv.us.i67 = zext i8 %113 to i32
  %114 = load i32, ptr %i_scale.i, align 4, !tbaa !104
  %mul.us.i = mul nsw i32 %114, %conv.us.i67
  %115 = load i32, ptr %i_denom.i, align 16, !tbaa !103
  %sub.us.i = add nsw i32 %115, -1
  %shl.us.i = shl nuw i32 1, %sub.us.i
  %add.us.i68 = add nsw i32 %shl.us.i, %mul.us.i
  %shr.us.i69 = ashr i32 %add.us.i68, %115
  %116 = load i32, ptr %i_offset.i, align 8, !tbaa !108
  %add8.us.i70 = add nsw i32 %shr.us.i69, %116
  %tobool.not.i.us.i = icmp ult i32 %add8.us.i70, 256
  %117 = icmp sgt i32 %add8.us.i70, 0
  %shr.i.us.i = sext i1 %117 to i32
  %cond.i.us.i = select i1 %tobool.not.i.us.i, i32 %add8.us.i70, i32 %shr.i.us.i
  %conv.i.us.i = trunc i32 %cond.i.us.i to i8
  store i8 %conv.i.us.i, ptr %arrayidx.us.i66, align 1, !tbaa !15
  %indvars.iv.next95.i = add nuw nsw i64 %indvars.iv94.i, 1
  %exitcond98.not.i = icmp eq i64 %indvars.iv.next95.i, %wide.trip.count97.i
  br i1 %exitcond98.not.i, label %for.cond2.for.cond.cleanup4_crit_edge.us.i, label %for.body5.us.i, !llvm.loop !176

for.cond2.for.cond.cleanup4_crit_edge.us.i:       ; preds = %middle.block200, %for.body5.us.i
  %inc12.us.i = add nuw nsw i32 %y.088.us.i, 1
  %add.ptr.us.i71 = getelementptr i8, ptr %dst.addr.087.us.i, i64 %idx.ext.i65
  %exitcond99.not.i = icmp eq i32 %inc12.us.i, %i_height
  br i1 %exitcond99.not.i, label %cleanup.loopexit152, label %for.cond2.preheader.us.i, !llvm.loop !121

if.else:                                          ; preds = %entry
  %weightfn24 = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 5
  %118 = load ptr, ptr %weightfn24, align 16, !tbaa !101
  %tobool25.not = icmp eq ptr %118, null
  br i1 %tobool25.not, label %if.else27, label %if.then26

if.then26:                                        ; preds = %if.else
  %119 = load i32, ptr %i_dst_stride, align 4, !tbaa !13
  %i_denom.i72 = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 2
  %120 = load i32, ptr %i_denom.i72, align 16, !tbaa !103
  %cmp.i73 = icmp sgt i32 %120, 0
  %cmp185.i74 = icmp sgt i32 %i_height, 0
  br i1 %cmp.i73, label %for.cond.preheader.i106, label %for.cond17.preheader.i75

for.cond17.preheader.i75:                         ; preds = %if.then26
  br i1 %cmp185.i74, label %for.cond23.preheader.lr.ph.i76, label %cleanup

for.cond23.preheader.lr.ph.i76:                   ; preds = %for.cond17.preheader.i75
  %cmp2477.i77 = icmp sgt i32 %i_width, 0
  %i_scale31.i78 = getelementptr %struct.x264_weight_t, ptr %weight, i64 0, i32 3
  %i_offset33.i79 = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 4
  %idx.ext43.i80 = sext i32 %119 to i64
  %idx.ext45.i81 = sext i32 %i_src_stride to i64
  br i1 %cmp2477.i77, label %for.cond23.preheader.us.preheader.i82, label %cleanup

for.cond23.preheader.us.preheader.i82:            ; preds = %for.cond23.preheader.lr.ph.i76
  %wide.trip.count.i83 = zext nneg i32 %i_width to i64
  %121 = add i32 %i_height, -1
  %122 = zext i32 %121 to i64
  %123 = mul i64 %idx.ext43.i80, %122
  %124 = add i64 %123, %wide.trip.count.i83
  %scevgep219 = getelementptr i8, ptr %dst, i64 %124
  %125 = mul i64 %idx.ext45.i81, %122
  %126 = add i64 %125, %idx.ext8
  %127 = add i64 %126, %idx.ext
  %128 = add i64 %127, %wide.trip.count.i83
  %scevgep220 = getelementptr i8, ptr %1, i64 %128
  %scevgep221 = getelementptr i8, ptr %weight, i64 44
  br label %for.cond23.preheader.us.i84

for.cond23.preheader.us.i84:                      ; preds = %for.cond23.for.cond.cleanup26_crit_edge.us.i101, %for.cond23.preheader.us.preheader.i82
  %y16.082.us.i85 = phi i32 [ %inc42.us.i102, %for.cond23.for.cond.cleanup26_crit_edge.us.i101 ], [ 0, %for.cond23.preheader.us.preheader.i82 ]
  %dst.addr.181.us.i86 = phi ptr [ %add.ptr44.us.i103, %for.cond23.for.cond.cleanup26_crit_edge.us.i101 ], [ %dst, %for.cond23.preheader.us.preheader.i82 ]
  %src.addr.180.us.i87 = phi ptr [ %add.ptr46.us.i104, %for.cond23.for.cond.cleanup26_crit_edge.us.i101 ], [ %add.ptr9, %for.cond23.preheader.us.preheader.i82 ]
  %min.iters.check234 = icmp ult i64 %wide.trip.count.i83, 12
  br i1 %min.iters.check234, label %scalar.ph233, label %vector.memcheck218

vector.memcheck218:                               ; preds = %for.cond23.preheader.us.i84
  %bound0222 = icmp ult ptr %dst, %scevgep220
  %bound1223 = icmp ult ptr %add.ptr9, %scevgep219
  %found.conflict224 = and i1 %bound0222, %bound1223
  %stride.check225 = icmp slt i64 %idx.ext43.i80, 0
  %129 = or i1 %found.conflict224, %stride.check225
  %stride.check226 = icmp slt i64 %idx.ext45.i81, 0
  %130 = or i1 %129, %stride.check226
  %bound0227 = icmp ult ptr %dst, %scevgep221
  %bound1228 = icmp ult ptr %i_scale31.i78, %scevgep219
  %found.conflict229 = and i1 %bound0227, %bound1228
  %stride.check230 = icmp slt i64 %idx.ext43.i80, 0
  %131 = or i1 %found.conflict229, %stride.check230
  %conflict.rdx231 = or i1 %130, %131
  br i1 %conflict.rdx231, label %scalar.ph233, label %vector.ph235

vector.ph235:                                     ; preds = %vector.memcheck218
  %n.mod.vf236 = urem i64 %wide.trip.count.i83, 4
  %n.vec237 = sub i64 %wide.trip.count.i83, %n.mod.vf236
  br label %vector.body240

vector.body240:                                   ; preds = %vector.body240, %vector.ph235
  %index241 = phi i64 [ 0, %vector.ph235 ], [ %index.next247, %vector.body240 ]
  %132 = add i64 %index241, 0
  %133 = getelementptr inbounds i8, ptr %src.addr.180.us.i87, i64 %132
  %134 = getelementptr inbounds i8, ptr %133, i32 0
  %wide.load242 = load <4 x i8>, ptr %134, align 1, !tbaa !15, !alias.scope !177
  %135 = zext <4 x i8> %wide.load242 to <4 x i32>
  %136 = load i32, ptr %i_scale31.i78, align 4, !tbaa !104, !alias.scope !180
  %broadcast.splatinsert243 = insertelement <4 x i32> poison, i32 %136, i64 0
  %broadcast.splat244 = shufflevector <4 x i32> %broadcast.splatinsert243, <4 x i32> poison, <4 x i32> zeroinitializer
  %137 = mul nsw <4 x i32> %broadcast.splat244, %135
  %138 = load i32, ptr %i_offset33.i79, align 8, !tbaa !108, !alias.scope !180
  %broadcast.splatinsert245 = insertelement <4 x i32> poison, i32 %138, i64 0
  %broadcast.splat246 = shufflevector <4 x i32> %broadcast.splatinsert245, <4 x i32> poison, <4 x i32> zeroinitializer
  %139 = add nsw <4 x i32> %137, %broadcast.splat246
  %140 = icmp ult <4 x i32> %139, <i32 256, i32 256, i32 256, i32 256>
  %141 = icmp sgt <4 x i32> %139, zeroinitializer
  %142 = sext <4 x i1> %141 to <4 x i32>
  %143 = select <4 x i1> %140, <4 x i32> %139, <4 x i32> %142
  %144 = trunc <4 x i32> %143 to <4 x i8>
  %145 = getelementptr inbounds i8, ptr %dst.addr.181.us.i86, i64 %132
  %146 = getelementptr inbounds i8, ptr %145, i32 0
  store <4 x i8> %144, ptr %146, align 1, !tbaa !15, !alias.scope !182, !noalias !184
  %index.next247 = add nuw i64 %index241, 4
  %147 = icmp eq i64 %index.next247, %n.vec237
  br i1 %147, label %middle.block232, label %vector.body240, !llvm.loop !185

middle.block232:                                  ; preds = %vector.body240
  %cmp.n239 = icmp eq i64 %wide.trip.count.i83, %n.vec237
  br i1 %cmp.n239, label %for.cond23.for.cond.cleanup26_crit_edge.us.i101, label %scalar.ph233

scalar.ph233:                                     ; preds = %vector.memcheck218, %for.cond23.preheader.us.i84, %middle.block232
  %bc.resume.val238 = phi i64 [ %n.vec237, %middle.block232 ], [ 0, %for.cond23.preheader.us.i84 ], [ 0, %vector.memcheck218 ]
  br label %for.body27.us.i88

for.body27.us.i88:                                ; preds = %for.body27.us.i88, %scalar.ph233
  %indvars.iv.i89 = phi i64 [ %bc.resume.val238, %scalar.ph233 ], [ %indvars.iv.next.i99, %for.body27.us.i88 ]
  %arrayidx29.us.i90 = getelementptr inbounds i8, ptr %src.addr.180.us.i87, i64 %indvars.iv.i89
  %148 = load i8, ptr %arrayidx29.us.i90, align 1, !tbaa !15
  %conv30.us.i91 = zext i8 %148 to i32
  %149 = load i32, ptr %i_scale31.i78, align 4, !tbaa !104
  %mul32.us.i92 = mul nsw i32 %149, %conv30.us.i91
  %150 = load i32, ptr %i_offset33.i79, align 8, !tbaa !108
  %add34.us.i93 = add nsw i32 %mul32.us.i92, %150
  %tobool.not.i72.us.i94 = icmp ult i32 %add34.us.i93, 256
  %151 = icmp sgt i32 %add34.us.i93, 0
  %shr.i73.us.i95 = sext i1 %151 to i32
  %cond.i74.us.i96 = select i1 %tobool.not.i72.us.i94, i32 %add34.us.i93, i32 %shr.i73.us.i95
  %conv.i75.us.i97 = trunc i32 %cond.i74.us.i96 to i8
  %arrayidx37.us.i98 = getelementptr inbounds i8, ptr %dst.addr.181.us.i86, i64 %indvars.iv.i89
  store i8 %conv.i75.us.i97, ptr %arrayidx37.us.i98, align 1, !tbaa !15
  %indvars.iv.next.i99 = add nuw nsw i64 %indvars.iv.i89, 1
  %exitcond.not.i100 = icmp eq i64 %indvars.iv.next.i99, %wide.trip.count.i83
  br i1 %exitcond.not.i100, label %for.cond23.for.cond.cleanup26_crit_edge.us.i101, label %for.body27.us.i88, !llvm.loop !186

for.cond23.for.cond.cleanup26_crit_edge.us.i101:  ; preds = %middle.block232, %for.body27.us.i88
  %inc42.us.i102 = add nuw nsw i32 %y16.082.us.i85, 1
  %add.ptr44.us.i103 = getelementptr inbounds i8, ptr %dst.addr.181.us.i86, i64 %idx.ext43.i80
  %add.ptr46.us.i104 = getelementptr inbounds i8, ptr %src.addr.180.us.i87, i64 %idx.ext45.i81
  %exitcond93.not.i105 = icmp eq i32 %inc42.us.i102, %i_height
  br i1 %exitcond93.not.i105, label %cleanup.loopexit151, label %for.cond23.preheader.us.i84, !llvm.loop !113

for.cond.preheader.i106:                          ; preds = %if.then26
  br i1 %cmp185.i74, label %for.cond2.preheader.lr.ph.i107, label %cleanup

for.cond2.preheader.lr.ph.i107:                   ; preds = %for.cond.preheader.i106
  %cmp383.i108 = icmp sgt i32 %i_width, 0
  %i_scale.i109 = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 3
  %i_offset.i110 = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 4
  %idx.ext.i111 = sext i32 %119 to i64
  %idx.ext13.i112 = sext i32 %i_src_stride to i64
  br i1 %cmp383.i108, label %for.cond2.preheader.us.preheader.i113, label %cleanup

for.cond2.preheader.us.preheader.i113:            ; preds = %for.cond2.preheader.lr.ph.i107
  %wide.trip.count97.i114 = zext nneg i32 %i_width to i64
  %152 = add i32 %i_height, -1
  %153 = zext i32 %152 to i64
  %154 = mul i64 %idx.ext.i111, %153
  %155 = add i64 %154, %wide.trip.count97.i114
  %scevgep249 = getelementptr i8, ptr %dst, i64 %155
  %156 = mul i64 %idx.ext13.i112, %153
  %157 = add i64 %156, %idx.ext8
  %158 = add i64 %157, %idx.ext
  %159 = add i64 %158, %wide.trip.count97.i114
  %scevgep250 = getelementptr i8, ptr %1, i64 %159
  %scevgep251 = getelementptr i8, ptr %weight, i64 44
  br label %for.cond2.preheader.us.i115

for.cond2.preheader.us.i115:                      ; preds = %for.cond2.for.cond.cleanup4_crit_edge.us.i136, %for.cond2.preheader.us.preheader.i113
  %y.088.us.i116 = phi i32 [ %inc12.us.i137, %for.cond2.for.cond.cleanup4_crit_edge.us.i136 ], [ 0, %for.cond2.preheader.us.preheader.i113 ]
  %dst.addr.087.us.i117 = phi ptr [ %add.ptr.us.i138, %for.cond2.for.cond.cleanup4_crit_edge.us.i136 ], [ %dst, %for.cond2.preheader.us.preheader.i113 ]
  %src.addr.086.us.i118 = phi ptr [ %add.ptr14.us.i139, %for.cond2.for.cond.cleanup4_crit_edge.us.i136 ], [ %add.ptr9, %for.cond2.preheader.us.preheader.i113 ]
  %min.iters.check264 = icmp ult i64 %wide.trip.count97.i114, 8
  br i1 %min.iters.check264, label %scalar.ph263, label %vector.memcheck248

vector.memcheck248:                               ; preds = %for.cond2.preheader.us.i115
  %bound0252 = icmp ult ptr %dst, %scevgep250
  %bound1253 = icmp ult ptr %add.ptr9, %scevgep249
  %found.conflict254 = and i1 %bound0252, %bound1253
  %stride.check255 = icmp slt i64 %idx.ext.i111, 0
  %160 = or i1 %found.conflict254, %stride.check255
  %stride.check256 = icmp slt i64 %idx.ext13.i112, 0
  %161 = or i1 %160, %stride.check256
  %bound0257 = icmp ult ptr %dst, %scevgep251
  %bound1258 = icmp ult ptr %i_denom.i72, %scevgep249
  %found.conflict259 = and i1 %bound0257, %bound1258
  %stride.check260 = icmp slt i64 %idx.ext.i111, 0
  %162 = or i1 %found.conflict259, %stride.check260
  %conflict.rdx261 = or i1 %161, %162
  br i1 %conflict.rdx261, label %scalar.ph263, label %vector.ph265

vector.ph265:                                     ; preds = %vector.memcheck248
  %n.mod.vf266 = urem i64 %wide.trip.count97.i114, 4
  %n.vec267 = sub i64 %wide.trip.count97.i114, %n.mod.vf266
  br label %vector.body270

vector.body270:                                   ; preds = %vector.body270, %vector.ph265
  %index271 = phi i64 [ 0, %vector.ph265 ], [ %index.next279, %vector.body270 ]
  %163 = add i64 %index271, 0
  %164 = getelementptr inbounds i8, ptr %src.addr.086.us.i118, i64 %163
  %165 = getelementptr inbounds i8, ptr %164, i32 0
  %wide.load272 = load <4 x i8>, ptr %165, align 1, !tbaa !15, !alias.scope !187
  %166 = zext <4 x i8> %wide.load272 to <4 x i32>
  %167 = load i32, ptr %i_scale.i109, align 4, !tbaa !104, !alias.scope !190
  %broadcast.splatinsert273 = insertelement <4 x i32> poison, i32 %167, i64 0
  %broadcast.splat274 = shufflevector <4 x i32> %broadcast.splatinsert273, <4 x i32> poison, <4 x i32> zeroinitializer
  %168 = mul nsw <4 x i32> %broadcast.splat274, %166
  %169 = load i32, ptr %i_denom.i72, align 16, !tbaa !103, !alias.scope !190
  %broadcast.splatinsert275 = insertelement <4 x i32> poison, i32 %169, i64 0
  %broadcast.splat276 = shufflevector <4 x i32> %broadcast.splatinsert275, <4 x i32> poison, <4 x i32> zeroinitializer
  %170 = add nsw <4 x i32> %broadcast.splat276, <i32 -1, i32 -1, i32 -1, i32 -1>
  %171 = shl nuw <4 x i32> <i32 1, i32 1, i32 1, i32 1>, %170
  %172 = add nsw <4 x i32> %171, %168
  %173 = ashr <4 x i32> %172, %broadcast.splat276
  %174 = load i32, ptr %i_offset.i110, align 8, !tbaa !108, !alias.scope !190
  %broadcast.splatinsert277 = insertelement <4 x i32> poison, i32 %174, i64 0
  %broadcast.splat278 = shufflevector <4 x i32> %broadcast.splatinsert277, <4 x i32> poison, <4 x i32> zeroinitializer
  %175 = add nsw <4 x i32> %173, %broadcast.splat278
  %176 = icmp ult <4 x i32> %175, <i32 256, i32 256, i32 256, i32 256>
  %177 = icmp sgt <4 x i32> %175, zeroinitializer
  %178 = sext <4 x i1> %177 to <4 x i32>
  %179 = select <4 x i1> %176, <4 x i32> %175, <4 x i32> %178
  %180 = trunc <4 x i32> %179 to <4 x i8>
  %181 = getelementptr inbounds i8, ptr %dst.addr.087.us.i117, i64 %163
  %182 = getelementptr inbounds i8, ptr %181, i32 0
  store <4 x i8> %180, ptr %182, align 1, !tbaa !15, !alias.scope !192, !noalias !194
  %index.next279 = add nuw i64 %index271, 4
  %183 = icmp eq i64 %index.next279, %n.vec267
  br i1 %183, label %middle.block262, label %vector.body270, !llvm.loop !195

middle.block262:                                  ; preds = %vector.body270
  %cmp.n269 = icmp eq i64 %wide.trip.count97.i114, %n.vec267
  br i1 %cmp.n269, label %for.cond2.for.cond.cleanup4_crit_edge.us.i136, label %scalar.ph263

scalar.ph263:                                     ; preds = %vector.memcheck248, %for.cond2.preheader.us.i115, %middle.block262
  %bc.resume.val268 = phi i64 [ %n.vec267, %middle.block262 ], [ 0, %for.cond2.preheader.us.i115 ], [ 0, %vector.memcheck248 ]
  br label %for.body5.us.i119

for.body5.us.i119:                                ; preds = %for.body5.us.i119, %scalar.ph263
  %indvars.iv94.i120 = phi i64 [ %bc.resume.val268, %scalar.ph263 ], [ %indvars.iv.next95.i134, %for.body5.us.i119 ]
  %arrayidx.us.i121 = getelementptr inbounds i8, ptr %src.addr.086.us.i118, i64 %indvars.iv94.i120
  %184 = load i8, ptr %arrayidx.us.i121, align 1, !tbaa !15
  %conv.us.i122 = zext i8 %184 to i32
  %185 = load i32, ptr %i_scale.i109, align 4, !tbaa !104
  %mul.us.i123 = mul nsw i32 %185, %conv.us.i122
  %186 = load i32, ptr %i_denom.i72, align 16, !tbaa !103
  %sub.us.i124 = add nsw i32 %186, -1
  %shl.us.i125 = shl nuw i32 1, %sub.us.i124
  %add.us.i126 = add nsw i32 %shl.us.i125, %mul.us.i123
  %shr.us.i127 = ashr i32 %add.us.i126, %186
  %187 = load i32, ptr %i_offset.i110, align 8, !tbaa !108
  %add8.us.i128 = add nsw i32 %shr.us.i127, %187
  %tobool.not.i.us.i129 = icmp ult i32 %add8.us.i128, 256
  %188 = icmp sgt i32 %add8.us.i128, 0
  %shr.i.us.i130 = sext i1 %188 to i32
  %cond.i.us.i131 = select i1 %tobool.not.i.us.i129, i32 %add8.us.i128, i32 %shr.i.us.i130
  %conv.i.us.i132 = trunc i32 %cond.i.us.i131 to i8
  %arrayidx10.us.i133 = getelementptr inbounds i8, ptr %dst.addr.087.us.i117, i64 %indvars.iv94.i120
  store i8 %conv.i.us.i132, ptr %arrayidx10.us.i133, align 1, !tbaa !15
  %indvars.iv.next95.i134 = add nuw nsw i64 %indvars.iv94.i120, 1
  %exitcond98.not.i135 = icmp eq i64 %indvars.iv.next95.i134, %wide.trip.count97.i114
  br i1 %exitcond98.not.i135, label %for.cond2.for.cond.cleanup4_crit_edge.us.i136, label %for.body5.us.i119, !llvm.loop !196

for.cond2.for.cond.cleanup4_crit_edge.us.i136:    ; preds = %middle.block262, %for.body5.us.i119
  %inc12.us.i137 = add nuw nsw i32 %y.088.us.i116, 1
  %add.ptr.us.i138 = getelementptr inbounds i8, ptr %dst.addr.087.us.i117, i64 %idx.ext.i111
  %add.ptr14.us.i139 = getelementptr inbounds i8, ptr %src.addr.086.us.i118, i64 %idx.ext13.i112
  %exitcond99.not.i140 = icmp eq i32 %inc12.us.i137, %i_height
  br i1 %exitcond99.not.i140, label %cleanup.loopexit, label %for.cond2.preheader.us.i115, !llvm.loop !121

if.else27:                                        ; preds = %if.else
  store i32 %i_src_stride, ptr %i_dst_stride, align 4, !tbaa !13
  br label %cleanup

cleanup.loopexit:                                 ; preds = %for.cond2.for.cond.cleanup4_crit_edge.us.i136
  br label %cleanup

cleanup.loopexit151:                              ; preds = %for.cond23.for.cond.cleanup26_crit_edge.us.i101
  br label %cleanup

cleanup.loopexit152:                              ; preds = %for.cond2.for.cond.cleanup4_crit_edge.us.i
  br label %cleanup

cleanup.loopexit153:                              ; preds = %for.cond23.for.cond.cleanup26_crit_edge.us.i
  br label %cleanup

cleanup:                                          ; preds = %cleanup.loopexit153, %cleanup.loopexit152, %cleanup.loopexit151, %cleanup.loopexit, %for.cond2.preheader.lr.ph.i107, %for.cond.preheader.i106, %for.cond23.preheader.lr.ph.i76, %for.cond17.preheader.i75, %for.cond2.preheader.lr.ph.i, %for.cond.preheader.i, %for.cond23.preheader.lr.ph.i, %for.cond17.preheader.i, %pixel_avg.exit, %if.else27
  %retval.0 = phi ptr [ %add.ptr9, %if.else27 ], [ %dst, %pixel_avg.exit ], [ %dst, %for.cond17.preheader.i ], [ %dst, %for.cond23.preheader.lr.ph.i ], [ %dst, %for.cond.preheader.i ], [ %dst, %for.cond2.preheader.lr.ph.i ], [ %dst, %for.cond17.preheader.i75 ], [ %dst, %for.cond23.preheader.lr.ph.i76 ], [ %dst, %for.cond.preheader.i106 ], [ %dst, %for.cond2.preheader.lr.ph.i107 ], [ %dst, %cleanup.loopexit ], [ %dst, %cleanup.loopexit151 ], [ %dst, %cleanup.loopexit152 ], [ %dst, %cleanup.loopexit153 ]
  ret ptr %retval.0
}

get_ref.print-before.loop-vectorize.ll

; *** IR Dump Before LoopVectorizePass on get_ref ***
; Function Attrs: nofree norecurse nosync nounwind memory(readwrite, inaccessiblemem: none) uwtable
define internal ptr @get_ref(ptr noundef %dst, ptr nocapture noundef %i_dst_stride, ptr nocapture noundef readonly %src, i32 noundef %i_src_stride, i32 noundef %mvx, i32 noundef %mvy, i32 noundef %i_width, i32 noundef %i_height, ptr nocapture noundef readonly %weight) #0 {
entry:
  %and = and i32 %mvy, 3
  %shl = shl nuw nsw i32 %and, 2
  %and1 = and i32 %mvx, 3
  %add = or disjoint i32 %shl, %and1
  %shr = ashr i32 %mvy, 2
  %mul = mul nsw i32 %shr, %i_src_stride
  %shr2 = ashr i32 %mvx, 2
  %add3 = add nsw i32 %mul, %shr2
  %idxprom = zext nneg i32 %add to i64
  %arrayidx = getelementptr inbounds [16 x i8], ptr @hpel_ref0, i64 0, i64 %idxprom
  %0 = load i8, ptr %arrayidx, align 1, !tbaa !15
  %idxprom4 = zext i8 %0 to i64
  %arrayidx5 = getelementptr inbounds ptr, ptr %src, i64 %idxprom4
  %1 = load ptr, ptr %arrayidx5, align 8, !tbaa !9
  %idx.ext = sext i32 %add3 to i64
  %add.ptr = getelementptr inbounds i8, ptr %1, i64 %idx.ext
  %cmp = icmp eq i32 %and, 3
  %mul7 = select i1 %cmp, i32 %i_src_stride, i32 0
  %idx.ext8 = sext i32 %mul7 to i64
  %add.ptr9 = getelementptr inbounds i8, ptr %add.ptr, i64 %idx.ext8
  %and10 = and i32 %add, 5
  %tobool.not = icmp eq i32 %and10, 0
  br i1 %tobool.not, label %if.else, label %if.then

if.then:                                          ; preds = %entry
  %arrayidx12 = getelementptr inbounds [16 x i8], ptr @hpel_ref1, i64 0, i64 %idxprom
  %2 = load i8, ptr %arrayidx12, align 1, !tbaa !15
  %idxprom13 = zext i8 %2 to i64
  %arrayidx14 = getelementptr inbounds ptr, ptr %src, i64 %idxprom13
  %3 = load ptr, ptr %arrayidx14, align 8, !tbaa !9
  %add.ptr16 = getelementptr inbounds i8, ptr %3, i64 %idx.ext
  %cmp18 = icmp eq i32 %and1, 3
  %idx.ext20 = zext i1 %cmp18 to i64
  %add.ptr21 = getelementptr inbounds i8, ptr %add.ptr16, i64 %idx.ext20
  %cmp29.i = icmp sgt i32 %i_height, 0
  br i1 %cmp29.i, label %for.cond1.preheader.lr.ph.i, label %pixel_avg.exit

for.cond1.preheader.lr.ph.i:                      ; preds = %if.then
  %4 = load i32, ptr %i_dst_stride, align 4, !tbaa !13
  %cmp227.i = icmp sgt i32 %i_width, 0
  %idx.ext.i = sext i32 %4 to i64
  %idx.ext12.i = sext i32 %i_src_stride to i64
  br i1 %cmp227.i, label %for.cond1.preheader.us.preheader.i, label %pixel_avg.exit

for.cond1.preheader.us.preheader.i:               ; preds = %for.cond1.preheader.lr.ph.i
  %wide.trip.count.i = zext nneg i32 %i_width to i64
  br label %for.cond1.preheader.us.i

for.cond1.preheader.us.i:                         ; preds = %for.cond1.for.cond.cleanup3_crit_edge.us.i, %for.cond1.preheader.us.preheader.i
  %y.033.us.i = phi i32 [ %inc17.us.i, %for.cond1.for.cond.cleanup3_crit_edge.us.i ], [ 0, %for.cond1.preheader.us.preheader.i ]
  %dst.addr.032.us.i = phi ptr [ %add.ptr.us.i, %for.cond1.for.cond.cleanup3_crit_edge.us.i ], [ %dst, %for.cond1.preheader.us.preheader.i ]
  %src1.addr.031.us.i = phi ptr [ %add.ptr13.us.i, %for.cond1.for.cond.cleanup3_crit_edge.us.i ], [ %add.ptr9, %for.cond1.preheader.us.preheader.i ]
  %src2.addr.030.us.i = phi ptr [ %add.ptr15.us.i, %for.cond1.for.cond.cleanup3_crit_edge.us.i ], [ %add.ptr21, %for.cond1.preheader.us.preheader.i ]
  br label %for.body4.us.i

for.body4.us.i:                                   ; preds = %for.body4.us.i, %for.cond1.preheader.us.i
  %indvars.iv.i = phi i64 [ 0, %for.cond1.preheader.us.i ], [ %indvars.iv.next.i, %for.body4.us.i ]
  %arrayidx.us.i = getelementptr inbounds i8, ptr %src1.addr.031.us.i, i64 %indvars.iv.i
  %5 = load i8, ptr %arrayidx.us.i, align 1, !tbaa !15
  %conv.us.i = zext i8 %5 to i16
  %arrayidx6.us.i = getelementptr inbounds i8, ptr %src2.addr.030.us.i, i64 %indvars.iv.i
  %6 = load i8, ptr %arrayidx6.us.i, align 1, !tbaa !15
  %conv7.us.i = zext i8 %6 to i16
  %add.us.i = add nuw nsw i16 %conv.us.i, 1
  %add8.us.i = add nuw nsw i16 %add.us.i, %conv7.us.i
  %shr.us.i = lshr i16 %add8.us.i, 1
  %conv9.us.i = trunc i16 %shr.us.i to i8
  %arrayidx11.us.i = getelementptr inbounds i8, ptr %dst.addr.032.us.i, i64 %indvars.iv.i
  store i8 %conv9.us.i, ptr %arrayidx11.us.i, align 1, !tbaa !15
  %indvars.iv.next.i = add nuw nsw i64 %indvars.iv.i, 1
  %exitcond.not.i = icmp eq i64 %indvars.iv.next.i, %wide.trip.count.i
  br i1 %exitcond.not.i, label %for.cond1.for.cond.cleanup3_crit_edge.us.i, label %for.body4.us.i, !llvm.loop !144

for.cond1.for.cond.cleanup3_crit_edge.us.i:       ; preds = %for.body4.us.i
  %add.ptr.us.i = getelementptr inbounds i8, ptr %dst.addr.032.us.i, i64 %idx.ext.i
  %add.ptr13.us.i = getelementptr inbounds i8, ptr %src1.addr.031.us.i, i64 %idx.ext12.i
  %add.ptr15.us.i = getelementptr inbounds i8, ptr %src2.addr.030.us.i, i64 %idx.ext12.i
  %inc17.us.i = add nuw nsw i32 %y.033.us.i, 1
  %exitcond36.not.i = icmp eq i32 %inc17.us.i, %i_height
  br i1 %exitcond36.not.i, label %pixel_avg.exit.loopexit, label %for.cond1.preheader.us.i, !llvm.loop !100

pixel_avg.exit.loopexit:                          ; preds = %for.cond1.for.cond.cleanup3_crit_edge.us.i
  br label %pixel_avg.exit

pixel_avg.exit:                                   ; preds = %pixel_avg.exit.loopexit, %if.then, %for.cond1.preheader.lr.ph.i
  %weightfn = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 5
  %7 = load ptr, ptr %weightfn, align 16, !tbaa !101
  %tobool22.not = icmp eq ptr %7, null
  br i1 %tobool22.not, label %cleanup, label %if.then23

if.then23:                                        ; preds = %pixel_avg.exit
  %8 = load i32, ptr %i_dst_stride, align 4, !tbaa !13
  %i_denom.i = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 2
  %9 = load i32, ptr %i_denom.i, align 16, !tbaa !103
  %cmp.i = icmp sgt i32 %9, 0
  br i1 %cmp.i, label %for.cond.preheader.i, label %for.cond17.preheader.i

for.cond17.preheader.i:                           ; preds = %if.then23
  br i1 %cmp29.i, label %for.cond23.preheader.lr.ph.i, label %cleanup

for.cond23.preheader.lr.ph.i:                     ; preds = %for.cond17.preheader.i
  %cmp2477.i = icmp sgt i32 %i_width, 0
  %i_scale31.i = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 3
  %i_offset33.i = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 4
  %idx.ext43.i = sext i32 %8 to i64
  br i1 %cmp2477.i, label %for.cond23.preheader.us.preheader.i, label %cleanup

for.cond23.preheader.us.preheader.i:              ; preds = %for.cond23.preheader.lr.ph.i
  %wide.trip.count.i61 = zext nneg i32 %i_width to i64
  br label %for.cond23.preheader.us.i

for.cond23.preheader.us.i:                        ; preds = %for.cond23.for.cond.cleanup26_crit_edge.us.i, %for.cond23.preheader.us.preheader.i
  %y16.082.us.i = phi i32 [ %inc42.us.i, %for.cond23.for.cond.cleanup26_crit_edge.us.i ], [ 0, %for.cond23.preheader.us.preheader.i ]
  %dst.addr.181.us.i = phi ptr [ %add.ptr44.us.i, %for.cond23.for.cond.cleanup26_crit_edge.us.i ], [ %dst, %for.cond23.preheader.us.preheader.i ]
  br label %for.body27.us.i

for.body27.us.i:                                  ; preds = %for.body27.us.i, %for.cond23.preheader.us.i
  %indvars.iv.i62 = phi i64 [ 0, %for.cond23.preheader.us.i ], [ %indvars.iv.next.i63, %for.body27.us.i ]
  %arrayidx29.us.i = getelementptr inbounds i8, ptr %dst.addr.181.us.i, i64 %indvars.iv.i62
  %10 = load i8, ptr %arrayidx29.us.i, align 1, !tbaa !15
  %conv30.us.i = zext i8 %10 to i32
  %11 = load i32, ptr %i_scale31.i, align 4, !tbaa !104
  %mul32.us.i = mul nsw i32 %11, %conv30.us.i
  %12 = load i32, ptr %i_offset33.i, align 8, !tbaa !108
  %add34.us.i = add nsw i32 %mul32.us.i, %12
  %tobool.not.i72.us.i = icmp ult i32 %add34.us.i, 256
  %13 = icmp sgt i32 %add34.us.i, 0
  %shr.i73.us.i = sext i1 %13 to i32
  %cond.i74.us.i = select i1 %tobool.not.i72.us.i, i32 %add34.us.i, i32 %shr.i73.us.i
  %conv.i75.us.i = trunc i32 %cond.i74.us.i to i8
  store i8 %conv.i75.us.i, ptr %arrayidx29.us.i, align 1, !tbaa !15
  %indvars.iv.next.i63 = add nuw nsw i64 %indvars.iv.i62, 1
  %exitcond.not.i64 = icmp eq i64 %indvars.iv.next.i63, %wide.trip.count.i61
  br i1 %exitcond.not.i64, label %for.cond23.for.cond.cleanup26_crit_edge.us.i, label %for.body27.us.i, !llvm.loop !145

for.cond23.for.cond.cleanup26_crit_edge.us.i:     ; preds = %for.body27.us.i
  %inc42.us.i = add nuw nsw i32 %y16.082.us.i, 1
  %add.ptr44.us.i = getelementptr i8, ptr %dst.addr.181.us.i, i64 %idx.ext43.i
  %exitcond93.not.i = icmp eq i32 %inc42.us.i, %i_height
  br i1 %exitcond93.not.i, label %cleanup.loopexit153, label %for.cond23.preheader.us.i, !llvm.loop !113

for.cond.preheader.i:                             ; preds = %if.then23
  br i1 %cmp29.i, label %for.cond2.preheader.lr.ph.i, label %cleanup

for.cond2.preheader.lr.ph.i:                      ; preds = %for.cond.preheader.i
  %cmp383.i = icmp sgt i32 %i_width, 0
  %i_scale.i = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 3
  %i_offset.i = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 4
  %idx.ext.i65 = sext i32 %8 to i64
  br i1 %cmp383.i, label %for.cond2.preheader.us.preheader.i, label %cleanup

for.cond2.preheader.us.preheader.i:               ; preds = %for.cond2.preheader.lr.ph.i
  %wide.trip.count97.i = zext nneg i32 %i_width to i64
  br label %for.cond2.preheader.us.i

for.cond2.preheader.us.i:                         ; preds = %for.cond2.for.cond.cleanup4_crit_edge.us.i, %for.cond2.preheader.us.preheader.i
  %y.088.us.i = phi i32 [ %inc12.us.i, %for.cond2.for.cond.cleanup4_crit_edge.us.i ], [ 0, %for.cond2.preheader.us.preheader.i ]
  %dst.addr.087.us.i = phi ptr [ %add.ptr.us.i71, %for.cond2.for.cond.cleanup4_crit_edge.us.i ], [ %dst, %for.cond2.preheader.us.preheader.i ]
  br label %for.body5.us.i

for.body5.us.i:                                   ; preds = %for.body5.us.i, %for.cond2.preheader.us.i
  %indvars.iv94.i = phi i64 [ 0, %for.cond2.preheader.us.i ], [ %indvars.iv.next95.i, %for.body5.us.i ]
  %arrayidx.us.i66 = getelementptr inbounds i8, ptr %dst.addr.087.us.i, i64 %indvars.iv94.i
  %14 = load i8, ptr %arrayidx.us.i66, align 1, !tbaa !15
  %conv.us.i67 = zext i8 %14 to i32
  %15 = load i32, ptr %i_scale.i, align 4, !tbaa !104
  %mul.us.i = mul nsw i32 %15, %conv.us.i67
  %16 = load i32, ptr %i_denom.i, align 16, !tbaa !103
  %sub.us.i = add nsw i32 %16, -1
  %shl.us.i = shl nuw i32 1, %sub.us.i
  %add.us.i68 = add nsw i32 %shl.us.i, %mul.us.i
  %shr.us.i69 = ashr i32 %add.us.i68, %16
  %17 = load i32, ptr %i_offset.i, align 8, !tbaa !108
  %add8.us.i70 = add nsw i32 %shr.us.i69, %17
  %tobool.not.i.us.i = icmp ult i32 %add8.us.i70, 256
  %18 = icmp sgt i32 %add8.us.i70, 0
  %shr.i.us.i = sext i1 %18 to i32
  %cond.i.us.i = select i1 %tobool.not.i.us.i, i32 %add8.us.i70, i32 %shr.i.us.i
  %conv.i.us.i = trunc i32 %cond.i.us.i to i8
  store i8 %conv.i.us.i, ptr %arrayidx.us.i66, align 1, !tbaa !15
  %indvars.iv.next95.i = add nuw nsw i64 %indvars.iv94.i, 1
  %exitcond98.not.i = icmp eq i64 %indvars.iv.next95.i, %wide.trip.count97.i
  br i1 %exitcond98.not.i, label %for.cond2.for.cond.cleanup4_crit_edge.us.i, label %for.body5.us.i, !llvm.loop !146

for.cond2.for.cond.cleanup4_crit_edge.us.i:       ; preds = %for.body5.us.i
  %inc12.us.i = add nuw nsw i32 %y.088.us.i, 1
  %add.ptr.us.i71 = getelementptr i8, ptr %dst.addr.087.us.i, i64 %idx.ext.i65
  %exitcond99.not.i = icmp eq i32 %inc12.us.i, %i_height
  br i1 %exitcond99.not.i, label %cleanup.loopexit152, label %for.cond2.preheader.us.i, !llvm.loop !121

if.else:                                          ; preds = %entry
  %weightfn24 = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 5
  %19 = load ptr, ptr %weightfn24, align 16, !tbaa !101
  %tobool25.not = icmp eq ptr %19, null
  br i1 %tobool25.not, label %if.else27, label %if.then26

if.then26:                                        ; preds = %if.else
  %20 = load i32, ptr %i_dst_stride, align 4, !tbaa !13
  %i_denom.i72 = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 2
  %21 = load i32, ptr %i_denom.i72, align 16, !tbaa !103
  %cmp.i73 = icmp sgt i32 %21, 0
  %cmp185.i74 = icmp sgt i32 %i_height, 0
  br i1 %cmp.i73, label %for.cond.preheader.i106, label %for.cond17.preheader.i75

for.cond17.preheader.i75:                         ; preds = %if.then26
  br i1 %cmp185.i74, label %for.cond23.preheader.lr.ph.i76, label %cleanup

for.cond23.preheader.lr.ph.i76:                   ; preds = %for.cond17.preheader.i75
  %cmp2477.i77 = icmp sgt i32 %i_width, 0
  %i_scale31.i78 = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 3
  %i_offset33.i79 = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 4
  %idx.ext43.i80 = sext i32 %20 to i64
  %idx.ext45.i81 = sext i32 %i_src_stride to i64
  br i1 %cmp2477.i77, label %for.cond23.preheader.us.preheader.i82, label %cleanup

for.cond23.preheader.us.preheader.i82:            ; preds = %for.cond23.preheader.lr.ph.i76
  %wide.trip.count.i83 = zext nneg i32 %i_width to i64
  br label %for.cond23.preheader.us.i84

for.cond23.preheader.us.i84:                      ; preds = %for.cond23.for.cond.cleanup26_crit_edge.us.i101, %for.cond23.preheader.us.preheader.i82
  %y16.082.us.i85 = phi i32 [ %inc42.us.i102, %for.cond23.for.cond.cleanup26_crit_edge.us.i101 ], [ 0, %for.cond23.preheader.us.preheader.i82 ]
  %dst.addr.181.us.i86 = phi ptr [ %add.ptr44.us.i103, %for.cond23.for.cond.cleanup26_crit_edge.us.i101 ], [ %dst, %for.cond23.preheader.us.preheader.i82 ]
  %src.addr.180.us.i87 = phi ptr [ %add.ptr46.us.i104, %for.cond23.for.cond.cleanup26_crit_edge.us.i101 ], [ %add.ptr9, %for.cond23.preheader.us.preheader.i82 ]
  br label %for.body27.us.i88

for.body27.us.i88:                                ; preds = %for.body27.us.i88, %for.cond23.preheader.us.i84
  %indvars.iv.i89 = phi i64 [ 0, %for.cond23.preheader.us.i84 ], [ %indvars.iv.next.i99, %for.body27.us.i88 ]
  %arrayidx29.us.i90 = getelementptr inbounds i8, ptr %src.addr.180.us.i87, i64 %indvars.iv.i89
  %22 = load i8, ptr %arrayidx29.us.i90, align 1, !tbaa !15
  %conv30.us.i91 = zext i8 %22 to i32
  %23 = load i32, ptr %i_scale31.i78, align 4, !tbaa !104
  %mul32.us.i92 = mul nsw i32 %23, %conv30.us.i91
  %24 = load i32, ptr %i_offset33.i79, align 8, !tbaa !108
  %add34.us.i93 = add nsw i32 %mul32.us.i92, %24
  %tobool.not.i72.us.i94 = icmp ult i32 %add34.us.i93, 256
  %25 = icmp sgt i32 %add34.us.i93, 0
  %shr.i73.us.i95 = sext i1 %25 to i32
  %cond.i74.us.i96 = select i1 %tobool.not.i72.us.i94, i32 %add34.us.i93, i32 %shr.i73.us.i95
  %conv.i75.us.i97 = trunc i32 %cond.i74.us.i96 to i8
  %arrayidx37.us.i98 = getelementptr inbounds i8, ptr %dst.addr.181.us.i86, i64 %indvars.iv.i89
  store i8 %conv.i75.us.i97, ptr %arrayidx37.us.i98, align 1, !tbaa !15
  %indvars.iv.next.i99 = add nuw nsw i64 %indvars.iv.i89, 1
  %exitcond.not.i100 = icmp eq i64 %indvars.iv.next.i99, %wide.trip.count.i83
  br i1 %exitcond.not.i100, label %for.cond23.for.cond.cleanup26_crit_edge.us.i101, label %for.body27.us.i88, !llvm.loop !145

for.cond23.for.cond.cleanup26_crit_edge.us.i101:  ; preds = %for.body27.us.i88
  %inc42.us.i102 = add nuw nsw i32 %y16.082.us.i85, 1
  %add.ptr44.us.i103 = getelementptr inbounds i8, ptr %dst.addr.181.us.i86, i64 %idx.ext43.i80
  %add.ptr46.us.i104 = getelementptr inbounds i8, ptr %src.addr.180.us.i87, i64 %idx.ext45.i81
  %exitcond93.not.i105 = icmp eq i32 %inc42.us.i102, %i_height
  br i1 %exitcond93.not.i105, label %cleanup.loopexit151, label %for.cond23.preheader.us.i84, !llvm.loop !113

for.cond.preheader.i106:                          ; preds = %if.then26
  br i1 %cmp185.i74, label %for.cond2.preheader.lr.ph.i107, label %cleanup

for.cond2.preheader.lr.ph.i107:                   ; preds = %for.cond.preheader.i106
  %cmp383.i108 = icmp sgt i32 %i_width, 0
  %i_scale.i109 = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 3
  %i_offset.i110 = getelementptr inbounds %struct.x264_weight_t, ptr %weight, i64 0, i32 4
  %idx.ext.i111 = sext i32 %20 to i64
  %idx.ext13.i112 = sext i32 %i_src_stride to i64
  br i1 %cmp383.i108, label %for.cond2.preheader.us.preheader.i113, label %cleanup

for.cond2.preheader.us.preheader.i113:            ; preds = %for.cond2.preheader.lr.ph.i107
  %wide.trip.count97.i114 = zext nneg i32 %i_width to i64
  br label %for.cond2.preheader.us.i115

for.cond2.preheader.us.i115:                      ; preds = %for.cond2.for.cond.cleanup4_crit_edge.us.i136, %for.cond2.preheader.us.preheader.i113
  %y.088.us.i116 = phi i32 [ %inc12.us.i137, %for.cond2.for.cond.cleanup4_crit_edge.us.i136 ], [ 0, %for.cond2.preheader.us.preheader.i113 ]
  %dst.addr.087.us.i117 = phi ptr [ %add.ptr.us.i138, %for.cond2.for.cond.cleanup4_crit_edge.us.i136 ], [ %dst, %for.cond2.preheader.us.preheader.i113 ]
  %src.addr.086.us.i118 = phi ptr [ %add.ptr14.us.i139, %for.cond2.for.cond.cleanup4_crit_edge.us.i136 ], [ %add.ptr9, %for.cond2.preheader.us.preheader.i113 ]
  br label %for.body5.us.i119

for.body5.us.i119:                                ; preds = %for.body5.us.i119, %for.cond2.preheader.us.i115
  %indvars.iv94.i120 = phi i64 [ 0, %for.cond2.preheader.us.i115 ], [ %indvars.iv.next95.i134, %for.body5.us.i119 ]
  %arrayidx.us.i121 = getelementptr inbounds i8, ptr %src.addr.086.us.i118, i64 %indvars.iv94.i120
  %26 = load i8, ptr %arrayidx.us.i121, align 1, !tbaa !15
  %conv.us.i122 = zext i8 %26 to i32
  %27 = load i32, ptr %i_scale.i109, align 4, !tbaa !104
  %mul.us.i123 = mul nsw i32 %27, %conv.us.i122
  %28 = load i32, ptr %i_denom.i72, align 16, !tbaa !103
  %sub.us.i124 = add nsw i32 %28, -1
  %shl.us.i125 = shl nuw i32 1, %sub.us.i124
  %add.us.i126 = add nsw i32 %shl.us.i125, %mul.us.i123
  %shr.us.i127 = ashr i32 %add.us.i126, %28
  %29 = load i32, ptr %i_offset.i110, align 8, !tbaa !108
  %add8.us.i128 = add nsw i32 %shr.us.i127, %29
  %tobool.not.i.us.i129 = icmp ult i32 %add8.us.i128, 256
  %30 = icmp sgt i32 %add8.us.i128, 0
  %shr.i.us.i130 = sext i1 %30 to i32
  %cond.i.us.i131 = select i1 %tobool.not.i.us.i129, i32 %add8.us.i128, i32 %shr.i.us.i130
  %conv.i.us.i132 = trunc i32 %cond.i.us.i131 to i8
  %arrayidx10.us.i133 = getelementptr inbounds i8, ptr %dst.addr.087.us.i117, i64 %indvars.iv94.i120
  store i8 %conv.i.us.i132, ptr %arrayidx10.us.i133, align 1, !tbaa !15
  %indvars.iv.next95.i134 = add nuw nsw i64 %indvars.iv94.i120, 1
  %exitcond98.not.i135 = icmp eq i64 %indvars.iv.next95.i134, %wide.trip.count97.i114
  br i1 %exitcond98.not.i135, label %for.cond2.for.cond.cleanup4_crit_edge.us.i136, label %for.body5.us.i119, !llvm.loop !146

for.cond2.for.cond.cleanup4_crit_edge.us.i136:    ; preds = %for.body5.us.i119
  %inc12.us.i137 = add nuw nsw i32 %y.088.us.i116, 1
  %add.ptr.us.i138 = getelementptr inbounds i8, ptr %dst.addr.087.us.i117, i64 %idx.ext.i111
  %add.ptr14.us.i139 = getelementptr inbounds i8, ptr %src.addr.086.us.i118, i64 %idx.ext13.i112
  %exitcond99.not.i140 = icmp eq i32 %inc12.us.i137, %i_height
  br i1 %exitcond99.not.i140, label %cleanup.loopexit, label %for.cond2.preheader.us.i115, !llvm.loop !121

if.else27:                                        ; preds = %if.else
  store i32 %i_src_stride, ptr %i_dst_stride, align 4, !tbaa !13
  br label %cleanup

cleanup.loopexit:                                 ; preds = %for.cond2.for.cond.cleanup4_crit_edge.us.i136
  br label %cleanup

cleanup.loopexit151:                              ; preds = %for.cond23.for.cond.cleanup26_crit_edge.us.i101
  br label %cleanup

cleanup.loopexit152:                              ; preds = %for.cond2.for.cond.cleanup4_crit_edge.us.i
  br label %cleanup

cleanup.loopexit153:                              ; preds = %for.cond23.for.cond.cleanup26_crit_edge.us.i
  br label %cleanup

cleanup:                                          ; preds = %cleanup.loopexit153, %cleanup.loopexit152, %cleanup.loopexit151, %cleanup.loopexit, %for.cond2.preheader.lr.ph.i107, %for.cond.preheader.i106, %for.cond23.preheader.lr.ph.i76, %for.cond17.preheader.i75, %for.cond2.preheader.lr.ph.i, %for.cond.preheader.i, %for.cond23.preheader.lr.ph.i, %for.cond17.preheader.i, %pixel_avg.exit, %if.else27
  %retval.0 = phi ptr [ %add.ptr9, %if.else27 ], [ %dst, %pixel_avg.exit ], [ %dst, %for.cond17.preheader.i ], [ %dst, %for.cond23.preheader.lr.ph.i ], [ %dst, %for.cond.preheader.i ], [ %dst, %for.cond2.preheader.lr.ph.i ], [ %dst, %for.cond17.preheader.i75 ], [ %dst, %for.cond23.preheader.lr.ph.i76 ], [ %dst, %for.cond.preheader.i106 ], [ %dst, %for.cond2.preheader.lr.ph.i107 ], [ %dst, %cleanup.loopexit ], [ %dst, %cleanup.loopexit151 ], [ %dst, %cleanup.loopexit152 ], [ %dst, %cleanup.loopexit153 ]
  ret ptr %retval.0
}

x264_pixel_sad_16x16.print-after.slp-vectorizer.ll

; *** IR Dump After SLPVectorizerPass on x264_pixel_sad_16x16 ***
; Function Attrs: nofree norecurse nosync nounwind memory(read, inaccessiblemem: none) uwtable
define internal i32 @x264_pixel_sad_16x16(ptr nocapture noundef readonly %pix1, i32 noundef %i_stride_pix1, ptr nocapture noundef readonly %pix2, i32 noundef %i_stride_pix2) #4 {
entry:
  %idx.ext = sext i32 %i_stride_pix1 to i64
  %idx.ext8 = sext i32 %i_stride_pix2 to i64
  br label %for.cond1.preheader

for.cond1.preheader:                              ; preds = %entry, %for.cond1.preheader
  %y.025 = phi i32 [ 0, %entry ], [ %inc11, %for.cond1.preheader ]
  %i_sum.024 = phi i32 [ 0, %entry ], [ %op.rdx, %for.cond1.preheader ]
  %pix1.addr.023 = phi ptr [ %pix1, %entry ], [ %add.ptr, %for.cond1.preheader ]
  %pix2.addr.022 = phi ptr [ %pix2, %entry ], [ %add.ptr9, %for.cond1.preheader ]
  %0 = load <16 x i8>, ptr %pix1.addr.023, align 1, !tbaa !14
  %1 = zext <16 x i8> %0 to <16 x i32>
  %2 = load <16 x i8>, ptr %pix2.addr.022, align 1, !tbaa !14
  %3 = zext <16 x i8> %2 to <16 x i32>
  %4 = sub nsw <16 x i32> %1, %3
  %5 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> %4, i1 true)
  %6 = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> %5)
  %op.rdx = add i32 %6, %i_sum.024
  %add.ptr = getelementptr inbounds i8, ptr %pix1.addr.023, i64 %idx.ext
  %add.ptr9 = getelementptr inbounds i8, ptr %pix2.addr.022, i64 %idx.ext8
  %inc11 = add nuw nsw i32 %y.025, 1
  %exitcond.not = icmp eq i32 %inc11, 16
  br i1 %exitcond.not, label %for.cond.cleanup, label %for.cond1.preheader, !llvm.loop !30

for.cond.cleanup:                                 ; preds = %for.cond1.preheader
  ret i32 %op.rdx
}

x264_pixel_sad_16x16.print-before.slp-vectorizer.ll

; *** IR Dump Before SLPVectorizerPass on x264_pixel_sad_16x16 ***
; Function Attrs: nofree norecurse nosync nounwind memory(read, inaccessiblemem: none) uwtable
define internal i32 @x264_pixel_sad_16x16(ptr nocapture noundef readonly %pix1, i32 noundef %i_stride_pix1, ptr nocapture noundef readonly %pix2, i32 noundef %i_stride_pix2) #4 {
entry:
  %idx.ext = sext i32 %i_stride_pix1 to i64
  %idx.ext8 = sext i32 %i_stride_pix2 to i64
  br label %for.cond1.preheader

for.cond1.preheader:                              ; preds = %entry, %for.cond1.preheader
  %y.025 = phi i32 [ 0, %entry ], [ %inc11, %for.cond1.preheader ]
  %i_sum.024 = phi i32 [ 0, %entry ], [ %add.15, %for.cond1.preheader ]
  %pix1.addr.023 = phi ptr [ %pix1, %entry ], [ %add.ptr, %for.cond1.preheader ]
  %pix2.addr.022 = phi ptr [ %pix2, %entry ], [ %add.ptr9, %for.cond1.preheader ]
  %0 = load i8, ptr %pix1.addr.023, align 1, !tbaa !14
  %conv = zext i8 %0 to i32
  %1 = load i8, ptr %pix2.addr.022, align 1, !tbaa !14
  %conv7 = zext i8 %1 to i32
  %sub = sub nsw i32 %conv, %conv7
  %2 = tail call i32 @llvm.abs.i32(i32 %sub, i1 true)
  %add = add nsw i32 %2, %i_sum.024
  %arrayidx.1 = getelementptr inbounds i8, ptr %pix1.addr.023, i64 1
  %3 = load i8, ptr %arrayidx.1, align 1, !tbaa !14
  %conv.1 = zext i8 %3 to i32
  %arrayidx6.1 = getelementptr inbounds i8, ptr %pix2.addr.022, i64 1
  %4 = load i8, ptr %arrayidx6.1, align 1, !tbaa !14
  %conv7.1 = zext i8 %4 to i32
  %sub.1 = sub nsw i32 %conv.1, %conv7.1
  %5 = tail call i32 @llvm.abs.i32(i32 %sub.1, i1 true)
  %add.1 = add nsw i32 %5, %add
  %arrayidx.2 = getelementptr inbounds i8, ptr %pix1.addr.023, i64 2
  %6 = load i8, ptr %arrayidx.2, align 1, !tbaa !14
  %conv.2 = zext i8 %6 to i32
  %arrayidx6.2 = getelementptr inbounds i8, ptr %pix2.addr.022, i64 2
  %7 = load i8, ptr %arrayidx6.2, align 1, !tbaa !14
  %conv7.2 = zext i8 %7 to i32
  %sub.2 = sub nsw i32 %conv.2, %conv7.2
  %8 = tail call i32 @llvm.abs.i32(i32 %sub.2, i1 true)
  %add.2 = add nsw i32 %8, %add.1
  %arrayidx.3 = getelementptr inbounds i8, ptr %pix1.addr.023, i64 3
  %9 = load i8, ptr %arrayidx.3, align 1, !tbaa !14
  %conv.3 = zext i8 %9 to i32
  %arrayidx6.3 = getelementptr inbounds i8, ptr %pix2.addr.022, i64 3
  %10 = load i8, ptr %arrayidx6.3, align 1, !tbaa !14
  %conv7.3 = zext i8 %10 to i32
  %sub.3 = sub nsw i32 %conv.3, %conv7.3
  %11 = tail call i32 @llvm.abs.i32(i32 %sub.3, i1 true)
  %add.3 = add nsw i32 %11, %add.2
  %arrayidx.4 = getelementptr inbounds i8, ptr %pix1.addr.023, i64 4
  %12 = load i8, ptr %arrayidx.4, align 1, !tbaa !14
  %conv.4 = zext i8 %12 to i32
  %arrayidx6.4 = getelementptr inbounds i8, ptr %pix2.addr.022, i64 4
  %13 = load i8, ptr %arrayidx6.4, align 1, !tbaa !14
  %conv7.4 = zext i8 %13 to i32
  %sub.4 = sub nsw i32 %conv.4, %conv7.4
  %14 = tail call i32 @llvm.abs.i32(i32 %sub.4, i1 true)
  %add.4 = add nsw i32 %14, %add.3
  %arrayidx.5 = getelementptr inbounds i8, ptr %pix1.addr.023, i64 5
  %15 = load i8, ptr %arrayidx.5, align 1, !tbaa !14
  %conv.5 = zext i8 %15 to i32
  %arrayidx6.5 = getelementptr inbounds i8, ptr %pix2.addr.022, i64 5
  %16 = load i8, ptr %arrayidx6.5, align 1, !tbaa !14
  %conv7.5 = zext i8 %16 to i32
  %sub.5 = sub nsw i32 %conv.5, %conv7.5
  %17 = tail call i32 @llvm.abs.i32(i32 %sub.5, i1 true)
  %add.5 = add nsw i32 %17, %add.4
  %arrayidx.6 = getelementptr inbounds i8, ptr %pix1.addr.023, i64 6
  %18 = load i8, ptr %arrayidx.6, align 1, !tbaa !14
  %conv.6 = zext i8 %18 to i32
  %arrayidx6.6 = getelementptr inbounds i8, ptr %pix2.addr.022, i64 6
  %19 = load i8, ptr %arrayidx6.6, align 1, !tbaa !14
  %conv7.6 = zext i8 %19 to i32
  %sub.6 = sub nsw i32 %conv.6, %conv7.6
  %20 = tail call i32 @llvm.abs.i32(i32 %sub.6, i1 true)
  %add.6 = add nsw i32 %20, %add.5
  %arrayidx.7 = getelementptr inbounds i8, ptr %pix1.addr.023, i64 7
  %21 = load i8, ptr %arrayidx.7, align 1, !tbaa !14
  %conv.7 = zext i8 %21 to i32
  %arrayidx6.7 = getelementptr inbounds i8, ptr %pix2.addr.022, i64 7
  %22 = load i8, ptr %arrayidx6.7, align 1, !tbaa !14
  %conv7.7 = zext i8 %22 to i32
  %sub.7 = sub nsw i32 %conv.7, %conv7.7
  %23 = tail call i32 @llvm.abs.i32(i32 %sub.7, i1 true)
  %add.7 = add nsw i32 %23, %add.6
  %arrayidx.8 = getelementptr inbounds i8, ptr %pix1.addr.023, i64 8
  %24 = load i8, ptr %arrayidx.8, align 1, !tbaa !14
  %conv.8 = zext i8 %24 to i32
  %arrayidx6.8 = getelementptr inbounds i8, ptr %pix2.addr.022, i64 8
  %25 = load i8, ptr %arrayidx6.8, align 1, !tbaa !14
  %conv7.8 = zext i8 %25 to i32
  %sub.8 = sub nsw i32 %conv.8, %conv7.8
  %26 = tail call i32 @llvm.abs.i32(i32 %sub.8, i1 true)
  %add.8 = add nsw i32 %26, %add.7
  %arrayidx.9 = getelementptr inbounds i8, ptr %pix1.addr.023, i64 9
  %27 = load i8, ptr %arrayidx.9, align 1, !tbaa !14
  %conv.9 = zext i8 %27 to i32
  %arrayidx6.9 = getelementptr inbounds i8, ptr %pix2.addr.022, i64 9
  %28 = load i8, ptr %arrayidx6.9, align 1, !tbaa !14
  %conv7.9 = zext i8 %28 to i32
  %sub.9 = sub nsw i32 %conv.9, %conv7.9
  %29 = tail call i32 @llvm.abs.i32(i32 %sub.9, i1 true)
  %add.9 = add nsw i32 %29, %add.8
  %arrayidx.10 = getelementptr inbounds i8, ptr %pix1.addr.023, i64 10
  %30 = load i8, ptr %arrayidx.10, align 1, !tbaa !14
  %conv.10 = zext i8 %30 to i32
  %arrayidx6.10 = getelementptr inbounds i8, ptr %pix2.addr.022, i64 10
  %31 = load i8, ptr %arrayidx6.10, align 1, !tbaa !14
  %conv7.10 = zext i8 %31 to i32
  %sub.10 = sub nsw i32 %conv.10, %conv7.10
  %32 = tail call i32 @llvm.abs.i32(i32 %sub.10, i1 true)
  %add.10 = add nsw i32 %32, %add.9
  %arrayidx.11 = getelementptr inbounds i8, ptr %pix1.addr.023, i64 11
  %33 = load i8, ptr %arrayidx.11, align 1, !tbaa !14
  %conv.11 = zext i8 %33 to i32
  %arrayidx6.11 = getelementptr inbounds i8, ptr %pix2.addr.022, i64 11
  %34 = load i8, ptr %arrayidx6.11, align 1, !tbaa !14
  %conv7.11 = zext i8 %34 to i32
  %sub.11 = sub nsw i32 %conv.11, %conv7.11
  %35 = tail call i32 @llvm.abs.i32(i32 %sub.11, i1 true)
  %add.11 = add nsw i32 %35, %add.10
  %arrayidx.12 = getelementptr inbounds i8, ptr %pix1.addr.023, i64 12
  %36 = load i8, ptr %arrayidx.12, align 1, !tbaa !14
  %conv.12 = zext i8 %36 to i32
  %arrayidx6.12 = getelementptr inbounds i8, ptr %pix2.addr.022, i64 12
  %37 = load i8, ptr %arrayidx6.12, align 1, !tbaa !14
  %conv7.12 = zext i8 %37 to i32
  %sub.12 = sub nsw i32 %conv.12, %conv7.12
  %38 = tail call i32 @llvm.abs.i32(i32 %sub.12, i1 true)
  %add.12 = add nsw i32 %38, %add.11
  %arrayidx.13 = getelementptr inbounds i8, ptr %pix1.addr.023, i64 13
  %39 = load i8, ptr %arrayidx.13, align 1, !tbaa !14
  %conv.13 = zext i8 %39 to i32
  %arrayidx6.13 = getelementptr inbounds i8, ptr %pix2.addr.022, i64 13
  %40 = load i8, ptr %arrayidx6.13, align 1, !tbaa !14
  %conv7.13 = zext i8 %40 to i32
  %sub.13 = sub nsw i32 %conv.13, %conv7.13
  %41 = tail call i32 @llvm.abs.i32(i32 %sub.13, i1 true)
  %add.13 = add nsw i32 %41, %add.12
  %arrayidx.14 = getelementptr inbounds i8, ptr %pix1.addr.023, i64 14
  %42 = load i8, ptr %arrayidx.14, align 1, !tbaa !14
  %conv.14 = zext i8 %42 to i32
  %arrayidx6.14 = getelementptr inbounds i8, ptr %pix2.addr.022, i64 14
  %43 = load i8, ptr %arrayidx6.14, align 1, !tbaa !14
  %conv7.14 = zext i8 %43 to i32
  %sub.14 = sub nsw i32 %conv.14, %conv7.14
  %44 = tail call i32 @llvm.abs.i32(i32 %sub.14, i1 true)
  %add.14 = add nsw i32 %44, %add.13
  %arrayidx.15 = getelementptr inbounds i8, ptr %pix1.addr.023, i64 15
  %45 = load i8, ptr %arrayidx.15, align 1, !tbaa !14
  %conv.15 = zext i8 %45 to i32
  %arrayidx6.15 = getelementptr inbounds i8, ptr %pix2.addr.022, i64 15
  %46 = load i8, ptr %arrayidx6.15, align 1, !tbaa !14
  %conv7.15 = zext i8 %46 to i32
  %sub.15 = sub nsw i32 %conv.15, %conv7.15
  %47 = tail call i32 @llvm.abs.i32(i32 %sub.15, i1 true)
  %add.15 = add nsw i32 %47, %add.14
  %add.ptr = getelementptr inbounds i8, ptr %pix1.addr.023, i64 %idx.ext
  %add.ptr9 = getelementptr inbounds i8, ptr %pix2.addr.022, i64 %idx.ext8
  %inc11 = add nuw nsw i32 %y.025, 1
  %exitcond.not = icmp eq i32 %inc11, 16
  br i1 %exitcond.not, label %for.cond.cleanup, label %for.cond1.preheader, !llvm.loop !30

for.cond.cleanup:                                 ; preds = %for.cond1.preheader
  ret i32 %add.15
}