1 ; RUN: llc -regalloc=greedy -arm-atomic-cfg-tidy=0 < %s | FileCheck %s
3 ; LSR shouldn't introduce more induction variables than needed, increasing
4 ; register pressure and therefore spilling. There is more room for improvement
7 ; CHECK: sub sp, #{{40|36|32|28|24}}
13 target datalayout = "e-p:32:32:32-i1:8:32-i8:8:32-i16:16:32-i32:32:32-i64:32:32-f32:32:32-f64:32:32-v64:32:64-v128:32:128-a0:0:32-n32"
14 target triple = "thumbv7-apple-ios"
16 %struct.partition_entry = type { i32, i32, i64, i64 }
18 define i32 @partition_overlap_check(ptr nocapture %part, i32 %num_entries) nounwind readonly optsize ssp "frame-pointer"="all" {
20 %cmp79 = icmp sgt i32 %num_entries, 0
21 br i1 %cmp79, label %outer.loop, label %for.end72
23 outer.loop: ; preds = %for.inc69, %entry
24 %overlap.081 = phi i32 [ %overlap.4, %for.inc69 ], [ 0, %entry ]
25 %0 = phi i32 [ %inc71, %for.inc69 ], [ 0, %entry ]
26 %offset = getelementptr %struct.partition_entry, ptr %part, i32 %0, i32 2
27 %len = getelementptr %struct.partition_entry, ptr %part, i32 %0, i32 3
28 %tmp5 = load i64, ptr %offset, align 4
29 %tmp15 = load i64, ptr %len, align 4
30 %add = add nsw i64 %tmp15, %tmp5
33 inner.loop: ; preds = %for.inc, %outer.loop
34 %overlap.178 = phi i32 [ %overlap.081, %outer.loop ], [ %overlap.4, %for.inc ]
35 %1 = phi i32 [ 0, %outer.loop ], [ %inc, %for.inc ]
36 %cmp23 = icmp eq i32 %0, %1
37 br i1 %cmp23, label %for.inc, label %if.end
39 if.end: ; preds = %inner.loop
40 %len39 = getelementptr %struct.partition_entry, ptr %part, i32 %1, i32 3
41 %offset28 = getelementptr %struct.partition_entry, ptr %part, i32 %1, i32 2
42 %tmp29 = load i64, ptr %offset28, align 4
43 %tmp40 = load i64, ptr %len39, align 4
44 %add41 = add nsw i64 %tmp40, %tmp29
45 %cmp44 = icmp sge i64 %tmp29, %tmp5
46 %cmp47 = icmp slt i64 %tmp29, %add
47 %or.cond = and i1 %cmp44, %cmp47
48 %overlap.2 = select i1 %or.cond, i32 1, i32 %overlap.178
49 %cmp52 = icmp sle i64 %add41, %add
50 %cmp56 = icmp sgt i64 %add41, %tmp5
51 %or.cond74 = and i1 %cmp52, %cmp56
52 %overlap.3 = select i1 %or.cond74, i32 1, i32 %overlap.2
53 %cmp61 = icmp sgt i64 %tmp29, %tmp5
54 %cmp65 = icmp slt i64 %add41, %add
55 %or.cond75 = or i1 %cmp61, %cmp65
56 br i1 %or.cond75, label %for.inc, label %if.then66
58 if.then66: ; preds = %if.end
61 for.inc: ; preds = %if.end, %if.then66, %inner.loop
62 %overlap.4 = phi i32 [ %overlap.178, %inner.loop ], [ 1, %if.then66 ], [ %overlap.3, %if.end ]
63 %inc = add nsw i32 %1, 1
64 %exitcond = icmp eq i32 %inc, %num_entries
65 br i1 %exitcond, label %for.inc69, label %inner.loop
67 for.inc69: ; preds = %for.inc
68 %inc71 = add nsw i32 %0, 1
69 %exitcond83 = icmp eq i32 %inc71, %num_entries
70 br i1 %exitcond83, label %for.end72, label %outer.loop
72 for.end72: ; preds = %for.inc69, %entry
73 %overlap.0.lcssa = phi i32 [ 0, %entry ], [ %overlap.4, %for.inc69 ]
74 ret i32 %overlap.0.lcssa