1; RUN: llc < %s -mtriple=thumbv7-apple-darwin -mcpu=cortex-a8 -relocation-model=dynamic-no-pic -disable-fp-elim | FileCheck %s
2; RUN: llc < %s -mtriple=thumbv7-apple-darwin -mcpu=cortex-a8 -relocation-model=pic -disable-fp-elim | FileCheck %s --check-prefix=PIC
3; rdar://7353541
4; rdar://7354376
5
6@GV = external global i32                         ; <i32*> [#uses=2]
7
8define void @t1(i32* nocapture %vals, i32 %c) nounwind {
9entry:
10; CHECK: t1:
11; CHECK: bxeq lr
12
13  %0 = icmp eq i32 %c, 0                          ; <i1> [#uses=1]
14  br i1 %0, label %return, label %bb.nph
15
16bb.nph:                                           ; preds = %entry
17; CHECK: movw r[[R2:[0-9]+]], :lower16:L_GV$non_lazy_ptr
18; CHECK: movt r[[R2]], :upper16:L_GV$non_lazy_ptr
19; CHECK: ldr{{(.w)?}} r[[R2b:[0-9]+]], [r[[R2]]
20; CHECK: ldr{{.*}}, [r[[R2b]]
21; CHECK: LBB0_
22; CHECK-NOT: LCPI0_0:
23
24; PIC: movw r[[R2:[0-9]+]], :lower16:(L_GV$non_lazy_ptr-(LPC0_0+4))
25; PIC: movt r[[R2]], :upper16:(L_GV$non_lazy_ptr-(LPC0_0+4))
26; PIC: add r[[R2]], pc
27; PIC: ldr{{(.w)?}} r[[R2b:[0-9]+]], [r[[R2]]
28; PIC: ldr{{.*}}, [r[[R2b]]
29; PIC: LBB0_
30; PIC-NOT: LCPI0_0:
31; PIC: .section
32  %.pre = load i32* @GV, align 4                  ; <i32> [#uses=1]
33  br label %bb
34
35bb:                                               ; preds = %bb, %bb.nph
36  %1 = phi i32 [ %.pre, %bb.nph ], [ %3, %bb ]    ; <i32> [#uses=1]
37  %i.03 = phi i32 [ 0, %bb.nph ], [ %4, %bb ]     ; <i32> [#uses=2]
38  %scevgep = getelementptr i32* %vals, i32 %i.03  ; <i32*> [#uses=1]
39  %2 = load i32* %scevgep, align 4                ; <i32> [#uses=1]
40  %3 = add nsw i32 %1, %2                         ; <i32> [#uses=2]
41  store i32 %3, i32* @GV, align 4
42  %4 = add i32 %i.03, 1                           ; <i32> [#uses=2]
43  %exitcond = icmp eq i32 %4, %c                  ; <i1> [#uses=1]
44  br i1 %exitcond, label %return, label %bb
45
46return:                                           ; preds = %bb, %entry
47  ret void
48}
49
50; rdar://8001136
51define void @t2(i8* %ptr1, i8* %ptr2) nounwind {
52entry:
53; CHECK: t2:
54; CHECK: mov.w [[R3:r[0-9]+]], #1065353216
55; CHECK: vdup.32 q{{.*}}, [[R3]]
56  br i1 undef, label %bb1, label %bb2
57
58bb1:
59; CHECK-NEXT: %bb1
60  %indvar = phi i32 [ %indvar.next, %bb1 ], [ 0, %entry ]
61  %tmp1 = shl i32 %indvar, 2
62  %gep1 = getelementptr i8* %ptr1, i32 %tmp1
63  %tmp2 = call <4 x float> @llvm.arm.neon.vld1.v4f32(i8* %gep1, i32 1)
64  %tmp3 = call <4 x float> @llvm.arm.neon.vmaxs.v4f32(<4 x float> <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, <4 x float> %tmp2)
65  %gep2 = getelementptr i8* %ptr2, i32 %tmp1
66  call void @llvm.arm.neon.vst1.v4f32(i8* %gep2, <4 x float> %tmp3, i32 1)
67  %indvar.next = add i32 %indvar, 1
68  %cond = icmp eq i32 %indvar.next, 10
69  br i1 %cond, label %bb2, label %bb1
70
71bb2:
72  ret void
73}
74
75; CHECK-NOT: LCPI1_0:
76
77declare <4 x float> @llvm.arm.neon.vld1.v4f32(i8*, i32) nounwind readonly
78
79declare void @llvm.arm.neon.vst1.v4f32(i8*, <4 x float>, i32) nounwind
80
81declare <4 x float> @llvm.arm.neon.vmaxs.v4f32(<4 x float>, <4 x float>) nounwind readnone
82
83; rdar://8241368
84; isel should not fold immediate into eor's which would have prevented LICM.
85define zeroext i16 @t3(i8 zeroext %data, i16 zeroext %crc) nounwind readnone {
86; CHECK: t3:
87bb.nph:
88; CHECK: bb.nph
89; CHECK: movw {{(r[0-9])|(lr)}}, #32768
90; CHECK: movs {{(r[0-9]+)|(lr)}}, #0
91; CHECK: movw [[REGISTER:(r[0-9]+)|(lr)]], #16386
92; CHECK: movw {{(r[0-9]+)|(lr)}}, #65534
93; CHECK: movt {{(r[0-9]+)|(lr)}}, #65535
94  br label %bb
95
96bb:                                               ; preds = %bb, %bb.nph
97; CHECK: bb
98; CHECK: eor.w {{(r[0-9])|(lr)}}, {{(r[0-9])|(lr)}}, [[REGISTER]]
99; CHECK: eor.w
100; CHECK-NOT: eor
101; CHECK: and
102  %data_addr.013 = phi i8 [ %data, %bb.nph ], [ %8, %bb ] ; <i8> [#uses=2]
103  %crc_addr.112 = phi i16 [ %crc, %bb.nph ], [ %crc_addr.2, %bb ] ; <i16> [#uses=3]
104  %i.011 = phi i8 [ 0, %bb.nph ], [ %7, %bb ]     ; <i8> [#uses=1]
105  %0 = trunc i16 %crc_addr.112 to i8              ; <i8> [#uses=1]
106  %1 = xor i8 %data_addr.013, %0                  ; <i8> [#uses=1]
107  %2 = and i8 %1, 1                               ; <i8> [#uses=1]
108  %3 = icmp eq i8 %2, 0                           ; <i1> [#uses=2]
109  %4 = xor i16 %crc_addr.112, 16386               ; <i16> [#uses=1]
110  %crc_addr.0 = select i1 %3, i16 %crc_addr.112, i16 %4 ; <i16> [#uses=1]
111  %5 = lshr i16 %crc_addr.0, 1                    ; <i16> [#uses=2]
112  %6 = or i16 %5, -32768                          ; <i16> [#uses=1]
113  %crc_addr.2 = select i1 %3, i16 %5, i16 %6      ; <i16> [#uses=2]
114  %7 = add i8 %i.011, 1                           ; <i8> [#uses=2]
115  %8 = lshr i8 %data_addr.013, 1                  ; <i8> [#uses=1]
116  %exitcond = icmp eq i8 %7, 8                    ; <i1> [#uses=1]
117  br i1 %exitcond, label %bb8, label %bb
118
119bb8:                                              ; preds = %bb
120  ret i16 %crc_addr.2
121}
122