1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -disable-lsr -ppc-asm-full-reg-names -verify-machineinstrs \
3; RUN:   -mtriple=powerpc64le-unknown-linux-gnu -mcpu=pwr9 < %s | FileCheck %s
4
5; long long foo(char *p, int n, int count) {
6;   int j = 0;
7;   long long sum = 0;
8;   for (int i = 0; i < n; i++) {
9;     sum += *(unsigned long long *)(p + j + 5);
10;     sum += *(unsigned long long *)(p + j + 9);
11;     j += count;
12;   }
13;   return sum;
14; }
15
16define i64 @foo(i8* %p, i32 signext %n, i32 signext %count) {
17; CHECK-LABEL: foo:
18; CHECK:       # %bb.0: # %entry
19; CHECK-NEXT:    cmpwi r4, 1
20; CHECK-NEXT:    blt cr0, .LBB0_4
21; CHECK-NEXT:  # %bb.1: # %for.body.preheader
22; CHECK-NEXT:    addi r6, r3, 5
23; CHECK-NEXT:    addi r3, r4, -1
24; CHECK-NEXT:    extsw r5, r5
25; CHECK-NEXT:    clrldi r3, r3, 32
26; CHECK-NEXT:    addi r3, r3, 1
27; CHECK-NEXT:    mtctr r3
28; CHECK-NEXT:    li r3, 0
29; CHECK-NEXT:    .p2align 5
30; CHECK-NEXT:  .LBB0_2: # %for.body
31; CHECK-NEXT:    #
32; CHECK-NEXT:    ld r4, 0(r6)
33; CHECK-NEXT:    add r3, r4, r3
34; CHECK-NEXT:    ld r4, 4(r6)
35; CHECK-NEXT:    add r6, r6, r5
36; CHECK-NEXT:    add r3, r3, r4
37; CHECK-NEXT:    bdnz .LBB0_2
38; CHECK-NEXT:  # %bb.3: # %for.cond.cleanup
39; CHECK-NEXT:    blr
40; CHECK-NEXT:  .LBB0_4:
41; CHECK-NEXT:    li r3, 0
42; CHECK-NEXT:    blr
43entry:
44  %cmp16 = icmp sgt i32 %n, 0
45  br i1 %cmp16, label %for.body.preheader, label %for.cond.cleanup
46
47for.body.preheader:                               ; preds = %entry
48  %0 = sext i32 %count to i64
49  br label %for.body
50
51for.cond.cleanup:                                 ; preds = %for.body, %entry
52  %sum.0.lcssa = phi i64 [ 0, %entry ], [ %add5, %for.body ]
53  ret i64 %sum.0.lcssa
54
55for.body:                                         ; preds = %for.body.preheader, %for.body
56  %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ]
57  %i.019 = phi i32 [ 0, %for.body.preheader ], [ %inc, %for.body ]
58  %sum.018 = phi i64 [ 0, %for.body.preheader ], [ %add5, %for.body ]
59  %add.ptr = getelementptr inbounds i8, i8* %p, i64 %indvars.iv
60  %add.ptr1 = getelementptr inbounds i8, i8* %add.ptr, i64 5
61  %1 = bitcast i8* %add.ptr1 to i64*
62  %2 = load i64, i64* %1, align 8
63  %add = add i64 %2, %sum.018
64  %add.ptr4 = getelementptr inbounds i8, i8* %add.ptr, i64 9
65  %3 = bitcast i8* %add.ptr4 to i64*
66  %4 = load i64, i64* %3, align 8
67  %add5 = add i64 %add, %4
68  %indvars.iv.next = add nsw i64 %indvars.iv, %0
69  %inc = add nuw nsw i32 %i.019, 1
70  %exitcond.not = icmp eq i32 %inc, %n
71  br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
72}
73
74; char foo1(char *p, int n, int count) {
75;   int j = 0;
76;   char sum = 0;
77;   for (int i = 0; i < n; i++) {
78;     sum += *(p + j + 1000);
79;     j += count;
80;    }
81;   return sum;
82; }
83
84define zeroext i8 @foo1(i8* %p, i32 signext %n, i32 signext %count) {
85; CHECK-LABEL: foo1:
86; CHECK:       # %bb.0: # %entry
87; CHECK-NEXT:    cmpwi r4, 1
88; CHECK-NEXT:    blt cr0, .LBB1_4
89; CHECK-NEXT:  # %bb.1: # %for.body.preheader
90; CHECK-NEXT:    extsw r5, r5
91; CHECK-NEXT:    sub r3, r3, r5
92; CHECK-NEXT:    addi r6, r3, 1000
93; CHECK-NEXT:    addi r3, r4, -1
94; CHECK-NEXT:    clrldi r3, r3, 32
95; CHECK-NEXT:    addi r3, r3, 1
96; CHECK-NEXT:    mtctr r3
97; CHECK-NEXT:    li r3, 0
98; CHECK-NEXT:    .p2align 4
99; CHECK-NEXT:  .LBB1_2: # %for.body
100; CHECK-NEXT:    #
101; CHECK-NEXT:    lbzux r4, r6, r5
102; CHECK-NEXT:    add r3, r4, r3
103; CHECK-NEXT:    bdnz .LBB1_2
104; CHECK-NEXT:  # %bb.3: # %for.cond.cleanup
105; CHECK-NEXT:    clrldi r3, r3, 56
106; CHECK-NEXT:    blr
107; CHECK-NEXT:  .LBB1_4:
108; CHECK-NEXT:    li r3, 0
109; CHECK-NEXT:    blr
110entry:
111  %cmp10 = icmp sgt i32 %n, 0
112  br i1 %cmp10, label %for.body.preheader, label %for.cond.cleanup
113
114for.body.preheader:                               ; preds = %entry
115  %0 = sext i32 %count to i64
116  %add.ptr = getelementptr inbounds i8, i8* %p, i64 1000
117  br label %for.body
118
119for.cond.cleanup:                                 ; preds = %for.body, %entry
120  %sum.0.lcssa = phi i8 [ 0, %entry ], [ %add, %for.body ]
121  ret i8 %sum.0.lcssa
122
123for.body:                                         ; preds = %for.body.preheader, %for.body
124  %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ]
125  %i.013 = phi i32 [ 0, %for.body.preheader ], [ %inc, %for.body ]
126  %sum.012 = phi i8 [ 0, %for.body.preheader ], [ %add, %for.body ]
127  %add.ptr1 = getelementptr inbounds i8, i8* %add.ptr, i64 %indvars.iv
128  %1 = load i8, i8* %add.ptr1, align 1
129  %add = add i8 %1, %sum.012
130  %indvars.iv.next = add nsw i64 %indvars.iv, %0
131  %inc = add nuw nsw i32 %i.013, 1
132  %exitcond.not = icmp eq i32 %inc, %n
133  br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
134}
135