xref: /llvm-project/llvm/test/CodeGen/ARM/intrinsics-overflow.ll (revision 854980341b5fecd5b19b39dc802a20edb23a67dc)
1; RUN: llc < %s -mtriple=arm-linux -mcpu=generic -verify-machineinstrs | FileCheck %s --check-prefix=CHECK --check-prefix=ARM
2; RUN: llc < %s -mtriple=thumbv6m-eabi -verify-machineinstrs | FileCheck %s --check-prefix=CHECK --check-prefix=THUMBV6
3; RUN: llc < %s -mtriple=thumbv7-eabi -verify-machineinstrs | FileCheck %s --check-prefix=CHECK --check-prefix=THUMBV7
4
5define i32 @uadd_overflow(i32 %a, i32 %b) #0 {
6  %sadd = tail call { i32, i1 } @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
7  %1 = extractvalue { i32, i1 } %sadd, 1
8  %2 = zext i1 %1 to i32
9  ret i32 %2
10
11  ; CHECK-LABEL: uadd_overflow:
12
13  ; ARM: adds r[[R0:[0-9]+]], r[[R0]], r[[R1:[0-9]+]]
14  ; ARM: mov r[[R2:[0-9]+]], #0
15  ; ARM: adc r[[R0]], r[[R2]], #0
16
17  ; THUMBV6: movs    r[[R2:[0-9]+]], #0
18  ; THUMBV6: adds    r[[R0:[0-9]+]], r[[R0]], r[[R1:[0-9]+]]
19  ; THUMBV6: adcs    r[[R2]], r[[R2]]
20  ; THUMBV6: mov     r[[R0]], r[[R2]]
21
22  ; THUMBV7: adds  r[[R0:[0-9]+]], r[[R0]], r[[R1:[0-9]+]]
23  ; THUMBV7: mov.w r[[R2:[0-9]+]], #0
24  ; THUMBV7: adc   r[[R0]], r[[R2]], #0
25}
26
27
28define i32 @sadd_overflow(i32 %a, i32 %b) #0 {
29  %sadd = tail call { i32, i1 } @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
30  %1 = extractvalue { i32, i1 } %sadd, 1
31  %2 = zext i1 %1 to i32
32  ret i32 %2
33
34  ; CHECK-LABEL: sadd_overflow:
35
36  ; ARM: add r[[R2:[0-9]+]], r[[R0:[0-9]+]], r[[R1:[0-9]+]]
37  ; ARM: mov r[[R1]], #1
38  ; ARM: cmp r[[R2]], r[[R0]]
39  ; ARM: movvc r[[R1]], #0
40
41  ; THUMBV6: mov  r[[R2:[0-9]+]], r[[R0:[0-9]+]]
42  ; THUMBV6: adds r[[R3:[0-9]+]], r[[R2]], r[[R1:[0-9]+]]
43  ; THUMBV6: movs r[[R0]], #0
44  ; THUMBV6: movs r[[R1]], #1
45  ; THUMBV6: cmp  r[[R3]], r[[R2]]
46  ; THUMBV6: bvc  .L[[LABEL:.*]]
47  ; THUMBV6: mov  r[[R0]], r[[R1]]
48  ; THUMBV6: .L[[LABEL]]:
49
50  ; THUMBV7: movs  r[[R1]], #1
51  ; THUMBV7: cmp   r[[R2]], r[[R0]]
52  ; THUMBV7: it    vc
53  ; THUMBV7: movvc r[[R1]], #0
54  ; THUMBV7: mov   r[[R0]], r[[R1]]
55}
56
57define i32 @usub_overflow(i32 %a, i32 %b) #0 {
58  %sadd = tail call { i32, i1 } @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
59  %1 = extractvalue { i32, i1 } %sadd, 1
60  %2 = zext i1 %1 to i32
61  ret i32 %2
62
63  ; CHECK-LABEL: usub_overflow:
64
65  ; ARM: subs    r[[R0:[0-9]+]], r[[R0]], r[[R1:[0-9]+]]
66  ; ARM: mov     r[[R2:[0-9]+]], #0
67  ; ARM: adc     r[[R0]], r[[R2]], #0
68  ; ARM: rsb     r[[R0]], r[[R0]], #1
69
70  ; THUMBV6: movs    r[[R2:[0-9]+]], #0
71  ; THUMBV6: subs    r[[R0:[0-9]+]], r[[R0]], r[[R1:[0-9]+]]
72  ; THUMBV6: adcs    r[[R2]], r[[R2]]
73  ; THUMBV6: movs    r[[R0]], #1
74  ; THUMBV6: subs    r[[R0]], r[[R0]], r[[R2]]
75
76  ; THUMBV7: subs    r[[R0:[0-9]+]], r[[R0]], r[[R1:[0-9]+]]
77  ; THUMBV7: mov.w   r[[R2:[0-9]+]], #0
78  ; THUMBV7: adc     r[[R0]], r[[R2]], #0
79  ; THUMBV7: rsb.w   r[[R0]], r[[R0]], #1
80
81  ; We should know that the overflow is just 1 bit,
82  ; no need to clear any other bit
83  ; CHECK-NOT: and
84}
85
86define i32 @ssub_overflow(i32 %a, i32 %b) #0 {
87  %sadd = tail call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
88  %1 = extractvalue { i32, i1 } %sadd, 1
89  %2 = zext i1 %1 to i32
90  ret i32 %2
91
92  ; CHECK-LABEL: ssub_overflow:
93
94  ; ARM: mov r[[R2]], #1
95  ; ARM: cmp r[[R0]], r[[R1]]
96  ; ARM: movvc r[[R2]], #0
97
98  ; THUMBV6: movs    r[[R0]], #0
99  ; THUMBV6: movs    r[[R3:[0-9]+]], #1
100  ; THUMBV6: cmp     r[[R2]], r[[R1:[0-9]+]]
101  ; THUMBV6: bvc     .L[[LABEL:.*]]
102  ; THUMBV6: mov     r[[R0]], r[[R3]]
103  ; THUMBV6: .L[[LABEL]]:
104
105  ; THUMBV7: movs  r[[R2:[0-9]+]], #1
106  ; THUMBV7: cmp   r[[R0:[0-9]+]], r[[R1:[0-9]+]]
107  ; THUMBV7: it    vc
108  ; THUMBV7: movvc r[[R2]], #0
109  ; THUMBV7: mov   r[[R0]], r[[R2]]
110}
111
112declare { i32, i1 } @llvm.uadd.with.overflow.i32(i32, i32) #1
113declare { i32, i1 } @llvm.sadd.with.overflow.i32(i32, i32) #2
114declare { i32, i1 } @llvm.usub.with.overflow.i32(i32, i32) #3
115declare { i32, i1 } @llvm.ssub.with.overflow.i32(i32, i32) #4
116