xref: /llvm-project/llvm/test/CodeGen/X86/uadd_sat.ll (revision 7b3bbd83c0c24087072ec5b22a76799ab31f87d5)
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -mtriple=i686 -mattr=cmov | FileCheck %s --check-prefix=X86
3; RUN: llc < %s -mtriple=x86_64-linux | FileCheck %s --check-prefix=X64
4
5declare i4 @llvm.uadd.sat.i4(i4, i4)
6declare i8 @llvm.uadd.sat.i8(i8, i8)
7declare i16 @llvm.uadd.sat.i16(i16, i16)
8declare i32 @llvm.uadd.sat.i32(i32, i32)
9declare i64 @llvm.uadd.sat.i64(i64, i64)
10declare <4 x i32> @llvm.uadd.sat.v4i32(<4 x i32>, <4 x i32>)
11
12define i32 @func(i32 %x, i32 %y) nounwind {
13; X86-LABEL: func:
14; X86:       # %bb.0:
15; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
16; X86-NEXT:    addl {{[0-9]+}}(%esp), %ecx
17; X86-NEXT:    movl $-1, %eax
18; X86-NEXT:    cmovael %ecx, %eax
19; X86-NEXT:    retl
20;
21; X64-LABEL: func:
22; X64:       # %bb.0:
23; X64-NEXT:    addl %esi, %edi
24; X64-NEXT:    movl $-1, %eax
25; X64-NEXT:    cmovael %edi, %eax
26; X64-NEXT:    retq
27  %tmp = call i32 @llvm.uadd.sat.i32(i32 %x, i32 %y)
28  ret i32 %tmp
29}
30
31define i64 @func2(i64 %x, i64 %y) nounwind {
32; X86-LABEL: func2:
33; X86:       # %bb.0:
34; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
35; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
36; X86-NEXT:    addl {{[0-9]+}}(%esp), %eax
37; X86-NEXT:    adcl {{[0-9]+}}(%esp), %edx
38; X86-NEXT:    movl $-1, %ecx
39; X86-NEXT:    cmovbl %ecx, %edx
40; X86-NEXT:    cmovbl %ecx, %eax
41; X86-NEXT:    retl
42;
43; X64-LABEL: func2:
44; X64:       # %bb.0:
45; X64-NEXT:    addq %rsi, %rdi
46; X64-NEXT:    movq $-1, %rax
47; X64-NEXT:    cmovaeq %rdi, %rax
48; X64-NEXT:    retq
49  %tmp = call i64 @llvm.uadd.sat.i64(i64 %x, i64 %y)
50  ret i64 %tmp
51}
52
53define zeroext i16 @func16(i16 zeroext %x, i16 zeroext %y) nounwind {
54; X86-LABEL: func16:
55; X86:       # %bb.0:
56; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %ecx
57; X86-NEXT:    addw {{[0-9]+}}(%esp), %cx
58; X86-NEXT:    movl $65535, %eax # imm = 0xFFFF
59; X86-NEXT:    cmovael %ecx, %eax
60; X86-NEXT:    # kill: def $ax killed $ax killed $eax
61; X86-NEXT:    retl
62;
63; X64-LABEL: func16:
64; X64:       # %bb.0:
65; X64-NEXT:    addw %si, %di
66; X64-NEXT:    movl $65535, %eax # imm = 0xFFFF
67; X64-NEXT:    cmovael %edi, %eax
68; X64-NEXT:    # kill: def $ax killed $ax killed $eax
69; X64-NEXT:    retq
70  %tmp = call i16 @llvm.uadd.sat.i16(i16 %x, i16 %y)
71  ret i16 %tmp
72}
73
74define zeroext i8 @func8(i8 zeroext %x, i8 zeroext %y) nounwind {
75; X86-LABEL: func8:
76; X86:       # %bb.0:
77; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
78; X86-NEXT:    addb {{[0-9]+}}(%esp), %al
79; X86-NEXT:    movzbl %al, %ecx
80; X86-NEXT:    movl $255, %eax
81; X86-NEXT:    cmovael %ecx, %eax
82; X86-NEXT:    # kill: def $al killed $al killed $eax
83; X86-NEXT:    retl
84;
85; X64-LABEL: func8:
86; X64:       # %bb.0:
87; X64-NEXT:    addb %sil, %dil
88; X64-NEXT:    movzbl %dil, %ecx
89; X64-NEXT:    movl $255, %eax
90; X64-NEXT:    cmovael %ecx, %eax
91; X64-NEXT:    # kill: def $al killed $al killed $eax
92; X64-NEXT:    retq
93  %tmp = call i8 @llvm.uadd.sat.i8(i8 %x, i8 %y)
94  ret i8 %tmp
95}
96
97define zeroext i4 @func3(i4 zeroext %x, i4 zeroext %y) nounwind {
98; X86-LABEL: func3:
99; X86:       # %bb.0:
100; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
101; X86-NEXT:    addb {{[0-9]+}}(%esp), %al
102; X86-NEXT:    movzbl %al, %ecx
103; X86-NEXT:    cmpb $15, %al
104; X86-NEXT:    movl $15, %eax
105; X86-NEXT:    cmovbl %ecx, %eax
106; X86-NEXT:    movzbl %al, %eax
107; X86-NEXT:    retl
108;
109; X64-LABEL: func3:
110; X64:       # %bb.0:
111; X64-NEXT:    addb %sil, %dil
112; X64-NEXT:    movzbl %dil, %eax
113; X64-NEXT:    cmpb $15, %al
114; X64-NEXT:    movl $15, %ecx
115; X64-NEXT:    cmovbl %eax, %ecx
116; X64-NEXT:    movzbl %cl, %eax
117; X64-NEXT:    retq
118  %tmp = call i4 @llvm.uadd.sat.i4(i4 %x, i4 %y)
119  ret i4 %tmp
120}
121
122define <4 x i32> @vec(<4 x i32> %x, <4 x i32> %y) nounwind {
123; X86-LABEL: vec:
124; X86:       # %bb.0:
125; X86-NEXT:    pushl %ebx
126; X86-NEXT:    pushl %edi
127; X86-NEXT:    pushl %esi
128; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
129; X86-NEXT:    movl {{[0-9]+}}(%esp), %edi
130; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
131; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
132; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
133; X86-NEXT:    addl {{[0-9]+}}(%esp), %ecx
134; X86-NEXT:    movl $-1, %ebx
135; X86-NEXT:    cmovbl %ebx, %ecx
136; X86-NEXT:    addl {{[0-9]+}}(%esp), %edx
137; X86-NEXT:    cmovbl %ebx, %edx
138; X86-NEXT:    addl {{[0-9]+}}(%esp), %esi
139; X86-NEXT:    cmovbl %ebx, %esi
140; X86-NEXT:    addl {{[0-9]+}}(%esp), %edi
141; X86-NEXT:    cmovbl %ebx, %edi
142; X86-NEXT:    movl %edi, 12(%eax)
143; X86-NEXT:    movl %esi, 8(%eax)
144; X86-NEXT:    movl %edx, 4(%eax)
145; X86-NEXT:    movl %ecx, (%eax)
146; X86-NEXT:    popl %esi
147; X86-NEXT:    popl %edi
148; X86-NEXT:    popl %ebx
149; X86-NEXT:    retl $4
150;
151; X64-LABEL: vec:
152; X64:       # %bb.0:
153; X64-NEXT:    movdqa {{.*#+}} xmm2 = [2147483648,2147483648,2147483648,2147483648]
154; X64-NEXT:    movdqa %xmm0, %xmm3
155; X64-NEXT:    pxor %xmm2, %xmm3
156; X64-NEXT:    paddd %xmm1, %xmm0
157; X64-NEXT:    pxor %xmm0, %xmm2
158; X64-NEXT:    pcmpgtd %xmm2, %xmm3
159; X64-NEXT:    por %xmm3, %xmm0
160; X64-NEXT:    retq
161  %tmp = call <4 x i32> @llvm.uadd.sat.v4i32(<4 x i32> %x, <4 x i32> %y)
162  ret <4 x i32> %tmp
163}
164