xref: /llvm-project/llvm/test/CodeGen/X86/fold-tied-op.ll (revision 7b3bbd83c0c24087072ec5b22a76799ab31f87d5) 
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -verify-machineinstrs -mtriple=i386--netbsd < %s | FileCheck %s
3
4; Regression test for http://reviews.llvm.org/D5701
5
6%struct.XXH_state64_t = type { i32, i32, i64, i64, i64 }
7
8@a = common global i32 0, align 4
9@b = common global i64 0, align 8
10
11; Function Attrs: nounwind uwtable
12define i64 @fn1() #0 {
13; CHECK-LABEL: fn1:
14; CHECK:       # %bb.0: # %entry
15; CHECK-NEXT:    pushl %ebp
16; CHECK-NEXT:    .cfi_def_cfa_offset 8
17; CHECK-NEXT:    .cfi_offset %ebp, -8
18; CHECK-NEXT:    movl %esp, %ebp
19; CHECK-NEXT:    .cfi_def_cfa_register %ebp
20; CHECK-NEXT:    pushl %ebx
21; CHECK-NEXT:    pushl %edi
22; CHECK-NEXT:    pushl %esi
23; CHECK-NEXT:    subl $12, %esp
24; CHECK-NEXT:    .cfi_offset %esi, -20
25; CHECK-NEXT:    .cfi_offset %edi, -16
26; CHECK-NEXT:    .cfi_offset %ebx, -12
27; CHECK-NEXT:    movl $-1028477379, %ecx # imm = 0xC2B2AE3D
28; CHECK-NEXT:    movl $668265295, %esi # imm = 0x27D4EB4F
29; CHECK-NEXT:    movl a, %edi
30; CHECK-NEXT:    cmpl $0, (%edi)
31; CHECK-NEXT:    je .LBB0_2
32; CHECK-NEXT:  # %bb.1: # %if.then
33; CHECK-NEXT:    movl 8(%edi), %ecx
34; CHECK-NEXT:    movl 12(%edi), %edx
35; CHECK-NEXT:    movl %edx, %eax
36; CHECK-NEXT:    shldl $1, %ecx, %eax
37; CHECK-NEXT:    orl %edx, %eax
38; CHECK-NEXT:    leal (%ecx,%ecx), %edx
39; CHECK-NEXT:    orl %ecx, %edx
40; CHECK-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
41; CHECK-NEXT:    movl 16(%edi), %ebx
42; CHECK-NEXT:    movl 20(%edi), %edx
43; CHECK-NEXT:    movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
44; CHECK-NEXT:    shldl $2, %ebx, %edx
45; CHECK-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
46; CHECK-NEXT:    shldl $31, %ebx, %ecx
47; CHECK-NEXT:    shll $2, %ebx
48; CHECK-NEXT:    orl %ecx, %ebx
49; CHECK-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload
50; CHECK-NEXT:    shrl %ecx
51; CHECK-NEXT:    orl %edx, %ecx
52; CHECK-NEXT:    addl {{[-0-9]+}}(%e{{[sb]}}p), %ebx # 4-byte Folded Reload
53; CHECK-NEXT:    movl %ebx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
54; CHECK-NEXT:    adcl %eax, %ecx
55; CHECK-NEXT:    movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
56; CHECK-NEXT:    movl 24(%edi), %eax
57; CHECK-NEXT:    movl %eax, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill
58; CHECK-NEXT:    movl $-1028477379, %ebx # imm = 0xC2B2AE3D
59; CHECK-NEXT:    imull %eax, %ebx
60; CHECK-NEXT:    mull %esi
61; CHECK-NEXT:    movl %eax, %ecx
62; CHECK-NEXT:    addl %ebx, %edx
63; CHECK-NEXT:    movl 28(%edi), %edi
64; CHECK-NEXT:    imull %edi, %esi
65; CHECK-NEXT:    addl %edx, %esi
66; CHECK-NEXT:    movl $1336530590, %edx # imm = 0x4FA9D69E
67; CHECK-NEXT:    movl {{[-0-9]+}}(%e{{[sb]}}p), %ebx # 4-byte Reload
68; CHECK-NEXT:    movl %ebx, %eax
69; CHECK-NEXT:    mull %edx
70; CHECK-NEXT:    imull $-2056954758, %ebx, %ebx # imm = 0x85655C7A
71; CHECK-NEXT:    addl %edx, %ebx
72; CHECK-NEXT:    imull $1336530590, %edi, %edx # imm = 0x4FA9D69E
73; CHECK-NEXT:    addl %ebx, %edx
74; CHECK-NEXT:    shrdl $3, %esi, %ecx
75; CHECK-NEXT:    sarl $3, %esi
76; CHECK-NEXT:    orl %edx, %esi
77; CHECK-NEXT:    orl %eax, %ecx
78; CHECK-NEXT:    movl $-66860409, %ebx # imm = 0xFC03CA87
79; CHECK-NEXT:    movl %ecx, %eax
80; CHECK-NEXT:    mull %ebx
81; CHECK-NEXT:    movl %eax, %edi
82; CHECK-NEXT:    imull $326129324, %ecx, %eax # imm = 0x137056AC
83; CHECK-NEXT:    addl %edx, %eax
84; CHECK-NEXT:    imull $-66860409, %esi, %ecx # imm = 0xFC03CA87
85; CHECK-NEXT:    addl %eax, %ecx
86; CHECK-NEXT:    xorl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Folded Reload
87; CHECK-NEXT:    xorl {{[-0-9]+}}(%e{{[sb]}}p), %edi # 4-byte Folded Reload
88; CHECK-NEXT:    movl %edi, b
89; CHECK-NEXT:    movl %edi, %eax
90; CHECK-NEXT:    mull %ebx
91; CHECK-NEXT:    imull $326129324, %edi, %esi # imm = 0x137056AC
92; CHECK-NEXT:    addl %edx, %esi
93; CHECK-NEXT:    movl %ecx, b+4
94; CHECK-NEXT:    imull $-66860409, %ecx, %ecx # imm = 0xFC03CA87
95; CHECK-NEXT:    jmp .LBB0_3
96; CHECK-NEXT:  .LBB0_2: # %if.else
97; CHECK-NEXT:    xorl b+4, %ecx
98; CHECK-NEXT:    xorl b, %esi
99; CHECK-NEXT:    movl $1419758215, %edx # imm = 0x549FCA87
100; CHECK-NEXT:    movl %esi, %eax
101; CHECK-NEXT:    mull %edx
102; CHECK-NEXT:    imull $93298681, %esi, %esi # imm = 0x58F9FF9
103; CHECK-NEXT:    addl %edx, %esi
104; CHECK-NEXT:    imull $1419758215, %ecx, %ecx # imm = 0x549FCA87
105; CHECK-NEXT:  .LBB0_3: # %if.end
106; CHECK-NEXT:    addl %esi, %ecx
107; CHECK-NEXT:    addl $-1028477341, %eax # imm = 0xC2B2AE63
108; CHECK-NEXT:    adcl $-2048144777, %ecx # imm = 0x85EBCA77
109; CHECK-NEXT:    movl %eax, b
110; CHECK-NEXT:    movl %ecx, b+4
111; CHECK-NEXT:    addl $12, %esp
112; CHECK-NEXT:    popl %esi
113; CHECK-NEXT:    popl %edi
114; CHECK-NEXT:    popl %ebx
115; CHECK-NEXT:    popl %ebp
116; CHECK-NEXT:    .cfi_def_cfa %esp, 4
117; CHECK-NEXT:    retl
118entry:
119  %0 = load i32, ptr @a, align 4, !tbaa !1
120  %1 = inttoptr i32 %0 to ptr
121  %2 = load i32, ptr %1, align 4, !tbaa !5
122  %tobool = icmp eq i32 %2, 0
123  br i1 %tobool, label %if.else, label %if.then
124
125if.then:                                          ; preds = %entry
126  %v3 = getelementptr inbounds %struct.XXH_state64_t, ptr %1, i32 0, i32 3
127  %3 = load i64, ptr %v3, align 4, !tbaa !8
128  %v4 = getelementptr inbounds %struct.XXH_state64_t, ptr %1, i32 0, i32 4
129  %4 = load i64, ptr %v4, align 4, !tbaa !9
130  %v2 = getelementptr inbounds %struct.XXH_state64_t, ptr %1, i32 0, i32 2
131  %5 = load i64, ptr %v2, align 4, !tbaa !10
132  %shl = shl i64 %5, 1
133  %or = or i64 %shl, %5
134  %shl2 = shl i64 %3, 2
135  %shr = lshr i64 %3, 1
136  %or3 = or i64 %shl2, %shr
137  %add = add i64 %or, %or3
138  %mul = mul i64 %4, -4417276706812531889
139  %shl4 = mul i64 %4, -8834553413625063778
140  %shr5 = ashr i64 %mul, 3
141  %or6 = or i64 %shr5, %shl4
142  %mul7 = mul nsw i64 %or6, 1400714785074694791
143  %xor = xor i64 %add, %mul7
144  store i64 %xor, ptr @b, align 8, !tbaa !11
145  %mul8 = mul nsw i64 %xor, 1400714785074694791
146  br label %if.end
147
148if.else:                                          ; preds = %entry
149  %6 = load i64, ptr @b, align 8, !tbaa !11
150  %xor10 = xor i64 %6, -4417276706812531889
151  %mul11 = mul nsw i64 %xor10, 400714785074694791
152  br label %if.end
153
154if.end:                                           ; preds = %if.else, %if.then
155  %storemerge.in = phi i64 [ %mul11, %if.else ], [ %mul8, %if.then ]
156  %storemerge = add i64 %storemerge.in, -8796714831421723037
157  store i64 %storemerge, ptr @b, align 8, !tbaa !11
158  ret i64 undef
159}
160
161attributes #0 = { nounwind uwtable "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
162
163!llvm.ident = !{!0}
164
165!0 = !{!"clang version 3.6 (trunk 219587)"}
166!1 = !{!2, !2, i64 0}
167!2 = !{!"int", !3, i64 0}
168!3 = !{!"omnipotent char", !4, i64 0}
169!4 = !{!"Simple C/C++ TBAA"}
170!5 = !{!6, !2, i64 0}
171!6 = !{!"XXH_state64_t", !2, i64 0, !2, i64 4, !7, i64 8, !7, i64 16, !7, i64 24}
172!7 = !{!"long long", !3, i64 0}
173!8 = !{!6, !7, i64 16}
174!9 = !{!6, !7, i64 24}
175!10 = !{!6, !7, i64 8}
176!11 = !{!7, !7, i64 0}
177